Bateman, Ian J. - Nottingham ePrints - University of Nottingham

Loading...
Bateman, Ian J. (1996) An economic comparison of forest recreation, timber and carbon fixing values with agriculture in Wales: a geographical information systems approach. PhD thesis, University of Nottingham. Access from the University of Nottingham repository: http://eprints.nottingham.ac.uk/11312/1/320019_vol1.pdf Copyright and reuse: The Nottingham ePrints service makes this work by researchers of the University of Nottingham available open access under the following conditions. This article is made available under the University of Nottingham End User licence and may be reused according to the conditions of the licence. For more details see: http://eprints.nottingham.ac.uk/end_user_agreement.pdf

For more information, please contact [email protected]

An economic comparison of forest recreation, timber and carbon fixing values with in Wales: agriculture information systems approach a geographical

by Ian J. Bateman, B.Soc.Sci., M. A.

Thesis submitted to the University of Nottingham for the degreeof Doctor of Philosophy, October, 1996

,., T4

LCF(

BEST COPY

AVAILABLE Variable print quality

CONTENTS Abstract Thanks and acknowledgments Chapter 1: Introduction

SECTION A: FORESTRY Chapter 2: Recreation: Valuation Methods Chapter 3: Recreation: Previous Valuation Studies Chapter 4: Recreation: New Valuation Studies Chapter 5: Recreation: Estimating and Valuing Demand

Chapter 6: Timber Valuation Chapter 7: Timber: Modelling Yield Class Chapter 8: Carbon Fixing in Trees, Products and Soil: Valuation and Modelling

SECTION B: AGRICULTURE Chapter 9: Modelling Agricultural Output Values

SECTION C: COMPARISON OF FORESTRY AND AGRICULTURE Chapter 10: Comparing Forestry and Agricultural Values Chapter 11: Conclusions

11

APPENDICES Appendix 1:

0 Recreation: Previous Valuation

Studies

Appendix 2: Recreation: New Valuation Studies Appendix 3: Recreation: Estimating and Valuing Demand Appendix 4: Timber Valuation Appendix 5: Timber: Modelling Yield Class Appendix 6: Carbon Fixing in Trees, Products and Soil: Valuation and Modelling Appendix 7.* Modelling Agricultural Output Values

iii

ABSTRACT Theresearch examinesthefinancialandeconomicviability of transferring landpresentlyunderagriculturaluseinto multipurposefarm-forestryin Wales. Tbreewoodlandbenefitstreamsareexaminedin detail:thevalueof open-access by recreation;the productionof timberand;thenetcarbonstoragegenerated afforestation.Modellingof the spatialvariabilitydeterminingtheproductionof by thenovelapplicationof a geographical information thesebenefitsis enhanced system(GIS).Monetaryevaluationof non-marketrecreationbenefitsis achieved by referenceto boththecontingentvaluationandtravelcostmethodswith prior By contrastcarbonstorage studiesbeingreviewedandnewwork presented. benefitsarevaluedpurelyby referenceto theexistingliterature.Bothof these analyses yield socialvalueswhereasour studyof timberproductionproducesboth shadowandmarketvaluations. Our GIS-based methodologyis alsoappliedto themodellingof agricultural valuesfor thetwo majorfarm sectors(mainlysheepandmainlymilk production) of thestudyarea.Againbothsocialandfinancialvaluesarecalculated. By comparisonof thevariousvaluesestimatedacrosstheaboveanalyses both the financial and social valuesassociatedwith potential transfers estimate we from into land farm-forestry.The financial values conventional agriculture of generatedby our analysissupportthe presentlow levels of conversionout of agriculture. However, the social valuesestimatedsuggestthat the presentsituation constitutesa significant market failure, particularly in the mainly sheepfarming benefit cost analysissuggeststhat substantialnet social benefitscould sectorwhere be generatedthrough conversionsinto multi-purposewoodland.

iv

THANKS AND ACKNOWLEDGMENTS This researchtook about sevenyearsto completeduring which time I changedcareerstwice, jobs four times, lived in four counties in eight houses(not simultaneously)three of which I bought (two of which I eventually and with great relief sold), met the woman of my dreams,married her and had three wonderful children (at the last count). Life has therefore been "eventful" during this period and no doubt this thesis would have been finished a lot sooner had it been otherwise.I thank God (literally) that this was not the case. FOR Fiona, Ben, Freya and Natasha.In truth I would not have had the strength to complete this without your love and support. I love each and every one of you tremendouslyand will try my hardestto make up for the time we have lost while I was doing this.

VERY SPECIAL THANKS There is one person I should like to single out for very special thanks: my friend and colleague Andrew Lovett. I will always be grateful to you for your unstinting help and for your constantsupportand ftiendship.

SPECIAL THANKS

, Julii Brainard (for being a great listener as well as one of the most talented

researchersI have ever encountered),Ian Langford (for intellectual stimulation, statistical wizardry, a sterling attempt to keep me fit - one day I will beat you - but most of all for being an excellent friend), Kerry Turner (for both starting and ever (for UEA), Randell Frances typing the supporting my career at since indecipherable,incompatible and at times impossible).I also send lots and lots of love and thanks to my Mum, Dad and sisters,Hepzibah and Seeseana,for giving

V

happy home (and for in life: best the a picking up the pieces and start me very apologisingon my behalf ever sinceI left).

MY SUPERVISORS I feel very fortunate to have had suchexcellent supervisors.ProfessorTony Rayner and Professor Chris Ennew allowed me the freedom to develop this researchand have consistently picked me up throughout the dark nights that a friends induces. length I them think as good this of now project of and solitude City if future Manchester in (and I hope the to will probably whom see more of keep on playing the way they havebeenlately).

THANKS ALSO TO The inherently interdisciplinarynature of this project meant that I neededa lot of help from a lot of people. Those listed below helped tremendouslyeither directly with the researchor by relieving me from certain of those areasof reality which tend to impinge upon one's ivory tower. I am very grateful to you all. Apologies to thoseI have doubtlessomitted. John Aitchison, Roger Backhouse,Tony Ballance,GrahamBentham,JaquieBerry, Nancy Bockstael, John Bowers, Kevin Boyle, Ian Bradley, Mick Bray, Richard Broadhurst, Mike Burton, Richard Carson, Marc Carter, Nigel Chapman, Anna Chylack, Rosemary Clarke, Keith Clayton, David Colman, Bill Corbett, Rosie Cullington, John Dalton, Emily Diamand, Peter Doktor, Cathy Dowson, Frances Elender, Doris Ellis, Bob Farmer, Ross Ferguson, Nick Flores, Jim Ford, Vivien Foster, Guy Garrod, Stavros Georgiou, Simon Gerrard, Simon Gillarn (and others Shirley Gilmore, Edinburgh), Commission Statistics Branch, Forestry the at Andreas Graham, Colin Green, Ian Goatman,Des Hammond, Michael Hanemann, Nick Hanley, Elanor Harland (and all at the Forestry CommissionLibrary), David Harley, Tim Harrod, Dick Hartnup, David Harvey, Robin Haynes, Debbie Hazelton, Norm Henderson, John Hoehn, Mike Holland, Alan Hounsell, John Hunt, Mark Hunt, Oliver Hunt, Tim Jenkins (and all the surveyors at FBSW, Aberystwyth), David Jenkinson, Bob Jones, Tom Jones, Geoff Kerr, Andrew Kitchen, Suzanne Kraft, Bengt Kristrom, Iain Lake, John Loomis, Loop Guru, Barbara Maher, Heidi Mahon, Robert Matthews (and others at the Forestry Commission Research Station, Alice Holt), Duncan MacLennan, Douglas MacMillan, Helen McCloud, Steve McGrath, Debbie McGurk, Chris Mellor, vi

Alistair Munro, Charles Norman (and others at the Timber Trades Association), Roger Oak, Tim O'Riordan, William Page, David Pearce, Jason Pearce, Mick Pearson,GeorgePeters,Phil and Joy, Sally Pidgeon,Neil Powe, Colin Price, Brian Pulford, Chris Quine (and others at the Forestry CommissionNorthern Research Station, Roslin), Tahir Rehman,Bruce Rhodes, Noel Russell, Andrew Schuller, Kerry Smith, The Smiths, Dorothy Snaddon,Mike Stabler, David Stafford, Chris Stanner, Jon Stewart, Bob Sugden,Colin Thirde, Arthur Thomasson,Transglobal Underground, Sylvia Tunstall, Fred Vine, David Walker, Jo Wall, Martin Whitby, Adrian Whiteman (and others at the Forestry Commission Headquarters, Edinburgh), Julie Whittaker, Soren Wibe, Ken Willis, Rob Willis, Michaela Windschuffel, Rick Worrell, Chris Wright, JohanneYates. I am also tremendouslygrateful to the Farm BusinessSurvey in Wales,Ibe Forestry Commission and to the Soil Science and Land Research Centre for providing me with the data necessaryto conduct this research.Quite simply this thesis could not have beenwritten without their support. I also wish-to thank agenciessuch as the ESRC, English Nature, ETSU, the CEC and many others who have funded other areas of my research.and so given me the licence to pursue this my own personal researchtopic. I hope that you the reader find it of interest. If so pleasefeel free to make any commentsor (onceit has beenpassed)exposeany glaring errors. Ian Bateman School of Environmental Sciences, University of East Anglia, Norwich, NR4 7TJ United Kingdom. i. [email protected] ac.uk

August 1996

vi'

Chapter 1: Introduction 1.1

THESIS Perhapsthe most often quoted definition of an economist is of someonewho knows

the price of everything and the value of nothing'. Such a description is sadly true of most of the discipline. However (if I may briefly stray away from that third person who seemsthe obligatory, impersonal author of all modem academicpapers), to me it is an awarenessof the distinction between value and price which separatesout the true economist from the glorified book-keepersand accountantswho so often masqueradeunder such a title. Recent years have seena proliferation of badge-engineeringin which so-called new disciplines such as environmental or ecological economics have risen to prominence. However, whilst these are appealing titles, in, essencethey represent not a radical departure but rather a very welcome return to the basic principles and domain of economics- the analysis of true value. 1.1.1 THE NATURE OF VALUE It is one of thesebasic principles which underpins this study: namely the assumption that values can be measured by the preferences of individuals.3 The interaction of preferenceswith the various servicesprovided by a commodity generatesa variety of values. Many economistshave studied the nature of thesevalues, however, a useful starting point is the concept of aggregateor total economic value (TEV) (Pearceand Turner, 1990; Bateman, 1991a; Bateman and Turner, 1993; Turner et al., 1994; Bateman, 1995a/b; Turner, forthcoming). Figure 1.1 shows how TEV can be broken down into its constituent parts and

illustratesthesewith referenceto certainof the valuesgeneratedby theprincipalcommodity underconsiderationin this study;woodland. 'This derives from a corruption of Oscar Wilde's definition of a cynic in Lady Windermere'sFan (Act III). However, given the perceivedsimilarity betweenthe two groups it is easy to seehow such a confusion may have Olvar Bergland, Colin Price and others regarding this.) (With to thanks arisen. 2It was not always so. In reading the papers of one of our most eminent economists (and I use the term most correctly) John Hicks, the readeris addressedby a real person who talks directly, plainly and with a clarity dear reader I do not possess(as you will soon discover). sadly which expression of 3Speculationsupon this issue and in particular about whether individuals have definite preferencesare presentedby Sugden (forthcoming).

1.1

Figure I. I:

The total economic value of woodland

HUMAN VALUES

NON-HUMAN VALUES

Total Economic Value Use Value Utilitarian --I II Market Priced Non-Market

(Timber]

Non-Use Value 17 Option

Ppeecn Access] reation

rs 0,net PeFulure recreation]

Landscape ameni ty

Bequest

Existence

Intrinsic Value

Future

[t SaI 1:. 1, I 13 c on hat is, . is there

V81USOf the r source in it: own right

generation benefits

I

Source: Adapted from Bateman (1995a)

The bulk of economic analysesconcentrateupon the instrumental or use values of a commodity. Most prominent amongsttheseare the direct use values generatedby private and quasi-private goods (Bateman and Turner, 1993) which are often partly reflected by market prices, and those indirect use values associatedwith pure and quasi-public goods (ibid) which generally have no market price description. A unifying characteristic of these values is that they are all generatedby the present use of the commodity by the valuing individual. An extension of the temporal frame allows for the possibility of individuals valuing the option of future use (Weisbrod, 1964; Cicchetti and Freeman, 1971; Krutilla and Fisher, 1975; Kristr6m, 1990). Related to this is the notion of bequestvalue wherein the valuing individual gains utility from the provision of use or non-use values for present and/or future others. Pure non-usevalues are mostly commonly identified with the notion of valuing the continued existenceof entities such as certain speciesof flora and fauna or even whole ecosystems. As before this is generally both an intra- and inter-generationalvalue and becauseof the lack of an instrumental element hasproved problematic to measure.Nevertheless,the theoretical case for the 'existence of existence value' is widely supported (e.g. Young, 1992). 1.2

Wider definitions of value have been argued for. An important issue concerns the extent of the 'moral referenceclass' (Turner et al., 1994) for decision making. One question here arises from the treatment of other (both present elsewhere and future) humans while another concernswhether animal, plant and ecosysteminterestsshould be placed on an equal footing with human preferences. The modern origins of such a view can be traced to Goodpastor (1979) and Watson (1979) who take the Kantian notion of universal laws of respect for other persons and extend this to apply to non-human others. Watson feels that those higher-animals such as chimpanzees (which he argues are capable of reciprocal behaviour) should be accordedequal rights with humans. Hunt (in Perman et al., 1996) and Rollston (1988) build upon the land ethic of Leopold (1949) to extend this definition of moral referenceeven further to include all extant entities, an approachwhich Singer (1993) defines as the 'deep ecology' ethic. Such a paradigm arguesthat theseentities possessan 'intrinsic' value separatefrom anthropocentricexistencevalues. A further departurefrom conventional utilitarianism is proposedby Turner (1992 and forthcoming) who arguesthat all the elements of TEV can be seen as secondary to a primary environmental quality value which is a necessaryprerequisitefor the generationof all subsequentvalues. Sidesteppingthe theoretical casefor such philosophical extensions,a practical problem with thesenon-TEV values is that they are essentially beyond the scope of conventional, anthropocentric, preference-based economic valuation. Given that in this study we constrain the moral referenceclass to present humans alone, this in turn defines TEV as the appropriate extent of value definition. However, this still leaves the problem of how such values should be measured.

1.1.2 FROM VALUES TO APPRAISALS: DIFFERING PARADIGMS One solution to the problem of valuation might be to abandon conventional neoclassical economic analysis in favour of modified or alternative appraisal and decision making strategies. One such alternative is to base decisions upon expert judgement and restrict the role of economics to the identification of least cost methods for achieving stated aims (see, for example, OECD, 1991). Such a cost-effectivenessapproach may be optimal for a risk aversesociety faced with high risk, high uncertainty problems such as the treatment of persistent pollutants (Opschoor and Pearce, 1991). Here a useful decision guide is (see, for by 'ecological by the principle advocated economics' precautionary provided example, Costanzaand Daly, 1992; Toman, 1992; Turner et al., 1995). 1.3

However, in situations where the precautionaryprinciple does not apply (particularly for low risk, low uncertainty decisions) then a cost-effectiveness approach may entail avoidable and, in some cases,major net welfare losses compared to a solution basedupon cost-benefit analysis (CBA).

Such a position is adopted by those who argue for an

&environmentaleconomics' paradigm (see, for example, Pearceet al., 1989; Department of the Environment, 1993; Pearce,1996). Here supportersacceptpreferencebasedvalues as the basis of decision making but argue for full assessmentof TEV as opposed to the concentrationupon market basedmeasureswhich appearsto dominate much presentpractical decision making.

This choicebetweenecologicalandenvironmentaleconomicscan be characterised as one between principle or pragmatism. The argument for an ecological economics approach 0 is that nothing less will preserve the environmental integrity which is vital if the present 'cowboy economy' (Boulding, 1966) is to attain a state of sustainabledevelopment. The environmental economic critique is that such a rigid approach fails to recognise the mechanisms through which present day decision making operates and thereby risks being ignored. In the absenceof hindsight it is impossible to know which strategy is most likely to influence the presently unsustainablecourse of economic growth. Our own position is that the two paradigmsneednot be in conflict and that a modified precautionary principle can be used to assesswhich approach is appropriate for any given decision situation. Furthermore, we seea role for public preferenceswithin this process. For caseswhere expert assessmentand/or informed public opinion identifies high potential risks or uncertainties from a given strategyor decision then a precautionary,ecological economics approach would appear justifiable.

In instances where this is not the case then an

environmental economicsanalysis seemslikely to be optimal. Both are significantly superior to simple market-basedappraisals.

1.2

THEORETICAL, METHODOLOGICAL OF THE STUDY

AND EMPIRICAL BASIS

We therefore need to select the appraisal paradigm which is most appropriate for the

land for This thesis the of examines economic potential conversions subjectunderanalysis. 1.4

use from conventional agriculture into woodland in Wales. Two points are immediately important here. First we are interestedin the full range of economic valuesgeneratedby such a change in land use. Second, following initial review (Bateman, 1991a and b, 1992), it becameapparentthat large scale unquantifiable risks or uncertaintieswere not a major factor in such an analysis. Given this, the adoption of an environmental economic CBA paradigm seemeddefensible. CBA is, in effect, an appraisalof the social worth of a project and the study presented in this thesis attempts to move beyond simple market related assessmentsof value to a more complete analysis of TEV. In assessingwoodlandswe attempt to be comprehensivealthough in practice focus falls upon timber production, open-accessinformal recreation, and the value of carbon sequestration(i.e. global warming abatement). This is compared to an appraisal of the social value of agriculture which takes account of items such as the subsidy transfers currently paid by society to farmers. However, while such an economic analysis is of use in informing decision makers and shaping optimal policy change, it cannot of itself predict farmers response to that change unless the impact upon farm incomes are also known. Consequently the study also examines farm gate incomes under presentand potential future policy scenarios. The ultimate objective of this study is therefore to provide a policy analysis tool. However, whilst the theoretical CBA framework of the research is conventional, the extent of application and the methodology employed is innovative. The depth of analysis is, we feel, more rigorous than in previous studies. Furthermore, the methods developed involve a spatial analytic framework which, to our knowledge, is unique. Regarding this latter point we make extensive use of geographical information systems

(GIS) throughout this study. A GIS is a software packagecapable of holding, interrogating facility Through data digital this we can spatially such as maps. and manipulating referenced combine environmental and other spatial data with more conventional variables into the stochastic economic models which underpin this study. As we demonstrate through the contexts of modelling timber yield, carbon sequestration,recreationaldemandand agricultural productivity, the ability to integrate diverse datasetsyields a substantial improvement in the is important However, the their consequentvalues. equally modelling of such variables and superior display and interrogation of resultant models permitting the decision maker to readily It is dual improved impact this the alternative policy choices. capability of of comprehend 1.5

modelling and display which we feel establishesthe potential of a GIS for significantly improving economic modelling. 1.2.1 THE COST AND BENEFITS OF WOODLAND: LIMITATIONS STUDY

OF THE

Figure 1.2 illustrates the complexity of internal and external costs and benefits which are generatedby woodland. Here the internal costs and benefits are shown in shadedboxes. These items all have market prices from which shadow values may be derived. Certain external items also have related market prices from which values may again be estimated; these are shown in the broken line boxes in figure 1.2. However, the remaining externalities do not have related market prices thereby making valuation problematic. Our study sets out to provide a full environmental economic assessment of all the values associated with the proposed conversion of agricultural land use into woodland. However, we have to recognise certain limitations to this study. First, methods for the monetary evaluation of preferences for non-market goods and services are not uniformly developed for all value types. In particular, methods for the evaluation of non-use benefits such as existence values have been the subject of sustained criticism in recent years (see Chapter 2). Our study reflects these reservations by concentrating upon use-values. Secondly, time constraints and data availability problems meant that even our treatment of all use values is somewhat uneven. Thirdly, we are only considering a conversion from C)

agricultural land to woodland and not any other alternative use. Strictly speaking this contravenesthe principles of CBA which state that the appraisalof opportunity costs should include the assessmentof a wide range of feasible alternative resource uses (Pearce, 1983; Bateman et aL, 1993). A fourth issue is that of equity and its root: ethics.

1.2.1.1. A note on ethics' Ethics and economicshaveoften be presentedas strangebedfellows. Indeedmany proponents of the 'positive economics, which has dominated so much of twentieth century economic analysis argue that the two concepts cannot be related "in any form but mere

juxtaposition" (Robbins, 1935: p. 148). However, this has not always been a widely held

'rhis discussionreliesheavilyon Permanel al. (1996),KneeseandSchulze(1985)andPearceandTurner (1990). Relevantdiscussionsare alsopresentedin Beauchamp and Bowie (1988)and Sen(1987).

1.6

Figure 1.2: Costs and benefits of woodland

UK FORESTRY

I BENEFITS

COSTS

I

INTERNAL I T'Imber Granw Subsidies Commercial Recreation sporting Venison

I

EXTERNAL

1

ac

EXTERNAL

I

I

Land

----------Grants/Subsidies

Employment/Rural Support

Labour Capilaf

Reducing Agricultural Surpluses Import Savings L--------------------------

Expertise

Alternative output (Opportunity Costs) L- ---------------------

. security ------------S766ýly

............

i

Recreatiorv education Amenity Wildlife Habitat Ecology Value Biodiversity Soil Stability Hydrological: Regulation Storage Carbon sinks/ macroclimate regulation Microclimate regulation Berries/ Game etc. Fuel Shelter

I Market priced hems i n an intemaý,

INTERNAL

Amenity loss HabrIat loss Ecological loss Biodiversity loss Soil degradation Hydrological: Water traps Acidification

-------------------Exlemal impacts

Non market external impacts.

Pricing via monetary evaluation methods where feasible

priced by reference marke-ts _--to. .....

Notes: 1. 2. 3.

Here 'internal' refers to forest operator costs and benefits whether Private Woodland or Forestry Commission. 'External'ireters to costs and benelits accruing directly to society. Shadow pricing techniques must be applied to all values. Not all the items listed may be valid (eg. import substitution argument).

Source: Bateman (1992) 1.7

belief.

Indeed the early great economists were explicitly concerned with morality and

' 6 ethics. Two ethical positions which have had a major impact upon the development of economic thought are the libertarian and utilitarian schools of thought. The libertarian view, which may be traced from John Locke and Adam Smith to Robert Nozick (1974), emphasises respect for the rights of individuals.

A principal concept here concerns the just acquisition

of property. This has been interpreted as emphasising both the rights of ownership and also the requirement of appropriate payment or transfer in return for acquisition.

However,

libertarianism makes no prescriptions concerning the outcome of any trade or transfer. In particular such a view would almost always condemn any redistributive policy, whether between present or to future people (intra and intergenerational transfers) unless they are freely entered into by all groups including donors.7 This focus than

upon processesrather

outcomes differs from the utilitarian view (which derives from the writings of David Hume, Jeremy Bentham and most notably John Stuart Mill (1863)), which explicitly highlights the ethical consequencesof actions. Classical utilitarianism judges actions upon whether they are 'good' for society, with 'good' being defined (by Mill) in terms of happinessor utility. Actions which promote utility are therefore good and should be judged by the amount of utility created. However, for utility to be cardinally measurableindividuals must be able to expressutility in terms of a numeric value. Furthermore, in order to assessthe social utility of an action we have to assumethat we can compare and add utility across individuals. These strong assumptionsmake Classical utilitarianism of little use for the practical

economicanalysisof projects. The neoclassicalutilitarianism(Kneeseand Schulze,1985) (Layard which underpinsmodernwelfareeconomicsrequiresrelatively weakerassumptions andWalters,1978;Varian, 1987). In particulara commonassumptionunderpinningCBA is that the marginalutility of consumptionis equalacrossall individuals. If this is so we can ignore distributive issues(which are vital under Classicalanalysis)as any action which createsnet benefits unambiguouslyraise social welfare. However, in reality such an assumptionseemsunlikely to hold, promptingsomeCBA analyststo explicitly considerthe

31nterestinglyAdam Smith's post at the University of Glasgow was as Professor of Moral Philosophy. 'Reviews of the work of Marx, Marshall, Pareto, Keynes and others are presentedin Schumpeter(1952). 'This would conventionally rule out any governmental action towards the enforced provision of such transfers.

1.8

equity implications of their analyses(e.g. Squire and van der Tak, 1975). For many years such commentators were a generally inconspicuous minority within the profession of economics. However, since the 1960's concerns regarding the effects of environmental degradationupon presentand future generation,and the issueof North/South inequality have meant that discussionsconcerningthe ethical basisof economicshave grown. Theseconcerns regarding the need to consider equity as well as economic efficiency have recently coalesced within what has beentermed the SustainableDevelopment (SD) debate(WCED, 1987;Pearce et al., 1990). Both intra- and inter-generationalequity issuesare central to the SD debatewhich has in essenceproposed an alternative to utilitarianism as a new ethical basis for economics. Pivotal to this has been the work of Page (1977) and in particular Rawls (1972). Rawls' Theory of Justice can in fact be seenas a direct developmentof Kants universal laws. Here the individual enjoys common liberties compatible with equal rights for others, while valid inequalities only result from personal attributes which are accessibleto all (e.g. work and learning as opposed to sex and creed). This latter prescription has important consequences for equity as Rawls arguesthat under such a systemthe optimal allocation of resourcesis one that is made behind a 'veil of ignorance' as to their intra- and intergenerational incidence. This can be seen as being in direct conflict with the individual maximisation principle of ' utilitarianism. This is perhaps most clearly demonstratedin the recent literature regarding sustainability. Turner and Pearce(1993) identify four alternativepositions ranging from 'very weak' to 'very strong sustainability'. Each definition moves further from a conventional Utilitarian to a Rawlsian position on equity, steadily imposing more constraintsupon resource use (most notably natural capital).

The ethical position adopted in this study As discussedabove there are a number of ethical positions which could be adopted for this study. Despite our own sympathy with the Rawlsian/Strong Sustainability view, our

"Theeconomicimplicationsof ClassicalandNeoClassicalUtilitarianandRawlsianethicalpositionscanbe throughconsequent expressed socialwelfarefunctions(SWF). ClassicalUtilitarianismimpliesan additiveSWF URwhereW= socialwelfare;U, U' = thetotal utility enjoyedby individualsA and of the form: W=0,1' + P2, B respectively;0,0, = weightsusedto calculateW. Neo ClassicalUtilitarianismrelaxesthe assumptionof additivity suchthat W= W(UA,U"). Finally, following Solow(1974).the Rawlsianpositioncanbe expressed asthe maxi-minfunctionin which we maximiseW= min (UA,UB). NotethatPermanet at. (1996)suggestthat Rawlsmay havestronglyobjectedto the latter utilitarianreformulationof his work.

1.9

self-assessmentis that the study is essentially neoclassically utilitarian in its ethical basis. The definition of values inherent in the TEV conceptremains anthropocentricand is therefore consistent with the extended utilitarian view discussedby Perman et al. (1996). The most non-Rawlsian characteristic of this study is the absenceof an explicit incorporation of any precautionary principle or equity constraint. Some commentators may argue that the sensitivity analysis acrossvarious discount rates (discussedin chapter 6) which we apply to our CBA effectively addressesthe issueof intergenerationalequity. However, as Hanley and Spash (1993) highlight, such an approach will not ensure equality of wellbeing across generations. Similarly we do not include explicit considerationsof distributional effects nor do we include any analysis which could be construed as compatible with a Rawlsian maximin criteria. Our approachtherefore is, in theoretical terms (and in terms of the ethical basis of that theory), essentially conventional. It is only in the practice of this analysis that we have attempted to improve upon convention. This theoretical standpoint should not be taken as implying a wholesale rejection of the Rawlsian or strong sustainability positions. Rather it is a pragmatic extension of accepted decision-analysispractice. 1.2.2 SELECTION OF THE CASE STUDY AREA While the fundamental objective of this study was the comparison of woodland with agricultural values, a supplementary initial goal was to see how this comparison varied spatially over a diversity of sites. Accordingly an initial researchframework envisioned three case studiesat two sites in England (one stretching from the Exe Valley acrossDartmoor and the other located in North Norfolk) and one in Wales (a lOkm wide transect running from Aberystwyth to Newtown). These sites were chosen to reflect a diversity of environments ranging from lowland areas yielding high agricultural productivity, to extreme upland locations where only marginal farming activities are feasible. However, in the event it proved impossible to obtain a full set of the data necessary

to modelthe diversity of woodlandandagriculturalvaluesassociated threeareas. with theS'e Specifically the Ministry of Agriculture, Fisheries and Food (MAFF) refused to release the

farm level agricultural data for England which we felt was necessaryto exploit the spatial

1.10

9 by GIS. Accordingly attention was turned exclusively to analytic capabilities afforded a Wales where the relevant authorities were highly supportive of our work (for which we are very grateful). However, to compensate for the lack of English data it was decided to expand the Welsh study area to encompass the entire principality, thereby including high productivity lowland as well as upland areas.10

1.2.2.1 Data sources This researchhas drawn upon a variety of data from a number of sources. All data was provided free or for a reasonable handling charge. We are very, very grateful to a number of people for this cooperation without which the research could not have been undertaken. Data on farm level agricultural activities, costs and revenues was obtained from the Farm Business Survey in Wales (FBSW). We are indebted to the enlightened attitude of the FBSW who, by being prepared to enter into a confidentiality agreement whereby no farm level results were reported, facilitated a highly substantial improvement in the ability to model agricultural production and its value by allowing us to link farm level decisionmaking to the local environment through the grid reference coordinates of the farm. Environmental data was provided in the form of the LandIS database kindly loaned by the Soil Survey and Land Research Centre (SSLRC), Cranfield.

This is the premier

repository of land infon-nation data for England and Wales. When used in conjunction with the FBSW data this provided the highest quality combination of information possible for modelling agriculture in the study area.

This high quality was maintained in our final principal data source; the Forestry Commission's (FC) Sub-CompartmentData Base (SCDB). This is the most extensive and comprehensivesourceof woodland data in the UK and is again spatially referencedto a high degreeof accuracypermitting synthesiswith the environmental data contained in the LandIS database.

9The only data profficred was the Parish Census database. This both fails to identify individual farm locafions (thus rendering accurateproduction modelling unfeasible)and does not report certain key profitability variables. "Nevertheless, given the rclafively low population density of Wales we do regret not being able to include the more populous areasof England in our study.

1.11

A number of other data sources were employed to provide specific variables. Prominent amongst these was data on windiness provided by the FC" and digital maps of Environmentally Sensitive Area borders provided by MAFF. The principal supplementary data source was surveys conductedfor this project and reported subsequentlyin this thesis, the structure of which we now consider.

1.3

STRUCTURE OF THE THESIS

This thesisis divided into three sectionsconcerningrespectively woodland, agriculture and a CBA comparison of the two. Section A opens with a consideration of the recreation value of woodland. This is subdivided into an appraisal of methods for the monetary evaluation of woodland (chapter 2), a review of previous evaluation studies (chapter 3), presentationof our own studies (chapter 4) and GIS-basedanalysis transferring results from thesevarious evaluations to the casestudy area (chapter5). The focus of attention then shifts to timber as an evaluation model is constructed (chapter 6) and applied to newly estimated yield models (chapter 7). The section, and with it our analysis of woodland values, is concluded by extending the definition of values to include the net benefits of carbon sequestrationprovided by forests (chapter 8). Section B shifts the focus of attention to agriculture presenting models of both the farm gate and social values of production (chapter 9). Section C opensby synthesisingthe precedingchaptersand comparing woodland with agricultural values. Both market and social perspective assessmentsare presented(chapter 10). This analysis identifies a number of interesting results from which policy implications 0 and conclusions are drawn and presented (chapter 11).

"in the person of Chris Quine at the FC's NorLhem ResearchStation, Roslin to who we are grateful for cooperation, accommodation and hospitality.

1.12

REFERENCES Bateman,IJ. (1991a) Recentdevelopments in the evaluationof non-timberforestproducts:the extended CBA method,QuarterlyJournal of Forestry,85(2):90-102. Bateman,1J. (1991b) Placingmoneyvalueson the unpricedbenefitsof forestry,QuarterlyJournal of Forestry,85(3):152-165. Bateman,1J. (1992) The UnitedKingdom,in Wibe, S. and Jones,T. (cds)Forests:Marketand Intervention Failures,Earthscan,London. Bateman,IJ. (1995a) Environmentaland economicappraisal,in O'Riordan,T. (ed.) EnvironmentalScience for EnvironmentalManagement, Longmans,Harlow. Bateman,1J. (1995b) ResearchMethodsfor ValuingEnvironmentalBenefits,in Dubgaard,A., Bateman, I.J. and Merlo, M. (eds)EconomicValuationof Benefitsfrom CountrysideStewardship, Wissenschaftsveriag Vauk, Kiel, pp.47-82. Bateman,IJ. andTurner,R.K. (1993) Valuationof the environment,methodsand techniques:the contingent valuationmethod,in Turner,R.K. (ed.) SustainableEnvironmentalEconomicsand Management: Principlesand Practice,BelhavenPress,London. Bateman,IJ., Turner,R.K. and Bateman,S.D. (1993) Extendingcostbenefitanalysisof UK highway proposals:environmentalevaluationandequity,ProjectAppraisal,8(4):213-224. Beauchamp, T.L. and Bowie,N.E. (1988) Ethical Theoryand Business,,3rd ed., PrenticeHall, Englewood Cliffs, NJ. Boulding,K. (1966) The economicsof the comingSpaceship Earth,in JarCLt, H. (ed.) Environmental Quality in a GrowingEconomy,JohnsHopkinsUniversityPress,Baltimore,MD. Cicchetti,CJ. and Freeman,A. M. 111(1971) Optimaldemandand consumer'ssurplus:furthercomment, QuarterlyJournal of Economics,85:528-539. Costanza,R. and Daly, H. (1992) Naturalcapitalandsustainable Conservation development, Biology 6:3746. Departmentof the Environment(1991)Policy Appraisaland theEnvironment,HMSO,London. Goodpaster, K.E. (1978) On being morallyconsiderable, TheJournal of Philosophy,75:308-325. Hanley,N.D. and Spash,C. (1993) Cost-BenefitAnalysisand the Environment,EdwardElgar,Aldershot. Kneese,A.V. and Schulze,W.D. (1985) Ethicsand environmentaleconomics,in Kneese,A. V. and Sweeney,J.L. (eds) Ilandbookof Natural Resourceand EnergyEconomics,Volume1, Elsevier SciencePublishers,B.V. Kristr6m,B. (1990) W. StanleyJeavons(1988)on option value,Journal of EnvironmentalEconomicsand Management,18:86-87. Krutilla, J.V. and Fisher,A.C. (1975) TheEconomicsof Natural Environments:Studiesin the Valuationof for the Future), Commodityand AmenityResources, JohnsHopkinsUniversityPress(for Resources Baltimore,MD. Layard,P.R.G. and Walters,A.A (1978) MicroeconomicTheory,Ist ed., McGraw-Hill,Maidenhead. Leopold,A. (1949) A SandCountyAlmanacand Sketches Ilere and There.Oxford UniversityPress,New York. Mill, J.S. (1863) Utilitarianism, in Smith,J.M. and Sosa,E. (eds)(1969)Mill's Utilitarianism: Textand Criticism,Wadsworth,Belmont,CA. Nozick, R. (1974) Anarchy,Stateand Utopia,JohnsHopkinsUniversityPress,Baltimore,M.D. Opschoor,J.B. and Pearce.D.W. (eds)(1991) PersistentPollutants:Economicsand Policy, Kluwer AcademicPublishers,Dordrecht. Organisationfor EconomicCo-operationand Development(1991) EnvironmentalPolicy: How to Apply EconomicInstruments,OECD,Paris. Page,T. (1977) Conservationand EconomicEffliciency,JohnsHopkinsUniversityPress,Baltimore,MD. Pearce,D.W. (1983) Cost-BenefitAnalysis,2nd ed., Macmillan,London. Pearce,D.W. (1996) Economicvaluationandecologicaleconomics,plenaryaddressto the EuropeanSociety for EcologicalEconomicsInauguralInternationalConference:Ecologle,Socielie,Economie, Universityof Versailles,Guyancourt,May 23-25,1996. Pearce,D.W. and Turner,R.K. (1990) Economicsof Naiural Resources and the Environment,Harvester Wheatsheaf, Hemcl Hempstead. Pearce,D.W., Barbier,E.B. and Markandya,A. (1990) SustainableDevelopment:Economicsand

1.13

Environment in the Third lVorld, Earthscan,London. Pearce,D.W., Markandya, A. and Barbier, E.B. (1989) Blueprintfor a Green Economy, Earthscan,London. Perman, R., Ma, Y. and McGilvray, J. (1996) Natural Resourceand Environmental Economics, Longman, Harlow. Rawls, J. (1972) A Theory of Justice, Oxford University Press,Oxford. Robbins, L. (1935) An Essay on the Nature and Significance of Economic Science,2nd ed., Macmillan, London. Rollston, H. (1988) Environmental Ethics, Temple University Press,Philadelphia. Schumpeter,J.A. (1952) Ten Great Economists: From Marx to Keynes, George Allen and Unwin Ltd, London. Sen, A. (1987) On Ethics and Economics, Blackwell, Oxford. Singer, P. (1993) Practical Ethics, 2nd ed., Cambridge University Press,Cambridge. Solow, R.M. (1974) Intergencrational equity and exhaustible resources,Review of Economic Studies, S:2946. Squire, L. and van der Tak, H. (1975) Economic Analysis of Projects, Johns Hopkins University Press, Baltimore. Sugden,R. (forthcoming) Public goods and contingent valuation, in Bateman, I.J. and Willis, K. G. (eds) Contingent Valuation of Environmental Preferences:AssessingTheory and Practice in the USA. Europe, and Developing Countries, Oxford University Press. Toman, M. A. (1992) The difficulty of defining sustainability, Resources,106:3-6. Turner, R.K. (1992) Speculationson weak and strong sustainability, CSERGEGlobal Environmental Change Working Paper 92-96, Centre for Social and Economic Researchon the Global Environment, University of East Anglia and University College London. Turner, R.K. (forthcoming) The place of economic values in environmental valuation, in Bateman, I.J. and Willis, K.G. (eds) Contingent Valuation of Environmental Preferences:AssessingTheory and Practice in the USA, Europe, and Developing Co. Turner, R.K. and Pearce,D. W. (1993) Sustainableeconomic development: economic and ethical principles, in Barbier, E.B. (ed.) Economicsand Ecology: New Frontiers and SustainableDevelopment,Chapman and Hall, London. Turner, R.K., Pearce,D.W. and Bateman, I.J. (1994) Environmental Economics: An Elementary Introduction, Harvester Wheatsheaf,Hemel Hempstead. Turner, R.K., Peffings, C. and Folke, C. (1995) Ecological economics:perspectiveor paradigm? CSERGE Global Environmental Change Working Paper 95-17, Centre for Social and Economic Researchon the Global Environment, University of East Anglia and University College, London. Varian, H.R. (1987) Intermediate Aficrocconomics,2nd cd., Norton, New York. Watson, R.A. (1979) Self-consciousnessand the rights of non-humananimals, Environmental Ethics, 1(2):99. Weisbrod, B.A. (1964) Collective - consumption services of individual consumption goods, Quarterly Journal of Economics, 78:471-477. World Commission on Environment and Development (1987) Our CommonFuture, Oxford University Press, Oxford. Young, M. D. (1992) Sustainable Investmentand ResourceUse, UNESCO/Parthenon,Carnforth.

1.14

SECTION A: FORESTRY

Chapter 2: Recreation: Valuation Methods 2.1: INTRODUCTION At the heart of Cost-Benefit Analysis (CBA) theory lie two basic principles (Pearce, 1984): firstly that, as far as possible, all the costs and benefits arising from a project should be assessed;and, secondly, that they should be measuredusing the common unit of money. While these seemcommonsenseprecepts,in application both principles raise highly complex problems. The issueof complete appraisalis, when taken to the extreme,ultimately insoluble in a world ruled by the laws of thermodynamicswhere, as noted by commentatorssuch as Price (1987) and Young (1992), everything affects everything else. For real world decisionmaking, practical rules regarding the limits of appraisal are needed. Such rules are the stuff of numerousproject appraisalguidelines, for example the Treasury's 'Green Book' (H.M. Treasury, 1991), whereasour researchfocussesupon the secondprinciple of monetary evaluation. In discussingapproachesto the monetary evaluation of environmental preferenceswe first identify a wider global family of monetary assessmentmethods (see figure 2.1). This below family discussed both formal 'valuation' and a quite separate the comprises methods 'valuation' 'pricing' In 'environmental theoretical terms techniques'. and of ad-hoc pricing' individuals based former Whereas dissimilar. the are upon preferences approachesare quite (hence 'valuation' term the methods), measures and yield conventional, neoclassical,welfare A typical example 'pricing' to price observations. market the techniquesare much more akin of such a technique is given by the use of assetreplacement,restoration or transplantation (Buckley, 1989). it has been While involving assets in environmental costs project appraisals for the heuristic tools appraisal of projects, policies or argued that such methods provide 1992), Brooke, techniques pricing Bateman reflect the costs (Turner, and coursesof action benefits. In of considering only prices of environmental protection alone to the exclusion incorrect Certainly danger in making choices. of such decisionmakers are rather than values, 'Critical reviews of thesepricing approachesare given (in ascendingdetail) in Bateman (1992,1995a and 1995b). 'As an interesting recent example of how pricing methodsmay give little practical guidance to a decision, Medley (1992) refers to the Departmentof Transportfspricing of a mOtorwaytunnel to avoid a cutting through this too considered was L70 expensive At and abandoned million in Hampshire. SSSI Twyford Down the being ' undertaken. alternative benefits an of such without any appraisal of the

2.1

=

t-

.M

cu 5ýC)

0

cli

iz

rz 412

&A

-. +

-0

t)

gn = eu -

E ce 2ý

ID

(b 0

LD

0 *Z:

ýO J

r_

9. ---UV

u) 4)

>

ý:

I

to *C r_ tu



.E00

00 ce .

4-

cu

E

ýK c2. f2 .2 020" (12

bAD

u2

t)

b. j2

ýa .ý3 0

4.)

tn

Z

9

0

0

ö Ici

9-

g

W 10

col)



0:

-

U

42 -=

7vý 0 u2ZE

Le -Z -E9 -

1.0.

:Dý

-A 10

2-

Z

FU

;. ä2

rz Z, ý

2 -8 78

cli

'5

ce3

19

Nce 10 E-

R

bo -2

0. s

LE "2u

rý. >

02

0

c -

2.2

10

:3

V

vi

information is insufficient for adequateCBA appraisals. We therefore reject the use of 'pricing' techniquesand turn to consider the more theoretically rigorous valuation methods. The valuation or demand-curve (Bateman, 1992) methods all ultimately rely upon individuals preferences. However, within this genre two distinct categoriesof approachcan be defined; methods based upon preferences which are revealed through purchases by individuals of market-pricedallied goods; and methodswhich rely upon expressedpreferences elicited through questionnaire surveys. Reliance upon market observationsmeans that the revealed preference techniques yield Marshallian demand curves and consumer surplus welfare estimates while expressed preference methods should, in theory, give incomecompensated,Hicksian demand curves associatedwith compensated(true) welfare measures (see subsequentdiscussion)". This researchconcentratesupon the use of one method from eachof thesefundamental valuation approaches:the contingent valuation (CV) method (expressedpreference) and the travel cost (TC) method (revealedpreference). Both of thesemethods are highly appropriate and have been extensively used for the valuation of woodland recreation externalities. Consideration was also given to the use of the hedonic pricing (HP) method and a theoretical/methodological paper was prepared (Bateman, 1993a) and a literature review undertaken". The HP method is most appropriatefor assessingthe landscapeamenity value of woodland and the author is currently undertaking such a study with colleagues. However, due to difficulties regarding obtaining data this work has only recently commencedand will therefore be incorporatedinto subsequentextensionof this research. Considerationwas given to the use of existing researchby others, however, to date only one major HP study of UW woodland has been completed (Garrod and Willis, 1992) and, after discussion with the authors, it was not felt appropriate to extrapolate these results (which referred to national averages) to the research in hand.

Similarly use of the Stated Preference technique

(Adamowicz et al., 1994) was not consideredappropriateat this stage. While promising, this is a relatively new approach which has only come to prominence in this field during the in be Further the to the method given the consideration will present research. course of

3While some discussion of this theoretical basis is given below, this is limited by spaceand further details (1993); Hanemann (forthcoming). Turner (1989); Bateman Carson in Mitchell and and and are given "Unpublished; available from author. 'The author is sceptical regarding the validity of extrapolating results acrossnational, economic and cultural borders.

2.3

future. The remainder of this chapter presentstheoretical and methodological reviews of the two chosen valuation techniques;the contingent valuation and travel cost methods. In both of thesereviews we concentrateselectively upon thoseareasof theoretical and methodological interest to this particular research. For wider ranging assessmentssee, for the contingent for Turner (1992,1993); (1990) Bateman Hanley the travel and and and valuation method; cost method, Bateman, Garrod and Willis (1992) and Bateman (1993a, 1993b).

2.2: THE CONTINGENT VALUATION METHOD 2.2.1: METHOD AND THEORETICAL

BASIS

2.2.1.1: Introduction

Hanley (1990) identifies six distinct phasesinvolved in the practical application of CV which we have interpreted as follows: Stage 1: Preparation i.

Set up the hypothetical market asking individuals either how much they are willing to pay (WTP) or willing to accept (WTA) in respect of the proposed change in provision of the good in question. Define the elicitation method. In a WTP study the major alternatives are: Open ended (OE); "how much are you willing to pay?".

This approach

least be bid therefore analysed using may and variable produces a continuous squaresapproaches(OLS). Dichotomous choice (DQ; "are you willing to pay EX", the amount X being individuals' to test the responses to sample systematically stepped across different bid levels. This approach produces a discrete bid variable and requires logit-type analysis.

A variant upon the dichotomous approachis to supplementthe initial question (see Hanemann (double-bound) iterative et al., question secondround with an 1991). Further bounds may also be used (Langford, Bateman and Langford, 1996). A further variant is to use an iterative bidding (IB) game moving from an 2.4

initial suggestedbid level to a final open-endedresponsefor which continuous

variableestimationmethodsare appropriate. Other elicitation methodsinclude the use of payment cards although these are less common in recent studies. Provide information regarding: the quantity/quality changein provision of the good who will pay for the good who will use the good. iv.

Define the payment vehicle, for example: higher taxes entrancefees donation to a charitable trust

Stage 2: Survey Obtaining responsesto the questionnaire. Interviews can be either on-site (face to face; users only), house to house (face to face; users and non-users) or by mail/telephone (remote; users and non-users). Stage 3: Calculation Calculate the mean WIP (or WTA) from responses.Somepractitioners omit 'protest' votes"',and/or use trimmed meansat this stage. In a dichotomous choice format experiment the mean is obtained by calculating the expected value of the dependentvariable (WTP or WTAY.

Stage 4: Estimation A bid curve can be estimatedto investigate the determinantsand thereby validity, of

WTP bids. For a continuousquestionformatOLS estimationtechniquesareoftenemployed. Typically, in WTP scenarios,the bid curvewill relatebids to visits, income,socioeconomic

"Respondents who refuse to state a WTP or WTA for an asset (or state extreme amounts) are commonly termed 'protest voters'. They should not be confused with the respondents who state a considered zero valuation for the good in question. A high proportion of protest votes may well signify a fundamental weakness in a study (see Sagoff, 1988; Eberle and Hayden, 1991; and discussions of strategic bias below). 7Sce, for example, Kristrom (1990a) or Bateman et al. (1995a).

2.5

factors, and other explanatory variables. A parameterto accommodateenvironmental quality of the site may also be estimated. There is no theoretically correct form for the bid function. However, if a log-log form is chosenthen the coefficients are elasticities. In such a casethe bid curve allows us to estimatechangesin meanWTP arising from changesin environmental quality. Indeed if the other relationships are sufficiently stable then we can use this curve to evaluate changesto other strongly related environmental goods, for example, the impacts of tree quality change upon overall woodland quality. If a dichotomous payment format has been used then a logit or simila? approach is required, relating the probability of a yes answer to each suggestedsum to the explanatory variables listed above. Stage 5: Aggregation This is required in order to extrapolate from sample mean WTP to total value. This

entails decisions about, for example, moving between household and individual data, and distinguishing the relevant population. Stage 6: Appraisal Was the CV successful? To answer the question posed in stage 6 we need to consider the theoretical acceptability of the evaluation estimatesproduced by CV. 2.2.1.2: Welfare change measures and the CV: a theoretical overview In estimating monetary valuesfor environmentalresourceswe are concernedwith how changes in the provision of environmental public goods impact upon individuals utility. Traditionally the welfare gain or loss from suchchangesof provision have beenapproximated by changes in consumer surpluslo; the area underneaththe ordinary (Marshallian) demand

$Alternativelya probit approachmay be used,seeCameronandJames(1987),Cameron(1988). 913ateman and Turner (1993) also briefly addresswider issuesof institutional,practicaland financial acceptability. "Referencesto 'consumersurplus' throughoutthis (and subsequent) chaptersrefer to the Marshallian consumersurplusmeasure.

2.6

curves and above the price level". The Marshallian demand curve tracks the 'full price effect' which occurs when the provision of a good changes. Typically it has been used to show how much the quantity consumed of a normal good increaseswhen its price falls. A practical problem therefore arisesin estimating the Marshallian demandcurve for an unpriced environmentalpublic good. Without private property characteristics,such as rival consumption and excludability, a good cannot be traded in a market and the price/consumptioninformation required to estimate the Marshallian demand curve will not be directly observable.One solution is to investigate a suffogate market, for example, analysing incurred travel costs as a proxy for the recreational value of an open-accessleisure site and indeed the TC (discussedsubsequently)is a consumer surplus method. However, a further theoretical problem remains in that the presenceof income effects mean that consumer surplus itself can give an inaccurate measure of the welfare changeresulting from a changein good provision. In the case of environmental public goods the individual is usually faced with a quantity rather than a price constraint, the good often being unpriced. Furthermore, these goods often have higher income elasticities than those associatedwith many market goods (Bateman et al., 1992; Kristr6m and Riera, 1996). The consequently large income effect arising from a changein quantity provision may undermine the consumersurplus measureof welfare change. In order to move from the ambiguity of consumersurplus to a theoretically more accurate measureof welfare change we therefore need to compensatefor the income effect by holding real income constant, i. e. moving from using the ordinary Marshallian demand curve to the compensated(Hicksian) demand curve. The Hicksian approach evaluates welfare change as the money income adjustment necessaryto maintain a constantlevel of utility before and after the changeof provision. Two The 'Compensating feasible for such an approach. change measures are such welfare Variation' (CV) is the money income adjustment (welfare change) necessaryto keep an individual at his initial level of utility (UO)throughout the change of provision, while the 'Equivalent Variation' (EV) is the money income adjustment (welfare change) necessaryto maintain an individual at his final level of utility (UI) throughout the provision change.

"Price may be zero or positive dependentupon the property rights of the good. In the caseof environmental introductory For faced text see constrained, public goods. quantity an unpriced, with goods we are usually Johansson(1991) and for further reading sce Just ct al. (1982) and Johansson(1987).

2.7

We therefore have two approaches to measuring welfare changes. Furthermore these changes can be either positive (a welfare gain) or negative (a welfare loss) giving us four possible scenarios. For a proposed welfare gain (i. e. a change in provision which increases utility, e.g. more recreation; less pollution; etc.) the CV measure tells us how much money income the individual should be willing to give up (WTP) to ensure that the change occurs" , while the EV measure tells us how much extra money income would have to be given to an individual (WTA) for them to attain the final improved utility level in the absence of the provision change occurring".

For a proposed welfare loss (i. e. a change in provision which

decreases utility, e.g. less recreation; more pollution; etc.) the EV measure will now show how much an individual

is WTP to prevent the welfare loss occurring"'

while the CV

measure now shows individuals WTA compensation for allowing the welfare loss tooccu? 5.

These variation measures(CV and EV) only strictly apply where the consumeris free to vary continuously (i. e. non-discretely) the quantity of the good consumed. Where the consumer is constrained to consume only discrete or fixed quantities (as for most environmental public goods) then we should consider compensating surplus (CpS) and equivalent surplus (ES) measuresin place of CV and EV respectively. Bateman and Turner (1993) discussin more detail the relationship betweenwelfare measuresfor price and quantity constrained goods. The upper panel of figure 2.2 shows a utility curve analysis of welfare gain and loss measuresin the context of an unpriced,quantity constrainedenvironmentalgood X,. Provision of X, is shown on the horizontal axis while the vertical axis shows income as a moneycomposite of all other consumption X0. BecauseX, is unpriced, the budget line is shown as R the horizontal line with initial consumption of X, being quantity rather than price constrained at Q0 corresponding to point A on initial utility curve UO". Suppose that a

'i. e. the loss of money income which, after the increasein provision, returns the individual to his initial lower utility level.

"i. e. the increasein moneyincomewhich raisesthe individual to the samefinal utility level as if the foregonewelfaregain in provisionhadoccuntd. "i. e. the maximumamountof moneyincomewhich the individual is preparedto give up to preventthe welfarelossoccurring,leavinghim as well off as if it hadoccurred(at the final, lower utility level). "i. e. the increasein moneyincomewhichreturnstheindividualto his initial (higher)utility levelgiventhat the welfarelosschangein provisiondoesoccur. 'Note thatequilibriumis not achievedat a point tangentialto a utility curve. Althoughtheindividualwould prefer to be at sucha point (i.e. morealongX from point A to a tangentialpoint with a higherutility curve), consumptionof X, is exogenouslyconstrainedat Q0.

2.8

welfare gain is proposed,increasingprovision of X, from Q0to Q1. This is shown as a move from point A on UOalong the budget line to point B on U,. This correspondsto the full price effect shown by the Marshallian demandcurve DD in the lower panel and the corresponding increase in consumer surplus shown by the shaded areas b+c.

Despite X, being itself

unpriced, its increasedprovision will still have an income effect by releasing some of that income previously spentupon priced goods(e.g. if Q is recreationthen its increasedprovision relieves spending upon other priced recreation goods). Consumer surplus is therefore only an approximate measure of the true welfare change. We can compensatefor the income effect and obtain a correct welfare change measureby asking how much the individual is WTP to ensurethat the increasein provision does occur. The individual should be prepared to give up the amount of income BC which returns him to point C on his initial utility curve UObut with the increasedprovision Q1.The correspondingcompensateddemand curve h0ho is shown in the lower panel and it is the shaded area c under this curve which correctly measuresthe welfare change for this scenario (CpSwTp). Now supposethat the sameproposedwelfare gain (QOto Qj) is not implemented. The authorities could still raise the individual's utility from UOto U, by increasingmoney income by the amount AD (the equivalent value of extra income which individuals are WTA to forego the welfare gain changein provision). This moves the individual to point D on U, and maps out the compensateddemandcurve hh, in the lower panel. The correct welfare measure for this scenario is therefore the equivalent surplus ESwTA(the shadedarea a+b+c in the lower panel). Note then that for a welfare gain we have CpSwTP < consumersurplus < ESWTA, in short WTP < WTA. Now consider a proposed welfare loss, say a decreasein the provision of the same

unpricedenvironmentalgoodfrom Q, to Q0. Herethe individualwill startat point B andthe initial utility curve will be U1. Facedwith a fall to point A on new utility level UOthe individual will be WTP the amountBC to avoid the loss (ESWrp)17.However,if the welfare loss changein provisiondoesoccur,then the authoritiescan still compensate the individual by giving him extraincomeAD to returnhim to his initial utility level U, (CpSwiA)". Note

"'This is an equivalent surplus measureas the welfare changeis measuredfrom the new utility curve, here UO. "Similarly this is a compensatingsurplus measureas the initial utility level, U1, is our welfare measure reference.

2.9

that for the welfare loss we now have ESwTp< consumer surplus < CPSWTA.Therefore, for

either gains or losses, the WTA measureexceeds WTP, however, the derivation of these measures (i. e. CpS or ES) changes'9.

Figure 2.2:

ComPensated welfare change measures for an unpriced quantity constrained good

Consumption of

ail

other

X

goods (income) Welfare n __Ga. WTA (ES)

Welfare _Lo _S_ WTA (CPS)

ý (Budget

Line)

WTP (CPS; ý WTP (ES)

o',

0,

0,

Q,

-Environmental X, Good

Price (E)

P,, (unpriced)

Source:

Bateman

xI Environmental

Good

& Turner (1993)

In summary, as we have seen, there are theoretical problems with the consumer surplus measure of welfare change. Yet, because of the impossibility

of mapping utility

functions, consumer surplus measures have often been calculated as best practical estimates

"Bateman and Turner (1993) present an expenditure function approach to assessmentof these welfare measures.

2.10

of welfare change. Ibe CV approach,in eliciting explicit statementsof how much income consumersare WTP to ensurethat a welfare gain occurs (or prevent a welfare loss occurring) or how much income they are WTA to endure a welfare loss (or forego a welfare gain) is, in theory, directly estimating the true Hicksian welfare measures of these changes (see Bateman and Turner (1993) for formal proof). Although in later sectionswe addressseveral important methodological criticisms of the empirical method, this theoretical ability to estimate true welfare measures represents a considerable potential advance over other approachesand deservesemphasis. 2.2.1.3: Theoretic and empirical asymmetry of CV welfare measures: WTP v. VVTA In planning the empirical CV research presented in later chapters an initial fundamental question arises regarding which of these alternative welfare measures is most appropriate for the assessment of open-access woodland recreation benefits. In the majority of CV studies this has translated into the problem of whether WTP or WTA measures are best fitted for such a task. At first glance we might have expected there to be no difference in the amount which consumers would be VvrrP for a specific welfare gain compared to the amount which they would be WTA in compensation for an equivalent loss, indeed certain aspects of neoclassical utility theory might well lead us to expect such a result (Schoemaker, 1982). However, as figure 2.2 illustrates, there is a theoretical asymmetry between WTP and WTA measures such that WTP for the welfare gain (i. e. move from A to Be; CpSwTP)is exceeded by WTA compensation for the welfare loss (i. e. move from B to A; CpSwTA).

In his seminal articles, Willig (1973,1976) showed that, for priced normal goods in most plausible situations, the deviation between compensating and equivalent variation measuresshould be relatively small (thus promoting consumer surplus as a valid welfare measure).The Willig limits suggestthat Hicksian WTP and WTA measuresshould generally lie within 2% either side of the Marshallian consumer surplus. These results using Hicksian analysis were formulated for price changesand Hicks (1943) shows that this asymmetry is, in theory, slightly more pronounced for unpriced goods subject to quantity constraints (see Raternanand Turner, 1993). Nevertheless,theselimits in no way provide a theoretical explanationof the very wide WTP/WTA asymmetry found in empirical testing. Table 2.1 shows that in practice CV

studieshave Tccordedvery wide divergencebetweenWTP and WTA raising considerable 2.11

concern about the validity of the method. We therefore need to consider whether such a is indicative pronounced empirical asymmetry of a fundamentally flawed methodology or whether it has any theoretical plausibility. Table 2.1: Empirical divergenciesbetween WTP and WTA Study

WTAIWTP

Knetsch & Sinden (1984) Coursey, Schulze & Hovis (1983)

Brookshire, Randall & Stoll (1980)

4.0 (i)

3.8

(ii)

1.6

(i)

1.6

(ii)

2.6

(iii)

6.5

Bishop & Heberlein (1979) Banford, Knetsch & Mauser (1977)

4.8 (i)

2.8

(ii)

4.2

Hanunack & Brown (1974)

4.2

Source: adaptedfrom Pearce& Markandya (1989). Reverting back to variation measuresto avoid discontinuity problems, the Willig formulae can be approximated as follows (Varian, 1984)": cs

-

cv

----------1CS1 where

I CS1

71

(2.1)

------2 Y'

CV

= compensatingvariation

CS

Marshallian consumersurplus =

11

income demand of elasticity =

YO

initial income (expenditure) =

2"Theapproximation formula is only valid if I Cý /Y* is less than 0.9 (Boadway and Bruce, 1984: pp.216220), i. e. expenditure on the good (and the associatedwelfare measures)cannot be too high relative to the consumersincome if this is to hold.

2.12

Willig (1973,1976) shows that, for the priced good case,such errors are likely to be small2l. However, this error will clearly increase with greater income elasticity'.

More

importantly in the environmental context, when we consider unpriced goods then the income

elasticity of demand term is not strictly relevant. Randall and Stoll (1980) show that income elasticity (TI) should be replaced by the 'price flexibility

of income' (c) and reformulate the

Willig limits as (for a welfare gain i.e. CV is given by WTP): cs - WTP -------------cs

ccs =

(2.2)

----2Y

and ECS2

WTA - WTP - ------y

(2.3)

where: CS

=

Marshallian consumer surplus

WTP

Willingness to pay: CV for a welfare gain (CpSwT?for a noncontinuous consumption function)

WTA=

Willingness to acceptcompensation:EV for a welfare gain (ESwTAfor a non-continuousconsumption function)

Y

Mean respondentsincome

C

Price flexibility of income

Randall and Stoll (1980) estimate c in a manner analogous to an ordinary income elasticity by regressing WTP upon the quantity of the good, income and other significant explanatory variables. From this they estimatedthat, under reasonableassumptions,measures of WTP and WTA for quantity constrained goods should be within 5% of each other. CV practitioners concluded from this that the wide empirical divergenceof WTA aboveWTP was merely a methodological glitch which could effectively be ignored and that WTP sums were

21Aresultconfirmedby Justet al. (1982)who alsoshowthat theWillig approachmaybe gencralised, to the multiple price changecase(pp.375-86). 'rhis error will alsoincreasefor aggregate populationswherethereare largevariationsin incomeand/or incomeelasticityof demandbetweenconsumers.

2.13

valid approximations to WTA (e.g. Desvousgeset al., 1983). In a significant re-analysisof theory, Hanemann(1986,1991) shows the Randall and Stoll (1980) derivation of the price flexibility of income (e) to be inexact, demonstrating instead that:

(2.4) cr where Tj = income elasticity for the environmental good a= elasticity of substitution between this and all other goods Using what they term "the not-too-unreasonable values of, say, 71= 2, and cy = 0.1", so that F,= 20, Mitchell and Carson (1989) apply the above formulae to their earlier empirical work on the evaluation of water quality improvements (Mitchell and Carson, 1981). In this work they found an average WTP of $250 (with average income = $18,000).

We can

therefore rewrite equation (2.2) as: ECS' - 2Y. CS + 2Y (WTP) = 0. Substituting in values

for C, Y and WTP gives consumer surplus (CS) = $300 and substituting this into equation (2.3) gives WTA = $350. On the basis of theseassumptionsWTA is shown to be some40% larger than WTP in this example. Furthermore, while they state that higher values of Tj are unlikely, Mitchell and Carson (1989) state that "much smaller values of a for a number of public goods are quite plausible". Using the same empirical data we can deduce that, for WTA to be double WTP, requires cr = 0.0625; while for WTA to be triple WTP requires cr (some Such describe 0.05. substitution progressively superior goods elasticities = environmental goods appearto fit this profile rather well). In an important extension of his work in this area,Hanemann(1991) simulatesWTP 23 levels for WTA CES and a generalised utility model under a variety of assumptions . Hanemannconfirms the inverserelationship betweenthe elasticity of substitution measureand the WTA/WTP ratio, i. e. for unique and irreplaceableenvironmental goods (Hanemanncites Yosemite National Park as an example) with very low elasticity of substitution. In this

'See Deatonand Muellbauer(1980)for detailsof this and otherutility systems.

2.14

context, we should expect WTA to be much greaterthan WTP. Hanemannalso demonstrates that the sameresult still holds with a much higher elasticity of substitution ((T- 1) where the ratio of WTP to income is high, i. e. where the proposedchangemattersa lot to the individual concerned. Under both these scenarios,Hanemann demonstratesthat standard theory can explain levels of WTA more than five times the magnitude of WTP. Furthermore the Hanemann formula confirms that, where elasticity of substitution is not low and the WTP/income ratio is not excessively high (a scenario typical of many market priced private goods), then WTP and WTA will not diverge very significantly. These findings extend rather than refute the original Willig limits. Indeed they show that the observedWTPIWTA asymmetrydoes have a theoretical basis and we should expect such asymmetry to occur where we are evaluating environmental goods which are in some significant way unique, irreplaceableor lacking substitutability. Such asymmetry,rather than being a methodological glitch, should actually be interpreted as theoretical backing for the internal consistencyof the CV. While there appearstherefore to be a strong case in economic theory for observed empirical CV results, we also recognisethat other argumentsstemming from the literature of psychology have been put forward to explain the apparent WT? /WTA asymmetry. We highlight three such argumentsbefore formulating our conclusions. z) Rejection of the WTA property right In a WTP format experiment, respondentsmay feel that they (or the CV researchers) have no right to, in effect, sell the environmental good being valued. This result in either a refusal to give a WTA sum, i. e. a 'protest vote' (Sagoff, 1988; Eberle and Hayden, 1991),or an inflation of that sum so as to indicate that no level of compensationis acceptable,thereby preventing any loss of the good. Bishop and Heberlein (1979) note that such protest votes are far less common in 'real' WTA situations where respondentsare actually offered cash compensation, indicating that this may in some way be a methodological artifact of hypothetical markets (see subsequentdiscussionsof hypothetical bias). Nevertheless, there is evidence that respondentsdo perceive public goods such as environmental assets in a different manner to their treatment of private goods. Turner (1988a/b) arguesthat individuals possessboth private and public preferences. This arisesout of the complex array of diverse serviceswhich environmental goods can exhibit. The Total 2.15

Economic Value concept (Pearceand Turner, 1990a)incorporatesconventionalutilitarian usevalues with option' and non-use (bequest and existence) values. Turner (1988a/b) argues that the combination of private and public preferences inherent in the evaluation of environmental goods is fundamentally distinct from the market pricing of a private good. In particular the respondent will value the continued preservation of environmental assetsfor enjoyment by future generations. This may in turn causea rejection of the compensation-forloss principle inherent in the WTA question. U) Inexperience and risk aversion Market prices are the result of consumersrepeatedevaluations of goods. However, the CV scenario effectively gives respondentsonly one opportunity to evaluate what is often do have Respondents high-preference the advantageof past experience therefore not good. a (1987) Hoehn Randall in determining their and argue that in such valuations. to call upon information, WTA imperfect tend to aversion will raise respondents risk of situations Such by in the them to continued provision of good. ensure a an attempt responses (1986) it in Coursey WTA by tests that al. et where was shown supported was proposition decline CV i. trials, for tended to over repeated e. as valuation good a particular sums into fed back the evaluation process. experience iii) Prospect theory The neoclassical Willig-type divergence between CV (WT? ) and EV (WTA) is illustrated as the smooth 'standard evaluation curve' in figure 2.3. Here an individual is initially at the origin with income Y' and an initial non-zero allocation of an environmental q. Q) for increment in Q (from QO Under Q theory, to the an to standard equal good individual has a WTP equal to the distanceYWTP, while for an equal decrementin Q (from QOto Q) the individual has a WTA equal to the distance YOWTA. The relevant factor here is the smooth nature of the standard evaluation curve and the consequentrelatively small

divergencebetweenWT? and WTA. In their 'Prospectlbeory', Kahnemanand Tversky (1979) postulatethat individualswill havea psychologicalaffinity for the statusquo such

24Typicalreferencesto option value include Wcisbrod (1964); Cicchetti and Freeman (1971); and Krutilla (1990b) Kristr6m Umt However, (1975). the concept of an option value can be traced back Fisher argues and to jeavons in 1888.

2.16

that, whilst they may be willing to pay for increments,they are very unwilling to contemplate 25 initial in in allocation of the good question . In such a model the Prospect a reduction their Theory evaluation curve is kinked at the initial allocation 'reference point' such that WTA is related not to WTP but to that referencepoint and exceedsWTP very significantly. In such for from be losses transfer traded off, example, a one readily cannot a system gains and individual to another would, given common preferences and endowments, always lower collective utility. Figure 2.3: Valuation of changesin the provision of an environmental good INCOME LOSSES (WTP)

STANDARD EVALUATION CURVE

DECREMENTS IN 0

PROSPECT THEORY EVALUATION CURVE

WTP

0-

()

45*

INCREMENTS IN 0

/0 x

0/I

WTA INCOME GAINS (WTA)

Source:

Bateman (1995a) adaptedfrom Brookshire et al. (1980), Kahnemanand Tversky (1979),Jones-Lee(1989)

2Me roots of such an idea may be traced back to Adam Smith (1790) who statesthat "We may suffer more, it has already been observed, when we fall from a better to a worse situation, than we ever enjoy when we rise from a worse to a better".

2.17

These ideas are developed by Kahneman et al., (1990) and Tversky and Kahnernan

(1991) into a model of reference-dependent preferences.In recentwork by the authorand colleagueswe tested this model using real tradeswith private goods, i.e. removing the hypotheticaland public goodsissueswhich were suspectedto be the causeof thesenonstandardresults. However,our findings clearly showedthat even under thesecarefully controlled conditionssignificantreferencepoint effectscontinuedto occur (Batemanet al., 1995band forthcominga).26 Given this we can no longer identify the natureof the CV Rather these this seemsa generalproblem for the cause sole of effects. as experiment throughpoorCV design)whichwe address microcconomictheory(whichmaybeexacerbated in other researchand is not within the remit of this study. 2.2.1.4: Determining

the appropriate

welfare measure

The above review suggeststhat both economic and psychological theory provides This WTP if WTA commonality of effect sums. causes may exceed problems reasons why As CV have two the a consequence measures. practitioners using to results compare we wish (Mitchell Carson, 1989; formats Harris WTP for scenarios provision gain and and and argued Brown, 1992) as a method of reducing psychological effects and thereby enhancingeconomic validity. While this generally seemssensiblethere is a potential credibility problem where the is format. is demonstrably the expected and compensation payment negative provision change Consequently in our subsequentresearchwe generally use a VTP for gain' approachwhen is in for loss' VTA However scenario a small experiment a used assessingrecreationalists. levels for farmers providing such woodlands. compensation required assessing 2.2.2: METHODOLOGICAL

ISSUES

2.2.2.1: Introduction Those methodological issuesmost pertinent to the CV can be Toughly divided into

(Bateman 1991). Validity degree bias the to al., et refers categories and validity, reliability indicates 'true' CV the correctly the evaluation value of the asset under to which investigation, bias being a common causeof low validity. Reliability refers to the consistency 26Notethat Petersonet al. (1996)producecontraryresultsin which transitivity is not violated. However, it different is as comparespublic with privatedecisionmaking. somewhat their experiment

2.18

or repeatability of CV estimates. Of coursereliability and validity need not be synonymous, for example, a particular CV instrument may, in repeated trials, yield a consistent value if However, for thesetrials are all subject to a bias then the results asset. estimate a particular will not be valid. We begin this review by a brief considerationof reliability issues. Problems of bias interest in in depth Ile the this academic area. greater reflecting then are addressed issue is then the concluded with an assessment of of validity. review methodological 2.2.2.2: Reliability In CV surveys reliability is associatedwith the degree to which the variance of WTP inversely being be to with reliability error, related to the random attributed responsescan degree of non-randomness.Notice that reliability saysnothing about the validity of estimates. Variance in WTP responsesderives from three sources;true random error; sampling itself (instrument True is variance). random the error questionnairrfinterview procedure and is induced sampling procedure error a potential to the process while statistical essential be by in inherent and can acceptably usually minimised statistical survey any problem is instrument is It is size variance used. which significant sample that statistically a ensuring here. of most concem Assessing Reliability Several commentators (Mitchell and Carson, 1989; Kristr6m, 1990a; Hanley, 1990) CV test the a of scenario as reliability of particular of a the retesting subsequent advocate is (1990a), "If Kristr6m According initial the from to same experiment test. an estimates different careful statistical analysis samples and reveals no times with of number a repeated is flag" indicating low between then this the collected a warning variables correlation high due have been Few to the tests out, resource carried mainly replicability such reliability. Carson is by (1995) However, involved. al., et who exception provided one notable costs 1992) WTP Prince William (Carson findings to conceming al., protect et retest their earlier Sound, Alaska from future oil spills like that from the grounding of the Exxon Valdez on 24th March, 1989. The two studies involved independentsamplesof interviews taken over two distributions did differ between the that these response not significantly concluding apart years by Loehman Heberlein De (1982); (1986); Loomis (1989, Other and studies two samples. 2.19

1990); Carson and Mitchell (1993) and Epp and Gripp,(1993) generally support the reliability ' Unfortunately time and resourceconstraintsprohibited retesting of our CV instruments. of is further consideration given to this area of research. empirical results and so no 2.2.2.3: Bias Is-sues is inherently CV is an expressed-preference and as such susceptible valuation method instrument into have bias cognitive, subdivided procedural and which we to various types of bias.

General Bias' i. Introduction: The Valuation Process Standard neoclassical economic theory is based upon a relatively simple model of "rationale economic person", dominated by the concerns of self interest. Under such a be donations seenas enlightened self-interest to groups can environmental utilitarian theory individuals impure towards the altruism quest of a glow while charitable acts contribute warm function (Andreoni, 1990). Here income-constrained his valuesare simply utility to maximise from individual the those precepts and measure of arising such preferences the of reflections in for individuals to the in question. or service good pay willingness the values Many commentatorshave criticised this rather simplistic definition of human nature. highlighted have the multifaceted discipline critics From the of environmental economics by the Many those environment, provide much provided particularly goods, nature of values. in (Pearce implicit instrumental basic models and simple utilitarian values use the than more from fundamental the crossover of environmental A comes 1990a). critique Turner, more individuals of models economics with psychology and resultant attitude/statement/behaviour 2.4. figure in illustrated as

"Mitchell and Carson(1989)alsohighlightsimilar findings from non-CVMsurveyre-tests.

2.20

?5 .2

C) A

0. (n

(L) (2

C13

1m4,1

c

(1) -1 c:

a' 11

E

.E

CA Gn

CL a, >

ci cii

cö mm a)

cil >

I

U) (L)

Q)

(0

0c

00 ZE

A,

Ici

1e OM

.2EE > (0

25 ci (D c

C). Z

Oui .2 Z >, cm ZEm Co

-Q G) 10 ID (U

12

c0C

ýE

cm CD

-1

1

73 m

1

cu

0

rýM

E0 Z.

2

1 o5 (D 0

00 Z r-

=

4b-

«ý;

'ýK 1 -

'o(

0 'ý; «a (D =

ý--

cu Co :l «a

a) *=5

0) M

'Zi ö0 .Z

(L) Z (L) > rr Q ;:

en m

> CL >0 Ce , ý cu :3

.5

=0=!! cu 0,'0z > ý; 13). CL

0

-2 0 N *Z

cu E 02£2 Cl250 CL) Mm

CL =

03 CoZ

I

I,

0

Gor :i

cn xE

0

CA

The attitude/statement/behaviourmodel illustrated in figure 2.4 draws upon a number of sources(Fishbein and Ajzen, 1977; Hoehn and Randall, 1987; Mitchell and Carson, 1989; Harris and Brown, 1992; and Bateman and Turner, 1993). It presentsa more complex and realistic view of the individual than that underpinning 'rational economic person' allowing both CV the to the to pre-existing cognitive state of survey, and respondents prior us consider how that state alters as a result of the experiment. In our attitude/statement/behaviourmodel the pre-survey base-stateof the individual is formed from the set of prior information held by that person. This base-stateis influenced by two dialectics through which information is interpreted: (i) positive versus normative beliefs (what is and what should be); and (ii) the individual between the the of preferences private and those of the public citizen. Ibis schism latter factor may be one of considerableconflict but is of major relevance to CV research few 'citizen those approaches capable are of that other potentially estimating given ignored by (Blarney, 1996). If the market we generalisesomewhatwe so often preferences' influence beliefs these that citizen preferences will normative public while private can argue 28 beliefs impact (Peterson have 1996). individual preferenceswill upon positive et al., most In the CV experiment the respondentis presentedwith new information which will be used to update the belief set. These beliefs will then form the individuals attitudes and into feed beliefs Information, behaviour. then all motivation. and attitudes norms concerning from bequest) (existence non-usemotives such as arise is values and It arguable that non-use from beliefs drawing 1987) arise whereas use values upon normative altruism (Randall, be likely However, to the of main routes beliefs these are while attitudes. and positivist ideas instrumental imagine positivist goods and influence we can also norms concerning concerning non-use values. These use and non-use motives combine and are expressedas the WTP sum within CV below). T'his (discussed the experience and statementof value the CV valuation process into directly (an behaviour back feeds more usually, itself then actual payment) or, either via beliefs. individuals normative positive and the The transition to formulated and then statedvalue is the subject of theoretical analysis CV facing Here is (1987). Randall the two respondent stage seen as a by Hoehn and first (ii) In formation (i) the the statement. stage respondent value value process valuation individuals different functions for 2OPeterson (1996) that work suggesting empirical use utility report al. et decisionmakers. to they asked are act as public agency to those when used private choices

2.22

attempts to determine his YV7Pfor a good. Hoehn and Randall consider two factors inherent in CV style surveys: (i) imperfect information (and resultantuncertainty); (ii) time constraints. In the presenceof either factor the individual will formulate a WTP which is lower than true WTP given perfect information and no time constraint. In moving from formulated to stated in has the to a variety of strategic behaviour opportunity engage the respondent value including

both understatement and overstatement of

formulated WTP (discussed

being Randall Hoehn these strategies as chosenaccording to and see various subsequently). Accordingly focus being (011; DC; one of the empirical etc). used the elicitation method is impact in the of elicitation method upon responses. chapters work presented subsequent In summary we would argue that our attitude/statement/behaviourmodel is more less but amenableto simple predictions than consequently and also more complex realistic The "rational theory. standard neoclassical underlying need to person" economic the that of formalise such complex models and thereby improve the credibility of economic models of behaviour is an ongoing area of economic research(as witnessedby the contemporaryvogue for experimental economics). However, such research lies beyond the remit of this study focusses (1987), Randall Hoehn the following upon valuation stageof our model and which, in CV include formulated to between and studies on actual our and stated value spanning behaviour in our TC research. In the remainder of this section we consider the various general biases which may formulated from to statedvalue. transition affect the ii. Hypothetical Bias Irrespective of the chosen elicitation method, a basic question concerns whether the is itself induces bias into There the question. hypothetical nature of the contingent market debate about the very nature of any such hypothetical bias. Freeman (1986) seesthe impact increased bid Mitchell being hypothetical and increasingly while variance, as scenario an of Carson (1989) extend this to reject the entire notion of hypothetical bias referring instead to low (discussed bias subsequently) and model reliability. and situations of recognised specific However, many commentators(Schulzeet al., 1981; Bishop et al., 1993; Randall et al., 1983) in hypothetical than the markets can certain that use of real rather are convinced distinct bias its own problems. produce circumstances Our opinion is that discussionof a distinct hypothetical bias is unhelpfully imprecise. 2.23

What is clear is that the hypothetical market presentsrespondentswith the opportunity to shift their stated WTP away from formulated WTP. The real issue, we feel is whether the strategiesinvolved conform to a perhapsextendedview of economic theory or to some other competing psychological theory. In the former case we can accept stated WT? as having if link true the latter is true CV results are acceptable with value while some economically benefit indicator. some cost as not admissible Researchinto the predictive ability of hypothetical markets has followed two paths: studies of the attitude/statement/behaviourrelationship; and experiments examining the for hypothetic markets. real substitution of Market research, political polls, consumer surveys and our own attitude/statement/ behaviour model all operateon the premise that statedattitudes or intention are significantly in behaviour. In influences indicators the model, each stage the our cycle of actual reliable for is influence However, this example, attitudes may not perfectly not perfect, next. dynamic loops feedback behaviour, the provide a adjustment system so while predetermine that, for example, a recent visit to the countryside may well affect a respondent's WTP to however, is This, (transformation habitat value). a reflection of reality preserve wildlife in present the consumption of all goods,marketedor not, and neednot pose a specialproblem for the CV technique. Ajzen and Fishbein (1977) develop three hypothesesfrom their model indicating how best behaviour Firstly, link be attitude will predict maximised. can the attitude-behaviour for improvement WTP So, a general environmental asking where the two closely correspond. improving for higher indicator WTP the taxes be water quality of a specific of poor a will interviewer by by the Scenario as constructed or as perceived either misspecification, river. fewer intervening between bias. Secondly, the the stages the respondent,will obviously cause behaviour, in the the the that greater predictive and power component model of component a (Mitchell and Carson, 1989), i. e. statedvalue is a better predictor of behaviour than attitudes because in fewer beliefs both better behaviour than cases predictors while attitudes are influence relationships are involved. In a study of unleadedpetrol consumption,Heberlein and Dlack (1976) found an attitude-behaviourcorrelation of just 0.12 but a statedvalue-behaviour Thirdly, be behaviour better 0.59. attitude will a predictor of of when the correlation

2.24

respondent is dealing with familiar behaviour situationS29.Hanley (1990) sees this as a source of error with respect to environmental goods where, unlike marketed goods, there is by he feels in learn However, to that, the main, this experience of purchasing. no opportunity be WTA be scenarios, respondents will associatedwith where very unfamiliar with error will the selling rather than purchasing role, and less significant in WTP situations with which both have greater experienceand empathy. respondents Bateman and Turner (1993) review a number of studiesexamining the potential effect 1976; Bishop CV (Schuman Johnson, hypothetical and of markets andHeberlein, the nature of 1979; Hill, 1981; Rowe and Chestnut, 1983; Bishop et al., 1984; Bishop and Heberlein, 1985; Heberlein and Bishop, 1986; Brookshire and Coursey, 1987;Dickie et al., 1987; Mitchell and Carson, 1989; Kealy et al., 1990). While there is empirical evidence for a divergence between formulated and statedvalue may be significant (particularly in WTA formats) there hypothetical bias. be Following to these the some that attributed could evidence was no clear Hoehn Randall (1987), key factor the theoretical the of and analysis assumptionsunderpinning in eliminating pure hypothetical bias is the credibility of the CV scenario and payment (e. free-riding; biases holds Where may still operate g. this see recognised obligation. have i. but discussion) theoretical the still obtained will validity e. they measures subsequent is However, WTP. but there true biased to a credibility gap and/or where still related are discussion) CV (e. biases effects; see subsequent anchoring operate g. where psychological indicators. benefit-cost be valid results will not In formulating appropriateguidelines, Rowe and Chestnut(1982) arguethat a credible CV instrument must bOo: informative; clearly understood; "realistic by relying upon "have institutions"; legal behaviour to and application all uniform and of patterns established for CV from further The these scenario moves precepts, that a particular respondents". its is familiar less the the the construct of valuation, with good or respondent example, the have low Realism instrument it is likely and credibility. that will an such then the more familiarity are therefore at a premium in undertaking CV studies and have been a major

29AjzenandPeterson(1988)extendtheseattitude-behaviour thatbehaviourmustbe criteriaby emphasising intention lags between that the the of measurement andpredictionof respondent, of control the under volitional of intentionand behaviourshould behaviourwill be problematicand that levelsof generalityin the measures assetasa predictorof intentionstoward be identical,i.e. theextrapolationof intentiontowardsoneenvironmental dubious. is highly set asset a wider Conditions' Optimal Cummings (1986). 3OSee 'Reference of the et al. also

2.25

objective in our design of the field experiments discussedsubsequently. However, even if credibility is attained other biasesmay still arise, to which we now turn.

W. Understatementof WTP If an individual feels that a good will be provided irrespective of his responseto a WTP question, or that the payments of others will be sufficient to secureprovision then he interest in less have "pretend to a given collective activity than he really has" will (Samuelson, 1954) and will understatehis WTP for that good, i. e. he will free-ride (MarweIl 1982). A be 1981; Brubaker, feel Ames, obtained similar result will where respondents and that actual paymentswill (or should) be related to cost-sharesrather than to WTP (Hoehn and Randall, 1987). Here respondentswill state the expectedcost if this is less than WTP and zero otherwise. There have been a number of empirical investigations of the free rider theorem. Brookshire, et al., (1976) and Schulze et al., (1981) argue that, assuming that true YVTP bids free-riding disturb in distributed, WTP this then should causing, a are theoretically normally Using Brookshire (1976) bias distribution towards such an approach, al., et zero. scenario, a bias. However, (1980) Rowe for the et al., criticise the of strategic presence test and reject bimodal distributions be that test can a stating posited upon underlying assumption of such in Certainly large income the a sample population. recent respondent characteristics of the income bid highly distribution. found (1992) Bateman skewed and et al.,, experiment,

A more

bid for (1982) by Brubaker is to asked where respondents were a adopted typical approach $50 shopping voucher under three scenarios, S, in which the n highest bidders were be for in S2 told that provided were vouchers would respondents which guaranteed a voucher, in S3 WT? long specific amount, and which of all respondents exceeded a as the total all as WTP that those told giving any positive all would receive a voucher. respondents were Brubaker assumed that S, would provide the true WT?, while S. had a weak incentive to freeThe WTP free-ride S3 mean results expected. response was where a strong ride compared to $27.07>(S3 $23.96). bear These $33*99)ý4% (SI to appear results out the = = = were behaviour. However, with further analysis, only the first two, S, and strategic of expectations S2. are significantly different at the 5% confidence level. This experiment tends to indicate it does less be free-riding to occur, appears prevalent than standard neoclassical that, while invalidate CV Table 2.2 from not may and compiles exercises. results theory would predict

2.26

a number of these studies.

Table 2.2:

Stated WT? as a percentage of true WTP in the presence of a free-rider incentive Study

Percentageof true WTPl

Schneider and Pornmerehne(1981)'

96

Marwell and Ames (1981)2

84

Brubaker [Sj (1982)2

80

Christiansen (1982)'

79

Bohm (1972)2

74

Brubaker S3] (1982)2

71

Schneider and Pornmerehne(1981)'

61

Notes:

1. 2. 3.

The true WTP being measured in an auction where the winning bid(s) received the good. In these experiments a group threshold WTP was required for provision, i. e. there was a relatively weak free rider incentive. In these experiments provision of the good was guaranteed irrespective of the (non-zcro) WTP sum offered by the respondent, i. e. there was a relatively strong free-rider incentive.

Source: Adapted from Mitchell and Carson (1989) Table 2.2 indicates that where respondentswere told that a certain threshold total WTP (weak free-rider incentives) before from the the good was provided population was required i. WTP free is between 71-96% WTP true the of e. extent of riding was then stated

lead (Mitchell Carson, 1989). less to theory than might us predict and standard considerably The fact that stated WTP is still somewhat below 'true' WTP in such situations is not Hoehn Randall (1987). Not the theoretical of and conclusions with accords surprising and irrespective WTP in guaranteed was of stated those provision experimentswhere surprisingly (i. e. strong free rider incentive) a larger deviation between stated WTP and true WTP was Barnett Yandle (1973) Such to the appear support conclusions of and results and observed.

Garrod andWillis (1990)that free-ridingshouldbe addressed Yia a propertyrights approach in which respondents receiveprovisionof a goodrelativeto their WTP. However,as these limited by the characteristicsof many environmental strategies are such out, point authors

2.27

public goods"'. An important caveat to table 2.2 is that all the results presented appertain to OE elicitation formats. As previously discussed,Hoehn and Randall (1987) show that, even in the absenceof any free-riding, the lack of an overstatementincentive, imperfect information in time result will understatementof WTP. Interestingly, in the same paper and constraints 2. formats for DC is using as a method of combatting such understatemene a case made Providing that respondentsbelieve that they will pay the DC bid level proffered to them (i. e. conditional upon instrument credibility), then they will only refuse a bid level if it exceeds their formulated value, i. e. there is no theoretical incentive to understateWTP. Indeed, as the individually (i. it, "in DC model with parametric referendum costs a policy e. put authors format), truth telling is the optimal strategy" (Hoehn and Randall, 1987; parenthesesadded). Clearly the potential for deliberateunderstatementof VVTPexists although it appears formats. Accordingly in OE investigation WTP likely elicitation to of occur more in our applied work. priority understatementwas made a research iv. Overstatementof WTP Bateman et al. (1995a) identify five factors which may induce a respondent to discuss further below: CV in WTP of which we each experiment, a overstate L

Strategic overbidding (all elicitation formats)

ii.

The 'good respondent' (all elicitation formats)

iii.

Upward rounding (DC formats)

iv.

Anchoring (DC formats)

V.

Starting point effects (IB formats).

Strategic Overbidding:

In an important empirical paper, Bohm (1972) argues that

in free-riding, WTP hypothetical their of respondents may overstate the prediction to contrary feel factual 'strategic that Such their respondents may occur where overbidding' markets. individual payment will be related to some sample measuresuch as mean WTP rather than if In formulated WTP, WTP then such a case, exceeds expected mean their own statements. in improve inflate WTP (up to to the stated the may expected an mean) effort the respondent

"Note that a further factor influencing the apparentlack of free-riding in table 2.2 might be the public citizen in individuals by felt respectof environmental goods. obligations 32SiMilarclaims are made by Loomis (1987) and Kristr6m (1990a).

2.28

probability of provision. Apart from Bohm's original study there is little empirical evidence regarding the strength of strategic overbidding tendencies. Consequentlythis was made an objective of our applied work. 'Good' Respondents:Orne (1962) points out that the relationship betweenanalyst and respondent is an interactive process with the respondentseeking clues as to the purpose of the experiment. If this purpose is inadequately conveyed then the respondentmay react in two ways, either he will not give the questionsdue considerationor he will attempt to guess the 'correct' answers,i. e. he will try to be a 'good respondent' and give the answerswhich he feels that the analyst wants. The problem of low consideration can be assessedby recording and analysing the numbersof respondentswho refuse to take part in the survey and the length of interview. The 'good respondent' problem may be exacerbatedwhere the interviewer is held in high esteem by the respondent (Harris et al., 1989) resulting in Desvousges from (1983) found little differ to true pay. et al., willingness which responses be it but bias noted that this study employed professional should a evidence of such interviewers, a potential solution to such problems. Tunstall et al., (1988) further recommend be follow interviewers the the that questionnaire of exactly wording and respondents that presented with a choice of preparedresponsesso as to minimise over or understatementof hypothetical bias (discussed designed Approaches to combat above) may true evaluations. here. be relevant also In our own empirical work considerableemphasishas been placed upon minimising (including Experienced design bias the practitioners certain of those stage. at such sourcesof referred to above) were consulted regarding the construction of questionnairesand execution in details Further given subsequentchapters. are of surveys. Upward Rounding: Batemanet al., (1993) arguethat, in DC formats, respondentsmay have an incentive to accept bids which are in excessof true WTP if the difference between by deviation is The such an effect will only operate caused the two amounts relatively small. in an upward manner, i. e. respondentwill not refuse to pay a bid level which is just below believes in However, WTP. that the the payment obligation provided respondents their true (i. e. he/she does not engagein strategicoverbidding) this should be a relatively minor effect further to subject made analysis. and was therefore not Anchoring: Kahnemanet al., (1982) among others have arguedthat respondentsfaced (particularly is described) the situation where good also not well will with an unfamiliar 2.29

interpret the DC bid level to be indicative of the true value of the good in question (Kahneman and Tversky, 1982; Roberts et al., 1985; Kahneman, 1986; Harris et al., 1989). Here the introduction of a specific bid level raisesthe probability of the respondentaccepting that bid.

This 'framing' or 'anchoring' effect may arise where a respondent has not

previously considered his/her WTP for a resource (which is likely with regard to public or is in In their their true mind about own valuation. such and/or unclear quasi-public goods) level bid may provide the most readily available point of reference onto the cases proposed is direction latches. There the about no a-priori presumption of such the respondent which bid-vector it has bid levels Positioning that more a such on the upper an anchoring effect. tail of the true WTP distribution should lead to anchoring increasingmean WTP. Conversely lower distribution depress bid to the tail the as of should the so emphasise vector positioning mean WTP. A related problem in DC (and potentially other) formats is the phenomenaof "yeadecides "nea-saying" to the ex-ante answer positively or respondent whereby or saying" Detection bid irrespective the of anchoring and related effects actual presented. of negatively is clearly important and was therefore made a researchpriority. Starting Point Effects: Severalstudieshave suggestedthat the use of an initial starting influence final bid, for (113) bidding the iterative in example, may significantly games point low WTP (see (high) Desvousges leads (high) low to mean a starting point the choice of a Green Navrud, 1989a; Green 1985; 1990; 1985; Boyle Roberts et al., 1983; et al., et al., et al., While 1991). the use of starting points may reduce non-responseand variance Tunstall, and lead to take short-cuts respondents cognitive may that an approach such argue commentators WT? (Cummings decision their true than thinking et al., seriously about rather to arrive at a informing been It has 1990). Loomis, 1989; Carson, that also Mitchell noted 1986; and respondentsas to the constructioncosts associatedwith a proposedenvironmental changecan is One 1982). Herzeg, bids (Cronin to to this approach problem and resultant also affect from Unfortunately bid shown on a payment card. a range to a allow the respondent choose bids "anchoring" produces of within the range given on the of necessity such an approach "correct" the that contains valuation and assuming such a range card with most respondents ignored 1982; Roberts Thompson, (Kahneman Tversky, being and and effectively outliers 1983; Kahneman, 1986; Harris et al., 1989). Given these concerns an IB format was investigated for evidence of starting point 2.30

effects. In summary we can seethat elicitation format provides a common theme in our review

incentives. Accordingly both and overstatement our empiricalinvestigation of understatement of these incentives was facilitated through an analysis of the effects of employing alternative elicitation formats upon CV responses. v. Mental Accounting Problems A further researchpriority was to assessthe extent to which respondentsconsidered income and expenditure constraints in determining their WTP. This problem is addressedin the theory of two-stage budgeting (Deaton and Muellbauer, 1980; Tversky and Kahneman, 1981; Kahneman and Tversky, 1984) where total income is, in the first stage, allocated to housing, food; in broad then, the e. g. recreation etc., and of expenditure, categories various items forest the amongst specific which constitute each category, e. g. subdivided stage, second in CV A if, because problem may arise studies etc. potential of recreation, water recreation, fail to consider all relevant hypothetical the. market, respondents of underlying nature the budget. Willis Garrod (1991a) 1972) (Slovic, the category such as particular and material Yorkshire National in CV Dales Park. Here the their study of one this point address "total budget for WTP is to their the to question, calculate prior yearly all asked, subsample (the donations including issues that those and subscriptions respondent) environmental ... WTP for Comparing (ibid). have that the remainder of with resultant made" might already face did such a mental account question, the authorsreport no significant not the sample who difference at the 1% level. This suggeststhat mental accountingproblems may not be severe. However, it was decided to test such a hypothesis in certain of our empirical work.

Bias Whole Part vi. Tversky and Kahneman (1981), in considering decision rationality, argue that

individuals see groups of goods, rather than specific goods, as the basis for utility (1992a/b) Kahnernan Quiggin Knetsch (1991), Extending this, and and contend maximisation. individual's flawed by 'part-whole' WTP fatally be bias, CV occurring an where may that between is 'part') (the distinguish fail the to specific good which under analysis and responses 'whole') into (the goods which that specificgoodfalls (seealso Kneese, the wider group of 1984;and Hoevenagel,1990,1996).If this werethe casethen,"whenrespondents are asked 2.31

to value someenvironmentalgood they may in fact make that valuationon the basisof a much wider rangeof environmentalgoods"(Willis and Garrod,1991a). The potential for part-wholebias is well documented(e.g. Walbert, 1984;Thaler, 1985; Hoevenagel,1990,1996). However,the major recent empirical supportfor such a by Knetsch (1992a). Here Kahneman is and respondents were askedtheir criticism provided WTP to maintain the quality of fishing in lakes in Ontario. The authorsreported no lakes (about for I% between WTP difference of the total small number of mean a significant lakesin Ontario)andmeanWTP for all lakesin Ontario.KahnemanandKnetschconcluded from this that the WTP statementselicited in CV studiesreferredto a 'purchaseof moral for (see Andreoni, 1990), 'warm than rather a payment a glow of giving'I satisfaction'or a good. An initial criticism of the Kahnemanand Knetschpaperwas providedby Mitchell (1991) who pointed out that this particular study relied upon both a poor instrument design (using telephone surveys thus relying upon a weak medium of description and dialogue with information (a low for to the single high survey); and poor commitment respondent potential a information thus description eliciting a and arguably providing vague used, was sentence 'part'). 'whole' knowledge based the than the rather of upon vague valuation, potentially is it by Smith (1992) the question that concludes who These criticisms are expandedupon framing itself, rather than some underlying theoretical problem, which results in the reported framing by Kahneman Smith Indeed, the that question used claims part-whole phenomena. (see Tversky, 1982) Kahneman Kahneman "does that and the Knetsch criteria satisfy not and interpret how in his develop people valuation questions" (Smith, helped to earlier researchon 1992). While the criticisms of Mitchell and Smith may be sufficient to discount the particular insufficient bias (1992a), Knetsch to that they Kahneman are prove part whole and results of Willis Garrod (1991a) described Yorkshire Dales In and the previously, study cannot occur. budgets WT?, by their stated with subsequent comparing respondents address this problem i. e. the environmental category budget is taken as a measureof the 'whole' while the 'part' is taken to correspond to WTP for the Dales alone. 7le authors report a highly significant difference (at the 1% level) between WT? and the overall budget arguing that, even if partindicates it is insignificant in Bateman is bias that a result et extent. such occurring whole Where (1991) that account and are somewhat mental part-whole effects similar. note al., 2.32

for their to account state mental a category of goods they are in effect respondentsare asked valuing the 'whole'. The empirical evidence for both part-whole and mental account effects is mixed. While Rae (1982), Burnesset al. (1983), Tolley and Randall (1983) and Strand and Taraldset (1991) confirm evidence of part-whole effects, Brown and Green (1981), Schulze hypothesis. In light (1987) Rahmatian (1983) the a of subsequent such reject and et al. link between it important feel is the findings to part-whole and emphasise we empirical by Tolly Randall the effect observed and and ordering problems especially mental accounting (1983) and Hoevenagel (1990) where it was noted that a good will elicit a higher WTP if it is be list is it than if to top evaluated, the goods valued after a of at of placed response inclusion If the of a mental accountingor part-whole question affects subsequent other goods. be if CV (1982) Rae's that only considered may valid results WTP this will violate criterion does in further the inclusion not significantly alter questionnaire goods environmental of the WTP values. Our own view derives from recently completed experimental research into the These in trades experiments goods. of private real occurrence of part whole effects in (Bateman 'overvalue' individuals to wholes parts relation demonstrated that consistently is Our b). to forthcoming therefore that similar concerning 1996a conclusion and et al., CV (although is to the environment peculiar that this not result reference-dependentutility: for is it) but calls an extension design a phenomena which rather exacerbate may poor again Accordingly behaviour. do individual we basic of model to the neoclassicalmicroeconomic instead to the to issues the refer reader this to preferring research central not make part-whole issue. for this of our view above results

Vii.

Information Bias Does the quality of information presented to 'service' a hypothetical market affect the

(1985,1986) Samples is The compared al. et certainly yes. almost answer responses received? information levels an from of regarding given varied two groups experimental responses from humpback (the a control group those received responses with whale) species endangered increased by WTP information found It increased information. that mean was given constant between 20-33% however statistical tests showed that while this test was significant at the

20% confidence level it was not significant at the 5% level. in the study by Mitchell et al. (1988) two groups were given differing information

2.33

regarding four sites of Special Scientific Interest (SSSI). Again additional information raised mean WTP but did not show this to be a statistically significant increase. A similar weak information bias result is found by Hanley and Munro (1991) in two CV experiments regarding WTP for heathland and woodland preservation.They postulate a threshold effect of information build-up below which no bias is detectablebut above which a weakly positive effect is found. A stronger result is provided by Bergstrom et al. (1985) whose study of bids to preserveprime farmland in the USA produced a 1% confidence interval test that additional information had resulted in higher bids. However such a finding is firmly challengedon both by Boyle (1989). In an experiment regarding WTP for theoretical grounds and empirical brown trout fisheries in Wisconsin, Boyle found no significant difference betweenmeanWTP information levels bid for fell three of although variance significantly as statements information increased. Boyle states that "the argument that changes in accurate or true in framing description the of CV questions will change value estimates is commodity blanket statement". a as unwarranted A less extreme view is adopted by Randall et al. (1983), Carson (1989), Kristr6m (1990a) and Hanley (1990) who argue that, since individuals do have preferencesregarding funding, distribution information their then and provision, goods, will always environmental from is different i. but WTP that this any other no good, priced or not, e. this is an affect Turner (1993) input Bateman important information that the and contend effect. expected issue is therefore to ensure that such information is seen to be true, constant across the bias induce implicit designed to towards a particular result; polemic and value sample, and not judgements being inadmissible. We have attemptedto adherecarefully to such guidelines in Given Samples (1985, this, the subsequently. work of presented et al. the research applied all 1986) indicates that inherent information bias should not be an overriding problem.

Procedural and TnstrumentRelated Bias i. Aggregation and Truncation of Welfare Measures

A particularproblemin theestimationof total economicvaluesumsfor spatiallyfixed forests ignore is that as the non-usevaluesheld such goods on-site surveyswill environmental by non-visitors. We argue that such surveys can only claim to estimate user values and that (off-site) sample remote random surveys are necessaryto estimate non-use supplementary (e. Brookshire 1982) have Such et g. al., studies shown that when aggregatedover the values. 2.34

larger non-visitor populations, total non-use value may be significant and may even exceed total use values by a significant factor". Consequentlywe undertake separateuser and nonuser surveys in our research.

The aggregationprocedureitself can inducebias.An importantissueis to definethe relevant population at the pre-survey stageand then conduct standarddiagnostics to validate the sample collected as being representativeof the population34. However the connection between the sample and the population is rarely perfect and certain adjustments may be justified, choice of adjustmentprocedure can however have a major impact upon aggregate Loomis (1987) In 2.25 to adjustment procedures experiment varied produce a one estimates. times difference in the range of aggregatebenefit estimates. A more fundamental question arises is the choice of an appropriate welfare measure for aggregation. If the distribution of WT? bids is non-normal (e.g. Poisson, binominal etc) (usually been by have the tail affected major upper) of the then the sample mean will distribution. However, such skewnessof itself does not indicate bias. Only where this is as In biases may a problem occur. overbidding such such as strategic a result of recognised justified. be In DC bidders thought experiments explicit may of strategic truncation cases is by WT? is this of calculating mean as given part a necessary truncation option choice of is issue discussed further" distribution. This in the probability cumulative the area under investigated. truncation options are of our empirical work where a number ii. Interviewer Effects Clearly the character of the interviewer may affect responses either directly by indirectly by (or light favourable in the unfavourable) or the a particularly good portraying Evidence of such an effect is mixed, being supportedby impression given to respondentS36 .

"An important point to note here is the criticism that when non-users,unfamiliar with an environmental WTP for token that they to as their asset, may state some small sum a of charitable preserve asked are asset, 1992a). If Kahneman Knetsch, (i. such sums are acceptedas and argument of satisfaction the moral concern e. (which is likely be large) to the may produce entire non-visitor population over true evaluations, aggregation be However, in far the amounts established, such can of of user value. validity excess until considerable sums be treated with caution. should such aggregations '"See Mitchell and Carson (1989); Hanley (1990). This will be a particular problem for mail surveys where interest in it is likely be biased low that towards those the responses will with a particular as are response rates below 40% Response the of general population. rates significantly are therefore unrepresentative good and Consequently face-to-face in our applied work adopts survey techniques. surveys. such common "See also Bateman et al., 1993,1995a; Ungford and Bateman, 1993. 36Clearly'good' respondenteffects (discussedpreviously) are again relevant here.

2.35

the findings of Walsh et al., (1990) and rejected by Desvousgeset al., (1983). One obvious is design to to this clear, unambiguousquestionnairesand train interviewers approach problem extensively in the art of presenting a neutral survey experience. In our empirical work the 'Total Design Method' advocatedby Dilman (1978) proved useful as did the discussion of '7 (1986). by Converse Presser issues and presented wider survey iff. Payment Vehicle Bias Rowe et al. (1980) found that WTP to preservelandscapequality was higher when an income tax increasewas suggestedthan when an entrancefee was proposedconcluding that debasement Many for fee-paying the experience. other studies, as a of viewed respondents Brookshire Coursey (1987) Navrud (1983), Desvousges or more recently and al. et example, (1989b), have reported a similar effect. Tunstall et al. (1988) feel that efforts should be made i. does WTP. 'neutral' e. one which vehicle not affect payment to adopt a We have addressedthis issue in two ways. Firstly we feel that the temporal unit both have Vn?, tested per visit and per annum vehicles. accordingly we employed may affect Secondly, as indicated above, the method of payment may be significant, consequentlywe have, in various studies, employed a number of differing payment routes including: national donations. local charitable and taxes; user-fees; and

Summarv In the preceding sections we have highlighted a variety of effects and biases which fall free-riding) (e. Some CV theory in these economic while within g. of studies. occur may (e. In theory that detected, g. anchoring). all cases the if of validity question would others, bias if to and thereby maximise the validity issues minimise are require addressing we these bias), (hypothetical Issues those as of scenario credibility such CV welfare estimates. of

information and interviewereffectshavebeenaddressedby adheringto recogniseddesign in fieI&I. issues The the of experts acknowledged consultations with and recommendations Hoehn 1987) Randall, (following WTP and of were understatementand overstatement "Detailed discussionof our approachto surveydesignandadministrationis givenin Batemanct,al. (1992). "In the courseof this researchthe authorhasdiscusseddesignmatterswith (alphabetically):Kevin Boyle; Green; Michael Colin Hanemann;Nick Hanley; Pcr-OlovJohannson;Bengt Garrod; Guy Carson; Richard Kristrom; John Loomis; Jim Opaluch;Sylvia Tunstall;Kerry Turner:Ken Willis; and many otherrecognised CV field in research. the of authorities

2.36

impacts the of of altering elicitation method while mental accounting addressedvia analysis inclusion in the the survey of specific questions examined via and part-whole problems were instrument. Truncation and paymentvehicle options were also explicitly addressedwhile both in in between these that were carried out surveys order variations value user and non-user groups could be considered. 2.2.2.4: Validity Mitchell and Carson(1989) identify three categoriesof validity testing for CV studies: Content; Criterion; and Construct, the latter of which may be subdivided into convergentand

theoreticalvalidity. Content Va idity Content validity is a concern over whether the measureestimated (WTP) can be said investigation (the Pearce fully to the construct). object under and correspond to accurately and Turner (1990b) point out both that the true construct (the Hicksian measure) will not be directly observable and that the pure public good nature of certain environmental goods (e.g. clean air) will make the necessarily subjective assessmentsof content validity extremely Analysts for decide in manner. must or replicable difficult to undertake any structured "the in has CV asked questions an right questionnaire themselves whether a particular for is "what if WTP a pay would actually respondents the measure and manner" appropriate 1989). Carson, (Mitchell it for if and existed" public good a market (1990) Willis Garrod literature, conclude that a general and In reviewing the been has "content has design that in not validity meant improvement survey questionnaire have As in consulted widely above we noted years". recent too a problem great regarded as in an attempt to maximise the content validity of our questionnaires.

Criterion Validity is CV One method used to assessthe validity of estimates to compare thesewith the in For (the many environmental goods such a test the question. good Atrue' value criterion) of is being CV is the undertaken. experimentation reason why is of course unfeasible and discussed do by Heberlein, Bishop those previously, and However, experiments such as indicate formats WTP These that test. generally will provide more provide us with such a 2.37

accurateestimatesof true behaviour than will WTA approaches. Accordingly, with the one exception of our farmers study (which we justify subsequently), we have adopted WTP formats throughout our empirical work. Construct Validity . One approachto validity testing is to examine whether the measuresproduced by CV relate to other measures as predicted by theory. Two variants of this construct validity approach can be identified: theoretical validity, testing whether the CV measureconforms to theoretical expectations;and convergentvalidity, testing whether the CV measureis correctly in the of good question. other measures correlated with Tests of theoretical validity have mainly centred upon examination of bid curve functions to see if they conform to theoretical expectations,for example, whether elasticities feasible In have Knetsch test theoretical sizes. an early of and validity, signed are corTectly for forest bid (1966) Davis curves recreation concluding that "the economic estimated and be high". A to the appeared similar approach was of responses rationality and consistency in for (1990a) WTP in Whittington by Haiti. an examination of water al. services et adopted Tests of the significance of explanatory variables found them to conform to standard by defined economic theory. as expectations A further variant of this approach is to examine the explanatory power of bid functions. However the cross sectional nature of CV and similar social survey data tends to R2 (1990) 0.2 R2 Hanley low that a minimum recommends value of should statistics. produce be used while Mitchell and Carson (1989) suggesta value of 0.15. However, psychologists R2 the techniques that social nature of survey to out make very statistics point are at pains is if A to limited test to they specific examine relationships stronger see much use. of So, for example,we should expect a significant, positive theoretical expectations. to conform income between WTP, diminishing and relationship with a similar relationship and marginally between visits to a site and total VV7P for it. The significance of coefficients can then be judged via simple T statistic tests. Studies which do not establish significant relationships indicates be they therefore treated with suspicion". should exist, must theory where

"Interestingly the absenceof suchrelationshipscanbe usedto testour earlierassertionthat non-userWTP ratherthangenuine valuationsfor poorly perceivedpublic goodsmayexhibit smallsumcharity type responses valuations.

2.38

A further theoretical test can be performed where measuresof consumer surplus are be T'hese measures can compared to the WTP estimatesobtained and, by available. surplus implied discussed Willig the earlier elasticity valuescan be calculated. equations manipulating Results can then also be compared with theoretical expectations and empirical findings. Similarly Bateman et al., (1994) compare CV results acrossa variety of goods as a test of internal consistency. Here it is shown that mean WTP varies logically with the availability be dismissed i. (perhaps imperfect) test as a should not of e. reasonableness of substitutes; CV (1991) Smith Finally testing to convergent a of et al., propose variant using validity. CV for demand the or programmes. commodities results can then actual marketed measure be compared with real world outcomes. Convergentvalidity may at first seemreminiscentof criterion testing. However, in this be 'truer' The than to the any other. most measures can claim comparative context none of from CV is those to revealedpreferencetechniques measures with compare common approach Cummings detail four (1986) (HP) hedonic (TC) methods. et al. or pricing such as travel cost based HP CV TC two studies,reporting that the value property and with with comparisonsof 60% different by of each other with some the were within approaches estimates produced being much closer. Mitchell and Carson (1989) report on a further nine comparisons and concur with this conclusion'O. A significant problem with such convergent validity testing is that the methods CV in For different theory example, while should constructs. measuring usually are compared be providing estimatesof aggregateuse plus non-usevalues, the TC only estimatesuse value. An important further distinction for site specific environmental goods (e.g. forest recreation CV is just is benefits the study carried out whether with an of clean air) compared with the is (the first 'remote' also population utilised should son-site' sample or whether an off-site hold the secondshould mainly exhibit non use values). while use values mainly theoretically Clearly in the latter case,comparisonwith TC results is questionableparticularly where nonis TC further be A that, theoretical to while problem significant". use values are thought CV from derive HP provides ex-ante measures,positing a ex-post situations, estimates and in the comparison of these measures. inconsistency information potential

Smith (1986). (1982) 4OSee Brookshire and et al. al. et also "'Many studieshavereportedhighly significantnon-usevalues,e.g. Walshet al. (1990)reportsexistence Other have 25% (see toLd Mitchell of value. this studies exceeded and over estimate as representing values Carson,1989).

2.39

_V

Validity: Conclusion

All of the validity testsreviewedcanbe criticised.Thecontentvalidity testis in many its however, (as be formalised and subjective fundamental, cannot operation yet) respects judgement is the underlying operand.Criterion validity as expressedthrough the comparison of hypothetical with real markets provides perhaps the most substantivetest of validity and indeed such tests indicate that WT? format questions can provide valid estimates of true is Unfortunately to this test the applicability of pure public goods empirical restricted value. be by inference from tests goods can only extended or quasi-public private upon and results to give validity to CV estimatesof pure public good values. Ile convergencetesting form However, been has to application. considerable practical we subject of construct validity between by CV, TC (and HP) degree the measures obtained on the comparability of question different Construct theoretical underlying constructs. they that measuring are the grounds does, feel, defensible testing, test of the a theoretical we provide validity analysis through is the and obtained employed throughout our empirical of results theoretical appropriateness work. 2.2.3: CONCLUSIONS: THE CONTINGENT VALUATION

METHOD

CV is a widely applicable and widely applied monetary evaluation method. It has the than the for goods any of other main to of environmental range a wider application potential basis CV believe We theoretical that a possesses strong with techniques. monetary valuation Furthermore income it compensated welfare measures. that estimates the unique advantage is consistent with many of the empirical basis demonstrated this theoretical have that we WTP/WTA (notably in the asymmetry of measures) observed which practice obtained results in have flaw technique, to being the appear considerable a of symptomatic than rather justification. theoretic Becauseof its nature as an expressed-preferencesurvey technique, CV is susceptible instill into it is bias indeed to bias easy responsesthe task of minimising such while to and

bias to an acceptablelevel is, we recognise,one which requiresconsiderableskill. We have discussed the major causes of bias and presented a programme for investigating theseissuesas part of our empirical examination of the value of forest recreation be integral Validity testing will also an part of this applied work. externalities.

2.40

COST METHOD

2.3: THE TRAVEL 2.3.1: INTRODUCTION

The original idea behind the travel cost method (TC) can be traced back to a letter from Hotelling (first reported in Prewitt, 1949) to the Director of the US National Park Service in which he suggestedthat the costs incurred by visitors could be used to develop a measure of the recreation value of the sites visited. However, it was Clawson (1959) and Clawson and Knetsch (1966) who first developedempirical models along these lines. TC is a survey technique. A questionnaireis prepared and administered to a sample of visitors at a site in order to ascertaintheir place of residence;necessarydemographic and information frequency information; to this trip of and other sites; and visit such as attitudinal data, From length, this etc. costs, associated visit costs can be calculated and purposefulness, factors, frequency demand to that visit so a relevant relationship may be other related, with demand function be In this then the case simplest can used to estimate the established. in be the site, while more advanced studies, of whole attempts can made to value recreation develop demand equationsfor the differing attributes of recreation sites and values evaluated for these individual attributes. 2.3.2: THEORETICAL

ISSUES

2.3.2.1: Welfare measures The demandfunction estimatedby the TC is an uncompensatedordinary demandcurve incorporating income effects and the welfare measure obtained from it will be that of Marshallian consumer surplus (shown by the area b+c in the lower panel of figure 2.2).

2.3.2.2: Basis of the method in essencethe TC evaluatesthe recreationaluse value for a specific recreation site by (measured its (measured for demand that to site as site visits) price as the costs of relating be defined by 'trip-generation TC function' (tgf) A can model a simple such as; a visit).

V=f

(C,X)

(2.5)

where V=

visits to a site

c=

visit costs

2.41

X=

other socioeconomicvariables which significantly explain V.

The literaturecan be divided into two basicvariantsof this model accordingto the dependent 'Individual V. The definition Travel Cost Method' (ITC) the of particular variable dependent defines the variable as the number of site visits made by each visitor over simply Cost 'Zonal Method' The Travel (ZTC) on the other hand, say one year. a specific period, into from the area which originate a set of visitor zones and then entire visitors partitions defines the dependent variable as the visitor rate (i.e., the number of visits made from a divided by in the population of that zone). 'Me ZTC approach a period particular zone

redefinesthe tgf as; Vl,I.Nh ý--

f (Ch9Xh)

(2.6)

where Vj,j NI,

Visits from zone h to site

q

Population of zone h Visit costs from zone h to site j

Xh

Socioeconomicexplanatory variables in zone h

The visitor rate, Vt,/.Nh, is often calculated as visits per 1,000 population in zone h. The underlying theory of the TC is presentedwith referenceto the zonal variant, and

discussionof thedifferencesbetweenthis andtheindividualvariantis presentedsubsequently before considerationof more generalissues. 2.3.2.3: The zonal travel cost method (ZTC) Discussion of the ZTC is illustrated by referenceto a constructedexample detailed in hypothetical The 2.3 the of a site. recreation value method proceedsas table which estimates

follows: (i)

Data on the numberof visits madeby householdsin a period(sayannually)andtheir is collected via on-site surveys. origin

2.42

9e 9

%D

0.

4q;

00 c:ý

r-

ýq

00

-ci

CM 6

ý M

44

r-

'

q

-

;ý !?

0

.00 92 2

2 g C'2ý -'

Q ý ýn m

0

0

9

0 ...

(D

le

ce .2

C-)

9 bk

%0

8

em

2

CD

Ci

Co) cn

0

e

4-

CA z3 ein ci .

A

&-0 K

kn

CD

p

ei

> io

to) (A

VI 4 ý

cr M

ý, 1 r-

vi 4

So

92.

"0

40.

ul

e0 -ci GA :j4

u U "Ci

10

ci Z

d.) vi

0

-2

rA

ý

0 25 .0



e=9

Ici 0=

t2

*ý z3 -CJ

GO2

ci Gn

c: 0

U

;ju==U

ri v2 E3 Ei 52 000

ei JD

-i

(11)

The areaencompassingall visitor origins is subdivided into zonesof increasingtravel (number households) in 2.3) I (column table the total population of each and of cost 2). (column noted zone

(iii)

Household visits per zone (column 3) is calculated by allocating sampled household

(iv)

visits to their relevant zone of origin. The householdaveragevisit rate in each zone (column 4) is calculatedby dividing the 3) by in (column (number household the zonal population each zone visits number of be 2). Note households; that this often not a whole number and will column of commonly less than one.

(V)

The zonal average cost of a visit (column 5) is calculated with reference to the distance from the trip origin to the site.

(vi)

A demand curve is then fitted relating the zonal averageprice of a trip (travel cost) demand for household. This curve estimates per to the zonal averagenumber of visits just In time the "whole than spent on-site. our recreation experience" rather the hypothetical example this demand is explained purely by visit cost and the curve has in (2.7). form linear (unlikely) equation given the (2.7)

V,,/.Nj = 1.3 - 0.3 Ch where

Vt,/Nj = visit rate (averagenumber of visits per household) from each zone

C',

from eachzone costs = visit

demand illustrates 2.5 curve.The experience this Figure particularwhole recreation in households implicit involves that the all assumption this curve of estimation distance zonesreact in a similar manner to visit costs. They would all make the same identical have i. faced if to they the are assumed same costs e. with number of trips tastes regarding the site. (vii)

In each zone the household consumer surplus for all visits to the site (column 6) is

demand (cost) (equation between integrating (2.7)) by the the curve price calculated from that the at each and price which visitor rate would made zone actually of visits

2.44

fall to zero (i. e. the vertical intercept of the demandcurve at point P in figure 2.5y'. Households in zone 3 for example would have a consumersurplus equal to area ABP for all their trips to the site i.e.:

P Consumer surplus for zone 3f

(1.3-0.3 Ch).dCh

(2.8)

q=B

(viii)

In order that annual total consumer surplus for the whole recreation experience can

be estimatedin eachzone,total householdconsumersurplusmust firstly be divided by the zonal averagenumberof visits madeby eachhouseholdto obtain the zonal household (column 7). This per canthenbemultiplied surplus visit consumer average by the zonal averagenumberof visits per annum(column3) to obtain annualzonal 8). (column surplus consumer (ix)

Cumulating annual zonal consumer surplus (column 8) across all zones gives our for the whole recreational experience annurn total per surplus consumer estimate of

of visiting the site. One immediate problem with the above approachis that it yields value estimatesfor (zero-priced) day trip to the a recreation site rather entire of the whole recreational experience information (1979) Freeman that the out points gathered the site alone. than an evaluation of for demand defines in fact the the on-site recreational curve in a TC survey only one point on but is for by incur their their Many consumption, the price travel set cost a goods experience. is therefore the zero sum of all visits across However the of recreation price market market. for demand with a zero admission recreation price. the on-site represents all zones

42Severaltexts make the simplifying assumptionthat consumersurplus for the marginal user (here the most Hufschmidt 1983). This 1979; lead Worrell, (Sinden typically et al., is to some will and remote zone) zero underestimateof true consumer surplus.

2.45

Figure 2.5: Demand curve for the whole recreation experience

Zý -vi 5 k

-0

0.5

025

zonal average nunbe(

0.75

1.0

125

of household visits pa

Key-(D - zonenumberI

Source: Bateman (1993a) In estimating consumersurplusfor the on-site recreationexperience,many earlier texts (e.g. Sinden and Worrell, 1979;Hufschmidt et al., 1983) follow Clawson and Knetsch (1966) first by increases in that to assuming people surplus would react consumer estimate and increases in i. in they to their the travel would react as same way costs e. the admission price demand curve function staysas estimatedfor the whole experiencebut each zonestravel cost is increased by an incremental admission cost and visits from each zone re-calculated Summing demand the curve. visits acrossall zonesat each admission to estimated according demand Integrating between this the curve. the under experience curve on-site out cost maps initial zero admission price and that admission price at which visits in all zones fall to zero for A the total surplus on-site consumer recreational experience. worked example estimates in is Bateman (1993a). given of such a calculation The weak link in the Clawson-Knetsch approachto on-site valuation is the need to 2.46

assumethat individuals will react in the same way to admission fees as they do to travel costO'. If individuals have different willingness to pay for an environmental good because is the of method of payment which used then it is likely that such an assumptionmay well be violated. In practice many TC studied have rejected the Clawson-Knetschapproachto on-site valuation, preferring modification of the whole-experience demand curve.

A common

approach (adopted in our empirical research)is to ask visitors to evaluate how much of the is due to the on-site experience.Typically visitors the experience recreation whole utility of information This to the off-site experience. to points on-site and percentage allocate are asked how incurred (i. be travel to costs e. evaluate much of reduce costs can used either can then justifiably be said to have been purely related to the on-site experience)or the information function into be directly the trip generating as a separatecontinuous explanatory entered can Garrod, Willis 1991b). In (for and see either case the whole-experience example an variable demand function will be altered. The resultant curve will not be the same as the on-site demand curve as defined by Clawson and Knetsch above. However, its validity may well be it does in not the defensible that rely upon previous assumption of travel cost effects more duplicating admission price effects. perfectly 2.3.2.4: The individual travel cost method (ITC) The fundamental difference between the ZTC and ITC is that the latter defines the dependentvariable as Vij, the numberof visits madeper period (annum) by individual i to site j (Brown and Nawas, 1973; Gum and Martin, 1975). We can thereforerewrite the simple tgf its ITC (2.5) equivalent; as of equation Vii =f (Cij, XI)

(2.9)

where

Cij

i individual j by to made per of visits year site number individual by i faced j to visit site visit cost

X,

factors determining individual i's visits other all

VO

"'rhe problems of vehicle bias, usually discussedwith regard to the contingent valuation method (Bateman here. 1992), Turner, are pertinent and

2.47

The demand curve produced by this model relates individual's annual visits to the costs of those visits (i.e. there is no requirement to convert from zonal visitor rate to actual visits as in the ZTQ. The above tgf relates to the whole recreational experiencebut may be adjusted to relate the on-site experiencevia either the Clawson-Knetsch(1966) or Willis and Garrod (1991) approachesoutlined previously (the latter being adopted in our work). The move from a zonal to an individual basis allows the specification of a number of individual-specific explanatory variables, for example, we could respecify our ITC tgf as; Vii

f (Cij, Eij, Si, A, Y1,Hi, NI, Mi, TI, Qj)

ViJ Cii

i j by individual to made site of visits per year number individual's total visit cost of visiting site j

Eij

individual i's estimateof the proportion of the day's enjoyment which

where

was contributed by the visit to site j Si

individual i's assessmentof the availability of substitute sites

A, Y,

age of individual i income of individual i's household

Hi

size of individual i's household

N,

i's individual party size of

Mi

dummy variable; whether individual i is a member of an outdoor or environmental organisation

Tj

activity undertakenon site

Qj

vector of environmental attributes of site

A number of permutations of equation (2.10) are possible. We discuss detailed

in below. However, Cq one approach which we adopt specificationof the cost variable Eij* So, is later to this the on-site combine utility variable with empiricalwork, certainof our 60% individual if that of the daysenjoymentwasdue to the on-site for example, an assesses 0.60 is ACij 0.6 Cij, Such Eij the adjusted and utility cost variable then an = experience for derived from the problem of allowing addresses other sitesor utility approachexplicitly be defined in S, Other be itself. journey may variables numerous ways. may either the definedasa binarydummyor asa categoricalvariableor a continuousvariableof the number 2.48

of substitute sites specified or a vector of distance costs to those sites. Similarly several definitions of Mi and T, may be used although our experienceindicates that simple dummy variables are often effective. The environmental attribute vector q. is discussedin further detail subsequently. Further explanatory variables are plausible, for example, BqJ6 (1985) includes a dummy variable for the mode of transport used which, in an empirical test, he finds statistically significant. The demandcurve for the site will be defined by the bVi/Wij relationship as illustrated in figure 2.6. Integrating under this curve gives us our ITC estimate of consumer surplus per individual. Our estimate of consumer surplus for the site is then obtained by multiplying by ie; individuals the site annually", visiting the number of f Total consumer surplus = Nj. f(Cij, X). dCij

(2.11)

where j individual to site per year visits = number of (Cij, X) = defined as per equation (2.9)

Nj

2.3.3: METHODOLOGICAL

ISSUES

Here we review the principle methodological issues arising in the application of the in have how these our practical studies. addressed TC and show we 2.3.3.1: The central assumption indication in be taken as an that costs can some way The underlying assumption visit Gibson (1978) In that study notes an early where qualification. requires value of recreational into be (e. to to have close a site g. moving as so their residency individuals place of changed becomes be trip then the of a and price endogenous site) to near a recreation area a country demand lie is In the such case estimated curve will a the central assumption violated". below the true demand curve and consumer surplus will be underestimated. Very few empirical studieshave taken account of this potentially important criticism.

"Care has to be takenin the aggregationprocedureas data may well havebeengatheredin the form of data is total annual visitor usuallyheldasnumbersof individuals. Household whereas householdor party visits (or, on occasion,doublecounting). data must be convertedto individualvisit datato avoid underestimation in H. M. Treasury (1972). is 4ST-hiS earlier studies, e. g. noted also problem

2.49

However in a recent study, Parson (1991) argues that the endogeneity may be eliminated job (place instrumental of approach work, characteristics,etc). A simple variables using an include importance be to the a survey question regarding this of proximity would variant of to the recreation site in deciding place of residency. A dummy variable could then be used to split up responseswith significance tests determining the importance of this factor. A related problem ariseswhere the on-site time is not the only or even major objective 'pure define Stabler (1976) Cheshire the three categories of visitor, visitor' the trip. and of 'meanderers' 'transit is trips; make multi-visit and visitors' who site orientated; who strongly journey itself'. While from the pure visitors pose no theoretical who gain utility primarily be how journey to the costs are allocated amongst of problem problem, transit visitors pose is by This the time to on-site where also applies meanderers problem the sites visited. definition only a side issue in the trip decision and where travel time in particular may not from i. travel time to the range of may negative e. utility cost, true opportunity a represent issues in discussed below latter These the context are categories. these visitor across positive

of time costs. 2.3.3.2: Calculating visit costs

costs,

Total visit costs can be defined as the sum of money expenditureon travel (e.g. petrol (usually time the time travel cost of on-site and opportunity etc), the opportunity cost of

(for 1717C define More an studyf; can we exactly zero). Cij

PTCij.Dij + PTrj,. TTj, + PSTij.STj

Cij

Total cost to individual i of visiting sitej

PTCjj

Money expenditure on travel per mile/km

Dij

Distance travelled by individual i to site j

PTTIJ

Individual i's opportunity cost per hour of travel time to site j Individual i's journey time to site j (hours)

(2.12)

where

TT, j PSTjj

Individual i's opportunity cost per hour of on-site time at site j

461na related study, Christensenet a]. (1983) discussthe problem of disaggregatingholiday from single visit costs. 'For a ZTC equivalent see Bateman (1993a).

2.50

Individual i's length of on-site time at site j (hours).

STjj

One basic problem in evaluating Cij is that PTCij is unlikely to be a constant for all segmentsof the journey. Rather the variable quality of roads used in a journey will lead to varying per mile travel expenditure rates. Travel time will also vary in a similar manner. Both of these issues are consideredbelow. i. Travel Costs In calculating travel costs, BqJ6 (1985) simply multiplied household size by the is fare. However, less applicable to car travel, a simple such approach economy class rail where three cost calculation options exist; Petrol costs only (marginal costs) (1) (2) (3)

Full car costs; petrol, insurance,maintenancecosts, etc. Perceived costs as estimated by respondents.

Clearly using option (2) will raise visit costs above that of (1) and ultimately increase Common (1987) both Hanley apply and options to the same estimates. surplus consumer forest recreation data finding that option (2) gave a consumer surplus estimate more than twice as large as option (1). Willis and Benson (1988) obtained a similar result in a study of visitors to wildlife in 2.4 for Results Yorkshire. table in the of sites studies are given showing that the one areas definition defining from to travel a of petrol plus standing charges costs as petrol only move (same functional form difference the to the of explanatory power model made no significant (highly impact in both the cost coefficient significant upon a minor cases) only retained); and i. e. both assumptions had equal statistical validity. However this translated through into a (over bigger 70% for in full increase the per visitor consumer surplus cost assumption) major and thereby to total site consumer surplus.

2.51

Table 2A

Impact upon estimated consumer surplus (CS) of alternative travel cost specifications

Travel cost

Travel cost

Specification

cocfficicnt

Petrol only

-2.667 (6.73)

Petrol. plus standing charges

-2.605 (6.49)

CS/visitor

Model R2

0.59

0.83

Visitors

Total CS

p. a.

estimate (;C)

15,235

9,001 1

1.02

0.83

15,235

15,574

1

1

1

Notes

Casestudy : Wildlife visitors to Skipwith Common,Yorkshire Method: ZTC Functional form : Double log throughout CS/visitor roundedto nearestpenny Figures in bracketsare t-statistics

Source:

Abstracted from Willis and Benson (1988).

is (1983) Christensen that the (1983) correct cost measure that which argue Price and be It that are at perceiving visitors poor to the may well visit. as relevant visitors perceive daily insurance and maintenance cost equivalents or that they see these as sunk costs which do not enter the tgf, i. e. they only consider the marginal cost of a visit, equating this with

marginal utility. in approach analysis our a sensitivity As a result of this apparentconflict we adopt definitions. the three cost above of testing all empirical work, Time Costs

function (2.12), in throughthe travel the time cost indicated visit enters As equation 1975; Freeman, (McConnell, However, theoretical analysis time and on-site time variables. 1993a/b Bateman, 1987; Johannson, 1980; shows that the relevant and Wilman, 1979; items. for be two these hour the same not need per costs opportunity determination of these opportunity costs raises considerableproblems.

Furthermore,

in difficult that, as noted previously, we to particularly analyse are Travel time values is If travel time definite whether utility positive or negative. have no a-priori notion about individuals (i. travel the their has as of recreational enjoy part e. utility positive time travel defined) 'meanderers' time travel then cost as previously some general using g. e. experience, 2.52

figure to price this will overestimatethe consumer surplus of a visit. Bojd (1985) does not include a travel time cost (i.e. implicitly he gives such time an opportunity cost of zero) on the grounds that 80% of survey respondentsexpresseda positive utility for travel time to the ignoring leads This that travel time residual costs only approach assumes site under analysis. to a minor underestimateof the true consumersurplusý" However, the BqJ6 approachis far from standard. Indeed static optimisation of any indicate income (subject function time that the to and constraints) would conventional utility leisure (i. between labour e. the value of recreational travel and marginal rate of substitution individuals However, is are'not able to completely vary the to when wage Tate. time) equal becomes for hours time the constrainedand the money substitution of worked the number of direct relation between the value of time and the wage rate breaks down (Johnson, 1966; McConnell, 1975). Early applied investigations of the actual relationship of wages to travel time were These (1970; 1976). Cesario Knetsch (1976) Cesario by papers examined and and undertaken implicit (and from to to costs) estimate an relevant transport and work commuters choice of date, "that, basis Cesario to the the of evidence collected on time. concluded travel value of half is between the travel to one quarter and one of nonwork time respect with value of (individuals) wage rate" (Cesario, 1976), and subsequentlyused a value of one-third the wage Nelson (1977) is An that of calculated who a time. approach travel alternative to price rate district housing data business for implicit the to with central price of proximity marginal Washington DC, from which he derived a value of time which, when related to wage rates, falls within the Cesario range. However, as he recognisedat the time, Cesario's analysis only is the there reason why marginal utility obtained time no necessary and commuter considers time. travel be to recreation applicable should Common (1973) and McConnell and Strand (1981) used an iterative processwhereby into final being determined tgf the the the choice where are substituted time values successive (1983) Desvousges is the (R) applied value the et al. maximised. model of explanatory power full McConnell Strand (1981) (1976), Cesario and and a wage rate of time results of individual in 23 ITC visitation of patterns at model water sites to recreation assumption an

"'Johansson(1987)pointsout that,if time costsareignoredthen"theestimatedcurvewill be locatedinside for 'true' living those the except possibly one, less very closeto the recreationsite,sincethe than be steep and in distancefrom the visitors zoneof origin". increases to relation costs of underestimation

2.53

Testing at the 10% confidence level, Desvousges et al. (1983) rejected the McConnell and Strand (1981) approach, while both the Cesario (1976) and full wage the USA.

assumptionsperformed equally well, both being rejected in roughly 7 of the 23 cases.On the basis of these results Smith and Desvousges(1986) concluded that "for practical purposes, there is no clearcut alternative to our using the full wage rate as a measureof the opportunity cost. Even though it may overstatethe opportunity costs ... none of the simple adaptationsare superior". Similar results are obtained in a completely different cultural setting by Whittington in in Kenya. (1990b) Here two the time collecting of a study of value spent water et al. both indicate of a value of time approximately which separate approachesare employed, labour. for However, to the activities such as collecting water rate unskilled wage equivalent from In different TC those their associated with recreation. study of UK are qualitatively forest recreation, Benson and Willis (1992) employ three value of time assumptions,justified as follows: i)

0%; this assumes that visitors would not benefit from some alternative recreation activity.

ii)

25%; the UK Department of Transport's value of non-working time used in CBA assessments of road proposals up to 1987.

iii)

43%; the value of time used by the UK Department of Transport following their review of non-work time in 1987 (Department of Transport, 1987).

The 0% figure initially appears difficult to defend. However, if visitors cannot vary forgone few (i. hours there there costs) and are no wage are competing e. or extend their work be low. The Department time then the two cost of opportunity well opportunities recreational Cesario-type based both latter (43%) figures the Transport analyses upon of which are of appears the more rigorous. While the Cesario approach is, on the surface, theoretically and practically appealing, a deeper analysis of the complexities of the work/leisure important problems.

highlights relationship

some

In a thorough analysis, Bockstael et al. (1987) note two major issues:

hours, for example, a second job may pay a lower rate than (i) wage rate may vary with work does a first; (ii) individuals face uneven time constraints, i. e. they may be restricted to work jobs. As in be hours the a result particular rate may an appropriate measure wage specific interior (at fully for hours, it but those solutions) their who can vary time costs work will of

2.54

be inappropriate for those who cannot (at comer solutions). While Bockstael et al. provide a theoretically plausible approach to the valuation problem by incorporating time and income constraints into a utility function, the empirical application of such a technique is problematic. In particular the data requirements of such a model, including information regarding each individuals time constraints, are- highly exacting. For these reasonssuch complex approacheshave not been widely adopted and no has UK study attempted such an analysis. published Shaw (1992) provides a number of suggestionsregarding how the value of time problem might be addressedin a practical study. One suggestionis to use CV-type questions to elicit WTP for Tecreationtime" while another is to accept that there is likely to be some rather unclear link with wage rate and to therefore usea sensitivity analysisapproachutilizing fractions. of wage a wide variety Turning to consider the unit value of on-site time, if the length of time spent on-site were a constant for all visits to a particular site, then such costs could effectively be ignored increase imply in they only an absolutevisit costs but not in marginal relationships would as (McConnell, 1992; Bateman, 1993a). Furthermore, in an empirical analysis, Bqj6 (1985) finds no evidence to refute an assumptionof constant on-site time costs while Bockstael et its from because (1987) time their on-site empirical of potentially ambiguous analysis omit al. function from its demand both the and constraints. arising entry within utility effect upon iii. Summary: treatment of travel and time costs The treatment of travel and time costs within the tgf is one of the most crucial issues

in operationalisingthe TC. The approachwe haveadoptedin this studyis as follows: issue One fundamental Measurement: of linear and concernsthe measurement a) temporal distance. We believe that our use of GIS manipulated digital road networks (incorporating road length quality and average travel time by individual road section) in certain of our TC studies,considerably enhancestheir accuracyof measurementcomparedto that in most other published studies.

"We employa similar approachin our TCM studyof theNorfolk Broads(unpublished).Hererespondents were askedWTP to reducetravel time. However,many gave a zero responseindicatingthat the journey contributedpositivelyto trip utility. Furtherdirect questionsconfirmedthis finding.

2.55

b) Travel costs: Following the above review we adopt three definitions of monetary travel costs: petrol only; petrol plus standing charges (insurance, depreciation, etc); respondentsperceived travel cost. c) Time costs: We adopt the suggestionof Shaw (1992) and perform a wage rate sensitivity analysis upon travel time. Four wage rate values are employed; 0% (following the argument of Benson and Willis, 1992); 43% (the UK Department of Transport's value of time); 100% (following the empirical findings of Smith and Desvousges, 1986); and that best fit percentage which provides a of the data (our preferred option). rate variable wage We recognise the limitations of such an approach and that the labour supply method of Bockstael et al., (1987) is theoretically superior. However, such an analysis is both complex data in Given limited demanding terms requirements. resourcesour approachshould of and provide a reasonableapproximation while yielding an analysis which is more rigorous than In line have UK such studies. with studies, we omitted on-site time from other contemporary the cost function (although suchdata was collected and analysed),following the argumentthat this may not significantly affect consumersurplus". d) Total costs: Given that travel and time costs are both a function of distance, their independent

inclusion

multicollinearity.

within

the tgf

is

likely

to create significant

problems

of

Accordingly (and for additional reasons reviewed subsequently) we follow

Brown and Nawas (1973) and Gum and Martin (1975) in using the ITC which we adapt following

Cesario and Knetsch (1970) by adding together travel and time costs to produce

total visit costs. in relevant studies we then multiply this by the respondents stated proportion of the in for to the days thereby that proportion site question allowing attributable total enjoyment journey itself. from This derived is the days sites and other adjusted cost visit the utility of then entered as a single explanatory variable within the tgf.

2.3.3.3: Site attributes (environmental quality and multicollinearity) The trip generating function described in equation (2.10) highlights several independent variables as explanatory of visits, one of which is the site environmental quality function is In Q.. the single stage analysis simple entire a estimated as one with variable, "Following the analysis of McConnell (1992) who shows how on-site time may, in certain circumstances, be a significant factor (and proposesa solufion to its treatment),we intend to incorporate this into future studies.

2.56

conventional significance and other statistical testing being carried out. While this is a common approach it is, strictly speaking, only universally valid for quantiflably unidimensional sites, that is, sites which possessonly a single environmental quality attribute in be which can measured a quantitative manner. The reason for this is that, where sites possessmultiple attributes,theseattributes should enter the tgf as separatevariables.However these attributes may themselves be highly correlated, i. e. a potential multicollinearity or $suppressorvariables" (Conger, 1974) problem exists making single stage OLS estimators invalid. In reality recreation sites very often provide multi-attribute services. For example Vaughan and Russell (1982) include the explanatory variable Qkj, the level of quality k be If kA j be k then there one may or more. may multiple site where at characteristic influencing factors 'Ilese factors be significantly visit rate. may well quality environmental large have forests but both for are may also many access which routes, example, collinear, A be factors to related visits. worked example of the suppressor positively may these of is in (1993b). Bateman in TC given studies variable problem In reality a number of possible outcomes may arise from a suppressor variable Furthermore Coefficients the significance signs. alter radically, even changing may problem. increase (see Langford, disturbed 1992). becomes and may even spuriously of parameters However, despite the potentially serious nature of this problem, no single definitive solution has yet been found'.

Clearly a first step is to test for the presence of such a problem by

is further A to test tables. estimate single explanatory variable correlation of calculation how for and coefficients and significance examine variables significant regression models levels alter in subsequent multiple regression models. If such a problem is confirmed one index is attribute to site variables a all with single of site replace proposed course (Talheim, 1978; Ravenscraft Dwyer, 1978). thus collinearity and removing attractiveness, However, such an index cannot be adequately set up without full knowledge of the functional is individual dictated by As demand between this relationship and site attributes. relationship index is infeasible. different for truly the of a creation representative attributes, preference Ideally we would wish to respecify the individual's utility function in terms of the attributes

(1988) "some for Maddala that comments the muldcollinearity solutions often suggested -"Interestingly The disease" lead track. than on a wrong the suggested us cures are sometimes worse actually problem can (p. 224).

2.57

of sites. Morey (1981,1984,1985) adopts various functional forms for the utility function which include site attributes and levels of use. By assuming budget constrained utility maximisation we can obtain estimating equationsfrom which parameters(including those for site attributes) can be estimated.However, the need to specify the form of the utility function in link this approach. A further approachis to use a two stagegeneralised constitutes a weak least squaresapproach' (the generalisedtravel cost method). Such an approach is adopted by Smith and Desvousges(1986) in their study of lakeside recreation in the US. Here the authors postulate a 'true' trip generatingfunction containing both socioeconomicand various latter being highly the collinear. The authors employ a two stage variables, site attribute first the stage of which consists of omitting all potentially collinear site attribute approach, variables and estimating a tgf containing the remaining socioeconomicvariables. The resultant coefficients are then regressed against the remaining site attribute functions demand for in to those attributes. the produce stage generalised second variables In effect then the tgf thereforerelatesvisits to socioeconomicvariables and site attributes, the latter being expressedas subfunctions of the socioeconomicvariables. In order to calculate the relationship (partial derivative) betweenvisits and a particular is information the the socioeconomic variables of regarding values required. site attribute, Smith and Desvousges(1986) addressthis problem by using mean values of these variables within the subfunctions of the tgf. By adopting such an approachSmith and Desvousges(1986) produce estimatesof the impact of individual site attributes upon visits, i. e. they estimate attribute demand functions. demand in facility TC important that such to the estimatesallow the This provides an extra demand to investigate site overall and thus to most contribute which attributes analyst to functions attribute, or combination of In to which specify allow us effect such welfare. facilitate the thus individuals optimum planning and site at a recreation most enjoy attributes, this Further approach of development and a worked consideration creation. and of site example is given in Bateman (1993b).

52Analternative two stageapproachis to cmploy factor analysisor principal componentsanalysisapproaches formation the Of combination variables These 1978). on (Goddard and Kirby, 1976; Johnston, approachesrely While this be weighting can adjusted to the explanatory variables. made up of weighted combinations of defy variables combination often the practical the model, resulting of the statistical significance maximise interpretation thereby greatly reducing the usefulnessof the model in any predictive economic analysis. Such is literature here. and not Pursued in has been the context of planning urban an approach used more

2.58

While the Smith and Desvousgesapproach is interesting it is but one method of addressinga problem which, to date, has no definitive solution". Furthermore, it has several drawbacks notably its complexity, the difficulty of determining an appropriate measurefor environmental attributes, and the relatively low power of resultant models'. A simpler approachis to confine analysis to similar sites. In effect this is the strategy adoptedby Willis and Benson in their TC studies of UK forest recreation. As an initial part of this study a cluster analysis was undertakengrouping forests into sets of similar attributes (Benson and Willis, 1990; discussedin subsequentchapters). Our task is made yet simpler by the fact that forests. Therefore, by than concerned with potential rather actual are we ultimately in forests field which have those attributes which we are interested in work conducting our (i.e. recreational facilities) and conducting tests of the transferability of our estimates to have in locations for forests (if not to to can, we some extent, claim controlled other similar Such problem. a strategy underpins much of our applied the variable suppressor solved) in discussed subsequentchapters. as valuation work 2.3.3.4: Weighted observations (heteroskedasticityand sampling bias) The observations used for estimating the demand curve in ZTC analyses represent a

from have themselves size zones which will of varying often varying samples of series have degrees As these of precision; they may observations may varying such populations. have non-constant variance (i. e. subject to heteroskedasticity).71is means that (Maddala, 1988): the least squaresestimatorsare unbiasedbut inefficient and estimatesof the variances invalidating biased significance tests. thus are

A commonapproachto heteroskedasticity problemsis to transformthe databy logs's, however Snedecor and Cochran (1976) state that a general approach "is to weight each i. least in its (VVLS) inversely a e. weighted squares variance" approach as which estimate Such is low by Bowes low an approach given weight. adopted are precision of observations be directly by (1980) Loomis that observations suggest weighted zonal population who and

"The multi-level modelling approach (Jones, 1991; Langford, Bateman and Langford, 1996) appears here in are currently and we employing such an approach our ongoing evaluation work. promising particularly lakes Smith Desvousges 22 US R2 between 0.02 and of report their values study and 0.54 with a simple -'In mean of 0.22. 55Maddala(1988) outlines a variety of possible approaches(p. 161 et sec.).

2.59

(i.e. large populations should produce more precise observations).However, Christensenand Price (1982) point out that heteroskedasticitYarises not only because of differing zonal differing from but also populations zonal sample sizes which in turn derive from differences in visitation rate across zones (an individual's visitation rate is likely to be higher in zones in Following this to the argument, a subsequentpaper, Price et al. (1986) weight site). nearer demand curve observationsby zone population/visit rate. Lucas (1963) shows that a weighting approach may also be appropriate where sampling bias arises in the presenceof a correlation between length of stay and travel cost further for (higher individuals travel travel stay at sites costs) shorter periods than who e.g. those who come from nearby and thus the former group are less likely to be sampled.Lucas individual by length in travel the the costs reciprocal of of cases, weighting that, such argues (1986) Price bias. this al. et present various permutations of a forest stay will correct both heteroskedasticity their combines own of which weighting and recreation model, one Lucas's sampling bias weighting. Potential heteroskedasticityproblems appear more likely to occur in ZTC than ITC logarithmic (as functional forms Combining well as other) models with and such models. heteroskedasticity testing, appearsto us a sensibleapproachto this problem and one explicit in investigate our applied work. which we 2.3.3.5: Substitute'sites Substitute sites should impact upon visit demand in three ways: the visit price of the fees; In quality at substitute sites. their and environmental practice entrance substitute sites; forms (e. Smith in Desvousges, included 1986), the estimated g. and rarely are such variables data involved. high In TC being difficulty the costs effect a survey would major practical have to be performed at all significant substitute sites in order to provide the full data 1991). (Bockstael al., et requirement The presence of substitute sites deflates recorded demand. The further away that higher from live the probability that there are substitute sites closer to them the a site people i. demand is below in depressed the trip observed e. curve the true question site the than demand curve at higher travel costs. By concentratingupon site loss (i.e. the sites 'contribution value' to the total recreation TC Price (1975,1978) Connelly than conventional site value, rather and sites) and all of value 2.60

Price (1991) show that, in fact, the presence of substitute sites leads the TC to either systematically over or underestimate true consumer surplus dependent upon the spatial relationship between sites and population centres. Assuming that population is randomly distributed, Price et al. argue that if recreation sitesare clusteredthen the loss of one site will, on average, make little difference to the general proximity of population to sites i.e. the conventional TC site value will overestimatethe value of site loss. Conversely if sites are systematically-spaced(particularly relevant for man-maderecreation areas) then the loss of for induce TC the the site-proximity change nearby population and a major value one site will loss. Only (as true the of site sites well as population) are value where will underestimate TC distributed these the over and on average cancel out underestimations and will randomly in loss. However, true the a simulated model test of value of site represent value accurately (1991) found fitting Price Connelly hypothesis, that errors could more than and curve this (see discussion functional form). impacts sites of the of substitute outweigh A number of solutions to the substitute sites problem have been put forward. Price (1979a) addressesthe problem "by the simple expedientof basing visit rates, not on visits per but 1000 for is 1000 this the population on visits per year per whom population, per year is best lower-boundary However, its facility this type". at a partial, approach, of nearest ignoring distant visitors who presumably value their visits highly. Burt and Brewer (1971) usetheir subjectivejudgement to identify presumedsubstitute homes from distances to these sites as explanatory variables respondents' the enter and sites in the tgf. Such an approachis admittedly subjective however a more fundamental criticism is that it implicitly assumeshomogeneity of sites, an improbable assumption.Greig (1977) imposes a predetermined, utility-based model linking visits to site characteristics. Such an for lack information both be the of adequate prior regarding criticised also may approach define A hybrid the to need and site characteristics. relationship of the utility appropriate Greig/Burt and Brewer approachcould theoretically be constructedif data were available on data Burt Brewer Given could and such run a sites. we to substitute actual visits homogeneity the predicted compare and visitor rates under model substitute-distance figures Differences between actual visit rates. recorded actual and predicted assumption with information be to regarding the utility characteristicsof the sites. used provide could then Connelly and Price (1991) suggesta fundamental change to the Clawson procedure by asking visitors hypothetical questions regarding their expected visit pattern if the site in 2.61

question was to be closed. These responsescould then be fed into the TC model as proxy variables regarding substitute sites. An interesting attempt to formulate such a substitute availability index is given in BqJ6 (1985)5'. Here the degreeof substitutability of a site was measuredby questioning respondents as to their preferencesfor substitute sites. Unfortunately his field experiment found that the majority of respondentsall named one and the same site as their preferred substitute and it became impossible to operationalisethe index. In a recent literature review, Bateman (1993a) concludes that the lack of adequate in sites remains a many TC models. Given the substitute weakness of consideration lack issue the this this and apparent a readily available solution, problem of of pervasiveness in We this study. addressed acknowledge this as a problem which we was not explicitly in in this study (Batemanand Lovett, 1996; Bateman et presented not tackle recent research al., 1996b). 2.3.3.6: Congestion A site becomes congestedwhen the number of visitors at a site rises to the point becomes (i. that the of site characteristics restricted e. the presenceof the of supply where In diminishes the of other users). extreme cases congestion will utility marginal users invalidate a TC study as the observed visits correspond not to the standard demand demand intersect but to the curve with an unknown supply of an undefined constrained system becomes i. under-identified. the system curve e. While Vaux and Williams (1977) feel that this problem is not of "overriding importance", in an early experiment Stankey (1972) records that 82% of his sample felt that "solitude - not seeing many other people except those in your own party" was desirable. Johannson (1987) states that site visitor numbers (YQ may be a separateargument in the individual's utility function. Furthermore,Bateman(1993a)points out that this argumentmay be complex in that, where X,, is very low (or zero), utility may be impaired as people feel lonely or intimidated at the site (this will obviously not be so for all individuals). As X, increases to a small number so utility may rise with the possibility of social interaction. However, as X, becomes large utility may again decline as congestion sets in. The visit

(1993a). in Bateman discussion -'See

2.62

decision may therefore well be dependent upon individual's expectations of X,. Such differences between expected and actual X, might also prove significant in CV studies. The presenceof congestion(or excessdemand)meansthat the observeddemandcurve is an underestimateof true demand. Tle classic treatment of this problem is presentedby Fisher and Krutilla (1972) and summarisedin Bateman (1993b). Ile presenceof congestion depressesdemand for the site such that the TC estimateddemand curve lies below that for the uncongested site. Christensen (1983) suggests that the true demand curve may be revealed by placing successivequantity restrictions upon visitors and re-estimating the TC demand curve at each iteration. However, it is difficult to envisage how such a quantity restriction could be placed upon visitors while still maintaining zero priced visits. Accordingly, in practice, simpler approacheshave been adopted to the quantification of congestion effects. Smith and Desvousges(1986) attempt to account for potential site congestion by level to the managers the site as of site congestion. On the of recreation opinions eliciting basis of received responsesthey concludedthat congestionwas not a significant factor at the it from further Freeman (1979) lists consideration. omitted and several sites studied drawn from the regional population of travel cost the non-visitor samples to of use references if how the many present site non-users would use environmental quality to zones examine 1971; Brown Nawas, 1973; Gum Brewer, (Burt Martin, improved be and and and were to 1975) and such an approach could be extended to the analysis of congestion. However, in (1979,1980,1981,1983) Price that, concludes casesof severe through a series of papers, be techniques than preference may rather revealed more appropriate. congestion, expressed Our own approachhas been to assesscongestionboth subjectively and through inspection of during 71iis have the to survey period. simple approach appears car-parking of the availability been adequatein our empirical studies.

2.3.3.7: Functional form Analysts are faced with a variety of functional forms under which the tgf can be linear, log-log). None has (typically quadratic, semi-log these and of strong specified However the over others. specification of a linear form exhibits a first theoretical ascendancy

derivativewhich will be a constantandis thereforetheoreticallyproblematic.Log formsmay be useful for elasticityestimatesandhavethe advantageof avoidingnegativevaluesfor the 2.63

dependentvariable". However, the double log form may also be criticised on theoretical grounds as its asymptotic properties imply infinite visits at zero costs, an attribute which is particularly unlikely for on-site experiencedemand curves (see Everett, 1979). An altered functional form (even if it has similar explanatorypower) can have a highly significant impact upon the demandcurve and resultant consumersurplus estimates.In a ZTC study of recreational fishing in Grafham Reservoir'(UK), Smith and Kavanagh (1969) found that both semi log (dependentvariable) and double log functions fitted the data very well (R2 However demand 0.97 0.91 the resultant curves were examined respectively)". when and = it was found that, at a zero admissionprice, while the semi log form pmdicted 54,000 annual log form 1,052,000 double predicted over annual visits with obvious consequences the visits for consumer surplus estimates. Subsequent re-estimation made little difference to this divergence. In theory the most appropriatefunctional form may be evaluatedby examining relative W However, tests are strictly non-comparablewhere the dependent degrees of explanation. is by A test to the model with compare visitor rates predicted valid variable changes. more large Wilcoxon either a sample, signed rank tese9 or a rates using visitor actual observed 61 Mann Whitney U test' as appropriate . Because of its large potential for disturbing consumer surplus estimates, we see the functional form issue as one of the most serious problems affecting the TC (as pointed out it may potentially have far more impact than substitute site or congestion effects). Consequently we have made this a priority issue in our applied research. We investigate a below) (see forms functional procedures as well as performing tests estimation and variety of of actual versus predicted arrivals. 2.3.3.8: Estimation procedure Pearceand Markandya (1989) point out that a truncation bias may be introduced where least squaresestimation techniquesare employed. 7be normal error distribution inherent in

"See, for example: Ziemer et al. (1980); Vaughan et al. (1982); Desvousgeset al. (1983); Smith and Desvousges(1986); Hanley (1989); and Benson and Willis (1990). 51Seesubsequentcomments Te.R' figures for ZTC studies. 19Wilcoxon (1945), see Mendenhall et al. (1986), p.806. 'Mann and Whitney (1947). see Kazmier and Pohl (1987), p.496. 6'Box-Cox approachesto fitting functional forms are discussedwith referenceto the hedonic pricing method in Bateman (1993b).

2.64

this technique allows the estimation of continuous and negative visitor rates rather than its discrete non-negativereality. This problem will not be fully solved by simply resorting to log dependentvariable functional forms. OLS estimation is, strictly speaking, inappropriate for TC models and should be replaced by procedures such as maximum likelihood (ML) estimation. Empirical studies come to differing conclusions regarding the extent of variance between OLS (truncated) and ML (non-truncated)estimates of consumer surplus. In a TC found OLS ML (1988) Balkan Kahn hunting deer that and estimates and quality, study of differed by relatively small amounts.On the other hand Garrod and Willis (1991) found that, OLS ML forest surplus and consumer similar sites produced relatively recreation some while factor differing by (one different a of nearly site results estimates,other sites produced very 20). Smith and Desvousges(1986) comparedOLS and ML estimatedTC models for 33 water both Estimates compared, approaches were mean obtained under variance of sites. recreation Using indicating high differences truncation this taken effects. highly as were significant and from further identified highly 33 truncated 11 omitted and were as sites were approach of the investigations. Other work fundamentallyquestionsthe appropriatenessof switching to ML estimation Smith (1988) (1987,1988) Kling bias. Both that, suggest while and truncation to as a counter (once OLS trimmed to remove techniques theoretically appropriate, more ML techniques are surplus estimates. consumer accurate more produce actually may visits) negative predicted in OLS ML both techniques have estimation various of debate and Given this employed we our TC studies. 2.3.3.9: Zonal v individual TC studies Throughout this chapter we have referred to both the zonal and individual variants of in literature is "there (1990) to the Hanley as which no consensus out TC points as the and, both However to is approaches applied are theoretical when grounds". on preferable option disturbingly different Table data of producing results. capable two are the methods the same Using forest joint UK ZTC/ITC to a regard study sites. illustrates with of six 2.5 this point definition (OLS) (full and cost running costs) throughout, procedure the same estimation by ZTC between 40% less to the ranged almost produced surplus consumer of estimates by ITC. As larger those the five than produced all cost coefficients produced times almost 2.65

by both methods are statistically significant this points towards some seriousmethodological problems for one or both of these approaches.

Table 2.5: ZTC/ITC consumersurplusestimatesfor six UK forests ZTC Travel Cost Coefficient

Forest

Brecon Buchan Cheshire Lome New Forest Ruthin Notes:

-0.384 -0.444 -0.525 -0.694 -0.702 -0.396

ITC

CS/visitor (E) 2.60 2.26 1.91 1.44 1.43 2.52

Travel Cost Coefficient -0.358 -0.996 -1.259 -0.327 -0.215 -0.386

CS ratio CS/visitor (E)

ZTCATC

1.40 0.50 0.40 1.53 2.32 1.29

1.86 4.52 4.78 0.94 0.62 1.95

All coefficients produced via OLS techniques and significant at 5% level Travel cost defined as full running costs Consumer surplus estimates at 1988 prices N= 21 for all forests

Sources: Garrod and Willis (1991), Willis and Garrod (1991b). There are a number of methodologicalproblems associatedwith the use of an average it The is impossible dependent that of a rate to use visitor means variable. zonal value as a For individualexample membership of an variables. explanatory specific specify be highly may a well significant predictor of pursuits association outdoor or environmental individual in ZTC information However be the characteristics such cannot visits. recreational is for likely be highly inefficient to such a average variable zonal used, and a constructed (Brown and Nawas, 1973). Similarly, intra-zonal variation is to a considerabledegreelost in dominate in An inter-zonal ZTC, average effects curve-fitting. extremecaseof this may the as occur where concentric zones are used; outer zones may encompass areas which are For from different example, supposethat we were to carry each other. geographically very Malvern Hills (Worcestershire, ZTC the the recreation of estimating value study out a England) using 25 mile wide distance bands.Here the distance band between 100 and 125 Hills both Snowdonia Malvern Mountains North Wales from the encompasses the of miles (see Eastern England figure 2.6). is Fenlands It likely flat of therefore that anyone the and (as for hills Malvern the to visitors presumably have) would have far with a predisposition 2.66

more substitute sites if he lived in Snowdonia than if he lived in the Fens. However, ZTC approachescan at best only construct comparisonsof the attributes of the studied site with those of all sites perceived by the analyst as substitutes, irTespective of the distance individuals would have to travel to reach such substitutes.Such variables will always be weak individualto the specific substitute variables which can be employed by the ITC. compared Figure 2.6: Concentric distance zones around the Malvern Hills

Source: Bateman (1993a) 2.67

Figure 2.6 also highlights a problem with the ZTC if straight line distancesare equated directly with both travel and time costs. Both Snowdonia and the Fens have relatively poor road links with Malvern whereas Leeds (in the same distance zone) has a direct motorway link. Therefore both time and travel costs from Leeds will be considerably less than those for distinction in be lost the a may any zonal average. others, which either of A further problem for the ZTC, which again does not afflict the ITC, is that RI be biased. This arisesas a natural consequenceof aggregating always upwardly statistics will individual responsesacross zones and so reducing the number of curve fitting points to the illustrates 2.7 Figure Panel A individual this the point. shows spread of of zones. number in hypothetical TC being by a survey, each point represented a number recorded observations distance band fitting from demand is defined by In in the turn a away a site. curve the which ITC would employ all these observations as individual points. In panel B these individual into in for ZTC. been The have zonal converted averages a use number of observations 6) (here in been has to the turn thus number of reduced zones which will observations of W fitted line. increase the the spuriously W high Consequentlythe very valuesrecordedin many ZTC studiesshould be treated is indicators has Their as of which only real validity model relatively with extreme caution. higher explanatory power within any particular functional form, their absolute value should be disregarded (and even not reported as it may well be misleading). This criticism does not in figures R2 for ITC this respect,unbiased. are, which apply to the A final criticism of the ZTC approach arises from the methods by which zones are defined. Zones are conventionally defined as concentric circles. However, this need not boundaries. (1985) The definition Bojo for be county uses of the So62' example necessarily influenced by is typically the either arbitrary or availability of of zones number and width implies definition different In data. each possible of effect zones a aggregation population different This in imply in turn Tate. certainly a visitor will almost of population and practice different demand in thereby consumer surplus estimates. and curve changes the estimated Therefore, in practice, it is almost certain that an analyst could respecify zones so as to either inflate or reduce valuation estimates as required. The extent to which such a change is

62Furthermore zonesmaybe cut off at somefinite distancealthoughtheouterbandmay be infinite. Englin in from (1991) tourism their study of rainforest Mendelsohn analyse visits all countries. and

2.68

possible is uncertain and the subject of our ongoing research9. Figure 2.7: R' bias in TC studies A: ITCMand R2 Cost

6 6 666 66

66

66 566

per 6a, 6 go. a 66 6 $a , V1311 as 66 66 a66 "66a66 (C)

666

BY'S

SS'S,

a

a 65 55

565

5,

353 5554 54454 4444444444444 44 44 444 41 44 44444" 41 44 4 44 44 444444444 44 .44444 44 44A44 44 44 44 444 44444333 34 33333 343 3433 33 33 33 3333333333 3 33 3333 33 3333 333 333 33 3333 33 33333 33 3333333 3323333232333 3333333 33 2222222 222223 222222 22 22 i2222 22222 21 22 22222222 222222 2222222222 22 2222 222 222 222 a222 2222222 12 22222 a222,2

44

3

2

2

Number of visits

B: ZTCM and R Cost per visit (C)

Number of visits

Source:Bateman(1993a) '30ne of the few examinationsof the impact of zonal respecification is given by Price et al. (1986) wherein impact being, in the compared, system are upon consumer this case surplus 10 concentric a6 zone zone and a (1983) impact Christensen division the examines In the of changing work, zonal population related minimal. from visits per 100,000population to visits per 1000population. Ilie author is currently examining the problem of zonal rcspecificadon.

2.69

Brown and Nawas (1973) argue that the ZTC is therefore, at best highly inefficient and therefore prefer the use of the ITC, a sentimentechoed by Gum and Martin (1975) and Bowes and Loomis (1980). Indeed the US literature has over the past two decadesslowly moved from use of the ZTC to employing the ITC. However the ITC is not without problems. Dobbs (1991) points out that most ITC studies to date have incorrectly estimated consumer surplus in that they have ignored the inherently discrete nature of the dependent integration demand function may lead to significant In the such cases of a smooth variable. bias in consumer surplus estimates.However, Dobbs develops a programmable approach to the computation of discrete dependentvariable benefits which overcomesthis problem. A more fundamental problem for the ITC occurs where a high proportion of visitors first time visitors (Freeman,1979; Bowes and Loomis, are annum or per one visit make only 1980). In such casesstatistical techniquesused in the ITC may not have a sufficient spread Ironically, have technique the those to operational. make sites the which observations of highest proportions of repeat visitors are also those which are most likely to be attracting a high proportion of locally basedvisits who walk to the site and incur zero monetary travel 1992). Bishop, (e. g. costs In conclusion the decision to use either zonal or individual TC approachesis likely impact While have the there appears no theoretical upon results obtained. to a significant discussion highlights for the this ahead other, one approach of a number of preferring reason both ZTC ITC. Our the the application of associated with and problems methodological ZTC is ITC. the that the that the of evidence against use of exceeds weight of conclusion This is not an ideal criterion for method selection but it seems to be the basic reason for in ITC US UK literature. the the applications current preference and general underlying We are particularly concernedabout the ZTC zonal specification problem and this is avoided by the ITC.

Consequently we employ the latter in our travel cost valuation studies.

However, we do feel that our use of GIS software provides a route for tackling the most demonstrate ZTC. In the this potential we use a visitor rate to order serious problems of dependent variable in our visitor arrivals prediction models. Here we use GIS routines to define true travel time zonesadjustedfor road quality. These extend along high quality road in the ZTC than unrealistic concentric using zones employed conventional rather corridors have Although not extended this approach up to the point of producing ZTC we studies. ITC) future (preferring this our values) provides an obvious extension to this work valuations 2.70

be pursuing. shall which we 2.3.3.10: Non-use values and the relationship with CV estimates TC measuresonly the 'use value' of recreation sites. Underestimation of the total be if due the non-usevalue truncation to the of non-visitors would made worse value of a site is TC both not capable of producing any total was significant. of visitors and non-visitors items in it that non-use such as existence value. cannot estimate economic value estimate This is becausethe basis of the technique is the level of use-basedcosts incurred by visitors in visiting a site. If non-use values are thought to be significant then an appropriate be CV) (e. to these employed capture must the method, valuation contingent g. methodology values. Comparison of CV and TC-derived valuesmay therefore be difficult given the various CV Three perception of potential scenarioscan questions. possible permutation of respondents be envisaged". (i)

Respondentsmay perceive CV questions as relating to their total willingness to pay (WTP) for the use value of the good in question". In such a situation CV and TC divergence here some may expect given even we although comparable measuresare TC (Hicksian) income-compensated the CV measures whilst welfare produces that the 66

(Marshallian) consumer surplus measures produces uncompensated Respondentsmay perceive CV questions as relating to their total YV7P for both the investigation. In a situation, such once we the good under of value non-use and use sated measures have made any adjustment with regard to the compensated/uncomPen

W, (which then still apply), we would expect a residual will scenario of problem

if (1993b). In differences in Bateman to these occur 64Further addition may also scenariosare considered transformation effect. a value CVm up pick surveys post-visit four feasible CVM (ii) (i) the 61several of question: are possible given variants and scenario of permutations for a welfaregain or welfareloss. Seeearlierdiscussion VM or willingnessto accept(WTA) compensation (1992) losses Turner complications surrounding additional asymmetry where of gains and are Bateman and and also reviewed. "As noted previously,whereasstandardeconomictheory would lead us to expectthat the divergence have be theoretical measures welfare will small, advances recent uncompensated and betweencompensated for i. be by those the the case public goods such the as environment, e. provided not this may that shown underscenario(i) maynot beinsignificant(seealsoBatemanand divergencebetweenCVM andTCM measures Turner, 1993).

2.71

difference between CV and TC measuressuch that CV > TC measure. (iii)

If the CV payment vehicle asksrespondentsfor their WTP per visit, frequent visitors in lower than sums occasional visitors an effort to reduce the formers may well offer total annual payments. Such strategiesmay arise from a variety of motivations. For derive visitors may greater marginal utility from a visit than do example occasional frequent face Alternatively, higher burden of overall as visitors may a regular visitors. in incentive indulge free-riding have behaviour. Whatever to they a greater may costs the motivation, if such a result pertains we will find an inverse relationship between CV per visit WTP and the TC measure,which has little to do with their Hicksian and Marshallian roots. These are interesting relationshipsand consequentlycomparisonof CV and TC welfare

focus of our applied research. a made was measures 2.3.3.11: Variants of the TC Apart from those already discussed, a number of variants on the TC have been developed (see review in Bateman, 1993a). Most relevant to this study is the Hedonic TC (HTQ approachdevelopedby Brown and Menddsohn(1984) in which the property prices of 'Me HTC travel hedonic costs. uses travel costs to with are replaced method price the A for than two stage site rather whole values. site attributes separate estimate values independent (including is In the variables one stage employed. estimation process in k for levels separate the are regressed, estimations, attributes) of site each characteristic be implicit The then of each attribute can calculated. price" travel CoStS68. against total Stage two involves the estimation of demandcurves for each attribute by regressingimplicit Summing level the and other explanatory variables7'. over attribute of price on the observed function demand for from an aggregate each attribute which consumer gives all observations has had HTC While be the method considerable recent obtained. surplus estimates may

6'Note that the adjustment arising from scenario(i), which still applies, may feasibly outwcigh that arising from scenario (ii). "Brown and Mendelsohn (1984) only consider travel time, i. e. they assumethat on-site time is a constant, be if (on time this equation cost) would required site were not the case. regression a separate "The implicit price tells us the value of a marginal improvement in attribute i. 70Theseinclude income, other socio-economicvariables and the predicted number of trips from each zone, TC from tgE derived being a standard latter the

2.72

application" its extreme data requirementshave cast doubt upon its practical decisionmaking

asto whetherthe relevantsitecharacteristics may applicability.In particularit is questionable be identified a priori and accuratelymeasured. In chapter 3 we consider an application of the HTC method (Hanley & Ruffell, 1992). However, as this review reveals the extremeproblems of the HTC approachwe do not pursue it in our own applied work. 2.3.4: CONCLUSIONS: THE TRAVEL COST METHOD The TC method is a potentially useful evaluation tool producing uncompensated

is best defined It to the of evaluation well applied consumersurplusestimatesof usevalue. has highlighted forest This those survey several providedat sites. recreationresourcessuchas during TC These the the of arise practical application method. may which potentialproblems include; i.

The decision whether to use zonal and individual approachesand variation in between these methods. results

ii.

Calculation of the cost elements and in particular determination of the opportunity cost of on-site and travel time. Multicollinearity between explanatory variables especially site environmental levels. characteristic

iv.

Problems of heteroskedasticity.

V.

Treatment of substitute sites.

vi.

Accounting for potential congestion effects.

vii.

Choice of the appropriate functional form and its impact upon consumer

Viii.

surplus estimates. Truncation bias and the choice of appropriate estimation technique.

We have outlined our approachto each of theseproblems. In subsequentchapterswe

our effectiveness discussthe practicalimplementationandresultsof our TC studiesandassess in addressingthesemethodologicalissues.

include Loomis HTC (1986); Bell andLeeworthy(1990);Bowes the 71Further approach et al. of examples (1991); Mendelsohn Hanley Ruffell (1992). latter The Englin (1989); threestudiesall and and and Krutilla and describeapplicationsto forestry.

2.73

2.4: VALUATION

METHODS: CONCLUSIONS

This chapterhasprovidedan overviewof methodsfor the monetaryassessment of detailed the two theoretical of and methodological review and a more goods environmental in for subsequentempirical work: the contingent valuation, and application methods chosen have how indicated have We addressed the various we also travel cost methods. 3 In implementation by these the of methods. chapter we methodological problems raised discuss and present results from our applied evaluation work using these methods.

2.74

REFERENCES Adamowicz, W. L., Boxall, P.C., Louviere, J.J., Swait, J. and Williams, M. (forthcoming) Stated preference methods for valuing environmental amenities, in Bateman, I.J. and Willis, K. G. (eds) Contingent Valuation of Environmental Preferences:Theory and Practice in the US, Europe and Developing Countries, Oxford University Press,Oxford.

Adamowicz,W.L., Louviere,J. andWilliams, M. (1994)Combiningstatedand revealedpreferencemethods for valuing environmentalamenities,Journalof EnvironmentalEconomicsand Management, 26:271292. Ajzen, 1.and Fishbein,M. (1977)Attitude-behaviour relations:a theoreticalanalysisand reviewof empirical research.PsychologicalBulletin 84(5):888-918. Aizen, 1.and Peterson,G.L. (1988)Contingentvaluemeasurement: the price of everythingand the valueof nothing?in Peterson,G.L., Driver B.L. & Gregory,R. (eds.), AmenityResourceValuation:Integrating Economicswith OtherDisciplines. VenturePublishingInc., pp.65-76. Andreoni,J. (1990)Impurealtruismanddonationsto public goods:a theoryof warm-glowgiving. EconomicJournal 100:464-477. Balkan,E. and Kahn,J.R. (1988)The valueof changesin deerhuntingquality: a travel-costapproach, Applied Economics,20:533-539. Banford,N., Knetsch,J. andMauser,0. (1977)"Compensating and equivalentmeasures of consumers Department Economics, Simon FraserUniversity,Vancouver. further of survey results". surplus: Barnett,A. H. and Yandle,B. (1973)"AllocatingEnvironmentalResources".Public Finance28:11-19. Bateman,I.J. (1991)"A critical analysisof COBA andproposalsfor an extendedcost benefitapproachto transport transportdecisions",in Hanna,J. (ed.) "Whatare roadsworth?:Fair assessmentfor 2000, Foundation/Transport London. New Economics expenditure"t Bateman,I.J. (1992)The economicevaluationof environmentalgoodsand services,Integrated EnvironmentalManagement14:11-14. revealedpreferencemethods, Bateman,I.J. (1993a)Valuation of the environment,methodsand techniques: PrinciplesandPractice,Belhaven in TurnerR.K. (ed.) SustainableEconomicsand Management: press,London,pp.192-265. Bateman,I.J. (1993b)Valuationof the environment:a surveyof revealedpreferencemethods,Global EnvironmentalChangeWorkingPaper93-06,Centrefor SocialandEconomicResearchon the Global Environment,Universityof East Anglia andUniversityCollegeLondon,pp.84. Bateman,I.J. (1995a)Researchmethodsfor valuingenvironmentalbenefits,in Dubgaard,A., Bateman,I.J. Countryside Stewardship, ) Valuation Beneflitsfrom (eds. Economic M. Merlo, of and Vauk, Kiel, pp.47-82. Wissenschaftsverlag Bateman,I.J. (1995b)Environmentalandeconomicappraisal,in O'Riordan,T. (ed.) EnvironmentalScience Longmans,Harlow. for EnvironmentalManagement, Bateman,I.J. and Bryan,F. (1994)Recentadvancesin the monetaryevaluationof environmental Sustainable Economics, Management Countryside in Environmental (conference the and preferences, Cardiff. Network, Recreation Countryside proceedings), Bateman,I.J. and Lovett, A.A., (1996)EvaluatingRecreationDemandfor NaturalAreas:A GIS Benefit TransfersApproach,End of AwardReportto theEconomicand SocialResearchCouncilfor award Sciences, School Environmental University East Anglia. L320223002, of of number Bateman,I.J. and Turner,R.K. (1992)Evaluationof the environmentthe contingentvaluationmethod, Global EnvironmentalChangeWorkingPaper92-18,Centrefor SocialandEconomicResearchon the Global Environment,Universityof EastAnglia and UniversityCollegeLondon,pp.93. Bateman,I.J. and Turner,R.K. (1993)Valuationof the environment,methodsand techniques:the contingent ) Sustainable in R. K. (ed. Environmental Turner, Economics and Management: method, valuation principles and Practice,BelhavenPress,London. Bateman,IJ., Brainard,J.S., Lovett, A.A. andGarrod,G.D. (1996b)"Modelling WoodlandValuesUsing GeographicalInformationSystems:A Note ConcerningRecentWork", presentedat the SeventhAnnual Meetingof the EuropeanAssociationof Environmentaland ResourceEconomists,UniversidadeNova de Lisboa,Portugal,June27th-29th,1996. G. (1992) K. Willis, An introductionto the estimationof non-pricedrecreation G. Gaffod, IJ., and Bateman, Countryside ESRC Change Initiative WorkingPaper36, Universityof method, travel-cost the using

2.75

Newcastle upon Tyne. Bateman, I. J., Green, C., Tunstall, S. and Turner, R.K. (1991) "The Contingent Valuation Method". Report to the Transport and Road ResearchLaboratory, Departmentof Transport, Bracknell. pp.91. Bateman, I. J., Langford, I. H., Willis, K. G., Turner, R.K. and Garrod, G.D. (1993) The impacts of changing willingness to pay question format in contingent valuation studies: an analysis of open-ended,iterative bidding and dichotomous choice formats, Global Environmental Change Working Paper 93-05, Centre for Social and Economic Researchon the Global Environment, University of East Anglia and University College London, pp.50. Bateman, I. J., Langford, I. H., Tumer, RX, Willis, K. G. and Garrod, G.D. (1995a) "Elicitation and Truncation Effects in Contingent Valuation Studies". Ecological Economics, 12(2):161-179. Bateman, I. J., Munro, A., Rhodes,B., Starmer, C. and Sugden,R. (1995b) A test of the theory of referencedependentpreferences,CSERGEGlobal Environmental Change Working Paper 95-19, Centre for Social and Economic Researchon the Global Environment, University of East Anglia and University College London, pp.29. Bateman, I.J., Munro, A., Rhodes,B., Starmcr, C. and Sugden,R. (forthcoming a) "A test of the theory of Quarterly Journal Economics. of preferences", reference-dependent Bateman, I. J., Munro, I.J., Starmer, C. and Sugden,R. (1996a) "Does part-whole bias exist? An experimental investigation", CSERGEGlobal Environmental Change Working Paper 96-03, Centre for Social and Economic Researchon the Global Environment, University of East Anglia and University College London. Bateman, I. J., Munro, A., Starmer, C. and Sugden,R. (forthcoming b) "Does part-whole bias exist? An Journal. Economic investigation", experimental Bateman, Ii., Willis, K. G. and Gaffod, G.D. (1994) Consistencybetween contingent valuation estimates:a Studies, 28(5): 457-474. National Parks, Regional UK two of studies comparison of Bateman, Ij., Willis, K. G., Garrod, G., Doktor, P., Langford, I.H. and Turner, R.K. (1992) "A Contingent Valuation Study of the Norfolk Broads". Report to the National Rivers Authority, pp.403. for beach by days, Journal Recreational demand V. (1990) tourists saltwater Leeworthy, of F. Bell, and Environmental Economics and Management, 18:189-205. benefits Forestry (1990) The K. G. Willis, the F. of recreation of J. aggregate value non-priced Benson, and Commission estate. Report to the Forestry Commission, Dept. of Town and Country Planning, University of Newcastle, Newcastle. Benson, J.F. and Willis, K. G. (1992) Valuing informal recreation on the Forestry Commission estate, Forestry CommissionBulletin 104, HMSO, London. "Public Environmental Amenity Benefits Private (1985) Stoll, J. R. L. B. Dillman, C., of J. and Bergstrom, Agricultural Economics 17(l). Southern Journal Land", Agricultural Prime Case of Land: the of Forests: Community benefits Assessing (1992) an evaluation of the recreational use the of K. Bishop, benefits of two urban fringe woodlands,Journal of Environmental Planning and Management, 35(l): 63-76. indirect biased? " "Measuring (1979) T. Hebericin, are of extra market goods: measures values R. Bishop, and 926-930. 61(5): Economics Agricultural Journal American of Study" Sandhill 1984 Report (1985) "Progress T. A. the Heberlein, of preliminary report to C. Bishop, R. and in Carson (1989). Mitchell Resources, Natural Department and Wisconsin of the "Hypothetical in Contingent (1983) Bias Valuation: Results MJ. Kealy, T. A. Heberlcin, C., and R. Bishop, 23: 619-633. Journal Resources Natural Market". Simulated from a (1984) "Does Contingent Valuation R. A. Baumgartner, M. P. Welsh, T. A., Heberlein, C., and Bishop, R. Work?: A Report on the Sandhill Study". Paper presentedat the Joint Meeting of the Association of Environmental and ResourceEconomists and the American Economics Association, August 1984, Cornell University, New York. Blarney, R.K. (1996) Citizcns, consumersand contingent valuation: clarification and the expression of citizen P.C., Luckert, M. K.. Phillips, W. E. and White. Roxall, W. L., in Adamowicz, issue-opinions, values and W. A. Forestry, Economics and the Environment, CAB International, Wallingford, Oxon. Boadway, R. and Bruce, N. (1984) WeVareEconomics, Basil Blackwell, Oxford. Strand, (1991) in I. E. Recreation, Braden, E. J.B. and Kolstad, C.D. K. McConnell, E., and N. Bockstael, (eds.) Measuring the Demandfor Environmental Quality, North-Holland, Amsterdam. (1987) Hanemann, W. M. "Time Jr. I. E. Strand, E., and the recreational demand model, N. and Bockstacl,

2.76

American Journal of Agricultural Economics, 69(2):293-302. Bohm, P. (1972) Estimating Demand for Public Goods: An Experiment". European Economic Review 3:733-736. BqJ6, J. (1985) A cost-benefit analysis offorestry in mountainous areas: the case of Valatialen. Stockholm School of Economics, Sweden. Bowes, M. and Krutilla, J. (1989) Multiple-use Management: The Economics of Public Forestlands. Resourcesfor the Future, Washington D.C. Bowes, M. D. and Loomis, J.B. (1980) A note on the use of travel cost models with unequal zonal populations, Land Economics, 56:465-70. Bowes, M. D. and Loomis, J.B. (1982) A note on the use of travel cost models with unequal zonal 408410. Land Economics, 58: Reply, populations: Boyle, KJ. (1989) "Commodity Specification and the Framing of Contingent Valuation Questions", Land Economics 65:57-63. Boyle, KJ. and Bishop, R.C. (1988) "Welfare MeasurementsUsing Contingent Valuation: A Comparison of Techniques", Journal of the American Agricultural Association, February, pp.20-28. Boyle, K. J., Bishop, R.C. and Welsh, M. P. (1985) "Starting Point Bias in Contingent Valuation Surveys". Land Economics 61: 188-194. Brennan, A. (1992) Moral Pluralism and the Environment, Environmental Values 1:15-53. Brookshire, D.S. and Coursey, D.L. (1987) "Measuring the value of a public good: an empirical comparison Economic Review 77: 554-566. American procedures", of elicitation Brookshire, D.S., Ives, B.C. and Schulze,W. C. (1976) *The Valuation of Aesthetic Preferences".Journal of Environmental Economics and Management 3(3):325-346. Brookshire, D.S., Randall, A. and Stoll, J.R. (1980) "Valuing increments and decrementsin natural resource Agricultural 62(3): 478488. Economics Journal American flows". of service Brookshire, D.S., Thayer, M. A., Schulze, W. D. and d'Arge, R.C. (1982) "Valuing public goods: a Review 71: 165-177. American Economic hedonic approaches", and of survey comparison hedonic Economics Statistics, The Review (1984) R. Mendelsohn, travel G. cost model, of and Brown, and 66:427433. Brown, R.A. and Green, C.H. (1981) "Threats to Health and Safety: PerceivedRisk and Willingness to Pay". Social Science and Medicine 15(C):67-75. Brown, T. C. and Slovic, P. (1988) Effects of Context on Economic Measuresof Values. In. Peterson,G.L., Driver, B.L. and Gregory, R. (eds) Amenity ResourceValuation: Integrating Economics with Other Disciplines, Venture State College, PA, pp-23-30. Brown, W. G. and Nawas, F. (1973) Impact of aggregationon the estimation of outdoor recreation demand functions, American Journal of Agricultural Economics, 55(2):246-249. Brubaker, E. (1982) "Sixty-Eight PercentFree Revelation and Thirty-Two Percent Free Ride? Demand Disclosures Under Varying Conditions of Exclusion" in V. L. Smith, ed., Researchin Experimental Economics, Vol. 2, JAI Press,Greenwich, Conn. Buckley, G.P. (ed.) (1989) Biological Habitat Reconstruction, Belhaven Press,London. S. (1983) M. "Valuing Policies Which Reduce Walbert, A. F. G., R. Mehr, Cummings, S., H. and Burness, Environmental Risk". Natural ResourcesJournal 23:675-682. from benefits Econometrica, (1971) Estimation D. Brewer, of net social outdoor recreation, O. R. Burt, and 39(5): 813-827. Goods Non-Market Using Refendum Data". Journal of Valuing for New Paradigm "A (1988) T. Cameron, Environmental Economics and Management. Cameron, T. and James, M. (1987) "Efficient Estimation Methods for Use with Closed Ended Contingent Valuation Survey Data". Review of Economics and Statistics 69:269-276. Carson, R.T. (1999) "Constructed Markets". Mimeo, in KristrOrn (1990a). Carson, R.T. and Mitchell, R.C. (1993) "A Reestimationof Bishop and Hcbcrlcin's Simulated Market Hypothetical Markets - Travel Cost Results Under Alternative Assumptions", Discussion Paper D-107, Resourcesfor the Future, Washington, D.C. Carson, R.T. and Mitchell, R.C. (1993) The value of clean water the public's willingness to pay for boatable, fishable and swimmable quality water, Water ResourcesResearch,29(7):2445-2454. W. M., Kopp, C., Hanemann, R.J., Presser,S. and Ruud, P.A. (1992) A contingent R. Mitchell, T., R. Carson, from Exxon Valdez lost Report the values resulting use to the passive oil spill, of study valuation

2.77

Attorney General of the State of Alaska, Natural ResourceDamage Assessment,Inc., San Diego. Carson,R.T., Hanemann,W. M., Kopp, RJ., Krosnick. J.A., Mitchell, R.C., Presser,S., Ruud, P.A. and Smith, V. K. (1995) Temporal stability of estimatesfrom contingent valuation. Discussion Paper 9537, Resourcesfor the Future, Washington, D.C. Cesario, F.J. (1976) Value of time in recreation benefit studies", Land Economics, 55:32-41. Cesario, F.J. and Knetsch, J.L. (1970) Time bias in recreation benefit estimates", Water ResourcesResearch, 6(3):700-704. Cesario, F.J. and Knetsch, J.L. (1976) A recreation site demand and benefit estimation model, Regional Studies, 10.97-104. Cheshire, P.C. and Stabler, MJ. (1976) Joint consumption benefits in recreational site 'surplus'; an empirical estimate, Regional Studies, 10-343-351. Christainsen, G.B. (1982) "Evidence for Determining the Optimal Mechanism for Providing Collective Goods". American Economist 26(l): 57-61. Christensen,J.B. (1983) An economic approach to assessingthe value of recreation with special referenceto forest areas,Ph.D Thesis,University College of North Wales, Bangor. Christensen,J.B. and Price, C. (1982) A note on the use of travel cost methods with unequal zonal populations: Comment, Land Economics, 58(3):395-399. Christensen,J.B., Humphreys, S.K. and Price, C. (1983) A revised Clawson method: One part solution to in disaggregation problems recreation evaluation. Journal of Environmental multidimensional Management, 20:333-346. Cicchetti, CJ. and Freeman,A. M. 111(1971) "Optional Demand and Consumer's Surplus: Further Comment". Quarterly Journal of Economics 85(3):528-539. Clawson, M. (1959) Methods of measuring the demandfor and value of outdoor recreation, Reprint series, Washington D. C. Future, for Resources 10, the no. Clawson, M. and Knetsch, J.L. (1966) Economics of outdoor recreation. Resourcesfor the Future, John Hopkins Press, Baltimore. Common, M. (1973) A note on the use of the Clawson method, Regional Studies, 7:401-406. Conger, AJ. (1974) A revised definition for suppressorvariables: a guide to their identification and interpretation, Educational and Psychological Measurement,34:35-46. Connelly, D. S. and Price, C. (1991) The Clawson method and site substitution: hypothesisand model, College North Wales, Bangor. University of manuscript, Converse, J. M. and Presser,S. (1986) "Survey Questions: Handcrafting the StandardisedQuestionnaire", Sage, Beverley Hills, California. Coursey, D.L., Hovis, J. and Schulze, W. D. (1986) "The disparity between willingness to accept and Quarterly Journal Economics 679-690. 102: to of of measures value". pay willingness Coursey, D.L., Schulze, D.W. and Hovis, J. (1983) "A comparison of alternative valuation mechanismsfor University Wyoming. of commodities". non-market Cronin, F.J. and Herzeg, K. (1982) "Valuing Nonmarket Goods through Contingent Markets" given in Cummings ct al. (1986). Cummings, R.G., Brookshire, D. S. and Schulze,W. D. eds. (1986) Valuing Environmental Goods: A State of Contingent Method, Rowman Allenheld, Totowa, NJ. Deaton, A. and Assessment Arts the and of the Muellbauer, J. (1980). Economics and Consumer Behaviour, Cambridge University Press,New York. Consumer (1980) J. Economics Behaviour, Cambridge University Press, Muellbauer, and A. Deaton, and Cambridge. Department of Transport (1987) Valuesfor journey time savings and accident prevention, Department of Twsport, London. Desvousges,W. H., Smith, V. K. and Fisher, A. (1987) Option Price Estimates for Water Quality Improvernem Journal of Environmental Economics and Management 14:248-267. Desvousges.W. H., Smith, V. K. and McGivney, M. P. (1983) A comparison of alternative approachesfor , benefits improvements. Report no. EPA-230-05-83and related of water quality recreation estimating 001 Office of Policy Analysis, U. S. Environmental Protection Agency, Washington, D.C. Dickie, M., Fisher, A. and Gerking, S. (1987) "Market transactionsand hypothetical demand data: a Journal American Statistical Association 82: 69-75. the of study", comparative Dilman, D. A. (1978) Mail and TelephoneSurveys: The Total Design Method, Wiley, New York. Dobbs, I. M. (1991) The individual travel cost method: estimation and benefit assessmentwith a discrete and

2.78

possibly grouped dependentvariable, Countryside Change Unit Working Paper 17, University of Newcastle upon Tyne. Duffield, J.W. and Patterson,D.A. (1991) "Inference and Optimal Design for a Welfare Measure in Dichotomous Choice Contingent Valuation". Land Economics 67(2):225-239. Eberle, W. D. and Hayden, F.G. (1991) "Critique of Contingent Valuation and Travel Cost Method for Valuing Natural Resourcesand Ecosystems"Journal of Economic Issues 25:649-685. Ehrlich, P.R. and Ehrlich, A. G. (1992) The Value of Biodiversity, Ambio 21:219-226. Englin, J. and Mendelsohn, R. (1991) A hedonic travel cost analysis for valuation of multiple componentsof site quality: the recreation value of forest management,Journal of Environmental Economics and Management, 21:275-290. Epp, D. and Gripp, S. (1993) Test-retestreliability of contingent valuation estimatesfor an unfamiliar policy in forests, J. (compiler), W-133 Benefit Costs Bergstrom, Transfers in tropical of and valuation choice: Natural Resource Planning, Department of Agricultural and Applied Economics, University of Georgia, Athens, GA. Everett, R.D. (1979) The monetary value of the recreational benefits of wildlife, Journal of Environmental Management, 8:203-213. Fishbein, M. and Azjen, 1. (1975) Belief, Attitude. Intention and Behaviour: An Introduction to Theory and Research, Addison-Wesley, Reading, Mass. Fisher, A. C. and Krutilla, J.V. (1972) Determination of optimal capacity of rmurce-based recreation facilities, in Krutilla, J.V. (ed.) Natural Environments: Studies in Theoretical and Applied Analysis. Resourcesfor the Future, John Hopkins University Press,Baltimore, pp.352. Freeman, A. M. 111(1979) The benefits of environmental improvement: theory and practice, Resourcesfor the Future. John Hopkins University Press,Baltimore and London. Freeman, A. M. (1986) "On Assessingthe State of the Arts of the Contingent Valuation Method of Valuing Environmental Changes" in Cummings, R.G., Brookshire, D.S. and Schulze, W. D. eds., Valuing Environmental Goods, Rowman and Allanfield, Totowa, NJ. Garrod, G.D. and Willis, K. G. (1990) "Contingent valuation techniques:A review of their unbiasedness, Change Working Paper No. 10. Countryside Countryside Initiative: consistency", efficiency and Change Unit, University of Newcastle upon Tyne. Garrod, G.D. and Willis, K. G. (1991) Some empirical estimatesof forest amenity value, Working Paper 13, Countryside ChangeUnit, University of Newcastle upon Tyne. Garrod, G.D. and Willis, K. G. (1991) "The hedonic price method and the valuation of countryside 14, Change University Newcastle Upon Unit Working Paper Tyne. Countryside of characteristics". Garrod, G.D. and Willis, K. G. (1992) The environmental economic impact of woodland: a two stage hedonic in Applied Economics 24: forestry Britain, 714-728. the of amenity value of pricing model Gibson, J.G. (1978) Recreation land use, in Pearce,D.W. (ed.). The Valuation of Social Cost, Allen and Unwin, London. CATMOG factor introduction 7, Geo Books, Norwich. (1976) An A. Kirby, to J. analysis, Goddard, and Green, C.H. and Tunstall, S.M. (1991) "The Evaluation of River Quality Improvements by the Contingent Valuation Method", Applied Economics 23: 1135-1146. Green, C.H., Tunstall, S.M., N'Jai, A. and Rodgers, A. (1990) "The Economic Evaluation of Environmental Goods". Project Appraisal 5(2):70-82. in demand Forecasting to the (1977) changes response recreational site characteristics,in; Eisner, J. P. Greig, G.H. (ed.) State of the Art Methodsfor Research,Planning and Determining the Benefits of Outdoor Recreation, USDA Forest Service, Washington, D.C., 11-22. in (1975) E. Problems W. Martin, estimating the demand for and value of rural L. solutions and Gum, R. and American Agricultural Journal Economics, 57: 558-566. of outdoor recreation, (1974) G. M. Jr. Waterfowl and wetlands: toward bioeconomic analysis, Johns Brown, J. Hammack, and Hopkins University Press (for Resourcesfor the Future), Baltimore. "Willingness how (1986) M. to W. to pay and willingness accept: much can they differ? ", mimeo, Hanemann, Department of Agricultural and ResourceEconomics, University of California, Berkeley, in Mitchell (1989). Carson and Pay "Willingness (1991) M. to W. and Willingness to Accept: How Much Can they Differ? ", Hanemann, American Economic Review 81(3):635-647. The (forthcoming) M. W. economic theory of WTP and WTA, in Bateman, I.J. and Willis, K.G. Hanemann,

2.79

(eds.) Contingent Valuation of Environmental Preferences:AssessingTheory and Practice in the USA, Europe, and Developing Countries, Oxford University Press. Hanemann,W. M., Loomis, J. and Kanninen, B. (1991) "Statistical efficiency of double-boundcddichotomous choice contingent valuation", American Journal of Agricultural Economics 73(4): 1255-1263. Hanley, N. D. (1989) Valuing rural recreation benefits: an empirical comparison of two approaches,Journal of Agricultural Economics, 40:361-74. Hanley, N. D. (1990) "Valuation of Environmental Effects: Final Report - Stage One". Industry Department of Scotland and the Scottish Development Agency, Edinburgh. Hanley, N. D. and Common, M. S. (1987) Evaluating the recreation wildlife and landscapebenefits of forestry: preliminary results from a Scottish study, Papers in Economics Finance and Investment, no. 141. University of Stirling, Scotland. Hanley, N. D. and Munro, A. (1991) "Design Bias in Contingent Valuation Studies: The Impact of Information", Discussion Paper in EconomicsNo. 13, Departmentof Economics, University of Stirling. Hanley, N. D. and Ruffell, R. (1992) The valuation of forest characteristics,Working Paper 849, Institute for Economic Research,QueensUniversity, Kingston, Ontario. Harris, C.C. and Brown, 0. (1992) Gain, Loss and PersonalResponsibility: The Role of Motivation in Resource Valuation Decision-Making, Ecological Economics 5:73-92. Harris, C.C., Driver, B.L. and McLaughlin, M. J. (1989) "Improving the Contingent Valuation Method: A Psychological Approach", Journal of Environmental Economics and Management 17:213-229. Heberlein, T. A. (1986) "Measuring ResourceValues: The Reliability and Validity of Dichotomous Contingent Valuation Measures",paper presentedat the American Sociological Association Meeting, New York. August. Heberlein, T. and Bishop, R. (1986) "Assessingthe Validity of Contingent Valuation: Three Experiments". Science of the Total Environment 56:99-107. Heberlein, T. A. and Black, J.S. (1976) "Attitudinal Specificity and the Prediction of Behaviour in a Field Setting". Journal of Personality and Social Psychology 33:434479. H. M. Treasury (1972) Forestry in Great Britain: An Interdepartmental Cost/Benefit Study, HMSO, London. H. M. Treasury (199 1) Economic Appraisal in Central Government:A Technical Guide for Government Departments, HMSO, London. Hicks, J.R. (1943) "The four consumer surpluses".Review of Economic Studies 11:3141. Hill, R.J. (1981) "Attitudes and Behaviour" in Morris Rosenbergand Ralph H. Turner, eds., Social psychology: Sociological PerspectivesBasic Books, New York. Hoehn, J.P. and Randall, A. (1987) "A satisfactory bcnefit cost indicator from contingent valuation". Journal 226-247. Management 14(3): Economics Environmental and of Hoen, H.F. and G. Winther (1991) "Attitudes to and willingness to pay for multiple-use forestry and in from Department Forestry, forests Norway". Report Agricultural the of coniferous of preservation University of Norway. Hoevenagel, R. (1990) "The validity of the Contingent Valuation method: Some aspectson the basis of three Dutch studies". Paper presentedat die first annual meeting of the European Association of Environmental and ResourceEconomists (EAERE), Venice, April 17-20 1991. Hoevenagel, R. (1991) "A Contingent Valuation (CV) experiment to test 'part-whole bias'". Paper presented Stockholm, June 10-14 1991. EAERE, meeting of annual at the second Hoevenagel, R. (1996) The validity of the contingent valuation method: perfect and regular embedding, Environmental and ResourceEconomics, 7(l): 57-78. Hufschmidt, M. M., James,D.E., Meister, A. D., Bower, B.T. and Dixon. J.A. (1983) Environment, Natural Systemsand Development: An Economic Valuation Guide. John Hopkins University Press,Baltimore and London. Johansson,R-0. (1987) The Economic Theory and Measurementof Environmental Benefits. Cambridge University Press. Cambridge. Johansson,P-0. (1991) An Introduction to Modern We6rareEconomics, Cambridge University Press. Johnson, B. M. (1966) Travel time and the price of leisure, WesternEconomic Journal, 4: 135-45. Johnston, R.J. (1978) Multivariate Statistical Analysis in Geography. Longman, London. Jones, K. (1991) Muld-level models for geographicalresearch,CATMOG 54, Institute of British Geographers,Norwich. Jones-Loe, M. W. (1989) The Economics of Safety and Physical Risk, Basil Blackwell, Oxford.

2.80

Just, R.E., Hueth, D.L. and Schmitz, A. (1982) "Applied Welfare Economics and Public Policy", Prentice Hall, Englewood Cliffs, NJ. Kahneman, D. (1986) "Comments" in Cummings et al. (1986). Kahneman D. and Knetsch, J.L. (1992a) "Valuing Public Goods: The Purchaseof Moral Satisfaction" Journal of Environmental Economics and Management 22(l): 57-70. Kahneman, D. and Knetsch, J.L. (1992b) "Contingent Valuation and the Value of Public Goods: Reply". Journal of Environmental Economics and Management22(l): 90-94. Kahneman, D. and Tversky, A. (1979) "Prospect theory: an analysis of decisions under risk". Econometrica 47(2): 263-291. Kahneman, D. and Tversky, A. (1982) "The Psychology of Preferences",Scientific American, January. pp.2136-141. Kahneman, D. and Tversky, A. (1984) "Choices, Values and Frames", American Psychologist 39:341-350. Kahneman, D., Knetsch, J.L. and Thaler, R.H. (1990) Experimental tests of the endowment effect and the Coase theorem, Journal of Political Economy, 98: 1325-1348. Kahneman, D., Slovic, P. and Tversky, A. (1982) Judgement Under Uncertainty: l1euristics and Biases, Cambridge University Press,New York. Kazmier, LJ. and Pohl, N.F. (1987) Basic Statisticsfor Businessand Economics, McGraw-Hill, Singapore. Kealy, MJ., Montgomery M. and J. F. Dovido (1990) "Reliability and predictive validity of contingent values: Does the nature of the good matter?" Journal of Environmental Economics and Management 19.244-263. Kling, C.L. (1987) A simulation approachto comparing multiple site recreation demand models using ChesapeakeBay survey data, Marine ResourceEconomics, 4:95-109. Kling, C.L. (1988) Comparing welfare estimatesof environmental quality changesfrom recreation demand models, Journal of Environmental Economics and Management, 15:331-340. Kneese, A. V. (1984) "Measuring the Benefits of Clean Air and Water", Resourcesfor the Future, Washington, D. C. Knetsch, J.L. and Davis, R.K. (1966) "Comparisons of Methods for Recreation Evaluation", in Allen V. Kneese and StephenC. Smith, eds., Water Research (Baltimore, The Johns Hopkins University Press for Resourcesfor the Future). Knetsch, J. and Sinden, J. (1984) "Willingness to pay and compensationdemanded:Experimental evidence of an unexpecteddisparity in measuresof value", Quarterly Journal of Economics 99:507-521. Kristr6m, B. (1990a) "Valuing environmental benefits using the contingent valuation method - an econometric analysis". UmeA Economic Studies No. 219, University of UmeA, Sweden. Kristr6m, B. (1990b) "W. Stanley Jevons (1888) on option value". Journal of Environmental Economics and Management 18:86-87. Kristrom, B. and Riera. P. (1996) Is the income elasticity of environmental improvements less than one?. Environmental and ResourceEconomics, 7(l): 45-55. Krutilla, J.V. and Fisher, A. C. (1975) The Economics of Natural Environments: Studies in the Valuation of Commodity and Amenity Resources,Johns Hopkins University Press(for Resourcesfor the Future), Baltimore. Langford, 1. (1992) Childhood leukaemia and infections, PhD thesis, School of Environmental Sciences, University of East Anglia. Langford, I. H. and Bateman, I. J. (1993) Welfare measuresfor contingent valuation studies: estimation and reliability, Global Environmental Change Working Paper 93-04, Centre for Social and Economic Researchon die Global Environment, University of East Anglia and University College London, pp.30. Langford, I. H., Bateman, I.J. and Langford, H. D. (1996) "A Multilevel Modelling Approach to TripleBounded Dichotomous Choice Contingent Valuation". Environmental and ResourceEconomics, 7(3): 197-211. Lochman, E. and De. V. H. (1982) "Application of stochasticchoice modelling to policy analysis of public improvements", Review of Economics and Statistics 64:474-480. air quality of study case a goods: Loomis, J.B. (1987) "Balancing public trust resourcesof Mono Lake and Los Angeles' water right: An Water Resources Research 23: 1449-1456. approach". economic Loomis, J.B. (1989) "Test-retest reliability of the contingent valuation method: a comparison of general American Journal Agricultural Economics 71:76-84. response", and visitor of population Loomis, J.B. (1990) "Comparative Reliability of the Dichotomous Choice and Open-EndedContingent

2.81

Valuation Techniques" Journal of Environmental Economics and Management 18(l): 78-85. Loomis, J., Sorg, C., and Donnelly, D. (1986) Economic losses to recreational fisheries, due to small-head power development, Journal of Environmental Management,22:85-94. Lucas, R.C. (1963) Bias in estimating recreationistslength of stay from sampling interviews, Journal of Forestry, 61:912-914. Lumsden, J. (1976) "Test Theory", Annual Review of Psychology 27:251-280. Maddala, G. S. (1988) Introduction to Econometrics, Macmillan, New York. Magnussen,K. (1991) "Valuation of reduced water pollution using the contingent valuation method methodology and empirical results". Paper presentedat the secondannual mecting of EAERE, Stockholm, June 10-14 1991. Mann, H.B. and Whitney, R. (1947) On a test of whether one or two random samplesis stochastically larger than the other, Annals of Mathematical Statistics, 18:50-60. Mannesto, G. and Loomis, J.B. (1991) "Evaluation of Mail and In-Person Contingent Value Surveys: Results Economics Management Boaters". Journal Environmental Recreational Study and of of of a 32: 177-190. Marwell, G. and Ames, R.E. (1981) "Economists Free Ride, Does Anyone Else? Experiments on the Provision of Public Goods", pt 4, Journal of Public Economics 15:295-310. Maslow, A. (1970) Motivation and Personality, Harper and Row, New York. McConnell, K. (1975) Some problems in estimating the demand for outdoor recreation, American Journal of Agricultural Economics, 57(2):330-334. McConnell, K. E. (1992) On-site time in the demand for recreation. McConnell, K. E. and Strand, 1. (1981) Measuring the cost of time in recreation demand analysis: an Journal Agricultural 63(l): Economics, 153-56. fishing, American to of sport application McCullagh, P. and Nelder, J.A. (1983) GeneralizedLinear Models, Chapmanand Hall, London. Medley, G. (1992) Nature, the environment and the future, presentedat The University of the Third Age at Cambridge: 1992 International Symposium, The Challenges of the Future, l3th/18th September 1992, Kings College. Cambridge. Mendenhall, W., Reinmuth. J.E., Beaver, R. and Dunhan, D. (1986) Statisticsfor Management and Economics, 5Lh ed., Duxbury Press,Boston. Milon, J.W. (1989) "Contingent valuation experiments for strategic behaviour". Journal of Environmental Economics and Management 17:293-308. Tests in Empirical Bias Contingent Review (1988) "A Benson, J. F. G. K. Willis, A. R., of L. and and Mitchell, Valuations of Wildlife Resources".Working Paper No. 5, Departmentof Town and Country Planning, University of Newcastle upon Tyne. Mitchell, R.C. (1991) Comments regarding Kahnemanand Knctsch; made at the 2nd Annual Conference of Stockholm June 11-14,1991. Economists, Resource Environmental Association European and of the in for "An determining (1981) R. Carson, to C. pay R. experiment willingness national water Mitchell, and Washington, Protection Agency, U. S. Environmental D. C. draft improvements", to the report quality Goods: Contingent Valuation "Using Surveys Value Public (1989) R. Carson, to the C. R. Mitchell, and Method". Resourcesfor the Future, Washington D.C. Morey, E.R. (1981) ne demand for site-specific recreational activities: a characteristicsapproach,Journal of Environmental Economics and Management, 8(4):345-71. Economic Review, 74(l): 163-73. American Confuser (1984) R. E. surplus, Morey, Morey, E.R. (1985) Characteristics,consumer surplus and new activities: a proposed ski area,Journal of Public Economics, 26:221-36. Navrud, S. (1989a) "Estimating Social Benerits of Environmental Improvements from Reduced Acid Depositions: A Contingent Valuation Survey" in Folmer, H. and van lerland, E., Valuation Methods Economics, Elsevier, Environmental Amsterdam. in Making Policy and in Estimates Environmental Decision Making: Case Study on Benefits Use "The (1999b) S. of Navrud, Norway", OECD. Navrud, S. (1991) "Willingness to Pay for Preservationof Species:An Experiment with Actual Payments" Conference 2nd Annual European Association the the of at of Environmental and paper presented Resource Economists, Stockholm, June 11-14,1991. Nelson, J.P. (1977) Accessibility and the value of time in commuting, Southern Economic Journal, 43(3): 1321-1329.

2.82

Orne, M. T. (1962) "On the Social Psychology of the Psychological Experiment", American Psychologist 17:776-789. Parson,G.R. (1991) A note on choice of residential location in travel cost demand models, Land Economics, 67:360-364. Pearce,D.W. (1984) Cost Benefit Analysis, 2nd ed., Macmillan, London. Pearce,D.W. and Markandya. A. (1989) The Benefits of Environmental Policy, Organisation of Economic Cooperation and Development, Paris. Pearce,D.W. and Turner. R.K. (1990a) Economics of Natural Resourcesand the Environment. Harvester Wheatsheaf, Hemel Hempstead. Pearce,D.W. and Turner, R.K. (1990b) The Use of Benefits Estimates in Environmental Decision-Making. Report to the Environmental Directorate, OECD, Paris. Peterson,G.L, Brown, T. C., McCollum, D.W. Bell, P.A., Birjulin, A. A. and Clarke, A. (1996) Moral responsibility effects in valuation of WTA for public and private goods by the method of paired comparison, in Adamowicz, W.L., Boxall, P.C., Luckert, M. K., Phillips, W.E. and White, W. A. Forestry, Economics and the Environment, CAB International, Wallingford, Oxon. Prewitt, R.A. (1949) The Economics of Public Recreation - An Economic Survey of the Monetary Evaluation Washington National Park Service, DC. National Parks, in Recreation the of Price, C. (1975) Right use of land in National Parks, PhD thesis, Departmentof Forestry, University of Oxford. Price, C. (1978) Landscape Economics. MacMillan Press,London, 168pp. Price, C. (1979) Public preferenceand the managementof recreational congestion,Regional Studies, 13:125-139. Price, C. (1980) Public preferenceand the managementof recreationalcongestion: a modification, Regional Studies, 14:427-428. Price, C. (1981) Are extra recreation facilities unproductive?,Journal of Environmental Management, 12:1-5. Price, C. (1983) Evaluation of congestion and other social costs: implications for systemsof recreation parks, Sistemi Urbani, 1:119-139. Price, C. (1987) Does shadow pricing go on for ever?, mimeo, School of Agricultural and Forest Sciences, University College of North Wales, Bangor. price, C., Christensen,J.B. and Humphreys, S.K. (1986) Elasticities of demand for recreation site and recreation experience,Environment and Planning A, 18:1259-1263. Quiggin, J. (1991) "Total Valuation for Kakadu National Park", Departmentof Agricultural and Resource Economics, University of Maryland. Rae, D. A. (1982) "Benefits of Visual Air Quality in Cincinnati". Report to Electric Power ResearchInstitute, Charles River Associates,Boston. Rahmatian, M. (1987) "Component Value Analysis: Air Quality in the Grand Canyon National Park", Journal of Environmental Management24:217-223. Randall, A. (1987) "The Total Value Dilemma" in Peterson,G.L. and Sorg, C.F. (eds) Toward the Measurement of Total Economic Value, USDA For. Serv. Gen. Tech. Rep. Rm-148. Rocky Mountain Forest and Range Experimental Station, Fort Collins, CO, pp.3-13. Randall, A. and Stoll, J.R. (1980) "Consumer's surplus in commodity space".American Economic Review 70(3):449-455. Randall, A., Hoehn, J.P. and Brookshire, D.S. (1983) "Contingent Valuation Surveys for Evaluating Environmental Assets", Natural ResourcesJournal 23:635-648. Ravenscraft, DJ. and Dwyer, J.F. (1978) Reflecting site attractivenessin travel cost-basedmodels for Report 78-6, Dept. Forestry, University Research Illinois Forestry benefit of of estimation, recreation Urbana, Illinois. Urbana-Champaign, at Roberts, K. J. and Thompson, M. E. (1983) "An Empirical Application of the Contingent Valuation Method to Value Marine Resources"given in Cummings et al. (1986). Roberts, KJ., Thompson, M. E. and Pawlyk, P.W. (1985) Contingent valuation of recreational diving at Fisheries American Society, Mexico, Transactions 114: 155-165. Gulf the of of rigs, petroleum Rowe, R.D. and Chestnut, L. G. (1982) The Value of Visibility: Economic Theory and Applications for Air Pollution Control. AN Books, Cambridge, Mass. Rowe, R.D. and Chestnut, L. G. (1983) "Valuing Environmental Commodities:- Revisited". Land Economics 59:404-410.

2.83

Rowe, R., d'Arge, R. and Brookshire, D. (1980) "An Experiment on the Economic Value of Visibility". Journal of Environmental Economics and Management7: 1.19. Sagoff, M. (1988) "Some Problems With Environmental Economics", Environmental Ethics 10:55-74. Samples, K. C., Dixon, J.A. and Gower, M. M. (1985) "Information disclosure and endangeredspecies valuation". Paper presentedat the Annual meetings of the American Agricultural Economics Association, Iowa State University, Ames, Iowa, 4-7 August 1985. Samples, K., Dixon, J. and Gower, M. (1986) Information disclosure and endangeredspeciesvaluation, Land Economics, 62:306-312. Samuelson,P. (1954) "The Pure Theory of Public Expenditure", Review of Economics and Statistics 36:387-389. Schneider, F. and Pommerehne,W.W. (1981) "Free Riding and Collective Action: An Experiment in Public Microeconomics". Quarterly Journal of Economics 97.689-702. Schoemaker,P.J.H. (1982) "The expectedutility model: its variants, purposes,evidence and limitations". Journal of Economic Literature 20:529-563. Schulze, W. D., d'Arge, R.C. and Brookshire, D.S. (1981) "Valuing environmental commodities: some recent experiments", Land Economics 57: 151-169. Schulze, W. D., Brookshire, D.S., Walther, E.G., MacFarland, K. K., Thayer, M. A., Whitworth, R.L., BenDavid, S., Malm, W. and Molenar, J. (1983) "The Economic Benefits of Preserving Visability in the National Parklands of the South-West", Natural ResourcesJournal 23: 149-173. Schuman, H. and Johnson, M. P. (1976) "Attitudes and Behaviour", in Alex Inkeles, ed., Annual Review of Sociology 2: 161-207, Annual Reviews, Inc. Scip, K. and J. Strand. (1990) "Willingness to pay for environmental goods in Norway: A contingent valuation study with real payments". Paper presentedat the first annual conferenceof the European Association of Environmental and ResourceEconomists,April 17-20 1990, Venice, 29pp. Shaw, W. D. (1992) Searching for the opportunity cost of an individual's time, Land Economics, 68(l): 10715. Shogren, J.F., Shin, S.Y., Hayes, DJ. and Klicberstein, J.B. (1994) Resolving differences in willingness to 255-270. Economic Review, 84: American to accept, pay and willingness Sinden, J.A. and Worrell, A. C. (1979) Unpriced Values: Decisions without Market Prices. John Wilcy, New York. Slovic, P. (1972) "From Shakespeareto Simon: Speculations- and Some Evidence - About Man's Ability to Process Information". O.RJ. ResearchMonograph Volume 12. Eugene. Smith, A. (1790) "The Theory of Moral Sentiments". in Kristrdrn (1990a). Smith, RJ. and Kavanagh, NJ. (1969) The measurementof benefits of trout fishing: preliminary results of a Authority, Ouse Water Great Huntingdonshire,Journal of Leisure Research, Water, Grafbarn study at 1:316-32. Smith, V. K. (1988) Selection and recreation demand,American Journal of Agricultural Economics, 70:29-36. Smith, V. K. (1992a) Non Market Valuation of Environmental Resources:An Interpretative Essay, draft copy of unpublished manuscript. Smith, V. K. (1992) "Arbitrary Values, Good Causesand PrematureVerdicts", Journal of Environmental Economics and Management22(l): 71-89. Smith, V. K. and Desvousges,W. H. (1986) Measuring Water Quality Benefits. Kluwer and Nijhoff, Boston. Smith, V. K., Desvousges,W. H. and Fisher, A. (1986) "A comparison of direct and indirect methods for Journal Agricultural Economics 280-289. American 68: benefits", of estimating environmental Smith, V. K., Liu. J.L., Altaf, A., Jamal, H. and Whittington, D. (1991) "How reliable are contingent draft Department in developing Economics, for North countries". report, of policies surveys valuation Carolina State University. Snedecor, G.W. and Cochran, W.G. (1967) Statistical Methods, 6th ed., The Iowa State University Press, Ames, Iowa. Stankey, G. (1972) A strategy for the definition and managementof wilderness quality in Krutilla, J.V. (ed.), Applied Analysis, Resourcesfor the Future, John Theoretical in Studies Environments: and Natural Hopkins University Press, Baltimore. Strand, J. and A. Taraldset (1991) "The Valuation of Environmental Goods in Norway: A Contingent Memorandum 2/91, Testing". Department Bias Multiple Study no. Valuation of Economics, with University of Oslo.

2.84

Talheim, D.R. (1978) A general theory of supply and demand for outdoor recreation and recreation system, manuscript, Department of Agricultural Economics, Michigan State University. Thaler, R. (1985) "Mental Accounting and ConsumersChoice", Marketing Science4(3): 199-214. Tolley, G.S. and Randall, A., with G. Blomquist, R. Fabian, G. Fishelson, A. Frankel, J. Hoehn, R. Krumm and E. Mensah (1983) Establishing and Valuing the Effects of Improved Visibility in the Eastern United States, Interim Report to the U. S. Environmental Protection Agency, quoted in Garrod and Willis (1990). Tunstall, S., Green, C.H. and Lord, J. (1988) "The Evaluation of Environmental Goods by the Contingent Valuation Method", report by the Flood Hazard ResearchCentre, Middlesex Polytechnic. Turner, R.K. (cd.) (1988a) "SustainableEnvironmental Management:Principles and Practice, Belhaven, London. Turner, R.K. (1988b) "Wetland Conservation: Economics and Ethics", in Collard, D., Pearce,D.W. and Ulph, D. "Economics, Growth and SustainableDevelopment", Macmillan, London. Turner, R.K. (1992) Speculationson Weak and Strong Sustainability, GEC Working Paper 92-26, CSERGE, University of East Anglia, Norwich and University College London. Turner, R.K., Bateman, I.J. and Brooke J.S. (1992) Valuing the benefits of coastal defence: a case study of the Aldeburgh sea defence scheme,in Coker, A. and Richards, C. (eds) Valuing the Environment: Economic Approaches to Environmental Evaluation, Belhaven Press,pp.77-100. Turner, R.K., Bateman, I.J. and Pearce,D.W. (1992) United Kingdom. In Navrud, S. (ed.) Valuing the Environment: The European Experience, ScandinavianUniversity Press, Oslo. Turner, R.K., Pearce, D.W. and Bateman,I.J. (1994) Environmental Economics: An Elementary Introduction, Harvester Wheatsheaf, Hemel Hempsteadand the Johns Hopkins University Press,Baltimore. Tvcsky, A. and Kahneman, D. (1981) "The Framing of Decisions and the Psychology of Choice". Science 211:453-458. Tversky. A. and Kahneman, D. (1991) Loss aversion in riskless choice: a reference-dependentmodel, Quarterly Journal of Economics, 106:1039-1061. Varian, H.R. (1984) Microeconomic Analysis. Norton, New York. Vaughan. W. J. and Russell, C.S. (1982) Freshwater Recreational Fishing: The National Benefits of Water Pollution Control, Resourcesfor the Future, Washington D.C. Vaughan, W.J., Russell, C.S. and Hazilla, M. (1982) A note on the use of travel cost models with unequal Land Economics, 58(3): 400-407. comment, populations: zonal Vaux, H.J. and Williams, N. A. (1977) The costs of congestion and wilderness recreation, Environmental Management, 1:495-503. Walbert, M. S. (1984) "Valuing Policies which ReduceEnvironmental Risk: An Assessmentof the Contingent Valuation Method". PhD thesis, University of New Mexico, in Mitchell and Carson (1989). Walsh, R.G., Bjonback, R.D., Aiken, R.A. and Rosenthal, D.H. (1990) "Estimating the Public Benefits of protecting Forest Quality". Journal of Environmental Management 30: 175-189. Weisbrod, B.A. (1964) "Collective-consumption services of individual-consumption goods". Quarterly journal of Economics 78:471-477. Whittington, D., Briscoe, J., Mu, X. and Barron, W. (1990a) "Estimating the Willingness to Pay for Water Services in Developing Countries: A Case Study of the Use of Contingent Valuation Surveys in Southern Haiti". Economic Developmentand Cultural Change 38:293-311. Wilcoxon, F. (1945) Individual comparisonsby ranking methods.Biometrics, 1:80-83. Willig, R.D. (1973) "ConsumersSurplus: A Rigorous Cookbook", Technical Report No. 98, Institute for Mathematical Studies in the Social Sciences,Stanford University. Willig, R.D. (1976) "Consumer's surplus without apology". American Economic Review 66(4):587-597. Willis, K. G. and Benson, J.F. (1988) A comparison of user benefits and costs of nature conservation at three Studies, 22: 417428. Regional reserves, nature Willis, K. G. and Garrod, G.D. (1991a) "LandscapeValues: A Contingent Valuation Approach and case study National Countryside Park", Dales Change Working Paper 21, Countryside Change Yorkshire the of Unit, University of Newcastle upon Tyne. Willis, K. G. and Garrod, G. (1991b) An individual travel-cost method of evaluating forest recreation. Journal 42(l): 33-42. Economics, Agricultural of in The (1980) time E. A. of recreation benefit studies,Journal of Environmental Economics value Wilman, 272-286. 7: Management, and

2.85

Whittington, D., Mu, X. and Roche, R. (1990b) Calculating the value of time spent collecting water: some 269-280. Kenya, World 18(2): Development, for Ukunda, estimates Young, M. D. (1992) SustainableInvestmentand Resource Use, ParthenonPublishing, Carnforth, Lancs and UNESCO, Paris. Young, T. and Allen P.G. (1986) "Methods for Valuing Countryside Amenity: An Overview, Journal of Agricultural Economics 37(3)349-364. Ziemer, R., Musser, W.N. and Hill, R.C. (1980) Recreationaldemand equations: functional form and 62: Economics, 136-41. Agricultural American Journal of surplus, consumer

2.86

Chapter 3: Recreation: Previous Valuation Studies 3.1: LITERATURE

REVIEW

In conducting an extended CBA of ago-forestry conversion, a principle aim of this researchwas to incorporate monetary evaluationsof the recreationalvalue of woodland. Ibis both through analysis of existing estimates produced by others and to achieve aimed we through our own original research.This chapter presentsresults from the former review of investigation for literature the of our potential and using these results to provide existing benefit transfer measuresof recreation value for use in our CBA model. In the UK there have been more applications of the CV and TC to the evaluation of Indeed factor to than this any other open-access recreational good'. recreation woodland heavily influenced our decision to conduct an extendedCBA of land conversion to woodland Our detailed literature is such as to resources wetlands, etc. natural other review as opposed 1 Tlis in to this study. presents commentary and critique of over 40 appendix presented 100 Due in to this monetary evaluation estimates2. space over restrictions, papers containing briefest bringing this the summary of review concentrating only upon chapter we provide benefit the to transfer a consideration of possibilities allow of estimates together valuation analysis. Section 3.2 (and subsections)presentsvaluation estimatesfrom reviewed papersin a

form. both These to the evaluation are subdivided according tabular estimates condensed Brief type the of elicited, values e. g. annum; specific per per visit; and etc. methodemployed feasibility benefit the transfer of conducting analyseson eachcategory regarding comments 3.3 (and Section subsections)presentsour attemptsto conduct of results are also given. benefit transferanalysesuponthosecategoriesof evaluationhighlightedfor suchwork in the Ve haveexcludednon-UK studiesas we believethat the uncertaintiessurroundingrelevantculturaland USA (where between differences the themajorityof evaluationwork hasbeen countries such as socio-economic highly UK dubious Loomis (1996) the such extrapolations of make value. providesa review and conducted) forestry benefits CV the preservation of conducted using method. evaluations of non-UK 'In additiona considerable numberof relatedpapersarereviewed.The 100-plusevaluationestimates referred to abovedoesnot includea plethoraof sensitivityanalysisestimateswithin thesepaperswherethe impactof is altering assumptions assessed.

3.1

3.4 Finally section section. surnmarisesand concludes this chapter. preceding

3.2: SUMMARY RESULTS TABLES In this section we summarise,in tabular form, results from the papers reviewed in is into four follows: divided 1. This subsections as summary appendix i.

Annual national values for the entire Forestry Commission estate obtained from TC studies;

I

Capitalised per household sums obtained by asking for once-and-for-all in CV studies; payments

iii.

Annual per person values obtained from CV studies;

iv.

Per person per visit amounts obtained from both TC and CV studies.

To allow comparison between studies all results are given both as reported and indexed to 1990 Prices (the baseyear for our subsequentanalysisof agricultural values) using indices given in CSO (1981,1992). Where appropriatewe have included summary results from our own studies (detailed in chapter 4) to permit comparison with those from other UK authors. 3.2.1: ANNUAL NATIONAL VALUES

3.2.1.1: H. M. Treasury TC study One of the earliest attemptsto evaluate,in monetary terms, the open-accessrecreation interdepartmental 1972 CBA in Treasury's Forestry is the the of given value of woodland Commission (H. M. Treasury, 1972). This built upon an earlier unpublished study of visitor by Department Environment. Ile CBA the the of patterns conducted visit arrivals and both hybrid TC to produce national and conservancyestimates approach therefore adopted a in 3.1. is important The Details table these estimates are study of given of recreation value. both because it marked the official acceptance of monetary evaluations of forestry because funding dominated the estimates produced thinking official and and of externalities'

3AIthoughit is interestingto notethattheTreasuryhasseemedreluctantto acceptsuchvaluationswhenthey Willis, (Ken 1994). "in-house" pers. comm. conducted are not

3.2

the Forestry Commission until at least the late 1980's (NAO, 1986; PAC, 1987). However,

interest, (appendix 1) historical and policy our the review are of produced estimates while dubious assumptionsand methodological shortrather they obtained using that were shows for benefit transfer purposes. therefore suitable not considered cuts and are Table 3.1:

H. M. Treasury (1972) TC recreation values for the Forestry Commission (FC) estate and by conservancy. Unit

Value M

Treasury (1972)

All FC forests

1,113,200

1970

7,575,928

ibid

England NW

99,400

1970

676,471

ibid

England NE

UK (FC) aggregate Conservancy aggregate Conservancy aggregate

100,800

1970

685,999

ibid

England E

161,300

1970

1,097,734

ibid

England SE

Conservancy aggregate Conservancy aggregate

125,300

1970

852,734

ibid

England SW

Conservancy aggregate

154,100

1970

1,048,734

ibid

New Forest

Forest aggregate

129,600

1970

881,998

ibid

Dean

Forest aggregate

27,400

1970

186,472

ibid

Scotland N

Conservancy aggregate

51,100

1970

347,763

ibid

Scotland E

Conservancy aggregate

23,000

1970

156,527

ibid

Scotland S

Conservancy aggregate

15,800

1970

107,528

ibid

Scotland W

Conservancy aggregate

89,300

1970

607,735

ibid

Wales'N

61,200

1970

416,499

ibid

Wales S

Conservancy aggregate Conservancy aggregate

74,900

1970

509,735

L

3.3

Base Year

1990 Value (L)

Forest

Study

3.2.1.2: Other ZTC and ITC studies The National Audit Office economic analysis of Forestry Commission operations highlighted an apparentinability of benefits to match current grant-in-aid but recognisedthat in this appraisal (which were based upon the estimates of recreational value used H.M. Treasury, 1972) might be a substantialunderstatementof true values. Accordingly it was forestry justify its Commission that the to existence continued attempt as part of a wider supporteda number of evaluation studies,most notably the per visit and national value studies brings 3.2 Table together estimatesof the national recreational value of the Willis et al. of figures ZTC All for Willis Commission Forestry the except via are estimated estate. entire is (1972) for H. M. Treasury ITC. The (1991) Garrod the result repeated employ who and in is indices CSO (1981,1992) 1990 Indexing given to undertaken using prices comparison. discussed being base to the comparison allow agricultural values with chosen year with the in subsequentchapters. Table 3.2:

National recreation values for the entire Forestry Commission estate, from ) ITC (f from ZTC studies pa. and studies various Method

Value W

1990 Value (1)

Base Year

Study

Forest

Treasury (1972)

All FC forests

ZTC

1,113,200

1970

7,575,928

Grayson et al. (1975)

All FC forests

ZTC

1,050,000

1971

6,530,924

NAO (1986)

All FC forests

ZTC

10,000,000

1986

12,891,021

Willis and Garrod (1991)

All FC forests

ITC

8,665,000

1988

10,221,296

Benson and Willis (1992)

All FC forests

ZTC

52,999,000

1988

62,517,997

Benson and Willis (1992)'

All FC forests

ZTC

Note:

I 39,615,414'

1988

46,730,624'

Re-estimatebaseduponour recalculatedresult ([3*1, seeappendixI and section3.3.2 in this chapter)using EIA8 forest findings Benson Willis (1992). to the UK applied visit of of per and value mean an all

Table 3.2 highlights a considerable controversy regarding the national recreational

3.4

value of the Forestry Commission estate.Early studies (up to and including the NAO report) can be discounted becauseof the rather crude assumptionsused in their preparation.However there is a clear disparity between estimatesproduced by Benson and Willis (1992) using the ZTC and those of Willis and Garrod (1991) using the ITC. This is particularly disturbing given that both are using the samedataset,i. e. we cannot put this difference down to survey design discrepancies.In the final row of table 3.2 we report our reworking of the Bensonand Willis (1992) results based upon what we see as more defensible and data supported I for definitions (see details). This results the appendix of cost variable assumptionsregarding in a considerable reduction in the estimate of national recreational value. However, the discrepancy with respect to the ITC results remains. In our review of the Willis et al. ZTC 1) (see ITC we show that, while the specification of explanatory appendix paper's and latter indefensible in ITC is this theoretical the study, paper uses a and superior variables form. As in 4, in functional functional linear the show we changes chapter statistically weak form of the trip generationfunction (tgf) estimatedin TC studiescan have very major impacts have in Therefore, 2 surplus. while we argued chapter consumer of estimates upon resultant for a general superiority of ITC over ZTC approaches,in this case we have considerable doubts regarding the estimatesproducedby this particular ITC study. This of coursedoes not do do ZTC these the as a result of stand and problems we not of not mean that our criticisms feel that the estimates produced in table 3.2 can be disaggregated for benefit transfer

purposes. As a postscript to this work, in a recent appraisalH.M. Treasury rejected all the above in Forestry El figure the the calculating recreational value of per visit of estimates and used a is how 1994). It Willis, (Ken this figure was arrived comm., unclear pers. Commission estate However, be "guesstimate". 26 likely is it to with visits estimated at over a simply at and 1992), Willis, (Benson this and would provide an evaluation which sits annum per million discussed ZTC ITC between above. the analysis those and of quite nicely

3.2.2: PER HOUSEHOLD CAPITALISM VALUES The evaluation estimates detailed in table 3.3 were produced using CV household (rather than on-site) surveysemploying a per household,capital sum (once-and-for-all) WTP interviewed households have We these to they studies according whether classified measure.

in (the distance definition the to resource question or not of proximal which were proximal 3.5

being, admittedly, subjective), the implication being that proximal studies interview more users and potential users than do non-proximal surveys.

Table 3.3:

Capitalised (once-and-for-all) household recreation values from CV studies (Mousehold pa.)

Study

Forest

Distance to forest

Value W

Base Year

1990 Value (L)

Hanley and Munro (1991)

Birkharn Wood

proximal

12.89

1990

12.89

Hanley and Ecotec (1991)

Central Scotland Woodlands

proximal

9.34

1990

9.34

Hanley and Craig (1991)

Flow Country

non- proximal

3.27'

199

1

3.27'

Notes: Proximal householdsare thosewhich live nearby to the forest, whether existing (e.g. Birkham Wood) or Woodlands), Scotland households Central do (e. to as opposed non-proximal who not. g. planned I We have recalculatedthesefigures by including thosewho refused to pay as zero bids (see details in is, but This 1). the reduces mean considerably reported we argue,consistentwith a conservative appendix design (as per Arrow et al., 1993)and is thereforetheoreticallymore defensible.

Given the socioeconomicdifferencesbetweenhouseholdssurrounding Birkham Wood (near Knaresborough,Yorkshire) and the Central ScottishWoodlands Project and the fact that is both in former is the latter existence and encompasses new wood whilst a projected the is Similarly two the these the of results as expected. ordering woodland, certain ancient

is in for WTP in difference and non-proximal proximal woodlands reflected expected for Country for both Flow those the the above with study'. comparisonof results Comparing these results with the annual payments expressed by users in on-site lower 3.2.3) likely the (see that than the can see above we are capitalised section surveys is indicating by This that sums stated such again as expected, users,. annual equivalents of that WTP exceeds of even a proximal sample of of users entirely composed group of a the

4Thisdistancedecayeffect accordswith that foundby Batemanet a]. (1992)in their non-usersurveyof WTP to preservethe Norfolk Broads. is givenin thehigh implicit recreationaldiscountrate 'However,somesupportfor this apparentdiscrepancy in (1992) CV WTP formats. their Bateman by comparison of annual a]. and capitalised ct question reported

3.6

householdswhich will contain both users and non-users.However, the work of Bateman et find formats indicates (1992) difficult to comprehendand that respondents once-and-for-all al., often fail to answer such questions (a finding echoed in the studies listed in table 3.3). We from be the therefore accuracy of estimates cautious regarding obtained such studiesand must do not pursue their use for benefit transfer purposes. 3.2.3: PER PERSON ANNUAL VALUES Table 3.4 brings together results from CV surveys employing annual, per person, measures.All results were gatheredusing on-site surveys and therefore representthe values Respondents by were asked to state either their use value or their use + users. expressed We indicated. Details the regarding payment employed are also given. vehicle as option value have included summary results from two of our studies using per person per annum instruments. Of these,only the Thetford 2 result is strictly comparableas the Thetford 1 study informed respondentsof the existing tax payments for forestry, a factor which subsequent (details both had have found anchoring effect upon responses significant of to a analysis by (1992) Tranter 4). The Maxwell (1994), in and et studies al., chapter given studies are from following 1, in table the as neither specifies are omitted appendix earlier reviewed household. or per are per person whether the per annum results reported The ordering of results is as expected. The Whiteman and Sinclair (1994) studies (1992) Bishop + use option value and as expected,reports examines while values use examine by fact be Bishop difference This the that exacerbated higher whereas may well values. the for Sinclair WTP Whiteman for and examine proposed woods, existing values addresses forests. We would expect the lack of experienceand uncertainty regarding the likelihood of Ordering for lower WTP to lead as opposed existing woodlands. to proposed provision to income higher logical higher with areas recording appears also studies of within these sets throughout. to pay willingness

Comparisonwith our own studiesis interestingandgivessomecross-studyvalidation. Bishop's 2 the value alone, relationship Thetford examines use with use + option study As the Similarly, both 2 Whiteman Thetford is the although and expected. and as value results Sinclair studies examine use value alone, as our study is of an existing forest while their's disparity is the observed of values again in accordancewith is of a proposed woodland, expectations.

3.7

Table 3.4: Forest users annual recreation values from CV studies (E/personpa.) Study

Forest

Value type payt. vehc.

Value (1)

Base Year

1990 Value M

Whiteman and Sinclair (1994)

Mercia

use TAX

7.70

1992

7.12

ibid.

Thames Chase use TAX

9.79

1992

9.06

ibid

Gt. Northern Forest

use TAX

8.66

1992

8.01

Bishop (1992)

Derwent Walk

use+option SUB

18.53

1989

20.28

ibid.

Whippendell Wood

use+option SUB

27.03

1989

29.59

Bateman'

Thetford 2 (tINB)2

use TAX

12.32

1993

11.22

3

1990

5.141

ibid

Notes:

Thetford I

5.14

use TAX

1. Full details of all our studiesare given in chapter4.

2. The Thetford 2 study conducts a number of split sample experiments investigating mental account and in is 3.4 The to the tlNB table the only one refers subgroup reported which result effects. ordering Full in included in details table. this are given the chapter 4. studies other comparable with 3. In the relevant subgroup of the Thetford I study, respondents were informed of current average annual This (E2.70 found Forestry Commission have biased for to household pa). was tax woodlands payments per WTP responses downwards towards this amount. Full details arc given in chapter 4. Terms in capitals in the third column indicate the payment vehicle used. These are as follows: SUB = subscription to private ownership shares allowing free entry to the wood; TAX = payment via direct income tax.

Our Thetford I annual use value study differs significantly from all the others informed level in here their that were of present respondents of tax expenditure surnmarised in respect of Forestry Commission woodlands (;C2.70pa in the 1990 study year). This appears to have significantly, downwardly biased mean WTP relative to all other studies and we doubts have the regarding validity of this result. serious consequently All the above studies employed a taxation payment vehicle with the exception of Bishop (1992) who uses subscriptionsto private ownership sharesallowing free entry to the

3.8

wood. Evidence exists to support the use of tax rather than private subscription vehicles. Bateman and Turner (1993) argue that, where a tax vehicle is liable to be the actual route for funding, selection of an alternative vehicle may exacerbatehypothetical bias in responses. Furthermore, in empirical comparisonsof payment vehicle effects, Bateman et al., (1993) report that non-tax vehicles (particularly those proposing annual donations similar to those disproportionately from high refusal to pay rates. It is regrettablethat by Bishop) suffer used Bishop (1992) does not report such ratesý 3.2.3: PER PERSON PER VISIT VALUES The per person (and per party) per visit measurewas found to be highly useful in our due CBA. In this to the wider availability of studies using such a was part subsequent in I The contain relevant results estimatedvia ITC, ZTC appendix studies reviewed measure. and CV studies. 3.2.3.1: ITC studies The ITC studies reviewed give per person per visit estimates of the Marshallian The in 3.5 table of recreation. results value woodland reported all employ surplus consumer Ordinary least full (OLS) fitting) travel (best assumption. cost squares and maximum a likelihood (ML) estimation techniquesare used as indicated. As table 3.5 shows there have been very few per person per visit ITC studies of date. bulk in UK The to the of theseare contained within the conducted woodland recreation Willis and Garrod (1991) paper and, as detailed in section 3.2.1.2 above, we have certain form in functional (detailed the tgf these the of used studies criticism reservations concerning is left to appendix 1). This meansthat we are left with just the point estimatesgiven in our 4. detailed in While ITC chapter as we can use theseas stand alone estimatesthis studies own does rule out the possibility of cross-study ITC 'meta-analyses', such as that conducted by Smith and Kaoru (1990), for the foreseeablefuture.

6A further problem arising from this non-reporting is that we are not certain as to the treatment of refusals in the calculation of means. We argue that all non-protest refusals should be included as zeros in such in assumption all our CV studies. an such adopt calculations and

3.9

Table 3.5:

Forest users per person per visit recreation values from ITC studies (f/person/visit)

Forest

Study

Regression method

Value M

Base year

1990 value M

Willis and Garrod (1991)

Brecon

OLS

1.40

1988

1.65

ibid

Buchan

OLS

0.50

1988

0.59

ibid

Cheshire

OLS

0.40

1988

0.47

ibid

Lome

OLS

1.53

1988

1.80

ibid

New Forest

OLS

2.32

1988

2.74

ibid

Ruthin

OLS

1.29

1988

1.52

ibid

Brecon

ML

0.66

1988

0.78

ibid

Buchan

ML

0.20

1988

0.24

ibid

Cheshre

ML

0.06

1988

0.07

ibid

Lome

ML

0.96

1988

1.13

ibid

New Forest

ML

0.12

1988

0.14

ibid

Ruthin

ML

0.88

1988

1.03

Bateman'

Thetford

ML

1.32

1993

1.20

Thetford

13,4

OLS

1.07

1990

1.07

Thetford

13,5

OLS

1.19

1990

1.40

ibid ibi ibid Notes:

Thetford

13,6

OLS

1.34

1990

1

1.58

1. Full details of all our studiesare given in chapter4. 2. Thetford 2 result is for best fitting (variablevalue of time) model. 3. All Thctford I estimatesuse the bcst fitting double log model 4. Makes no valuation distinction betweenan adult and a child (under 16) visitor 5. Weights one child as 0.5 adult visitors 6. omits child visitors

3.2.3.1: ZTC studies ZTC studiesbut is dominatedby themultiTable 3.6 givesresultsfrom threeseparate Willis (1992). The figures Benson for and this particular study are of reported site analysis

3.10

from their 'Standard Model' [SM] where travel expenditure is calculated upon full costs of

33p per n-fileand travel time is valued at 43% of wage rate (see discussionof tgf cost definitions in chapter 2 and review in appendix 1).

Table 3.6:

Forest users per person per visit recreation values from ZTC studies (E/Person/visit) Value (f)

Forest

Study

Base Year

1990 Value (1)

Benson and Willis (1992)

New Forest

1.43

1988

1.69

ibid

Cheshire

1.91

1988

2.26

ibid

Loch Awe

3.31

1988

3.91

ibid

Brecon

2.60

1988

3.07

ibid

Buchan

2.26

1988

2.67

ibid

Durham

1.64

1988

1.94

ibid

N York Moors

1.93

1988

2.28

ibid

Aberfoyle

2.72

1988

3.21

ibid

South Lakes

1.34

1988

1.58

ibid

Newton Stewart

1.61

1988

1.90

ibid

Lome

1.44

1988

1.70

ibid

Castle Douglas

2.41

1988

2.85

ibid

Ruthin

2.52

1988

2.98

ibid

Forest of Dean

2.34

1988

2.76

ibid

Thetford

2.66

1988

3.14

ibid

Mean (all forests)

1.98

1988

2.34

Han ey (1989)

Aberfoyle

1.70

1987

2.14

0.41

1976 L

1.30

Everett (1979)

Dalby 1

1

1

In addition to the studies given in table 3.6, Christensen (1985) produces a ZTC E0.37 in 1980 Indexing 1990 to per visitor group of prices. surplus consumer of estimate

is Given is 0.70. f that this than this a per group rather per person estimate, gives a value of 3.11

significantly lower than any of the above studies. Christensengives a warning regarding the poor quality of data used in what was primarily a methodological investigation, and

analyses. consequentlythis result hasbeenomittedfrom our table and subsequent An initial inspection suggeststhat the studiescollected in table 3.6 might be sufficient I to form the basis of some sort of benefit transfer analysis.This is given further consideration subsequendy. 3.2.3.1: CV studies All the results summarisedin table 3.7 were derived from CV WTP studies of per A fee variety of elicitation an entrance payment vehicle. employing person recreation value formats indicated. 'use' direct 'use both + as and option' value methods were used as were Table 3.7:

Forest users per person per visit recreation values from CV studies (f/person/visit)

Study

Forest

Value type/ Elicit. method

Value (1)

Base Year

1990 Value (f)

Whiteman and Sinclair (1994)

Mercia

use OE

1.00

1992

0.93

ibid

Thames Chase use OE

0.71

1992

0.66

ibid

Gt. Northern Forest

use OE

0.81

1992

0.75

Hanley and Ruffell (1992)

Mean of 57 forests

use OE

0.93

1991

0.88

ibid

Mean of 57 forests

use OE

0.84'

1991

0.79'

Hanley and Ruffell (1991)

Aberfoyle

use OE

0.90

1991

0.85

ibid

Aberfoyle

use IB

1.21

1991

1.14

ibid

Aberfoyle

use PC

1.39

1991

1.31

ibid

Aberfoyle

use DC

1.49

1991

1.41

I

3.12

Study

Forest

Value type/ Elicit. method

Value W

Base Year

1990 Value (1)

Bishop (1992)

Derwent Walk

use OE

0.42

1989

0.46

ibid

Derwent Walk

use+option OE

0.97

1989

1.06

ibid

Whippendell Wood

use OE

0.54

1989

0.59

ibid

Whippendell Wood

use+option OE

1.34

1989

1.46

Willis and Benson (1989)

New Forest

use OE

0.43

1988

0.47

ibid

Cheshire

use OE

0.47

1988

0.51

ibid

Loch Awe

use OE

0.50

1988

0.55

ibid

Brecon

use OE

0.46

1988

0.50

ibid

Buchan

use OE

0.57

1988

0.62

ibid

Newton Stewart

use OE

0.73

1988

0.80

ibid

Lome

use OE

0.72

1988

0.79

ibid

Ruthin

use OE

0.44

1988

0.48

ibid

Mean (all forests)

use OE

0.53

1988

0.58

ibid

New Forest

use+option OE

0.88

1988

0.96

1988

0.79

1988

0.83

I

ibid

Cheshire

use+option OE

0.72

ibid

Loch Awe

usc+option OE

0.76

use+option OE

0.66

ibid

Brecon

3.13

1988

I 1

I 0.72

Study

Forest

Value type/ Elicit. method

Value (P

Base Year

1990 Value (f)

use+option OE

0.79

1988

0.86

use+option v lu

1.18

1988

1.29

1988

1.12

ibid

Buchan

ibid

Newton Stewart

ibid

Lome

use+option OE

1.02

ibid

Ruthin

use+option OE

0.63

1988

0.69

ibid

Mean (all forests)

use+option OE

0.82

1988

0.90

Hanley (1989)

Aberfoyle

use OE

1.24

1987

1.53

ibid

Aberfoyle

use PC

1.25

1987

1.55

Willis et al (1988)

Castle Douglas

use

0.37

1987

0.46

ibid

South Lakes

use OE

0.39

1987

0.48

ibid

North York Moors

use OE

0.53

1987

0.66

ibid

Durham

use OE

0.31

1987

0.38

ibid

Thetford

use

0.23

1987

0.28

ibid

Dean

use OE

0.28

1987

0.35

ibid

Castle Douglas

use+option OE

0.80

1987

0.99

ibid

South Lakes

use+option OE

0.86

1987

1.06

ibid

North York Moors

use+option OE

1.03

1987

1.27

ibid

Durham

use+option OE

0.56

1987

0.69

3.14

Study

Forest

Value type/ Elicit. method

Value M

Base Year

1990 Value (1)

ibid

Thetford

use+option OE

0.41

1987

0.51

ibid

Dean

use+opdon OE

0.63

1987

0.78

Bateman'

Thetford 2 (f2NB)3

use OE

0.52

1993

0.47

ibid

Thetford 1

use PCL

1.21

1990

1.21

ibid

Thetford I

use PCH

1.55

1990

1.55

Notes: Elicitation methods are as follows: OE open ended iterative bidding IB PC payment card PCL = payment card (low range) PCH = payment card (high range) DC = dichotomous choice

Valuation categoriesinvestigatedare as follows: use = use value option = option value (the extra WrP to ensureconservationof the site for future use). Notes:

1. Derived from a per householdvalue of L2.00 (1990prices)convertedto a per personper visit value using 2.53 in CSO (1990). UK household size of persons given an average 2. Full details of all our studiesare given in chapter4. 3. ne Thetford 2 study conductsa numberof split sampleexperimentsinvestigatingmental accountand in 3.7 f2NB is The to the table refers the only one reported subgroup result which effects. ordering in Full included details in table. this the studies other are given chapter4. with comparable

With the exception of our Thetford I experiment, all the studieslisted in table 3.7 vary follows: factors three as to major according

i.

Questionnairedesign(for which we can useauthorshipas a proxy);

ii.

Whether the study addressed'use' or 'use + option' value;

iii.

Elicitation method.

In the following section we consider the potential of conducting benefit transfer analyses upon our reviewed studies.

3.15

3.3:

BENEFIT TRANSFER ANALYSIS OF REVIEWED VALUATION STUDIES

To what extent can the results obtainedin thosestudiesreviewed (and indeed our own issue benefit This has in be transfer to of areas? recent years other woodland studies) applied developed into a major area of research'. The advantagesof a rigorous approachto benefits individual financial both The temporal, and of conducting costs, transfer are clear. Consequently involved in US the are prohibitive. a planning policy site each at evaluations Environmental Protection Agency and, more recently, several UK government departments interest in However, have this of research'. avenue as the considerable shown and agencies in formulating US the raised problems and commentatorsacknowledge, numerous eminent (Desvousges 1992; benefits transfer are extreme et al., numerous and conducting a successful Smith, 1992; McConnell, 1992; Atkinson et al., 1992). Despite the interest in benefit transfer work, few major applied studiesexist. Of these (1990) Kaoru Walsh (1992). Both Smith those and et al., claim and of are the most notable is designed This technique a statistical have approach9. adopted a meta-analysis to if However, for the we examine rigorous across studies. results synthesising specifically date, (Wolf, 1986) that, to the can see raw material we meta-analysis of requirements beenunsuitablefor such a technique. has benefit individual by generally evaluations provided Meta-analysis was originally designed for cross-assessmentof medical trials conducted to data deviates from further The formats. the that such specifications raw precisely replicated highlight Glass (1981) inferences et al., a variety of the of cross-analysis. are suspect more being base important that the ones studies often employ most meta-analyses, with problems from designed definitions that studies are not excluded poorly and of variables varying

consideration". The rigorous demandsof full meta-analysishave yet to be met in any benefits transfer

large, derived, lack because is the consistently and a sufficiently This of of primarily study. 7LOOMiS (1992)actuallytracesresearchinto benefitstransferbackto 1962. However,it is only in the late (1994). Willis Carrod by focus See became major of research. and a review 1990's that this

IThc author has on separate occasions been approached by the Department of the Environment, Department Authority for bcnerits Rivers National to the transfer work. potential conducting with regard Transport and of For reasons explained in the text, such offers have not been pursued. Tor a lucid introduction to meta-analysis see Wolf (1986). 100ther major problems are that insignificant results may not be published leading to a bias towards from be individual, independent treated the that results same study may as multiple significant results, and observations.

3.16

Our individual UK dataset studies. review of existing studiesalso of evaluation comparable design in insufficiently terms these reported of questionnaire that are often of many revealed The US demands data true the meta-analysis". approachto this to of a statistics satisfy and has been 1992) Walsh 1990; Kaoru, to perform a simplified partial (Smith et al., and problem in benefits focusing the each source study and relating measure reported mean upon analysis detailing: binary) (usually the evaluation variables explanatory this, to a series of simple the the unit; elicitation method type measurement the studied; of resource employed; method derives directly from 'benefit Our such an transfer' articles study of reviewed used; etc. do Given recreation studies, we obviously that considering woodland are only we approach. 12 Remaining detailing define type the explanatory evaluated of good variables not need to . binary definition (as be intended treated of relevant variables. above) to via factors were feasible for below, discussed those for only proved such an approach However, reasons be had ZTC ITC Reviewed CV. to the assessed upon and studies using the adopting studies basis. a more ad-hoc 3.3.1: ITC Studies insufficient UK ITC been have 3.2.3.1 to perform a in indicated there As section 1), (appendix in detailed discussed Furthermore, we review our as benefit transfer analysis. derived from reviewed estimates the point validity of have considerablereservationsregarding for descriptive 3.8 For statistics all studies date. table reports completeness ITC studies to excluding our own. in ITC indicates studies 3.8 reviewed surplus mean consumer that, on average, Table 3.9 Table ML OLS techniques. provides to for estimation higher opposed as those using was but for trend this confirms general two studies which these of groups variance of an analysis have We is difference mean results significant. reproduced not statistically this that shows former While ITC the this table. ML the OLS of end studies at from our own estimated and Given latter does CI 95% the not. our considerable lies within the of reviewed studies

"As part of our review we attemptedto gather full descriptive statisticson all results. However, the standard issue full incomplete The to this task. to complete allow us of consistent and and too variable of reporting was in UK. is be if benefit to the transfer be ever successfully undertaken work tackled reporting must 121na separatestudy we present a simple analysis of valuations across differing recreational experiences both logically to the the substitutability of environmental concerned and resource related were that results noting (Bateman, Garrod, in Willis 1994). considered and provision the change the magnitude of

3.17

reservationsconcerning results from reviewed UK ITC studies (see appendix 1), we do not feel that extrapolation of such findings to other forests is justified and are confident in preferring our own ITC results here. Table 3.8: Descriptive statistics for reviewed ITC studies (mean CS in f/person/visit) OLS ML Studies Studies n mean median tr. mean st. dev s.e. mean minimum maximum Q1 Q3

Table 3.9:

6 1.462 1.585 1.462 0.840 0.343 0.470 2.740 0.560 2.035 1

6 0.565 0.510 0.565 0.472 0.193 0.070 1.130 0.123 1.055 1

Oti ML

95 % Cl for mean(f)

vel

1 dev mean st n 6 1.4617 0.8403 6 0.5650 0.4718

Nnjýtervals I OLS studies ML studies all studies 1

n 6 6 12

mean 1.462 0.565 1.013

----------------------------------

---------------------------------0.00 0.60 1.20

Pooledst dcv = 0.6814

Our studies:

12 1.013 0.905 0.935 0.801 0.231 0.070 2.740 0.297 1.6H-JI

Analysis of variance: reviewed ITC studies (mean CS f/person/visit) Analysis of variance

F

All Studies

st dev 0.840 0.472 0.801

se mean 0.343 0.193 0.231

95 % cl 0.580 0.070 0.504

Thetford 1 (OLS study) meanconsumersurplus= LI. 07/person/visit Thetford 2 (ML study) mean consumersurplus= LI. 20/person/visit

3.18

1.80

2.344 1.060 1.522

3.3.2: ZTC Studies The ZTC studies of Benson and Willis (1992) detailed in table 3.6 provide an

interestingset of internally consistentstudies.While the numberof studiesis considerably less than would nonnally be used for benefit transfer work our review (appendix 1) shows that thesestudieswere conductedto a high standardand are not subject to the methodological and theoretical problems besetting reviewed ITC studies. However, the same review does in Benson Willis' definition the the of cost and that variable used Preferred 'Standard show Model' [SM] is questionable.In an early paper the authorsreport findings from a sensitivity definitions detailed in 3.10. the table those of cost variable across analysis Table 3.10: Cost specifications.for standardand alternative models

Assumptions

Model Travel Cost

Time Cost

(p/mile)

(% wage rate)

Standard model [SMI

33

43

Alternative [11

33

25

Alternative [2]

33

0

Alternative [31

Visitors estimate

43

Alternative [41

8

0

Alternative [51

8

43

Source:Willis and Benson(1989) Justificationfor the variouscostmultipliersis consideredin chapter2. Clearlywe will [SM] being [SMI [11 [21 [51 > > > and estimates the of consumer surplus order obviously get Table 3.11 details [41 cstimatesof consumersurplus are uncertain. relationships while other > forests in the each of calculated under varying cost table assumptions across given per visit 3.10. Here we can see a consistent ordering of results for all forests being [SM] > [11 > [2] it is inspection On had [4]. little impact [51 that time [31 clear varying costs > relatively > > had impacts. The travel while estimates varying surplus costs significant upon consumer

3.19

authors point out that, a priori, time may be expectedto have a more significant effect but that this may be diminished by its treatment as a subsetof total costs (Cij) rather than as a separatevariable within the tgf.

Table 3.11:

Consumer surplus per visit by forest under varying assumptions (f/visit; 1988 prices) [SM]2 Travel=33ppm Timc=43%

Forest district

V] Travcl=33ppm Time=25%

(21 Travel=33ppm Timc=O%

[31 Travel=estimate Timc-43%

[41 Travel=gppm Time--O%

151 Travct=8ppfn Time=43%

New Forest

1.43

1.40

1.36

0.93

0.33

0.40

Cheshire

1.91

1.97

1.81

1.25

0.44

0.54

Loch Awe

3.31

321

3.05

1.92

0.73

1.00

Brecon

2.60

2.56

2.50

1.70

0.61

0.71

Buchan

2.26

2.22

2.16

1.67

0.52

0.63

Durharn

1.64

1.77

1.73

0.54

N. York Moon

1.93

1.87

1.84

0.59

Abcrfoyle

2.72

2.59

2.37

0.61

South Iakes

1.34

1.30

1.27

0.41

Newton Stewart

1.61

1.57

1.53

1.24

0.36

0.45

Lome

1.44

IAO

1.35

1.10

0.33

0.42

CastleDouglas

2.41

2.36

2.54

Ruthin

2.52

2A7

2.40

Forest or Dean

2.34

2.19

2.13

0.69

Thetford

2.66

2.62

2.55

0.76

0.95

0.72 1.72

0.59

0.70

Weightedmean= fl. 98' Notes:

Willis, L2.00 & ) for (Benson 1992; their cs/visit of page to mean v. The I. a weighted refer often authors figure (and for is V. 983. We However [SM]. the that actual used aggregation) calculate model standard in benefits in C430,000. figure higher an overstatement TCSUlt of aggregate excess of the would ; that use of

Choice of the-optimal model is problematic but clearly important. Unlike the ITC the be ZTC fit tgf the may affected by the choice and number of zones which will degree of of (see 2) fit The the chapter statistic. study authors advocatethe upwardly-biased already affect from in [SMI the estimates standard model preferenceto alternative surplus use of consumer

(Willis Benson, 1989; Benson Willis, 1992): following grounds and and the modelson 3.20

i.

They argue that respondentsperceptions and statementsregarding travel costs are based upon full rather than marginal (petrol only) cost per mile;

I

They argue that respondentsdo not adopt differential costs toward recreationas opposed to non-recreationtravel;

iii.

Whilst respondents value the whole trip experience, the forest visit was valued journey, (of less than the that time than the car so more a positive cost proportionately full wage rate) can be justified, with the time cost used in the standardmodel being that most recently advocatedby a governmentdepartment(in this casethe Department of Transport);

iv.

"Given that entry fees at many National Trust, English Heritage and similar properties (which include gardens, parks, woodlands and forests) are closer to our higher for forest figure (L2.00), this seems realistic and car-borne plausible visits of estimate the kind studied in this project". Thesejustifications are open to somecriticism. Arguments i and ii, which are similar,

however be true, while respondentsperceptions of travel cost well exceed pure may well 8p/mile, travel the of site average expressed costs range reported of costs petrol marginal (from 17.7p/mile to 27.1p/mile with a mean for all sites of 22.8p/mile) does not support the 'standard 3.12 in Table details 33p/mile the model'. gives used the assumption adoption of for forest districts the eight costs car running where this of of respondents estimates information was elicited. Given the uncertainty surroundingthe value of time, and in particular leisure time (see is iii to 2) although one not would want make such reasonable above argument chapter However, iv is the argument open result. criticism. strong any the of mainstay uncertainty invalid. Informal forest is is recreation of the essence Firstly the comparison of goods its be its indeed to nature may endogenous good value; comparisonswith public unpriced and Secondly, if between therefore cheese. such a comparison chalk and goods priced goods are TC disappear. Surely three the of undertaking a necessity year study would then were valid, this is not an argument which the authors would push too stronglyl .

3.21

Table 3.12: Respondentsestimatesof car running costs Forest district

Estimated car

confidence

travel cost

interval (±)

(p/mile)

(p/mile)

Brecon

20.85

9.31

47

Buchan

23.77

11.02

135

Cheshire

21.77

7.51

128

Loch Awe

17.74

11.85

38

Lome

27.08

19.36

119

New Forest

21.02

14.45

266

Newton Stewart

25.21

15.75

150

Ruthin

21.92

10.91

61

All sites

22.79

13.91

944

Sample size

Source: Willis and Benson (1989) While we can accept the authors choice of functional form, it would seem that their [SM] is less 'standard defensible. Indeed, the the as accurate results most model' of choice following the authors own reasoning (see argument i above), it would seem that the most logical model is that using time costs valued at 43% of wage rates (as this is a government figure) (average travel calculated costs as and visitors perceived costs of recognised.

22.8p/mile)i.e. model[3] in table3.11. Oneproblemwith this approachis thatdatafor such However, be derived by certain sites. collected at an only approximation can was an analysis first calculatingthe ratio of consumersurplusresultsfor models[3] and[SM] for thesesites. The weightedaverageof this ratio can then be usedto extrapolatefor the remainingother 3.13 ([31/[SMI) Table 0.690 this calculates weighted ratio as and usesthis to sites. seven estimate consumer surplus per visit at the sevensites where perceived costs were not elicited.

Theseresultstogetherwith thoseof model [3] from consumersurplusestimatesfor all sites 43% travel of perceived costs and the wageratetime costsandarerecorded assumption under in table 3.13 as model [3*]. An all sites weightedaverageconsumersurpluswas then 3.22

calculated as being fl. 48 per visit. We would argue that this representsa more defensible result than the weighted average of fl. 98 obtained from the [SM] model and preferred by Willis and Benson 13 .

Table 3.13:

Calculation of the whole sampleweighted mean consumersurplus per visit for model [3] (producing model [3*])

Forest district

Sample size

% of total sample

New Forest Cheshire Loch Awe Brecon Buchan Durham North York Moors Abcrfoyle South Lakes Newton Stewart Lome Castle Douglas Ruthin Forest of Dean Thetford

316 324 56 241 201 481 387 1148 322 213 201 66 310 276 254

6.59 6.76 1.17 5.03 4.19 10.03 8.07 23.94 6.71 4.44 4.19 1.38 6.46 5.75 5.30

Total

4796

100.00

Model 13] total

1862

Weighted mean Notes:

I. 2. 3.

% Of model [3) sample 16.97 17.40 3.01 12.94 10.79

11.44 10.79 16.65

CS/Visit ism] (E)

Cstvish [31 W

Ratio 131/ ISM]

CS/visit 13*1 (f)

1.43 1.91 3.31 2.60 2.26 1.64 1.93 2.72 1.34 1.61 1.44 2.41 2.52 2.34 2.66

0.93 1.25 1.92 1.70 1.67

0.650 0.654 0.580 0.654 0.739

1.24 1.10

0.770 0.764

1.72

0.683

0.93 1.25 1.92 1.70 1.67 1.133 1.333 1.881 0.922 1.24 1.10 1.663 1.72 1.612 1.842

100.00 1.98

1

0.690'

1.483

Calculated by multiplying the ratio 13]/[SM) by the decimal % of model [31 sample column and then finding the (weighted) mean. Calculated by multiplying the site [SMI CS/Visit for sites where perceivedcosts were elicited by the weighted average of the ratio [31ASMI to 7 decimal places (0.6901678) and then rounded. Calculated by weighting the site [3*1 CS/visit by its decimal % of total sample (across all sites). Calculated to 7 decimal places (1.481469) and then rounded.

Clearly the resultsgiven in table 3.13 are far from an ideal meta-analysisof benefit benefit function linked Furthermore, than transfer to site a producing rather transfer values.

features,etc. we haveonly produceda singlemeanestimateof recreationalvalue.Ibis is, we freely admit, far from ideal. However,as our reviewsof variouspapers"'have shown,the

"Interestingly, following this adjustment,the authors agreed(pers. comm.) that their original estimateshad bound assumptions. upper used somewhat 24Veryfew papers have attempted to addressthe issue of attribute rather than site valuation, even though benefit if be Even is transfers to and sensitive successful are those studies which undertaken. vital such work have attempted such disaggrcgadonshave not been successfulin so doing (seeour review of Hanley and Ruffell, 1991,1992).

3.23

its in UK is in formative feel This being evaluation studies still stages. so state-of-the-arts we that the estimation of mean values does provide useful - estimates of the magnitudes of be As for benefit transfer can, appropriate caution, used with work. which recreation value E1.48 defensible for provides a our study mean of starting point such, our reviewed-ZTC wider research. 3.3.3: CV Studies Our cross analysis of UK CV studies of woodland recreation studies best conforms Walsh Smith Kaoru (1990) (1992) benefit transfer and and et al., although of approach to the (as with these) it still falls short of an ideal meta-analysis. Our review shows that these CV into be three types: categorised studies can per householdcapitalised valuations; per person per annum valuations; per person per visit valuations. In section 3.2.2 we raised serious questionsconcerning the validity of those type (i) been insufficient (ii) have 3.2.3 type that there Furthermore studies showed section studies. been focused has Consequently those upon our attention justify analyses. cross-study to any 3.7 details Table 9 3.2.3.1) (section person per visit value. (iii) per yielding studies type is less by Smith This 48 than those used considerably estimates. evaluation studies yielding Walsh (120 400 35 to (77 or et estimates) al., roughly yield Kaoru used studiesof which and CV This 129 287 the the underlines method). used of which estimates yielding studies used UK in US in the and and reinforces our opinion that difference available, comparablestudies benefit transfer in this country is currently premature. The analysis we conducted here was definitive. illustrative be than intended to rather therefore

One early considerationconcernedthe extentto which our Thetford I studieswere in detailed 3.7. We table that the the were concerned of use others with not, or compatible, into it impossible incorporate to these results our two types of payment card may make benefit transfer study. An analysis of variance suggestedthat these fears were well founded

from 4) in these two details our wider analysis. and excluded results were (see chapter

3.24

Our remaining database of evaluation estimates yielded the following simple explanatory variables: WTP

Study mean willingness to pay (;E/person/visit)

OPTION

I if the study asked WTP for use + option value; 0 if the study asked WTP for use value alone

ELICITAT

Elicitation method (categorical variable): I= open ended; 2= iterative bidding; 3= payment card; 4= dichotomous choice

OE

I if open-endedelicitation method; 0 if other elicitation method

AUTHOR

Authorship (categoricalvariable): I= Hanley and Ruffell; 2= Bishop; 3= Willis and Benson (1989); 4= Hanley (1989); 5= Willis et al., (1988); 6= Bateman; 7= Whiteman and Sinclair.

Following Glass et al., (1981) an early concern" was to ensure the comparability of from had been 3.7 due A table studies to design, excluded of reviewed number studies. implementation or gross reporting problems (see appendix 1). To some extent further identified by AUTHOR been identified have the analysis of variable may which problems individual study designs. Although a generalised linear model analysis did reveal some differences, these were highly correlated with the OPTION and OE variables and the AUTHOR variable had to be omitted from further analysis.Analysis of unusualdesign effects identifying detailed (as below). by outliers therefore conducted was Clearly the variables ELICITAT and OE cannot be included within the samemodel. Analyses of variance showed that the numbers in categories 2,3 and 4 of the ELICITAT I for individual (1,2 to treatment. too and respectively) allow small meaningful variable were However, when these categories were amalgamated to form the OE variable, a highly from between level) (5% difference these and the open-endedstudies was results significant latter in 3.14. Details this table analysis are given of observed.

"Our first concernin compilingour data-base was to excludethosestudieswhich we felt wereof a poor designstandard(seeappendixI for details).

3.25

Table 3.14: Analysis of variance in WTP as a result of the OE variable Analysis of variance

OE level

n

mean

041.3535 42

1

0.7571

95 % Cl for mean WTV

st dev 1.2945

-----------------------------------(* (.

0.0819

_*....

)

)

1

Pooled st dev = 0.2862

0.90

1.20

1.50

Source

df

SS

ms

F

p

OE

1

1.2945

1.2945

15.81

0.000

error

44

3.6029

0.0819

total

45

4.8974

Notes:

1. Basedon pooled st dcv. F and p values are calculatedas per the default method used by the Minitab statisticspackage(seenote to Table 4.21 subsequently).

Following these preliminary analyses we concluded that the most conservative it just investigate VV`TP OPTION OE to the to a simple model of relating and was approach from initial 3.15 details findings from Table results our such a model and variables. statistically unusual observations. Studying the unusual observations from our initial regression model, we have no large influence those a observations with as these all relate to the non-OE problem with in form (Bishop, 13 1992; However, the two there clear outlier of observations are studies. OE use + option value for Whippendell Wood) and 32 (Hanley, 1989; OE use value for Aberfoyle).

Further analysis confirmed these to be highly unusual' results indicating

in design from by Of differences that methodology adopted the two, other authors. significant the Hanley (1989) OE use value result is the most unusual. In our review of this study (see in details) for 1 that the this study was likely to lead to a argue we approach used appendix WTP. This be by to of conclusion appears overstatement the above supported significant deleted from WTP Both these results were our of model of studies which was then analysis.

3.26

16-

re-estimated. Results for our best fitting model are given in table 3.16 .

Table 3.15:

Initial multiple regressionmodelof reviewedwoodlandCV studies (E/person/visitvalues)

Dependent variable = study mean WTP (f:/person/visit) Predictor

Coef

Stdev

Constant

1.3525

0.1241

10.90

0.000

OPTION

0.30720

0.07801

3.94

0.000

OE

1 1

-0.7197 R' = 45.9%

0.2482

t-ratio

0.1336

p

1 1

0.000

-5.39

R2(adj) = 43.4%

n= 45

Unusual observations Obs.

Option

WI?

Fit

Stdev.Fit

7 8

0.00 0.00

1.1400 1.3100

1.3525 1.3525

0.1241 0.1241

9 13 32 33

0.00 1.00 0.00 0.00

IA100 IA600 1.5300 1.5500

1.3525 0.9400 0.6328 1.3525

0.1241 0.0602 0.0496 0.1241

Residual

St.Resid

-0.2125 -0.0425

x -0.99 X -0.20

0.0575 0.5200 0.8972 0.1975

R denotes an obs. with a large st. resid. X denotes an obs. whose X value gives it large influence. for Whippendell Wood ' OE (1992) Bishop + Notes: use option value Hanley (1989) OE use value for Aberfoyle

"Rudimentaryfunctionalform analysisconfirmedthe linear form of the bestfit model.

3.27

0.27 X 2.16R' 3.69R2 0.92 X

Table 3.16: Final multiple regressionmodel of reviewed woodland CV studies (;E/person/visitvalues)

Dependentvariable= studymeanWT? (f/person/visit) Predictor

Coef

Stdev

Constant

1.3525

0.09634

14.04

0.000

OPTION

0.31208

0.06219

5.02

0.000

OE

-0.7571

1 1

t-mtio

0.1041

p

1

0.000

-7.2ý8ý

R'= 61.1%R2(adj) = 59.2%n= 43 Analysis of variance on WTP SOURCE

DF

SS

NIS

F

p

Regression

2

2.3886

1.1943

32.17

0.000

Error

41

1.5222

0.0371

Total

43

3.9108

1 1

ISOURCE

DF

SEQ SS

OPTION

1

0.4234

OE

1

1.9652

j

A number of interesting observations arise from table 3.16. The overall fit of the data) 60% dealing (given is that socioeconomic about are with with of total we model good is finding The the constant, a explanatory strongest variable which variation explained.

both from CV bid functions (e. Bateman 1992) our own g. et al., and other many accordswith is interesting believe indicate The the constant strength of and may, we a othersresearch. disturbing determinant of stated WTP. It may well be that a major factor affecting WTP individuals level. 'social is This a socially of appropriate perceptions payment responses i. be linked, to properly, of existence quite perceptions value e. the value of the norm' may from However, it be linked the separate valuers personal also good use. may assetas a public (and in influences fee less as the such particularly case of an entrance to valid payment

3.28

fees (open-air) the experience comparable of at attractions, e.g. car respondents vehicle) parking fees". Figure 3.1 illustrates how such a regressionconstant may be formed.

Figure 3.1: Fonnation of the WTP regressioncoefficient Experience of Fees

Social Norm II

Value of VVT? RegressionConstant The extent to which these 'social norm' and 'fee experience' factors influence WTP investigation. However, the is specific potential require and would uncertain responses influence of such factors upon the validity of CV estimates marks this out as an area deserving of further research. All the observations were from on-site users of forest resources. These users can is feel We types that, the of question valuation not where express a variety of values". dominated by WTP likely it is their that statements are use respondents most raised, explicitly have broaden discussed, However, to attempted authors several (i. as value e. usersusevalue). been done by has This include definition asking respondents, to option values. this value in be WTP how to WTP they addition initial secure to would much state response, their after OPTION by As dummied Such the above. variable can questions are future use of the site. higher WTP in sums. significantly be seen,responsesto such questionsresult We have not asked such questionsin our studies as we are sceptical of their validity. is likely WTP just to occur. have and some anchoring their value use Respondents stated Equally importantly we feel that such questionnaire structures put the respondent under a Such WTP 'improve' questions may, their previous response. to on obligation psychological inadequate. is bid It WTP feel also their that previous was respondents make we argue, for invoke concerns continued access regarding unclear whether or not such questions will fairly feel In bequest that a such questions elicit etc. all we values; others to the resource; higher response. meaningless

171nterestingly,since completion of the Thetford 2 study (detailed in chapter 4), the Forestry Commission has introduced a car park fee of 50p at Lynford Stag (open air walks area) and fl, at High Lodge (open air walks plus visitor centre), amounts which are well within our estimatesof per person recreational value. "Subsumed within the catchall concept of Total Economic Value (see chapter 1).

3.29

The regressionequation also shows a significant relation betweenWTP responseand OE that techniques are significantly lower such responses elicited using elicitation method than those obtained by other methods. This echoes our findings from earlier research (Bateman et al., 1993). We have shown that, in the absence of highly developed better OE WTP do than techniques, may provide a estimate" of other methods questionnaire 'use Given the + option value' measures,we this and our concernsregarding approaches". have less reservations" about predictions of use value elicited using OE techniquesthan we do about the other values implied by our best fitting model (table 3.16). The fitted mean in 3.17. Accordingly from table this our preferred are summarised model values estimated benefit transfer estimate from our databaseof reviewed CV studies of per person per visit for is that valuations

OE (alone) techniques, namely estimated value via use

LO.60/person/visit.

Table 3.17:

Predicted users WTP response for a variety of CV questionnaire types (L/Person/visit) Elicitation Method

Value type

se Value Use + Option Value

OE

Other

0.60 0.91

1.35 1.66

3.4: CONCLUSIONS Our review of UK monetary evaluations of woodland recreation suggeststhat such

body is far from In its infancy, the of still mature. particular out of arguably while research, benefit does for transfer, high meta-analysis not advanced papers necessary quality consistent, is is it (although this true the date arguable even strictly of advanced whether more to exist

"Here we define a better estimateas one which minimises elicitation bias. An error from formulated WTP 2). (see chapter may still remain 2'Seediscussion of relevant non-forestry researchat start of chapter 4 and Bateman et al., forthcoming. 21NOtethe phraseology; we do not claim that the OE/use value estimate is correct, just that it is likely to be less biased than other measures.

3.30

US literature). Consequentlywe have had to conduct fairly simple cross-studyanalyses.Even these have only been feasible upon per person per visit ZTC and CV studies, the former yielding a best estimate of f: 1.48 and the latter giving a value of f:0.60. These provide magnitude estimatespermitting a simple sensitivity analysis. However, these estimates are insensitive locational factors. being We to site and attempt to addressthese crude, admittedly factors in our own work which examines forests with common recreational features and issue locational (if the through the spatial prediction of number of arrivals models explicitly This individual in values). work also provides new valuations of woodland not variation in in to those this chapter. cases, reviewed we prefer certain which, recreation

3.31

REFERENCES Arrow, KJ., Solow,R., Portney,P.R., Learner,EX., Radner,R. and Schuman,E.H. (1993)Reportof the NOAA Panelon ContingentValuation,FederalRegister,58(10):4602-14. Atkinson,S.E., Crocker,T.D. and Shogrcn,LF. (1992) Bayesianexchangeability, benefittransfer,and Research,28(3):715-722. researchefficiency,WaterResources Bateman,IJ. andTurner,R.K. (1993)Valuationof the environment,methodsand techniques:the contingent valuationmethod,in TurnerR.K. (ed.) SustainableEnvironmentalEconomicsand Management: Principlesand Practice,BelhavenPress,London,pp.120-191. Bateman,I. I., Langford,I.H. and Rasbash, J. (forthcoming)Willingness-to-pay questionformateffectsin contingentvaluationstudies,in Bateman,IJ. and Willis, K.G. (eds.) ValuingEnvironmental Preferences:Theoryand Practiceof the ContingentValuationMethodin the US,EC and Developing Countries,Oxford UniversityPress. Bateman,IJ., Langford,I.H., Willis, K.G., Turner,R.K. and Garrod.G.D. (1993)The impactsof changing iterative willingnessto pay questionformatin contingentvaluationstudies:an analysisof open-ended, bidding and dichotomouschoiceformats,GlobalEnvironmentalChangeWorkingPaper93-05,Centre for Socialand EconomicResearchon the GlobalEnvironment,Universityof EastAnglia and UniversityCollegeLondon,pp.50. Bateman,11, Willis, K.G., Garrod,G.D., Doktor,P., Langford,1.and Turner,R.K. (1992)Recreationand environmentalpreservationvalueof the Norfolk Broads:a contingentvaluationstudy,Reportto the National RiversAuthority,pp.403. Bateman,Ii., Willis, K.G. and Garrod,G.D. (1994)Consistencybetweencontingentvaluationestimates:a National UK Parks,RegionalStudies,28(5):457-474. two studies of of comparison Benson,LF. and Willis, K.G. (1992)Valuing informalrecreationon the ForestryCommissionestate,Bulletin 104, ForestryCommission,Edinburgh. Bishop,K.D. (1992) Assessingthe benefitsof communityforests:an evaluationof the Recreationaluse benefitsof two urbanfringe woodlands.Journalof EnvironmentalPlanningand Management, 35(l): 63-76. Christensen,J.B. (1985)An economicapproachto assessing the valueof recreationwith specialreferenceto forestareas,PhD thesis,Departmentof ForestryandWood Science,UniversityCollegeof North Wales,Bangor. W.H., Naughton,M.C. andParsons,G.R. (1992) Benefittransfer conceptualproblemsin Desvousges, Research,28(3):675-683. estimatingwaterquality benefitsusingexistingstudies,WaterResources Everett,R.D. (1979)"The monetaryvalueof the recreationalbenefitsof wildlife". Journal of Environmental Management,8:203-213. Garrod,G.D. and Willis, K.G. (1994)The transferabilityof environmentalbenefits:a review of recent WorkingPaper12,Centrefor Rural Economy,Universityof researchin waterresourcesmanagement, NewcastleuponTyne. Glass,G., McGaw,B. and Smith,M.L. (1981) Metal-Analysisin SocialResearch,Sage,BeverleyHills, California. Grayson,AJ., Sidaway,R.M. andThompson,FY. (1975)Someaspectsof recreationplanningin the ForestryCommissionin Searle,G.A.C. (ed.). RecreationalEconomicsand Analysis:paperspresentat the Symposiumon RecreationalEconomicsand Analysis,LondonGraduateSchoolof BusinessStudies, January 1972,Longman,Essex. Journal Hanley,N.D. (1989) Valuing rural recreationbenefits:an empiricalcomparisonof two approaches, 40(3): 361-374. Economics, Agricultural of Hanley.N.D. and Craig, S. (1991) Wildernessdevelopmentdecisionsand the Krutilla-Fishermodel:the Country. Flow EcologicalEconomics,4: 145-164. Scotlands caseof Hanley,N.D. and EcotecLtd (1991) The ValuationofEnvironmentalEffects:StageTwo Final Report. The Scottishoffice IndustryDepartmentand ScottishEnterprise,Edinburgh. Hanley,N.D. and Munro,A. (1991) Designbias in contingentvaluationstudies:the impactof information, DiscussionPaperin Economics91113,Departmentof Economics,Universityof Stirling. Hanley,N.D. and Ruffell, RJ. (1991) Recreationalusevaluesof woodlandfeatures.Reportto the Forestry Commission,Universityof Stirling. WorkingPaper849, Institute Hanley,N.D. and Ruffell, RJ. (1992) The valuationof forestcharacteristics, for EconomicResearch,Queen'sUniversity,Kingston,Canada.

3.32

H.M. Treasury (1972) Forestry in Great Britain: An Interdepartmental CostlBenefit Study, HMSO, London. Loomis, J.B. (1992) The evolution of a more rigorous approach to benefit transfer: benefit function transfer, Water ResourcesResearch,28(3):701-705. Loomis, J.B. (1996) Measuring general public preservationvalues for forest resources:evidence from contingent valuation surveys, in Adamowicz, W. L., Boxall, P.C., Luckert, M. K., Phillips, W.E. and White, W. A. Forestry, Economics and the Environment, CAB International, Wallingford, Oxon. Maxwell, S. (1992) Valuation of rural environmental improvements:a case study of the Marston Vale Community Forest Project using contingent valuation methodology, MSc Thesis, Departmentof Land Use, Silsoe College, Cranfield Institute of Technology. McConnell, K. E. (1992) Model building with judgement: implications for benefit transfers with travel cost models, Water ResourcesResearch,28(3):695-700. NAO (National Audit Office), (1986) Review of Forestry CommissionObjectives and Achievements. Report by the Comptroller and Auditor General, National Audit Office, HMSO, London. (PAC) Public Accounts Committee, House of Commons (1987), Forestry Commission: review of objectives and achievements. Twelfth Reportfrom the Committeeof Public Accounts, HMSO, London. Smith, V. K. (1992) On separatingdefensible benefit transfers from "smoke and mirrors", Water Resources Research, 28(3):685-694. Smith, V. K. and Kaoru, Y. (1990) Signals or noise? Explaining the variation in recreation benefit estimates, American Journal of Agricultural Economics, 72(2):419-433. Tranter, R.B., Bennett, R.M. and Board, N.F. (1994) Valuing woodland walks, Countryside Recreation Network News, 2(2):4-5. Walsh, R.G., Johnson, D. M. and KcKean, J.R. (1992) Benefits transfer of outdoor recreation demand Research, 28(3): Water Resources 707-713. 1968-1988, studies, Whiteman, A. and Sinclair, J. (1994) The Costs and Benefits of Planting Three Community Forests: Forest of Mercia, Thames Chase and Great North Forest, Policy Studies Division, Forestry Commission, Edinburgh. Willis, K. G. and Benson, J.F. (1989) Values of user benefits of forest recreation: some further site surveys. Report to the Forestry Commission,Department of Town and County Planning. University of Newcastle upon Tyne. Willis, K. G. and Garrod, G.D. (1991). An individual travel cost method of evaluating forest recreation. Journal of Agricultural Economics, 42(l): 33-41. Willis K. G., Benson, J.F and Whitby, M. C. (1988) Values of user benefits of forest recreation and wildlife. Report to the Forestry Commission,Departmentof Town and County Planning, University of Newcastle upon Tyne. Wolf, F.M. (1986) Meta-analysis: quantitative methods for researchsynthesis,Quantitative Applications in Sage, Hills, California. No. 07-059, Beverley Sciences Social the

3.33

Chapter 4: Recreation: New Valuation Studies 4.1

INTRODUCTION This chapter presentsour own relevant recreation evaluation work using the CV and

TC methods. '11iisis subdivided into studiesof woodland, and studies which looked at nonforest resourcesbut derive results which are of direct relevance to this research. The chronology of our work is important as we feel that the robustnessof study design In defence improved this the research. over course of of our early considerably and analysis in UK However, to practice valuation contemporary work. closely these adhered studies from US for introduces techniques studies allowing more advanced more subsequentwork in later have Indeed, work we attemptedto exceed the specification sophisticated analysis. US studies. such of This chapter does not, as might be expected, end by drawing together general

both 5 brings Instead together these chapter studies and these studies. across conclusions findings. feel However, do to of an overall assessment we provide those reviewed previously here different discussed theoretical and empirical problems. addresses the studies that each of We therefore provide brief commentary and conclusionswithin our discussionof each of the in this chapter. presented studies

4.2

RELEVANT

NON-WOODLAND

RESEARCH

In collaboration with others we have concluded a variety of non-forestry recreation Of direct during these, the this of most group of studies the research'. course of evaluations CV the of environmental recreational and analyses our this was to research relevance 1992,1993,1995a/b (Bateman Broads Norfolk et al., and the preservation value of forthcoming; Batemanand Bryan, 1994; Batemanand Langford, forthcoming a; Langford and from 1994,1996). Findings Langford, Bateman Langford, this work 1993; and Bateman, in design CV two that they influenced answered of subsequent studies woodland our strongly

10ther CV studiesinclude studiesof recyclingin Norfolk and NorthernIreland;watersiderecreationat Rutland Water was recently(for the NRA)-,eutrophicationin the Baltic Sea0ointly with CSERGE,Harvard the Beijer Institute(Stockholm)andWarsawUniversity)andlaboratory Instituteof InternationalDevelopment, disparities WTP/WTA (with Sugdcn, Professor Robert UEA). Other includes TC and work testingof part-whole Broads. Norfolk the a study of

4.1

important questions which we had previously identified as problem areas for empirical analysis: first, what payment vehicle should be used and; second,what elicitation method to adopO. 4.2.1 PAYMENT VEHICLE EFFECTS As indicated in chapter 2, one of the questions we wished to addressin our research in Prior how to the to the main of payment changes method used. react may respondents was Norfolk Broads survey, a smaller sampleof 433 respondentswas collected through face-toface interviews with usersof Broadland. Following Boyle and Bishop (1988), it was decided initial in this survey should be undertakenusing expectations, that the absenceof any a-priori be follows: three tested OE that should payment vehicles as and method elicitation an 1. 2.

An unspecified charitable donation (DONATE). Payments to a hypothetical charitable fund specifically set up to facilitate flood defence work in Broadland (FUND). Payments via direct taxation (TAX).

Other alternatives were considered to lack credibility for this particular study. Specifically entrance fees were, given the nature of the resource (large area; considerable be UK thought to credible. not precedent), resident population; no Table 4.1 details WTP and relatedresults from this study acrosseachpayment vehicle. Analysis of table 4.1 reveals the reasonsfor rejecting both the DONATE and FUND from bids (46.5%) disproportionately WTP DONATE The zero suffered vehicle vehicles.

definition felt It this that the of vehicle the was vague other vehicles. to of either compared led respondentsto be uncertainthat their donationswould be effectivelyused. In shortsuch in hypothetical did therefore the market and rejected. was credibility engender not a vehicle

2Further design decisions can also be attributed to this work. For example, responserate problems in our led Broads decision face-to-face interviewing to the to a subsequent techniques of adopt of non-users mail survey in our woodland studies.

4.2

Table 4.1:

Norfolk Broads study: payment vehicle analysis

Payment vehicle

N

WTP---O WTP--O WTP>500 WTP>500 mean M (number) (number) M WTP(f)

DONATE

157

73

46.5

0

0

FUND

65'

15

23.1

2

TAX

211

25

11.8 1

10

Payment vchicle

St. dev.

coeff. of variation

DONATE

39.81

FUND TAX

1

median WTP (L)

trimmed mean(L)

25.60

10.00

20.34

3.1

47.60

10.00

22.00

4.7

89.22 1

40.00

65.06

S.C. mean

min. value (c)

max. value M

lower quartile (f)

upper quartile (f)

156

3.18

0

250

0

50

140.70

296

17.40

0

1000

1

50

144.95

162

9.98

0

1000

10

100

1 Excludes one outlier (seetext).

Mean bids for the FUND vehicle were heavily upwardly biased by the presenceof a E10,000 4.1 bid Table FUND that the ornitted. of which was shows outlier single extreme in high bid (23.1%) badly terms of rate and also performs zero vehicle still perfort'ns in bid. DONATE TAX Ile terms the than or vehicles of either markedly worse

FUND

vehicle was therefore also rejected. The TAX vehicle produced by far the lowest zero-bid rate (11.8%) almost half that The TAX better DONATE FUND the vehicle vehicle. also performed of quarter and one of in terms of bid variability than the FUND vehicle and about as well as the DONATE vehicle. As no vehicle produced excessive evidence of strategic bidding (large numbers of TAX deemed bids) high thus the the this and a problem vehicle not was was unreasonably if fact flood favoured by defence be It that, the to also works were was preferred choice. built, such works would in reality be paid for out of taxes rather than trust-fund donations.

immediate had TAX the and applicability,an The of realism advantages vehicle therefore forestry to our that applied studies. also advantage 3Aftcr questioning by Ole interviewer this particular bid was judged not to be strategic behaviour because landowner income from Broads the a major within an annual with commercial actually was the respondent E300,000, income an exceeded which would be put at risk by flooding in Broadland. As Broadland recreation different a categorically expressing type of value to the rest of the sample and was was such the respondent in from 4.1. the table analysis reported therefore omittcd

4.3

All respondentswere asked why they had respondedin the way they had. Many of thosepresentedwith the FUND (and especiallyDONATE) vehicles commentedthat they were not confident that paymentsvia such vehicles would be fully channelledtowards preservation work (trust funds were not to be trusted!). Furthermoremany of thoserespondingto the TAX vehicle commentedthat, while they disliked paying extra taxes,they had confidencethat such money would be spent efficiently upon any flood defence scheme. One potential criticism of the TAX vehicle is that, while most respondentsseea TAX be taxpayers, to all others may unsure about this. Furthermore other sum as applying latter be I'his be to non-taxpayers. criticism can some extent currently may respondents WTP increase by to tax that the question refers an absolute rather than out countered pointing it Overall felt increase that the statistical advantages payments. existing was on a proportional Consequently disadvantages. it TAX the outweighed any vehicle was of and realism for based WT? future tax per annum should a on woodland use that research concluded payment vehicle'. 4.2.2

ELICITATION

EFFECTS

in chapter 2 we presented a variety of, in some cases conflicting, economic and in the the changes elicitation method may effect which regarding arguments psychological have upon WTP responses. Differing context and respondentcognition and motivation may lead to stated WT? being either in excessof, or below, formulated value. We therefore saw Fortunately, investigation objective of our research. as a major of such elicitation effects the from NRA (although for funding the the that obtained with proviso was such work significant large Broads) Norfolk focus had the allowing a series of sampletestsacross to upon the study CV in is Europe date largest The to the experiment and study undertaken elicitation methods. is comparable with major US studies. It is also one of only a handful of studies worldwide Arrow, Panel" drawn by Kenneth "Blue Ribbon NOAA US guidelines up to the to conform Robert Solow and others regarding the conduct of CV studies (Arrow et al., 1993).

'Another issue may be that as the TAX vehicle involved money being taken away before it is received then individuals may find this less 'painful' than the FUND/DONATE vehicles which involve visable losses. sClearly such a vehicle is inappropriate for our per visit analyses;see subsequentdiscussion.

4.4

4.2.2.1 The elicitation methods The Norfolk Broads study investigates three WT? elicitation methods: open-ended

(OE); dichotomouschoice (DQ; and iterative bidding (IB). The OE approachrequiresa separatesub-samplebut the DC and IB studies are linked in a way so as to allow a further 'multi-level modeling' (MLM) dichotomous choice analysis to be performed as follows:

1.

Prior to all (OE/DC/IB) WTP quesdons,r-espondents were asked whether or not they in for (the 'payment the to some amount good question pay principle'). were willing This legitimised a negativeresponsewhich may have beeninhibited by going straight to the WTP question. Answers to this question are usually analysedseparatelyfrom involves (as latter WTP the to a subsampleof those answeringthe questions responses MLM be However, techniques this to allow question). analysedas payment principle WT? full level first the of responses. set of the

2.

Respondents who answer the payment principle question positively are then presented

initial bid level which they are asked whether or not they with a randomly selected forms binary The DC be to this the to question answer response. pay. willing would 3.

All such respondentsare then given a supplementarydichotomousquestiondetermined by their initial DC response. If, for example,a respondentagreedto pay an initial bid in it doubled be EX, the second round question while would this was amount of halved if the initial responsewas negative. This forms the double-bounddichotomous (2DC). response'

4.

This process is then iterated again to produce a 3DC response.

5.

Finally respondentswere asked to state, in an open-endedmanner, their maximum WTP. This formed the IB response,so called becauseit is the final answer from the iterative process.

Figure 4.1 illustratesthe possiblebidding pathwaysarising from just one initial DC bid level (here 000).

Turther details of the operation of each method are given in chapter 2. 7Details regarding determination of the number and absolute value of DC bid levels are given in Bateman be It that (1993). a variety of other analyseswere conducted as part of this research emphasised should et al. (1992). in Bateman al. et detailed as sFollowing the nomenclatureof Hanemannet al. (1991).

4.5

Figure 4.1:

DC, IB and ML responsesarising from a single initial bid level DC response /0%^NPN I st Bound

IB response 2nd Bound

3rd Bound

Yes

Max WTP?

No

Max WTP?

Ile vv It, E400? Yes WTP E200? Yes WTP anything at all?

Yes INITIAL BID --0' C100 e.g.

No

Max WTP?

Yes

Max WTP?

WTP C100? No

No

i Ii I

WTP E50? No

Yes

Ma x WTP?

No

Max WTP?

WTP E25? I\

MULTI-LEVEL (HIERARCHICAL) ANALYSIS

response

2nd response

3rd response

4th response]

Source:Batemanand Bryan (1994) 4.2.2.2 Analysis of data

from the OE and IB formatsis simple as both yield The analysisof WTP responses is for Although both trivial. which calculation of means are truncated continuous variables indicated distributions least-squares bid that regression analysis of of analysis at zero, functions was not inappropriatefor thesefortnats. Analysis of the binary DC (initial bid level

bid Here data logit (or curve straightforward. estimation relatively also via was response) WTP being derived from is bid the mean with the required area under curve analysis similar) (cumulativeprobability distribution (CPD) function). Further details of OE, IB and DC 4.6

analyses are given in Bateman et al. (1993,1995a) and Langford and Bateman (1993). Analysis of the multi-level data is somewhatmore complex. Hanemannet al. (1991) in their double-boundeddichotomous choice experiment, employ a multinominal logistic design to estimate likelihoods for the probabilities of each of the possible set of responses(i.e. an extension of our DC analysis). However, such an approach becomes almost infeasibly complex when set against the diversity of possible responsesin a triple-bounded design. Therefore we applied the hierarchical,multilevel statistical techniquesdevelopedby Goldstein (1987) in the context of educationalresearch. A considerableadvantageof a MLM approach is that the prior question concerning the principle of paying anything at all can be analysed function bid the rather than separatelyas with all other elicitation formats. Full as part of details of this analysis are given in Langford, Bateman and Langford (1994,1996). 4.2.2.3 Results and discussion i.

The OE Experiment

In total 862 interviews were completed using the OE elicitation method. Of these 'no' All 131 to the answered principle pay question. payment such respondents some had The to then they state why asked given such an answer. most common were respondents income (almost 40% for to related and existing commitments was of non-payment reasons followed by free-rider OE 6% the the total to sample) pure of reply non-payers, equivalent (e. the someone else g. the government) should pay (almost was valued, area that, although 25% of non-payers, equivalent to 4% of the total OE sample). Whilst income constraints does free-riding downward bias in here, OE to the a the effect point possible pose no problem importantly free-riders indicate More WTP. this of extreme small group the may estimate of larger group of respondentswho, whilst still stating some non-zero sum, a of existence below free-rider WT? WTP incentive. true their the as a stated result of reduced nevertheless However, attempts to quantify such a strategy would have required a significant extension to laboratory-type (and possibly controls) and were consequently not the questionnaire undertaken". Evidence of 'protest bidding', in the senseof a refusal to participate in the valuation

'It is interesting to note that recent reviews have indicated that free riding behaviour may result in a (very WTP between 6G-95% to WTP depcnding approximately) true of stated of upon the strength of reduction 6 See free incentive chapters and 7 of Mitchell and Carson (1989) and Milon (1989). to ride. the

4.7

process,was conspicuously absent. The possibility of such a responsewas directly catered for by listing a refusal to value the Broads as an explicit option amongstreasonsfor refusing to pay. However, only 30 respondents(1% of the total OE + DC/IB sample) gave this as their reasonfor refusal. 'Mis finding strongly contradictsthe assertionby somecommentators that CV studies are pervasively invalidated by the prevalenceof protest bids (Sagoff, 1988). Alongside evidence of free-riding, other respondentsin the OE sample appearedto exhibit strategic overbidding. Table 4.2 details univariate OE WTP statistics for a variety of Mean for WTP the entire OE sample of 846 respondentswas truncation tail points. upper f67. l9(95%CI=E59.53; f:74.86). However, omission of just the single highest bid (0.11% OE WTP fall E65.79 (a OE to to mean caused the reduction of over 2%). sample) of Similarly truncating the top 1% of bids causesa reduction in the mean of nearly 10%. In themselves such statistical effects are not conclusive proof of strategic overbidding as a skewed distribution may simply reflect the socioeconomic and preferencecharacteristicsof it found inspection by However, that the those at the upper was sums stated upon the sample. infeasible distribution bid given the ability of these respondentsto pay. appeared tail of the Several of the highest bidders stated WTP sums which exceeded their entire annual (in by factor 5). and environmental goods some cases all recreational a of upon expenditure We therefore conclude that there is strong evidence for a degree of strategic overstatement by a small number of respondentsin the OE experiment. Validity testing was applied to all three elicitation methods following the criteria set in Content Carson (1989). & Mitchell the main, carried out prior to the by was, validity out in fields the recognised authorities of of meetings with a number of consisted survey and These surveys and psychology. consultancies addressedall social economics, marketing, design the on of the questionnaire, associated emphasis the particular study with aspectsof information and survey sampling strategy. Criterion validity testing (comparison with actual WTP for the good) was not feasible and therefore a major effort was made to establish (i. One testing to e. whether results conformed expectations). simple construct validity WTP (convergent that mean with of other studies validity), was only comparing approach, feasible for the OF, study as other formats have had few applications 'in the UK to. date. Results from the OE experiment were contrasted with those from 28 comparable UK use-

4.8

be logically This factors: to to two analysis showed results related according value studies'O.

i)

the numberof adequatesubstitutesitesavailable;

ii)

the magnitude of the proposedchange in provision.

Table 4.2:

Truncation effects - open ended WTP study' 0

1

8

42

84

126

168

211

% of upper tail truncated

0%

0.1%

1%

5%

10%

15%

20%

25%

N

846'

845

838

804

762

720

678

635

Mean WIV

67.19

65.79

60.89

46.76

37.38

32.57

28.39

25.54

Median WTP

30.00

30.00

30.00

25.00

25.00

20.00

20.00

12.00

113.58

106.10

90.08

55.19

38.64

33.69

30.10

24.41

3.91

3.65

3.11

1.95

IAO

1.26

1.16

0.97

1250.00 1000.00

500.00

250.00

150.00

100.00

100.00

100.00

5.00

5.00

5.00

5.00

2.13

2.00

1.00

100.00 100.00 100.00

60.00

50.00

50.00

50.00

50.00

No. of upper tail truncated

I

SL Dev. S.E. Mean Maximum Bid'

5.00

Lower Quartile I

Upper Quartile

Notes:

1

I

.

All rows, exceptthe upper three,are measuredin L's

Total sample of 862 interviews included 16 incompleted questionnaires (omitted from calculation of mean) 3 Includes, as zeros, those who refused to pay anything at all " Minimum bid = zero throughout.

Most other UK studies have looked at sites with some or many substitutes,facing

fact Accordingly in the this that study estimates a provision. changes marginal relatively logically higher than correct. WTP seems others most value mean The theoretical validity of OE responses was examined via estimation of the bid form investigated. Functional full A was function. range of explanatory variables was initial but linear forms (although theoretically an undesirable), were uncertain a-priori, be (a degree high to indicated unlikely achieved was explanation that a of overall analysis detailed Ilerefore (e. Box-Cox) form functional OE g. analysis of studies). characteristic of

forms. by double The best favour in model standard was provided a by-passed of using was 'OFUllresultsare given in Bateman,Willis and Garrod(1994).

4.9

log form which is reported in equation (4.1). Ibis model narrowly outperformed a semi-log (dependent)form which contained the sameexplanatory variables. LWTP(OE) = 0.1934 + 0.2920 LINC + 0.2695 RELAX + 0.2473 ENV (3.32) (4.15) (3.93) (0.22)

(4.1)

where: LWT? (OE) = Natural log of open ended WTP response income log (continuous Natural LINC of respondents variable) = if (---0 RELAX to often area relaxlenjoy scenery respondent visits otherwise) =I is if (--0 ENV a respondent member of an environmental group otherwise) =I R25.29% 80011 Total d. f. Figures in brackets are t-statistics.

The exPlanatory variables given in equation (4.1) are all significant at the 99% level, 95% level. feature The further the at even significant major of this were variables while no 12 degree its is 'best model' of explanatory power , which although more very poor overall

is OE Therefore, in trait this of many studies. a characteristic case, while than usual extreme indicates logical that economic theory responses observed across studies mean of ordering the is adequate to explain results at such a level, the poor performance of the model given in (4.1) suggeststhat further consideration of the motivations underlying individual responses is required here (see below). The DC experiment As with the OE survey, those interviewed using the DC/IB questionnairewere asked, In WTP taxes. to they total any willing pay extra or not were whether questions, the to prior 240 of the 2070 DC/IB respondentsanswered'no' to this question (11.6%). Tests showed

there to be only one significant predictorof a positive responseto this question,namely All to refused pay any extra who respondents environmental groUp13. an of membership involved before, As the common most reasons to specify a reasonwhy. taxeswere asked incomeconstraintsandexistingcommitments(33%of non-payers;3.9%of the totýl sample) "Equation (4.1) omits all responsesfor which information on any explanatory variable was missing. 11o ensure that no errors had beenmade, statistical analysiswas carried out independentlyat the University University Newcastle-upon-Tyne. Both Anglia the of analyses confirm the weak explanatory East at and of 'best' model. the of power "Significant at cc= 1%. No other significant factors at a= 5%.

4.10

closely followed by the pure free-riding response(31.7% of non-payers; 3.7% of the total failed Analysis to reveal any significant factors determining the reason for sample)". non-payment. Those respondentswho indicated that they were preparedto pay at least some amount bid levels, to the then one pay of selected at random. The mean DC WTP is were asked calculated by integrating the CPD function between appropriate truncation limits. Accurate is function because incorrectly bid fitted function will therefore the an vital, of estimation give a spurious estimateof the mean. Both linear and log models were testedusing both logit Log functions. better fit link than linear specifications. models gave a markedly and probit The choice between link functions was more difficult as both logit and probit approaches log-logistic However, better fit a model gave a marginally well'5. and similarly performed it for further has been elsewhere was preferred extensively analysis. In all cases this used as high feature the the of estimated models was very explanatory power of the most remarkable Equation WTP log-logistic (4.2) determining in level bid the response. presents model the LBID; logarithm bid level (E) from the the variable natural of the explanatory single resulting presented to respondents. LOGIT

0.9939 LBID + -4.932 (18.39) (-19.74)

(42)

Deviance change -594.4 1325.7 Residual deviance 1624 d.f. Figures in brackets are t-statistics

LOGIT 7; = where:

r

In I LJ

7; = probability of an individual saying 'no' to the bid level"'

"Note that as this question was asked prior to any WTP question, this responsedoes not refute the earlier inhibit free-riding relative to OE approaches. formats DC may that suggestion "Full details in Bateman et al. (1993). "Readers should be aware that this meansthat 'positive' relationships (e.g. between WTP and income, etc) will have a negative sign and vice versa.

4.11

As can be seenfrom equation (4.2), a log logistic model with the single explanatory

variable LBID fits the dichotomouschoice datasetextremelywell". Furtherexplanatory variableswere then addedto this model in an attempt to improve the fit". The best log-logistic model is given as equation (4.3). LOGIT

1.026 LBID 0.0907 LINC + -3.736 (-1.34) (-6.23) (18.40) 0.3756 RELAX 0.3126 ENV BOAT -0.5888 (-2.58) (-2.22) (-3.35)

(43)

Deviance change -622.9 1297.2 Residual deviance 1620 d.f. Figures in brackets are t-statistics where Natural logarithm of respondentshouseholdincome (continuous variable) LINC 1 if respondentdoes participate in some boating activity (=O otherwise) BOAT I if respondentvisits area to relaxlenjoy scenery often (=O otherwise) RELAX I if respondentis a member of an environmental group (=O otherwise) ENV defined as previously other variables Although not significant at (x = 5%, the variable LINC is included in equation (4.3) income finding the that, expectations, complying with respondents although the to underline in determining dichotomous insignificant) (statistically to response part plays a very weak 'price' lead While the to WTP theory expect us economic would variable questions. choice dominance its degree be over other variables, particularly of (LBID) to the most significant, income, is of interest. We comment on these findings subsequently. In calculating mean WTP from a DC experiment an important issue is the choice of in in integration issue is detail Bateman CPD. This the to analysed of prior truncation option Langford in Bateman (1993). For these (1993,1995a) and reasons given and paperswe et al. "As an ancillarytestof this result,individualmodelswerefitted for the datawithin eachbid level (214:5 inevitable for All is This level). the for of eight the models 227 produced exceptionally were weak. 5 each n: registeredrefusalsi.e. very little variation. Howevereventhebest lower bid levelswherevery few respondents level) in bid ESO deviance being deviance (for the only recorded a change of with residual these models -24.65 of 223.73. Theseresultsconfirm the key role of the bid level in determiningresponses. "Alternative modelsare consideredin Batemanet al. (1993).

4.12

follow Hanemann (1989) in preferring a non-negative but positive untruncated approach. Applying this approach to our simple bid function (4.2) gives a mean WTP estimate of E143'9 while for our 'best fit' model the mean is very similar at E140 underlining the relative unimportance of the non-BID variables. iii.

Comparing OE and DC results

Do our findings from the OE and DC experimentsconfonn to economic theory or is there evidence of the psychological biases discussed in chapter 2? The most common has been through comparison of means and our study effects elicitation assessmentof OE from finding DC (Sellar that the mean exceeding experiment et of the general confirms little 1990; 1993). Kristr6m, However, 1989; Walsh this 1985; above result gives et al., al., indication regarding the validity of either approach. Indeed, as Kristr8m (1993) points out, imply distribution. this if the need not similarity of same even means were Figure 4.2 presentsboth DC and OE responsedistributions in the form of survival functions for those WTP at least some amount (i.e. excluding, for both formats, all those DC Here the of respondents to all). proportion anything at pay respondents who refused OE is bid level the of respondents with proportion at each compared giving positive responses bid level. In the than that WTP absenceof any elicitation or greater equivalent sums stating bid However, the across we can see vector. roughly coincide effects theseproportions should is distribution format DC shifted outwards which generates a response apparently that the OE the that approach'. to of compared Figure 4.2 suggests that it is more likely that a respondent will agree to pay a bid level, DC X than that amount as a rather with via an presented when amount particular OE experiment. This discrepancy can be viewed from either an economic or psychological format provides no incentive for OE Economic that the theory suggests perspective. free-riding be 1987) Randall, (Hoehn to or cost and may subject expected and overstatement incentive WTP. Furthennore, to the given both an understate will give which of effects

Calculation 1995,70 C1calculatedasper thegeometricmethodof LangfordandBateman(1993)is M-Ml. is (4.3) difficult 'best fit' for CI andnot attemptedasresultswouldobviously 95% complex model the more of a be very similar. "it is interestingto notethat this figure is very similar to that reportedby Kristrom(1993)in his studyof forests. Swedish This be if WTP for to similarity even would greater we extend our were preservationvalues DC bid OE had higher DC their include overbidders/yea-sayers and expected strategic counterparts even to axis levels beenused.

4.13

necessary assumptions (discussed in chapter 2), truth-telling is the optimal strategy in a DC The observed discrepancy between OE and DC results is therefore not

format (ibid).

inconsistent with economic theory. However, such results can also be explained in terms of

in 2. biases discussed Here, commentatorshave seen 'psychological' chapter the certain of OE/DC divergence as evidence of some sort of anchoring effect in the DC responses (Kahneman and Tversky, 1982; Roberts et al., 1985; Kahneman, 1986; Harris et al., 1989) for DC biases the the term this effect of overall various potential as a generic can use and we identified.

Figure 4.2:

Survival functions for OE and DC responses

100 90 80 response (%)

70 60 50 40 30

DC responses

20 10

................ .... ......... 1 1

5

10

........................ ......... ......................... 20

50

100

OE responses 200

500 WTP

Testing for such anchoring is problematic. Kristr6m (1993) discounts comparisons based upon means becauseof their implicit assumptionsregarding distributional form. fie based between OE DC distance tests the the non-parametric upon and of the use suggests However, there the functions. complications associated simultaneous are with various survival data. Kristr6m discrete therefore to approach uses a an and continuous such of application do OE DC from that to the the same test show and not come responses simple chi-square distribution.

This is supplemented by a somewhat unusual test of an anchoring hypothesis

in which responses from the OE sample are compared with supplementary OE responses

4.14

distance Although, Kristr6m is DC known to by the test states, the sample. as used given have low power," it is interesting to note that the computed statistic actually rejects the anchoring hypothesis (a = 5%). However, comparative analysis can be used to illustrate the strength of the apparent In OE DC between difference to order to underline this and questions. responses cognitive difference it was decided to treat all the OE responsesas if they had come from DC El, 0 'yes' bid levels ClO bid WTP OE i. to taken and as a response of; was questions, e. an ElO and a 'no' responseto all others. Here the (optimal) log-logistic model gave a very much it did for freedom) 6 degrees 227.72 deviance (residual data than fit of of with to the poorer in format DC indicates 6.24). deviance This (residual data that the DC used of the genuine WTP, fixed framework their well very of a evaluating within respondents places questioning described by the log-logistic model, whereas OE respondents appear to be undergoing formulating in their different responseS22. and stating processes cognitive significantly iv

The IB experiment

into WTP DC initially then entered question were All respondents presentedwith a 'reasons for Discussion and refusal' the bidding question of payment principle game. the IB for the IB sample are therefore as for the DC experiment. IB WTP the the of proceduregave a mean end The open-ended question presentedat in OE E80.55). However, E69.27; CI the (95% experiment, this amount E74.91 as WTP of = bids'. On-dssion 5% higher WT? the of upper truncation the of highly to responsive was in OE E52.41. As in 30% decline in the to the for mean of responses, example, resulted a in overbidding strategic that engage respondents certain the then, possibility experiment

be cannot ruled out. in OE bidding iterative in bid to WTP an final given response the game was As the

but be in bid truncated dependent continuous curve estimation will any variable question,the between DC the function association Bid strong positive a revealed analysisquickly at zero. IB (which the the to point of game)and constituted starting bid level presented respondents This 113 by the the WTP relationship process. respondents at end of amountstated the final

21TheKolomogorov-Smirnofftest; seeKanji (1993).

'Another approach might be to compare common covariates in the OE and DC bid functions. 23Full results in Bateman et al., 1993.

4.15

was strongest when both the final WTP responseand the initial bid level were expressedas natural logarithms. The optimal model is given as equation (4.4). LWTP(IB) = 2.104 + 0.3733 LBID + 0.000005 INC (22.18) (19.79) (1.86) 0.1758 BOAT + 0.1720 ENV - 0.1222 FIRST (3.67) (4.70) (-2.89)

(4.4)

21.86% R2= Total df = 1634 log final in WTP IB LWTP(IB) the of respondents natural statement game. where: = log bid LBID the of amount offered to respondents natural = household income (continuous variable) INC respondents = is if his/her first (=O FIRST to on the respondent visit area otherwise) =I Other variables as previously defined. Signs on the explanatory variables of equation (4.4) are as expected.The variable INC is included for interest although it is only significant at the 90% confidence level. Interestingly when tested, the variable LINC was found to be significantly weaker. As far by ignoring the most powerful explanatory variable was the (log) the constant, expected, bid level first presentedto respondents.This appearsto have strongly anchoredrespondents into a corresponding range of final WTP bids, i. e. a classic starting point effect. In their theoretical analysis,Hoehn and Randall (1987; p.237) appearto imply that DC identical from bid levels, IB should yield similar mean results. started when approaches, and Clearly this has not occurred in this case. Our IB format can be viewed as an amalgam of it is OE DC such and as not surprising that we see evidence of several approaches the and in formats IB The initial bid those the of reflected responses. power of the characteristics of level, so dominant in the DC bid functions, is clearly apparent. However, the IB approach 'understatement' free-riding OE for traits such or expected-coststrategiesto as now allows in WT?. the reduced estimate of mean reflected as emerge

The MLM experiment An interesting characteristicof the WTP data was identified by comparing responses first, developed dichotomous the bounds. Here it was noticed across second third and they as

4.16

that, at all bid levels, respondentsexhibited a certain unwillingness to accept a doubling of a previously accepted amount. This trend was, to varying extents, apparent whether that intensify) E10 E500, (and to appear or and continued at successive was previous amount bounds. At the second (or third) bound respondentsappear to view their previous accepted bid as more or less representingtheir total WTP and therefore resist the further doubling of this amount. This meansthat, at a given bid level, we are likely to have a lower proportion bound (because doubling first from those than the the second an at up of recorded refusals at initial lower bid level may refuse to pay this amount). This will mean that the discrete fall between bounds, WTP will a result which accoids with the mean of variable estimate findings of Hanemannet al. (1991). The bid function for the MLM experiment was estimated by Ian Langford and is in Langford, (full details Langford, 1994, here Bateman and upon therefore not reported 1996). However, this model confirms our DC finding that, in such experiments,the bid level far influence initial Our is by the strongest response. upon presented to respondents in bid levels is DC bias downward this confirmed analysis. also across a of observation Possible reasonsfor such a bias are interesting. Carson et al. (1994) suggesttwo routes by which such an effect may operate: Respondentswho agree to the initial DC amount and would, a-priori, have paid the A. (higher) 2DC amount may still refuse the latter if they feel that the governmentwould

wastethe extramoney. B.

Respondentswho refuse the DC bid but would, a-priori, have paid the (lower) 2DC if lower latter this they the amount with a reduction equate amount may still refuse in either the quality of the good or its probability of provision. Both these response patterns arguably accord with economic theory regarding

A However, initial DC by formed that type the argue reactions may we amount. expectation initial for influences by the status quo be preferences psychological regarding augmented also initial feeling DC response by 1993) that the Knetsch. a consequent and respondents (as per increase. become they then to to which attached and unwilling representedan agreedprice These effects will intensify betweenthe 2DC and 3DC questionand are reflectedin the i. 3DC is WTP for E82, ML the mean about e. of response, which, estimate overall for our analysis of responsesto the initial DC bid level. lower that than considerably

4.17

Summary and conclusions Table 4.3 summarisesmean WTP results from our analysis of elicitation effects. While there are some similarities acrosselicitation methods(i. e. optimal bid functions contain a number of common explanatory variables and the confidence intervals of the mean WTP estimates also overlap), there are more dominant differences. The possibility of conflicting (high free-riding as considerable strategic overbidding, as well uncertainty and effects such as if likely. However, OE many, seems not all, of these responses variability) within for be accounted within econornic theory. can characteristics Table 4.3:

Mean WTP results from four elicitation methods

Elicitation Method

Mean WTP (f.)

95% Confidence Level Lower (f)

Upper (f)

DC

143.18

75.00

261.00

IB

74.91

69.27

80.55

MLM (3DC)

81.65

44.32

118.97

OE

67.19

59.53

74.86

1 1

The disparity betweenOE and DC resultsmight also be explained by economic theory. influence be here. The large of the bid valid However, psychological argumentsmay also level within the DC bid function can be interpreted either as an expected economic price bias. Because the number of potential exacerbating,conflicting of anchoring an as or effect, formats, have doubts both discussed to the we about respect with effects and confounding Rather OE DC between to we choose results. and comparisons simple of usefulness

if it from data derived DC OE treated as were test the which emphasisethe results of difference is highly indicates This that there cognitive significant response a questioning. dependingon which questionformatis beingused. Thesedifferencesin interpretationappear initiated by include these indicate questions certainquite separate processes the that mental to be Such both to reinforced conclusions seem economic and psychological. elements,probably by the findings of our MLM analysis which echoesour DC findings but introduces further influences upon respondents. economic/psychological

4.18

The IB approachcan be seenas a hybrid of both the DC and OE formats and as such demonstratesa mix of the effects associatedwith both. The dominance of the bid level, so characteristic of the DC approach,is clearly evident as a classic starting point bias (Roberts et al., 1985). It appearsthat, once the initial (DC) responseis elicited (possibly raising stated WTP), the ensuing respondentcontrol may engenderOE-type 'understatement' strategies. In conclusion, our elicitation effect study has raised important issues regarding the bulk for The of economic theory appearsto suggest remaining research. our optimal strategy that DC approacheswill come nearestto eliciting 'true' WTP' while OE methods will lead to some understatementof such amounts. Conversely, psychological theory suggeststhat, in from both OE suffer under and overstatement separaterespondents, may responses while DC methods will generally exhibit overstatementas a result of the group of influences we have labelled as anchoring. Furthermore, we have shown that these argumentsand others both IB MLM degrees in the to and approaches. varying applying We are therefore forced to adopt a pragmatic solution to this quandary25. Ile US NOAA panel guidelines on CV (Arrow et al., 1993) clearly state that a conservative design is likely Such be WTP. WTP to to to likely one overestimate preferred to underestimate more interpretation Atlantic by UK this the the side on of of reflected guidance seems Consequently have by H. M. Treasury". OE adopted we evaluations environmental likely for that these the on grounds the of our research are remainder elicitation techniques in interpreting final However, WTP. lower-bound the on results, estimates to give on-balance be WTP should such amounts not overlooked. true that may exceed possibility

4.3

WOODLAND

RESEARCH

Three separatewoodland recreationevaluation studieswere conductedwhich we shall

is defensible 2'Sugden (forthcoming)fundamentally questions whetheror nottheconceptof 'true' preferences bias, that CV the of others argue minimisadon the as see main problem researchers that while arguing just it being how is (rather to the than question asked asked)and cannottherefore according preferencesalter be.stafically 'true' in the senseimplied by CV studies. 'rhe alternative- to abandonour valuationexercise- being rejectedon the groundsthat non-economic For failed have the to assess of environmental adequately recreation goods. a contraryview value approaches diversity the (1994) that monetary of valuesgenerated Adams evaluation argues methods cannot capture who see by environmentalpublic goods. 'Conversation with numerousseniorcolleagues(who would doubtlessprefer to remainanonymous)at H. M. for Treasury lower-bound in institutions the preference UK of confirm assumptions suchstudies. various The nearestthing to written confirmationof this is given in Whiteman(1994).

4.19

refer to as: Thetford 1 (carried out in 1990); Wantage (1991); and Thetford 2 (1993). The first two of these were conducted prior to the Norfolk Broads study discussedpreviously. Fortunately the approachof the Thetford 1 and Wantagestudiesdoes not contraveneour later findings.

However, the experience gained during the Norfolk Broads and subsequent

later is Thetford 2 designed higher that to our means study studies a much evaluation specification than our earlier woodland research. in the remainder of this section we present summariesof our applied work. Where in 2. details full given appendix are relevant 4.3.1 THE THETFORD I CV/TC STUDY The Thetford I study was conductedin the summer of 1990 and consistedof a series both interviews" face-to-face users and non-usersregarding the recreational value with of The in Norfolk. Forest Thetford overall sample was sub-divided to permit a number of of differing CV analyses in addition to an ITC study of users. The structure of sub-sample follows: as analyseswas A.

CV studies: Users (forest) surveys i.

WTP via annual payment: tax vehicle; OE elicitation method

ii.

WTP via per visit payment: entrance fee vehicle; elicited using low

rangepaymentcard iii.

WTP via per visit payment: entrance fee vehicle; elicited using high range payment card

2.

Non-users (Norwich city) surveys i.

WTP via annual payment: tax vehicle; OE elicitation method

ii.

WTP via annual payment: poll-tax2s vehicle; with OE elicitation

method iii.

WTP via per visit payment: entrance fee vehicle; elicited using low range payment card

iv.

WTP via per visit payment: entrance fee vehicle-,elicited using high range payment card

27Theauthor is grateful to JoanneWall (UEA student) who conducted interviews for this exercise. 28MOreproperly termed the 'Community Charge'.

4.20

B.

ITC study

Questionnaires for the on-site(CV/TC) and Norwich (CV) studiesare reproducedin appendix 2. 4.3.1.1 Thetford 1: CV studies All CV sampleswere collected using face-to-face interviewing of randomly selected respondents. In all the per annum (but not per visit) CV evaluation studies it was decided, inform WTP to to question, respondentsof the current averagelevel of annual per any prior household payments to support the Forestry Commission which was estimated at approximately E2.60 pe.

Ibis approachfollowed contemporarypractice in UK CV studies

in Turner Brooke (1988), the of work as pioneered and a study which had particularly 30) (as CBA by H. M. Treasury. been a part of wider approved recently A number of socioeconomicvariables were collected in all surveys. In the caseof the included: forest home interviews these users with address; sex; age; employment; on-site income; interview interviewee location; a pensioner, precise the was preference for whether

history frequency forest to the and of visits specific recreation; site and or urban natural Similar WT113'. from and use value time on site; spent variables elicited the were entirety; knowledge the of addition questions regarding respondents of the forest with non-user samples and integral visitor sites. In all studies WTP responseswere investigated by regression analysis of underlying bid functions. Here a variety of analysesconcerning the specification and functional form for brevity However, best fitting bid reasons of was undertaken. only curves models are of

if differences be Results to then across studies see compared were could explained. reported. i.

The per annum payment studies: results In this section we report results from those studies in which respondents were

for WTP for questions their evaluation asking the recreational with per annum presented

29BasM upon Forestry Commission (1985). 3017ora review of the overall project (including the CV study) see Turner, Bateman and Brooke (1992). "A use + option value WTP was also elicited, but following our criticisms of chapter 3, this is not further. considered

4.21

facilities provided by Thetford Forest. As noted, all respondents to these studies were informed of the present level of mean annual per household tax contributions in respect of Forestry Commission grant-in-aid (E2.60 pa).

On-site survey: WTP direct tax per annum

A.

Responseswere elicited using an OE WT? question and a generalincome tax payment vehicle.

A useable sample of 46 interviews was collected32 of which 40 (87%) were WTP

at least some amount for the recreational facilities provided at Thetford Forest. WT? bids E10.00 EO. 10 bids from to three two per annum notable exceptions; of with generally ranged E50.00 and one of ;C52.00. Ibis gave an all sample mean WTP of E5.14" but a 5% trimmed Univariate WTP statistics are presented subsequently in comparison with

0.20. of mean

those from the other per-annum studies. Estimation of a bid function for such a skewed distribution was problematic.

However, a log (dependent) functional form satisfied an n-

(MINITAB, distribution test scores normal

1991) and the best fitting model is given in

equation 4.534.

InVvrTPftx=

1.146 - 0.652 STAY210M + 0.490 DAYS12 (1.71) (5.28) (-2.31)

20.1%

R(adj) = 16.5%

(4.5)

n= 46

where: natural log of WTP responseof forest users to per annurn (tax vehicle) question 1 if respondents average length of visit was at least 120 STAY120M minutes; =0 otherwise I if respondentvisited forest at least 12 times per annum; =0 DAYS12 otherwise Figures in brackets are t-statistics InWTPftx

Equation 4.5 is not particularly strong and reflects, we believe, the rather crude nature

"A further 4 interviewswereincomplete. 33Zerobids were not excluded.

3'We accept criticisms that, strictly speaking, OLS techniques should not be used in cases of discrete However improvement little distribution blocky the zeros. not particularly and some was with observations highly data. by technical be solutions such upon using gained would

4.22

it Nevertheless does satisfy the overall fit requirements of some of our early evaluation work. CV commentators (see chapter 2) and the individual relationships described seem plausible. It appears that regular/short stay visitors have higher annual WTP than do irregular/long stay former group seem straightforward (WTP rising with use). Further analysis

visitorS35. Ile

of the latter occasional visit/long-stay group revealed that these individuals had generally travelled relatively long distancesto the forest. Accordingly they were more likely to have a wider range of substitute recreational options for such visits than do the 'regular visitor' is lower WT? logical. therefore again a annual and group Remote survey: WTP direct tax per annum Although this survey was conductedin the centre of Norwich, some 25 miles remote be it Forest, Ibetford misleading to think of this as a survey of pure non-usevalue would of 41 knew (77%) forest, 25 (53%) had 53 the of while respondents, visited the sampleof as, of it. 49 fully completed questionnaireswere collected of which 41 (84%) were V; TP at least facilities forest. bids WTP for the the provided at generally ranged recreational some amount from fo. 10 to El 0 with three exceptions:two bids of ;C20and one of E52 (coincidentally the This bid). WTP E3.51 5% highest sample gave an all mean of with a the on-site same as detailed WTP N-scores E2.22 (univariate testing statistics are subsequently). trimmed meanof; best fitting is (dependent) be log function bid the to normal and model given the confirmed in equation 4.6.

InWTpntx

R2

0.260 VISARB 2.07 HOME + + -2.33 (2.04) (-0.93) (2.04) 14.7%

R2(adj) = 11.0%

(4.6)

n= 49

where:

InWTPntx

naturallog of WTP responseof Norwich subsampleto per annurn(tax vehicle)question

1 if respondentshome addressis in Norfolk or Suffolk; 0= otherwise HOME I if respondenthad visited Thetford Forest (Arboretum site). VISARB Figures in brackets are t-statistics.

3.STeStS of potential collincarity betweenthe explanatory variables (correlation coefficients and impact upon in is turn) this one variable suggest omitting not a significant problem. of coefficients

4.23

Equation 4.6 is arguably even weaker than that for the forest users (4.5), a finding lower level is the given generally of site knowledge of the Norwich which unsurprising individual However, the with explanatory variables is logical relationship subsample., indicating that those who live closer to the forest as well as those who actually visit it have a higher annual WTV. Remote survey: WTP poll tax per annum

C.

This subsamplewas collected in a manneridentical to that describedat B aboveexcept that the payment vehicle was altered to the Community Chargeor 'poll' tax. This was a local taxation system which had recently been imposed upon local councils by the government of the day. The tax was the subject of extreme controversy at the time of our survey. While in CV to controversial attempt avoid a payment vehicle, a survey would any normally factors be held could other relatively stable and the strength test such as ours, comparative of reaction to the vehicle alone, assessed. it was quite clear from responses that interviewees reacted very strongly to the use of the poll tax vehicle.

Refusals to pay increased dramatically such that only 23 respondents

(49% of the total subsamPle of 47) were willing to pay anything at all (compared to 84% of distribution However, bids direct the Norwich tax of non-zero respondents). was much the less concentrated upon low amounts than in the other per annum experiments with far more in an all-sample mean of E7.09 and a 5% This being bids high resulted recorded. relatively both decision bid Statistical E5.19. that the to analysis revealed and the of trimmed mean income bid than correlated with respondents strongly any more much were of magnitude No in been had the experiments. other variable proved significant other explanatory variable in explaining WTP responses under the poll tax vehicle, and the best fitting bid function is 37 4.7 given as equation .

InWTPnpoll

R' = 23.6%

(4.7)

0.000098 INCOME + -0.129 (-0.37) (3.72)

R(adj) 21.9%

3&rr

n= 47

for multicollinearity showed no problem here. 'StS 37ASbefore an n-scorestest confirmed the suitability of the log-dependentfunctional form.

4.24

where: natural log of WTP response of Norwich subsample to per annum.(poll tax vehicle) question INCOME respondentshouseholdannual income (E) Figures in brackets are t-statistics. InWTPnpoll

It appearsfrom equation 4.7 that high income individuals react positively to the polltax vehicle by increasing their stated WTP while the reverse is true of low income WTP state a of whom zero many under the poll tax vehicle. If we characterise respondents, its Government the and policies as typically having aboveaverageincomes then supportersof this result can be interpreted as reflecting political preferences. D.

Comparison across the per annum studies Analysis of responsesacross our three per annum studies reveals some interesting

Our first consideration was to investigate the socioeconomic characteristics of for factors Summary for to check confounding subsamples etc. statistics across respondents findings.

4.4. in table this analysis are given Comparison of the two Norwich subsamplesshows a reassuring similarity between factor The is facing tax tax the questions. only and poll which significantly different those between the two groups concerns the 'home' variable as all those facing the poll tax vehicle

is 'niis by higher income Norfolk/Suffolk from the somewhat offset the area. slightly came in the direct tax sample, although this latter difference is statistically insignificant. While the two Norwich subsamplesseemvery similar the Thetford Forest sample is from likely Ile 'home', different to a separate underlying come appears population. and very 'income' and (obviously) 'visForPa' variables appear to be quite dissimilar (although 95% least While do data intervals to the extent). some at visits comes as no overlap confidence in income Thetford Forest differences the that to suggest variable visitors enjoy a surprise, by Norwich do higher to than the our pay population represented ability subsamples. generally This factor, combined with the higher use rate of on-site interviewees supports the observed higher mean WTP (and WTP distribution characteristics)of this group than those faced with in Norwich (direct Table 4.5 WTP the tax) question survey. gives same univariate the faced for WTP three subsamples with per annum questions. all statistics

4.25

Table 4.4: variable

Socioeconomiccharacteristicsof the Thetford 1 per annum WTP subsamples n

mean

st. dev

95% C1

se mean

lower limit

upper limit

Thetford survey: WTP direct tax sex age home income knowFor visForlO visForPa

50 50 50 50 nja n/a so

0.62 1.94 0.76 15800 n/a n/a 5.98

OA9 0.74 0.43 9793 n/a n/a 10.25

0.07 0.1 0.06 1385 n/a n/a 1.45

OA8 1.73 0.64 13016 nja n/a 3.07

0.76 2.15 0.88 18584 n1a n/a 8.89

0.59 1.68 0.917 12679 0.77 0.47 1.89

0.50 0.73 0.30 9243 0.42 0.50 4.47

0.07 0.10 0.04 1270 0.06 0.07 0.61

0.45 IA8 0.82 10131 0.66 0.33 0.65

0.72 1.88 0.99 15228 0.89 0.61 3.12

0.58 1.72 1.00 11175 0.78 0.50 2.21

0.50 0.78 0.00 6524 0.42 0.51 5.53

0.070 0.11 0.00 923 0.06 0.07 0.78

0.44 1.5 1.00 9320 0.66 0.36 0.64

0.72 1.94 1.00 13030 0.9 0.64 3.78

Norwich survey: WTP direct tax sex age home income knowFor visForlO visForPa

53 53 53 53 53 53 53

Norwich survey: WTP poll tax sex age home income know r visForIO visForPa

50 50 50 50 50 50 50

wherc:

sex age home income knowFor visForlO visForPa Notes:

0 female if if male; =I 2= (I young; middleaged;3= old) category variable = = =I if homeis in Norfolk or Suffolk;0 otherwise annualincome(L) = household knowsof ThetfordForest;0 otherwise =1 if respondent hasvisitedThetfordForcst;0 otherwise =I if respondent = averagenumberof visitsto 7betfordForestper annum

1. All values identical (all 'home' respondents) n/a = questionnot applicableto on-site survey

Table 4.5 also shows the dramatic impact of changing payment vehicles. With the

higher direct the tax vehicle, use rate and ability to pay of Thetford visitors (FTAX) common leads to a mean WTP in excess of that for Norwich interviewees (NTAX). However, (NPOLL) to the poll-tax controversial vehicle reversesthis situation. Figure 4.3 switching

4.26

clarifies what has happened by plotting out the WTP distributions

derived from each

subsample. The switch to the poll tax vehicle increases the number of zero bids but, more

importantly (with respectto impact upon mean WTP), also raises the number and magnitude of relatively large bids.

It appears that in all three cases there are small groups of

respondentspreparedto pay proportionately very high amounts but that the poll tax vehicle considerably inflates this trend. Table 4.5

Group

Univariate WTP statistics for per annum subsamples mean (f pa)

n

FrAX NTAX NPOLL

median (f pa) 2.00 0.70 0.00

5.14 3.51 7.09

46 49 47

3.20 2.22 5.19

WTP Forest Thetford FTAX asked subsample = NTAX = Norwich subsample asked WTP pa via NPOLL = Norwich subsample asked WTP pa via Number of incomplete questionnaires (refused to give WTP Minimum bid is zero for all subsamples (not excluded from Lr. mean = 5% trimmed mean Notes:

Figure 4.3

st. dcv (L pa)

Lr. mean (f pa)

sc mean (f pa)

12.35 8.26 15.02

1.82 1.18 2.19

95% CI (f pa) lower

upper

1.49 1.13 2.68

8.81 5.88 11.50

pa via direct tax direct tax poll tax bid) as follows: FTAX = 4; NTAX = 0; NPOLL = 3. calculation of mean).

Distribution of WTP responses across the per annum evaluation Subgroups

(A

cCD 'a c 0 cl V) CD 0 0

z

5

10

15

20

25

30 WTP

Notes:

WTP Forest FTAX tax pa survey, = NTAX = Norwich survey, WTP tax pa NPOLL = Norwich survey, WTP poll tax pa

4.27

35 (2)

40

45

50

55

60

Returning to table 4.5, anotherimportant finding is that )VTP confidence intervals for the FTAX and NTAX groups indicate that meanWTP is not significantly dissimilar from the informed of as being their current annual paymentsfor this amount which respondentswere good. Comparison of the FTAX results with those for our subsequentThetford 2 study (which included a very similar subsample)show a marked difference in WTP distribution". We conclude that many respondentsdid use this information regarding existing paymentsas an anchoring point on which to base their WTP response. We feel that this conclusion also fact NPOLL that the to the that the E2.60 information point falls subsample and applies CI for 95% this group merely underlines the dramatic strength of the payment the outside by distinct halves (most likely this two group of very exhibited effect reflecting the vehicle individual in is turn respondents of which proxied by the income political persuasion variable). in conclusion, the Thetford I per annum experimentsyield some interesting findings information Ignoring vehicle and payment effects. regarding such effects would particularly lead to considerablebias in WTP results. The study was useful in that it gave us reasonsto information in both the tax the and of present-payment vehicle use our subsequent poll reject for However, dubious the same reasons, are we regarding the validity exercises. evaluation from in this to them produced particular exercise and estimates not use the our evaluation of later benefit transfer work.

ii.

The per-visit payment studies: results

None of the following studiesinformedrespondentsof their existing levels of tax do forestry Here for and criticism of such an so our previous approach not apply. payments had two prime researchobjectives: we i)

ComparisonbetweenThetfordForestand Norwich respondents

ii)

Comparisonbetweenlow-rangeandhigh-rangepaymentcards. The latterpoint refersto the useof two subsamples within eachof theThetfordForest

'low-range' Norwich surveys, one of which per-visit was prcsented with a paymentcard and 'high-range' detailed in figure 4.4. a payment card as the other with and "Differences in meanWTP betweenthesestudiesand other per annum analyses(not informing respondents illustrated levels) in our review of per annurn evaluations in chapter 3. most are vividly of current payment

4.28

Figure 4.4

Per-visit studies: payment card ranges

low-range payment card ff): 0

0.50

1.00

1.50

2.00

2.50

3.00

Other (specify)

3.50

4.00

4.50

5.00

Other (specify)

high-range payment card (P) 2.00

2.50

3.00

A further considerationwas to compareper visit with per annurnmeasures. However, in the absolute the light of regarding value of per annum. estimates reservations our given information from level the regarding present of tax payments, such a anchoring apparent Accordingly dubious both validity. of per visit and per annurn measures was comparison Wantage informed in study at subsequent where respondents our were not of were used levels. tax payment present A.

On-site surveys: WTP entrancefees Comparison of the low-range and high-range payment card subsamples reveals

interesting differences and similarities. In both casesWTP was negatively correlatedwith the length but during the time correlated with of spent on-site positively visits. number of visits Both results are highly logical, indicating that those who take frequent but short visits are incur high fee this as the will a to vehicle overall cost given their visiting entrance averse infrequent but long duration Conversely those make who visits would see the pattern. for higher fee good value money, and consequently relatively as stated relatively entrance WTp sums. The inverse correlation between trip frequency and visit duration (essentialfor in both However, line the presenceof clearly subsamples. was evident reasoning) of such a be into bid that these two meant the variables could not entered same multicollinearity such function. Table 4.6 lists zero-order Pearsoncon-elationcoefficients for these relationships. It is interesting to note that, in every case, the relevant coefficient for the low-range for its high-range is This than that equivalent. suggeststhat the low-range weaker subsample

4.29

payment card has restricted the range of bids statedby respondentseven though, in theory, the payment card allowed for any bid. Table 4.6

Variables

Correlation coefficients: on-site WTP fees and relevant explanatory variables

low-range payment card InWTPffl

InVISUS LONG

InY; TPffl InWTPffh InVISITS LONG =I

where:

-0.128 0.254

high-range payment card

InVISITS

InV;TPffh

InVISITS

-0.166

-0.704 0.404

-0.335

natural log WTP entrance fees forestry subsample presented with low-range payment card natural log WTP entrance fees forestry subsample presented with high-range payment card natural log of number of visits per annurn of respondents in respective subsample if respondents average time on-site was at least 150 minutes per visit; =0 otherwise

Given that, due to multicollinearity, the visit rate (InVISITS) and visit duration (LONG) variables could not simultaneously be included in any bid function39, specifications is low-range in In determined. data the the terms of the event specified subsample were LONG variable while the high-range subsamplecontains the InVISITS variable. This derives from the somewhat larger proportion of long-duration visitors in the low-range (54%) than

high-range(46%) subsample.Equation4.8 gives the best-fittingmodel of the low-range payment card responses: InWTPffl

R' = 24.6%

0.706 + 0.179 LONG -0.381 PENSION (-3.37) (2.05) (10.52) R(adj) = 21.4%

(4.8)

n= 50

where: PENSION =I if respondentof pensionableage; 0= otherwise Other variables defined in notes table 4.6. Values in brackets are t-statistics.

"The correlation matrix (table 4.6) shows that this is not a clear cut decision particular for the low-range Omitted did tests variable not add much clarity to this decision. responses. card payment

4.30

The relationshipsof equation4.8 are as expectedwith the PENSION variable possibly interestingly, for to this was stronger than an income ability pay although, acting as a proxy

best fitting high-range 4.9 Equation the the model of gives variable. paymentcardresponses. 1.26 - 0.237InVISrr (15.28) (-6.87)

InWTPM

(49)

R(adj) = 48.5%

R' = 49.6%

n= 50

Variables as defined in notes to table 4.6. Values in brackets are t-statistics. The strength of equation 4.9 is, by CV standards,quite remarkable. However, its factors is little (such that suggesting many standard economic worrying, a as simplicity income) have little relevance to these responses (although further analysis showed all if low be significance). to signed of correctly relationships

We would suggestthat the apparentpower of both equations 4.8 and 4.9 may have been inflated by two related factors: i)

Respondentsconcepts of a socially reasonableamount to pay per visit (as per our discussion of the concept of a 'social norm' value in chapter 2).

ii)

The restrictions upon bids as perceived by respondents facing the payment card vehicles. Evidence for such conclusions is given by the strength of the constant in both bid

functions.

While this itself questions the economic validity of such responses, it is

in high-range the that, experiment where the perceived restriction upon to encouraging note bids is looser, the consequent increased variation in WTP is logically related to the InVISIT. variable explanatory Turning to consider the magnitudeof bids, it is interesting to note that bid distributions

but highly in bimodal less than the are relatedto the payment studies per-annum are much Figure 4.5 illustrates these to subsample. points and compares each presented ranges card for (upper for distributions those our on site with survey panel) our remotesurvey per-visit in Norwich 0ower panel).

4.31

Thetford I study: WTP response distributions for per-visit evaluations (upper panel = on-site survey; lower panel = remote survey)

Figure 4.5

20 18 c: (D 16 -0 14 12 cn

M 10

08 d6 Z4 2 0.0

0.5

1.0

1.5

2.0

2.5 WTP

V)

3.0

3.5

4.0

4.5

5,0

3.5

4.0

4.5

5.0

(2)

20 18

70

16

c 0 14 Cl. U) 12

10 0 0 Z6

8

4 2 0.0

0.5

1.0

1.5

2.0

2.5

WTP Notes:

WTPffh WTPffl WTPnfh WTPnn

= = = =

3.0 (E)

Forest survey, WTP fees (high range payment card: E2-f5) Forest survey, WTP fees (low range payment card: LO-O) Norwich survey, WTP fees (high range payment card: E245) Norwich survey, WTP fees (low range payment card: LO-O)

This apparent anchoring of responsesto the payment card range is also reflected in discussed in from WTP the those are subsequently comparison which with results univariate Norwich per-visit evaluations which we now turn to consider. 4.32

B.

Remote survey: WTP entrancefees In one senseasking a group of respondentsin Norwich about their WTP entrancefee's

to Thetford Forest may seem strange in that they might not ever wish to visit the forest. However, our exercise was intended to see to what extent our 'social norms' theory might increase in hypothetical the to such an as as see effect nature to group well what a such apply low/high-range have. The CV payment card effects were also might previous questions of here. pertinent To a certain extent the entrancefee proposition was not irrelevant to the Norwich per80% low high-range knew both In the and over of subsamples respondents visit subsamples. had it. knowledge forest 50% Interestingly forest the of visited while was nearly while of the )&7P, as per our respective on-site per-visit samples, visitation rate with correlated positively The best fitting bid functions for the low-range and high-range

was negatively related. 4.10 4.11 in and respectively. equations subsamples are given

InVvrTPnfl

0.633 + 0.331 KNOW - 0.390 PENSION (-3.18) (2.43) (5.14) W(adj) = 19.8%

23.0% InWTPnfh

(4.10)

n= 50

1.102 + 0.255 KNOW - 0.0257 VISITS (-2.64) (2.01) (14.88)

RI = 15.6%

where:

InWTPnfl lnWTPnfh

KNOW PENSION VISITS =

n= 50

R(adj) = 12.0%

faced Norwich (fee) low log WTP with per-visit respondents of natural range payment card (fee) faced high log Norwich WTP per respondents of visit with natural range payment card I if respondentknows forest; 0= otherwise I if respondentis of pensionableage; 0= otherwise number of visits made by respondentto forest per annurn

Tests of multicollinearity confinned that the inclusion of the KNOW and VISITS variables 4.11 was valid. within equation Figure 4.5 shows that, as with the forest subsamples,the distribution of bids from Norwich respondentsappearsto have been strongly anchoredwithin the particular payment to respondents. presented card range

4.33

Univariate WTP statistics are discussed in comparison with those from the Ilietford Forest per-visit evaluations below.

C.

Comparison of Thetford Forest and Norwich entrancefee studies Socioeconomic analysis of the per-visit subsamplesrevealed very similar findings to

4.4), (see for table analysis our per-annum namely that respondentsin our those reported Thetford Forest subsampleshad markedly higher incomesetc., than did those of the Norwich subsamples. Consideration of the bid distributions illustrated in figures 4.5 and 4.6 clearly shows that the ranges specified on WT? payment cards do significantly affect YVrM response. Increasing the bid range used increasesmean WTP in a way which we feel invalidates the in Consequently the we abandoned cards techniques. of subsequent use payment use of such research. Analysis of reported bid functions reveals some interesting findings. We argue that functions (consistently in the strongest predictor) suggests that the all constant the strength of influenced having fee a a socially notion of reasonable entrance respondents theory of our (see chapter 3) has some validity.

Other explanatory variables seem logical and consistent

functions bid for both However, the of theory. weaker nature considerably with economic Norwich as opposed to both Thetford Forest subsamples, suggests to us that the former group fees in determining WTP did their than entrance more uncertainty considerably experienced the latter.

We argue that this arises because of the inherently more hypothetical nature of

Although knew forest Norwich the to the of most sample. and such questions when asked increase is in It likely few it that this the time, had visited regular visitors. were at some hypothetical nature of the contingent market and consequent decrease in the likelihood of such fees, has led Norwich having the to to entrance respondents such pay respondents actually it for Thetford WT? Forest to the that that the true exceeds respective their extent overstating subsamples.

Given the more affluent

socioeconomic

characteristics of the 7letford

have been doubtful We to this a strong therefore effect. appears are of the subsamples, for Norwich Table 4.7 details WTP WTP the subsamples. estimates univariate validity of for all the per-visit subsamples. results

4.34

Table 4.7

Univariate WTP statistics for per-visit evaluations n

FFLOW FFIR NFLOW NFHI

50 50 50 50

mean median

1

1.21 1.55 1.45 2.37

1

1.00 1.25 1.25 2.00

1

tr. mean 1.17 IA2 1.35 2.34

St. dev 0.78 1.29 1.01 1.37 1 1

se mean

max

0.11 3.00 0.18 5.00 0.15 5.00 0.19 1 5.00 1

Q1

Q3

0.50 1.50 0.50 2.13 0.69 2.00 1.00 1 3.13 1

95%Cl lower

upper

0.99 1.19 1.15 1.98 1

IA3 1.92 1.75 2.76

Notes: FFLOW FFHI = NFLOW NFHI =

Forestsubsample,fee (per visit) vehicle, low rangepaymentcard Forestsubsample,fee (per visit) vchicle, high rangepaymentcard Norwich subsample,fee (per visit) vehicle, low rangepaymentcard Norwich subsample,fee (per visit) vehicle, high rangepaymentcard

No incomplete/refusalquestionnaires Minimum bid = zero throughout(not excludedfrom mean) tr. mean= 5% trimmed mean

iii.

Thetford 1 CV studies: conclusions

The Thetford I CV studies provided many valuable pointers towards better study design. These are surnmarisedas follows:

1.

Choice of payment vehicle can have a significant impact upon respondents' WTP. Despite the apparent attractivenessof a local tax vehicle, when this is politically (as per our poll tax subsample)responsesrelate to the vehicle rather than controversial the good. However, the direct tax vehicle appearsto have worked well.

2.

The payment card elicitation method appearsto anchor responseswithin the range shown.

3.

The use of per-visit measuresfor sampleswith high proportions of respondentswho be using the resource, appears to seriously reduce the credibility of the will not contingent market. Our subsequentstudies were designed with these findings in mind. One further

be is that measures per-visit may somewhat subject to influence from "social suggestions However, feel valuations. appropriate that the Ibetford I experiment we regarding norms'

4.35

was not sufficiently controlled to convincingly isolate such a finding and therefore did not rule out the use of per-visit questionnairesfrom subsequentresearch. We do feel however that further investigation of such a proposition is justifiable. Regarding our specific results from the CV studies as a whole, we feel that once the caveatsraised above are addressed,remaining results appearlogical and valid. In particular the relationships betweenvisit rate, visit duration and WTP are interesting. WTP was related positively to visit rate in the per annum studies but negatively in the per-visit studies. This having high total value but relatively low marginal value for visits a visitors reflects regular comparedto those of infrequent visitors. Converselyinfrequent visitors (proxied by high visit duration) have relatively lower total but higher marginal values than regular visitors. Such a result seemsto point to the underlying validity of our experiment once we allow for the biases we have identified in its course. Our findings from this study concerning the design of CV experiments fed directly into our subsequentCV experiment in the town of Wantage. 4.3.1.2 Thetford 1: ITC Study The Thetford I on-site survey also collected data for an ITC study of woodland This (compared a simple experiment to our subsequent was relatively recreation values. Thetford 2 ITC study) which used OLS40estimation techniquesto focus upon the impact of in form (and ITC As functional gain experience conducting work). even such a changing details involves complex series of analyses, a are presented along with the simple study in A brief is 2. here. this summary of appendix work presented questionnaire survey A sample of 129 parties representingapproximately 400 individuals was interviewed

duration; distance, data substitutes;andsocioeconomic cost and variables regardingvisit and dependent Initial the the correct specification analysis considered of variable were elicited. for our trip generatingfunction (tgf). A seriesof correlationand simple regressiontests dependent log This decision that clearly superior. a variable was wasnot so clearconfirmed Leading from discussion the of cost the variable was considered. specification on cut when in chapter2, all permutationof the travelexpenditureandtravel time costdefinitionsdetailed in table 4.8 were consideredin both linear and log form.

OLS investigated in detail in the Thetford2 ITC study. 40problems techniques were applying with

4.36

Table 4.8

Thetford I ITC study: travel expenditureand travel time cost definitions (both linear and log linear investigated)

Variable

Definition

Cost

Travel expenditure

1. marginal(petrol)cost

8p/mile

2. petrol and insurance

23p/mile

3. full runningcosts

33p/mile

1. zerocost (enjoystravel)

09o'wagerate

2. Dept. of Transportrate

43% wagerate

3. full time cost

100%wager! te

Travel time cost

Source: Based upon approach of Willis and Benson (1989) and Benson and Willis (1992). See discussion in chapter 2.

Detailed analysis of the complete set of cost permutationsrevealed that a marginally by defining logarithmic function (In COST) follows: fit a cost given as was superior In COST = In Courney cost @ 33p/mile + zero time costs) A considerableadvantageof using a cost function which is not (via time costs) linked income be is the that may entered as a separateexplanatory variable visitors to wage rates inducing collinearity problems. without Further explanatory variables were investigated through stepwiseregression analysis in Of full the collected these only the of socioeconomic variables survey. the range of income finding This household proved significant. again echoesthe results of respondents (Willis Benson, 1988; 1989) TC UK and studies which report tgfs relating visits to earlier

Equation 4.12 details best-fitting indicator tgf. of socioeconomic status. our cost and some

'. 37

InVISFOR = -5.548 - 0.9422 InCOST + (-1.30) (-8.41)

s=1.378

R2= 45.1%

1.0135 InINCOME (3.50)

R2(adj) 44.2%

(4.12)

n= 129

where: InVISFOR InCOST InINCOME

natural log of number of party visits to Thetford Forest per annurn cost variable (as previously defined) natural log of household annual income

All explanatory variables were defined in pence. Figures in brackets are t-statistics.

The overall explanatory power of our tgf is very satisfactory, considerably exceeding (1991) ITC Garrod Willis for studiesand higher than all but two of the 22 OLS the and that tgfs reported by Smith and Desvousges(1986) in their ITC studiesof water basedrecreation in the US. The impact of changing the functional form of the tgf was investigated by estimating linear Table 4.9 details for (dependent) models. and regression equations all three semi-log functional forms as well as giving consumer surplus estimates per party visit and per individual visit. The latter results are subdivided to consider different treatments of child

visitors. The valuation estimates given in Table 4.9 accord well with prior expectations. Clearly mis-specification of functional form leads to significant error in consumer surplus estimates.

Furthermore the issue of defining the individual visitor is highlighted by the

definitions. to alternative of valuations responsiveness

We feel that this is a potentially

has been date. Our to which not confusion and error properly addressed of case serious in is to concentrate upon the party as which adopt subsequent we work, solution, proposed individual level decisions basic thus avoiding subjective regarding values. the unit of valuation

4.38

43 0

cn 41

$

-! 41

1--ý

-ýr CD

.0



r

f

CD

00

Nt

01ý

00

C>

.0j

*zi

Ci

W

w,

00 'r,

v

00 I

I

-c 9 .0

id

Ir

40.

V:

C4 %0 114: Cý

OP rz

C?

lý v4ý

C4

9

0

r

9

9

C> %-0

E4-ý

00

t cý

t-

CO)

00 C4

(4

C? N

f9

0%

Id: .0

1

to '0 .2

Thetford I ITC studv: conclusions This study seemsto have succeededin providing a defensible valuation of woodland fitting best Our E3.37 tgf gives a consumer surplus estimate of per party visit recreation.

(fl. 07 per individual per visie'.) 4.3.1.3: Thetford 1 CV/ITC study: conclusions The Thetford I CV provided many important pointers towards improved study design in our subsequentCV work. However, the biases highlighted in this study means that we for benefit Conversely ITC transfer these purposes. our study estimates use readily cannot defensible have and provides estimates of woodland well to reasonably worked appears

recreational value.

4.3.2: THE WANTAGE

WTP/WTA

CV STUDy42

This study set out to assessvaluations of a proposed (hypothetical) community determine': Specific Oxfordshire. Wantage, to to aims were woodland schemenear 1.

The willingness to pay of the local community for the provision of a forest. This was is CV As household the site presently not available respondents study. achieved via a future rather than current users. potential are current

2.

The willingness to accept compensationof local farmers on whose land the proposed be located, feasibly thereby assessinguptake of recreational-access, woodland could woodland schemes.

4.3.2.1: Household WTP Survey: methodology Wantage is a rural town in Oxfordshire with a population of 11,495adults as recorded in the 1991 electoral register. It is 15 miles from any cities and although there are The distance facilities this there open-access within are no nearby woodlands. recreational

4'Treatingchildrenand adultsequally. 4IFulldetailsof this studyandaccompanying analysisaregivenin appendix2. The studyhasbeenpublished (1996a). l3ateman al. et as QA side issuewasto testdie feasibilityof applyingthe CV to a small scaleplanningissuesuchas this.

4.40

for discrete population sample which some demand for additional town therefore provides a is likely. facilities recreational The survey coveredthe four censuswards of the town, including the connectedvillage 400 households by Out Grove. these was selected targeting areasa stratified sampleof of of is This household the on electoral role. method consistent with that twenty-ninth every in CV (1988) Between by Tunstall their of sampling review procedure. et al. recommended July and Septemberof 1991 each selectedhouseholdwas visited and the 'head of household' If there was no responseon the first visit, the householdwas revisited on two day if being different time later the of and, necessary, at a second visit occasions; separate

interviewee.

29 later. Of 400 households least the visited, were one week a third was carried out at further further 37 to the and a questionnaire refused answer unobtainable after three visits, a 9 interviews yielded incomplete questionnaires. A useable sample of 325 responseswas therefore collected. Household questionnairedesign An initial questionnairewas testedin a pilot survey of 30 householdsnot selectedfor in The to: order in undertaken pilot was the study. main use Clarify the meaning of the contingent market description with respect to the it, in order to avoid market mis-specification of respondents' understanding (Mitchell and Carson, 1989). At this point set responsesto certain questions developed. the were scenario market regarding 2.

Assess the level of non-response to an OE valuation question as a

(Eberle Hayden, had highlighted this and as a problem contemporaryarticle 1991). Levelsof non-response werefound to be acceptableand thereforethe fonnat wasretained. Assess instrument bias. Initially only an annual trust fund payment vehicle fee, After the a per-visit entrance a was pilot second vehicle, was used. included to provide some comparison.

"Problems regardingthe definition of "headof household"are recognised.Selectionwas necessarilya it is not felt that any seriouserror was incurredhere. All those discretion interviewers and for the matter interviewedwereat least 18 yearsof age.

4.41

The main survey questionnairewas refined in the light of findings from the pilot and a full copy is given at the end of this section. Initial questions asked respondentshow long they had lived in the area. This was both to provide data on a potential explanatory variable interview to the to respondents process.Subsequentquestionsaskedrespondents and accustom to name sites of recreation that they had visited on a day trip basis during the last year and to state their preferenceswith respect to urban or rural sites. These questionswere included to encourageconsideration of preferencesfor competing recreation facilities and to establish familiarity with the proposed good. of a measure Following this the contingent market and payment vehicle were introduced via a 'constant information statement'which was read out verbatim to all respondents. Households be they not would preparedto pay towards provision of the wood. or then whether asked were Such a 'payment principle' question was included mainly as a way of validating zero bids as it was felt that directly presentingrespondentswith a WTP question might intimidate those 1989). Respondents 'no' (Harris hold to this question al., et answered who values zero who for their such a responsewhilst those who answeredpositively to reasons were asked state WTP the questions'-. were asked Two WTP questionswere used. Firstly respondentswere askedhow much they were WTP per household per annum (referred to subsequently as the 'per-annum' question). Secondly, respondentswere then askedhow much they would be WTP per adult per visit as 'per-visit' Here (referred fee to the then all as question). subsequently a car parking in both VTIT turn, the annual and some amount were presented with, were respondentswho for Ideally format to either separate samples we would want use each question. per-visit format or vary the order in which questionsare presentedso that any ordering or anchoring because However be an analysis was not undertaken such we were assessed. effects might in (this the size sample problem was rectified obtaining sufficient of a-priori uncertain 2 Thetford study). subsequent After the valuation questions,respondentswere asked to assesstheir expecteduse of included for This both to was a potential explanatory provide variable such a woodland. included indicate level function bid to of the and non-use and use valuation of the analysis of in willingness to pay figures. Ibis indirect method was considered preferable to asking 41ltwas subsequentlyfelt that the motivations behind positive responsesshould also be investigatedand such Thetford 2 CV into built the experiment. an analysis was

4.42

respondentsto divide their valuation into subcategoriesof existence,use and option value (as per Loomis et al., 1984) which we consideredto be a highly suspectprocedureliable to allow inflate to the altruistic motivations of their valuations. respondents Finally all respondentswere askedquestionsregarding their householdcharacteristics in order to establish socioeconomicfactors affecting willingness to pay. 4.3.2.2: Farm VVTA survey: methodology The study also examined the levels of payment required by local farmers for them to i. WTA The their the scheme compensation. woodland e. proposed relatively small undertake local fanning population posed an immediate problem regarding sample size. Farm addresseswere taken from the local telephonedirectory. Initially addresseswere in the three town those mile of to radius a order to maintain consistencywith within restricted in household However, the this failed to produce an acceptable survey. the scenariopresented finally Just forty farms was adopted. radius over a six-mile and were contacted size sample by mail to request a face to face interview. A considerableproportion of farms refused to be interviewed, the main reason being that, as interviews coincided with the harvest season (the surveys being conductedbetweenJuly and October 1991),farmersfaced heavy workloads interview". for Because for such refusals were reasonsunconnected available and were not from (as distinct household for the to say questionnaires refusals pay the of content with is for farmer the as a serious the rate not seen problem validity of the participation woodland) interviews Whilst farm In completed. total were we recognise and accept nineteen survey. highlight difficulty the of assembling a size, we would such a sample problems regarding large sample here and feel that the results can be acceptedas generally indicative of farmers attitudes. Farm questionnaire design Due to the limited availability of respondentsit was impossible to conduct a pilot Initial farms. questionswere related to the value of present agricultural production survey of data between 71is the expressedvalue of the provided comparison a and associatedcosts. in household the due the survey, and current to agricultural value given as woodland "A secondreason, given by four farmers, was that they had already participated in other researchsurveys further devote time. to unwilling and were

4.43

production. Furthermore, by initially establishing the value of the land on the farm, it was hoped to focus the farmer's attention on an acceptableand reasonablelevel of compensation for income loss due to the removal of land from presentproduction. Such an approachwas designed to minimise any tendency to overstate compensation requirements. After these questions the contingent market and payment conditions were introduced to the respondent in the following manner: "The purpose of this survey is to assessthe feasibility in this region of planting an for know, As recreational purposes. may woodland you under the Farm mixed areaof Woodland Scheme the government provides grants for planting areas of at least 3 hectares on fanns.

The scheme being examined in this survey would allow

farmers but in further to take these to up grants, addition receive participating local These from be trust. woodland a extra payments would conditional payments (with being to the accessible public the a small areaallocated for parking woodland on but land The remain your property you or your subcontractorwould would space). be expected to provide basic maintenance." This scenario proved to be similar to that embodied in the Forestry Commission's Scheme. Woodland Community subsequent The respondentswere then asked to state a minimum level in pounds per annurn per hectare (or acre), which would be acceptableto them in order to commit land into such a land how 71ey they would allocate to the scheme at the much also asked were scheme. be levels It level that the told noted respondents payment were not should stated. payment in This the to was order possibility of such schemes. avoid existing under available information providing an anchoring point for the valuations given. However, it was clear from the interviews that some of the farmers had prior knowledge of the schemeand levels have this may affected responses. of payment and

4.3.2.3: Householdsurvey: results i. Household characteristics

lengthof residenceTevealed Questions had regarding thatlessthan5%of thesample lived in Wantage for one year or less. The mean age of residence was 18.5 years. T'his

4.44

distribution was somewhat skewed with a 5% trimmed mean of 17 years and median The indicated 14 overall picture a high degree of familiarity with the of years. residency local environment. Respondentswere invited to list up to four recreation siteswhich they had visited over the past year and state averageannual frequency of visits to stated sites. Responseswere into three categories of recreation attraction: urban; park (i. e. nonsubsequentlyclassified fees); (open entrance and rural access). Responses indicate a with urban attractions frequency (a 8 higher ) to of sites visit rural mean of over visits/household p. a. significantly than either urban or park sites (means of 2.0 and 2.6 visits respectively). This trend was borne out by a direct question asking householdswhether, given the choice, they would prefer (indoor) 298 325 households (outdoor) the an urban recreation of or site. to visit a rural leaving just 27 (92%) to they that prefer visit a would rural/outdoor site stated surveyed households (8%) stating a preferencefor an urban/indoor site. Following the VV-M questions (discussedsubsequently47) to respondents were asked , Only households 11 (3.4%) how the they proposed wood annually. visit would often predict frequency just Mean the predicted visitation wood. was under they not visit would statedthat 15. Comparisonof responsesregardingexisting recreationvisits and expectedvisits to the is demand for high. Whilst the that wood relatively predicted revealed woodland proposed in favour be due difference to there over-enthusiasm of provision", may and this of some is clearly a rounding effect in predicting visits, this does demonstrate a very significant demand for the proposedwood. This is perhapsnot surprising given the notable absenceof land. in locality, the particularly of quality rural space open accesspublic Data detailing householdcomposition by age was also collected. Observationswere dependency (i. into to corresponding economic criteria roughly e. pregroups categorised income-earners, pensionable) and these categories proved young/mid/older school, school, is bid If in curve the subsequent analysis. adjustment made to recombine these useful bands domed distribution into the age width expected we observe roughly constant categories typical of a stratified sample.

47Thequestionnaireis reproducedas part of the detaileddiscussionof this study in appendix2. 48Analogousto the subsequentlydiscussedphenomenaof strategicoverbiddingin responsesto WTP questions.

4.45

Finally data was gathered regarding the economic characteristics of households. Principle amongst these variables was household income9. Assurancesof confidentiality and the use of information cards employing alphabetical income categories,appear to have allayed any resistanceto providing such information and a 100% responserate was achieved Sample income found to approximate a normal distribution about this question". was on the median E15,000-09,999 category. I

Refusals to pay

Prior to both the annual and per-visit format WTP questions,respondentswere asked in WTP for the proposed woodland. 'Mis they some principle, amount were, whether bid it felt in included to that, the absenceof a validate zero as primarily was question was for WT? inhibit bids their might respondents asking such and upwardly bias such a question, Such 'conservative design' which WTPthe accords with emphasis an approach upon mean design 'blue NOAA (Arrow 1993). All survey the ribbon' protocol those et al., underpins who responded negatively regarding the principle of payment were asked to specify their Details for of thesereasonsand overall refusal rates are given a response. such motivations in table 4.10. Table 4.10 indicates a relatively high refusal rate regarding the annual WTP question (24.3%). However, as an economic constraint (insufficient income, etc) was by far the prime do WTP for The theoretical sums such zero not pose a refusal, problem. residual motivation indicated include 'extreme format free-riding' for three this respondents who an refusals incentive as their underlying motivation. Such a strategy was expected to occur to some is indeed is level indicated lower The than that excessive and not considerably extent. (Bateman 1992). large Those in et studies al., scale user respondentswho refuse observed be bid the that the woodland should grounds open access could arguably be to upon interpreted as articulating a fundamental objection to the entire principle of the economic just (not monetary evaluation of environmental preferences),arguing appraisal of projects instead for a policy-led approachto decision making. If such responseswere widespreadthey

""Data was also gatheredregarding professionaland employment status. However, a logical categorisation information data achieved the satisfactorily and this not was was not used in bid curve analysis. of 'Ve view this as a good test of questionnairedesign. Similarly, Bateman ct al. (1992) record only a 6% in face face interview for to a question similar a situation. refusal rate

4.46

might provide a serious criticism of the basis of this.study. However, the observed scarcity of suchresponsescan be interpreted as a counter-argumentthat individuals recognisethe need

to allocatefinite resourcesin an economicallyefficient manner.

Table 4.10: Refusal reasonsand refusal rates for annual and per-visit WT? formats Reasonfor refusal

Annual WTP No. %

Insufficient income or other economic constraint

70

21.5

37

11.4

Access to woodland should be free

5

1.5

11

3.4

The Government should pay

3

0.9

0

0.0

The land should remain in agriculture

1

0.3

0

0.0

79

24.3

48

14.8

Total refusal numbers/rate Note:

Per-visit WTP No. %

Percentagesare basedupon the entire sampleof 325 households(all respondentspresentedwith both WTP formats).

The lower refusal rate for the per-visit format might be interpreted as reflecting a fees entrance over the more general annual payment vehicle. of use-related wider acceptance Whilst we suspectthat the difference betweenrefusal rates for the two formats is likely to be it be insignificant, could simply argued that respondents are expressing a statistically fees donations for than annual entrance rather which, amongst other use-related preference be A less favourable, interpretation less likely to to sensitive usage. second, attributes, are

include households do be sample that, will who not enjoy woodlandrecreation as our could fee households the the to state a persite, vehicle allows entrance such not visit and would (where fee) in knowledge WTP they to are the unwilling that, as nonpay sum an annual visit

does describe be logic If they also ultimately non-payers. such a significant will visitors faith in have less then to the per-visit the positiveresponses sample we should proportionof is It its fee household for that that notable not question. one stated reason refusing entrance intention it had it is Given likely that no that some such that of visiting. was to pay households were interviewed, this heightensconcernsregarding the per-visit measure. Such be by tempered to the observation that, within statedreasonsfor refusal, a conclusion needs

4.47

the majority centred upon economic constraints which themselves pose no theoretical

problems. iii. Mean Vv, 7P and analysis of distributions The Wantage CV study usedopen-ended(011)elicitation methods. In the light of our (see Norfolk Broads into technique the switching elicitation our of study effects research discussedat the start of this chapter) this seemsa valid approachalthough our findings indiGiven lower bound WTP. desire for OE of elicit estimates a general may that questions cate (Arrow 1993) desirable in CV design this seems studies et al., a potentially conservative feature of this study. Accepting these riders, table 4.11 gives univariate WTP statistics for foffnats. two to the responses

Summary WTP results (L's): per-annum (WTPpa) and per-visit (Y; TPfee) formats

Table 4.11:

Format

n

WTPpa

325

WTPfee

325

st.dev

Q3

tr.mean

9.94

10.00

8.64

10.66

0.591

50.00

2.00

15.00

0.82

0.75

0.79

0.64

0.036

3.00

0.50

1.00

se.mcan

max

Ql

median

mean

Note: All values in 1991prices. Minimum bid = zero for both formats (included in calculationof meanetc.).

Figure 4.6 illustrates WTP responsedistribution for both the annual and per visit first be At included there to All glance may appear as to zeros. are pay refusals questions. illustrated in figure 4.6, with the distributions between differences fundamental the certain Furthermore, the the than whilst per-visit values. skewed more seemingly responses annual declining distribution increase distribution the as values per-visit appears smoothly per-annum fl, E2, However, figures (50p, be closer etc). upon certain upon round to clumped appears inspection these distributions exhibit some similarities. Ilie characteristic of respondents in is, in the to the some extent, repeated scenario per-visit answers number giving round E5, LIO, typically responses were where etc. although examination of annual sum experiment in is distributions that this the per-visit shows rounding effect more pronounced the overall format question.

4.48

Figure 4.6:

Responsedistributions for annual and per-visit fonnat WTP question (WTPpa and WTPfee respectively).

0 CY C13

ul N V)

A c v

10 c 0 C6 U)

z

6 z

05

10 15 20 25 30 35 40 45 50 WTP (C)

0.00

0.50 1.00

1.50 ZOO Z50

3.00

wrp (E)

a) 'Wann-glow' altruism Further examinationof the two distributions showsthat, examining non-zerobids, both VV7P level increasesfrom zero to in 'positive' initial increase the as responses exhibit an distributions low tail off. This trend has been the after amount which some relatively indicate 1992) (Bateman may an effect similar to the 'warmand et al., observed elsewhere (1990) 'purchase by Andreoni the or of moral satisfaction, glow giving' phenomenaproposed idea put forward by Kahneman and Knetsch (1992). Andreoni (1990) discussesthe conceptof 'impure altruism' whereby individuals donate

'warm-glow' a Of giving. Therefore,in to charitablegood-causesso that they can enjoy (Probably state may some answeringour questionnaire,certainrespondents small) bid for This that suchrespondentsare genuinely provided poses no problem reasons. warm-glow be it that However, somerespondentsseethe CV may stated. to the amounts pay prepared hypotheticalscenarioas an opportunityto endowthemselveswith a warm-glowsatisfaction WTP be true to Such state a Of zero and will prefer' will unwilling respondents at no cost. 4.49

" bid. (again to state some probably small)

A related issue here is that some respondents

may have an aversion to stating a zero response. Motivations for such a responseare many interactive interview Ome discusses but (1962) the the upon process. centre and complex 'good respondent' who attemptsto please the interviewer by stating what they perceive as a $correct' answer. A zero bid is unlikely to be thought to conform to such specifications. Similarly the respondentmay hold the interviewer in high esteemand again 'try to please'. A further motivation may be the desire (either conscious or subconscious)to conform to a $socialnorm' WTP as discussedin chapter 3. All the above motivations are liable to lead respondentswho would not actually pay from interview bid from towards the one which arises mechanism. and a zero stated away Such a responsecannot necessarilybe attachedto the specific good in question i.e. we could 'good individuals (note, for those scale cause' and concerned not similar the any good change all respondents)would still give the same responSe52. Whilst it was not possible,without adopting extendedpsychological testing, to identify implication bidders, the 'warm-glow' to analysis undertaken examine of a simple was such fell level into 'warm-glow' below bids the Here that a certain all we assumed such strategies. dictated by limited but is This one which was resources. approach crude a clearly category. The distribution of bids under both formats were examined for evidence of any appropriate low bids The suggested certain category amounts observed earlier of rounding cut-off point. 'warm-glow' bidding. For format the to annual give under choose which respondentsmight let us assumethat the relevant bid threshold is E5 p.a. whilst for the per-visit question we can by bids WTP We 50. EO. to all mean setting can now recalculate up threshold of a assume 4.12 details the results of such an analysis. Table to including thresholds these zero. and Table 4.12 indicates that, for both formats, even if we adopt the very strong 'warm-glow' including bids the threshold chosen are responses to up and assumption that all be little then this (again, really zeros, should makes relatively assumption) a strong and for for format 11 % 17% declines the the annual and difference to the estimatedmean, which in fact, We they too format. that strong as omit are, such assumptions suggest would per-visit

by roundingeffectswhich,as Batemanet al. (1995a)argue,are likely 5IThisproblemwill be compounded direction. in upward to operate a generally 51lnshort,suchrespondents would statesucha bid for any similar goodcause,i.e. woodlands,the dogsetc. home,the donkey-sanctuary,

4.50

bids which are significantlynon-zero. Table 4.12: WTP format

Annual Per Visit

Note:

Impact upon estimatedmeansof truncating potential 'warm-glow' bids

truncation option'

mean WTP W

median WTP W

St. dev.

untruncated

9.94

10.00

10.66

truncated

8.85

10.00

11.36

untruncated truncated

0.82

0.75

0.64

0.68

0.75

0.63

1. Untruncated= all bids includedas received.Truncated= all per annumbids up to L5 (inclusive) set to included 50 (inclusive) All (n=325 LO. bids to to to set pay are as zero. refusals zero's up all per visit zero; throughout).

We conclude then that although 'warm-glow' bidding may be a feature of this and impact is the to this tendency CV study of any such not severe. surveys, with regard other b) Free riding The non-woodland researchdiscussedat the start of this chapter suggeststhat free OE We have in WTP to incentives questions. stated reduce responses somewhat may rider free-riding does be to that to not appear extreme particularly pay our analysis of refusals in downward free-riding, form less However, in the of a revision extreme this study. evident both 'warm WTP. If bids to a so as reduce mean glow, bids non-zero may operatewithin of in directions. However, in these then act opposite 'free-riding' would effect are operation and basis be the would, on of the paucity of self-cancelling might that effects such to suggest is both All be hand, that conclude either or can effects we seriously premature. evidence to degrees. in be to uncertain operation may

c) Strategicoverbidding Chapter2 discussedthe possibilityof certainrespondents overstatingtheir true WTP for strategicreasons.Extremestrategicoverbiddingwill be evidencedby uppertail outliers 4.7 in WT? In Figure WTP high to their omission. responsiveness mean and a consequent from lowest horizontal for highest been have to the that sorted along axis showing responses wererecorded. both paymentvehicles,a few relatively high responses Considerationof figure 4.7 suggeststhat, if strategicoverbiddingis present,thenit i's In both the small number of respondents. per-annurnand per-visit to relatively a confined highest few bids does the very causethe meanto fall rapidly, suggesting formats,omissionof 4.51

these are the extreme outliers indicative of strategic overbidding. However, the rate of decline slows rapidly once these most extreme bids have been removed. Clearly at some point we move from bids which are high becauseof (possibly) strategic behaviour, to bids interaction high because the of which are of preferencesand ability to pay. If we assumethat be identified by very disproportionately high bids, then figure 4.7 can strategic overbidding suggeststhat there are relatively few of these. We therefore conclude that strategic overbidding may occur in a small minority of cases. The impact of such bids will be relatively high and, may be responsible for inflating per-visit mean WTP by perhaps 10% and perby 20% WTP to anything up mean although, without carefully designed, specific annurn experimentation, such estimatesare merely ballpark figures. Figure 4.7: Potential strategic overbidding responses

48

Per visit .

..........

Annual

format format

(right (left

hand

hand

3.0

scale)

scale)

32.2.0 mean WTP E/household p. a.

mean WTP C/adult visit

16-

1.0 ...................

0

0.0

........ 0

60

120

1180

240

300

No. of respondents (sorted

by size of WTP bid)

The result that per-visit valuesseemlessresponsesto upperbid truncationcould be format behaviour. indicating Howto that this to are more resistant answers strategic taken as 'social follows hypothesis discussed in explanation our norms' chapter 3. alternative ever, an if responses to per-visit questions relate more to a notion of a 'reasonable' entrance fee for WTP lack the then this true to account apparent relative than of strategic would amount behaviour but in turn question the validity of such an approach.

c) Did curve analysis Validity testing was undertakenin part through bid curve analysis. Tbe,socioeconomic in data both linear and log-linear specificathe collected to survey was related and preference WTP per-visit and the response. per-annurn tions of

4.52

cd) The per-annum responses(WTPpa) Analysis showed that a log-linear specification of the dependent variable WTPpa better linear 4.13 Table than significantly versions. performed reports results from a forwardlog-linear dependent the analysis relating regression entry stepwise variable. InWTPpa, to significant explanatory variables. Table 4.13: Step Constant _ InINCOME t-ratio

Stepwise regressionof InWTPpa on 34 predictors 1 -5.397 0.755 9.79

InRURVIS t-ratio _ InPKVIS t-ratio

2

3

4

-5.096 0.683 9.06

-4.418 0.683 9.16

-4.214 0.647 8.54

-4.374 0.630 8.33

0.165 3.78

0.160 3.74

0.140 3.25

0.156 3.61

0.131 2.98

0.246 3.69

0.227 3.43

0.239 3.62

0.235 3.59

-0.56 -2.75 0.167 2.32

-0.52 -2.58 0.173 2.42

-0.59 -2.90

AGE 17-25 t-ratio _ InVISWOOD t-ratio -

21

6

-5.335 0.726 9.56

PREFrOWN t-ratio

S1

5

0.140 2.34 1.04

1.02

1.00

22.87

26.14

29.15

0.992 30.96

0.985 32.11

0.978 33.26

_R Notes: n= 325. Variable dcflnitions as follows: InWTPpa natural logarithm of householdsannualWTP (E) InINCOME natural logarithm of householdsgrossannualincome InRURVIS naturallogarithmof numberof visits madeby householdto rural sitesper annurn. InPKVIS natural logarithm or numberof visits madeto parks I if prefers town-bascdrecreation;=0 otherwise PREFrOWN AGE17-25 number of personsin householdaged 17-25years InVISWOOD natural logarithm of householdspredictednumberof annualvisits to proposed wood

The final equation reported in table 4.13 contains certain explanatory variables which inspection does However, be of coefficient values across steps to collinear. we might expect fairly they immediately as any obvious severe problems remain stable. reveal not Explicit tests for multicollinearity suggested that only the correlation between InRURVIS and InVISWOOD gave any real causefor concern. Accordingly the latter variable is best-fit 4.13. from dropped model our which reported as equation was

4.53

InWTPpa = -4.77 + 0.647 InINCOME + 0.156 InRURVIS (-6.70) (8.54) (3.61)

+0.239InPKVIS -0.556PREFTOWN+ 0.167AGE 17-25 (3.62) (-2.75) (2.32)

(4.13)

R2= 32.1% R(adj) = 31.0% n= 325 RegressionF= 30.17(p = 0.000) The bid curve model given in equation 4.13 fits the data well in comparison to most CV studies employing OE elicitation methods and satisfies the more stringent guidelines on theoretical validity testing (Mitchell and Carson, 1989; Bateman and Turner, 1993). More importantly the relationships suggested by individual explanatory variables are highly income It household is in that the expectations. a-priori appears with accord and significant WTP Responses dominant to the affecting responses per-annum question. consideration most linked to visits to rural or town park recreation sites while those who positively are also lower WTP. A final interestleisure levels significantly exhibit pursuits of town-based prefer ing factor is the positive influence upon WT? exerted by the presenceof householdmembers between the agesof 17 and 25. This may be due either to higher recreation demand or to an enhancedenvironmental awarenessamongst this group. In summarythe per-annumstudy appearsto haveelicited theoretically consistentWTP responses. fee) (WTP The ii) per-visit responses c. Per-visit WTP responses question were much less firmly linked to standard

bids. Regression WTPpa bid for the the than analysis of curve were variables explanatory While log-linear dependent this observation. a confirmed variable per-visit responses in data, bid detailed 4.14, fit best the the curve of resulting model, equation provideda degree low of overallexplanatory power. exhibitsa very LnWTPf. r%2

where

0.595 - 0.135PENSION - 0.00175VISWOOD (-2.26) (25.33) (-3.94)

(4.14)

5.7176 R2 (adj) = 5.1Clo n= 325 RegressionF=9.76 (p = 0.000)

InWTPf.,. PENSION VISWOOD

natural logarithm of stated WTP per visit number in household aged 65 years or over predicted number of householdvisits to the proposedwood per annum 4.54

The equation given in equation 4.14 takes a semi-log (dependent)functional form. Explanatory variable relationships are as expected. The negative sign on PENSION accords lower the expected visitation rate and ability to pay of this age group. The negative sign with on VISWOOD reinforces the relationship, observed in our Thetford I per-visit survey, of responsesindicating that regular visitors are more resistant to the per-visit payment vehicle than are occasional visitors. These factors provide the strongest support for the validity of our per-visit results. However, contrary evidence is suggestedboth by the poor overall fit of this model and the very strong nature of the constant. We believe that this latter factor for WTP further by that our contention per-visit responses are evidence affected provides social norm factors. iv. Summary results: household WTP studies It seems that responsesto the per-annum WTP questions were strongly linked to expected explanatory variables and therefore pass a simple test of theoretical validity". Responses to per-visit format questions were less strongly linked to such factors and, while justification have some they may still

as magnitude estimates, these results seem to support

hypothesis. norm our social

Convergent validity testing (seechapter 2) was not feasible for our per-annumformat UK literature. (remote the directly studies woodland exist within survey) comparable as no However, a within-format comparison across several different types of outdoor recreation logically WTPpa the to the substitutability, that mean was related above showed resources for determine factors WTP to seemed which results a sample uniquenessand provision change 1994). (Bateman, Willis et al., thirty studies of over Cross-study comparison of our WTPfee result was easier given the relatively high in literature. Our falls but WTPfee the studies mean above well numbers of comparable deviation UK the of mean of all other comparable studies-. within one standard

"In effect, responseswere in logical accordancewith economic theory. Wider questions regarding the (as in 2) CV reviewed chapter responses may still apply. of overall validity 54Meanof other per-visit OE use value studies = f:0.63; st.dev = LO.25. Full details of cross-studyanalysis 3. in chapter are given

4.55

4.3.2.4: Farm survey: results Responseswere elicited from nineteenfarmers using face to face interview techniques. Whilst we have already recognised problems associatedwith inferring from small sample difficult this given the necessarysteps to secure each sample proved even sizes, eliciting interview during the harvest season. We have no reason to supposethat those interviewed formed a biased sampleand therefore report percentageresponses(as well as numbers)as an in farmer to attitudes similar areas". expected approximate guide i. Fann characteristics The interview openedwith questionsregarding the generalcharacteristicsof the farm. Specifically farmers were asked to state the agricultural land use; farm tenure; and average 4.14 details individual farm Table hectare). (or responsesto theseand certain profit per acre other questions. Most farms (10 farms, equivalent to 53% of the total sample)were mixed agricultural The the a variety of other standard of with activities. remainder arable combining producers (7 farms; 37%), dairy farm one purely arable purely producers of sample consisted mainly Nearly in (5% farm the completed sample. all those setaside each) entirely and one interviewed owned their farms (17 farms; 90%). This may limit the applicability of results farms. tenure to Tented Fanmerswere asked to state their averageprofit", per acre under existing production. immediately following farmers the to to sensibly consider This was asked so as encourage financial levels to allow a comparison of compensation and acceptable regarding question between these two amounts. Mean statedprofit was E125/acre(009/ha). Individual stated in be farmS57 This due between to measure may some an considerably profit varied . farmers interviewer (three (16%) to this to the refused answer to profits reveal unwillingness indicate it However, in true turn may a wider understatement profit). was of question which felt that the majority of this variation was due to changes in economic efficiency and farms. across consequentproductivity

"We would expect participation rates to rise as per-acre agricultural incomes fall. Such conditions would Welsh hill farms. of studies apply to our subsequent

16'niesimpleterm 'profit' wasPreferredto any more technicaldefinition.

"Although only one farm lies Oust) outside the 95% confidence internal around the mean.

4.56

Table 4.14:

Farm characteristicsand farmers' willingness to acceptcompensationfor transferring from present output to woodland

Landuse

Tenure

Profit/acre (hectare)

I

Arable/ Sheep

Owned

L100 (U47)

2

Arable/ Beef

Owned

3

Arable/ Dairy

Owned

4

Arable

5

Farm

WTA/acre (hectare)

Allocation acres(ha)

Reasonfor non-allocation

L250 (L618)

0

Landshouldbe usedto produce food

L20,000 (L49.440)

0

Doesnot like government policy

L125 (L309)

L300 (L741)

0

Doesnot wantpublicaccessto the farm

Owned

;E30 (C74)

:C2OO (L494)

5 (2)

Arable

Owned

L105 (L260)

:E250 (E618)

30 (12)

6

Amble

Owned

L45 (L74)

L150 (L370)

2 (0.8)

7

Arable/ BcWLamb

Owned

8

Arable

Owned

9

Dairy

Rented

L85 (f-210)

10

Amble

Owned

LI 16 (L287)

11

Amble/ Sctaside

Owned

floo (L247)

12

Amble/ Beef

Owned

E186 (L459)

L100 (E247)

125 (50)

13

Arable/ Dairy

Owned

E186 (459)

L200 (L494)

too (40)

14

Amble/ Pigs

Owned

L163 (L402)

E250 (L618)

20 (8)

15

Arable/ Beef

Rented

L150 (L370)

L250 (;C618)

0

16

Amble

Owned

L280 092)

Amble

Owned

L145 (358)

Arable/ Dairy

Owned

L140 (;E346)

Setaside

Owned

17

19

L130 (021)

L300 (L741)

L600 (LI,483)

L250 (L617) E130 (021)

L250 (L617)

Mean

L57 (L141)

L121 (000)

4.57

Doesnot wantpublicaccessto the farm

0

Landnot suitableto growtrees upon

0

Doesnot wantpublic accessto the farm

0

Farmtoo smallfor the scheme

0

Doesnot wantpublicaccessto the farm

Doesnot wantpublic accessto the farm

3 (1.2)

L150 (L370)

Total

0

15 (6)

0

Farmtoo smallfor scheme

0

Faimertoo old to undertake long-termproject

0

Unwilling to undertakeanother schemeto Setaside

ii. Willingness to allocate land to the woodland proiect Twelve farmers (63%) initially

stated that they were unwilling

to allocate land for

Of these the most commonly stated reason for refusal woodland. public access recreational was that the farmer did not want to allow public access to the farm (5 farms or 42% of those refusing to enter the scheme). Concerns regarding a loss of rights following entrance to such founded. Repeated footpaths be public use of may well within a wood may lead a scheme to their classification as public rights of way. Furthermore, interviews with senior Forestry Commission staff revealed that current policy will not allow farmers to be granted felling licences unless equivalent areas of replanting are agreed".

In other words the decision to

into forestry irreversible. from be agriculture recreational may area well allocate a certain Such irreversibility forestry.

may perversely prove to be a considerable block to the extension of agro-

Other reasons for refusing to participation can be broadly classified as 3 (25%)

land 2 (farm (17%) disliked farm type); the particular size or which specific which were farmers These 2 (17%) the reflected particular preferences. categorisations which policy; and interviews differently. However, have these the of somewhat outcome as a classified might is is both It feel indication this that the tenure that acceptable. notable of Tented we rough farms declined to allocate land to the scheme". This may be because farmers felt that is from (which be legal have the to owners sought a requirement) or a permission would delayed However, disinclination the towards schemes. sample size precludes return greater drawn. being firm conclusion any

I

Seven farmers (37%) were initially woodland

scheme.

willing

to allocate land to the recreational

Given concerns regarding public

access this was felt to be an

just Oust 40 Mean 15 high over acres over allocation was rate. percentage encouragingly hectares)per participating farm. This mean falls to approximately 15 acres (about 6 hectares) if non-participating farms are also taken into consideration. Uptake amongst participating farms appears to be bimodally distributed with two farms willing to allocate 100 acres or into woodland and the remainder only willing to undertake small scale afforestation more is if is for Whilst to the objective aiding available schemes, grant scale small projects. be (unless discrete then they such small area pockets can recreational provide a viable, be Nevertheless by large the to suitable. agreement two not planting scale may combined) "Interview with Chief Forester, Santon Downham, Thetford Forest, 1993. "Subsequent analysis (see table A2.39) confirmed this as a statistically significant relationship.

4.58

farmers is encouraging particularly where the objective (as under the Forestry Commission Community Woodland Scheme)' is simply to ensure that the local community has access to a woodland recreation site within five miles of the community centre.

iii. Willingnessto acceptcompensation The majority of interviewees (14 farms; 74%) stated a sum which they would be into land for in of agriculture and public out allocating willing to accept annual compensation farms initially included 7 This (WTApa). those who rejected the of access woodland if indicate latter (58%). 7bis the that, to seems price was result allocation such of principle is However, there farms agriculture. of conventional out a move would consider right, such is C20,000/acre bid"' 'protest this only at; not subsample amongst which noticeable one very larger 30 deviations than the times the than 150 and above mean more than standard more farm be likely is but bid, to to the of equal annual highest magnitude entire net also next in had feasible is It this mind a discounted total net present value that income. respondent be Howin reasonable. a response would for case such the which project, the entirety of sum feel the same we magnitude, within even answers ever, given that no other respondentsgave likely. is strategy seems much more that such an explanation unlikely and a protest Excluding this one outlier, the meanstatedWTApa is; E250/acre(E617/ha). Restricting into initially they to the allocate willing were an area which stated those to who the sample to the of non-allocators has support validity this adding result, upon effect no scheme bids. being (and as valid the thereby entire sample)" responses Modelling WTApa

Analysis of responsesshowedthat statedcompensationlevels were strongly related further No farm. levels the significant the size of overall both and profit existing to is WTApa fitting identified best the of given model and regression were variables explanatory in equation,4.15:

discussionof grantschemesin chapter6. "'The authordislikesthe generalapplicationof this term to anyonewho doesnot give an expectedanswer However, this WTA) (WT? particularrespondent question. mustsatisfyall relevantrequirements bidding or to a 'protester'. of an archetypal `Excluding the single'protest' bid. 6ISee

4.59

WTApa

94.04 + 1.48 PROFIT 1.93 ACRES (1.81) (4.04) (-3.37)

(4.15)

R2 = 69.9% R2 (adj) = 63.2% n= 13 10.43 (p = 0.005) RegressionF= where

Farmers required compensation(f/acre) for entering the woodland = scheme PROFIT = Level of profit underexisting agriculture(4/acre) ACRES = The numberof acreswhich the farm is preparedto allocateinto the woodlandscheme. WTApa

The model presented in equation 4.15 fits the data well and reports logical dependent between Farms higher levels the and explanatory variables. with profit relationships from existing activities demand higher levels of compensation for entering the woodland scheme. Furthermore those who are only willing to consider small scale planting require higher per-acre payments. This implies, logically, that large scale plantations, which from benefit economies of scale, are considered viable alternatives at a will presumably lower per-acre subsidy Tatethan small scale woodlands. relatively The areaof land which farmerswere preparedto allocate into woodland was positively levels farm thence to acre so that a significant correlation size and per profit to related overall (r 0.359). Stepwise ACRES indicated PROFIT has between that analysis this and = exists in increase PROFIT We the t-value the coefficient and on variable. significant a caused in faith in the too precise coefficient place much estimates given therefore equation cannot 4.15. However, the observed multicollinearity is not strong enough to make such estimates invalid, rather they should be treated as having wide confidence intervals. The degree of explanation of the WTA bid curve is not affected by collinearity between explanatory variables. Even allowing for the small sample size, the degree of fit is OE for CV high WTA this an study, particularly as survey employed a question. exceptionally We can conclude that farmers' responseswere highly logically consistent and accord with funding This WTA to contrary theory. runs most studies and we consider reasons economic be so subsequently. this may why

4.60

4.3.2.5: Wantage CV WTP/WTA study: discussion and conclusions

i. Theoretical welfare measures

This studyhasaskedtwo separatequestions.Firstly, householders wereinterviewed Both WT? the to their provision of a welfare gain. ensure per-annumand per-visit regarding from in here. Values formats tested theory, estimate an exercise such should, were payment the compensatingsurplus measureof welfare gain. Secondly,farmers were askedto statethe in for forgoing (per WTA annum) compensation existing agricultural they were amount in latter This favour in exercise provides, of open-accessrecreational woodland. production loss. discussion Before the of surplus measure of compensating welfare theory, measuresof based these simple comparison analyses, we present a upon various the relative validity of implied by WTA WTP these results. sums and the aggregate ii.

AgUegate values

Aggregation of the household WTP measures Householderswere askedto stateWT? for a 100 acre block of recreationalwoodland. WTP (including format those who refused to pay a simple mean The annual question elicited has 11,495, Wantage The household. L9.94 town an adult population of of per as zeros) of derive (so household bound to if size as on a estimate take an extreme upper we so, even imply 2.57 (CSO, 1991)63 household WTP) this bound some would of lower estimate on E44,450 imply WTP in in Wantage turn households of per an aggregate 4,473 which would for the woodland. annurn

Turning to consider our per-visit measureof WTP, we elicited a WTP of EO.82 per

The including (again to estimated number those as mean pay zeros). refused who visit adult implying just 15 total (including those under annurn a per visit) was not who would of visits Grossing E12.29 fee across all adults"implies up expenditure of per adult. entrance annual fees E141,252. WTP entrance of a total annual

figure refers to average UK household size rather than the average number of adults per household. increase this would our estimate of household WTP. i. e. we have chosen a conservative, latter if the were used lower bound assumption. "Note that we have already accounted for non-visitors in the annual per-adult visit rate. "This

4.61

Aggregation of thefarmers WTA compensationmeasure The farm survey estimated a mean WTA compensationof E250/acrep.a. Given that our household survey scenarioelicited WTP for a 100 acre site, our estimatedWTA for such

E25,000 is per annum. a site Comparison of WTP and WTA measures Either measureof WT? exceedsour estimate of WTA to a considerabledegree. In the case of our annual format we have a simple'5 benerit/cost ratio of 1.78 whilst the

entrance fee format yields a ratio of 5.65. Such results point strongly in favour of the setting-up of such schemes. However, we WTP Another to the to approach sums. cautious a way of examining these is prefer retain to consider the minimum number of payments needed to meet the required aggregate format Using level. household the per-annum and our above estimate of size compensation implies that some 2,515 households(i.e. 56% of all those in Wantage) would need to pay the E9.94 mean WTP for the scheme to break even. Alternatively all households in Wantage for break E5.59 Using have the to the per-visit mean to scheme pa again even. pay would implies that 30,487 individual visits per annum would be required to pay for the forest, i. e. in Wantage 2.65 for individual to the site to would need make paying visits per annum each break even.

iii.

Discussion

At first glance this study appearsto have been a successand seemsto hold out the CV decision to the studies relatively application of small-scale wider making of possibility discrepancy between household However, the the and particularly relationship problems. is Our WTP disturbing. discussion bid for these somewhat of curves per-visit and annual format that to the answers per-visit questions representednot true WTP suggested measures "price" influenced by but a rather social norm expectationsof what respondents valuations, felt was reasonableto pay for a forest visit. Conversely we argued that answersto -theannual format question were, at least in some way, related to respondentstrue valuations. How then "aggregate "aggregate exceed value"? price" can `5'rheterm 'simple' refershereto the fact that this studyrepresentsonly a partial cost-bencritanalysisof sucha scheme.

4.62

One explanation of this discrepancyarisesfrom noting that we have reasonto believe both that our annual format WTP measuremay be downwardly biased by elicitation effects and that our per-visit measuremay be upwardly biased by a number of factors. Regarding the annual format measure,our elicitation effect studies (Bateman et al., 1993) indicate that an open-ended (OE) WTP question format (as used in all the Wantage experiments) will produce an estimatedmeanWTP significantly below that elicited from a dichotomous choice approach". Whilst we stress that the dichotomous choice format need not, per se, be 'true' WTP", the conclusion of this work was that OE formats of an estimate producing bound lower best, estimates of WTP. We have compounded this in our at produce, by further 'aggregate lower bound assumptions regarding adopting value' calculation of household size. In short 'true' WTP could lie some way above our per-annum estimate. Turning to the 'aggregate price' derived from our per-visit measure, a number of Firstly, household be our aggregation assumptions noted. regarding composition should points for lower-bound 'aggregate as our as aggressively value' estimate. Secondly,we have are not some reservationsregarding estimatedvisit rate and note that the adoption of the 5% trimmed in 22% fall in 'aggregate More for result a this price'. severereductions would variable mean in mean visit rate (which averagesacross the entire study population) are quite feasible further in "aggregate in Thirdly, reductions our corresponding price" estimate. as resulting (see 3), it is in discussed have that chapter probable elsewhere answering this question we for a social norm responseregarding a socially appropriateentrance searching are respondents fee. Considerations in forming such a responseare likely to include experience of other destinations fees to questions regarding other as responses recreation show, which, entrance includes many with significant fees. Fourthly, the rounding effect commentedupon earlier has a far greater relative impact upon answers to the per-visit question than the annual for Thus, example, many respondents said that they would pay "one question. payment Multiplying by leads through this to an predicted visits visit. rounding per often pound" in fee being that entrance payments given above responseto the annual estimate of annual WTP question. Finally, as an extension to this, it may be that the spreadingof paymentsvia is fee lump inherent in attractive to the when relatively compared sum payment an entrance

"See previous discussion of elicitation effects. 17Sucha strong conclusion is implied by Arrow et al. (1993) in their preferencefor dichotomous choice over open-ended approaches.

4.63

the annual format question. In conclusion, the disparity between 'aggregatevalue' and 'aggregateprice' may not be a problem although the above discussion does highlight the need to consider these interval. a confidence measuresas point estimateswithin wide Turning to considerfarmers' WTA responses,the most striking featureof this analysis is the comparatively very high explanatory power of the WTA bid function. This result from WTA have findings many previous studies", where respondents exhibited contradicts We believe in fact difficulty that the that such questions. our answering result reflects great UK farmers are well accustomedto making decisionsregarding schemesand products which decisions These levels differing are made with respectto the opportucompensation. of entail factor in function. Other bid WTA forgone a very well reflected our alternatives, nity cost of individuals have decisions interviewed have who no experience compensation of and studies in WTA answering uncertainty questions. extreme consequently exhibit Finally, even taking into consideration the various actual and potential problems with

households' WTP (by aggregate of whatever measure)over the the excess clear this study, WTA compensationamountsstatedby farmers does indicate that the implementation of such in in hypothesised the the generation questionnaire scenario may well result that a schemeas benefit!. social net of a significant 4.3.3: THETFORD 2 CV/TC STUDY Between 26th March and 25th April, 1993,351 parties of visitors to Lynford Stag, Thetford Forest, were interviewed in an on-site survey. Data was collected for both a CV and ITC study" with the latter eliciting sufficient data to employ GIS route analysis of travel distance and travel time'.

Many of the findings from our earlier studies influenced the

design of theseexperimentswhich, we feel, allow for a significantly improved and more degree of analysis. sophisticated "See review in Mitchell and Carson(1989). 69Amorecertainstatement concerningthetotal netbenefitsof sucha schemecanbe madeif we assumethat statements.This is feasible farmershaveincorporateddirect afforestationcostsinto their WTA compensation in fact that grants respectof many of thesecostsare available,suchan assumptiondoesnot the and, given appeartoo strong. 71Further detailsof thesestudiesand thejoint surveyquestionnaire are given in appendix2. 71ThiSdata was also used to estimatean 'arrivals function' (detailedin chapter5) which combines information from the surveywith detailsregardingpopulationdistributionand road networkavailability and to of visitors the otherspecifiedsites. number to predict quality

4.64

The Lynford Stag site was chosenprimarily for the transferability of its recreational 7' few While there are a other minor attractions, the main activity of the site is attributes. This meansthat many of the attribute related measuresof walking. recreational open-access, our analysis may be transferableto other sites. 4.3.3.1: Thetford 2: The CV studV73 i.

Study focii

In coming to this study we felt that our previous work, together with our benefit had studies, provided us with a good grasp of the range of of reviewed transfer analysis CV UK What from derived being typical study of recreation. a woodland we valuations here investigate to theoretically the extent which reasonablere-specifications was to wished impacted WTP In design CV upon response. particular we wished to address questionnaire of

two issues: (i)

The mental accounts question: In chapter 2 we discussed the extent to which individuals do, or do not, consider other demands upon disposable income when answering WT? questions. Payment scenario effects: In the Thetford I study separategroups were presented The Wantage experiment scenarios. or per payment annum visit with either per In first to the then the all scenario respondents. the per visit annum, and per presented Thetford 2 CV study we set out to see whether answers to these questions were to Specifically design. instrument to the we wanted to some extent endogenous investigate the possibility of an ordering effect, i. e. does the answer to one question depend upon prior responses?If so, to what extent can the inclusion, exclusion or reordering of questions affect responses? These two potentially additive or interactive effects were investigated through a split-

into divided in design two groups each of which was respondents were which study sample further divided into two subgroupsas follows:

"The site alsohasa carpark,an informationboardgiving detailsof walksat the site,a few picnic benches frame, However, toilcts. and swing wooden climbing and some our surveyconfirms barbecue child's a sites, and is recreationalwalking. far by activity the major that "This studywill be publishedas BatemanandLangford(forthcomingb).

4.65

Group B:

Prior to any WTP question,respondentswere askedto calculate and statetheir annual recreational budget.

Group NB:

No budget question asked prior to any WTP response.

Subgroup 1:

AIM per annum (tax) asked prior to WTP per visit (fee) question.

Subgroup 2:

WTP per visit (fee) asked prior to WTP per annum (tax) question.

We therefore have four subgroupseachof which provides both a tax (per annum) and WTP fee (per response. visit) a Following the findings of our previous research,an open-endedelicitation method was deriving WTP In to throughout conservative approach as a responses. addition to the used WTP questionsthe survey also elicited information regarding all relevant visit, socioeconomic for interview necessary subsequentvalidity analysis. condition variables and ii.

The

ayment i3rinciple question

Prior to the WT? (and budget) questions,respondentswere askedwhether or not they This 'payment included because to at all. principle' question anything was pay were willing did feel inhibited from interviewees felt to not wish pay might that who stating such a we directly how if In they to they asked much were were willing pay. such a casewe response felt that some of these respondents might state some non-zero sum because they felt inhibited their true, admitting about zero, willingness to pay. otherwise or embarrassed, 73% of respondentsstatedthat they were preparedto pay at least someamount for the

is lower Thetford Forest. This for facilities than at somewhat provided our study recreational increased (85% Broads Norfolk the and may reflect acceptance) number of sites which of the

Broadland. for Thetford to the compared almost unique of nature might substitute The determinants of the decision to respond positively to the payment principle investigated indicated (statistically This just through chi-square analysis. weak were question " income, duration, distance travel insignificant) positive relationships with and visit and a interest in relation with wargaming and other structured recreational similarly weak negative found for Significant three and positive relationships were activity groups: those pursuits. less 6.52)'5; Q2 than of two take walks short the miles those who often at site = who often 74AII factors which support the findings of our Thetford I study. 75CritiCalXI values with ldf = 3.84 (cc= 5%); = 6.64 (ot = 1%).

4.66

for (e the relaxation/enjoying scenery site use = 11.95); and those who sometimesor often enjoy nature watching at the site ()? = 8.13). Thesefactors are clearly interrelated and further analysis confirmed that the majority of those who statedthat they often relaxed and enjoyed the scenery at the site also stated that they sometimesor often enjoyed nature watching and Clearly factors be logit for then could not such entered separately a within short walks. went in (as Bateman 1992). Consequently responses principle used et al., an payment of model by created whose significance was maximised grouping together all was variable amalgam factors. least (which The two these three at of resultant variable we label those who exhibited in highly both be A) VISITOR to significant explaining responsesto the payment proved as in bid 16.54) (e and subsequent curve analysis. = principle question Reasonsfor both positive and negative responsesto the payment principle question direct Those investigated questioning of respondents. who refused to pay via were explicitly detailing a show-card with various set responsesregarding anything at all were presented best described for to their reason asked state which category and response refusaI76 reasons for refusal. Table 4.15 details results from this analysis.

Table 4.15: Respondentsstatedreason for refusing the principle of payment 1ý9

I 2 3 4 5 6 7 8 9

Reasonfor refusal'

No. of respondents

% of all refusals

% of total sample

Cannot afford to pay Does not like site Prefers natural state Refusesto value site Someoneelse should pay Pays too much tax already Rejects entrancefees Other Not stated

2 0 10 11 6 24 7 12 24

2.1 0.0 10.4 11.5 6.3 25.0 7.3 12.5 25.0

0.6 0.0 2.8 3.1 1.7 6.8 2.0 3.4 6.8

Totals

96

100.0

27.3

Notes: 1. Full details. show cardsand questionnairereproducedin appendix2 2. TOW samplesize = 351 Numbersrounded to one decimal place.

based W'Show were on our prior woodlandstudies. categories card and

4.67

Considering the reasons given for refusing to pay for the site we can see that the is issue one of economic constraint (inability to pay, reasons 1; and current specified major

6) in demands, reason although someways this might reflect a rejectionof the expenditure tax and fee vehicles (reasons6 and 7). Reason 5 (someoneelse, such as the government, free-rider The individuals in is the this response. small extreme number of should pay) but larger is to may nevertheless point a group of respondentswho, category encouraging Our WTP. Norfolk Broads their total to something, still study pay understate while prepared (Bateman et al., 1993) indicated that OE elicitation methodsmay suffer from a certain degree here low in This the although numbers may apply relatively refusal of understatement. in instance be in 5 too this this of that much a problem and will may not category suggest WTP. in total of predictions conservative any case result The one category which would suggestthat our study is fundamentally invalid is that for individuals who refuse to value the site (response4). Reasonsfor such a responsemay be diverse. However, even if we assumed that all such respondents fundamentally rejected

individuals in the this category suggests of small number the principle of monetaryevaluation, here. have do a problem not that we Respondentswho acceptedthe principle of paying at least someamount were similarly from for doing. Table 4.16 details this analysis. results so their reasons asked Table 4.16: Respondentsstatedreason for accepting the principle of payment Reasonfor acceptance'

No.

1 2 3 4 5 6 7 8 9 10

No. of respondents

% of all acceptances

% of total sample'

28 8 8 5 28 3 25 92 3 55

11.0 3.1 3.1 2.0 11.0 1.2 9.8 36.1 1.2 21.6

8.0 2.3 2.3 1.4 8.0 0.9 7.1 26.2 0.9 15.7

255

100.0

72.7

Reasonableamount to pay Similar price to equivalent sites Lives close to site Visits site often Keen on countryside Keen on forests Keen on wildlife/environment Preservefor future Other Not stated Totals

Notes: 1. Full details in show cardsreproducedwith questionnairein appendix2. 2. Total samplesize = 351 Numbersrounded to one decimal place.

4.68

Interpretation of some of the responsesin table 4.16 must be somewhat loose as several categories are overlapping (e.g. 5,6 and 7). However, the lack of respondentsin category 2 is encouraging7. Perhapsthe most interesting observation is the large number of respondents stating that their prime motivation in agreeing to pay something was to for future areas generations. Tle wording of this category was phrasedso as such preserve to separatethis from option value, although it is always possible that some respondentsmay have been influenced by such considerations. Neverthelessthe prime rationale behind such be bequest In to appear value. other papers we have been somewhat would a response in CV statements of altruism such studies (Bateman, 1992). However, the of suspicious feeling is interesting this within sample apparent remarkable and such raises an of strength his her bid. how WTP While it the respondent views or own seems regarding question (entrance fee) bids WTP that would relate to the pure use value of a visit, per-visit probable for 4.16 that table many people, responsesto per annum WTP (tax) suggest the results of be likely (bequest to a mixture of use plus non-use quite and existence) value. are questions In a fully informed, rational expectationsmodel of respondentbehaviour we would therefore be able to use the difference between the annual equivalent of per visit responseand stated Unfortunately WTP to non-use value. as equal we suspectthat problems regarding per annum future discounting expectations,measurementerror within the individual and (probably of the Nevertheless important) effects, may confound such a calculation. the payment vehicle most in 4.16 (and desirability table the of successful estimation of expressed strength of opinion is for future this that a worthwhile avenue suggests research. non-use values) iii.

Mean WTP and confidence intervals

(unbeknownto themselves) At the startof eachinterview,respondents wererandomly four described the to subgroups abovesuchthat roughly one-quarterof the of allocated one in However, these numbers were then randomly disturbed subgroup. each was total sample by those respondentswho stated that they were not willing to pay anything at all. As the WTP in the question preceded consequent any question, reduction principle payment is does invalidate in WTP and random not or any contaminate sizes way responses. subsample However, it does mean that we need to subsequentlyadjust for the differing subsamplesizes 77pardcularly asthis appearsso nearthetop of the showcardlist of options,i.e. it mightbe inflatedby any list-biaseffect (Oppenheim,1966).

4.69

by redistributing these zero bids equally amongst all subgroups". In the following subsectionswe present results from the analysis of, firstly,

when calculating mean WrP

WT? and, secondly, per visit responses. per annum

a)

Per annum (tax) WTP responses Table 4.17 details mean WTP per annum (via taxes) and 95% confidence intervals (in

brackets)for eachof our four subsamples. Table 4.17:

Mean WTP (tax) per annurn(f) and 95% confidenceintervals (in brackets)for (including subgroup payment principle refusals as zeros) each Payment ordering scenario

Budget question (tax then fee)

2 (fee then tax)

NB (not asked)

12.55 (8.11 - 16.99)

7.62 (2.87 - 12.37)

B (asked)

32.60 (21.76 - 43.43)

16.37 (11.19 - 21.55)

Consideration of the results of table 4.17 indicates that the inclusion or exclusion of

in budget the ordering of payment vehicle changes and/or question, the recreational impacts WV'9. inclusion The in and major upon stated of presentation, results consistent factor 2.60 for (tax) by WTP budget of annual a vehicle ordering question raised mean the 2.15 for 2 (fee by factor fee) (tax I then tax). then of vehicle ordering scenario and a scenario Given the magnitude of change this clearly raises major questions for CV research. The direction of change is also interesting. Most commentators (Mitchell and Carson, 1989;

Willis andGarrod,1993)discusscasesin which,a prior!, we would expectthat respondents' budget expenditure upon recreation and consequent constraints would annual of consideration lead to a reduction in statedWTP comparedto statementsmade without such consideration.

However,herewe observea very strongoppositeeffect wherebyrespondents who areasked higher WTP annual expenditure state signiricantly sums than those their present to calculate

WFurtherdetails in appendix 2. 79Notethat there is some overlap of confidence intervals for changesbetween certain subgroups(although impacts do have beendetected Nevertheless here. In from Ole to strong for appear case every mean others). not interval its horizontal just (i. falls the confidence of outside vertical or e. neighbour where we vary one subgroup have factors both contrary effects, tends to canccl each other out). factor; varying the as one

4.70

budget the question. prior asked not Why has this effect occurred? It seemsto us that two interpretationsare possible, one deriving from economic theory and the other influenced by psychological literature. An forced be to overtly consider their annual that respondents might argument economic

for find budget this that, a significantportion of their total accounts average, on recreational Certainly than they without such consideration. realised more annual expenditure, perhaps insignificant. The budget (E227.30) budgets mean was not were stated recreational in income) described (with distribution by this the as skewednature of considerably affected Nevertheless, the median value of E120 shows that most respondents had

table 4.18.

line budgets. Following the this of consideration of argument, upon recreational considerable in individuals' importance gave such respondents preference sets, recreation of apparent higher WTP sums than would otherwise have been stated. Table 4.18:

Variable

Budget income rne

Descriptive statistics for respondents' annual recreational budget and gross household income (f pa)

No. asked question

167 251 351

No

mean

answering question 152 349

227.30 18,247

120.00 17.500

SL dev.

trimmed mean

median

1

169.40 17,524

Q3

Q1

MAX.

95% ci lower

1

345.50 10,923

70.00 12.500

1

250.00 25.000

2500.00 55,000

171.90 17.106

upper_ 282.60 19.388

If we acceptthe economic argumentthen a supplementaryquestion arisesas to which is line This budget (with, correct. of reasoning the question) prior WTP measure or without, following formulated the consideration of available that answers to suggest seem would " budgets will be less susceptible to mental accounting problems and therefore preferable. interpretations by of our to that provided psychological However, this conclusion runs counter WTP is high (which budget to compared the Here relatively annual of the calculation results. (1982) Kahneman WTP for et al. statements. subsequent for the forest) acts as an anchor inexperienced individuals is likely and to are where occur most effect an that such suggest have individuals do Certainly in forming not their response. face considerable uncertainty '. to to then? prices reacting as opposed setting of experience much "Such a conclusion would imply that the bulk of the CV literature, which has not incorporated mental flawed. is account questions, slour own work (Bateman et al., 1993) suggeststhat respondentsexhibit greater uncertainty in answering DC WTP Use OE format than questions. of an may therefore exacerbatethis OE (as per this experiment) be ic. this exacerbation elicitation of should reasonably constant across the extent respondents although problem format In future findings. DC to this these work aim a we would experiment do repeat within not explain effects formats OE induce. level which may uncertainty of to reduce the

4.71

ClearlY this finding gives us pause for thought regarding the degree to which WTP responsesmay be manipulated by small and apparently defensible changesin questionnaire design. The responsivenessof stated WTP to the inclusion of the budget question is for future CV of significant concern a matter and studies. remarkable Turning to consider the impact of changing the order of payment questionsupon per annurnresponses:for those subgroupsnot given the prior budget question, asking for per visit WTP before the per annurn question lowered the latter to just 60.7% of stated annual WTP by For those subgroups who were given a prior a per visit question. not preceded when budget question, this disparity increasedso that annual WTP precededby a per visit question is just 50.2% of annual WTP not so preceded. Here an economic justification might be that such respondentswere taking prior per-visit payments and extrapolating them to produce a imply However, bids this that would per annum made prior to per-visit sud'. per-annum bids were in error. Here then the psychological argument that the relatively small per-visit WTP sums have anchored subsequent per annum statements, seems the most logical explanation of theseresults". Consideration of the rates of impact of the mental account (budget) and ordering (payment vehicle) effects suggestssome interaction. It appearsthat the use of a per visit diminishes impact WTP WTPpa the the to per annum response upon of prior question inserting the budget question. This is to be expected as inclusion of the per-visit question inclusion Furthermore budget responses. of a annum prior the of per question range restricts

WTPpabid andthe responseto the increasesthe disparitybetweenan otherwiseunpreceded by inclusion Here budget the the a per visit question. preceded of when question same WTP the subsequent annum of per responses. range question opens up Per visit (fee) WTP responses The mental account and payment vehicle ordering effects upon per visit WTP bids

4.19 Table details WTP fees for mean eachsubgroupand via per visit analysed. also were 95% confidence intervals. Here, as before, all payment principle refusals are included as between equally subgroups. zeros allocated

"Factors suchas discounting,uncertaintyand risk meanthat we would not expecta simplerelationship betweenper visit and per annumWTP. "Which in turn can only enhancethe anchoringinterpretationof the budgeteffect.

4.72

Table 4.19:

Mean per visit (fee) WTP (E) and 95% confidence intervals (in brackets) for each subgroup (including payment principle refusals as zeros) Payment ordering scenario

Budget question

L

1 (tax then fee)

2 (fee then tax)

NB (not asked)

0.45 (0.35 - 0.55)

0.20 (0.11 - 0.29)

B (asked)

0.78 (0.53 - 1.03)

0.46 (0.30 - 0.62)

Considering table 4.19 we can see that the design effects detectedin the per annum in been have because fee the repeated studies, per-visit although responseswere experiments bids, line is the ordering of than annum effects reversed. Here the prefixing of per smaller increases by bids. WTP Similarly, and WTP questions per-visit per annum questions per visit budgets inclusion leads before, of question to significant the a prior regarding recreation as increasesin subsequentper visit WTP responses.The economic and psychological arguments before feel influence that the these although are as effects we of 'social norm' surrounding (see 3) have diminished intensity the responses chapter may per visit upon of these pressures in This factor is the to those per-annum experiments. additional observed compared effects in 4.19 demonstrated the table contrast per-visit means when we with their per most clearly

both left hand 4.17. In lower in WT? table the cases cell represents when equivalents annum both positive (budget and ordering) effects are in operation resulting in the most extreme WTP responses. In the per-annumexperimentsthe positive budgeteffect (vertical shift to this by factor 2.6 WTP (horizontal the a of while positive ordering effect shift cell) raises mean factor 2.0. by In WTP increases the per visit experiments the of mean a to this cell) We 1.7. increases in both factor are casesroughly would argue that this relative equivalent diminishment of extreme effects by the per-visit vehicle are due to the 'social norm' pressures into 'fair' take account of a who notions socially entrance price exerted upon respondents (and/or experience of fees elsewhere)when formulating their per-visit WTP response.

iv.

Validation*bid function analysis Validation of our results was carried gut as for previous studieswith the main

investigation factors deten-nining WTP being the statistical of upon responses. emphasis

4.73

a)

Per-annum WTP responses

Examination of the most appropriate functional form was conducted as before and

againconcludedthat a naturallog specificationof the dependentvariablefitted the databest. Following this, considerationswitchedto identificationof significantexplanatoryvariables linear The and stepwise regression. one-wayanalysesof of variance analyses via one-way interesting Weakly ((X highlighted >0.05) a of significant number relationships. variance factors includeda negativerelationbetweenWT? and being a first time visitor or member day the of positive relation number visits to the site club and a with of a sports/Wi or other follows in (cc (figures brackets Strongly <0.05) variables as significant were are annually". from one-way analysesof variance): p values ORDER,,,,

1 if respondenthad been asked a per-visit (fee) question prior to per annum (tax) WTP; =0 otherwise (p = 0.000)

BUDGET

I if respondenthad beenaskedto stateannualrecreationalbudget prior to per annum (tax) WTP; =0 otherwise (p = 0.000)

NOTCAR

I if visitor did not arrive by car-, =0 otherwise (p = 0.003)

SUPERB

I if respondentrated scenery at the site as superb; =0 otherwise

RSPB

(p = 0.028) I if respondentwas a member of the Royal Society for the Protection 0.021) (p Birds; otherwise = of =0

TRUST

I if respondentwas a member of a wildlife trust; =0 otherwise (p = 0.040)

LOWINCOME

I if respondent'shouseholdincomewas below 0,000 per annum; 0 otherwise(p = 0.048)

All theseexplanatoryvariableshadsignificantlypositiveeffectsuponper-annumWTP LOWINCOME ORDEk,, variables which were negatively the and of the exception with bestfitting regressionmodel of per annumWTP. A GLM details 4.20 Table the related. between ORDER,.., interaction BUDGET the term test to and an variables. analysiswasused (p found be to not significant = 0.375). However, this was

940ther even weaker but correctly signed factors include: incomC(+ve); sunniness(+vc);tcmpcraturc(+vc); journey(-ve); length the of time of enjoyment on sitc(+vc); and whether respondentwas a trips(. ve); multi-site tax-payer(+vc).

4.74

Table 4.20:

Best fitting bid function for per annum WTP. Dependent variable = natural log of per annum WTP (In WTPpa)

variable Constant ORDER,.,, BUDGET NOTCAR SUPERB

coeff.

st. dev.

t-ratio

p

1.2573

0.1157 0.1359 0.1370 0.3259 0.2033

10.86

0.000 0.000 0.000 0.000 0.002

-0.6024 1.4668 1.1772 0.6226 W=

-4.43 10.71 3.61 3.06

2(adj)

31.8% R 31.0% = s=1.267 Regression F= 40.18 p=0.000 Variables as defined in text. Tests confirmed no multicollinearity problems. The overall predictive power of our best fit model is, by CV standards,quite good for is is What OE the confirmation of the strength of the vehicle ordering of concern study. an individuals' inclusion The responses. of a prior budget effects upon and mental accounting increases WTP subsequent per-annum significantly responseswhereasa prior very question Of WTP to the two the subsequent responses. acts reduce per-annum question per-visit budget effect is the greater both in terms of absolute magnitude and statistical significance, but both factors are very clearly at work here.

b)

Per visit WTP responses As before validity testing focussed upon estimation of a bid function for WTP

due lack investigations in Initial to that, a relative of showed variation per-visit responses. WTp responses, a linear dependent variable fitted the data better than a log-linear interpret dominated We to that this per-visit questions are as a sign responses specification. by our 'social-norm' factors rather than by standardsocioeconomicand visit characteristics. Simple data analysis techniques were used to identify potential explanatory interesting but WTP A quadratic weak statistically relationships of with number variables". initially WT? found distance (particularly Per to rise was with visit noticeable were noted. for be linked The this to 15 appear to the would reason purposefulnessof miles). about at have Visitors travel way specifically to considerable the some who visit site clearly the visit. its However, for distance becomes long, as attractions. particularly preference a strong become more by chance than design, i.e. such very long falls visits and purposefulness "Techniquesincludehistogramsandplots,calculationof correlationcoefficientsand one-wayanalysesof variance.

4.75

distancevisitors generally happenupon the site by accidentand stopjust to break the journey, their real destination being elsewhere. This interpretationis supportedby the positive relation in day's Thetford WTP terms their overall enjoyment and negative to of of rating of visitors' findings day. These to that travel sites other and of visits relationships with enjoyment in TC importance travel costs our subsequent enjoyment-adjusted the of using underscore for the site. consumers surplus overestimate we would the site, without which study of A second quadratic relationship was found with the number of day visits per annum. Here WTP is initially relatively small at low numbersof annualvisits. This is a function both few because those trips to and who make referred above, passers-by of the meanderersand increases As have trips because the do of available alternatives. number they many so may like but do have Here WTP. the does site not initially, respondentswho we per-visit so, (which distance because high trip availability substitute and of numbers of visits make very falls back WTP high However, per again. of visits, visit numbers be at very collinear). will few For have live substitutes. to the available and may site Such respondentsprobably close fee to they to cost are which considerable annual translate a would per-visit a them in WTP Such sums which rise annual reflected are observations resistant. understandably declining but rate. at an eventually rates visitation with A third quadratic, identified to some degree in all our empirical studies, is with age. (both lower bids WTP than the and per-annurn per-visit) to tend give old and Both the young distributions. income be likely to reflecting middle-aged, a result most A number of simple but statistically weak linear relationships were also identified. journey day's home having be found the to started at to negatively related V,rM per visit was (not holiday from to the of catered activities wargaming principle and address, a than rather dogwalkers live (for dogwalking tend to above; given reasons for at the survey site) and found Weak with picnicking and positive relations were locally and visit often). income. and with objectives, as principle visit scenery relaxing/enjoying identified follows (a <0.05) were as variables A number of statistically significant from brackets in one-way analysesof variance): p are values (numbers

ORDER&,

I if respondenthad beenaskeda per-annum(tax) question prior to per-

BUDGET

(p 0.033) (fee) WTP; otherwise = visit -_O I if respondenthad been askedto state annualrecreationalbudget prior

CAMPOFr

to per-visit (fee) WTP; =0 otherwise (p

0.024)

1 if respondentoften camps in the area;

0 otherwise (p = 0.007)

4.76

SUPERB

I if respondentrated sceneryat the site as superb; =0 otherwise (p 0.014)

STAY4

I if respondentspendsat least 4 hours on site per visit (p = 0.035) I if respondentstated that the prime reason for visiting the site was

BUSINESS

` business to a connected meeting; GREEN

=0 otherwise (p = 0.000) I if respondent is a member of at least one of the following:

any

wildlife trust, the National Trust, the Broads Society, Friends of the Earth, Greenpeace; =0 otherwise (p = 0.004) Unlike the per-annurnexperiment,the ordering variable (ORDERJ is now positively indicating (per-visit) WTP, that the asking of a prior per-annum question to stated related WTP bid. The WTP subsequent per-visit relationship of with BUDGET respondents raised is (as before) positive, as is that with CAMPOFT, SUPERB, STAY4 and BUSINESS. Ibis is all as expected. However, against our first expectations,the variable GREEN proved to be strongly negatively related to per-visit (fee) WTP although membershipof such groupswas WTI"'. to annual related positively

It seems that the members of such groups strongly

implicit in fee is It the the the of site the open-access nature of to vehicle. ending object interesting to note that the survey took place in the middle of a well publicised, year-long fears had Commission Forestry the raised of which wholesale privatisation of the of review

This loss fees by to open-access strong of current rights. objection consequent and the estate likely be (who this to the aware of review") may well reflect most were such respondents a deeperprotestagainstthe prospectof privatisation". All the variables listed above are simple, two-level, dummies. The number of in for 45 level BUSINESS the cases except all variable or above was at any respondents

86rhisinformationwaselicitedfrom interviewm' commentswhenspecifyingthe'other' categoryin answer forest. for to the the coming main reason to a questionregarding "Interestingly membership of non-environmental groupssuchassportsclubswas(weakly)positivelyrelated WTP. Ibis forest to WTP related per-annum negatively makes sense as such and respondents visit to per-visit by less therefore expenditure would minimise on such and recreation payingper useratherthan Sitesrelatively via a flat annualrate. IsMost of thesegroups,includingeventhe normally sedateNationalTrust, had lobbiedhardagainstthe Stirling, (see 1994). possibility of privatisation 89Suchprotestsdo, arguably,causeproblemsfor the validity of our pcr-visit valuations. However,this is in our considerationof answersto the paymentprinciplequesdon.Furthermore,the examined to someextent by orderingand mentalaccountingeffectsare of a greatermagnitude. raised theoreticalproblems

4.77

which had the value 1 for just two interviewees but proved to be highly significant. Table 4.21 reports the best fitting bid function" which included both of our focus variables (ORDERfý. and BUDGET) and any other significant (cc < 0.05) explanatory

variables. Table 4.21:

Bid function for per-visit WTP responses

Explanatory variable

coeff.

st. dev.

t-ratio

P

0.4647 0.0865 0.1224

0.0617 0.0685 0.0679 0.0767 0.0984 0.4505

7.53 1.26 1.80

0.000 0.207 0.072 0.004 0.001 0.000'

Constant BUDGET ORDERf,,, GREEN CAMPOFF BUSINESS

-0.2198 0.3175 5.1676

0.6281

R' = 33.4%

-2.87 3.23 11.47

R' (adj) = 32.4%

Regression F= 34.42 (p = 0.000) Note:

1. As noted in Table 3.14, the p-valuereportedhereare thoseproducedby default by the statisticspackage. In this instancethe small number(2) of observationson the BUSINESSvariablemeansthat the sampledf Using df =2 reducesthe p-value slightly to 0.005. is somewhatunrepresentative.

Table 4.21 reveals several interesting characteristicsof the per-visit VvrTPresponses. The focus variables ORDERf,..and BUDGET, while exerting pressureupon bids, are not the in Indeed determinants responses. neither satisfy a 5% per-annum exhibited highly significant (given our previous Other are expected as variables explanatory test. significance (which BUSINESS With variable the exception ofthe only applies to two discussions). is factor We believe that this, determining the far by constant. the strongest responses), fit (for OE CV bid function), degree of an gives strong overall the good with combined highly determined by individuals' are that responses per-visit support to our argument for level 'social-norm' to charging of acceptable entry such a site. a of conception common both their true their experience tempering own valuations with Respondents are, we argue, fees (e. their at comparable sites) through and car-parking conceptions fee of a g. paying of is, for traditionally, level an open-access just what good. payment of socially

from this regression.For detailsseeappendix2. omitted were 90rwo unusualobscrvations

4.78

V.

The Thetford 2 CV studv: conclusions This study has raised as many questionsas it has provided answers. By analysing the

bids be by design WT? to can manipulated variations we have raised questions extent which is design to as which permutation preferable. The variations testedare, we feel, all justifiable. The introduction of a prior budgetquestioncan be justified on the grounds that this addresses the possibility of mental accounting error and indeed studies have adopted such an approach (Willis and Garrod, 1991; 1993). Furthermore,as several studieshave adoptedboth per-visit 1992; Whiteman (Bishop, Sinclair, 1994), the possibility of and measures and per-annum by is Our in interacting their ordering controlled study shows that a way worrying. these by inclusion halved be WTP the of a prior per-visit question or can almost mean per annum budget (with budget by doubled the question effect somewhatoutriding the a prior more than included). Tle fact both if both less that these are priors effects are effect payment ordering is is hardly if bids (and WTP this comforting as a result of pronounced upon per-visit for) a social-norm conditioning of such answers. evidence The implications of thesefindings for our research(and for the wider use of CV) may depend upon the perspectiveof the individual researcher.We have experiencedvery differing further Some have findings. from taken them the these as evidence to of colleagues reactions have Conversely CV others pointed out that, while results results. &sheersubjectivity' of increased decreased by larger factor, be i. halved doubled they or a be could not e. or could interval of valuation arises. confidence a creating the possibility of

we

We have some sympathy for both interpretations of these findings. Certainly when (Bateman WTP the into elicitation of method the varying et al., upon effect account take

in design the presentstudy are certain to widen any resultant observed 1993), then the effects format for have OE by In this study, we an ensured dvaluation envelope'. adopting effect, To adopt a further design lower-bound to elicitation effects. respect with a conservative, here be design the to studied might somewhat effects lower-bound assumption with regards dangerous, certainly the lowest mean WTP sums of E7.62 per annum and 0.20 per visit do is beset This thorny to a highly problem, with uncertainties which we conservative. appear later. return

4.79

4.3.3.2: Thetford

2: The ITC Study9l

Alongside the CV experiments,the Thetford 2 study undertook a travel cost analysis of visitors recreation use-valuefor Lynford Stag. Following the discussionsof chapter 2 we again used an individual rather than zonal (Clawson-Knetsch)approachto the TC. Here we discussion and of this study. Full results are given in appendix results provide summary of A2.4.2. Three researchobjectives were defined for this study: To examine the application of geographicalinformation systems(GIS) to travel cost GIS felt It that the spatial analysis capabilities of were of considerable was studies. s92. to potential value such studie

2.

To conduct a full sensitivity analysis across a range of time cost and travel cost has impact The costs of such clearly considerable valuation upon assumptions. but, in discussed 2, is there estimates as chapter surplus some subsequentconsumer debate regarding both the absolute value and methodological approach towards valuation of these cost elements.

3.

To assessthe impact and validity of using ordinary least squares(OLS) or truncated Chapter 2 (ML) OLS likelihood techniques. that estimation showed maximum in fail invalid that to the truncation they technically recognise of nonare approaches OLS have (see TC I Nevertheless techniques studies used appendix many visitors. investigation I ITC Thetford a comparative and appearedtimely. study) and our own

The survey All 351 parties interviewed in the Thetford 2 CV study were also asked travel cost details are as before". Therefore sampling questions. Respondents itself. to were asked state: trip the upon

Several survey questions focused

9,Dctails of this study are published as Batemanet al. (1996b). 92Furtherdetails regarding the GIS exercise are given in Bateman et al. (1995c). "The common CV/ITC questionnaireused in the Iletford 2 study is reproduced in appendix A2A. 3.

4.80

0)

Home address,and trip origin if different to this (e.g. if on holiday away from home); How they travelled to the site; The perceived travel time and cost;

(iv)

The number of other sites visited during the days trip;

(V)

The proportion of the whole days enjoyment attributable to time spent travelling; time spent at the survey site; time spent at other sites.

Perceived and GIS calculated travel distance and duration As stated, a prime objective of this study was to examine the potential application of GIS spatial analysis techniquesto the TC. It was decided that a simple test of effectiveness distance duration travel be of and perceptions to with those respondents' compare would it is immediately GIS. A-priori the not clear which of these calculated through use of individual If is then these statements visitors' should reflect we use superior. approaches highlight In decisions they take travel visitors particular speeds. will who routes and routing increase journey. distance/duration However, to the so as of enjoyment shortest which are not interviewees' description journey is both distance the that of upon reliance a problem with from is This likely liable be to to duration effects. suffer rounding are estimates and journeys be for the rounding error may where well relatively shorter proportionately worse large. The GIS approach addressesthis problem directly by producing accurateestimatesof is highly drawback detailed However, duration. this that, the approach of unless distance and have be logical to trip routing which made regarding itineraries assumptions elicited, are trip due Nevertheless, deviations take to to those routes unusual a site. who may not capture Sidaway, (e. Colenutt have forest 1973) UK and g. suggested recreation previous studies of highly deten-ninants time travel significant provide minimum of visit as such that variables based based ITC GIS those Comparison perceived costs results upon with of upon rate. interesting an exercise. seemed calculations therefore Calculations of GIS trip costs first required accurateinformation regarding trip origin. from (i) data question above the national grid referenceof trip origin was collected Using the Ordnance Survey Gazetteer Great Britain (Ordnance Survey, the by of located consulting

4.81

1987)'. Figure 4.8 illustrates trip origins for the entire sample in relation to the survey site. This shows clearly the importance of spatial factors in the determination of visits. Trip origins were concentrated around Thetford, with over 90% of visitors having set out from within 100 miles of the site. However, straight line distance is clearly a rather crude determinant of visits and one of the principle advantagesof adopting a GIS approachwas that it allowed us to account for both the distribution and quality of the available road network. Digital road network details were extracted from the Bartholomew 1:250,000 scale database for the UK. distinguishing motorways.

This source provides information on the quality and width of roads,

15 road categories ranging from minor, single-track country lanes to six-lane Computing time and space limitations made it impractical to assemble a road

Thetford for the covering origins area of entire visitors (this ranged from Edinburgh network in the north to Hampshire in the south). We therefore defined a study area to include the Norfolk, of counties Lincolnshire

Suffolk

and Cambridgeshire, together with adjoining districts in

92% EsseX95. This encompassed over of the visitor origins. and

Figure 4.9

illustrates the resulting digital road network. Supplemental data for visitors from outside this Automobile by Association's 'Auto Route' the use of package". obtained region was

The classification and quality of individual roads is defined in the Bartholomew's database. By applying differential road speedsto thesedetails, travel times can be calculated for discrete sections of road. From these,travel times can be calculated for the whole of the for detailing differing Data travel average speeds categoriesof road network. available road The from derived from of sources'. a road speed variety estimates this obtained were in 4.22. detailed table exercise are

"Like all our data from all surveys, the recording accuracy of this data was checked by double-punching datasets. and comparing resultant all completed questionnaires 9"The Bartholomew's road coverage is stored in map tiles (100 km squares)on the national grid. ne boundary defined subsequently and the appended clipped using study area tiles were as above. map relevant Undershoots (common in the Bartholomew's data) were locatedand correctedwheneverpossible. Further details in Bateman ct al. (19950. 9ý3ecauscof the rather general digital road network used by Auto Route, this packagewas not suitable for decisions such as those taken by the majority of respondentswho have trip origins micro-routing analysing relatively near the site. 97Sourccsinclude DOT(1992,1993); Gattrell and Naumann(1992); the Automobile Association's 'Autoroute' knowledge in Further the of details in routes the authors study area. are and given appendix A2A2 software; (1995c). Bateman al. et and

4.82

Figure 4.8: Trip origin for visitor sample: Lynford Stag, Thetford Forest

Key:

ýgmý L. ýw; I: 3750 000

ThetfordSite

5 km

D*7

Visitor Origin County Boundafies

4.83

Figure 4.9: Digital road network for the Thetford 2 ITC study

c) Study Site Key: FIV, Motorway A road, multi-lane 20

g.

I:

30

40

50

A road,single-lane

km

Broad CountyBoundaries

1 100 000

4.84

Table 4.22: Road speedestimates Road Type

Average Road Speed(mph) Rural

Urban

Minor Road

14

11

A-Road Primary Single Carriageway

45

25

A-Road Single Carriageway

32

18

B-Road (with passing places)

24

12

B-Road Single Carriageway

24

12

A-Road Dual Carriageway

50

25

B-Road Dual Carriageway

36

18

A-Road Primary Dual CarTiageway

54

28

Motorway

63

35

A-Road Single Carriageway (under construction)

32

18

B-Road Single Carriageway (under construction)

24

12

A-Road Dual Carriageway (under construction)

50

25

otorwa (under construction)

63

35

GIS calculated travel times were then obtained by extracting an individuals travel

for (derined digital from speed distance network and adjusting road on each segment road our from trip origin to site9g. the route by road type) of

As noted at the start of this section,visit cost estimatesbasedupon GIS calculated over thosebasedon visitors distanceand durationhave both advantages and disadvantages GIS 4.10 distance Figure the to plots ratio of stated calculated againstthe perception. latter. the of absolutevalue

"Details of the stepsand GIS commandsusedin this operationare given in appendixA2.4.2.

4.85

Figure 4.10:

Graph of the ratio of stated to GIS calculated distance against calculated distance

5.0ý 1

1

4.04

Ratio

of to

stated GIS calculated distance

3.0-

31

2.0- - 3-1 2.243 1 604-6-0.1 1.015:.. 0.- -02 6 43 1 -,.. 33

upper 95% Cl

2121 3-2

11 1-1-1-2

1-2

-stated

-1-1 2

333

21111 2111

lower 95% Cl

--------------------------------

0

= c alculated

300

200

100

GIS calculated

400

500

travel distance (miles)

integer (1-9).

number of observations as shown '0 ý 10 or more observations n= 350 (one missing value) boundary of rounding errors possible = ...............

Examining figure 4.10 shows that, on average, both distance measures coincide between deviation low larger distance The the measures at comparatively reasonably well. in from derives, rounding error statementsregarding mainly argue, we is as expected and in (dotted lines) a cone of observations which may fit into drawn have We journeys. short this category.

Support for such a line of reasoning is given by noting that, for these

distances below travel the state as many respondents roughly 'rounding-error' observations, is based impedance distance GIS As on a minimum the algorithm GIS calculation as above. below (unity) line be the those estimates equality respondent must time), subject (minimum due is For to form an error which we argue rounding. observationswithin error, of to some basis better for distance GIS than may provide a cost calculated estimates the this category,

does stateddistance. 4.86

As the majority of respondentsfall into this category this is an encouraging finding. However, figure 4.10 also shows us that for a few respondentscalculated distance is likely to be a poor estimate of true distance. Six extreme casesare identified all lying above the upper 95% confidence interval around the mean. All of theseare for stateddistancesof less than 100 miles and it seems most likely that these respondentsare 'meanderers'9,;those whose main objective is enjoyment of the journey rather than time spent on-site. For these individuals the advantagesof removing rounding problems are more than outweighed by the error induced by the logical routing assumptionunderlying the GIS calculated distance. While the majority of observationsfall within our rounding error cone or close to the few for line the observations which the ratio of stated to calculated distance is large, unity do cause a problem. Overall it is difficult to decide, prior to our subsequentanalysis, which distance measureis superior. In hindsight we feel that our survey should have elicited more information upon route itinerary for meanderers.Integration of such information into our GIS distance and duration calculations should produce a superior measure. Definition of tiril2 generatinij functions (tjj ITC tgf's were estimatedby regressingthe number of visits which parties made to the Examination data indicated of explanatory variables. a of variety raw on plots annum site per fit Subsequent data best. dependent log the tests confirmed this variable would that a natural

defined follows: InVISIT was accordingly as and the variable InVISIT = ln(Q+I) where Q is the number of party visits per annum.

Travel cost was initially definedas the sumof time andjourney cost, both of which Time basis to calculated costs were sensitivity analysis. upon a wage rate with were subjected household income. from derived The being journey respondents latter trip return time the

(whether basedupon GIS calculationsor respondentstatements)was then monetisedby income Following discussions by 2 the calculated per minute. the of chapter multiplying factors then time conversion appliedto produceour various where rate/lCisure wage several The factors follows: conversion time cost. applied are as estimatesof

9'See chapter 2 for further details.

4.87

1.100% (assuming that leisure time is valued at the full wage rate); 2.43% (the Department of Transport appraisalrate'00); 3.0% (assuming that there is no opportunity cost of non-work time). 4. Best fit (data determined). Journey cost was also based upon return trips. Three valuation assumptionswere tested as follows: 1.8p per mile (Automobile Association estimate of averagepetrol costs"). 2.23p per mile (Automobile Association estimate of average total running costs"). 3. Perceived (unlike the previous two, this assumption was not related to distance travelled but instead set journey cost at that level stated by respondents in answer to a direct question).

The sum of time and journey cost was then divided through by a factor relating to the days by the enjoyment was attributed which respondentsto their time on-site of proportion for This fact Forest. the Thetford that not all of the trip costs could be made allowance at Such is important this site. allowance to particular especially when, as here, we attributed have evidence of meanderer's and multi-site visitors amongst the sample. This adjusted travel cost estimate formed the first of a considerable list of variables To (semi-log tgf analysis. our comparability within ensure a consistent considered were which dependent) functional

fonn and list of explanatory variables was used for all analyses,

folloWS103: being as explanatory variables

TC

Travel cost (as definedin text)

HSIZE

Householdsize Respondent on holiday (0-1)

HOLS LIVE

Respondentworking (0-1) lives nearsite (0-1) Respondent

RATING

Sceneryrating (1-4)

WORK

10OFromBenson and Willis (1992). 10'Frorn Benson and Willis (1992). INbid. 1030thervariables consideredbut rejected from the comparativemodels include: party size; agc<25; age>65; organisation; environmental membership of separateorganisations, other main activity any of membership dummies.

4.88

TAX

Respondentis a taxpayer (0-1)

NT

Respondentin the National Trust

MDOG

Main reasonfor visit is dog walking

An income variable was omitted from the above becauseof intercorrelation with the Such travel costs. a variable was tested within a separateset of tgf's time cost element of income but here insignificant. the time used, were variable proved costs where zero Four different approachesto tgf definition were investigated as follows: 1. ML and OLS analysis of tgf's basedupon GIS calculated distance and duration; 2. ML analysis of tgf's based upon respondents estimate of total journey cost (perceived cost) and GIS calculated duration; 3. ML analysis of tgf's basedon respondentsestimateof journey duration from which journey cost is also calculated. 4. ML analysis of tgf's based on respondentsestimates of journey duration and journey cost. Sensitivity analysesconcerningthe per unit value of journey cost and travel time were

options. appropriate all on out carried also duration distance GIS based and calculated upon Anal ýsis of tgf's

Here journey distancesand duration are as calculatedin our GIS analysis with the full being discussed. The journey time applied costs as and main of unit sensitivity range inherent in is it the that rounding errors counters approach respondents' an advantage of such disadvantage is inability duration, distance journey the the to main while and of estimates

detect meanderers. OLS analysiswas carriedout as discussedpreviously.TruncatedML analysiswas Here Garrod (1991)114. Willis of and we can rewrite our tgf as: based upon the approach InVISITi = ßXi +ei

independent is (as defined Xj individuals; indexes i explanatory variables our vector of where: P; disturbances independent, be vector and are ej assumed to coefficient previously) with 104Whichin turn is basedon Maddala (1983). We are very grateful to Guy Garrod (University of Newcastle for this assistance excellent with and analysis. Tyne) copious upon

4.89

identically distributed N(O,(;2). Given this model, the ML estimator is basedon the density function of InVISITj which is truncated normal as given in equation (4.16): f (1/cr)O[(InVTSIT. )/Gl if VISIT, >0 -DX. j (1-(D[-PX/.ol) Lo otherwise

f(InVISIT)

(4.16)

Goodness of fit measureswere given by R2 statistics for OLS regressionsand log likelihood values for ML analyses.

i. ML results Sensitivity analysis showed that a marginal journey cost assumption (8p/mile) fitted full based better data than on running costs (23p/mile). Furthermore, a zero an estimate the fitted better DOT (43%) full than the either or wage rate assumptions. assumption time cost Iteration revealed that a small wage rate (2'/2%) time cost assumptionprovided a superior fit best fitting 4.23 ML based GIS Table data. our reports model on calculatedjourney to the distance and duration. Table 4.23:

Best fitting ML model based on GIS estimates of journey distance and duration Courney cost @ 8p/mile; time cost @ 2.5% of wage rate).

Variable Constant TC HSIZE HOLS WORK LIVE RATING NT TAX MDOG gma

Std. Error

Coefficient

0.592317 0.024008 0.054196 0.533289 0.453372 0.394588 0.157927 0.241705 0.237004 0.246530 0.070205

-0.485323 -0.0776564 0.0718489 -1.47287 1.74084 2.27700 0.505034 -0.462887 0.441578 0.606602 1.17890

T-ratio -0.819 -3.235 1.326 -2.762 3.840 5.771 3.198 -1.915 1.863 2.461 16.792

The model given in table 4.23 fits the data reasonablywell and has expectedsigns and The is highly travel variables. cost explanatory variable all significant, easily on significance

4.90

indicating inversely 1% that test, and visits are related to the sum of journey and a passing

time costs. Table 4.24 gives travel cost coefficient, log-likelihood value and three consumer for household the entire range of sensitivity analysesfor ML models per visieOs surplus per basedupon GIS calculateddistanceand duration. Upper figures in the consumersurpluscells (1993) lower figures (in brackets) deflated in 1990 the to while study year are to value relate in 3. those other studies our and with reviewed with chapter to comparison values allow Table 4.24: Sensitivity analysis: ML models basedon GIS calculated distance and duration Travel cost (pence/mile)

Travel time (% of income)

8p

0%

8p

43%

8P

100%

8P

2.5%

23p

0%

23p

43%

23p

100%

23p

Notes: 2 3

6%4

Travel cost coefficient (t-value)

-0.084758 (-3.32) -0.031808 (-2.92) -0.016002 (-2.72) -0.077656 (-3.24) -0.031207 (-3.32) -0.020856 (-3.02) -0.013251 (-3.00) -0.029540 (-3.32)

Log likelihood value

Consumer surplus per

household per . (E)'-, visit

-455.46

3.62 (3.29)

-455.59

9.65 (8.77)

-456.28

19.18 (17.42)

-454.59

3.95 (3.59)

-455.36

9.83 (8.93)

-455.72

14.71 (13.36)

-455.35

23.16 (21.04)

-455.34

10.39 (9.44)

Upper valuesin eachcell areat 1993prices,lower values(in brackets)are at 1990prices.Deflator from CSO (1993). on averagehouseholdsvisited Thetford 14.65 times per annum. Best ritting wagerate with a 23p/milejourney cost.

8p/43%; 15% level: 8p/100%; ML entered at a significance variables all explanatory following models For the gp/2-5%. For all remaining ML models,all explanatoryvariableswith the exceptionof HSIZF, enteredat a 15clo level. significance

105AppendixA2.4.2 also gives consumer surplus per householdper annum and per person per visit for this and all other tgf specifications.

4.91

Examining table 4.24 we can seethat our best fitting model Oourneycost @ 8p/mile; time cost @ 2.5% wage rate) gives a per householdper visit consumersurplus of E3.95 (1993 far defensible This 1990 E3.59 than previous published seems more prices). value at prices; ITC estimates for UK woodland recreation as given in Willis and Garrod (1991). We feel functional form due be to the satisfactory more permitted by the larger sample this may well Thetford I Such ITC accord also reasonably well our earlier with results study. our size of is feel that the study superior"'. present we experiment although The most worrying finding from table 4.24 is the comparatively minor difference in fit between our best fit model and ones using differing journey and time cost assumptions. It is arguable that the deletion of just a very few observationsmight well reversethe ordering Given that that model appeared optimal. such another such statistics the goodness-of-fit of is imply this of our consumer surplus estimates revisions a substantial very changes would be found. Our However, argument can sensitivity analysis counter a matter of some concern. TC Although differing the the within multipliers variable. altering to simply amounts in differing considerably consumer surplus this results estimates, engenders values coefficient impacts have fit. (by their upon model particularly significant nature) such changes cannot be Nevertheless, between differences small. of necessity such models will even Therefore the imply for it the travel this still problems cost may method as argument if we accept such an be by in may engineered surplus estimates switching consumer changes that substantial means is issue for This a serious explanatory power. practical similar between models of quite involving for CBA implications such the assessments evaluations are as studies evaluation clearlY major.

ii. OLSresults used

Given the findings of our ML analyses,only zero and 43% wage rate time costs were I fitting best journey The OLS model used a unit analysis. cost value in the sensitivity

4.25 details Table this time model. cost. 8p/mile a zero and of All the explanatory variables in table 4.25 are correctly signed and generally of high 4.26 Table range of consumer analysis surplus gives our sensitivity statistical significance.

-----------

discussion). "In particularthe Thetford I studyrelied uponOLS estimationprocedures(seesubsequent

4.92

107 measures .

Table 4.25:

Best fitting OLS model basedon GIS estimatesof journey distanceand durationCourneycost @8p/mile;zerotime cost) Coefficient

Variable

0.574852

Constant TC HOLS WORK LIVE RATING

-0.0432747 -0.798169 1.40939 2.00810 0.334414

TAX MDOG

-0.305482 0.277334 0.425503 RI = 23%

Std. Error

T-ratio

0.352221 0.013387 0.264766 0.359360 0.319801 0.105753 0.156728 0.153215 0.179052

R2(adj)= 21%

1.637 -3.226 -2.984 3.923 6.282 3.169 -1.916 1.773 2.396

n= 351

based GIS distance duration OLS Sensitivity calculated and 4.26: on models analysis: Table Travel cost (pence/mile)

Travel time (% of income)

8p

0%

8p

43%

23p

0%

23p

43%

Travel cost coefficient (t-value)

R'

Consumer surplus per householdper visit (E)'-'

-0.046776 (-2.93)

21.72%

21.38 (19.42)

-0.011519 (-2.12)

20.79%

86.81 (78.87)

-0.016801 (-2.90)

21.69%

59.52 (54.07)

-0.008904 (-2.51)

21.21%

112.13 (101.87)

Uppervaluesin eachcell areat 1993prices,lowervalues(in brackets) areat 1990prices.Deflator from CSO(1993). 14.65 Thetford households timesper annum. on visited average 2 level. 15% at a significance entered variables For all OLS modclsall explanatory I Notes: I

107Forcomparative Purposestable 4.26 contains models with exactly the sameexplanatory variables as table between in in best fit OLS difference 4.25 for table coefficient estimates our model the slight 4.24 (this accounts 4.26). in table and its counterpart

4.93

These results confirm our prior ML findings that models using marginal journey costs (8p/mile) and very low (here zero) time costs fit the data best. Also, and for the same

is little difference in before, there comparatively overall degreesof explanation reasonsas acrossthesemodels"'. Comparison of our ML and OLS estimates can be conducted on three levels: On ML theoretical. the statistical grounds and models appearto have statistical; cross-study; better OLS Although data than their counterparts. the somewhat comparison of explained (log likelihood R') is degrees statistics values versus explanation of problematic, overall in directly higher in ML t-values comparable models generally than were variable explanatory OLS models, and invariably so with regard to the travel cost variable. Cross-study have better, ML best-fit that the models performed producing suggest also comparisons in line both are much more which with other studies and prior estimates consumer surplus by OLS However, those models. our such tests of validity are than produced expectations justification. by backed theoretical weak unless The theoretical casefor preferring ML over OLS estimatesis strong. Severalauthors (see chapter 2) have argued that OLS methods are inappropriate for analysing on-site individuals information do data elicit any on not surveys who choose not as such recreation OLS Balkan the truncation the variable neglect of methods visits at zero. the site. to visit OLS in in (1988) that circumstances methods result Kahn such an over-estimate of show and for ML Conversely techniques the truncated can explicitly allow absence surplus. consumer 4.24 with 4.26 suggeststhat the findings of Balkan and Comparison tables of of non-visitors. by (1988) our study". Kahn are confirmed

Consequently we adopt ML estimation

in analyses. techniques our subsequent GIS journey duration based and calculated cost perceived on Analysis, of tjzf's Here journey duration is calculated as before but journey cost (petrol, etc) is taken Such direct question. an approach goes part way towards to survey from responses a in However, statements of meanderers. relying upon respondents the problem addressing

ImMis is of coursecomparingrefinementswithin a commonfunctionalform. As notedin our ThetfordI functional forms be between can differences much morepronounced. rrC study, rind

77 TC Smith (1990) Kaoru studies, of and 1091n their meta-analysis $50. by over OLS estimates reduce

4.94

that adjusting for truncation could

some rounding errors may be reintroduced to the dataset. Given our prior findings regardinglikelY time costs,two wagerates were investigated,

former better fitting 43%, the the providing model which is detailedin table with zero and 4.27. Consumer surplus estimatesfor both perceived cost models are given in table 4.28. Table 4.27: Best fitting ML model basedon perceivedjourney cost and zero time costs Coefficient

Variable Constant TC HOLS WORK

Std. Error 0.554329 0.026318 0.536695 0.453857

-0.0917968 0.0836772 -1.53383 1.74738

2.17069

LIVE

0.461651

RATING NT TAX MDOG Sig na Log-likelihood

-0.518101 0.409297 0.601554 1.18499

T-ratio -0.166 -3.179 -2.858 3.850

0.397130

5.466

0.156705 0.242304 0.238369 0.245362 0.070902

2.946 -2.012 2.154 2.500 16.689

4 iterations; in defined (estimates text) after as variables converged value = -455.95

Our best fitting model basedon perceived costs perforins only marginally worse than This GIS produces similar consumer surplus estimates. based and very calculations upon that both As before, for to the approaches. and to validity some additional give appear would is fit between degree cost of perceived models similar. the overall reasons, same Table 4.28:

Sensitivity analysis: ML models based on perceived journey cost and GIS duration calculated

Travel cost (pence/mile)

Travel time (% of income)

Travel cost coefficient t-value)

Log likelihood value

Perceived

0%

-0.083677 0.18)

-455.95

Perceived

43%

-0.034485 0.03)

-456.60

Consumer surplus per household per visit (E)'-' 3.66 (3.33) 8.90 (8.08)

(in brackets)are at 1990prices.Dctlator in 1993 lower Upper prices, cell at values each are values r4otes: from CSO (1993). on averagehouseholdsvisited Thetford 14.65 times per annum. 2 15% level. a at significance entered All explanatory variables

4.95

Analysis of tgf's basedupon respondentsestimate of ioumey duration In these analyses both journey cost and time cost are derived from respondents duration. A journey specific question asked respondentsto state how statementsregarding long it had taken them to travel to the site. Theseresponseswere then doubled to give round trip journey times to which wage rate proportions could be applied to derive time costs. Implicit journey distance was calculated by assuming an averagespeedof 40 mph, a figure based upon our earlier GIS research. Applying our various per-unit rates gave us our Such journey an approachprovides an arguably more complete approach to cost. perceived it is however liable does the to the section, more rounding errors previous than meanderers induced by moving away from our GIS calculated measures. Sensitivity analysis showed that a zero time cost assumption fitted the data best, This causesa slight problem with regard to the rate"O. wage positive outperfon-ning any both 8p/mile time element, and 23p/mile journey cost joumey cost assumption as, with no fit (i. degrees identical they to model e. act as simple overall multipliers of an give will costs 8p/mile Given has better than identical term). an assumption travel performed cost otherwise is in best detailed table this model which as our preferred in our previous analyseswe chose 4.29.

Table 4.29:

Variable Constant TC HSIZE HOLS WORK LIVE RAIING NT, TAX MDOG Sigma

Best fitting ML model based on perceived journey duration from which journey costs are derived Ooumey costs @ 8p/mile; zero time costs) Std. Error

oefficient

0.589563 0.031096 0.054057 0.530947 0.452641 0.396713 0.155442 0.240327 0.235351 0.245218 0.069881

-0.247513 -0.106951 0.0631174 -1.40119 1.73693 2.14083 0.466641 -0.455569 0.389710 0.625690 1.17540

T-ratio -0.420 -3.439 1.168 -2.639 3.837 5.396 3.002 -1.896 1.656 2.552 16.820

in iterations; defined 4 (estimates text) as converged after variables = Log-likelihood value -453.93

"OVariablewagerateassumptions weretestedhere.

4.96

Table 4.30 details travel cost coefficients, overall fit and consumer surplus estimates for all the models estimatedin this analysis. The best fit per household per visit estimate of consumer surplus is a little over fl. higher than for the models basedon our GIS calculations and has a very marginally superior log-likelihood value. Table 4.30:

Travel cost (pence/mile)

Notes:

Sensitivity analysis: ML models based on perceived duration and derived distance Travel time

Travel cost

Log

(% of income)

coefficient (t-value)

likelihood

8p

0%

8p

43%

8p

100%

23p

0%

23p

43%

23p

100%

I 2

-0.106951 (-3.439) -0.032386 (-2.765) -0.0153005 (-2.499) -0.037200 (-3.439) -0.0220848 (-3.118) -0.013271 (-2.840)

value

Consumer surplus

per household per Visit

(; C)1.2

-453.93

4.86 (4.42)

-456.09

9.47 (8.60)

-456.90

20.06 (18.22)

-453.93

8.25 (7.50)

-454.92

13.89 (12.62)

-455.84

23.12 (21.00)

Upper values in each cell are at 1993 prices, lower values (in brackets) are at 1990 prices. Deflator from CSO (1993). on averagehouseholdsvisited Iletford 14.65times per annum.

For the following truncated ML models all explanatoryvariables enteredat a 15% significance level: 8p/43%; 8p/100%; 8p/2.5%. For all remaining truncatedML models,all explanatoryvariableswith the exceptionof HSIZE, level. 15% significance a entered at

journeY durition based on respondents estimates of Analysis of tgf's and cost In this analysis both journey duration (and hence time costs) and journey cost are

information from to this directly responses separate visitors questions eliciting as part taken before As behaviour an approach such should the capture survey. of of the on-site GIS calculationsbut is susceptibleto response-rounding better than our errors. meanderers 4.97

By comparing results from this approach to those from the previous analysis based solely duration, we can also assessthe relative accuracyof respondentsestimatesof upon perceived

joumey durationandjoumey cost. As previously we only estimated models for zero and 43% wage rate time costs, a 100% rate seeming unfeasible given prior results. Of those the zero time cost model in full in 4.31. is better table and reported performed marginally Table 4.31:

Variable Constant TC HSIZE HOLS WORK LIVE RAUNG NT TAX MDOG Sigma

Best fitting ML model basedon perceivedjourney duration and cost Ooumey costs as statedby respondent;zero time cost) Std. Error

Coefficient

0.593045 0.023003 0.054356 0.536713 0.454354 0.397103 0.156934 0.242462 0.238005 0.246816 0.070644

-0.489717 -0.0676986 0.0969824 -1.47050 1.82321 2.25116 0.469444 -0.484902 0.399381 0.649186 1.18268

T-ratio -0.826 -2.943 1.784 -2.740 4.013 5.669 2.991 -2.000 1.678 2.630 16.741

in dcrincd 4 iterations, text) (estimates 455A7 as variables convergedafter Log4ikclihood value =

fit details andconsumersurplusestimates 4.32 overall travel cost coefficients, Table fit Best in consumer surplus estimates are this analysis. for both of the models estimated is fit journey duration but based somewhat model perceived upon solely those to similar journey indicates This for 4.30 that respondents perceived comparison). (see table worse journey do they cost. than distance more accurately

4.98

Table 4.32:

Sensitivity analysis: ML models based on perceived journey duration and journey cost

Travel cost (pence/mile)

Travel time (% of income)

Perceived

0%

Perceived

43%

1 Notes: I 2

Travel cost coefficient (t-value) -0.023003 (-2.943) -0.011113 (-2.831)

Log likelihood value

Consumer surplus per household per visit

-455.47

4.53 (4.12)

-455.80

9.75 (8.86)

Upper valuesin eachcell are at 1993prices,lower values(in brackets)are at 1990prices.Deflator from CSO (1993). on averagehouseholdsvisited Thetford 14.65times per annum.

For the following truncated ML models all explanatoryvariables enteredat a 15% significance level: 8p/43%; 8p/100%; 8p/2.5%. For all remainingtruncatedML models,all explanatoryvariableswith the exceptionof HSIZE, level. 15% significance a at entered

Thetford 2 ITC studv: Conclusions. This study has examined three separateand very fundamental issues regarding the have both OLS ML Firstly, ITC. estimation examined and we methodsand the of application latter. Secondly, the the use of supporting regarding the valuation found convincing evidence full have time, applied a analysis travel sensitivity acrossa range of we and costs of journey journey low finding that costs and petrol-only very or zero time definitions consistently tgf based functions Travel fitting best cost upon respondents estimates models"'. us gave costs flat these than approaches rate and subsequent analysis journey worse performed cost of journey to their costs compared of perception unsure relatively of are that visitors suggested journey distance duration has been issue Thirdly, the and duration. addressedboth of journey through and estimates a novel application of respondents analysis conventional more through former is better have We the that, approach while to argued the suited GIs software. of GIs to the the take routes site, circuitous approachreduces who identification of respondents Comparison the amongst majority of endemic visitors. are of the the rounding errors which is interesting. As figure from 4.10 derived these two approaches tgf's statistical power of few distance to the compared numbers meanderers whose estimates very a there are showed 1"This resultgivesfurthersupportfor our questioningof theassumptions usedby BensonandWillis (1992) (see 3). thcir chapter of results for estimate revised our thereby and

4.99

may suffer from rounding error. However, the omission of thesefew meanderersin the GISbasedtgf's is likely to lead to a relatively large fall in overall fit compared to the impact of rounding errors upon tgf's basedon visitors responses. In the event our best fit GIS-based tgf has a log-likelihood value only slightly lower than the best fit response-basedtgf'2. These give per household per visit consumersurplus estimatesof E3.95 (1993 prices; E3.59 at 1990 prices) and E4.86 (f:4.42) respectively, amountsthat could defensibly be used to mark out an envelope of valuation. We strongly suspect that a measurementapproach which GIS itinerary information the approach of our with route accuracy elicited from combines basis for ITC superior a significantly studies. provide would respondents

4.4: CONCLUSIONS This chapter has presented three studies of the monetary value of open access has Each these to studies contributed of our understanding of the recreation. woodland the analyses of complexity of monetary evaluation evaluation monetary of complexity for design bias. Thetford I highlighted The in the potential study particular analyses and issues such as the anchoring effect of information on presentpaymentsand the potential for findings into These fed the subsequentWantage effects. vehicle highly significant payment CV study which appeared to be less subject to bias and showed interesting relationships be WT? period which may symptomatic of wider concernsTegarding between and payment is for The Wantage space. study also notable open establishing that the privatisation of public in principle farmers would be preparedto consider venturesinto the provision of recreational levels these although of compensation payments, were considerably appropriate at woodland indicating that crops a risk aversemotivation agricultural on standard returns above present here. be at work may While these early results are generally encouragingour later work in the Thetford 2 both CV TC difficult the optimal execution of regarding and questions studies study raises in by both the variability welfare measure estimates generated highlighting substantial in in the studies observed of previous variability ibis our review presented echoes methods.

"'Best fit GIS-basedtgf (8p/milejourneycostand2.5%timecost;seetable4.24)haslog-likelihoodvaluejourney (8p/mile fit tgf best cost and zero time cost'.see table 4.30) has logresponse-based 454-59 while likelihood value -453.93.

4.100

in both Thetford 2 CV TC 3. While the and analysescasescan be made regarding chapter have greater empirical or theoretical validity, this variability leads us to results which conclude that decisionmakersshould consider a sensitivity analysis acrossdiffering valuation results when taking decisionsregarding the provision of environmental resources. Following this we adopt such a sensitivity analysis approach in the subsequent chapter which is but into the then this translates estimation of visitor numbers with primarily concerned demand valuations.

4.101

REFERENCES Adams,J. (1994)Puttingalchemyto work for openspace?EnvironmentalEconomics,Sustainable Managementand the Countryside,CountrysideRecreationNetwork,Cardiff. Andreoni,J. (1990)Improvealtruismanddonationsto public goods:a theoryof warm-glowgiving, The EconomicJournal,100:464477. Arrow, K., Solow,R., Portney,P.R., Learner,E.E., Radner,R. and Schuman,E.H. (1993)Reportof the NOAA Panel on ContingentValuation,Reportto the GeneralCounselof the US NationalOceanicand for the Future,Washington,D.C. AtmosphericAdministration,Resources BalkanX. and Kahn,J.R. (1988) The valueof changesin deerhuntingquality: a travcl-costapproach, Applied Economics,20:533-539. Bateman,I.J. (1992) The UnitedKingdomin Wibe, S. andJones,T. (eds)Forests:Market and Intervention Failures, Earthscan,London. Bateman,I.J. and Bryan,F. (1994)Recentadvancesin monetaryevaluationof environmentalpreferences, EnvironmentalEconomics,SustainableManagement and the Countryside,CountrysideRecreation Network, Cardiff. Bateman,I.J. and Langford,I. H. (forthcominga) Non-userswillingnessto pay for a NationalPark:an applicationand critiqueof the contingentvaluationmethod,RegionalStudies. Bateman,I.J. and Langford,I.H. (forthcomingb) Budgetconstraint,temporalandorderingeffectsin contingentvaluationstudies,Environmentand PlanningA. Bateman,I.J., Brainard,J.S. andLovett, A.A. (1995c)Modellingwoodlandrecreationdemandusing geographicalinformationsystems:a benefittransfersstudy, CSERGEGlobal EnvironmentalChange WorkingPaper 95-06,Centrefor SocialandEconomicResearchon the Global Environment, University of EastAnglia andUniversityCollegeLondon,pp.65. Bateman,I.J., Langford,I.H. and Graham,A. (1995b)A surveyof non-uscrswillingnessto pay to prevent CSERGE Global Broads, in Norfolk EnvironmentalChangeWorkingPaper GEC flooding the saline 95-11, Centrefor SocialandEconomicResearchon the Global Environment,Universityof EastAnglia London, College University pp34. and J. (forthcoming)Willingness-to-pay Bateman,I.J., Langford.I.H. and Rasbash, questionformateffectsin in 11. Willis, Bateman, K.G. (cds)ValuingEnvironmental and studies, contingentvaluation Preferences:Theoryand Practiceof the ContingentValuationMethodin the US, EC and Developing Countries,Oxford UniversityPress,Oxford. betweencontingentvaluationestimates:a Bateman,I.J.. Willis K.G. and Garrod,G.D. (1994)Consistency Studies. Parks, Regional UK 28(5): National 457-474. two studies of of comparison Bateman,I.J., Diamand,E., Langford,I.H. andJones,A.P. (1996a)Householdwillingnessto pay and for establishinga recreationalwoodland, Journal of farmerswillingnessto acceptcompensation 39(l): 2143. EnvironmentalPlanningand Management, Bateman,I.J. and Turner,R.K. (1993)Valuationof the environment,methodsand techniques:the contingent Sustainable Environmental ) Economics in Turner K. (ed. R. Management: and method, valuation Principles and Practice,BelhavenPress,London,pp.120-191. (1996b) A. A. Measurement, Lovett, J. S. G. D., Brainard, Garrod, I. J., valuationand estimation and Bateman, issuesin the travel cost method:a geographicalinforinafionsystemsapproach,Journal of Agricultural Economics,47(2):191-205. K. G. Garrod, Willis, G. D. H., R. (1995a) I. Turner, K., Elicitationand truncation and Langford, J., I. Bateman, Economics, 12(2): 161-179. Ecological in valuation studies, effects contingent R. Garrod, K. G. D. (1993)The impactsof changing Turner, H., Willis, G., I. K. and J., Langford, I. Bateman, An iterative format in studies: analysis valuation of open-end contingent cd, to question pay willingness ' CSERGE Global Environmental Change formats, Working Paper93. dichotomous choice bidding and Global Environment,Universityof EastAnglia, Economic Research the Social on for Centre 05, and Norwich and University CollegeLondon. P., Langford, I. H. G., Garrod, G. D., Doktor, K. Willis, andTurner,RX (1992)Recreationand Ii., l3ateman, Norfolk Broads: Report a contingent the valuation study, value to the of preservation environmental Group, Appraisal Universityof EastAnglia. Environmental Authority, Rivers National informal Forestry G. (1992) Commission K. Valuing the recreation Willis, on F. J. estate,Bulletin Benson, and 104, ForestryCommission,Edinburgh.

4.102

Bishop, K. D. (1992) Assessingthe benefits of community forests: an evaluation of the Recreational use benefits of two urban fringe woodlands. Journal of Environmental Planning and Management, 35(l): 63-76. Boyle, KJ. and Bishop, R.C. (1988) Welfare measurementsusing contingent valuation: a comparison of techniques. American Journal of Agricultural Economics, February 1988,20-28. Carson, R.T., Mitchell, R.C., Hanemann,W.M., Kopp, RJ., Presser,S. and Ruud, P.A. (1994) Contingent valuation and cost passiveuse: damagesfrom the Exxon Valdez, Discussion Paper 94-18, Resources for the Future, Washington DC. Colenutt, RJ. and Sidaway, R.M. (1973), Forest of Dean Day Visitor Survey, Forestry CommissionBulletin 46, HMSO, London. CSO (1993) Internal Purchasing Power of the Pound: March 1993, HMSO, London. Department of Transport (1992) London traffic monitoring report 1992, Transport Statistics Report, Department of Transport, London. Department of Transport (1993) Vehicle speedsin Great Britain 1992, Statistics Bulletin 93 (30), Department of Transport, London. Eberle, W. D. and Hayden, F.G. (1991) Critique of contingent valuation and travel cost methods of valuing natural resourcesand ecosystems,Journal of Economic Issues, 25:649-685. Gatrell, A. C. and Naumann, 1. (1992) Hospital location planning. a pilot GIS study, paper presentedat the Mapping Awareness Conference,London, February 1992. Goldstein, H. (1987) Multilevel Models in Educational and Social Research,Charles Griffen, London. Hanemann, W. M. (1989) Welfare evaluations in contingent valuation experimentswith discrete responses: Reply, American Journal of Agricultural Economics 71: 1057-1061. Hanemann, W. M., Loomis, J. and Kanninen, B. (1991) Statistical cfficiency of double-boundeddichotomous choice contingent valuation, American Journal of Agricultural Economics 73: 1255-63. Harris, C.C., Driver, B.L. and McLaughlin, WJ. (1989) Improving the contingent valuation method: a Environmental Economics Journal Management 17: 213-229. of and approach, psychological Hoehn, J.P. and Randall, A. (1987) A satisfactory benefit cost indicator from contingent valuation, Journal of Environmental Economics and Management 14,(3):226-247. Kahncman, D. (1986) Comments, in Cummings, R.G., Brookshire. D.S. and Schulze, W. D. (eds) Valuing Environmental Goods: A State of the Arts Assessmentof the Contingent Method, Rowman and Allanheld, Totowa, NJ. Kahneman, D. and Knetsch, LL. (1992) Valuing public goods: the purchaseof moral satisfaction, Journal of Environmental Economics and Management,22:57-70. Kahneman, D. and Tversky, A. (1982) The psychology of preferences,Scientific American 246(l): 160-173. Kahneman, D., Slovic, P. and Tversky, A. (eds) (1982) JudgementUnder Uncertainty: Heuristics and Biases, Cambridge University Press,New York. Kanji, G. K. (1993) 100 Statistical Tests, SAGE Publications, London. in Adamowicz, Resource (1993) LL economics: persistent conventions and contrary evidence, Knetsch, W. L., White, W. and Phillips, W.E. (eds) Forestry and the Environment: Economics Perspectives, CAB International, Wallingford. Kristr6m, B. (1990) Valuing environmental benc!its using the contingent valuation method -an economic Studies UmcA, Sweden. No. 219, University Economic Umed of analysis, Kristrom, B. (1993) Comparing continuous and discrete valuation questions,Environmental and Resource Economics 3(l): 63-71. (1993) 1J. Estimation Bateman, H. I. and reliability of welfare measuresfor dichotomous and Langford, CSERGE Global Change Environmental Working contingent studies. ended valuation choice and open Paper 93-04, Centre for Social and Economic Researchon the Global Environment, University of East Anglia, Norwich and University College London. Langford, (1994) I. J. H. D. Multilevel modelling and contingent valuation part Bateman, H., J. and Langford, 1: a triple bounded dichotomous choice analysis. CSERGRGlobal Environmental Change Working Paper 94-04, Centre for Social and Economic Researchon the Global Environment. University of East Anglia, Norwich and University College London. Langford, I. H., Bateman, I.J. and Langford, H.D. (1996) A multilevel modelling approach to triple-bounded dichotomous choice contingent valuation, Environmental and Resource Economics, 7(3): 197-211.

4.103

Loomis, J.B., Walsh, R. and Gillman, R. (1984). Valuing option, existence and bequestdemandsfor 60(l): Economics, 14-29. Land wilderness, Maddala, G.S. (1983) Limited-Dependentand Qualitative Variables in Econometrics, Cambridge University Press, Cambridge. Milon, J.W. (1989) Contingent valuation experiments for strategic behaviour, Journal of Environmental Economics and Management 17:293-308. MINITAB (1991) MINITAB ReferenceManual PC Version Release8, Minitab Inc., Rosemont, P.A. Mitchell, R.C. and Carson, R.T. (1989) Using Surveysto Value Public Goods: The Contingent Valuation Method, Resourcesfor the Future, Washington, D.C. Oppenheim, A. N. (1966) Questionnaire Design and Attitude Measurement,Heinemann, London. Ordnance Survey (1987), Gazetteerof Great Britain, Macmillan, London. Ome, M. T. (1962) On the social psychology of the psychological experiment, American Psychologist 17:776789. Roberts, K. J., Thompson, M. E. and Pawlyk. P.W. (1985) Contingent valuation of recreational diving at Society Fisheries 114: 155-165. American Mexico. Transactions Gulf the of of petroleum rigs, Sagoff, M. (1988) Some problems with environmental economics,Environmental Ethics 10:55-74. Sellar, C., Stoll, J.R. and Chavas,J.-P. (1985) Validation of empirical measuresof welfare change: A 61: 156-175. Land Economics techniques, nonmarket comparison of Smith, V. K. and Desvousges,W. H. (1986) Measuring Water Qualhy Benefits, Kluwer-Nijhoff, Boston. Smith, V. K. and Kaoru, Y. (1990) Signals or noise? Explaining the variation in recreation benefit estimates, American Journal of Agricultural Economics, 72(2):419433. 3. Magazine, 72: Trust National The Sparc (1994) the A. axe, Stirling, in Bateman, I.J. and Willis, K. G. (eds) Alternative theories (forthcoming) of choice, R. Sugden, Contingent Valuation of Environmental Preferences:AssessingTheory and Practice in the USA, Europe, and Developing Countries, Oxford University Press,Oxford. by 7le J. (1988) Lord, the contingent C. Green, goods of environmental evaluation S., and Tunstall, Centre, Middlesex University. Research Ha7ard Flood Report, F11RC valuation method, Sea Scheme. Defence A Assessmentfor Aldeburgh (1988) Benefits S. J. Brooke, the K. R. Turner, and Environmental Appraisal Group, University of East Anglia. defence: Valuing (1992) benefits J. S. Brooke, J. I. the of coastal a case study of Bateman, K., and R. Turner, C. (cds) Valuing in Coker, A. Richards, Environment: defence the and Aldeburgh scheme, sea the John Belhaven Press: Wiley Sons, Evaluation, London. Environmental Approaches and to Economic in (1989) Issues J. R. McKean, M. D. nonmarket valuation and policy application: Johnson, G.. and Walsh, R. Economics 14(l): 178-188. Agricultural Western Journal of A retrospective glance. Economics, Sustainable Environmental future, into die Summary (1994) A. and glance Whiteman, Cardiff. Network, Recreation Countryside Countryside, the Management and Community Planting Three Benefits Forests: Forest Costs The J. (1994) Sinclair, of A. and Whiteman, and Studies Division, Policy Forestry Commission, Great North Forest, Chase Thames Mercia, and of Edinburgh. benefits (1988) A J. F. Benson, and costs of nature conservation at three of user G. comparison K. Willis, and Studies, 22: 417-428. Regional nature reserves, forest further benefits (1989) Values J. F. site surveys. Benson, of recreation: some of user Willis, K. G. and County University Planning. Town Commission, Department Forestry and of of Report to the Newcastle upon Tyne. (1991) Landscape G. D. Garrod, values: a contingent valuation approach and case study of G. K. and Willis, Change Initiative Working Paper 21, Department of Countryside Park, National Dales Yorkshire the Newcastle-upon-Tyne. University Food Marketing, Economics of and Agricultural Journal landscape: (1993) Valuing G. D. Garrod, approach, a contingent valuation of G. Willis, K. and Environmental Management, 37:1-22.

4.104

Chapter 5: Recreation: Estimating and Valuing Demand 5.1: INTRODUCTION In this chapter we utilize geographicalinformation systems(GIS) to model arrivals at a particular woodland site and test the efficiency of the resultant arrivals function in from Findings to other sites. our studies of the value of open-access predicting visits woodland recreation are then applied to our predicted visits surface to obtain a valuation of potential demand.

5.2: ESTIMATING AN ARRIVALS FUNCTION 5.2.1: PREVIOUS STUDIES In this chapter we are concerned with estimating overall visitation rates which are being individuals. By definition than to specific rather populations, across applicable have individual to only studies relevance visitors and tell us nothing valuation conventional incapable determining the absolutenumber of people who they that of are so of non-visitors, has be Therefore, to arrivals model composedof variables which our visitor will visit a site. have relevance acrossthe population. To date there has been relatively little researchregarding the level and determinants few in Furthermore, have UK. for those the of demand studies which recreation woodland of demand looked (Willis and Benson, have issue, recreational at national looked at this most

forest One 1991) that Whiteman', than site. at any particular notableexception rather 1989; (1973) Sidaway Colenutt by regardingthe modellingof demand thework and of is provided household Here (postal) Forest Dean. the a combined to on-site of and day for trip visits factors determining information trip the trips. origins to and regarding collect used was survey important factor determining by far data that the most revealed arrivalswas Analysis of this trip duration, to the effectiveexclusionof otherexplanatoryvariables. The Colenutt and Sidaway result is important as, if it were reconfirmed in our own

be function' 'arrivals travel time to the could estimated relating probability of a an analysis,

'Whiteman (1995)presentsnationalmodelsof both the demandand supplyof UK woodland.

5.1

2. visit taking place The spatial analytical power now provided by a GIS makes it possible to apply such a function to detailed population data, such as that provided in the UK Census, in order to predict arrivals at any existing or hypothetical MO. Obviously, in practice, the function for an arrivals estimated one site to another would need to be of applying validity in carefully assessed terms of the accuracy of the predictions made and such a cross site is test carried out and presentedsubsequently. actual predicted versus 5.2.2: RECREATION DEMAND: THE THETFORD FOREST STUDY The base data for our investigation was elicited as part of the Thetford 2 CV/ITC duration journey distance for Individual and measures were calculated use within the study. ITC valuation study as describedin chapter 4. These measureswere, by dint of their method for derivation, the availability and quality of the road network and are therefore adjusted of inherently superior to the simple, unadjustedequivalents used in the Colenutt and Sidaway inappropriate level for individual However, variables were use within our arrivals such study. function. We therefore neededto convert our travel time road network data into continuous have To to and non-visitors relevance visitors aRe. obtain this would time which travel zones date derived for individual the each segment of the road vector continuity of coverage be had to rasterised. network Rasterisation is a processof converting vector featuresto cells on a regular gri&. In this study the travel time values assignedto points along roads were reassignedto the grid 'majority filter' Rasterisation be to those allowed a points. run cells which contained in between fill the to gaps the roads, providing smoothly study area entire recursively across travel time for the surface a continuous centred and producing upon site cells grid all values 'For obvious reasonseconomistshave tendedto focus upon travel cost rather than travel time, indeeda major its incorporation within the has been travel time to the of enable monetary evaluation focus of such research have hand, both Geographers, distances (for function. the time examined on other and cost a cost overall visit have been limited latter investigations but 1983), the and most research Gatrell, of relatively empirical review sce Euclidean distance. This study seeks to has to tended of simply on measurements rely location theory on but time travel the of visits, also examining both overall costs as a predictor of quantity using concepts, examine in an attempt to place monetary value on the relevant recreation. 3English Nature and the ESRC have recently provided the author with funding to extend the work described in this chapter so that socioeconomic factors and the availability of substitute recreational facilities may be function. the incorporated arrivals within explicitly 4SecEnvironmental SystemsResearchInstitute (1993) for further details. 5Vcctor features (roads) were rasteriscdonto a 500 m grid. This equated to a total of 161195 cells for our filled directly 58364 being through the were the rasterisation process, remainder of which area study entire described in the text. the process through values assigned

5.2

and fanning out to fill the entire study area. The majority filter worked by means of a 'moving window' (usually eight by eight cells in extent)', where the centre-mostempty cell was assignedthe value held by the majority of already assignedcells in the specified window. The majority filter worked well for the vast majority of cells. However, a few gapsremained in areasvery remote from any roads where the filter window did not contain any cells filled directly by the rasterisation process. These grid cells were given the values of their nearest neighbours. Once all the grid cells had been assigned a value these were reclassified into convenient categoriee. Inspection of the calculated travel times showed that the extended 120 to all values up minutes. Within this range, 13 time zones encompassed road network innermost Given defined. the the the of origins concentration visit around site, zones were (between 0 and 30 minutes) were tightly defined at 5 minute intervals, after which 10 and 30 bands (between 60 15 were used and minutes and 60 to 120 minutes minute eventually illustrated in Certain figure 5.1. time travel the resulting of zones are respectively). Once travel time zoneswere defined the relevant zone for each survey respondentwas Arc/Info provides two routes of obtaining this information; the identity and Both identical tested were and commands. provided almost results. The addroutemeasure identified.

identity command appearedto suffer slightly more from the rasterisationanomaliesdiscussed from Results from this exercise are addroutemeasure were preferred. above and so results (1) 5.1. Here first limit in table two the the of column shows columns upper of each presented (2) (in travel to the site) and column of vehicle minutes records the number travel time zone during from (other the the the period of survey' to zone site each columns are of party visits

6A filtering window of 8x8 cells was used.This was held constantacrossour entire study area except for few (only feasibly 4 filled incorporated into to as as the could reduce cells cells window are where cells edge does but, large distortion in The the an edge exist given rilter). of the very number of cells used possibility the distortion would be extremely minor. dataset, Thetford such any entire

77he time zone classification procedure revealed a very minor but intractable problem in the rastcrisation GIS (version 6.1). This in Arc/Info by definition 0 the the two was evidenced of separate available algorithm Since 71ictford by the travel essentially to people would 5 time assuming we were travel zones. to minute to cells during the rasterisation process should have been the time assigned values possible, route quickest (giving by lower directed As table the a to time values) weight preference software manuals, smallest possible. implemented inconsistently by but There the this this, programme. was to ensure are references was spaified file, known_problems, Arc/Info in documentation but is the online no solution shortcoming offered. this to Furthermore, the effect of this inconsistency was compounded slightly by the majority filtering. Even so, as is 5.1, the overall impact of this problem was minor. figure from evident Me possibility of repeat visits was recognised. This was tested for and proved not to be a feature of the survey sample.

5.3

discussed subsequently). Of the total sample of 351 parties, 326 (92.8%) originated from time GIS by road network. our zones encompassed

This provided a sufficient sample to both

it beyond function the limits of our road network. and extrapolate estimate an arrivals Figure 5.1: Travel time zones for the Thetford Forest study (travel time in minutes)

Travel time (minutes) =

0-

5

M

0

40-50

5-10

50-60

10- 15

60-75

15-20

75-90

20-25

90-105

25-30

105-120

30-40

No Data

Study Site

17ý7/ Main Roads

10

I:

5.4

0

3p

4,0 mm

I 100 000

50

km

Table 5.1: Observed and predicted visitor rates Time Zone' (1)

Actual Visits' (2)

5

13

10

Zonal Pop'n 3 (3)

Observed Visit Rate (4)

Predicted Visit Rate' (5)

Predicted

954

0.0136268

0.0103972

9.9

31

21596

0.0014355

0.0027285

58.9

15

8

13326

0.0006003

0.0012476

16.6

20

10

14377

0.0006956

0.0007160

10.3

25

26

26811

0.0009698

0.0004655

12.5

30

38

58416

0.0006505

0.0003274

19.1

40

46

191009

0.0002408

0.0001879

35.9

50

65

405831

0.0001602

0.0001222

49.6

60

17

375134

0.0000453

0.0000859

32.2

75

48

776817

0.0000618

0.0000559

43.4

90

15

562508

0.0000393

22.1

105

7

253762

0.0000267 ' 0.0000276

0.0000292

7.4

120

0.0000225

150

0.0000147

180

0.0000103

210

0.0000077

240

0.0000059

300

0.0000038

360

0.0000027 0.0000014

Notes:

1. Upper limit of travel time zone measuredin minutesof vehicle travel. 2. 3. 4. 5. 6.

Number of party visits recorded during survey (no repeat visits in sample). Number of households within each travel time zone as recorded in the 1991 Census. Column (2) divided by column (3) Visit rate predicted from the best fitting arrival function (detailed Subscquently) Predicted visit rate multiplied by zonal population (number of visiting parties)

5.5

VisitS6

(6)

-

-

The desired arrivals function would predict visits as a function of travel time. However, to achieve this it was necessaryto account for varying population densities across in (i. to terms of party visits per capita). calculate a visit rate time e. needed we our zones Accordingly a population surfacewas createdwhich coincided in geographicextent with the in for Enumeration Districts (the finest level Totals resident persons usually road network. from 1991 Census data SASPAC detail the extracted using software were available) of (London ResearchCentre, 1992)and grid referencesfor centroidswere obtained from the U42 files held at Manchester Computing Centre. A check on the accuracyof grid referenceswas for distances by the Enumeration centres and mean standard calculating then conducted Districts within each Ward. This processrevealeda few grosserrors in grid referenceswhich were corrected. Allocation of residential populations to the 500 metre grid cells comprising the travel form SBUILD the through algorithm, using a of a volume-preserving time zones was achieved image A (1990). by Martin described to mask used prevent allocations was programme 6,675 initial input to the software consisted of centroids with a and outside the study area by SBUILD (after The 2,723,971. totals cell were rounded surface produced of population inspection 2,724,133. Detailed indicates integer) total of a population contained to the nearest the are represented only criticism which might areas well and of urban that the characteristics isolated 'unpopulated' is contain undoubtedly properties. classed as areas be made that some inevitable data for is, however, deficiency given on reliance areal virtually This type of is in Districts Enumeration this thought the to context of research not and as such aggregates problem. significant a represent Population totals for our defined travel time zones were straightforward to calculate

Arc/Info. By the to travel each of surveyed parties a allocating of module the grid within for the a visit zonal rate was calculated, using population, zonal allowing and time zone in from 5.1. Here (3) Results table this are shown exercise column zonalsumcommand?. from (4) divides Column as above. each extracted visits zone population the zonal records This (column (3)) to by rate. observed visit represents our give (2)) (column zonalpopulation (5) function. The (6) in contents of columns and are dependent arrivals our variable the

9To calculate the zonal visit rate, the zonalsumcommand was executed as follows; (where angpopgr), angpopgr = east anglia population). zonalsum(timezones, popzsgr = for its the time zone. contains sum population cells Each of the new output

5.6

described subsequently. Table 5.1 indicates a marked inverse relationship between travel time and visit rate. Note that the furthest time zone (120 minutes) has beenomitted from our observeddata. This (see figure by 5.1), the road network encompassed and the calculated completely zone was not for in initial Data this zone was consequently statistical analysis. rate appearedanomalous 3.1). function (see from the arrivals appendix the of calibration excluded An examination of the relationship betweentravel time and visit rate was undertaken, full details of which are given in appendix 3.1. This revealed that a double log model " data. Equation (5.1) fit the to the surnmarises resulting arrivals excellent an provided

function. In VR = -1.46 - 1.93 In TZ (-2.41) (-11.39) R2 (adj) = 92.1% Figures in parenthesesare t-values. where:

VR TZ

from i divided by (number zone of party visits observed visit rate zonal population) travel time zone (minutes)

investigations into potential omitted variables and correlations of residuals failed to " Given (5.1). the this strength of relationship with equation problems significant reveal any function distant in time to travel more zones. arrivals our felt extrapolating confident we (2) 5.1 list (5) columns predicted visitor rates, while table (4) and observed of Columns and function The respectively. arrivals numbers visitor and (6) report actual and predicted during first 12 from time the travel sampling 317.8 the zones period. visits party predicted 12 less 2%. i. 324, figure than e. an error of of This compares with an actual

One of

interviewed during the samplingperiod. function to those Our arrivals visitors refers Illetford Welsh for than at rather at a sitewas survey our conducting the mainreasons

(dependent) form model, other forms fitting the data log double outperformed "The a semi-log narrowly is similar to the findings of Colenuttand Sidaway (1973) who report results for both forms although This poorly. is it is not made clear which superior. "Detailed analysis reported in appendix 3.1. "Note that both these figures (actual and predicted) omit non-sampledvisitors (eg. those arriving at hours for (details in 3.2). These appendix subsequently are adjusted outside those sampled).

5.7

that it is one of the very few forests for which accuratedaily and weekly visitor records are available (weekly data being held for several years). This information enabled us to allow for those visitors to Thetford which we failed to interview during our sampling period and between that existed to stable relationship visits during our sampling a very also establish 3.2 full details (appendix This this analysis). gives of allowed us to visits and annual period function basis. Comparison an annual arrivals onto of predicted period extrapolate our sample less 2%. discrepancy than a of showed with actual annual visits

5.3:

APPLYING ARRIVALS

THE ARRIVALS IN WALES

FUNCTION:

PREDICTING

Our first concern was to test the validity of our arrivals function againstactual arrivals boundary defined A Welsh coincident was and road network study area sites. of sample a at in Thetford In to the to manner analysis. similar order a constructed surface and population Welsh border, the the distant to sites along for woodland study area travellers potential allow 13 boundaries Appropriate into England deep county defined to were reach as so was . data were then extracted, clipped and database". Road Bartholomew from the obtained Wales B-roads described outside minor class and roads of were previously. as corrected in Roads topology. that their significant road created gaps omission deleted, except where M6 included (notably defined the motorway outside also just the study area were outside were impact have likely if to on population accessto a significant Coventry) their absenceseemed illustrated is in figure 5.2. The network road resulting the road network. Population data and centroids for Enumeration Districts were again obtained from

30,311 Enumeration Districts Centre. The Computing encompassed area study Manchester had been Once 13,821,562. centroid grid references of population total resident a with 500 to Arc/Info generate a population surface at m programme was used sbuild the checked, for illustrates 5.3 the Figure the population the output, surface resulting resolution. grid cell being 13,821,361 integer) people. to the nearest (aftercell totalswererounded

OThe study area was defined as the following countiesand areas: Avon, Cheshire,Clwyd, Dyfed, & Worcester, Merseyside, Gwent, Gwynedd, Hereford Mid Glamorgan, Powys, Manchester, Greater Gloucester, West Midlands Glamorgan, West Anglesey Staffordshire, & Holyhead Glamorgan, and South Shropshire, Britain islands the of wereremoved. coast 14Minor off

5.8

Figure 5.2:

Digital road network for Wales and the English Midlands'

FZVI Motorway A-road, multi-lane A-road, single-lane B-road Minor (other) road Coastline 2ý

50

75

190

135 km

I: 2 000 000

from figure. minor roads the are ornitted reasons For Note: cartographic

5.9

In order to compare the recreational potential with the agricultural output results calculated subsequentlyin this study, it was highly desirable to present the former as a map. The monetary equivalent of this demand could then be evaluated using our findings from chapters 3 and 4, and compared with that for agriculture to see where a switch to woodland Such demand desirable. be a recreation surface required the estimation of predicted might arrivals for a regular grid over the entire extent of Wales. 'Ibis necessitatedrepetitive, intensive data processing. Furthermore, becausethe size of the Welsh study area and road for Thetford the that exceeded analysis, an alternative approachto the network considerably calculation of travel time zonesand zonal population around eachpotential site was necessary. Accordingly the road data was transformedfrom a vector to raster format to allow the use of iunction in the definition of time This zones. required generation of an costdistance a impedance surface, with the value of each grid squarerepresentingthe impedance involved during traversal of that cell. Overall, the costdistance approach seeks to minimise the impedance between origin and destinationlocations. Severalstepswere necessaryto generate the impedance surface.

class

Roads were rasterisedon a 500m regular grid. The value assignedto each cell was the in database) (as Bartholomew the recorded with the greatest the segment road of

As long length the through grid square. a consequence, a section of road running cumulative 5000 took cell just the a precedenceover a short segmentof road that of edge clipped that feature This length the the had grid square. was a of within rasterising the greatest actually Urban boundaries" be were rasterised and circumvented. readily algorithm and could not from The of to separation urban rural roads. allow the adjusted network road overlayed onto Thetford 2 ITC (chapter 4) the of as part study calculated speeds were urban and rural road However, impedance initial the travel scrutiny times of resultant values. on to calculate used is illustrated This for Welsh that too most were slow. suggested network road the rasterised journey 5.2 in both times 32 table the calculated where were using original routes of a sample Such be a contrast can attributed the version. rasterised subsequent and network road vector bias in the towards topology and classifying road upon rasterisation changes to unavoidable length longest the than those in rather with cumulative segments of of road terms cells the square. grid within extent greatest

"Also obtainedfrom the Bartholomewdatabase.

5.10

Figure 5.3:

Population density surface for Wales and the English Midlands (population in Arc/Info 1991 500 square: calculated sbuild grid using software m2 with each Census ED centroids)

Population Resident Estimated Cell Grid Metre 500 in 1991 per < 10 10 - 24 M Jim

7SZ]

25 - 49 50 - 99 100-249

250 -499 500 -999 >= 1000 2ý

_i..ý

5P



IqO

5 km

I: 2 000 00o

Welsh Border

Crown Copyright, ESRC/JISC purchase. Census, The 1991 data: Source of population SBUIFLD the density calculated software using values were The population Enumeration District centroids. Census 1991 with

5.11

Table 5.2: Calculatedjourney times for selectedroutes in the Welsh study area (in minutes).

Route Aberystwythto Machynlleth Aberystwythto Lampeter Lampeterto Caernarfon Newtown to Brecon Aberystwythto Knighton Caernarfonto Chester PembrokeDock to Aberaeron Aberystwythto Aberaeron Swanseato Brecon Cardiff to Welshpool Wrexhamto Newtown BlaenauFfestiniogto Llandudno Rhyl to Colwyn Bay Newtown to Welshpool Swanseato St. Davids Haverfordwestto Wrexham Pwllheli to Holyhead Pwllheli to Llangollen Cardiganto Brecon Cardiff to Liverpool Cardiff to Liverpool Airport Cardiff to Aberystwyth Coventry to Aberystwyth BirminghamAirport to Swansea Bristol to Caernarfon ManchesterAirport to Caemarfon Aberystwythto Caemarfon ManchesterAirport to Port Talbot ManchesterAirport to Liverpool Airport Port Talbot to Liveipool Airport Aberystwythto Shrewsbury Aberystwythto BirminghamAirport

5.12

Vector Time- -Raster Time 16 27 49 59 149 173 77 87 100 91 94 144 93 118 24 31 58 69 155 175 57 72 39 52 32 40 19 20 103 149 221 257 81 107 102 121 119 134 205 224 201 238 160 165 168 196 154 199 244 284 135 179 103 114 205 268 45 52 224 258 102 116 154 173

A regression analysis was conductedcomparing raster and vector travel times for the routes in table 5.2. No constant tenn was fitted and the results were as shown in equation (5.2).

VECTOR=

0.84 RASTER (68.74)

(5.2)

R2 (adj) = 97.2% Figures in parenthesesare t values. where: VECTOR = Calculated travel time (minutes) on vector network RASTER = Calculated travel time (minutes) on raster grid On the basis of the result given in equation (5.2), all the raster road speedswere did Cells have (i. 1/0.84). 1.190 (and by that through them not roads running e. multiplied impedance impedance that assumedan averagerural assigned an were value) therefore no into This 2.66 likely lack to take the value attempts account rnph`. of of walking speed between footpaths the roads and sample sites. straightline suitable With the Welsh travel time zone algorithm defined, an actual versus predicted test of for (as function time calculating as our methodology revised zones) was well arrivals our possible"'.

The Forestry Commission only holds visitor data for five sites in Wales. In

it became for these that two of closed apparent were unusually conversation with officials

long periods during the year. Furthermore a third contained several special attractions not for forest found those above numbers normally sites expected which raise visitor at normally (5.3) Equation to the location'. actual per annum visits simply relates prediction a such

function for to the time by these arrivals zones genemted our sites and applying obtained factor during Iletford conversion calculated the visitor period/annual the sample using survey".

"Equivalent to 4.29km/h or lkrn every 14 minutes. 17Fulldetails of this analysis are given in appendix 3.3. "These include a museum, catering facilities and a variety of organised recreational activities. 'Trior analysis confirmed that any constant was insignificantly different from zero.

5.13

ACTUAL = 0.903 PREDICTED (4.420)

(5.3)

= 83.0% where: Actual ACTUAL arrivals at site (party visits pa.) = PREDICTED = Predicted arrivals at site (party visits pa.)

Equation (5.3) indicates that the arrivals function performs reasonablywell, the slope coefficient for PREDICTED not being significantly different from 1. However, using dummy for factors highlighted by Forestry Commission officials the to site-specific account variables fit improved by the this model, of as shown equation (5.4). significantly

ACTUAL = 0.958 PREDICTED - 73692 CLOSED + 107397 SPECIAL (-2.23) (2.70) (7.10)

(5.4)

98.4% where: CLOSED =I for two sites closed for extendedperiods during the year; 0 otherwise. SPECIAL= I for one site with special attraction; 0 otherwise. Clearly, the use of dummy variables with such a small number of observationsis not ideal. However, given the reasonable strength of equation (5.3) we can conclude that the

function doesprovide an adequatepredictorof arrivalsat a typical woodlandsite (although function illustrates (5-4) the the to any particularnon-typical applying problcms of equation site). Given this result, the arrivals function can reasonablybe applied to a regular grid of

Walesý'. An to recreational potential expected visits woodland sites across to predict points important practicalissue,however,is theappropriategrid sizefor suchan analysis.Evenwith

2OSuchestimates do not take into account the substitution effects which would arise in any specific area if in locality. The is that identify the planted object of were current to exercise thoseareas woodlands of a number beneficial. be impact The a wood would of be of supply side changes the establishment will considered where subsequently.

5.14

the use of a raster structure and other efforts to shorten processing,determination of travel time zones for a representativegrid covering the entirety of Wales representeda significant Sparc SunI between 15 30 Using took each site and a workstation computational exercise. former Assuming (depending the time, calculation of aI on workload). minutes to process km grid surface for the entire area of Wales (some 20,500 cells) would take over 200 days Even for though three these such machines available were of continual processing. calculations, a courser sampling schemewas clearly required. The issue of grid size was investigated by defining two transectsacross Wales, one border, from Aberystwyth from to the the and second running a the near coast running east Swansea. 5.4 illustrates just Figure, these due transects. to outside a point south similar origin The horizontal transect consisted of 19 sites, the western-most 13 of which were km 5 The 18 km 2.5 transect by consisted spacing. the of at vertical and remainder separated in defined for Travel interval. the km these time sites same were all of zones a5 at all sites Zonal Thetford then for calculated and populations were expected visits study. the as manner 5.5 illustrate 5.6 Figures function. total annual visits and predicted the arrivals using estimated for the horizontal and vertical transectsrespectively. Examining the horizontal transect (figures 5.4 and 5.5) the overall pattern is highly 2, located at the western end of the transect,are close to the town 1 Sites and encouraging. in high is This compared numbers with predicted visitor relatively reflected Machynlleth. of in located the mid-westem, populated sparsely the are to east which sites neighbouring decline in in infrastructure Poor the the compounds areas upland Cambrian mountains. between lowest The roughly midway are estimated numbers predicted visitor numbers. Predicted in high Newtown, the arrivals rise sharply closer to mountains. Machynlleth and Thereafter is to the town. the site closest visitor at Newtown, and the peak value achieved Cambrian leaving because both the high and more uplands entering are we numbers stay infrastructure improved because that English towns and means lowland and areas, populous impact have begin distant, to being an arrivals. upon predicted now despite relatively cities, jump in by illustrated is 5 km krn the 2.5 from to predicted well resolution The switch km 14 (5 km (2.5 13 resolution), resolution) and between sites arrivals

5.15

Figure 5.4: I.Axation of two transects across Wales

LIVERPOOL "&Vfl

All,

MANCHESTER

y CHESTER jW,,

L

,\Z, IaI I's

CREWE

1k ON STOKE

wllhel

TRENT

b0ioellau

Barmouth't-

oakengates WOLVERHAk4PTON ®R f3riciogenorth

Ln Aberys

h

BIRMINGHAM

WORCESTER

Larnpete, FishQuarcJ

ýER;EFORD

V.

GLOUCESTER

Carmarine,

: DMllford Haven

a ko ro

/'

A mm an f n, C',

LýANELý-! kTenby

Nealr,

NEWPORT SWANSEA C RDIFF Sites

transect usecl in vertical (Northernmost - site 154

BRISTOL

0

Sites used in norizontal iransec! (westernmost) . site 117) 0 At)erystwytri ////SWANSEA

[email protected]( Large

towns iowns)c-lies

5.16

30km

Figure 5.5: Predicted visits (parties per annum): horizontal transect

123456189

10

11

12

13

14

15

16

17

18

19

SITE

NO.

Figure 5.6: Predicted visits (parties per annum): vertical transect

23456789

10 11 12 13 14 15 16 17 18

SITE NO

Findings from the vertical transect (figures 5.4 and 5.6) suggest that the 5 km grid does provide adequate sensitivity regarding changes in those factors detennining visit lies in low inaccessible, The point a relatively northemmost population density area. numbers. However, moving southwards, the transect crossesan estuary and passesclose to the town 5.17

4) Thereafter (site Aberystwyth arrivals the transectagain and predicted accordingly. rise of in fall Mountains Cambrian into off of predicted visitors. However, the resulting a climbs is decline Lampeter that this not as pronounced as for the horizontal ensure towns such as transect. Towards its southernend, the vertical transectnearsthe major population centresof Swanseaand Llanelli. Furthermore,the excellent infrastructureprovided by the M4 motorway is in large in This the to catchment. reflected steep a rise makes such sites accessible latter for the sites. arrivals predicted The detail afforded by the 5 km grid systemused in the vertical transectindicates that in in is the major contrasts predicted visitor numbers reflecting such a resolution adequate Clearly 2.5 km density by availability/quality. a and road grid would population engendered inevitably give further information regarding rates of change. However, given the very from 5 demands the the a grid, and of results of such acceptability considerable processing krn resolution sites, such an approach seemedunnecessary.Accordingly travel time zones Wales. for The base 5km for the entirety of map of grid points grid used a calculated were is illustrated in figure 5.7. potential surfaces visitor to generate subsequent Regardless of the chosen resolution, certain sampling problems are difficult to from the interaction of the road network with the sampling Inconsistencies arise alleviate. fell from kind how far Two depend Cell any a sampling point of upon road. values pattern. infrastructure might have from far road comparable and with population areas equally if in (and therefore one of the areas times numbers) predicted visit different estimated travel in far from fell the the sampling point was other any and on a road right point the sampling However, findings for this the is arbitrariness. around There way no straightforward roads. Wales) (and the of were area reassuringly entire sensible and subsequently transects two the had had impact. inconsistencies any these not significant that predictable, suggesting 5 krn follows. A window was for the sites grid as Travel times were calculated eachof An the rasterised. allocation process, using the cost and site site each derined around in linking find the the site and each other cell the shortestpath impedance grid, was run to locations impedance these then The to of each was assigned reach necessary surface. raster in in This travel, minutes time-surface of grid. provided, output a an cells to corresponding into Information for total time then on classified residents time zones. each was which output from in the surface and rasterised population recorded extracted a subsequently were zone iterated in 5 km This the then file. all sample across sites was grid. process separate 5.18

Figure 5.7: 5km grid points used to generate the predicted woodland visitors surface.

11

%1 IT, 0

1%, *

IV 4.

0

A,

11,

1%.

X1

%V k,

A,

11

11

A,

Al

0

41

A,

AN Al

A.

,

%, "

%,

lk

%

11

IV

0

4

Al

NV \N

A)

4

li,

1\1 1ý

0

q,

ýp

1\1 *

*

%N .10

*

0

\1

A,

IN,

4

A\

A,

1%,

k%,

\v

*

\v

0,

AN

a,

*

IN,

1%, 1\1 4 IN

\1

AN *

0

Al

%,

I&

It

11

A,

0

*4

AN

11 A,

0

0

%

%,

ýD

1\1 *

0

0

kv

%1 0

Al

4.

0

1%

IN,

*

\V

4

4.



.%N Al

ffi

1%, 0

ki

A,

AN

10

&

*

4.

0

*

A'

A,

1%,

A,

It

A,

4\

ol

-P

NV

4

41

4.

0

v

0

A'

'V.

4'

41

0

IV

19

*

*

0

-b

*



V,

4,

kv

4

AV

Al A% 1\1

4

4

4

0

A,

4

t

0

44

4 dD

41

4

.0

+

01

4

&

a

1%

41

4

Al

10

a,

4 A,

KeY: 5 km Grid SquareCentroids 5p

I:

7p

2 000 000

5.19

0

A.

4

AN

6,

1\1 Ax

kv

4

0

*

0

iv

IV

Is,

0.

%V



a,

A,

Ir

0,

A\

1%,

0

-D

A,

01

lvý

0

1%

\% A%

k\

25

a

DAMAGED

TEXT

N ORIGINAL

Once time zones and zonal populations had been calculated for all grid points, woodland recreation demand (in terms of total party visits per annum) was predicted using

the estimated arrivals function. Figure 5.8 illustrates the resulting predicted woodland visitors surface Figure 5.8: Woodland recreation demand in Wales: Predicted annual total party visits per site

PredictedVisit.,

WoodlandSitesper AnnLIM

Under 35,000

Em

35 to 59,999

Roads

60 to 99,999

F/7V Motorway

100 to 149,999 >= 150,000

7V ,

=

0

10 20 "()

-.ý()

Dual Carriageway Single Road

5.20

I:

I

-loo

ooo

ýo

k'ý,

Figure 5.9:

Woodland recreation demand in North Western Wales: Predicted annual total party visits per site

)2'Th

Key to Towns Caernarfon

C=

BF = Blaenau Ffestiniog B =Bala Dolgellau D 0 Machynlleth M A

Aberystwyth

Roads FZNZ

Dual Carriageway

=

Single Road

Predicted Visits to Woodland Sites per Annum Under35,000

05 tvt:jý-ý

35 to 59,999

10

15

10 2-5 kni -

1: 6 00 00 0

60 to 99,999

As expected, figure 5.8 strongly reflects population distribution in the prediction of In south Wales the influence of cities such as Swansea visits. woodland recreational and in 'valleys' densely populated the area, results Cardiff and relatively high visitor predictions.

5.21

As expected, figure 5.8 strongly reflects population distribution in the prediction of influence In Wales the south visits. of cities such as Swanseaand recreational woodland Cardiff and the densely populated 'valleys' area,results in relatively high visitor predictions. Similarly, in the northeast, the influence of nearby English cities such as Manchester and Liverpool is very clear. Conversely, in mid Wales and western coastal areas,the sparsity of Population impacts depressed in tend to estimates. visitor arrival severely population results be compounded by the distribution of higher quality transport infrastructure. 71is inflates the by large high the proximity of centres of population. generated numbers arrivals already However, infrastructure effects are perhaps best demonstrated in areas of relatively low Wales. Figure 5.9 in detail, density this and north mid shows area coastal, such as population Here that the we can see presenceof a major network. the road major relevant superimposing it individuals facilitates by from heightened corridor as visits potential visitor road creates a distant time travel zones. relatively

5.4: VALUING

PREDICTED DEMAND

TO WOODLAND: SYNTHESIS VISITS RECREATIONAL VALUING 5.4.1: While the visitor demand maps are interesting, they are of limited use for decision demand do this the therefore tell of and cannot they about value not us as making purposes forest In 3 in and management planning. of chapter analysis we directly assist cost-benefit In literature to UK visits woodland. of recreational our valuation regarding the reviewed Benson Willis (1992) ZTC the of and a reworking that present we chapter conclusions to CV Table 5.3 'meta-analysis' studies. reports of per-visit univariate cross-study a and studies from detailed in 4". for those own studies our chapter these alongside results WTP statistics

Table 5.3 illustratesa considerable rangeof valuesfor woodlandrecreation,although L5 indicating exceeds the reported per that estimates party of visit, some none it is noticeable by CV Estimates the are the of range valuation. produced appropriate regarding consensus from variants of the TC. This result is supported by the recent below those consistently in 3A, is to adjustusersWTP per annumby the number on appendix reported "An alternativeapproach, issue how further future discount However, the of respondents such an approach raises visits annum. per of visits WTP We have longer discountingmay term shown that responses. elsewhere or such forming annum per when be hyperbolic 1992) (Henderson (Bateman, than and may even rather heavy exponential et el., and b,p_very Bateman,1995).

5.22

findings of Carson et al. (1996) who assess83 studies providing 616 comparisonsbetween CV and revealed preference(RP; mainly TC) estimates. They report a sample mean CV/RP ratio of 0.8922. I'liese results are in themselves interesting as, in theory TC estimates consumer surplus while CV estimatestotal WT? (consumersurplus plus price paid), in other words we might expect the reverse relationship to hold. This problem is exacerbatedwhen be CV that may capturing users option and non-use values. Reasonswhy also consider we the observed relationship may hold are explored in Bateman (1993). Here we argue that CV responsesmay apply to the on-site experienceto the exclusion of the remainder of the whole trip experience such that travel and time costs are treated as sunk costs. Here the respondent is only considering the surplus over those sunk costs. In effect therefore expecting TC and CV results to be the same may be a category mistake, consideration of which justifies the observed relation of value estimates. Our Thetford 2 CV design effects experiment seems to have shown how far CV inflated deflated by be either or various re specifications of the survey can estimates Adopting a sensitivity analysis approach we can take the upper and lower questionnaire. bound results from this study as our first CBA valuation estimates. As expected our crossCV between is these produces of experiments a value this extremes and analysis study CV To central estimate. as our complement these we use our adjusted adopted accordingly Willis (1992) Benson ZTC from the and estimate of recreation value, and study as a values from the Thetford 2 StUd)r2l Table 5.4 summarises the ITC based GIS estimate our . in CBA (ordered by recreation our wider of woodland values used study range sensitivity value).

2795%confidenceinterval= (0.81- 0.96):median= 0.75. A weighteddatasetrecordsa similarmeanof 0.92 from 1.0. different longer is significantly this no although 23Althoughthe ITCM estimatebasedon perceivedjourneydurationprovidesa slightly betteroverallf it, the Furthermore highly theformermodelproducesconsumersurplusestimates is marginal. difference whicharevery Densonand Willis (1992)results.Accordinglywe prefer to indulgeour own adjusted our those of to similar in belief CIS-based in the thatfuturestudiescombiningperceptionswith CIS measures techniques interests will better models. even produce

5.23

Table 5.3: Valuing recreational visits to woodland: a synthesisof studies Study

Notes:

Method

f1person /visit ,

f/party/visiO me-an

upper 95% CI

lower 95% CI

Bensonand Willis (1992) (adjustedas per Ch.3)

ZTC

IA8

4.52

4.85

4.22

Cross-study (meta analysis)

CV

0.60

1.82

1.95

1.69

Thetford I (low range paymentcard)

CV

1.21

3.69

3.96

3.44

Thetford I (high range paymentcard)

CV

1.55

4.73

5.07

4AI

Thetford I (OLS)

ITC

1.07

3.37

3.61

3.14

Wantage (WTP/visit study)

CV

0.82

2.50

2.68

2.33

Thetford 2 (WTP/visit, no prececdingquestions)

CV

0.20

0.61

0.65

037

Thetford 2 (WTP/visit, after mental a/c questiononly)

CV

OA6

1.40

1.51

1.31

Thetford 2 (WTP/visit, after WTP pa. question only)

CV

OA5

1.37

1.47

1.28

Thetford 2 (WTP/visit after mental a/c and WTP pa questions)

CV

0.78

2.38

2.55

2.55

T'hetford 2 (ML model: GIS based time and Journeycosts)

ITC

1.20

3.59

3.85

3.35

Thetford 2 (ML model: basedon perceivedduration)

ITC

II

IA7

II

4A2

II

4.74

II

I 4.12

n/a = not applicable. 1. Figuresare bestestimatemeans(1990prices).Appendix MA reports95% O's and alternativeestimates basedon WTP per annum,studies. 2. per party per visit measureswere not explicitly reportedin the following studies:Bensonand Willis (1992); cross study CV meta-analysis;Thetford 1, Wantageand Thetford 2 CV studies.In thesecasesper from have been reported per person per visit measuresusing party calculated estimates visit party per in 53A below (adults table given and children being treatedequally in this analysis) statistics composition Tbetford 2 from in detailed 53A Such taken the table survey as as follows. were statistics rates.

Table5.3A:Descriptivestatisticsfor partysize:Thetford2 survey Partysize(no.of persons)

Measure

3.0523 3.2726 2.8468

mcan upper95%CI lower95%Cl

for skewby takinglogarithms, Note:All measures adjusted calculating meanandt-intervalsandthenf inding cxponentials.

5.24

Table 5A Sensitivity analysis woodland recreation values (f/party/visit; 1990 prices). Study

Method

f/party/visit

Benson and Willis (1992): adjusted

ZTC

4.52

Thetford 2 (ML model: GIS based time and journey costs)

ITC

3.59

Thetford 2 (WTP/visit; after mental a/c and WTP pa. questions)

cv

2.38

Cross study CV meta-analysis

cv

1.82

Thetford 2 (WTP/visit; no prior questions)

cv

0.61

5.4.2: MAPPING PREDICTED RECREATION VALUES The number of predicted party visits per annum (illustrated in figure 5.8) was simply demand by into multiplication with the various party visit of a value recreation transformed 5.4. This in in Tdrisi GIS. table the achieved was using scalar command the given values Figure 5.10 illustrates the range of values produced by this exercise. The distribution of values within each of the maps shown in figure 5.10 exactly is due base demand (figure 5.8) factors discussed the to the map and that of same mirrors in detailed in However, 5.4 is the table estimates wide variance value graphically previously. illustrated by figure 5.10. This is clearly a causefor someconcern. While it may be that the 'envelope of valuaflon' (Batemanet al., 1992) described here is sufficient to justify certain decisions, the uncertainty of values illustrated shows that we should be very cautious findings from interpretation Such of any one the study or even any one method. regarding below. upon expanded reservations are

5.25

Figure 5.10:

Predicted value of total annual recreation demand per site using five evaluation estimates

B) ITCM

A) ZTCM (adjusted)

D) CVM meta-analysis

C) CVM upper bound

E) CVM lower bound Predicted Value of Total Party Visits Per Annum Under E60,000 f. 60 - 99,999 000 199,999 E200 - 299,999 > X300,000

5.26

5.4.3: LIMITATIONS

OF THE PREDICTED RECREATION VALUES

While we feel that the recreation value maps illustrate the methodological potential of applying GIS techniquesto economic evaluation of alternative woodland planning options, we conclude this chapter with a brief discussion of a number of potential limitations and further issues which would have to be addressedbefore the full decisionmaking potential of this approach can be realised.

5.43.1: The supply side Our analysis only considersthe demandside of the woodland recreation 'market'. The demand for a typical woodland established the tells us about maps recreation recreation value base intersections (figure 5.7). It does not tell us about the 5krn the the of map grid at any of in 'neir two this major market. are ways of which the supply side interacts with supply side demand to determine actual visits. Firstly, the existing distribution of woodland will already have soakedup some of our predicted demand.Secondly,as new forests are planted and (with become demand becomes lag) services available, so time recreational satisfied. If supply some in demand demand for that non-congested such excess any one area supply exists, so exceeds forestry in nearby areas will be diverted into the surplus supply forest thus reducing latent demand in those diverted areas. Such supply induced substitution effects mean that our recreation demandvalue maps

information. In in isolation judged be of supply side ongoing research' we are cannot including information Bartholomew database, the of such sources remoteexaminingvarious Terrestrial Ecology (ITE) land Institute imagery, database (satellite) the of cover and sensed database Our Commission's Forestry sub-compartment and census of woodlands. eventual the demand both the is the side of and supply woodlandrecreationmarketthus to compare aim identifying areaswheresurplusdemandexists.In the absenceof suchanalysiswe have to distribution forest reasonably uniform apparently an of existing resourcesacross given assume, high demand likely locations for that to of areas are prime prove excess area, the study demandand thereforeforestestablishment.

by EnglishNatureand the ESRC. 24Fundcd

5.27

5.4.3.2: Applicability of the Thetford Forest period to annual conversion factor As part of our arrivals function calculations we had to convert from the survey period is here basis. One factor the concern whether conversion used is valid for onto an annual fully Thetford Forest. In ideally to to test this order we would or unique needdata other sites both Thetford distribution of the visits at and at any site we wish to annual regarding for it is data Thetford Unfortunately for. such while exists only currently being extrapolate Gillam (pers Welsh few for sites. comm)' suggeststhat seasonalitypatterns are a compiled likely to be roughly similar acrossEngland and Wales and only differ in very remote areas Scotland likely be North to where seasonal peaks of are the relatively more such as in information, basis On the absenceof any contrary evidence, this and the of pronounced. '. issue? defensible have approachto this adopted a we feel that we 5.4.3.3: Comparability of recreation in Thetford Forest with that in Wales The major demographic and infrastructure differences which separate Wales from our

East Anglian survey site are explicitly accounted for in out arrivals function which takes distribution. Two distribution density both and quality road and and population account of does our survey site provide similar recreational here. Firstly, issues pertinent are remaining definition, By here is because the map. answer potential visitor yes, our services to those of the looking major service sites wherein recreational the of similar attraction creation at are we is open accesswalking and it's associatedactivities. Analyses such as that given in equation in differences arrival rate which may occur at non-standardwoodland sites. the (5.4) underline differ between East does of recreation perception woodland the psychological Secondly, from In this the supply side problem Wales? this separate out must we considering Anglia and is distinction Once this see made we no reason why such a above. upon commented issues, did Although have no difference we assess not such occur. we should perceptual here, by Colenutt the assertion reinforced an earlier problem work of any reason to suspect Dean which reports similar visitation pattems to those in Forest (1973) the Sidaway of and

"Letter (9th August, 1993) from Simon Gillam, Chief Statistician, Forestry Commission. This letter also for Forest data function Thetford the be amongst arrivals the estimating this as to of was statcd the use supported UK Day Commission Visit Survey The Forestry during April a to undertook available. the most reliable information However, 1993. 1992 such was not available at the time of writing. and September of

'One ad hoc solutionto this problemmight be to conducta sensitivityanalysisto assesstheresponsiveness to assemptions. our predictions of

5.28

in observed our own analyses. 5.4.3.4: Limitations of the predicted recreation values: conclusions In conclusion, we recognisethat our study concentratesexclusively upon the demand if decisions be be issues to to need considered planning are optimal. that side supply side and Ongoing work is addressing this issue. However, we see no further problems with the in Wales function to the arrivals prediction of recreation and our arrivals application of our defensible The that this test estimates. provides valuation of suggests actual predicted versus have briefly here issues length in further demand which we considered and at raises such issue, have Because the surrounding valuation we adopted of uncertainties previous chapters. demand In of a number alternative value maps. producing a sensitivity analysis approach, future chapters we augment these with further forest values before comparison of aggregate in from the study area. agriculture conventional those with values

5.29

REFERENCES Bateman, IJ. (1993) Valuation of the environment methods and techniques: Revealed preferencemethods, in Turner, R.K. (ed.) SustainableEnvironmental Economics and Management: Principles and Practice, Belhaven Press,London. Bateman, IJ., Willis, K. G., Garrod, G.D., Doktor, P., Langford, 1. and Turner, R.K. (1992) Recreation and Norfolk Broads: the of a contingent valuation study, Report to the value preservation environmental National Rivers Authority, pp-403. Benson, J.F. and Willis, K. G. (1992) Valuing informal recreation on the Forestry Commission estate, Forestry CommissionBulletin 104, HMSO, London. Carson, R.T., Flores, N.E., Martin, K. M. and Wright, J.L. (1996) Contingent valuation and revealed preference methodologies: comparing the estimatesfor quasi-public goods, Land Economics 72(l): 8099. Colenutt, RJ. and Sidaway, R.M. (1973) Forest of Dean Day Visitor Survey, Forestry CommissionBulletin 46, HMSO, London. Environmental Systems ResearchInstitute (1993) Understanding GIS: The Arcl/n/b Method, Longman, London. Gatrell, A. C. (1983) Distance and Space: A Geographical Perspective, Oxford University Press,Oxford. Henderson, N. and Bateman, I. J. (1995) Empirical and public choice evidence for hyperbolic social discount discounting, Environmental for intergenerational implications the and ResourceEconomics, rates and 5:413-423 London ResearchCentre (1992) SASPACUser Manual, First Edition, Reprinted by ManchesterComputing Centre, Manchester. Martin, D. (1990) 'A suite of programs for socioeconomicsurface modelling', Technical Reports in GeoInformation Systems,Computing and Cartography 28, Wales and South West Regional Research Laboratory, Cardiff. Whiteman, A. (1991) 'An analysis of forest visitor numbers using household surveys 1987-1991', Research information Note (draft), Forestry Commission, Edinburgh. Whiteman, A. (1995) The supply and demand for timber, recreation and community forest outputs in Great Britain, PhD. Thesis, University of Edinburgh. Willis, K. G. and Benson, J.F. (1989) Values of user benefits of forest recreation:some further site surveys, Report to the Forestry Commission,Department of Town and Country Planning, University of Newcastle upon Tyne.

5.30

Chapter 6: Timber Valuation 6.1: INTRODUCTION In this chapter we assessboth the social and private (farmers') value of timber influences important is One the the outcome the upon of such most analyses of production. long delayed, (even increase Because any returns are small) timber. plantation of real price in real prices will have a major impact upon NPV sums. In order to assessthis, the chapter forestry in UK designed history brief the to acquaint the reader of commercial opens with a increase in domestic breaking (section 6.2). timber trend supply and major with the recent, In the subsequent section (6.3) both the supply and demand sides of the UK market are be drawn. future These balanced can conclusions are prices on view. that a modelled so reinforced by time-series analysesof price movements. Whilst timber value is clearly important, private planting decisions are often determined by the availability of shorter term grantsrather than long delayed felling benefits. In section 6.4 we review the various subsidiesschemesavailable. Section 6.5 brings together information discussions grants, with regarding plantation regarding prices and the preceding base timber the which our to models upon valuations rotation' produce tree growth costs and

are calculated. brings investments in inherent horizons long us to the vexed woodland The time discounting 6.6 discusses Section discounting. the of and principle provides a of question discount both 'correct' literature to the rate with respect social regarding brief review of the We investment conclude that, as no single, clearly correct appraisal. CBA and private identified, be discounting) (or so a sensitivity analysis can discount rate even met4od of for. is approach called Section 6.7 provides investment appraisal results from the viewpoint of a private

limited 6.8 CBA farmer) this to provide a (the social extends of the section while individual dealt in ignoring (i. those this externalities with elsewhere e. plantation a of product timber former being NPV both the In results are and annuity equivalent reported, cases research). being latter forest the comparablewith competing fare the economist while of the usual agricultural outPuts. 'A rotation is the full lifespanof a plantationfrom plantingto felling.

6.1

When commencing this analysis it soon becameclear that assessmentof all possible because both feasible time constraints and lack of data of tree not species was woodland indicated Furthermore, less that costs and preliminary analysis species. popular concerning benefits of different conifers would be reasonablysimila?, the same being (broadly) true of broadleaves. Iberefore, two 'indicator' species were selected for analysis: Sitka spruce (conifer), and beech (broadleaf).

6.2: HISTORICAL

BACKGROUND

6.2.1: PRE-1945 In terms of land use, British forestry has always been the poor cousin of agriculture. Although the prehistoric 'natural' condition of the land was primarily as forest, the influence land Even by been the to to has convert agricultural clear-fell and use. consistently of man 15% England Ibis Book Domesday trees3. of remained under only the time the of last heavy losses for the millennium particularly of with most downward spiral continued husbandry in century when of adoption advanced and seventeenth the sixteenth occurring land forestry to allowed agriculture confine of common enclosure techniques and subsequent latter being the operatedon a non-commercial often parklands, private to marginal areasand 4% 1976). By 1900 (Rackham, only of England and Wales basis for private amenity values being by far lowest forestry, levels in Ireland these the Scotland was under 2% and of and Europe Obid). dependent UK completely the twentieth almost the was upon By the start of century by German This the its exposed was for strategic weakness naval timber supply. imports input With World War. First during timber to the UK's vital a Britain the major blockade of domestic felt timber it supply the that of a strategic was essential creation industry was coal Commission in 1919, Forestry (FC) the the future and, country was of security to the initial FC's Although the this supply security of constituted objective strategic establishe&. further by timber, the the production of commercial as such aims supplemented was quickly benefits depopulation; in the and provision of public stimulation of employment areasof rural 2This is of course a relative statement. Differences do exist and are important at the micro level. However for be defensible for benefits to this the a assumption are similar enough purposes and costs of the magnitudes of this study. 3Pcrs.comm. Colin Price, Dept of Agricultural and Forest Sciences,UCNW, Bangor. *rhe decision to establish the Commission was approved in 1918.

6.2

such as open-accessrecreation and wildlife habitat' (Bateman, 1992). Public sector forestry in the UK has from the outset followed an erratic course. A strong initial political will to establish a securenational timber supply ensuredthat the 1920s were a period of major afforestation, reversing the trend (if not the effects) of the previous millennia. However, as memories of wartime shortagesreceded and world timber prices slumped, the 1930s saw planting figures fall well behind the 30,000 ha annual target FC. This the the of slump was offset to some extent by the creation at envisaged Commission's own promotion of forestry as a responseto rural depopulation trends and a Government initiative "to create a settled force of woodsmen and their families whose livelihood would be enhanced from their own tenanted smallholdings" (Philip, 1976). Nevertheless the 1930s still saw an overall contraction of new planting.

6.2.2: POST-1945 Figure 6.1 illustrates total, FC and private sector annual planting from 1945 to the present day. 6.2.2.1: Public sector forestry The end of the SecondWorld War marked the start of the most sustainedperiod of UK forestry expansion in recorded history (see figure 6.1). Initially, national security highlighted The high strategic policy objectives. again post-war adoption prices concerns and firm the to the prices the and economy, expansion of world timber approach planned of a 28,000 ha in FC to a over accelerated peak of the that per annum planting trade ensured The period from the mid 1950s to the early 1970s was 24,000 ha annually. This was fairly by at planting about sector public stable characterised FC favourably decision low the to Government to by allow operate at a helped rate of a decade following the war.

discount A investments. 3%" rate of only to state was required of the other return compared between 5% 10% for State and to of other rates owned enterprises7. Commission compared

'reduction its defended import has in FC existenceas a sourceof 51nrecent years the also savingsand import (1992) be invalid. the Bateman to substitution shows argument agricultural subsidy. 6EYenlower ratesof returnwere requiredfrom plantingscarriedout in NorthernIreland. From 1989the is but, 6% this as virtually unattainablewithout explicit valuationof nonof ]PCwas set a targetrateof return decisions investment be takenat a 3% rate with the resultant to Treasury new benefits, allowed the market (H. M. Treasury, Subsidy 1991; Annex Forestry G). Felling decisionsremainat being as off written shortfall FC (Adrian Whiteman, appraisal October, systems existing with pers. d. comm. compatibility 5% to retain r. a 1994). Trorn 1989this hasbeensetat 8% for commercialpublicsectorenterprises with a discretionaryrateof 6% benefits (H. M. Treasury, 1991). non-market significant with applied to projects

6.3

Cc 0

E P4 ci

eln



00 00 00

00 clý

rA

0 u

(s. ooo) vinNNV 143d S3HV133H

10

f.)

0 cn

1971 marked a significant peak for the FC with plantings exceeding28,000 ha p.a. for the first time since the early 1950s. However, that year also marked a turning point in the fortunes of UK public sector forestry, beginning a downward trend in planting which continues over two decadesto the presentday. The 1970swere a difficult period for the UK domestic (in the and economic oil crisis problems particular relatively high economy -with inflation and poor trade balances) leading to heavily depressedgrowth rates. This put finance FC immune. Contractions in FC to the which not of public was areas all on pressure in by 1990) 1979 (Thompson, accompanied reductions planting and annual employment i. ha 40% 11,800 1971 level. dropped had to the p. a., e. roughly of planting The election in 1979 of a Conservativegovernment, pledged to the reduction of the in decline favour in in that the meant enterprise, new planting private of seen public sector 1980s day. By been 1993, has the throughout to the 1970s and up present extended the less 1971 fallen However, had the than to of one-tenth peak. a more serious annual planting in 1981 FC arose operations of when an extensive programme of scale threat to the absolute land sales was implemented. In the following year and for the first time since its creation, it land Since date forced than that the overall extent of purchased. FC to more sell the was fallen. Between 1981 1994 150,000 ha FC land has and over FC consistently of estate the 80,000 ha forest. In light the of under sector, which was the the private to of sold were its Commission, in 1993/94 failure, FC the to the review of privatise recent government's has been disposals it is that the programme notable noticeably stepped estate at one stroke, back door'. Table 6.1 by details 'privatisation FC land facilitating the of up as a method Despite disposals the the programme. numerous ministerial period of holdings throughout land by FC has in to the sold access privatisation public safeguarding on pronouncements (Goodwin, 1995). I This is led the to the public"' exclusion of particularly cases all almost high been in has it FC the population areas of where that proportion of serious given been highest" (Lean, 1996). has woodlands privatised sSeestatement by the Secretaryof State reproducedin Appendix V of FC (1985b). 'During 1993/94 the government considered a variety of proposals for the future of the FC. National (at in least a such policy the term). untenable privatisation apparently made short of to the prospect opposifion I0jn the period from October 1991 to November 1995 of 35233 ha privatised only 506 ha (I A%) has had (Goodwin, 1995). guaranteed freedom of access "For example, between 1981 and 1996,91% of FC woodlands in West Yorkshire were privatised; 72% in in Essex (Lm, 1996). However, one countervailing trend in Humberside 43% 53% Kent; in and 67% Durharn; funded (although these are not always open-access)such as those charity woodlands of been the growth has Woodland Trust, which was recently awarded L6 million by the Millennium Commission for by the operated (Smith, 1996). Doorstep scheme Woods its on your

6.5

6.2.2.2: Private sector forestry From the outset, direct Governmentintervention through the agencyof a Stateforestry service has been complemented by the stimulation of a private forestry sector via the provision of tax relief and other incentives to private individuals who invest in timber 12 production .

Table 6.1: ForestryCommissionholdings:GreatBritain 1978-96('000 ha) YEAR

FC PLANTATION

862.5 868.2 884.0 895.7 905.5 908.7 901.7

1978 1979 1980 1981 1982 1983 1984 19852 1986' 1987 1988 1989

AWAITING PLANTING

SCRUB

TOTAL FOREST

TOTAL FC LAND'

945.9

1,253.2

952.2 962A 965.9 954.9 962.7 949.3

1,256.3 1,263A 1,264.0 1,258.7 1,250.9 la09.2

17.2

926.4 919.1 915A

1,156.4 1.149.4 1,144.2

11.2 9.8

909.0 902.8

1,139.5 1,133.1

895.8 887.6

1,127.5 1,115.4

873.8

1,099.5

n/a

1,082.8

etc 83.4

77.0 71.5 63.1 51.5 46.1 39.3

899.7 898.5 898.2

7.0 6.9 7.1 7.9 7.9 8.3

23.4 20.6

PRODUCTIVE

OTIMR3

1992 1993

863.5 858.5 855.3

34.3 34.5 34.8

845.4

37.1

1994

826.6

44.0

5.6 5.1 3.2

19964

n/a

n/a

n/a

1990 1991

Notes:

I= 2= 3= 4=

Total forest + Nursery land + Agricultural land + Unplantablc+ Forestry workers holdings Not available at time of compilation Recreationalland, etc. Not from official statistics

(1979,1985a, Commission 1989,1990,1993,1994a);Lean (1996) ForeStry SoUrCe:

12Detailsof these tax relief schemesare given in Bateman (1992).

6.6

Despite theseincentives,inexperiencemeant that initial private sectorinvolvement was very restrained. However, from the late 1950s a proliferation of firms specialising in facilitating private forestry investments considerably eased the practical problems of such investment. These companieslocated land, arrangedpurchases,planting and felling, and took care of the tax liability and refunding formalities thus allowing those for whom tax relief was an attractive proposition to becomeforestry owners without ever having to visit a plantation or see a tree. In this way post-war planting of private woodlands expanded at a consistently increasing rate from 1945 to the early 1970s(see figure 6.1). However, as with the FC, the 1970s were a period of relative decline for the private forestry sector. As the OPEC oil-shock UK's forest into the recession, so the economy owning elite no longer had the world sent forest into income divert tax-havens. However, these were just the people to taxable excess from Thatcherite boom benefitted the private sector of the 1980's and by 1989 the who its highest level. In at was ever the search for cheap woodlands private of planting destroyed land" (RSPB, 1987). This of many sites great ecological value were afforestable factor, and a national outcry against such tax-avoidance"', causedthe government to act and withdraw such tax-relief`ý The scrapping of tax-relief had an immediate impact upon private sector planting 1989-90. between The it did fall further halved due reason not was primarily almost which (discussed and maintenance planting subsidies subsequently)designed of various to a system in landowners farmers 7liese than those tax-havens. rather search of to and appear to appeal in have a constant annual expansion reasonably private woodland of the order generated to ha 15,000 1990-94. throughout the per annum period of approximately 6.2.3: HISTORICAL

BACKGROUND: SUMMARY

In forestry terms the UK has only recently expandedits domestic supply. Although

it grew rapidly in the post-warperiod,the FC now appearsto be contractingrapidly and the land purchasewas not tax deductible. This led investorsto plant on costs, planting other most , destroying habitats highly to producevery poor but areas. valuable wetland but unsuitable, natural often cheap (RSPB, 1987). plantations highly tax-deductible in a disparagingObserverfront pagemagazinefeatureon the 100 largestforest ownersin 14Culminating See 1988). The Times (1988) Bloom Rosie, (1988). (Lean also and and Britain "Announced in the Chancellor's1988budgetstatement(LTKParliament,1988)but not cominginto effect 1989. late until 23Unlikr

6.7

degree to which this is offset by increasesin private woodland is uncertain. However, the is After 75 years of growth only 10.3%16of the land area of for clear. expansion potential Great Britain is under woodland while 77% is under agriculture. This compares with EC averages of 25% and 60% respectively (FICGB, 1992). Given ongoing and planned in CAP, be therefore that there the conclude we may scope for continued contractions is domestic this timber although supply clearly going to be subject to of expansion direct intervention (both to regard and subsidies) and long term with policy government latter is It to the subject that we now turn. market conditions.

6.3: THE UK TIMBER MARKET

AND LONG TERM PRICES

6.3.1: SOFTWOODS At present the UK consumes some 53 million rný of timber" annually of which is based (FICGB, 1992). (83%) In UK 44 softwood timber comparison rný million nearly 0 6.7 further 8.2 (all at some million stands currently per annurn species) with a production (ibid). This fibre difference between very considerable production recycled of million m3 in being demand timber the fourth largest UK import category domestic and supply results in (ibid). With domestic demand forecast 1991 double in billion E6.3 to the next of at a value 60 years (ibid) and concern rising regarding acid-rain damage to softwood timber stocks (Bergen et al, 1992), some commentators (the 'pro-forestry school'; Doran, 1979) have forecast increasesin future real prices for timber. However, we see two major flaws in the level UK Firstly, the of production representsonly this present argument. supply aspect of domestic dramatic by supply expansion of the ongoing engendered an of the early stages inter-war in from the late 1940's levels the high the and years period of planting sustained into is the This to next century reaching an estimated peak of continue well 1970's. set to (as by 2020's FC tailing the the 20 off early a result of curtailing of rný million nearly 0 decades) 12 by the middle of the to two in a plateau of about million the past planting importantly, 6.2). Secondly, domestic figure (see this expansion of and more next century import increase in by been the of softwood has availability an supplies". echoed supply

'Mis decomposesinto 14.7% in Scotland, 12.0% in Wales and 7.4% in England (FICGB, 1992). in wood raw material equivalent (WRME). 17MCaSUred I&Mis trend is exemplified by the caseof Swedenwhere, since the 1930's, timber growth has consistently 1992). (Wibe, outstripped cutting

6.8

World coniferous roundwood production rose from 1096 million M3 in 197119to a peak of 1307 million m3 in 1986 slipping back only slightly to a level of 1295 million M3 in 1991 20

(Whiteman, 1995). When combined with arguments regarding ongoing technical change , these factors seem to suggest that real prices for softwood are unlikely to increase in the foreseeable future.

Figure 6.2: Actual and predicted UK domestic production of sawn softwood 1945-2055

UK domestic output 1, mber (million M3) 20

15

5

0

1950

1960

1970

1980

1990

2000

2010

2020

2030

2040

2050

Year

Sources:

Timber Trade Federation (1987); Bateman and Mellor (1990); FICGB (1992).

Such an argument would be invalid if the observed increase in world supply were

However, existing stocks. exploitation of while the total extent based upon non-sustainable decreased WWII (see has discussions), land forest since subsequent alarmingly of global ' in forest stocks the major conifer growing countrieS2 have increasedfrom 1593 million ha 1995)22. ha in 1991 (Whiteman, 1648 1971 in million to

'Thesee measurementsarc in underbark volumes.

2"Two forms of technical change can be identified (Bateman, 1988a): (i) improved plantation husbandry; (ii) increased availability of firnbcr substitutes (particularly in the construction industry, see Leigh and Randell, 1981). 2'Former USSR, Canada, USA, Sweden, Finland and Norway. 2'This argument uses a simple definition of sustainability, namely that overall resource size should be nonbe it that the conifer plantations underpinning these statisfics are degrading die natural However, may declining. in which thcy are grown. This is an important issue but introduccs a further level of coniplexity environments to address within this research. unable were we which

6.9

Concerns about acid-rain initially appearto be better founded. Table 6.2 details rates from defoliation European arising acid-min, showing that this is particularly serious in the of UK 23 .

While the defoliation problem appearswidespreadthere have been few studies of its 4 (1986) impact. Ewers et al. estimate a net present value? of expected future economic damagesto German forests of DM 1.2-1.5 billion (roughly E500 million) for the period from

1983to 2060. However,whencomparedto a presentsoftwoodproductionvalueof over flO billion annually from the six largest producersalone (Whiteman, 1995) such damagesdo not appear likely to undermine an otherwise expanding global softwood resource. Table 6.2: Acid rain defoliation rates in Europe 1986-921 1 Participating countries Belgium Denmark Fran=2 Gemany3 Greece Italy Luxembourg Neth rlands Portugal Spain United Kingdom

All s

ies, defoliation classes2-4

1988

1989

1990

1992

1992

No. of sample trees

23.0 9.7 17.3

18.0 6.9 14.9 17.0

14.6 26.0 5.6 15.9 12.0

16.2 21.2 7.3 15.9 12.0

7.9 21.4

10.3 18.3 1.3 7.0 25.0

12.3 16.1 9.1 3.3 28.0

12.3 16.1 9.1 3.3 28.0

17.9 29.9 7.1 25.2 16.9 16.4 20.8 17.2 29.6 7.3 56.7

16.9 25.9 8.0 26.0 18.1 18.2 20.4 24.5 22.5 12.3 58.3

2,384 1,558 10,113 103,422 1,912 5,857 1,152 32,875 4,518 11,088 8,856

1987

22.0

% change 199102 -1.0 -4.0 0.9 0.8 1.2 1.8 -0.4 7.3 -7.1 5.0 1.6

Notes: Percentagesof trees surveyed in defoliation classes24: damageclassesranging from moderate to severe 1. Change in sampling procedure in 1988 2. 3.1986-90 only includes Western Germany

Source: Pearce (1993) adaptedfrom ECE (1993) Given these factors, significant increases in real softwood prices seem, a-priori,

decreasing does Indeed the of We real prices possibility not seem unfeasible. unlikely. hypothesis in formulated tested constant a null of real prices which two ways: we therefore

23Notethat up until 1993/94the UK systemof defoliationclassificationdiffered from that usedin Europe. has by FC EC the the significantlyreducedtheapparentUK defoliationrate. classification of The recentadoption (which FC the the system retainsalongsidethe EC approach)defoliationof several old However,even using 1993/94 (FC, during 1994b). declined irnportantspecies 24Usinga real discountrateof 2%.

6.10

firstly, an econometric model of the UK softwood market was constructed and investigated; secondly, a time-series analysis of real price trends was conducted. 6.3.1.1: An econometric model of the UK softwood market

Models of UK demand,domesticsupplyand importswere formulatedfrom datafor the period 1946-86(full detailsof this analysisareprovidedin appendix4.1). Tlieseshowed that prices were linked directly to the world marketratherthan to domesticsupply which formed a minor part of overall consumption. As domesticsupply has increasedit has illustrated for imports25as in figure 6.3. substituted progressively The UK timber market:impactof increasesin domesticsupply

Figure 6.3:

World price

Domestic pnce (UK) --

World market

UKmarkat

S.

pQK

P.ý,

OUK

Ouantity (world) 0 HO Uptake of grant and tax incentives (time proxy)

h"

IVIC f,

Reduc in imports

Ouantity 4-of imports

10

OoTQuantity (UK)

H,

Increases In home production

0

HO

H,

Quantity of home p(oduction

Source: Bateman and Mellor (1990)

investigation it fitted data this that the showed of model well but that no statistical be found (details between in link 4.1). Our supply and prices could appendix significant

(1992) Showsthat if non-marketbcnefits areexcluded this intervendon-inducedimport substitudon 2513aternan failure. However, it is inclusion of non-market benefits reverses that the a market argued to constitute appears this result.

6.11

timber market analysis therefore fails to find any empirical (or strong theoretical) support for increasing real prices.

6.3.1.2: Time seriesanalysisOf sOftwoodprices A numberof time seriesanalyseswere carriedout to test whetherreal priceswere increasing or not (full details of these analysesare given in appendix 4.2). Inspection of long-run price indices27suggestedthat real prices had remained relatively stable during the for a supply side shock occurring in 1974 arising from a coincidence except post-war period of unrelated factorsý'. A simple initial text of this hypothesis was undertaken by fitting real prices against a constant. Ibis highlighted the unusual nature of the 1974 peak which was dummied to detailed in (6.1). equation the model produce 80.98 + 95.01 SHOCK (9.86) (36.65) 84.2% regressionF= 97.28

ISSRPI, k2(adj)

where

(6.1) p=0.000

Imported sawn softwood real price index in year t: 1946-86 (1975 = 100) in 1974 (commodity price boom); 0 otherwise SHOCK =I Figures in brackets are t-statistics. ISSRPI

=

An additional time trend variable was added to equation (6.1) but this proved to be highly insignificant (t-value = 0.75). Therefore, although simple, this analysis gives strong hypothesis during for the post-war period. the of constant prices real support

An alternativeapproachis to usetime seriesmodels(Pindyckand Rubinfeld,1981). Various autoregressive, moving average and ARIMA models were estimated, with the

in being (6.2). equation reported strongestmodel

211na similar recent supply and demandanalysis,Whiteman(1995) also concludesthat constantreal forecast. further In Whiteman (ibid) the most plausible a study, appears analysesplantingand softwoodprices from to the trend examine whether ongoing away exploitation and towardsmanaged costs management Again impact best is be the to that real priceswill remain real prices. estimate upon will shown plantations future. foreseeable into the constant "Preparedby Adrian Whitemanat the ForestryCommission,Edinburgh. included 2sFactors exporters major a particularlysevereScandinavian affecting winter,political unrestin the USSR and industrialdisputesin Canada(Colin Price,pcrs. comm.). Thesewere compoundedby increased OPEC the of result oil crisis. a as transportcosts

6.12

ISSRPI, = 56.49 + (22.61) Mean df Residuals: SS = NIS = where:

0.309 ISSRPI, + 0.523 + e, -, (1.44) (2.72) 81.75 (s.d. = 3.62) 37 3982.86 (back forecasts excluded) 107.64

(6.2)

in error year t e, = estimation

The moving averageelement (e, ) of equation (6.2) shows that ISSRPI, is related via -, a less than unitary coefficient of previous period prediction err-or. Furthermore the has (ISSRPýj a statistically insignificant impact upon present real element autoregressive is before As the strongest predictor provided by the constant, supporting the prices. hypothesis of constant real prices in the data period. A number of other commentatorshave examined this issue, the majority concluding in favour of a constantreal prices assumption(Doran, 1979; Price and Dale, 1982; Pearceand Markandya, unpublished). In a recent in depth analysis Whiteman (1995) undertakesa time from 1870 1989. Figure 6.4 illustrates to prices of real softwood this series series analysis WWII due WWI 1970's the to and and mid supply side shock referred peaks clear showing to earlier. Figure 6.4: The real price of sawn softwood imported into the UK' (1870-1989) 1t

1

co

x

W

a.

luuw

1. ýWw

1-

. --%,

Year Real price series 15 year moving average is by dictated UK domestic the world price the discussed As market previously 1. Note: Source: Whiteman (1995)

6.13

Whiteman's best fitting time series model of this period indicates stable real prices (excluding shocks) prior to VMI, undergoing a shift to a higher level during the war and higher, but again constant (excluding shocks), level after the wa?". at a remaining Whiteman's best estimate is therefore for a constant real softwood price for the foreseeable

futuroo. 6.3.1.3: Real prices for softwood- conclusions Both our theoretical and empirical analysesgive no support to the hypothesisof future increasing real prices for UK softwoods and an assumptionof constantreal prices is therefore highlighted be However, should caveat one regarding the useof time-seriesanalyses adopted. to support such assumptions. These analysesare only strictly valid for forecasting if both demand Future conditions remain supply and reasonably stable. shocks might overall destabilise the existing market causing unforeseenreal price changes. 6.3.2: HARDWOODS While global reservesof coniferous forest have beenreasonablystable or even grown decades, has decline in the two temperatehardwoods post-war era seen some the past over in hardwoods. fall Considering first dramatic UK hardwood highly tropical the case of a and has dramatic in decline the area of semi-natural seen period a the post-WWII reserves, hardwood woodlands. In England and Wales such woodland has more than halved from 142,000 ha in 1933 to 76,500 in 1983 (NCC, 1984). Ille bulk of this loss has been through losses the to conifer plantations with remaining mainly generally attributable to conversion (NCC, 1984; CPRE, 1992). However, in terms of overall area, encroachment agricultural broadleaved woodlands have actually increased since WWII as a result of new planting 6.3). in 1980's (see the table occurring particularly While newly planted broadleaved woodland does not have the ecological value of does However, it in trend. represent the case of an encouraging as ancient woodlands,

in hardwoods. far from is UK sufficient the self softwoods,

29Thisfinding concurs with our own supply and demandand time-series analyseswhich were TCStriCtrdto from Whiteman loss this WWII the cxpWns as arising shift wartime of the British Empire (and period. the post. inherent favourable terms of trade) which arose as a result of WWII. 3'Whiteman notes that, even if the best fit time seriesmodel is disregardedin favour of a simple straight line figure (see 6.4), dataset future 1/2% this would only support a modest the critire real rise of about p.a. through

6.14

Table 6.3:

High forest by general species:Great Britain ('000 ha)

Forest type Mainly coniferous high forest Mainly broadleaved high forest rTotal high forest . Notes:

1947

1965

1980

1994

397 380

922 352

1317 564

1516 615

r 777-

L 1274 1881

Figures for 1947,1965 and 1980are from the occasionalCensusof WoodlandsTC, 1987;reproducedin Pearce,1993). Figuresfor 1994arefrom FC (1994a)and include someextrapolationfrom the 1980Census.

Present levels of UK domestic hardwood (round and sawn) production are about 0.9 This 1992). (FICGB, compareswith present demand of approximately m3annually, million 1.9 million m7 per annum (ibid). While this representsa present self sufficiency rate of for higher (i. 50% than much softwoods), this still means that the UK market is e. nearly highly import dependentand consequentlysubject to fluctuations in the world market. Global stocks of hardwoodshave fallen dramatically in the post war period, primarily in developing, deforestation in the tropical the of countries of world which they result as a The is deforestation direct this cause of principle a as result of population are predominant. by LDC exacerbated widespread and growing poverty, has led to an which, pressures increased demand for fuelwood for basic energy needs(World ResourcesInstitute, 1994)31. Two further pressures upon supplies have been developed world demand for tropical hardwoods" (Whiteman, 1995) and forest burning for agricultural expansion33 (Myers, 1990). Table 6.4 details deforestation in 26 major hardwood producing tropical countries

(which includes estimatesof the carbon releaseengenderedby this deforestation- see discussions in chapter 8).

"World productionof non-coniferous roundwoodcurrentlystandsat roughly2 billion m' perann= (1991 r,j;u,cs; forecastto rise to 3 billion by 2011). Of this about`14is consumedin developingcountries,mainlyas fuelwood (figuresfrom Whiteman,1995). 32Majorexport marketsare (in orderof magnitude)the USA, Japan,ChinaandEurope(Collins, 1991). 33'rhislatter factor providesa significantsourceof greenhouse gasemissions.7beseare quantifiedon an 6.4. in basis table annual

6.15

Table 6A

Tropical deforestation to 1989 (inclusive)

Country(plusarea)

Original forestcover (km)

Present forestcover (km)

Present primary forest (km)

Current deforestation 1989 km2p.a.

Bolivia (1,098,581) Brazil (8.511,906) Cameroon(475,442) CentralAmerica(522,915) Colombia(1,138.891) Congo(342,000) EcuadorC270,670) GabonC267,670) (469,790) Guyanass India (3,297,000) Indonesia(1,919,300) Ivory Coast(322,463) Kampuchea(181,035) Laos (236,800) (590,992) Madagascar Malaysia(329,079) Mexico (1,967,180) MyanmO(696,500) Nigeria (924,000) Papua7(461.700) Peru(1,285,220) Philippines(299,400) Thailand(513.517) Venezuela(912,050) Vietnam(334.331) Zaire (2,344.886) Totals

90,000 2,860,000 220,000 500,000 700,000 100,000 132,000 240.000 500,000 1,600,000 1,220,000 160.000 120,000 110,000 62,000 305,000 400,000 500.000 72,000 425.000 700,000 250,000 435.000 420,000 260,000 1,245,000 1 13,626,000

CarbonReleasein 1989

million tonnes

70,000 45.000 1,500 (2.1) 2,200.000 1,800,000 50.000 (2.3) 164,000 60,000 2,000 (1.2) 90.000 55,000 3.300 (3.7) 278,500 180,000 6,500 (2.3) 90,000 80,000 700 (0.8) 76,000 44,000 3.000 (4.0) 200.000 100.000 600 (0.3) 410,000 370,000 Soo (0.1) 165.000 70,000 4.000 (2.4) 530,000 12.000 (1.4) 860,ODO 4.000 16,000 2,500(15.6) 20,000 67,000 500 (0.8) 25,000 68,000 1,000 (1.5) 24,000 10,000 2,000 (8-3) 84,000 157,000 4,800 (3.1) 110,000 166,000 7,000 (4.2) 245.ODO 80,000 9.000 (3.3) 28,000 10,000 4.000(14-3) 180,000 360,000 3.500 (1.0) 515,000 420,000 3,500 (0.7) 50,000 8,000 2,700 (5.4) 74,000 22.000 6,000 (8.4) 350,000 300,000 1.500 (0.4) 60.000 14,000 3.500 (5.8) 1.000.000 700,000 4,000 (0.4) 1 1 5.321,00(? 138,600(1.8) 7,783,5W

Notes: 1. 2.

Equals 97 per cent of estimated total original extent of tropical forests, around 14 million km2 Equals 97.5 per cent of presenttotal extent of tropical forests, 8 minion km'

3. 4. 5. 6. 7.

Equals67 pcr centof total remainingtropicalforests,9 minion km2 omits countriesnot on this list asminor FrenchGuiana,GuyanaandSurinam Burma papuaNew Guinea.

(% of totair

14 454 28 30 59 10 27 9 4 41 124 36 5 to 28 50 64 83 57 36 32 28 62 14 36 57

1.0 32.1 2.0 2.1 4.2 0.7 1.9 0.6 0.3 2.9 8.9 2.6 0.4 0.7 2.0 3.6 4.6 5.9 4.1 2.6 2.3 2.0 4.4 1.0 2.6 4.1

1.398

100.0

Source:Myers (1990)

Annual net hardwoodextractionrateshaverisen from 0.8% at the endof the 1970's (Doran, 1979)to 1.8%a decadelater (Myers, 1990);ratesthat meanthat by 2010only Brazil have (Collins, 1991). Given areas of rainforest Zaire significant any remaining that will and for biodivcrsity, has led to thi's the global environment richest rates already represent these in history (MacNeill, 1990; World the the of world unprecedented extinction of species Pearce Warford, 1993). 1994; institute, and Resources

6.16

Setting aside the terrible ecological consequencesof this destruction, the nonsustainability of such loss meansthat current hardwood supply levels cannot be maintained. Compared with a growth in UK demandof up to 3.5% per annum (Hart, 1987) the potential exists for increases in real price levels. In a review of such arguments, Bateman (1987) from 0.5% 4% 'pro-forestry' to ranging estimates annually'. reports

However, this is

balanced by opposing views such as that of Whiteman (1995), that while demand will increase, "it should be possible to improve forest managementto meet thesedemands,which Certainly keep hardwood timber the stable". then relatively prices rate of growth of would 6.3 in detailed table means that the current rate of UK self sufficiency will rise planting domestic buffer from future a element of some any providing reduction in global considerably supplieS35. This is clearly an area of uncertainty and disagreement. While we feel that there is increases for for hardwoods have than case real price softwoods, stronger we a considerably here adopted a zero real price rise assumption on the grounds that any alternative rate has long delayed benefits for hardwood timber the of plantations and that major consequences in light to the easily revise our us more calculations of subsequent allows rate using a zero improved information.

6.4: GRANTS 6.4.1: HISTORICAL

BACKGROUND

Given the long delayed nature of forestry returns, government incentives have always decisions. The incentives in UK earliest private-sector planting coincided played a major role in FC 1919, the of a scrub clearance and ground preparation when, with the establishment introduced A in 1927, introduced. scheme, grant second planting establishedan grant was broadleaves be for to given preferential subsidy rates over conifers"', trend enduring forestry objectives of non-strategic/production within recognition policy. early reflecting an Following WWII a variety of FC administered schemeswere introduced. Through

(1979); (1967); Doran Bumham(1985)and Hart (1987). Johnston from 3"Esdmates et al are 35Aithoughsuch pressureswould still act upon prices if global demandbecamesignificantly supply constraincd. 36ThiSplanting grantpaid f2/acre for conifersand Wacre for hardwoods(Johnsonand Nicholls, 1990).

6.17

examination of these we can see a gradual movement in forestry policy objectives from simply maximising timber production (e.g. the Dedication Scheme:Basis 1) to schemesgiving equal emphasis to timber, environmental and recreational goals (e.g. Dedication Scheme: Basis HI; Broadleaved Woodland Grant Scheme). Table 6.5 charts the developmentof these

schemes. Table 6.5: Forestry Commission administeredgrant schemes1948-85 Inaugurated

Closed

Dedication Basis 1 Dedication Basis 11 Dedication Basis 111

1948 1948 1974

1972 1972 1981

Small Woods Planting Grant Approved Woodlands Scheme small Woods Grant

1950 1953 1977

Forestry Grant Scheme Broadleaved,Woodland Grant Scheme

1981 1985

Grant

Structure

Annual grant Planting grant (from 1977 also a management grant) I 1971 For smaller areas 1972 Planting grant 1981 Woodland areas between 0.25 ha and 10 ha 1982 1988

Planting grant only Planting grant only

Source: Johnson and Nicholls (1991)

While grants were important, as discussedat the start of this chapter the overriding

force behind the expansionof privatesectorforestryin this periodwastax concessions.The has in 1988 Budgee' thrust the the these concessions of role of grants most of scrapping likely be is in it to the these which main motivators the are of any expansion and centre-stage foreseeable future.

6.4.2: PRESENT SITUATION 6.4.2.1: Forestry Commission administered grants

Throughoutthe 1980's the FC crnphasisedits reorientationaway from the simple (FC, 1985c). Such towards and output wider objectives timber policy was pursuit of

37See Lynch (1989)for a reviewof the presenttax situation.

6.18

introduction, in 1988, Woodland Grant in Scheme(FC, 1988b). This was the the of embodied designed to encouragethe multi-purpose use of woodland with the following objectives:

i.

to encouragetimberproduction;

ii.

to provide employment in areasof rural depopulation;

iii.

to enhancelandscape,createwildlife habitat and provide longer term recreation and sporting facilities;

iv.

to encouragethe conservationand regenerationof existing woodlands.

Rates of support under the WGS were revised in 1990 as detailed in table 6.6. Table 6.6: Woodland Grant Scheme(F-/ha)

Area planted 0.25-0.9 ha 1.0-2.9 ha 3.0-9.9 ha 10 ha+

Conifers

Broadleaves

1005 880 795 615

1575 1375 1175 975

Source: Johnsonand Nicholls (1991)

Payments under the WGS are made in three instalments: 70% at planting, 20% after

5 yearsand 10% after a further 5 years(subjectto satisfactoryestablishment).In addition for is Supplement (BLS) Land Better plantingon arable/improved grassland payable to this a is E400/ha 10 BLS for (including the previous years. conifers within ploughing) cultivated broadleaves, for all at planting. payable E600/ha or in introduction 1992 by the this Furtherenhancement of the of packagewasprovided This Grant (WMG). WGS, Management to the an annual provides addition Woodland in for down 10 first the return setting years of establishment the and execution after payable increase designed the to the environmental plans value of 5-yearly woodlands management of details WMG payments. 6.7 Table concerned.

6.19

Table 6.7: WoQdland ManagementGrants (:C/haper annum) Period of eligibility (age of wood in years)

Rate of grant (F-/ha/pa)

Standard WMG' Conifer Broadleaved

11-20 11-40

10 25

Special WMG'-2

II onwards

35

Supplement for small WoodS3 Standard: conifer Standard: broadleaf Speciae

11-20 11-40 II onwards

5 10 10

Type of WMG

Notes:

1.

2.

3. 4.

All these grants are also payable as additions where the owner is a farmer under the Farm Woodland Scheme,as compensationfor agricultural output foregone (not againstestablishment costs). Higher rates are available for woodlandsof special environmentalvalue (nature conservation, landscapeor public recreation). The owner will be expectedto maintain the wood's character. Thesegrantsareavailablefor any forest of any ageover 10 years,however.they may be extended to youngeror evenproposedforest if the ForestryCommissionis satisficd that thereis demandfor such a provision. Available as additionsfor all woodlandsof less than 10 ha (of correct age). Available for any woodland (over 10 years)of less than 10 ha where the woodland is of special environmentalvalue.

(1991) Nicholls Johnson Source: and

1991 also saw the FC introduce the Community Woodland Supplement (CWS), a

further additionto the WGS (andWMG) designedto promoterecreationalwoodlands"within in for the town opportunities an area where or city and woodland a 5 miles of the edge of (FC, 1993) has implementation been 1991). In (FC, limited" this pers. comm., recreation are just few broadly thousand that of a communities small people are so relatively very translated CWS. The justify to scheme consists of a single payment payment of sufficient considered for CWS WGS All allowed qualifying were and E950/ha planting. woodlands at payable of 'special' being latter rate. the enhanced paid at WMG, the

In additionto the above,from 1992the FC offereda singleflOO flat ratepaymentfor drawing (irrespective the on of size) provisional up of a managementplan woodland new each (FC,

1991)38.

Min additionto this theFC alsoprovidescertainothergrantpaymentsfor generalandcoppicemanagement, in Details Johnson Nicholls (1991). control. are given and squirrel grey and open spaces

6.20

6.4.2.2: Ministry of Agriculture, Fisheries and Food (MAFF) grant schemes In 1988 MAFF introduced the Farm Woodland Scheme (FWS) to provide annual income support" to farmers who establish woodlands on what was previously agricultural land (MAFF, 1987y.

The schemehas almost identical objectives to the FC's WGS (and

is payable concurrently) with the additional goal of reducing surplus agricultural production. As a consequencehigher rates of FWS are payable on better quality land. Although these broadleaf* between distinguish do conifer and woodlands, the period of annual not rates latter. for is longer the support Initial expectationsregarding the impact of the joint FWS/WGS packagewere mixed. While Kula (1989) felt that the new scheme would "no doubt encourage a lot of new investors into the industry", our own analysis(Bateman, 1988b) suggestedthat the initial rates In feel the out areas of agriculture". event marginal we very attract only of grant would (1990) Fearn few borne that, the reports out. exception of a with that our predictions were the scheme attracted very few farmers entering the scheme for non-economic reasonS42, initial its during period. applicants Consequently, in 1992, the FWS was replaced by the Farm Woodland Premium Scheme (FWPS) (MAFF, 1992a,b,c). Here farms first applied to the FC for planting grants CWS BLS, WMG, (including WGS the new and single woodland payments where the under for farm FWPS detailed MAFF If to the then payments could as apply appropriate). approved

in table 6.8. For woodlands consisting of less than 50% broadleavesthe FWPS is payable in each is 15 for 10 to first extended which years after planting, a period mainly years of the interest if land is However, are stipulated with grant repayments broadleaved woodlands4'. 30 for latter former (MAFF, 20 for the the and years years returned to agriculture within impact fully FWPS. it is the to the At too assess early of time still the of writing 1992b).

3Me FWS also pays planting grants but, since its revision in 1992 theseare identical to thoseoffered under both FWS and WGS planting grants. Farmers WGS. collect not may the 'qnitial rates of FWS and contemporary WGS levels are detailed in Bateman (1988b). 4'For example, the initial WGS was calculated to only outperform sheep-stockingdensities of less than one 1998b). (Bateman. ewc/ha "The prime reason for entry to the initial WGS was for ornamental planting around farm-gardens(Fearn, farmers favourcd have ironically who would probably rich 7bercfore. the undertaken such scheme 1990). planting anyway. OThis is considerably more front-loaded than the original FWS which provided lower annual sums but over 1988b). Bateman, (see a longer period

6.21

Table 6.8:

Annual payments under the Farm Woodland Premium Scheme' (f/ha per annum)

Lowlands

Presentuse

Lessfavouredareas DA

SDA

Arable improved grassland

250

190

130

Unimproved

n/a

60

60

Notes:

n/a = 1.

not available The following FWPSrestrictionsapply: (i) not more than50% of farm eligible; (ii) not more than 40 ha of unimprovedland per farm; (iii) eligibility for arable/improvedgrasslandrestrictedto land under suchusagewithin the previous three years;(iv) the FWPS as a whole is cashrather than in MAFF (1992b). Further details limited. are given area

Source: MAFF, (1992b)

Farms may also convert land into forestry under the Common Agricultural

Policy

(CAP) Set-Aside scheme. Set aside woodland is not eligible for either FWPS annual be WGS However, BLS. WGS may received concurrently standard payments or payments For high in this the the of study, purposes set-aside productivity areas. set-aside with indicate little Johnson Nicholls (1991) in that the and very of area under reported guidelines for WGS (Wales) and set-aside qualify concurrent payments and would we consideration further. do this particular any permutation pursue not consequently 6.4.2.3: Other schemes and regulations With respect to our Welsh study area, the creation of the Cambrian Mountains and

in 1986 (ESA's) 1987 Areas (respectively) Sensitive Environmentally and Peninsula Lleyn for broadleaved further Welsh Office woodland. grants the of possibility seemsto offer features Cambrian importance MountainsESA while the the of such within (1989a)stresses E45/ha Office, 1989b) (Welsh leaflet of payments per annurnare specified in a subsequent These are clearly specifiedas additionsto existing woodlands. such of for management Lleyn However, publications the regarding grants. subsequent planting and management (E15/ha lower ESA rates of grant per annum)restrictedto existingbroadleaf offer Peninsula b). Conversations Office, 1992a, both ESA indicate (Welsh with authorities woodlandalone 6.22

that these anomalies still persist today. Further grants towards the costs of promoting landscapeor countryside conservation Countryside Commission by the and Nature Conservancy Council paid are occasionally (Johnson and Nicholls, 1991) while ADAS can provide certain technical support. However,

the occasionalnatureof suchprojectsmeansthatthey arenot consideredfurtherin this study. A factor which may become of increasing importance is the developing planning framework around woodland. During conversationswith the FC", the author was informed felling be licences farmers granted subject to a replanting order. While future only that would farmers in land this that means once effect place change, under trees they may may policy bound legally farm in become to the an equal area of maintain woodland on perpetuity. well This irreversibility, if it became widely known, would we believe, be a major brake upon farm-woodland expansion. However, as such licences will not be sought for the'best part or information be lag in decades, this system. there a significant may two Another important development is the increasing involvement of the Department of in (DoE), forestry Environment the national planning authority, ultimate expansion. the Concernsregarding the aestheticand environmentalimpact of monocultureconifer plantations led to an announcementin March 1988 that planning permission for such plantations would for in England. The Wales be to sites possibility of extension granted and not normally Scotland still exists although the subsequentchangesin tax-law have lessenedthe pressure for such developments. This slowing of conifer expansion was also given a European dimension in the same year with the implementation of the Environmental Assessment (Afforestation) Regulations. These rules, derived from EC Directive 85/337, state that any be FC for to submit an environmental assessmentof the may required assistance applicant have become routine requirements for In forest. practice such assessments proposed Nature National ha National Reserves, Parks, Sites 100 affecting of plantations of over has for further EC funding Ile Interest, the Scientific etc. also provided Special of the Countryside Commission schemesmentioned above.

6.4.3: Grants; conc usions Farmers considering diverting land into forestry are eligible for a variety of grants and

"In particular with the Chief Forester, Santon Downham, Tbetford Forest, April 1993.

6.23

subsidies. These vary considerably according to which schemethey register under and, to a lesserdegree,upon locational factors. In the following section we incorporate thesesubsidies from arising plantation management. the wider costs and revenues within

6.5:

PLANTATION

COSTS AND REVENUES

6.5.1: CHOICE OF SPECIES Ideally one would wish to analyse all those specieswhich are likely to be used in a feasibility forestry. The investigated from to of such an analysis was agriculture conversion by use of the FC's Forestry Investment Appraisal Programme (FIAP). However, while this is an excellent tool for the managementof given stands,it was not amenableto the type of by Consequently it this the posed to questions research. answer was required modification decided that two representativespecies,one conifcr the other broadleaf, would be chosenfor study. Amongst the eight major speciesof conifer grown commercially in the UWI, the Sitka spruce stands out as by far the most dominant constituting 28% of total forest area, is (FICGB, 1992). The double capable of producing an species species other any more than 24mý/ha UK typical over an optimal with rotation productivity of yield average annual felling agescan be very dramatic This 12-160/ha. that optimal growth rate means averaging Choice little Sitka 45 60 to sites'6. from on good as as of ground years poor on years short, logical therefore a and often observed reflects conifer species spruce as a representative is be in However, thought to terms of this not optimal species choice. timber-productivitY

recreationvalue. Interestingly there is little empirical evidenceregarding a connection between species

Hanley few Ruffell In to this, the consider studies and one of valuation and recreationvalue. here. This isolate fail a significant relationship may mean that all woodland (1992)47 to for than rather outdoor, specifically values are observing studies valuation recreation 411ndmending order of total forest area, major conifers are: Sitka spruce (28%); Scots pine (13%); (2%); Douglas larch (6%): Corsican fir Norway (6%); (2%); European Japanese (7%)*, spruce pine lodgepole pine (4%),, (4%). (9%)-, beech The (4%); birches hardwood Major ash oak are: remaining area consists (2%). larch 1992). (FICGB, diversity species of of a "As discussed subsequentlyoptimal felling age is a function in part of discount rate rather than just of growing conditions. 47See reVieW in appendix 1.

6.24

woodland, activities. However, if we temporarily lurch from the empirical to the anecdotal, it is the authors firm belief that walkers do recognise and appreciatethe difference between the claustrophobic atmosphere produced by a species like Sitka spruce (with its dense entanglement of lower branches,tightly packed together to maximise timber yield, set in a bed of stultifying acid pineneedles)and, say, the much more airy and open feel of a Scots difference is An clearer evident when we then consider the gorgeous even pine woodland. foliage, beautiful trunks and verdant undergrowth of an oak woodland. and spaciousness, To allow for this difference we determined to extend our appraisal to consider a is difficult Here is hardwood. the the the most abundant choice more as oak representative broadleaf speciesbut is relatively slow growing and less productive than the beechwhich we decided to study as a more viable hardwood alternative. 6.5.2: SITKA SPRUCE COSTS AND REVENUES 6.5.2.1: Costs Irrespective of species,the majority of plantation costsoccur at the start of the rotation (planting, etc) and at felling. Here we make the common FC assumptionthat all cutting costs (both thinnings - the extraction of undersizedtrees at set points during the rotation so as to felling) by incur long and yield are either carried out contractors plantation or run maximise This implicit the costs upon plantation operator. allows us to use the standing contractor-level 49 ignore (see these subsequent sections) and effectively costs timber price-size curve Remaining costs are detailed in table 6.9. The costs detailed in table 6.9 will vary in individual casesaccording to spatial factors in input distance local (including infrastructure, to supply sawmills, variation prices such as incorporated for Typical are within the basedata of table labour), etc. suchparameters values 6.9 (FC, 1987). They will also vary according to the intensity of planting. Here, typical 2m being trees apart. planted with chosen are parameters

"As a side analysis we produced the following model of felling costs from data given in Hart (1991). 8.98 - 0.145 YC COST = where:

COST YC le(adj)

clearfelling cost (flm' at 1990/91 prices) yield class (see subsequentdefinition) 69.3% F= 21.33

6.25

p=0.000

Table 6.9: Conifer plantation costs (f/ha, 1990 prices) Cost

Years from plantin'

(f/ha)

0

647.45

1 2 3 4-6 7 8 9 10 11-19 20 21-(F-1)

174.86 101.08 68.19 22.06 114.94 54.42 22.06 60.17 22.06 109.20 22.06

Cost items

Construct maintenance roads, plough land, other land preparation, initial drainage, new fencing, purchasing plants, planting, initial fertilizer. Beating upý, weeding, maintenance Weeding, maintenance Weeding, maintenance Maintenance Cleaning, maintenance Subsequent fertilizer, maintenance Maintenance Respacing, maintenance Maintenance Pruning, maintenance Maintenance

in incurred first is (varies E285.03 the the thinning year preceding In addition a cost of forwarder for transporter and of roads points. construction rate) to growth according Notes: 12.

3.

Planting year = year 0; felling year = year F (F varies across growth rate and discount rate, see subsequentdiscussions). Costs such as road construction, ploughing, etc are based on average Base data be tasks that also considers necessary'. such will probabilities incidence of unplantable areas(rocky outcrops, etc). For explanation of this and other tenns see Hart (1987,1991).

(1987,1991). (1987); Hart Bateman (1987); FC Sources:

6.5.2.2: Timber revenues

As with mostenterprisesthe generalobjectiveof the forestmanagercan be assumed factors With two timber to output, are of regard particular be maximisation. profit to discount. in However, (ii) (i) the here: of rate order to the growth; rate of importance first hold both is it factors to impact constant these and consider useful a of the understand

single plantation. in for they ideal as are, economic material analysis microcconomic terms, Trees are increasing initially diminishing and subsequently systems exhibiting 6well-behaved' production 6.5 Figure MP (MP) typical timber shows a curves. and corresponding product marginal intersection defines The (AP) the point curve. maximum average annual product average

6.26

increment in volume which a stand can deliver (in this case l2m3/ha per annum), otherwise known as its yield class (YC).

Figure 6.5:

Marginal and averageproduct curves for an even age stand of YC12 trees

20 18 16m

14AP

10AP

64[. 2 OL 0

20

60

40

80

Years from planting

Notes:

MP AP

marginal product (otherwiseknown as the meanannualvolume increment:MAD averageproduct (otherwiseknown as the current annual increment:CAI).

Source: adapted from Edwards and Christie (1981)

YC thereforemeasuresthe productioncapacityof a particular stand. For planning by be height YC to the age measured relating plantation observed top can average purposes 6.6 illustrates Figure YC for in Sitka by that curves stand. trees spruce superimposed the of is This latter the average at annual product line age which maximised. curve showing a it faster tree the will maximiseaverageproduct. grows, sooner a shows that the

6.27

Figure 6.6: General yield class curves for Sitka spruce

YC 24

4 0.....

age of maximum mean annual volume increment

3 6-

22 20

.... ....

.......

18

3 2...

...

......... ....... 14 ......

2 8..... ...... ....

E

......... .......... 10

......

2

8 . .

CL 210ý2

.

...... ......

............ ...........

...... ...

..... .... ......... .6 ........

1 6-

2time of first thinning

10

20

30 Age

40 from

50

60

planting

70

80

90

(years)

Source: Edwards and Christie (1981) Since its inception in 1919the Forestry Commission hascollected data quantifying the These YC. 'yield differing have been growing at models' plantations now of characteristics (Edwards Christie, 198 1). Table regimes management and and species varying across collated for YC12 Sitka 2. Orn qpacing and thinned the model spruce illustrates planted yield at 6.10 Commission's Forestry standard guidelines. the under

6.28

Table 6.10:

Yield model for YC12 Sitka spruce (2.Om spacing; intermediate Yicld frm TIENNINGS

MA[NCROP aftef Th-g

(1) Age yrs

(5) BA

(6) Heam

1ha

(4) Heaý dbh

1ha

-1

(3) Trees

(2) TOP Ht

thinning)

CUMULATIVE PRODUCMON

NIA I

(7) Vol

(8) Trees

(9) Mean

(10) BA

(11) MWR

(12) Vol

(13) BA

(14) Vol

(15) Vol

(16) Age

1ha

Ika

dbh

1ha

wl

1ha

lha

Ika

1ha

yri

20 25 30

7.3 10.0 12-5

2309 1450 1057

11 15 is

24 25 28

0.03 0.06 0.12

66 91 131

0 799 393

0 12 15

0 9 7

0.00 0.05 0.11

0 42 42

24 34 44

66 133 215

3.3 5.3 7.2

20 25 30

35 40 45

14.9 17.2 19.2

877 678 571

22 25 29

32 34 36

0.22 0.34 0.49

180 231 279

230 150 1017

18 21 2.3

6 5 5

0.18 0.29 0.39

42 42 42

53 61 68

306 399 489

8.7 10.0 10.9

35 40 45

50 55 60

21.0 22-5 23.7

492 439 401

31 34 36

38 39 40

0.65 0-81 0.97

319 357 390

79 53 37

26 28 30

4 3 3

0.51 0.64 0.78

40 34 29

73 78 82

570 642 704

11.4 11.7 11.7

50 55 60

65 70 ' 7

24.8 25.7 263

373 351 332

37 39 40

41 42 43

1.12 1 26 1 140

418 443 465

28 22 1 19

32 33 35

2 2 2

0.93 1.05 1 1 10 .

26 23 21

85 '' 901

758 805 1 848

11.7 "' 3

65 70 75

Glossary of terms: Age: Top /it: MAINCROP after Thinning: Yield ftom THINNINGS: Treeslha: Mean dbh: BAlha: Mean vol: Vollha:

CUMULATIVF MAI (and YC):

Note:

PRODUCTION:

1

1

1

1

The number of growing seasons that have elapsed since the stand was planted. Top height; the average height of a number of 'top height trees' in a stand, where a 'top height tree' is the tree of largest breast height diameter in a 0.01 ha sample plot. All the live trees left in the stand, at a given age, after any thinnings have been removed. All the live trees removed in the thinning. The number of live trees in the stand, per hectare. The quadratic mean diameter (the diameter of the tree of mean basal area) in centimetres, of all live trees measured at 1.3m above ground-level. Basal area. The sum of the overbark cross-sectional areas of the stems of all live trees, measured at 1.3m above ground-lcvel, and given in square metres per hectare. The average volume, in cubic metres, of all live trees, including any with a breast height diameter of less than 7cm. The overhark volume, in cubic metres per hectare, of the live trees. In conifcrs, all Limberon the main stem which has an overbark diameter of at least 7cm is included. In broadleaves, the measurement limit is either to 7 cm, or to the point at which no main stem is distinguishable, whichever comes first. This is the main crop basal area or volume, plus the basal area or volume of the present and all previous thinnings. The mean annual volume increment (or average product); i. e. the cumulative volume production to date divided by the age. Ilere this peaks at 11.7m3. YC is by convention rounded to the nearest even number, i. e. this is a YC12 stand. All trees which die through natural mortality are excluded, except that in models of unthinncd stands the volume of dead trees, expressed as a percentage of the cumulative volume production, heading per cent mortality.

is given under the

Source: Edwards and Christie (1981). first in is YC for two columns of the yield model the The yield curve given each

left lists in (the 'maincrop') trees the the third of 6.10) the number column stand (table while in (9) (12). The details to the given columns are of which value of thinning, each after in by by is (7), felling the the multiplying given volume/ha, shown column at maincrop, in is function (given itself the tree mean volume per However, a of price column price/m3. Simply felling date. function in is put, when trees are thin they are of turn of a (6)) which increase is low. As in trees their price/M3 volume so their usefulnessand limited use and so 6.29

therefore price/0,

rises. This continues (at a diminishing rate) to the point where their girth

(see column (5) of table 6.10) is such that the tree can be used for sawn wood, telegraph After fairly this the point priceftný remains constantand the products. other poles and myriad

increases does. as much as onlY volume value of a stand Estimation of this 'price-size' curve has been the subject of repeatedstatistical investigation by the FC (Mitlin, 1987; Whiteman, 1990; Sinclair and Whiteman, 1992). In this study we adopted the findings of Whiteman (1990), primarily becausethis usesthe same base year as our wider study, but also becausethis analysis recognisesthat prices are higher in England and Wales than in Scotland and therefore provides a significantly better fit to the data (W = 87.5%) than Sinclair and Whiteman's subsequentunified analysis (R' = 74.7%). Figure 6.7 illustrates the price-size curve used in our analysis.

Figure 6.7:

Price

Price size curve for conifers in England and Wales (f.; 1990/91 prices)

W

(EI990/91 prices)

Volume

/ tree

in (1990) Whiteman data from drawn given Source:

6.30

(m3)

The value of thinnings is calculatedin a similar mannerexcept that, whereasmaincrop revenues are only collected once (at felling), once commenced thinning takes place on a regular (usually five yearly) basis. The FC yield models give information on when thinning should commence (approximately year 20 although this varies acrossYQ. The felling date therefore emergesas a key factor in determining the overall value of a stand. As mentioned this will vary according to both the YC concernedand the discount faster 6.6 As figure the showed, a tree grows the sooner it reachesits age of rate employed. increases, Therefore YC as average product. annual optimal felling age falls. This maximum is exacerbatedby discounting, i. e. the higher the discount rate the lower the optimal felling age. The impact of varying YC and discount rate upon optimal felling age was calculated FIAP by FIAP operates mentioned above. the software maximising the net present using value of a stand subject to severaluser-determinedparameters. Results from this analysis are 6.11. in table given With felling year established we can now calculate both maincrop and thinnings insufficiently it felt RAP flexible However, for that to conduct was stand. was each revenues for YC6-24 Sitka into further therefore and yield models spruce analysis were encoded our MINITAB (MINITAB, 1994). for A the database statistical within package particular use a design that the the to this approach was software allows researcher of custom advantage data facilitating complex analysis. and/or repetitive written macro's, 6.5.2.3: Combining timber revenues with subsidies and cost streams As discussedpreviously in this chapter, grants and subsidiesconstitute a major source due for to their relatively early receipt, outstrip the operator may, which woodland revenue of felling discussed The discounted array of available grants of revenues. value previously the 12 6.12 Table details to be possible subsidy payment stream permutations. simplified can for benefit YC/discount plantation costs timber and revenues one with rate these along discounted 6%, (YC24, felling 41 rate of at a real giving an optimal age of combination yearsY".

49ThiScombination is chosen for illustrative purposes(to ensurea reasonably short rotation). Note that as felling in 0, 40 (as 6.11). in occurs table year per year planting occurs

6.31

Table 6.11:

Optimum felling age for various discount rates: Sitka spruce, YC6-24

Yield Class (Sitka spruce)

234568

6

80

73

68

64

60

8

78

72

67

62

10

74

69

64

12

70

63

14

69

16

Discountrate 10

12

54

50

47

58

53

48

44

60

56

51

47

43

58

56

54

50

46

42

60

54

52

50

47

44

41

68

58

51

49

47

44

42

40

18

66

57

50

46

43

42

40

38

20

66

57

50

44

42

40

38

36

22

66

56

49

44

41

37

36

34

24

65

56

48

44

40

35

34

Notes.

1

optimal felling age maximisesNPV given the relevant discountrate (r) and YC combination. The above figures treat the planting year as year 0. The table was calculatedusing FIAP running at the Forestry CommissionHeadquartersat Edinburgh(exceptfor the Towfor r= 3% which was interpolated).The author is obliged to JaneSinclair and Roger Oakesat Edinburgh for assistance. The table is calculated according to the following assumptions: 2.00 x 2.00 Spacing: Line, MTT Thinning: None Delay on first thinning: 85% Stocking: Zero Successor crop NPV: G.B. conifer 1992 Price size curve: Thinning price differential (E 1992/93): 0.30 An" Charge per m3 (E 1992/93) : E3.68 le

6.5.3: BEECH COST AND REVENUES

6.5.3.1: Costs Information on hardwood planting costs is far less readily available than for conifers. Data was collected both from interviews with managersof broadleaf woodlands" and from

"Notably Fred Lewis, Kerswell, Exminster and Cyril Hart, Chenies, Dean.

6.32

c; c; c; c; c; c; 6 c; c; ci li 00 03 8 88 c. > 08 0. Is 0.

c;

Icý>99 00 Co!

099c;

6 ci 6 c; c; c; c; c; c; c; c; ýi ci 008008888888888 88 00 99 --

c; c; c; c; c; c; c;

.

. -99

....................

v.

0900 88S 00ccl! CO, 2 008 Co> ......

oooc

00 9 .9 .9 -9

C!

90

vo

d"

ý; ý;. ý ;;;;; c; vý t; c; u; c; ---------------

-; Q

c! cý 9 ce C! 9 4 c; ooo ýi c; 0 c; (ý c; e;ci ci c; ýi c; 000

9

C)

99... -------------

9>



9999

c; c; ci c; 6 c; <; 66 di c; c;

06c;

L-8

-0 51

-2 g

c8.0.0 c;c; c;0

ý9E ý, *ý ý, b* *.;

rA

99C!

ý; ;ýj;

....

ý; ;J= ......................

> LL) <>

g8

> r. Ei Ei 50

. In

.........

0

rq lý

999 c; c; c; 2,9021

c; c;

4 c; 46 c;

di C;

99999ce C; c; c= (; e;

9...

C! . ce ....

9 le

ce c;

66

66c66c;

m .-M Z .-. 99999 5,9

-----------

6

C? 9

c; x; c; c;

10

ce9c!

99

99

9

11

st

11

11

99 .99 Ln

;o9

rl

C!C!

C;C; 9...... 1! C! 1! 60C; 1ý 0000 0 c; 4; C; Iz g C; 0 C; 1; a C; C; C; 1; c; C;60C; C;C;C;C;C;4; C;C;C;C;-ý

0

ýý0

C! 1! C!C! ...... ; ; "0 00000 C; 4 0C 4; C; C; C; 0 0 C; C; C; 6 000c; c; C; C; C; ci v0o clo cocc; C;ci 0-% CA Coll

ýR Iz

0

.9'ý

6 0; 6 ; ; ; 6 ; ; ; C; ci 6 C C C; C gi C c; 4 c; C; 0C C; C; 66C; C; C; C; C; ci C; C; C; CIO C;C;C;

09

MMC 00

ts

C;C;4iC;C;C;co0

66 C; ci ci F: C; 0 46 ; ; 9; ; 4; ; C; C; C; C; C; ci C; C; 0 C;C;C;CC C C

00

40

n.

.0

2 o z

0

8

C4

>

o.

>

r.

(A

certain published sourcess'. Assumptionsregarding the incorporation of felling and thinning before full broadleaf cost stream for a the timber are as and prices costs within standing 6.13. in is detailed hectare table typical

Table 6.13: Broadleaf plantation costs (F-/ha;1990 prices)' Year from planting

Cost (E/ha)

Cost items

0

740.79

1 2 3-5 6-(F- 1)

22.06 151.81 45.65 22.06

Construct maintenanceroads, plough land, other land preparation, initial drainage, new fencing, purchasing plants, individual shelters and stakes(assuming 50% recycling rate), planting, initial fertilizer, initial herbicide. Maintenance Beating up, weeding, maintenance Herbicide, maintenance Maintenance

In addition a cost of L285.03 is incurred in the year before first thinning (varies accordingto YQ for construction of forwarder Toadsand transporterpoints. Notes:

I.

Generalassumptionsare as per table 6.9

1990). 1988), Hart (1987 (pers. Lewis comm., and pers. comm., Sources:

6.53.2:

Timber revenues

Figure 6.8 detailsYC curvesfor beechshowingthe long rotation periodstypical of broadleaved species. into YC formalised have FC these curves yield models (Edwarxis As with conifers, the

for all reasonablefelling ages,the volumeof timber produced detailing, 1981) Christie, and before As tree price varies positively with volume. clear-felling. at and thinning at each for broadleaves, Whiteman (199 1) in et al. show ]Furthermore, their study of price-size curves high for have costs extraction per M3 standing prices because thinnings relatively that, , for fell below Consequently 24% the timber. clear two price/O paid thinnings are on average for being (with third curve average of a reported ease estimated are curves price-size

slNotably Hart (1987).

6.34

generalised account rather than individual plantation assessment). As hardwood timber values vary considerably between species, price-size curves are estimated for each (unlike the generalised conifer relationship), with those for beech being illustrated in figure 6.9.

Figure 6.8: General yield class curves for beech

YC

40age of maximum mean annual volume increment

35-

10 8 ............... ....... . ....... 6

30-

E ... ........... ........... .............. ............

25

a) ..........

20-

4 ...................

.............. ... .........

........ ....... ......

a

...........

... .......... ............... ..... ....... ................... .... ... ............. ........... ... ............

15 10

time

of first

50

60

70

Age

from

planting

thinning

5

10

20

30

40

80

90

100

110

120

130

140

(years)

Source: Edwards and Christie (1981)

With volumes and prices establishedwe again simply require the optimal felling age is function As before YC discount define this a of the stream. revenue and rate and was to 6.14 details RAP Table from the this analysis. software. results evaluated using 6.5.3.3: Timber revenue, subsidy and cost streams Timber revenues and cost streams were now encoded into the MINITAB software be integrated to them Table 6.15 grants with available to allow and subsidies. package illustrates the resultant databasefor one yield class/discountrate combination, namely YCIO discount 6% rate. a with 6.35

Figure 6.9:

Price size curves for beech in Great Britain: 1990/91 prices

ear Tenings terage of all Ilings innings Price /m (E 1990/91 prices )

0.50

1.00

1.50

2.00

2.50

3.00

3.50

4.00

4.5U

Volume / tree (m 3)

Source: drawn from data given in Whiteman et al. (1991) discount for felling Optimum 6.14: rates: Beech, YC4-10 age various Table Yield Class (beech)

Discount rate M

10

468

2

125

120

119

118

3

105

99

95

93

4

91

85

80

78

5

81

75

71

69

6

75

69

65

62

8

65

59

56

53

58

52

48

47

47

43

42

10 12

ý7

1 L_53

_

NPV discount (T) YC The the felling given relevant rate and maximises combination. age above Optimal Notes: The FIAP 0. table calculated using was running at the Forestry as year the year figures treat planting CommissionHeadquartersat Edinburgh(exceptfor the row for r--3% which was interpolated). The author Oakes Roger Sinclair Jane at Edinburghfor assistance. and to is obliged following to the assumptions: is according 'rbe table calculated Thinning: Broadlcaved, intermediate 1.20 1.20 thin x Spacing: Stocking: 85% None first thinning: Delay on Price size curve: Broadleavesfor 1989/90T.R. Successorcrop NPV: Zero LO. 30 Irre (L Charge 1992/93): differential per m3 (L 1992/93): 0.68 /m3 Thinning price

6.36

-i

...............

c;

o-oo.

oo..

oo

oo .....

CD Ch

...

....

....

0 ci cn

........................

................

fto..

ce 5

..........

----------

00

o..

...........

u . ID

0 """""""""

-02 .......................

..............

1.2 6: ...........

0 c)

Q ...

0 .................................

0

Oe

e.

c

e.

5..

0

05000500

50



P4 0505

5505

P

50

50

*5

555

-..

e5

*5

.

es

es..

es..

Sa

S5"".

0: CA JD

b: o0aO000000000000.0.0000

X. 2

00.0000.0.0.0

iN

Iq z

.1 0

a.

...........

0-0-000060600000000000000:

aoao0o*oa*o**

................

1-11 ..............

0000.0.0000000000000090

.............

JD

eaeaeoe-a-0

0

..........

........................

Notes: t= age of standfrom planting in year 0 Revenuecodes asfollows: M maincrop

T trees Av Vol = Vol/ha = price ben costs

thinnings numberof trees averagevolumeper tree volumeper hectare(m) pricefinýfrom price-sizecurve revenuevalue total annualcosts

Subsidycodesasfollows: S subsidy I planting on improved grasslandor arableland U planting on unimprovedland planting is not in a disadvantagedarea nda da planting in a disadvantagedarea in disadvantaged area a severely planting sda planting given community woodlandgrant +CW grant given community woodland not planting -M

6.6: DISCOUNT RATES Any investment in forestry has to trade off initial costs againstdelayed benefits. This is conventionally achieved by calculating the NPV of the investment via a discount Tate(r), in defined the two terms time and of elements: pure preference; opportunity cost commonly Ulph (1995) have Pearce discount Adopting the the of and we standard notation of capital. follows": as rate equation

(6.3)

+99 where: discountrate 8= time preferencerate (the rate at which utility is discounted) g= elasticity of the marginal utility of consumption schedule g= expected growth rate of averageconsumption per capita.

This research has set out to examine two perspectivesregarding the decision about

"For a good introduction see Pearce (1986) and for further reading see Pearce and Nash (1981), Lind (19823,b) and Price (1993). the latter being of particular relevance to forestry.

6.38

whether or not agricultural land should be converted to forestry: that of the fanner; and that of society. However, there is good reason to suppose that these two will have differing discount rates5'. Put at its simplest, if we consider time preference,farmers are mortal while society is, at very least, much longer lived (we hope!). T'hereforesociety is likely to be more concerned about long delayed returns than will an individual farmer. Accordingly we might expect society to have a relatively lower rate of pure time preference. A similar result is obtained when we consider the opportunity cost of capital basis for discounting. For a riskimply this a relatively low social discount rate dictated by the rate of should averse society (government investments bonds, etc). However, for the private individual on riskless return the opportunity cost of capital should be relatively high due to the rates of return available from alternative investmentsm. In this section we examine evidence regarding agricultural and social real55rates of discount. However, before turning to this we need to addressone further complication, that forestry investments. Farmers commonly make agricultural and of the comparability of decisions on an annual timescale whereas the time horizon of a forester is usually a full from four decades for typically a minimum varies which of a stand, of conifers to rotation Comparison for hardwoods. is NPV of annual gross with margin rotation a century over Two approachesexist. Firstly agricultural margins could be assessed therefore problematic. length. Secondly, NPV be discounted a rotation over can woodland converted to an and i. (or 'annuity') length the the constant annual return e. which, over equivalent, of a annual be NPV After discussion the equally with standard sum. valued with relevant rotation, would former lacked farmers decided (who it that the option credibility as was are the experts'6 decisionmaking. decisionmakers) Therefore, to than are annual rather used rotational relevant (using for NPV's the relevant agricultural or social models our all yield after calculating discount rate), these were converted to annuity equivalents using the formula given as (6-4). equation

"For further discussion on the divergence of social from private discount rates see Baumol (1968), Goodin (1988) Pearce Sagoff (1982), Sen and Turner (1990). and (1982). '*rhis may be a less strong argument if re-investment is restricted to the agricultural sector where rates of low. historically return are 5-lie.inflation adjusted as opposedto nominal (unadjusted) discount rates. Willis, UCNW, Rob Bangor. Price Colin and -56NOtably

6.39

(I

Annuity =(

[(I

+ r)F + r)F]_,

* NPV

r

(6.4)

where: r= F=

discountrate (expressed as decimal,e.g. 6% rate expressedas 0.06) felling year (lengthof optimal rotation,in years)

NPV = net presentvaluecalculatedfor discountrate r and optimal felling yearF [. )= net presentvalueof an infinite sum of optimal rotations. 6.6.1: FARMERS' DISCOUNT RATES

6.6.1.1: Literature review A priori we would expect that the relatively lower rates of return exhibited by the industrial in (as to and commercial opposed equivalents) would result sector agricultural 8% for discount lower the than the rest of the government's rates estimate real somewhat little However, 1991)57 has been Treasury, in M. (H. explicit published work private sector . interest real rates examining of agricultural return or rates this area with most commentators discount rates per se. than rather The early work in this latter areais predominantly American, dating back to Melichar (1979) who proposed that real rates of return were determined by expectedrents and actual by (1980) Feldstein inflation theory this suggestingthat such modified rates. and expected land be driven by inflation Tanzi acting upon prices, while ultimately may a mechanism business in further link However, by to the cycle. this a an proposing (1980) extends rind long-run failed link between inflation Alston (1986) to a theories, these empirical test of in favour (1986) long Burt complex models such land rejects of a simple run prices and and 4% land a real rate of return estimate of %ýhich yields per annum. price approach equilibrium

Similar resultsare reportedby Cooper(1992)who usesa real interestrate approach (1989), La Due Brase 4.5% for UK to report a and mean of value the of work based on interest highly variable", 1964-90. While for agricultural rates the are period agriculture lending In by be to current practice. correspondence roughly echoed with seems result a such

17The8% estimateis "basedon averagereturnson assetsachievedin the private sectorfor activitieswith (H. M. Treasury. Rates 1991). be higher in otherareasof by to are expected variability" year low cyclical year the private Sector. 5'Annualaveragesrangefrom -13.01%(1976)to +10.08%(1990)in Cooper(1992).

6.40

the author the National Westminster Agricultural Office" (a major source of farm finance) interest 4% base in turn roughly rate of agricultural over real rates average which quote an (at inflation present). rates shadow A lower interest rate, averaging 2.44% above base rate, is reported by Cunningham (1990) in a study of MAFF surveys,while MAFF themselvescurrently assumean agricultural interest rate risk premium of 2.78% above baserate'. However, there are severalproblems if base discount. Firstly, fluctuate interest frequently, from to rates rates with extrapolating lags in the adjustment system may confound the analyst. Secondly, interest rates vary very 61 interest Thirdly, link between farms, time the and projects rates and significantly across . discount rates may be weak in that the former relate to returns on new investmentsrather than be lower). likely (which to are on total assets In addressingthis latter point, Harrison and Tranter (1989) analysethe period 1978/79 2.56%'. Positive time 1986/87, on all of return assets of real rate a mean reporting to be higher discount than this, that the rate might somewhat real a suggest would preference factor which gives further support to the findings of a recent study by Lloyd (1993). Here in England Wales for land the and prices period agricultural of model pricing asset capital a discount 3.6%. long derive is of rate a run real to 1946-89 used empirically These latter studies provide what we feel is the best evidence on agricultural real discount rates. However, neither study is specific to our Welsh study area and so our own rate of return analysis was undertaken. 6.6.1.2: Empirical work

a short

Two studies of agricultural rates of return in Wales were undertaken; the first being being 1987-92; the second the a cross sectional study period of time-series analysis

by Welsh division Farm data both In the base the of 1989/90 was provided cases year. of the

"Pers. comm. SueTrain, NWAO, and letters from Brian Montgomery, Senior Executive, NWAO, July 1993. farms in the highlighted and projects. For example a range rates across variation However, this correspondence for by Charles differing Morgan Chris Grote 0-5% times Farms, from and projects given of was of real rates Norfolk. 60pers.comm. Douglas Cooper, MAFF, 1993. OThis point was made in correspondencewith both NWAO (seeabove) and Paul Hill (Wye College) who 2% base for be higher interest they above rates good could were roughly risks, rates very much stated that while for risky investments.

Great Rates Britain. from 1.87%to 3.90%. 62Sample are quite consistent only ranging across extends

6.41

Business Survey63(FBS, 1988,1989,1990,1991,1992)

which defines the nominal return

as farm income expressedas a percentageof tenantscapital". i. Rates of return in Wales: 1987-92

Table 6.16 details nominalrate of return (RoR,) statisticsfor variouscategoriesof farm type identified during FBS surveys for the years 1987/88to 1991/92. These categories

farm by further size. subdivided are Statistical analysis was undertakenacrossall farm categoriesexcept those for pig and in farms Wales these as are minor activities and were not separately cropping poultry, and farms dairy This 1989. that showed specialist or mainly achievedsignificantly classified after higher RoR. than did other farms. Subsequentanalysis also isolated a quadratic relationship but diminishing RoR,, in BSU65) (measured that rose at a showing with size with size fluctuated (1988/89) found RoR,, although to only one annually year was also marginal rate. be significantly different from all others". A model was constructed encapsulatingthese relationships and was tested across a is fitting Our best (6.5). forms. Tests for functional reported as equation model variety of failed isolate heteroscedasticity to and any significant autocorrelation multicollinearity,

problems with this model. RoR,

18.616 + 7.683 TYPE + 9.566 HIYEAR + 1.1289 BSUt - 0.010508 BSUI2 (8.38) (-6.33) (6.53) (-9.06) (6.32)

(6.5)

where RoP, TYPE

(%) tenants capital on nominal net rate of return I for dairy farms (FBS specialist or mainly dairy categories) 0 for non dairy farms (other categories)

63TheFBS is an arm of MAFF (operating in Wales under the auspicesof the Welsh Office) which conducts farms Sample farms 734 the throughout country. sample of size averaged a representative per of annual surveys in for farms 3 however, The the 1987-92 sample retained about are years. many study period, annum over our is 2867. in farms the time series number of unique 64Dcrlnitionsare as follows (from FBS, 1989): 71c farm managementand investment income (Mll), which (and interest farmer's for "the the the spouse's) management tenant's capital employed and on reward represents farm input (the latter including between both difference by is and output values farm", the actual given on the farmer and spouse):Tenant's capital (Kp) is defined as "the value of tabour for income the the of and notional (including cultivations) and stores .... expressedas the averageof the opening and crops livestock, machinery. items"; is [(Mll/KF) Return for 100]. tenant's these therefore on capital x closing valuations "British stocking units. "Interaction terms were found to be insignificant.

6.42

HIYEAR =I BSUg

BSU2

RI

R2

F p dw

for 1988/89 0 otherwise Average size of farm type in year t where, here, farmtype is: I= Specialist dairy I in TYPE dummy 2= Mainly dairy 3= Hill sheep 4= Hill cattle and sheep 0 in TYPE dummy 5= Upland cattle and sheep J 6= Lowland cattle and sheep BSU, * BSU, 77.9% 76.7% 66.10 0.000 1.89 where dL = 1.39 1 no autocorrclation n= 79; exp.vars.=4 d,,,= 1.60

Note: numbers in brackets are t-values D fanns (denoted RoR RoRýDrespectively) for dairy RoR. and Average and non-dairy . -, for by be each group's mean values substituting evaluated now can period the study over (6.6)68: farms this dairy For (6.5). into equation yields equation explanatory variables67 RORP.

9.566(0.2) 7.683(l) + + -18.616 + 1.1289(34.02)- 0.010508(1590)

(6.6)

12.677658 12.68% For non-dairy farms this yields equation (6-7) IZND

Ro,

9.566(0.2) 7.683(0) + + -18.616 0.010508(918) 1.1289(24.78) + -

(6.7)

1.624998 1.62% dramatically RoR, that they be vary across of shows analysis simple As can seen, is highly indicting farrns dairy between the significant difference and non-dairy farms. The 67Meanswere used after examining variable distributions for skewness. Arguably mean values may not is farms. farms however than this actual of rather an analysis optimal sized optimal on conditions reflect term is data derived and consequentlyslightly differs from the square 68NOtethat the averagefor the BSLJ2, of the BSU, mean.

6.43

very considerablepositive impact which CAP milk quotashave had upon dairy farm incomes. Table 6.16:

Agricultural nominal rate of return on tenants capital: Wales 1987/88 - 1991/92 1982/89

1917/08

-T. -

type

&. d . 110

Dairy Speciallot Up to 25.9 OSU 16-23.9 BSU 24-39.9 SSU 40 9sU and over All Si.. S Dairy mainly Up to 23.9 &SU 850 24-39.9 4o sso and over All time Bill Shoop up to 13.9 BSU 16 SSO and --r All *1. -& Shoop sill cattle up to 13.9 SSU 11SU 16-23.9 BSU 24-39.9 40 Stu and over All 81--d G Shtep Upland Cattle up to 15.0 BSU 16 BSU and over All Sims Cattle A fhaaP JýIand L... k1l pig Sýdlpovltry 41 .1.. ram. Cropping All Nixon

1 ". " an 14"A TtiaoOd st.. d. rd D-S. I, r Qsartll. upper Q"rtiio Mani. -

Notes:

Me :. aaU,

I",

rate it)

a

MOOR . 1.. (35u)

1990/91 rate it)

a

mean . 1.. 4113u)

1991/92 rate it)

a

mean sit. (93u)

1987-92 rate it)

a

a"" fi.. issu)

rate it)

30 26 35 27 110

11.97 19.57 30.02 67.13 31.83

10.04 10.21 13.76 25.10 18.11

30 26 35 27 118

11.83 19.32 31.23 69.03 32.33

13.99 13.02 26.52 36.06 21.77

29 Is 30 31 115

11.37 19.98 30.93 67.10 34.21

4.94 14.29 17.01 27.37 21.16

20 14 34 36 104

10.42 19.27 31.63 63.21 3632

-0.13 4.27 13.30 19.69 15.56

17 20 28 31 96

10.15 19.40 3t. 13 60.70 34.34

-6.25 9.27 15.24 20.65 16.25

125 104 170 152 551

11.29 19.52 31.15 65.22 33.95

9.9566 10.6393 17.4441 23.3209 20.0099

14 13 If 47

14.14 31.45 56.01 35.73

6.6s 13.41 15.55 13.83

14 IS It 47

14.14 31.79 54.48 36.37

4.70 18.32 19.36 17.31

14 13 IS 45

13.32 31.91 59.09 37.96

0.01 13.72 16.01 13.24

2S 9 13 49

16.31 34.61 73.32 37.12

-1.12 13-60 10.05 8.11

15 11 16 42

14.15 34-52 72.21 41.20

-2.99 13.60 13.10 11.67

82 63 Is 230

15.02 22.61 62.63 37.59

1.0640 14.7173 15.1502 12.8127

24 27 51

10.13 32. S7 22.01

-3.94 12.96 10.14

24 27 51

10.35 31.67 21.73

1.34 20.06 13.99

25 24 49

10-03 33.68 21.62

-10.04 6.14 0.34

22 32 54

9.69 31.14 22.40

-16.34 -1.04 -3.94

21 32 33

11.39 34.21 25.20

-3.15 11.14 8.76

116 142 Ilse

10.33 32.63 22.62

-9.1123 9.7024 6.2636

39 29 24 14 108

10.30 19.07 30.14 S7.77 23.59

3.91 3.58 12.417 20.84 12.11

39 29 26 14 log

10.64 19.52 30.33 $7.36 23.82

9.49 12.21 17.70 20.12 15.05

35 32 29 13 110

10.97 19.67 29.92 10.36 26.13

-8.81 -2.3S 5.72 7.22 2.39

34 36 29 Is 111

11.32 19.63 31.37 76.13 241.86

-11.80 -4.06 2.57 -3.53 -2.95

2s 23 25 20 93

10.44 It. " 31.41 74.04 32.12

-5.96 4.70 8.29 6.70 S. 31

172 149 134 @1 536

10.75 19.40 30.61 68.13 26.79

. 1.9387 2.6594 9.2296 11.2064 6.4310

Is 20 36

9.33 26.21 18.71

-3.66 4.64 2.71

26 20 26

1.65 27.43 19.08

3.33 7.52 6.60

If Is 34

9.29 23.29 16.70

-7.42 -2.07 -3.57

19 21 42

S. 29 23.66 17.00

-17.65 -3.53 -6.57

19 25 44

7.56 30.43 20.55

-15.09 2.14 -0.91

of 106 192

41.So 26.92 19.63

-8.6379 1.6816 -0.5094

13

12.64

-1.50

13

12.68

1.38

1?

18.14

-5.03

31

22.04

-1.39

26

17.90

-0.06

100

18.11

-1.3826

6

29.77

3.96

6

22.64

12.94

12

26.20

8.4500

11

44.04

10.96

11

42.89

22

43.87

6.2500

I. S4

750

750 tl-

. 1SSU)

1989/90

28.07 27.11 13.92 3.32 9.33 14.14 32.57 1.13

9.54 0.43 7.40 1.54 -3.64 3.96 13.03 25.10

763

723 27.21 26.75 13.93 3.32 11.65 14.14 32.33 69.03

14.06 13.61 0.99 1.47 1.34 6.60 19.36 36.06

29.43 27.26 17.05 3.90 9.29 16.04 33.94 70.36

4.91 4.93 11.17 2.44 -18.04 -3.06 14.01 27.37

602 29.91 20.61 19.40 4.23 8.29 17.06 35.72 76.13

0.61 0.57 10.03 2.10 -17.65 -3.05 9.09 19.69

3660 30.16 29.04 10.98 4.14 7. S6 14.02 34.53 74.04

5.40 5.67 9.94 1.1's -IS. 09 -1.90 12.60 20.65

29.23 29.37 17.05 3.56 1.31 19.11 33.85 60.13

1.07 0.93 IAS 1.74 -8.64 1.04 12.81 25.32

1. The summarystatisticsare calculatedby omitting the "All sizes" categorymeans(except where this is the only entry for the category). 2. The 1987-92meanrate of return is weightedby annualnumbersof farms as is the averageBSU size 3. *= not available 4. n= number of farms in sample 5. rate = nominal rate of net return on tenantscapital, calculatedas follows: MII = Output - Inputs rate = (MIL/TC) * 100 and where: (i) Output = All returns from an enterprise,plus the market value of any of its productstransferredout to from the enterprise given to workers or the of any production market value plus enterprise, another livestock In livestock farm. the the value of the enterprises, purchased case of and the market on consumed deducted. in from All for livestock totals are transferred enterprise another are adjusted changes of value in valuation. (ii) Inputs = Feeds(purchasedconcentrates,homegrownconcentrates,purchasedbulk) + Tack and stock keep + veterinary and medicines+ other livestock costs+ fcrtilisers + sceds(purchasedand homegrown) (contract, labour (farmer fuels, + + casual) and spouse, paid, unpaid, machinery costs repairs, crop + other depreciation)+ generalfarming costs+ other land expenses+ rent/rcntal value + rates. Note that asa nominal farmer/spousclabourcost is included,we are calculatingnet ratherthangrossreturns. (iii) M11 = Managementand InvestmentIncome; The MII representsthe reward for the farmer's (and farm interest the tenants the on and capital employed on management spousels) (iv) TC = TenantsCapital; The value of livestock, machinery,crops (including cultivations) and stores.In is Survey Business expressedas the averageof the opening and closing tables, tenants Farm capital the items. for these valuations

from FBS (1988,1989,1990,1991,1992) data taken Sources:

6.44

Conversion to real rates of return (RoP, ) was achieved using retail price indices19

inflation (1993). This in CSO shows an average rate for the period 1987-92of 5.81% given RoRT = -4.18%. implying RoR' and ,=6.86%; One problem with our approachis that it usesmean BSU by farm type (dairy or nondairy), yet the summary statisticsgiven in table 6.16 suggestthat the distribution of farm sizes is positively skewed (mean significantly exceedstrimmed mean) rather than normal. An by is the the number to mean rates of return weighted simply examine approach alternative farms irrespective Ibis in to farms all of their size. gives equal weight each category. of Table 6.17 reports RoR, and RoR, for each of the farm categories shown in table 6.16 with before. from as nominal real rates adjusted Table 6.17:

Mean nominal and real rates of return on tenants capital: Wales 1987/881991/92

Farm type

Specialist dairy Mainly dairy All dairy Hill sheep Hill cattle and sheep Upland cattle and sheep Low and cattle and sheep All cattle and sheep

n

mean nominal rate of return RoR,, (%)

551 230 781

15.73 10.01 14.04

9.92 4.20 8.24

258 536 192 100 1086

1.74 3.83

-4.07 -1.98 -8.75 -7.19 -4.15

-2.94 -1.38 1.65

mean real rate of return RoR, (%)

(1988,1989,1990,1991,1992) from FBS data taken Source:

Comparisonof resultsfrom table 6.17 with estimatesfrom equations(6.6) and (6.7) farm farms to those equally produces similar all results obtained when size treating that show farins. in the is considered, Particularly case of non-dairy

Ratesof return in Wales:1989190 We were particularly interested in RoR. during our study base year. The previous

69Useof the RPI rather than some farm price index reflects the fact that ultimately investment funds could be moved out of the agricultural sector.

6.45

is this that not a significantly unusual year and should therefore be fairly analysis suggests included 1987-92 Furthermore, that the time previous given series one representative. five (1988/89), the might year averagefrom that study to we expect exceptionally good year be somewhat about that observed in 1989/90. In addition to the previous TYPE and BSU variables, a number of farm level variables (e.g. capitalization, livestock intensity, etc) were analysed. These were taken from the FBS individual farm record databaseand are discussedin further detail in chapter 9. These data TYPE The (MILK) definition the redefined variable showed variable. of permitted a superior derived from dairy had farm least 20% farm output a significantly of produce at that any with higher RoR, than other non-dairy farms. However, no further significant, non-collinear, farm defined. output variables were A number of physical environment variables (e.g. soil type, altitude, etc) were also investigated. These were obtained from the LandIS databaseof the Soil Survey and Land Research Centre, Cranfield (discussedin detail in chapter 7). However, no variables could inducing into introduced severe collinearity problems. Finally a the be model without Rudeforth (1984) found from derived tested to the al. of et was and work variable regional is derived in final level. The 10% to therefore that similar model insignificant be at the cc= (6.8): is as equation given and our previous analysis RoR, W=

(6.8)

In BSU + 12.115 TYPE 13.205 + -39.372 (9.51) (-9.66) R2

43.3%

n= 240

42.8% =

Where:

RoR. BSU TYPE Note:

1989/90 of return rate nominal farm sizein BSU I if Dairy 0 is Non-Dairy 'Non-Dairy' is defined as less Om 20% of farm output being milk [n (non-dairy) = 126 of which 124 had 7% (next farm 24% I had had 3% I had revenue; milk milk revcnue)). and revenue milk zero milk revenue;

(6.8) for into RoR. allows to the us Substituting variable means calculate equation

(6.10) (6.9) farms, and respectively. given as equations dairy and non-dairy 6.46

RoRP.

12.115(l) 13.205(3.2208) + + -39.372 15.273664 15.27%

(6.9)

RoRý.'

12.115(0) 13.205(2.7773) + + -39.372

(6.10)

-2.6977535 -2.70% Adjusting for inflation (which averagedover 9% in 1989/90) implies RoRD,= 5.81% These RoRý" results emphasizeour previous conclusions regarding the gulf and = -12.2%. between dairy and non-dairy farms in Wales. Indeed here we see the latter group even is in the a situation clearly non-sustainable return, which rates of nominal making negative

long run. 6.6.1.3:

Farm discount rates: summary

thatagriculturaldiscountrateswill While datais scarce,availableinfonnationsuggests Our literature the economy. survey suggests that general low of be relative to other sectors

defensible. However, in 3% low of tenms our analysis rates of quite return are real as as rates different between the of sectionsof performance highlights the great variability which exists Wales, disparity in between to the reference with and particular, community the agricultural farms indicates, dairy 6.16 As farms. the table of elite consistently record dairy and non-dairy figures, in double (and while, as our of, return subsequent rates real) sometimes nominal farms These Welsh show negative real rates highlight, of regularly return. non-dairy analyses long from in Welsh hill the term the and exodus latter rates are clearly non sustainable (FBS 1987-1992) That to seems set continue. over years recent farming consistently observed low) by definition (if in business farms record positive rates must, remain which those said, discount have rates. positive and return of The link betweenrates of return and discount rates is not simple involving as it does discount Ilis rates raise above time may rates of although preference. return of consideration be by (1993) Lloyd that this not a particularly suggest will such as consideration of studies 12% 6V dairy farms feel In that of and the of rates case we large amount.

should provide

7OArguablyour findings could support a slightly higher rate. However, as discussedin the following section, is for for is discount this the 6% comparative purposes as government's useful rate nonrate choice of a low risk activities. and/or commercial

6.47

best discount bound for Welsh dairy estimate and majority of real rates an upper respectively farms. For non-dairy farms rates will clearly be significantly lower with only the most After literature 6% to the consideration of our review, empirical rates. efficient aspiring feel discount rates of return, we a real rates negative analyses and the non-sustainability of for Welsh farms be 3% from 1.5% those to non-dairy appropriate should sensitivity range which do survive into the next century. 6.6.2: SOCIAL DISCOUNT RATES The Thatcherite assertion(implicit in much positive economics)that society is no more than the sum of its individuals suggeststhat we should not separateout social and private discount rates. However, upon closer examination we can see that social preference often it: (1993) Fankhauser As individual from puts diverges preference. "Drug legislation, safety regulations, speedlimits or statepension schemesare ignoring individual state preferences". all examples of a paternalistic Pearceand Turner (1990) seemto argue that the individual may contain two separate The decisions, the the optimal private and public. maps: preference and at times conflicting be discount these systems of preference may very therefore rates time-horizons and difference between in divergence the the Recognition underlies part of such a differenel. its investments 8% 'public 'commercial' for 'required of and service of return' rate Treasury's 1991). 6% (H. M. Treasury, discount of rate output' An important exception to this rule is the land acquisitions and new planting activities 3% is Commission. Here discount Forestry Enterprise the a rate Forestry of arm the of important to note that this is not as a result of any notion of public is it However, allowed. for the benefits of forestry, but rather as a subsidy so that the different being preference This hand be 6% absurd rather accountancy obtained. sleight of on paper, can, official rate of 6% if from that, rigorously almost a rate were enforced, no the reason simple about comes

-For example, with respectto transport my private preferencemay be to drive unimpeded upon open roads former to recognises the externality preference cost public of option home my while to place of work, from my Where internal is transport this system. available, public conflict not resolved we widely a reliable, prefers and see NIMBYism-

6.48

new planting would occur. By using a 3% rate some (although by historical standards does benefit-cost little) test and the shortfall between pass a simple planting comparatively revenue generatedby this planting and that necessaryto satisfy a 6% discount rate is written 2

off as Forest Subsidj .

We havearguedelsewhere(Henderson andBateman,1993)thata comparativelylower discount rate for forestry may be justified on the grounds of true social preference,and that, as the time horizons underlying such preferencesvary acrossprojects, so we would expect discount The Treasury's have to rates. unwillingness to admit such a multiple society farcical is Subsidy), by Forestry (evidenced the the cooking-the-books which arises, possibility horror decisionmaking having from between the to of choose projects quite understandably, (nevertheless feel differently discounted that the complexity of social we which are Consequently discount imply does to this and return rates subsequently). multiple preferences is derived from discount Treasury's rate empirical data averagedover a public service the discrepancy (and This a argue, wide of social preferences). would we sectors of variety wide for 2 basic discount from the is the of roughly each of elements of values rate calculated rate formula (equation (6.3)), i. e. r=5+

gg =2+

(2 * 2) = 6%. However, a wide variety of

these the of each of elements. value regarding exist views Perhaps most controversial is the value of 8, the pure rate of time preference in the immortal is (or be) ). If (r, to discount then, aspires society as very many eminent rate social 8 be low (Ramsey, 1928; Pigou, 1932; have or should very zero out, pointed commentators Solow, 1974,1992; Price, 1987,1993; Cline, 1992,1993; Broome, 1992; Fankhauser,1993, forthcoming). Arrow 1995; Ulph, Pearce et al., 1995; and

Such arguments have been

development. This has centred upon notions debate by sustainable the surrounding reinforced be decisions 1972) (Rawls, to truly the equitable, wherein, regarding Rawlsian use equity of human or natural capitaly' should be made involving (be they man-made, of resources impact. Such a view is ignorance' temporal 'veil to their with respect behind a of Commission Brundtland definition of sustainable the to often quoted fundamental

"The detail of the situation is more bizarre than this. As all the Forestry Commission's historical felling Commission 5% discount The has has been 3%, this therefore rate retained. rates of a at made decisions were its different to operation. 6% of awas applied 5% and 73Foran excellent overview of the key role of capital types in notions of sustainability see Pearceet al.. from (1993). While Pearce an NeoClassical perspective,more extreme (but very Turner radical and (1989) or in Herman Daly (Daly, 1977; Daly and Cobb, 1990). the of work given interesting) views are

6.49

development as "... development that meets the needsof the present without compromising the ability of future generationsto meet their own needs" (WCED, 1987). Price (1993) sees discounting for interpretable of abandonment global level social as an this as only decisionmaking. A more 'conventional' view is given by Fankhauser(1993) who sees the above as implying that 5= 0%, but not necessarilythat r, = 0%. Pearceand Ulph (1995) review an 8 favouring from 0-1.7% but (for literature empirical a range reporting on social extensive 8=1.4%. best high estimate of reasons) a relatively Turning to consider the elasticity of the marginal utility of consumption (9) Price (1993) reports a wide divergenceof private sector rates, generally ranging from 0.5 (Squire finds Stem 1974). (1977Y5 Mirlees, 3 (Little 1975)74 Tak, der many values to and and van in the region of 2, however, we would expect the social preferencevalue of g to be somewhat lower than that found in the market. This is borne out by Pearceand Ulph (1995) who report from 0.7-1.5. 0.8 g a range with a best estimate of social of The social value of g (the expectedrate of growth of averageconsumption per capita) is typically taken as being the real rate of growth of national income. Following such an However, 2%76 for b) the (1982a, sustainable Lind rate of g= maximum a argues approach, . development debatehas highlighted the problem that accountingmeasuressuch as GDP often in base losses) (frequently the capital and other non-market of the natural ignore changes Ulph Pearce (1995) Taking 1989R). these and suggest of (Repetto account al., et economy from 1.3-2.2%. 1.3% in UK for a the range with of g a best estimate

the

Taking best estimatesfrom Pearceand Ulph (1995) gives a central estimate of r, for from 0.9-5%. While this may seem 1.3]) * (= 1.4 [0.8 2.4% + with a range UK of about

forward is higher by it Treasury's that than recently put certain the rate' low with respect to

74gis negative but we report modulus values following the convention of Pearceand Ulph (1995). 75SteM(1977) reports one extreme value of p= 10. is lower in in less developed GDP that (1994) than countries often much growth 76, out real point rUmer et al. this and sometime negative. 77RepettoPuts forward an adjusted, sustainablenational income measure. See also Pearce et al. (1989), (1993). Pearce (1992) Warford and Pearce and 7"Fcarce and Ulph suggest that for policy purposes the Treasury should use a range from 24%. In David in 1994). Pearce Treasury (November, had that the they stated the meetings with author conversation with for be is but had This figures, that they policy reasons stated would not adopted. such of accepted the validity dictated by being Henderson damning rates policy than of official rather evidence preferences. the most perhaps is discount far from UK. (1993) that of rates to political rigging the such exclusive show and Bateman

6.50

other commentators,particularly with respectto the discounting of global warming damages (perhapsthe most potent challenge to intergenerationalequity in the history of man). While not stating any particular rate, Arrow et al. (forthcoming) do make explicit reference to the in his Cline (1992) by from 0-2% economic analysis of long-run climate change range used models.

Similarly, in his evaluation of the social costs of greenhouse gas emissions,

Fankhauser(1993) uses a central (mode) estimate of r, = 0.5% with a range from 0-3% (the for being comparative purposeswith other studies79). mainly end upper A further complication arisesfrom the issue of multiple discount rates; the notion that diverge between radically projects to the extent that a single discount may social preferences As Arrow (forthcoming) is oversimplification. of an et al. and many earlier rate somewhat here is factor i. key have the out, substitutability, e. the extent to which pointed commentators development benefits (often in terms of man-madecapital, K. ) be traded off against costs (generally in terms of natural capital, K. ). Assuming for the moment that sustainability is be in both desirable can that capital measured of sets somecomparablenumeraire and socially (presumably money), then perfect substitutability would mean that any project would simply have to pass a standardHicks-Kaldor hypothetical compensationtest" to be sanctioned. in 'very has been development this termed the literature weak sustainability' of sustainable the benefits 1993) (total Pearce, that, total (Turner states provided net which capital) are and rule be This be sanctioned. may perfect substitutability assumption a project may non-declining, (e. Sitka into K, /K,, for into thence swops g. spruce plantations paper some more acceptable (e. for destruction SSSI's back than the to to others g. of plantations) new make so and money irreversible. destruction is i. K. for e. some motorways"), way Bateman (1991) suggeststhat we can define a continuum of capital types from money (the purest form Of K. ) through various types of K, (trees, land, etc) to 'critical natural 82 the latter being tlýose services of the planet vital to life-support (climate and ), (K: capital' from layer, As this move away money along etc). we continuum, ozone control, atmosphere falls it for than staying constant, rather substitution, until reaches zero with so the potential Kn--

79Forexample Nordhaus (1991ab,c). (1986). Pearce for 8OSee text, example, cost-bcnefit almost any "As in the case of the M3 Twyford Down extension. from (1990). Pearce Turner is borrowed 92, and rhe term

6.51

Such a view causesproblems for cost benefit analysis if we feel that the building up of K. does not adequately compensatefor the loss of Kn. This is the view of the 'weak sustainability' rule (Turner and Pearce, 1993) which argues that stocks of Keeshould be inviolate, while Kn should be subject to some safe-minimum-standard(SMS), below which further interpretation, A be the 'strong-sustainability' rule, in effect prohibite&'. use should been has breachedand that any further use of K. should be SMS that already such a argues in by terms of shadow projects restoring, transplanting compensation physical actual offset in future Y, levels used projects". of any or recreating The divergence betweenbest estimatesof r, given by Pearceand Ulph (r, = 2.4%) and Fankhauser (0.5%) or Arrow et al. (implicitly 0-2%) can therefore be viewed as comparing KJK. substitutability with that of a non-substitutablegood: global climate. of rate a general The implication of such an analysis is that, becauseof the various rates of substitutability and irreversibility inherent in the differing capital baseof eachproject, society will have different discount rates for different projects. Furthermore, we could extend this line of reasoning to benefits in forestry individual that, single project of a so and our costs case study, UK the losses higher (for reversible) might attract are reasonably a r, than recreation which timber benefits (which arguably belong to a more depleted set of K,), which is more discountedthan K: Following (which to the stock climate of global services). this contributes carbon storage impact final in discount the analysis of results examine we of using multiple our argument forestry in case study. our rates In practice, the variance of r, within a project is clearly a decisionmaking nightmare for discount 'management' Indeed the avoidanceof the abuses. rate potential and opens up for In be the adopting a single coherent argument rate most policy. a review, may abuse Henderson and Bateman (1993) report numerous examples from around the world of both inter-and intra-project multiple discount rates. However, these appeared to be almost by than rather empirical evidence regarding objectives policy motivated exclusively discount Ile to management of rates give policy-favoured projects preferences. underlying

33Under weak-sustainability further us'e Of Ka up to the SMS must still be compensatedfor by reinvestment (savings) of the appropriate level of K. proceedsfrom each project Crurner and Pearce, 1993).

K. both in termsof K. savingsandby "Under strong-sustainability an individualprojectmustcompensate fund. Suchphysicalcompensation to compensation, an offset physical shadow project contributions appropriate Hicks-Kaldor hypothetical (re A (very the than be rejecting rule). still stronger view strong rather must actual have its K. (see that must own actual each project physical compensation states shadow project sustainability) Turner and Pearce,1993).

6.52

a spurious sheenof financial respectability is widespread and to be avoided. The desirability of a single rate is thereforeclear. The Pearceand Ulph (1995) results (central estimate r, = 2.4%; range = 2-4%) are useful here but we have to recognise that probably the recreation bencfits, and almost certainly the carbon sequestrationbenefits, of woodland would attract a lower than average rate of public pure time preference. Accordingly we have chosena sensitivity analysisfor r. which includes one rate (1.5%) which falls below the Pearce and Ulph range" and another which is the centre of that range (3%). For comparative purposes we have also employed the Treasury's (6%) rate throughout, Pearce Ulph "very difficult justify". that this the to of and sentiments seems echo although we 6.6.2.1: Hyperbolic social discount rates The standarddiscount function is most commonly expressedas the quotient shown in equation (6-11): DF,

where: DF,

discount factor in year t

r

discount rate

t

time in years from the start of project (t--0,1,2,,,F)

While this is perfectly adequatefor discrete time periods, discounting over continuous formula known the is equivalent of a mathematical quotient using as often performed time in (6.12): format equation'6 as shown the negative exponential

DF,

e-P'

c

2.718,the baseof naturallogarithms

where In (I+r)

(1993) Fankhauscr lower Arrow (forthcoming). 8-sThis the range estimates of and et al. reflects also gTrice (1993)discussesthe slight differencebetweenthe two definitionsof the discountfactorpresented in equations(6.11)and (6.12).

6.53

The exponential nature of this relationship is taught as a first principle in most basic microeconomics coursesand, if it is questioned at all, is usually justified with reference to the opportunity cost of capital link between the discount and interest rate, the latter being compounded at a positive exponential rate. However, basic textbooks in behavioural science teach, with equal certainty, that "researchindicates that the discount functions usually take the form of a hyperbola", (Rachlin, 1991)" as shown in equation (6.13): DF, =I

(6.13) (1+rt)

As indicated by this quotation, such a statement is based upon empirical findings. Indeed hyperbolic discount functions have been observed as the norm in the behaviour of adults (Mazur, 1987), children (Mischel and Baker, 1975; Mischel et al., 1989), non-human birds (Rachlin 1971) Green, (Menzel, 1972) suggestingthat this may and even and mammals be a genetic trait common to all sentient beings. Some behaviouralists have attempted to bridge the gap between their discipline and by be 1989) (see Rachlin, to a mixture of silence and dismissal". met only economics However, such xenophobia may now be breaking down as economists begin to run have behaviouralists those to carried out for more than two which experiments similar

decades. In one such experiment, Cropper et al. (1992)89attempt to estimate r, for future human lives. Some 3,200 US householdswere interviewed with respondentsbeing presented one only of which the government could hypothetical control programmes pollution two with lives Y X today Programme lives and programme fund. save would would to save at afford Although future. in in each respondent the fixed was presentedwith time some years point a programme, the future present-day acrossthe total sample to against weigh scenario only one five future alternatives were examined, fixed at 5,10,25,50 and 100 years in the future. be funded allowed calculation of an Answers to questions as to which programme should implied exponential discount rate which, rather than remaining constant,declined from 16.8% (1999). in Logue field introduction given to this "See also the excellent "Loewenstein and Thaler (1989) conclude that "Many economistsview the,researchon the psychology of decision making as a nuisance". (1992). Portney Cropper in and 'Turther results are reported

6.54

for lives saved in year 5 (against lives saved now), to 11.2% in year 10, to 7.4% in year 25, to 4.8% in year 50 and, finally, to 3.8% in year 100. Henderson and Bateman (1995)

examinethe mathematicalrelationshipsunderlyingtheseresultsand find that the hyperbolic function given in equation(6.13)providesa near-perfectfit (W(adi) = 99.6%)". Given the weight of evidencefrom behaviouraland now economicresearch,why is the economic profession generally so dismissive of hyperbolic discount function. Two first being the link between discount and interest rates via the reasons appear predominant, the opportunity cost of capital. As interest ratescompoundexponentially so we would expect This but it is discounting. for seems a sound argument perhaps strongest exponential market individual is by lifespan discounting the to emphasiseshort run constrained where and private do We discount the therefore exponential not attack nature of private costs. rates. opportunity However, the link with short run interest is less strong for public discount rates and this brings us to the second, and we believe underlying Neoclassical objection to hyperbolic discounting: the problem of preferencereversal. Table 6.18 considersa particular project commencing in 1995 and yielding annual net benefits thereafter. Each cell in columns (1) to (4) gives the ratio of the presentvalue of one pound received at the start of the specified year and one pound received at the end of that from (2) (1) Columns this the as seen give relationship project start year using, and year. hyperbolic discount functions (r 6% in throughout). and = exponential every respectively, is discounted the at start a pound received of a year of more than that of a value case the

(i. However, that than the e. all ratio of year values are greater end one). at pound received discounting this year start/yearend presentvaluerelationshipis constant exponential using for hyperbolic discount function lifetime the the the of project, whereas the throughout hyperbolic discount functions This declines that time. means over give relatively relationship do functions benefits discount delayed than the exponential of to net same more weight (2) do (1) for to This seem pose problems and not projectappraisal. columns said, rate". Such problems arise when we comparecolumns (1) and (2) with columns (3) and (4). discount latter again show year start/year end ratios but now assessmentis These columns

90optimumvalueOf Tis 21% (t value= 36.63). An interceptterm,allowedfor in our estimatingequation, insignificant from be zero. strongly proved to "The annualrateof presentvaluedeclineis progressivelyslowerunderhyperbolicdiscountingwhereasit (at discounting. higher exponential any give r) under remainsconstantly

6.55

in (3), before As in 2015. the exponential ratios, column are constant acrossall years made (4), decline in hyperbolic column with time. the ratios, while

This means that, using

between (3)) in 1995 (1) (columns discounting a choice projects and made will exponential hyperbolic in in 2015. However, the comparison of ratios remainoptimal whenreassessed be for (4) (2) this that so now the sameproject year not always may suggests columns and in different discount has different from time ratios. very points when assessed Table 6.18:

Year

Start/end of year discount ratios from exponential and hyperbolic discount functions (r = 6% throughout)

Project age in years W

0 20 30 50 70 90

1995 2015 2025 2045 2065 2085

Assessmentin 1995

Assessmentin 2015

Discount ratio (exponential discounting) (1)

Discount ratio (hyperbolic discounting) (2)

Discount ratio (exponential discounting) (3)

Discount ratio (hyperbolic discounting) (4)

1.06 1.06 1.06 1.06 1.06 1.06

1.0600 1.0288 1.0214 1.0137 1.0118 1.0096

n/a 1.06 1.06 1.06 1.06 1.06

n/a 1.0600 1.0340 1.0219 1.0152 1.0116

A benefit Project yields a single net real illustrate two of To this consider projects. in 000 benefit 2026. B 2025 in of a single real net million while project yields E97 million 1995 2015, two and details calculated assessment points, 6.19 at our present values Table 6% (discount discounting hyperbolic throughout). rate = both and exponential under

hyperbolic between differences interesting 6.19 and exponential Table revealssome discounting.

Because exponential discount factor curves decay much faster than their

initial in 1995 discount (at the assessment rate) shows any given hyperbolic equivalent if benefits A immediate favouring the the smaller of project while more curves exponential for delayed leads hyperbolic to the future a preference curves more of weighting enhanced is 2015, because B. When to the changed benefits of assessment point of project but larger discounting ratio, exponential curves still give the sameresult that project its constant annual decline hyperbolic initial its However, the of a to curve, steep compared favoured. is A

6.56

(see figure 6.10) is A that shape means now project preferred relatively shallow subsequent have i. B, reversed. to project e. preferences Table 6.19: Hyperbolic discounting and preferencereversal Project

A B

Yield year

2025 2026

Undiscounted net benefit

Presentvalue assessedin 1995 (r = 6%)

C97M ; CIOOM ;

Decision Project

A B

Yield year

2025 2026

Undiscounted net benefit

Exponential discounting

Hyperbolic discounting

f. 16.9M E16AM

C34.6M ; L35.OM

Prefer A

Prefer B

Presentvalue assessedin 2015 (r = 6%)

C97M ; MOM

Decision

Exponential discounting

Hyperbolic discounting

C54.2M ; f:52.7M

f.60.6M E60.2M

Prefer A

Prefer A

The apparentinconsistencyof suchpreferencereversalshave beenviewed with horror behaviouralist literature. Such have have the results considered by those economists who "Dynamic "Myopia Inconsistency"; "Anomalies"; headings and as such been presentedunder for implications Certainly Discounting"". "Myopic the project appraisal Inconsistency"; or finds literature, behaviouralist However, be the no which problem to major. appear would dependent be decisions their that change time and people might that might with the notion in the that of subsequently changing time', possibility notes closer come options mind as individuals deliberately Here 'commitment' leads strategies. to aaopt people often mind ones into for by for lock example entering reversal the preference possibilities to out paths choose 1988; (Logue, Rachlin, 1991). investment contracts long term repeated

"For a review see Hendersonand Bateman (1993). "An analogy (from Rachlin, 1991) is that of choosing betweenstudying for an exam or going to the cinema. However, for distant to the particular evening. on a study "am may wish student a When both choices are few (as days hence) human (with to the the a opposed still a neoclassical exam student arrives evening that when film Whether internal his than the rather studying. or not such and see mind change model) may well crack, is interesting do basis form the public policy choices an to of question we which individual rationality should answer. have ready a not

6.57

Figure 6.10: Discount rate sensitivity analysis: Discount factor curves for chosen hyperbolic discount and rates exponential

I

0.8

Co

0.6

ýz4 4.a

.

50.4 ý2

C:) 0.2

0

25

50

75

100 125 Years

150

175

200

Given that the adoption of'strategies to avoid preferencereversal implies that we wish inconsistency the the arguments surrounding positions, assessment of to retain original hyperbolic discounting seem to have more to do with defending standard economic behaviour. Consequently, in face the preference than public of observed assumptions rather (1992) Cropper from the and array vast of empirical studies that al. et of as evidence such implications it is feel the that pursuing worthwhile of adopting behavioural sciences,we the

hyperbolicdiscountfunctionsin our wider research. 6.6.3: DISCOUNT RATES: CONCLUSIONS

Given the major impact which alterationsin the discountrate will haveupon longfeel discussion highlights the need to adopt a sensitivity that forestry our we delayed returns, Considering issue. discounting first feel this exponential to that we real approach analysis 3% justified 1.5% here. discount are well and as a reasonable range rates of social included is for 6% Turning Treasury's comparative also to purposes. rate the Furthermore

discountrates,the 1.5%and 3% ratesare useful for assessing farmers private real consider 6.58

decisions in the Welsh non-dairy agricultural sector.

Conversely rates of 6% and 12%

limits describe the reasonable which we may wish to apply to dairy farms in Wales. roughly We feel that their is a sufficiently strong case for assessinga hyperbolic social (if not

have We hyperbolic function. discount chosen a rate of 6% as this givesa discount private) function curve which initially lies below the V/2% exponentialcurve but is then cut from become highly delayed latter long benefits by the to curve the which most values net above (see figure 6.10)"'. We feel that there is a strong theoretical and empirical case to suppose that this more accurately reflects public preferencesfor public goods in the long term.

6.7: THE PRIVATE VALUE OF TIMBER PRODUCTION From the discussionsof this chapter we can seethat the private value of a productive factors: four broad by determined is of categories plantation i. Plantation costs As iternised in tables 6.9 (for conifers) and 6.13 (for broadleaves). Plantation timber benefits felling. Crucial factors here are both thinnings maincrop and These arise through be to following assume (which we constant) and yield class. analyses, our future real prices

iii. Grants and subsidies farmer is permittedto to the schemes which discussed As thesewill vary according defined disadvantaged in is farm agriculturally a area; the and not or whether under, register 6.15 (for broadleaves) (for bring 6.12 and Tables together land. conifer) the the prior use of Welsh hectare of for Productive typical woodland. a benefit streams subsidy and cost,

be farms attaining even Howeve not 1.5%. may rates of return these of of "We recognisethat a number 'r, join

probably long will and term the from in the ongoing this exodus sustainable are not feel these many of we farms. non-dairy sustainable the to remaining, apply our therefore rates sector. factor here. background a 9-1he Treasury's 6% (exponential) rate was also

6.59

iv. Discount rate We argue that this will be significantly higher for dairy as opposedto non-dairy Welsh

farms. All these factors were brought together by inputting data from the FC yield models (Edwards and Christie, 1981) for Sitka spruce (YC6-24) and beech (YC4-10) to a series of MINITAB worksheets(MINITAB, 1994). This allowed easymanipulation of all assumptions (e.g. grant schemes,discount rates, optimal fefling age, etc) to produce a full range of from in full in 4.397. Results As these this are exercise reported appendix private values. just (exponential) discount here. Figure 6.11 one rate scenario reproduce we extensive are for for 3% discount full Sitka YC's the rate a range of spruce and equivalents annuity graphs in (detailed figure). 6.12 Figure feasible, to the notes registrations repeats grant scheme all this analysis for beech. For both Sitka spruceand beechwe can seethat, as expected,annualequivalent values fall discount 4.3). As Oust YC they rate; see appendix with subsidy schemesare as rise with between difference is linked the scheme payments timber constant within to productivity not YC. Comparison between Sitka spruceand beechis interesting as it shows that, holding YC broadleaves from higher for 10), 6,8 is This YC (Le, than are returns or conifers. constant due to higher prices and subsidy levels for broadleavesand despite the shorter felling age of higher YC broadleaves because However, than capable of much are conifers and, conifers. have high lower felling because (thus importantly, plantations much such yield ages more long broadleaves), discounting they rotation upon visited can provide the severe avoiding Furthermore, broadleaves. than typically as conifers higher perform equivalents annual much better (in YC terms) than broadleaveson any given piece of ground, the financial attractions broadleaves. those to of outstrip of conifers appear

broadleaves). 6.14 (for 6.11 (for 96SCt conifer) and tables as per

9'Appendix 4.3 reports NPV, perpetualsum NPV and annuity equivalents for eachdiscount ratc/YC/subsidy scheme permutation.

6.60

Figure 6.11: Farmers private values for Sitka spruce (annualisedequivalents of a perpetual series of optimal rotations: E/ha; r--3%). Various yield classesand subsidy types. 600 500Annuity Value 400E/ha (r=3%) 300-

. ..........

200-

"OOL

8

12

1,6

2,0

2L4

YIELD CLASS Slnda+CW SIda+CW SIsda+CW Slnda-CW

............. ---------

SUsda+CW SUda+CW SIda-CW SUnda+CW

........

Slsda-CW SUsda-CW SUda-CW SUnda-CW

------

--

Note: Subsidy permutationsare codedas follows: S1 = subsidy paid on land which was recently improvedgrassland/arable SU = subsidy paid on land which was formerly unimprovedgrassland disadvantaged area a not nda. = da, = disadvantagedarea disadvantaged area specially sda = supplement not paid woodland community = -CW +CW= community woodlandsupplementpaid For rates of subsidy paymentsseesection6.5.

The impact of discounting is fully documented in appendix 4.3.

However, an

in 6.20. Here for is highest Sitka table annualised equivalents given output spruce overview (YC 24) and beech(YC 10) underone subsidypermutationare given for all discountrates (including some,suchasthe hyperbolic6% rate,which areprobablyinappropriatefor private farm decisionmaking,but are includedfor comparativepurposes).

6.61

Figure 6.12:

Farmers private values for beech (annualisedequivalents of a perpetual series E/ha; Various r--3%). rotations: optimal yield classesand subsidy types. of ;

250

200

.

Annuity Value C/ha 150 (r=3%) 100

50

z----::

::: b

-

a

10

YIELD CLASS Slnda + CW Slda+CW Slsda+CW .............. Sinda-CW ----------

SIda-CW SUda+CW SUsda+CWj SIsda-CW ......................

SUnda+CW

SUsda - CW SUda -CW SUnda-CW

Note: Subsidy codesas per table 6.11.

Table 6.20:

Farmersprivatevaluesfor YC 24 Sitka spruceand YC 10 beech(annualised C/ha). Various discount of optimal rotations: series of a perpetual ; equivalents SUnda-CW. Subsidy option = rates.

Discount rate

Farmers private value (annualisedequivalent; f/ha) Sitka spruce (YC 24)

1.5% 3% 6% 12% Hyperbolic 69o'

496-30 388.46 219.36 19.45 864.80

Beech (YC 10) 103.54 80.68 31.21 9.59 335.68

in for disadvantaged grassland, previously unimproved not a areaand without subsidy SUnda-CV = Notes: community woodlandsupplement.

6.62

In subsequentchapterswe examine how such forest values compare with returns to forecast farmers likely conversion rates under present and possible and existing agriculture future subsidy schemes.

6.8: THE SOCIAL VALUE OF TIMBER PRODUCTION in movingfrom theprivateto the socialvalueof timberproductiona numberof issues basic The (thinnings benefits be timber costs and plantation and to maincrop) addressed. need in form. Unlike be defensibly an unaltered agricultural produce, timber prices are used can UK domestic is by intervened that the and, timber controlled given or otherwise price set not Appendix 4.1), (see we see no clear reason to embark on a market the competitive world do have However, to we subtract all grants and subsidiesas these exercise. adjustment price baseline This for benefits timber gives us our transfer social payments. value net are simply beech YC discount in Sitka for 4.4. detail across all spruce and and rates appendix we which As discussedin our opening chapter, the social value of a woodland is more than just (Bateman, identify 1992) In discuss therein. timber earlier work we and a the value of detailed set of environmental and non-environmental non-market costs and benefits which Here discussion by briefly from that we summarise afforestation. considering the may arise be items from to need considered may when moving which a private to a major non-market social assessmentof woodland. 6.8.1: NON-ENVIRONMENTAL

NON-MARKET

SOCIAL COSTS AND BENEFITS

Here we discussfour major issues"; national security; economicsecurity; import substitution; and employment.

j. National security

While this formedthe impetusfor the creationof the ForestryCommissionjust after important spur to planting post-WWII, the prospectof the UK being WWI, and was an for from timber supplies any extendedperiod seemsrather unlikely. We receiving blockaded benefits is be derived from there to that no significant security the national conclude therefore

"Further details in Bateman (1992).

6.63

domestic supply capability. a of expansion ii. Economic security While not of strategicimportance,uninterruptedsecurity of supply doesbring avoided-

issue, Pearce(1991)statesthat "an evaluationof the chances benefits. In this of a study cost interruptions increment in that suggests supply a small other and prices of 0.2 of embargoes to 0.8 per cent to reflect the shadow value of economic security would be justified". Accordingly timber benefits were increasedby 0.5% in our social evaluation models". Import substitution

Although timberformstheUK's fourth largestimport item (FICGB, 1992),the theory industries do that state-support of shows which not enjoy such advantage comparative of benefit. does inefficient is not constitute a social and advantage iv. Employment

It has beenarguedthat creatingjobs in forestry is a good way to stemthe ongoing depopulation and combat the psychological and other economic costs of rural trend of rural have forestry is However, that studies suggested numerous a relatively unemployment. inefficient of providing therefore method rural employment, particularly when and expensive 1972; Laxton (H. M. Treasury, Whitby, 1986; NAO, 1986; and to agriculture compared Johnson and Price, 1987; Evans, 1987). Forestry expansion could therefore be seen as is do feel likely in However, Welsh farmthis the not case we of costs'00. creating shadow in feel that, the absenceof a specific study, though should be ignored therefore forestry and benefits in indicating imbalance versus agricultural private woodland net the net any with benefits. flows level farm of social employment

can

In conclusion the only clearly valid non-environmental non-market social benefit we isolate is a small benefit due to increasedeconomic security of supply.

increase, is single not a compounding of an annual real price increase. across-the-board this an that is very small. effect the net Consequently 100'rhis may be becoming less true as Forestry Commission employment has been failing, and productivity Commission, 1979,1989,1994a; (Forestry Thompson, 1970's 1990; late FICGB, 1992). the since rising, "Note

6.64

6.8.2: ENVIRONMENTAL

NON-MARKET SOCIAL COSTS AND BENEFITS

Woodlands create a myriad of social benefits and costs of which we discuss the following major issues"':

recreation; carbon storage; non-user (bequest and existence)

impacts. values; and acidification

i. Recreation This is the major focus of our evaluation research as discussed in chapters 2-5. Because of the potentially significant problems of declining marginal utility",

we have

decided not to incorporate such benefits within the plantation value models presentedin this deal benefits timber Instead to these primarily with values models which recreation chapter. in are added subsequentchapters'03 . ii. Carbon sequestration As with recreation we have chosento deal with carbon sequestrationseparatelyfrom diminishing for (as is because in This of marginal utility, NPV explained not models. our later chapters) the likely levels of sequestrationwill not have a significant impact upon the issue which we feel deserve because but the this complexities of of rather globalC02budget, particular separateattention.

iii. Non-uservalues Travel cost and other revealed preference evaluation methods only address users difference between To for the this and the total extent some a resource. direct-use value be I in discussed can addressed via expressed chapter preference concept value economic However, to the site surveys are still restricted values valuation. contingent as methods such hold body to that suggest exists non-users Yet research may of a considerable of users. der Van Linden, 1987; Willis Benson, (Oosterhuis for and and woodlands significant values Diamand 1990; Bateman, Langford, 1995; Bateman Walsh 1990; and Kristr6m, al., et 1989;

IOTurtherdetailsin Bateman(1992). in increase in the to expect recreation an result opportunities expands would 102As we the areaof woodland in per hectarerecreationvalues. Givensupplyanddemandconditionswe wouldnot expect decline observable for timberproduction. be problem a this to that the monetaryevaluationsof woodlandrecreationare "As discussedelsewherethis implicitly assumes landscape. the present agricultural of value the amenity surplusesto

6.65

because These 1996). non-userssee woodlands as a store of indirect use arise values et al., (to both bequest in future), (landscape the value others now and amenity) and existence value habitat, (biodiversity, etc)104. wildlife value These are quite clearly significant values. We have spent some time considering the

landscapeamenity problem (Bateman,1994)" and our ongoing researchis examining be The issue is by quantified". might pure such values non-use more which methods from I A braver that than might suggest results our non-userstudy at man problematic. Wantagemight provideestimatesof this issueandindeedthis wasoneof the prior objectives interviewed in But the that experimentalmost out, sample as pointed of that experiment. Furthermore itself than rather as potential users actual non-users. our other uniformly viewed 1992,1995; Bateman Langford, forthcoming) (Bateman have CV and al., et surveys non-user here is be tool technique crude the a rather which may well eliciting the that suggested swarm-glow' type of responsediscussedin chapter 2. Certainly we feel that many of the in CV the the are most valid context of non-user surveys. approach of criticisms major So we are left with having to acknowledge a possibly significant deficiency here. While we feel that our analysis is relatively sophisticated and useful in a policymaking is Our into far from it this ongoing examining research evaluation perfect. remains context, issue"7 but we cannot be certain of successhere and only feel confident of predicting a in time. improvement over methods relative iv. Acidification In our detailed review (Bateman, 1992) we show that forests are both the victims and

While Forestry Commission damage"'. the that suggests acidification of perpetrators industrially for fixing medium emitted atmosphericacid forests tend to act as a catalytic is in the that that this 1987), part of story and only conifers particular suggest (Innes, others

"Generally thesevalueswill be thoughtof as positive(i.e. woodlandsproviding benefits)but they may landscape biodiversity Sitka (e. be reduce single age spruce plantations amenity and g. negative equally well RSPB, 1987,1988). 1: 1987,199 1; Price, 198 Moss, Newton and value: see (1990) in Willis (1992a, bc) Helliwell 1. Garrod and and 105See appendix of also review 1160urongoingresearchin this areaexaminesdie applicationof CIS viewshedandnoiseshed routinesto the is by data from by ESRC Professor Duncan This the sponsored and assisted research hedonicpriCingmethod. Glasgow) Ordnance Survey. University the (CHRUS, and of MacLennan is demand for here by the regarding 107particularly ongoing research our naturalareas(sponsored relevant indebted). ESRC to are we the whom Nature and English "Acidification of both waterwaysand soil is a recognisedproblem.

6.66

directly contribute to a lowering of pH levels (see Harriman and Morrison, 1982; Batterbee, 1984; Nisbet, 1990; and the particularly relevant papers in Edwards et al., 1990"). We take the position that whether or not forests actually generatethe acids concerned,they are in increased linked to of aquifers acidification non-buffered areasand therefore significantly do generate costs. Our research in this area has not progressed beyond the level of a literature survey; however, this has shown that the acidification problem is eminently intend GIS to proceed with shortly. to we analysis which amenable 6.8.3: NON-MARKET

SOCIAL COSTS AND BENEFITS: SUMMARY

Those items which we feel to be of major significance (recreation and carbon in Of dealt the remaining our model other outside rotation chapters. with are sequestration) justify benefits, to arguments a minor upward security seem economic social costs and issues insignificant benefit most other appear while values with the revision of social Both (and acidification) values. possibly of these are the subject of exception of non-user is this that that must remain a accept partial analysis until and we work ongoing research improvement defend Nevertheless the present study as a significant on we would complete. decisionmaking CBA considerable use. models and of existing 6.8.4: ANNUAL EQUIVALENT

SOCIAL VALUES

We can now calculatesocial net benefit values for our plantationmodels. These benefit the costs categories with and exceptionof recreation the social above all encompass dealt (which separately) with and non-user and acidification are sequestration carbon and In items dealt those effect, after omitting (which the of ongoing research). subject are values investigation, the those only value of economic security still of under and elsewhere with base benefits be Full to timber. is to social our net added of quantified supply sufficiently in form 4.5. As is in there tabular from appendix no subsidy appear this exercise results illustrate resultsacrossall YC and discountrateson can these calculations we dimensionto in figure 6.13, for figure dimensions) (in three two and and conifer, shown a single graphas 6.14 for broadleaves.

101hiscollectionof papersfocussesexclusivelyuponacidificationin Wales.

6.67

Figure 6.13:

Social value for Sitka spruce (annualised equivalent of a perpetual series of optimal rotations). Various yield classes and discount rates.

1000 800 600 400 200 0 -200

F' 6% HYP 2%

y k-Iui YC 14-

YC18

ý' 6% 12%

Yield Class

Discount Rate

1000 800 600 400 F4 200

0 -200

YC6

YC8

YCIO

YC12

YC14 YC16 YC18

YC20

YC22

YC24

Yield Class

6% HYP -m-2%

-0-3%

6.68

6% -c-1-

12% -vA-

Figure 6.14:

Social value for beech (annualised equivalent of a perpetual series of optimal rotations). Various yield classes and discount rates.

300 u zi 200 100 flyp

0 -loo -200

nt . Yield Class

300

200

loo

0

100 -

-200

YC4

2% HYP 6% -m-a*--

YC6 YC8 Yield Class: Beech

3% -a,--

6.69

6% -o-

YCIO

12% -za-

Comparison of figures 6.13 and 6.14 show relative relationships similar to those observed in the private sector evaluations. Again we see (on this restricted range of value types) conifers having the ability to outperform broadleaves. Given that we have excluded does such a result non-user values, not appearunusual. and recreation

6.9: CONCLUSIONS We have constructed rotation models which take into account plantation costs and benefits, real prices, grants and subsidies. We have also consideredthe difference between in differential both discount terms of assessments rates and with regard to social private and the differing range of values which either assessmentshould appraise. In subsequentchapters from incorporated derived this analysis are with. our assessments the private and social values to provide our overall assessmentof the values values sequestration carbon and of recreation These farm-forestry. for by then values are those compared with existing generated likely to conversion rates under a variety of scenarios. agricultural activities so as assess

6.70

REFERENCES Alston, J.E. (1986) An analysis of growth of U.S. farmland prices 1963-82,American Journal of Agricultural Economics,68: 1-9. Arrow, K. J. Parikh, J. and Pillet, G. (with contributions from Babu, P.G., Beltratti, A., Chichilnisky, G., Fankhauser, S., Faucheux. S., Froger, G., Gassmann,F., Haites, E.F., Hediger, W., Hourcade, J-Ch., Morgan, M. G., Parikh, H., Pearce,D.W., Peck, S., Richels, R., Schubert, R., Suarez,C. and Wellenmann, U.) (forthcoming) Decision Making Framework to Address Climate Change, report by IPCC Working Group III Writing Team H, Intergovernmental Panel on Climate Change,Geneva, Switzerland.

Bateman,I.J. (1987)Forestryin the UK: a standardand socialdiscountingapproachto investment evaluation,MA Thesis,Departmentof AgriculturalEconomics,Universityof Manchester. Bateman,I.J. (1988a)Futurereal timberpricesin the UK market,mimeo,Departmentof Economics, University of Exeter. Bateman,I.J. (1988b)The financialpotentialfor farm-basedforestryin the UK following recentgovernment 6(11): Management, 473482. Journal Farm developments, of policy BatemanI.J. (1991)Socialdiscounting,monetaryevaluationand practicalsustainability,Townand Country Planning,60:176-178. Bateman,I.J. (1992)The United Kingdom,in Wibe, S. andJones,T. (eds)Forestry:Market and InterventionFailures,Earthscan,London. Bateman,I.J. (1994) The contingentvaluationandhedonicpricing methods:problemsandpossibilities, LandscapeResearch,19(l):30-32. Bateman,I.J. and Langford,I.H. (forthcoming)Non-userswillingnessto pay for a NationalPark:an Regional Studies, in the method, contingent valuation press. of critique and application Bateman,I.J. and Mellor, C. (1990)The UK timbermarket:an econometricmodel,041ordAgrarian Studies. 18(l): 53-61. (1995) Households I. H. Langford, E. Diamand, J., I. willingnessto pay and farmers and Bateman, CSERGE for GEC Working a community establishing to woodland, compensation accept willingness Paper, Centrefor SocialandEconomicResearchon the GlobalEnvironment,Universityof East Anglia and UniversityCollegeLondon. A. P. (1996) Household I. H. Jones, E., Langford, Diamand, J., and willingnessto pay and I. Bateman, for establishinga recreationalwoodland, Journal of farmerswillingnessto acceptcompensation 39(l): 2143. EnvironmentalPlanningand Management, (1995) A Graham, A. H. I. Langford, J., surveyof non-uscrswillingnessto pay to prevent I. and Bateman, Global Environmental Change CSERGE Working Paper Norfolk Broads, GEC in flooding the saline 95.11, Centrefor Socialand EconomicResearchon the Global Environment,Universityof EastAnglia London. College University and P., Langford, 1. Turner, Garrod, G. R.K. (1992) Recreationand D., Doktor, G., K. Willis. and Ii., Bateman, EnvironmentalPreservationValueof theNorfolk Broads:A ContingentValuationStudy,Reportto the National Rivers Authority, ppA03. lakes. Philosophical Transactions Diatom (1984) the W. acidification of R. and analysis of the Batterbee, Royal Societyof London:SeriesB, 305:451-477. Economic American Review,58:788-802. discount, On (1968) the WJsocial rate of Baumol, in S. (eds) Germany, Wibe, G. (1992) Jones, T. Forestry: Pfister, Market M. and Behrndt, V., and Bergen, London. Earthscan, Failures, Intervention and Times, Financial 10th March 1988, London. Unlovely trec, (1988) as a B. Bloom, behavior literature, investment Discussion (1989) Farmer E. L. Due, the La a review of T. B. Brase, and Paper 8915,Departmentof AgriculturalEconomics,CornellUniversity,N.Y. White Warming, Horse Press, Cambridge. Global Costs Counting (1992) J. the of Broome, in land Town British Country June A Planning, 1985. (1985) turning use policies, P. point and Burnham, formula for farmland Econometric the capitalization (1986) of prices,American R. modeling O. Burt, journal of Agricultural Economics,68:10-26. Digest Statistics: (1993) Monthly (CSO) February 1993, No. 566. Office HMSO, of Statistical Central London. Global Warming, Institute for International Economics The Economics,Washington (1992) of R. W. Cline,

6.71

D.C. Cline, W.R. (1993) Give greenhouseabatementa fair chance,Finance and Development, 30(l): 3-5. Collins (1991) Atlas of the World, Collins, Edinburgh.

Cooper,D.N. (1992) An analysisof the agriculturalinput industries,UnpublishedPhD Thesis,Department of Economics,Universityof Nottingham. Councilfor the Protectionof Rural England(CPRE)(1992)TheLost Land: Land UseChangein England 1945-90,CPRE,London. Summer.1-4. Cropper,M. and Portney,P. (1992) Discountinghumanlives,Resources, Cropper-M., Aydede,S. andPortney,P. (1992) Ratesof time preferencefor savinglives,American Zonomic Review(Papersand Procee&ngs),82.469-472. Cunningham,S. (1990) An economicanalysisof depreciationand investmentin UK agriculturalmachinery. PhD Thesis,Departmentof Economics,Universityof Exeter. Daly, H. (1977)SteadyStateEconomics,Freeman,SanFrancisco. Daly, H. and Cobb,J. (1990)For the CommonGood,GreenprintPress,London. Doran,A. (1979)Economicsof forestry:a reviewandcritical appraisal,Bulletin No.168, Manchester University. EconomicCommissionfor Europeand the EuropeanCommunity(ECE)(1993),ForestConditionin Europe Survey,UN/ECE,Geneva. 1992 Results the of ForestryCommissionBooklet Edwards,P.N. and Christie,J.M. (1981)Yield modelsfor forestmanagement, 48, HMSO, London. Biologicae Edwards,R.W., Gee,A.S. and Stoner,J.H. (eds)(1990) Acid Watersin Wales,Monographiae Volume 66, Kluwer AcademicPublishers,London. Evans,A. (1987) The growth of forestryand its effectsuponrural communitiesin North EastScotland:the 310-313. 41: Scottish Forestry, Strathdon, caseof

Ewers, HJ. et al. (1986) Zur monetArenbewertung von urnweltschAden:mcthodische untersuchungam bcispiel der WaldschAden,UmwelibundesamtBericht 4186,Berlin.

issues Global (1993) S. andstateof the art, Global EnvironmentalChange economics: Fankhauser, warming WorkingPaper 93-28,Centrefor SocialandEconomicResearchon the Global Environment, University of EastAnglia andUniversityCollegeLondon. Earthscan,London. Fankhauser,S. (1995) ValuingClimateChange:TheEconomicsof the Greenhouse, Farm BusinessSurvey(1988) Farm BusinessSurveyin Wales:StatisticalResultsfor 1987188,Department University The Wales, Aberystwyth. Economics, Agricultural Economics of and of in Statistical Survey Wales: Resultsfor 1988189, (1989) Farm Business Department Survey Business Farm The University Wales, Aberystwyth. Economics, Agricultural Economics of and of Farm BusinessSurvey(1990) Farm BusinessSurveyin Wales:StatisticalResulisfor 1989190,Department University Wales, The Aberystwyth. Economics, Agricultural Economics of and of Statistical Wales: Resultsfor 1990191, Survey in Department Business (1991) Farm Survey Business Farm Wales, University Aberystwyth. The Economics, Agricultural of Economics and of Wales: Statistical 1991192, Survey in Resultsfor Department (1992) Farm Business Survey Business Farm Wales, University Aberystwyth. Economics, The Agricultural of Economics and of Forestry Commission's Plant Trees, Farmers First Motivating the (1990) to presented at paper H. Fcarn, University. September 1990, York Conference,, Research Economics land Inflation, the (1980) M. and corporatestock,AmericanJournal prices of choice and portfolio Feldstein, 62: 910-916. Economics, Agricultural of Commission, Figures, Forestry Edinburgh. Forestry Facts (1979) Commission and Forestry Forestry Commission, Facts Figures, Forestry Edinburgh. (1985a) Commission and Forestry Edinburgh. Accounts, 65th Annual Report (1985b) Commission and Forestry for broadleaved Policy The Paper Forestry (1985c) 5, Commission, policy woodlands, Commission Forestry Edinburgh. Census 1979-82, Bulletin 63, (1987) ForestryCommission, trees and Commission of woodlands Forestry Edinburgh. Forestry Figures, Commission, Facts Forestry Edinburgh. (1988a) Commission and Forestry (leaflet), Woodland Grant Scheme Forestry (1988b) Commission, Edinburgh. commission Forestry Commission, Forestry Facts Figures, Forestry Edinburgh. (1989) and Commission Forestry Forestry Facts Figures, Commission, Forestry (1990) Edinburgh. Commission and Forestry

6.72

Forestry Commission (1991) Community Woodland Supplement(leaflet WGS 12/91), Forestry Commission, Edinburgh. Forestry Commission (1993) Forestry Facts and Figures, Forestry Commission, Edinburgh. Forestry Commission (1994a) Forestry Facts and Figures, Forestry Commission, Edinburgh. Forestry Commission (1994b) Forestry Review: Highlightsfrom the Forestry CommissionAnnual Report 1993194,Forestry Commission, Edinburgh. Forestry Industry Committee of Great Britain (FICGB) (1992) The Forestry Industry Year-Book 1991-92, FICGB, London. Garrod, G.D. and Willis, K. G. (1992a) Valuing goods' characteristics:an application of the hedonic price 34: Management, 59-76. Journal Environmental attributes. of to method environmental Garrod, G.D. and Willis, K. G. (1992b) Ile environmental impact of woodland: a two-stage hedonic price 24: in Applied 715-728. forestry Britain. Economics, the of value amenity model of Garrod, G.D. and Willis, K. G. (1992c) The amenity value of woodland in Great Britain: a comparison of 2: Resource Economics, 1-20. Environmental and economic estimates. Goodin, R.E. (1982) Discounting discounting, Journal of Public Policy, 2:53-72. Goodwin, S. (1995) It's hard to get into the woods today, The Independent, 15th November 1995:7. Hanley, N. D. and Ruffell, R.J. (1992) The valuation of forest characteristics,Working Paper 849, Institute for Economic Research,QueensUniversity, Kingston, Canada. Harriman, R. and Morrison, B.R.S. (1982) Ecology of streamsdraining forested and non-forested 88: 251-63. Ilydrobiologia, Scotland in to acid precipitation, subject central of area an catchments Harrison, A. and Tranter, R.B. (1989) The changing financial structure of farming, CAS Report 13, Centre for Agricultural Strategy, University of Reading. Timber Costings, Guide British Prices Forestry A Woodlands: Private to (1987) published and E. C. Hart, by the author, Coleford, Dean, Gloucestershire. Hart, C.E. (1991) Practical Forestryfor the Agent and Surveyor, 3rd ed., Alan Sutton, Bath. Woodland, Arboricultural Association, Romsey. Trees Valuation Amenity (1990) R. and D. of Helliwell, discounting: (1993) Intergenerational I. J. Bateman, N. public choice and empirical evidence Henderson, and for Social Centre Working Paper 93-02, GEC CSERGE discount and Economic for hyperbolic rates, College University London. East Anglia University Environment, Global and of the Research on for hyperbolic social discount Empirical (1995) I. J. Bateman, N. public choice evidence and Henderson, and Resource Economics, in intergenerational discounting, Environmental for implications and the rates and press. CostlBenefit Study, HMSO, An Interdepartmental London. Great in Britain: Forestry (1972) H. M. Treasury Guidefor Government: A Technical Government Central in Appraisal Economic (1991) H. M. Treasury Departments, HMSO, London. Commission Bulletin 70, Forestry Commission, forestry, Forestry Air (1987) and L. pollution J. Innes, Edinburgh. Forest Management, Department Measures Land C. (1990) Fiscal D. Nicholls, and A. of J. Johnson, and Economy, University of Cambridge. intervention forest impact C. (1991) The D. Nicholls, of government on private A. J. Johnson, and 30, Forestry Commission, Paper Edinburgh. Occasional Wales, in England and management in depopulation Snowdonia National (1987) Afforestation, C. the Price, and employment A. J. Johnson, and 87(3): Studies, 195-205. Rural Journal Park, of Planning, Faber. (1967) Forest R. T. Bradley, AJ. Grayson, R., and Johnston, D. Umed Economic benefits Valuing the contingent valuation method, (1990) using B. environmental Kristrom, Studies No.219, University of Umek Sweden. in United Kingdom in forestry taxation the Recent and and their grants planting (1989) changes Kula, E. in C. Dawson, P. J. (eds) Whitby, M. Usefor Land and profitability, sector impact on private 2OLh Symposium European the Development, Rural the of Forestry of proceedings and Agriculture, Association of Agricultural Economics, Department of Agricultural Economics and Food Marketing. University of Newcastle upon Tyne. in Forestry in the Northern Region, report to the Employment C. (1986) M. Whitby, H. and Laxton, Countryside Commission, University of Newcastle upon Tyne. Sunday, The Independent 16th June 4-7. 1996: have Where the on gone?. woods all (1996) G. Lean, Observer Money, The Magazine, Forests 14th February 34-41. 0. (1988) 1988: of Rosic, G. Lean, and

6.73

Leigh, J.A. and Randall, A. G. (1981) Timber Trade Practice, 41h ed., Macmillan, London. Lind, R.C. (1982a) A primer on the major issuesrelating to the discount rate for evaluating national energy options, in Lind, R.C. (ed.) Discounting for Time and Risk in Energy Policy, Johns Hopkins, Baltimore. Lind, R.C. (1982b) Introduction in Lind, R.C. (ed.) Discounting for Time and Risk in Energy Policy, Johns Hopkins University Press, Baltimore. Little, I. M. D. and Mirlees, J.A. (1974) Project Appraisal and Planning for Developing Countries, Heinemann. Lloyd, T. (1993) Testing a present value model of agricultural land prices, paper presentedat the Agricultural Economics Society 1993 Conference,Department of Economics, University of Nottingham. Loewenstein, G. and Thaler, R. (1989) Intertemporal choice, Journal of Economic Perspectives,3: 181-193. Logue, A. W. (1988) Researchon self control: an integrating framework, Behavioural and Brain Sciences, 11:665-710. Lynch, T. D. (1989) The taxation of woodlands in the United Kingdom, Green and Son, Edinburgh. MacNeill, J. (1990) Sustainabledevelopment: meeting the growth imperative for the 21st century, in Angell, DJ. R., Comer, J.D. and Wilkinson, M. L. N. (eds) Sustaining Earth: Responseto the Environmental Threats, Macmillan, London. Mazur. J.E. (1987) An adjusting procedure for studying delayed reinforcement, in Mazur, J.E., Nevin, J.A. (eds) Quantitative Analysis Behaviour, Volume H. V. The Effect of Delay and of Rachlin, of and intervening Events on Reinforcement Value, Erlbaum, Hillsdale, NJ. Melichar, E. (1979) Capital gains versus current income in the farming sector, American Journal of Agricultural Economics, 61: 1085-1092. Menzel, E.W. (1971) Group behaviour in young chimpanzees:responsivenessto cumulative novel changes in a large outdoor enclosure,Journal of Comparative and Physiological Psychology, 74:46-51. Ministry of Agriculture, Fisheries and Food (MAFF) (1987) Farm Woodland Scheme,A consultation document by the Agriculture Departmentsand the Forestry Commission, London and Edinburgh. ministry of Agriculture, Fisheries and Food (MAFF) (1992a) Increasedincentives for farm woodlands: joint UK Departments by Agriculture the of and the Forestry Commission, News Release announcement 36192,31st January 1992, MAFF, London. Ministry of Agriculture, Fisheries and Food (MAFF) (1992b) The Farm Woodland Premium Scheme:Rules MAFF, London. Procedures, and Ministry of Agriculture, Fisheries and Food (MAFF) (1992c) Farm Woodlands: A Practical Guide, MAFF, London. MINITAB (1994) MINITAB ReferenceManual, Minitab Inc., Rosemont, P.A. Mischel, W. and Baker, N. (1975) Cognitive transformationsof reward objects through instructions, Journal 31: Social Psychology, 254-261. Personality and of Mischel, W.. Shode. Y. and Rodriguez (1989) Delay of gratification in children, Science,244:933-938. Midin, D.C. (1987) Price-size curves for conifers, Forestry CommissionBulletin 68, HMSO, London. Myers, N. (1990) Tropical forests, in Leggett, J. (ed) Global Warming: The GreenpeaceReport, Oxford university Press, Oxford. National Audit Office (NAO) (1986) Review of the Forestry Commission's Objectives and Achievements, Auditor General, London. HMSO, Comptroller by the and report Nature Conservancy Council (NCC) (1984) Nature Conservation in Great Britain, NCC, Shrewsbury. Newton, L and Moss, D. (1981) Factors affecting the breeding of Sparrowhawks and the occurrence of their in Last, F. in T., Jeffers, J. N. R., Buncc, R. G. H., Claridge, C.I., quoted woodlands, prey songbird Baldwin, M. D. and Cameron, RJ. Whither Forestry? The scene in AD 2025, in Jenkinson, D. (ed) (1986) Trees and Wildlife in the Scottish Uplands, Institute of Terrestrial Ecology and Natural Environment ResearchCouncil. Nisbet, T. R. (1990) Forests and surface water acidification, Forestry Commission Bulletin 86, Forestry Commission, Edinburgh. Nordhaus. W. D. (1991a) The cost of slowing climate change; a survey, Energy Journal, 12(l): 37-65. A (1991b) D. W. sketch of the economics of the greenhouseeffect, in: American Economic Nordhaus, Review, Papers and Proceedfngs, 81(2): 146-150. To (1991c) D. W. slow or not to slow: the economics of the greenhouseeffect, Economic Journal, Nordhaus,

6.74

920-937. 101(407): Oosterhuis,F.H. and van der Linden,J.W. (1987) Benefitsof PreventingDamageto Dutch Forests:An Applicationof the ContingentValuationMethod,paperpresentedto the conferenceEnvironmental The Netherlands. Policy in a MarketEconomy,September1987,Wageningen, Analysis,2nd ed., Macmillan,Basingstoke. Pearce,D.W. (1986) Cost-Benefit Pearce,D.W. (1991) Assessingthe returnsto the economyand to societyfrom investmentsin forestry,in ForestryExpansion:A Studyof Technical,Economicand EcologicalFactors,ForestryCommission occasionalPaper14,ForestryCommission,Edinburgh. London. Pearce,D.W. (1993)Blueprint3, Earthscan, ) An analysisof real pricesof UK timber,mimeo,University Pearce,D.W. and Markandya,A. (unpub. CollegeLondon. Pearce,D.W. and Nash.C.A. (1981) TheSocialAppraisalof Projects:A Textin Cost-BenefitAnalysis, Macmillan,Basingstoke. Pearce,D.W. and Turner,R.K. (1990) TheEconomicsof Natural Resourcesand the Environment,Harvester Wheatsheaf,HemelHempstead. Pearce,D.W. and Ulph, D. (1995)A socialdiscountratefor the United Kingdom,CSERGEGEC Working Paper 95-01,Centrefor SocialandEconomicResearchon the GlobalEnvironment,Universityof East Anglia and UniversityCollegeLondon. Pearce,D.W. and Warford,J.J. (1993)World WithoutEnd, Oxford UniversityPress,Oxford. Pearce,D.W., Markandya,A. and Barbier,E.B. (1989) Blueprintfor a GreenEconomy,Earthscan,London. on forestinvestment,in Grayson,A.J. (ed) Philip, M.S. (1976)The impactof policiesand fiscal measures Evaluationof the Contributionof Forestryto EconomicDevelopment,Bulletin 56, Forestry Commission,Edinburgh. Pigou, A.C. (1932) TheEconomicsof Wetfare,4th ed., Macmillan,London. Pindyck,R.S. and Rubinfeld,D.L. (1981) EconometricModelsand EconomicForecasts,2nd ed., McGrawHill, London. in for forestry framework developing The the the United Kingdom: (1987) evaluation economic of C. price, 38: 497-500. Economics, Agricultural Journal of a comment, price, C. (1987) TheTheoryand Applicationof ForestEconomics,Blackwell,Oxford. School Agricultural Decision Making, Valuation Landscape 1) (199 of C. and ForestSciences, and price, University Collegeof North Wales,Bangor. price, C. (1993) Time,Discountingand Value,Blackwell,Oxford. Journal Agricultural Price (1982) 1. Dale, economically affordable and area, of C. predictions Price, and Economics,33(l). Choice, New York. W. H. Freeman, Decision Judgement, (1989) H. and Rachlin, 3rd W. H. Freeman, New Modern Behaviourism, York. Introduction (1991) to ed., H. Rachlin, Journal Experimental Commitment, (1972) L. the Green, self-control, of choice, and H. Rachlin, and Analysisof Behaviour,17:15-22. Landscape, Dent, London. in British Woodland Trees the (1976) 0. and Rackham, Journal, 38: Economic 543-559. A theory (1928) F. P. of saving, mathematical Ramsey, Oxford. University Press, Oxford Justice, A Theory (1972) J. of Rawls, (1989) F. Wasting C. Rossini, Assets: Natural Well, M., Beer, Resources in W., Magath, and R., Repetto, Washington, C. Institute, D. Resources Accounts, World Income National the in (1987) Forestry Flows Caithness (RSPB) Sutherland, Birds Protection the for of of and Society the Royal ConservationTopic PaperNo.17, RSPB,Sandy,Bedfordshire. (1988) Evidence Birds (RSPB) Protection for to the 11ouseof CommonsSelect the of Society Royal Committeeon Agriculture,Inquiry into the ForestryIndustryand Land Use,RSPB,Sandy, Bedfordshire. E. Wright, T. R. P. S. (1984) Soils Thompson, Their Use Lea, J. W., in R., Hartnup, and C., C. and Rudeforth, Wales,Soil Surveyof EnglandandWalesBulletin No.11, Harpenden. Cambridge University Earth, Press, Cambridge. Economy The (1988) the M. of Sagoff, discount for benefit-cost Lind, R. C. rateS social Approaches to the (1982) choice of analysis, in A. Sen, Policy, in Energy JohnsHopkins,Baltimore. for Time Risk Discounting ), and (ed. for Forestry Commission (1992) Price-size A. Research Whiteman, conifers, curve J. Sinclair, and 226, FC, Edinburgh. Note Information

6.75

Smith, M. (1996) Where have all the woods gone?, The Independent Weekend,25th May 1996:7. Solow, R. (1974) The economics of resourcesor the resourcesof economics, American Economic Review, 64(2): 1-14. Solow, R. (1992) An Almost Practical Step Toward Sustainability, invited lecture on the occasion of the Fortieth Anniversary of Resourcesfor the Future, Washington, D.C. Squire, L. and van der Tak, H. (1975) Economic Analysis of Projects, Johns Hopkins University Press, Baltimore. Stem, N. (1977) The marginal valuation of income, in Artis, M. J. and Nobay, A. R. (eds) Studies in Modern Economic Analysis, Blackwell, Oxford. Tanzi, V. (1980) Inflationary expectations,economic activity, taxes and interest rates, American Economic Review, 70: 12-21. The Times (1988) Lords speechfor Prince on tax forests, The Times, Saturday 13th February 1988, London. Thompson, J. (1990) Forest employment survey 1988-89, Occasional Paper 27, Forestry Commission, Edinburgh. Timber Trade Federation (1987) Year Book of Timber Statistics, Timber Trade Federation, London. Turner, R.K. and Pearce,D.W. (1993) Sustainableeconomic development: economic and ethical principles, in Barbier, E.B. (ed.) Economics and Ecology: New Frontiers and SustainableDevelopment. Chapman and Hall, London. Turner, R.K., Pearce,D.W. and Bateman, I.J. (1994) Environmental Economics: An Elementary Introduction, Harvester Wheatsheaf,Hemel Hempsteadand the Johns Hopkins University Press,Baltimore. United Kingdom Parliament (1988) House of Commons Official Report: Written Answers 23 March, 1988, Weekly Ilansard, Issue 1443: Cols 137-139. Walsh, R.G., Bjonback, R.D., Aiken, R.A. and Rosenthal, D.H. (1990) Estimating the public benefits of protecting forest quality, Journal of Environmental Management,30: 175-189. Welsh Office (1989a) Cambrian Mountains Environmentally SensitiveArea: Guidelinesfor Farmers, Agricultural Department, Welsh Office, Llandrindod Wells. Welsh Office (1989b) Cambrian Mountains Environmentally SensitiveArea (leaflet), Agricultural Department, Welsh Office, Llandrindod Wells. Welsh Office (1992a) Lleyn Peninsula Environmentally SensitiveArea: Guidelinesfor Farmers, Agricultural Department, Welsh Office, Aberystwyth. Welsh Office (1992b) Lleyn Peninsula Environmentally SensitiveArea, Agricultural Department, Welsh office, Aberystwyth. Whiteman, A. (1990) Price-size curves for conifers, Forestry CommissionResearchInformation Note 192, FC, Edinburgh. Whiteman, A. (1995) The supply and demand for timber, recreation and community forest outputs in Great Britain, PhD Thesis, University of Edinburgh. Whiteman, A., Insley, H. and Watt, G. (1991) Price-size curves for broadleaves,Occasional Paper 32. FC, Edinburgh. Wibe, S. (1992) Sweden, in Wibe, S. and Jones,T. (eds) Forestry: Market and Intervention Failures, Earthscan, London. Willis, K. G. and Benson, J.F. (1989) Values of User-Benefitsof Forest Recreation: Some Further Site Surveys, report to the Forestry Commission, Department of Town and Country Planning, University of Newcastle upon Tyne. World Commission on Environment and Development (WCED) (1987) Our Common Future. Oxford University Press, Oxford. World ResourcesInstitute (1994) World Resources1994-95, Oxford University Press,Oxford.

6.76

Chapter 7: Modelling and Mapping Timber Yield and its Value 7.1

INTRODUCTION

'In this chapter we present various models of the production of timber for the two Sitka consideration: spruceand beech. In Section 7.2 we presenta brief review speciesunder of previous studies. These have exclusively been basedupon relatively small scale surveys have furthermore, they tree also generally been confined to comparatively small growth, of areas and often to one topographic region, e.g. upland areas. Our study differs from these it GIS large in to that a utilise scale existing databasescovering a very uses previous models large and diverse study area; the whole of Wales. Section 7.3 presentsdetails regarding the discusses how in datasets this these data were transformed for the and study used various Results from analysis. regression our models of Sitka spruce and subsequent of purposes beech growth rates are presented in Sections 7.4 and 7.5 respectively while Section 7.6 images GIS Finally Section 7.7 map of predicted created yield analyses class. and presents findings of the previous chapterto produce monetised,equivalents of theseresults. the applies

7.2

LITERATURE

7.2.1

Literature Review

REVIEW AND METHODOLOGICAL

OVERVIEW

Clearly tree growth rates will depend upon a variety of species, environmental and field in Early simple factors. on thumb this rules relied of work reliant upon silvicultural factors. 1974) (Busby, data Reviews of single analyses or little supporting across relatively the specification (YC) regarding of clues a of literature number yield a class provide this impact interest focus of elevation (Malcolm, the upon productivity An was of early model. Subsequent 1974). Blyth, papers considered the various routes by 1973; Mayhead, 1970; (Grace, 1977), including YC windiness slope affected and aspect elevation which factors impact of such as soil type, soil (Tranquillini, 1979). Other work examined the Blyth Macleod, 1970; and 1981; (Page, droughtiness Jarvis and and moisture transport 1981). However, Savill, (Kilpatrick and the estimation of Mullins, 1987) and crop age likely is full variables explanatory of the a range relatively across recent models statistical find no examples concerning the innovation. Amongst such investigations we could

7.1

believe beech the model presented subsequently to be the first such and of productivity investigation of this species. However, there has beenmore attention paid to the other species has been both by Sitka Richard Worrell separately spruce, which analysed analysis; under (then of the University of Edinburgh) and Douglas Macmillan (Macauley Land Use Research Institute, MLURI)1While there had beena numberof earlier considerationsof factors affecting the growth Studholme, Malcolm 1972; 1970; Mayhead, 1973; (Malcolm, Blyth, Sitka and spruce of 1974; Busby, 1974; Gale and Anderson, 1984), the work of Worrell (1987a,b) and Worrell being first is b) (1990a, the to adopt a multiple regression approach Malcolm as notable and These (including highly range of explanatory variables. elevation were: extensive across a bottom for hilltop dummy and valley sites); windiness; temperature; variables separate full dummies. However, (measured a range and of soil and cosine); as sine while this aspect Worrell's for exercise, our own modelling results are not transferable pointers gives us vital due is Scottish location This Worrell's to the Welsh partly upland of study. case to our focus his Worrell interested but the of of study. as a result was mainly primarily experiment in detecting the influence of elevation upon YC in upland areas2.To this end he selected 18 had took steep slopes, relatively and all of measurementsalong which sites', sample principal locating from By 50 600 to samples at sites site. ranging each at m transect m a vertical be level tendency relationship central with elevation strong, a very could above sea is only applicable to similar, steeply sloping sites However, a model such established. found Scotland), is those the within and of not generalisable subset only (strictly speaking, in found Wales. the an area conditions size of environmental of to the plethora

A similar, thoughless extreme,considerationpreventsus applying the findings of is Here (1991). the study geogaphically confined, this time to lowland again Macmillan Scotland, although the 121 sites used are not selected to emphasisethe influence of any lowland therefore somewhat more generalisable are and within variable particular explanatory be in this cases, adequate, many to However, with would, respect our study area while areas. based Wales lowland is data that a model means of purely upon variability the topographic Macmillan is interesting Nevertheless, for anotherreason for the paper insufficient our needs.

11am grateful to both Richardand Douglasfor extensivediscussionsof their work.

'An important question given that this is the location of much of the existing stock of Sitka spruce. 3'rhe number of individual tree measurementsis not reported.

7.2

in that it comprises multiple regressionwith a prior principal componentsanalysis (PCA) of final degreeof explanation of R2 = 36.8%4. a reporting explanatory variables, A short note regarding model fit is justified here. As discussedin the previous chapter YC is the average annual growth rate of a plantation is therefore

d in given

ha7l yf 1. However,

assessed over an optimal

YC values are rounded

rotation.

YC

to the nearest even

6 have YC that stands with we or 8 we do not have sites with YC 7. While while number so this does not invalidate measurement explanation

statistical

analysis, as YC is the dependent variable, this approach to

does induce variance difficult

to attain.

into the dataset and therefore

makes high degrees of

As such the absolute value of fit statistics such as R2 should

be treated with some caution and instead we should consider relative degrees of fit compared to those attained in other studies.

7.2.2: Overview of modelling approach These prior studies provide very useful indications regarding the likely explanatory in be The differences in our analysis. considered should which modelling approach variables decided investigate both interest PCA to consequently and we a and standard of also are in However, other respects the methods of Worrell and methodology. multiple regression Macmillan were not appropriateto the specific types of question asked in our research. Our identify Wales is be the entire surface to areas over of which might aim suitable for central into forestry. This development the and necessitated agriculture of out of a conversion for both lowland capable of producing estimates was upland and which areas methodology findings had the of extrapolating such capability across the entire surface area of and which GIS/Sub-Compartment Database (SCDB) To this adopt a end we approach to the country. from base data Forestry YC Commission (FQ SCDB This the takes our modellingý. information holds detail in on each discrete stand (sub(described subsequently) which 6. both upland and lowland sites, results from As FC's in this the covers estate compartment) described is Use SCDB than those has generalisable previously. more the of model such a

"Although not specified this appearsto be an unadjustedR2 StadSdC.

in interest GIS been has (Moxey, the to 1996) application of agricultural modelling there this recent -'While is the first GIS basedapplicationto timber productionutilising multiple data sourcesand variables. An data is Mapper by Landsat Thematic Ccmmell (1995). presented using alternativeapproach 6Weare greatly obliged to Adrian Whiteman,Chris Quine and the ForestryCommissionfor useof the SCDB.

7.3

the added bonus of massively increasing our sample size relative to previous studies. However, rather than relate YC to the environmental variables reported in the SCDB, we 7 database, LandlS (described from these subsequently),which has complete a separate extract SCDB has data for forest (unlike the which only areas). Our regression national coverage Wales, including be to those not all other areas then extrapolated of readily can results disadvantage forestry. The is that, unlike the one of such an approach presently under is directly by data but here by the the not collected researcher many others, previous studies, be While this can viewed as not entirely negative, subsequent period. extended often over an be had for induced by indicated taken to that allowances variance such an approach modelling to measurement.

7.3

DATA AND DATA MANIPULATION

This researchrelies upon a diversity of data sources. Two already mentioned are the Forestry Commission's Sub-CompartmentDatabase(SCDB) and the Soil Survey and Land Research Centre's LandIS database. In addition to this, further environmental and In from describe data this these of sources. section a variety we obtained was topographic data and how they were manipulated prior to consideration within the subsequentstatistical investigation of tree growth. It is important to rememberthat, while the SCDB holds detailed it does individual sites, not extend to the majority of Wales which plantation data regarding is unplanted. Therefore the environmental variables given in the SCDB are, for our purposes, land for YC these surface coverages as complete of variables are not predictions unsuitable for be to extrapolation of predictions therefore cannot used presently unplanted and available held in LandIS data described the The of and other variables coverages complete areas. for this to extrapolation therefore allow of regressionresults. needed are subsequently SCDB DATA

The SCDB is the Commission'scentralforestinventorydetailingobservationsfor all invaluable it As high dita. Some Estate. in source provides an such of quality the stands interest investigation internal to and was not of administration our the and so concerns this of

7We are greatly obliged to Arthur Thomasson,Ian Bradley and the Soil Survey and Land ResearchCentre IS. Land for of (Cranfield) use 8The FC were, as always, most willing to allow accessto their data, for which we are most grateful.

7.4

final list of variables extracted for this study was as detailed in Table 7.1. This also shows how certain of this data was manipulated to produce further (often dummy) variables. In doing this, one-way analysesof varianceon the dependentvariable (YQ were usedto identify likely significant divisions in the data. The SCDB alsocontainsa variety of sub-compartmentspecific environmentalvariables hazard Normally be type terrain these and windblow class. type, would altitude, soil such as ideal for modelling purposes. However, as the FC only holds such data for those grid squares in which it has plantations, and since these are not (with the exception of altitude) variables for which uninterrupted national coveragesexist, findings based upon such data would not form a suitable basis for extrapolation to other, currently unforestedareas. This is somewhat is data than that almost certainly obtainable more accurate this specific site as unfortunate from more general databasessuch as LandIS. Ibis means that the regression models data YC factor fit LandIS those the the as well as using site not will produced using information given in the SCDB. However, for the purposesof this research,the advantage Wales the entire across surface of and consider currently to out being extrapolate able of (which basis the subsequently argue, costs we on of such outweighs easily areas unplanted be likely to small). our results, are In all records for some6082 Sitka spruceand 766 beechsub-compartmentswere used increase This a significant over sample sizesused very represents in our regression analysis9. distributed These literature. in throughout were upland and observations the previously for basis Wales extrapolation of results to other, presently a good lowland providing unforested areas.

9Appendix 5 details descriptive statistics for variables used in our best fitting Sitka spruce and beech YC locations illustrates Sitka Chapter 10 of spruce sub-compartments detailed superimposed subsequently. as models 10.9). (figure map upon an elevation

7.5

Table 7.1:

Variables obtained from the SCDB (except where shown otherwise). Ordered database. the as per Values

Variable name

Notes and recodings (in italics)

Grid reference

Easting Northing

100 to resolution OS grid references

Land use/crop type

PIIF plantation high forest PVA3 - uncleared windblown area PRP research plantation

unckared -I if PWB =0 otherwise research =I if PRP -0 otherwise

Storey

Ia single Storey 2- lower storey 3- upper storey

singte -I -0

Species

SS - Silk& spruce BE - beech

Used to identify target species

Planting year

Discrete variable

plantyr. year in which stand was planted

Survey year

Discrete variable

survyr: year in which stand was surveyed'

Yield class

Even number

YC: tree growth rate: averageWlhalyear over an optimal rotation - the dependent variable

Productive forest am

Ha

Area: stocked area of the sub-compartment

Unproductive forest

Ha

Unprod'. the am within the sub-compartment which has a permanent affect upon the crop. e.g. rocky outcrops, etc.

area Rotation

I= Ist rotation on formerly nonland 2,3 etc. - 2nd, 3rd rotatiom etc. 9- historical woodland sites S- ancient, semi-natural woodland

forest

if single storey otherwise

Is# Rot -I for I at rotation; =0 otherwise 2nd Rot -I for 2nd rotation; a0 otherwise (Note for BE this Includes some subsequentrotations.) Historic I if historic site; .0 otherwise Seffi-nat I if ancicnt/serni-natural woodland a0 otherwise Mixed -I

if mixed crop; w0 otherwise

Mixture

P- single species crop M mixed species crop

Legal status

P L E

purchased by FC ]eased extra land, managed by FC outside legal boundary

Purchased aI if purchased; -0 otherwise Leased aI if leased; a0 otherwise Extra -I if extra; -0 otherwise

Landscape

I 2 3

National Park AONB/National Scenic Area ESA (where not included in I or 2 above)

NatPark =I if National Park; w0 otherwise AONBINSA aI if AONB/National Scenic Am& w0 otherwise OthESA =I If ESA area not included In above 0 otherwise

Forest Park

I

Forest Park

FPark aI

Conservation

SSSI (Site of Special Scicnfi fic Interest) I 2a NNR (National Nature Reserve) 3z Non-FC Nature Reserve

FC Conservation

Ancient monument/

S= scheduled ancient monument U= unscheduled ancient monument Wa ancient woodland

woodland

SSSI -I if SSSL n0 otherwise NNR -I if NNR, -0 otherwise NonFCNR I if Non PC nature reserve; 0 otherwise FCNR -I if Forest Nature Reserve; =0 otherwise FCcons -I if other FQ, -0 otherwise

Forest Nature Reserve Other FC conservation

I. 2-

if forest park; a0 otherwise

Ancient aI If S. U or W, a0 otherwise Momanent aI it S or U, ar 0 otherwise

Further mcodes from above: NpAonbSa -I If any of Nat Park or AONB/NSA a0 otherwise Com I if any of NNF, NonFCNR. FCNR, mons 0 otherwise Reserve aI if any of Cons, AONB/NSA. OdT_SA a0 otherwise Park -I if any of Nat Park, F Park. SSSI -0 otherwise

Research Station, FC Northern Roslin, Quine to Chris the by at whom we are very grateful. Supplied 1. Note:

7.6

7.3.2: LANDIS DATA 7.3.2.1: Background The first systematic attempt to analyze and record British soil information was the

"county series"of mapsinitiated by the Boardof Agriculturein the late 18thand early 19th source centuries. Until comparativelyrecentlythis remainedthe standardand unsurpassed Soil Survey England Wales 1940s (SSEW) began During data. the the of and a of soil detailedmappinginitiative. However,by the late 1970s,only one fifth of the country had beencovered. In 1979the SSEW,which in the late 1980'sbecomethe Soil SurveyandLand ResearchCentre(SSLRC),commenced a five-yearprojectto producea soil mapof the,whole describe distribution land in Wales to England soil and related quality and appropriate and of detail. The data collected in this exercisewas digitised, spatially referenced,and subsequently infort-nation (Bradley Knox, 1995). include and other environmental and to climate expanded database initially (LandIS) information land commissioned by the was The resulting system Ministry of Agriculture, Fisheries and Food, with the stated aim of "providing a systematic inventory capable of being used or interpreted for'a wide range of purposes including facets land for but the many of use planning and national also work, advisory agricultural However, 1984; (Rudeforth the added). although emphasis maps and et al., use" resource in 1984 there has never beenany major attempt since bulletins completed were accompanying The into incorporate them presentresearchtherefore representsone making. policy then to intended its land for LandIS originally purpose: national first to use use attempts the of planning'o. 7.3.2.1: The data data derivations the extracted from LandIS are presented Definitions, and accuracyof 7.2. Further details in Table LandlS data These the 5.1. of and in Appendix are surnmarised discussion (1985) Welsh Thomasson in Jones with of and conditions given therein are given is km data LandIS (1984). supplied at a5 resolution. by Rudeforth et al.,

loAgrcement to use the data was obtained from Arthur Tbomassonin 1987. However, at the time the SSEW becoming 'downsizing'. is now Cranficld being and part trauma privatiscd, of of what the was undergoing Jones RIA. SSLRC Ian Bradley for the to of We and subsequently honouring this grateful are University. Sciences/UEA for funding Environmental School data the the of transfer entailed to costs. and commitment

7.7

Table 7.2: Variables obtained from LandIS Variable name

Label

Definition

Accumulated temperature

Acclemp

Average annual accumulatedtemperature (in OQ above (rC

Accumulated rainfall

Rainfall

Average annual accumulatedrainfall (in min)

Available water

Avwalgra

Amount of soil water available for a grass crop after allowing for gravity induced drainage

Avwalcer

As per Avwatgra but adjusted for a cereal crop

Avwalpot

As per Avwatgm but adjustedfor potatoes

Avwaisb

As per Avwatgra but adjusted for sugarbeet

Mdefgra

The difference between rainfall and the potential evapotranspirationof a grass crop

Mdefter

As per Mdefgra but adjusted for a cereal crop

Mdefsbpt

As per Mdefgra but adjusted for a sugarbect/potatoescrop

Field capacity

Fcapdays

Average annual number of days where the soil experiencesa zero moisture deficit

Return to field capacity

Relmed

Median measurefrom a distribution of the number of days between the date on which a soil mums to field capacity and 31st Decemberof that year

Retwel

The upper quartile of the above distribution (measureof return to field capacity in wet years)

Retdry

Ile lower quartile of the above distribution (measureof return to field capacity in dry years)

Endmed

Median measurefrom a distribution of the number of days between the 31st Decemberand the subsequentdate on which field capacity ends

Endwet

The upper quartile of the above distribution (measureof the end of field capacity in wet years)

Enddry

The lower quartile of the above distribution (measureof the end of field capacity in dry years)

Workability

Workabil

A categorical scale indicating the suitability of the land for heavy machinery work in spring and auturnn

Spring machinery working days

SprMWD

I'lie averagenumber of days between I st January and 30th April where land can be worked by machinery without soil damage

Autumn machinery working days

AuIMWD

Ile averagenumber of days between I st Septemberand 31st December when land can be worked by machinery without soil damage

Soil type

See Table 7.3

SSLRC soil type classification code

Moisture deficit

End of field capacity

An immediateproblemwith the LandISdatawasthe plethoraof differing soil codes. hundreds lists (1983) large SSEW from many types, of separate which soil a Theseare taken far dataset. level detail in Welsh This that of exceeds used our present were of which number (1987b) Worrell dummies derived type YC who uses seven soil as such in previous studies in SCDB in FC turn the the on relies standard which information classification given from 7.8

in large LandIS are a problem both becauseof their The of soil codes given number of soils. implication for degreesof freedom in our subsequentregression analysis and becauseany little forester familiar be to the of practical use with an alternative and such results would in field Furthermore, the consultations an with expert of soil science and simpler system. forestry suggestedthat, for our purposes,many of the SSLRC soil codes could be merged information increase loss in clarity". Details of the and a substantial of with no effective final categorisation are given in Table 7.3. Table 7.3: Soil type codes F

" ly pel

Lowland lithomorphic Brown earths Podzols Surface water gley Stagnogley (perched watertable) (3round water gley Peats Upland lithomorphic Urban

Variable label

Subsumed SSLRC soil 2 codes

soil 1 so!12 soil 3 soil 4 soil 5

361 514,541,551,561,571, 572 611,631 651,654,711,712,713, 721 813 1011,1013 311 n/a

soil 6 soil 7 soil 8 n1a

1. In our analysis of farm outputs (Chapter 9) we additionally assumethat the dominant soil type on a farm is an adequate farm. that on sofls aU of representation 2. Here we have only listed categorisationsdown to the subgrouplevel (as defined in Avery, 1980). LandIS further subdivides these into nwnerous soil associationsas detailed in SSEW (1983).

Subsequentstatistical analysis suggestedthat, if anything, merging of soil codescould

havebeentakenevenfurther and somecombinationsof the variablesgiven in Table7.3 are later. considered 7.3.3: OTHER DATA 12 hazard Topex and wind 7.3.3-1: Topex is a measureof the topographical shelter of a site. It is usually detennined as "Dr 13ill Corbett of the School of Environmental Sciences,UEA, and formerly of the SSEW. kindly advised in the merging of soil codes to produce a simple eight-category system which groups together similar soils. 121km referenced data on topex and wind hazard were kindly supplied by Chris Quine at the Forestry Station, Roslin, Research Northern to whom we are very grateful. Commission's

7.9

the sum of the angle of inclination for the eight major compasspoints of a site (Hart, 1991). Here then a low angle sum (low topex value) representsa high degree of exposure. The resultant variable was labelled Topex I Ian. Blakey-Smith et al. (1994) define wind hazard on the basis of four factors": i. ii. iii. iv.

delimited GB a zone on wind map; elevation - high values increasing wind hazard; topex - as defined above; soil type - those which relatively speaking promote growth (brown earths, podzols, etc.) being low wind hazard while those which restrict growth (gleys, peats, etc.) are higher wind hazard. The resultant continuous variable (Wind I kn?) is inversely linked with tree

productivity and growth rates. 7.3.3.1: Elevation and associated variables The work of Worrell and Malcolm (1990a) shows that elevation and its associated However, YC. key is included in the such of a variable predictors are not characteristics LandlS databaseand the SCDB only gives heights for existing plantation sites. Clearly for is inadequate data this so an alternative and source of purposes was required. extrapolation digital GIS form in elevation model (DEM)14 The DEM is a the This was provided of a . Wales. This image digital topography of the was created from three principal GIS-based of

data sources: L

The Bartholomew 1:250,000 databasefor the UK. This gives 50 rn contours up to 1000 m, after which 100 rn intervals are reported.

ii.

These Bartholomew's from maps. heights paper Spot were panicularly useful for DEM for the and of addressing the, problems accuracy the predictive assessing

funnelling tends variable have discuss which to (1994) a higher values in valley also 1313jakey-Smith et al. bottoms. Julii Brainard by Andrew for and this Lovett of the School of research 14TheDEM was custom-created I gratefulUEA, most am to Sciences, whom Environmental

7.10

iii.

identifying mountain tops. with associated Spot heights of plantationsfrom the SCDB. 'Mis provided additional information used in the interpolation of heights between contours.

After exhaustive accuracy testing of the resulting elevation variable ffselvgr2), the it DEM to provide two further GIS surface variables: slope angle the also used authors of (DsI2) and aspect angle ffsaspgr2).

Data on all these variables was supplied at a 500 m

resolution. 7.3.4: Creating GIS surfaces for explanatory variables Prior to the regressionanalysis two fundamentalissueshad to be addressedregarding for definition the environmental variables. common and resolution common extent of a the While the geo-referenceddata obtained from the LandlS and non-SCDB sources detailed inspection into GIS surfaces, of these showed that the various data above were converted in its both differed geographicalextent and spatial resolution. obtained Data were supplied at a wide array of resolutions ranging from the (nominal) 100 m 5 km interpolation LandlS While SCDB tiles the to the the of the variables. accuracy of facilities available within the GIS made conversion to a common scale relatively 1-5choice of that scale was a matter for some deliberation. While straightforward, is interpolation (100 GIS, the the m) given capabilities smallest unit of a upon standardisation it did The 100 in SCDB feasible, choice. the not seem a sensible m reference used perfectly is, the FC admit, spuriously precise. Furthermore, use of a larger scale would, in the caseof likely DEM to of predictions which an averaging out was avoid problems entail the say However, km 5 to the aggregation up point estimates. single scale of the with associated in loss, be likely interesting felt detail. data to to a of result valid and was coarsest interpolated data km to this resolution. and all were grid settled upon was Consequently aI

The spatial extent of Wales was defined by rasterisingon to aI border England. data for Ilis the and coast with Bartholomew's vector

km grid the

in resulted a GIS

"'This is somewhat misleading. In reality Careful interpolation is a highly time consuming exMise involving interpolation decay weights with actual versus predicted verification. of a of range iterative reassessment the improved have the time which such analyses take, they are still in considerably speed processor Whilst advances issue is This length in Bateman, Lovctt and Brainard addressed to at undertake. arduous somewhat (forthcoming).

7.11

land 20,563 cells which was used as a mask file to extract I krn values of surface consisting for each of the variables in the LandIS and non-SCDB datasetsdescribed above. However, in undertaking this exerciseit was found that, with the exception of the custom written DEM least all variables were missing at virtually some observations. and associated variables, Given our principal aim of extrapolating regressionresults across the whole of Wales, this situation had to be rectified. In some casesthe problem of missing data was relatively minor. With respectto the topex and wind hazard data, which was supplied in a1 krn rasterisedform, just 103 of the being located 20,563 these all of missing, at the tips of various peninsula. cells were required Here interpolation from surrounding cells provided a ready solution to this problem.16 The missing dataproblem was more seriousin the LandlS databaseboth becausemore data tiles were missing and becauseof their larger, 5 km, resolution. Using the OS grid, Wales extends to some 942 of these tiles". Only three of the variables described in Table 7.2 had data for all of these tiles. Table 7.4 lists omissions from this database. Table 7.4: Omissions from the LandIS database No. of 5 km land tileS2 supplied

% of all Welsh 5 km land tiles

Acctemp- Growseas, Grazseas

942

100.0

Rainfall, Retmed, Retwet, Retdry, Endmed, Endwet, Enddry, Fcapdays

898

95.3

Mdefgra, Mdefcer,

858

91.1

Avwatgra, Avwatcer, Avwatpot, Avwatsb

812

86.2

Workabil, SprMWD, AutMWD, Soils

780

82.8

Variable label'

1. FromTable7.2. be in England (sorne land Welsh OS 5 krn mainly may or in the sea). square containing any grid includes 2. This any

"This and subsequentinterpolation operations were conducted by Andrew Lovett, to whom I am very grateful. "Note that coastal and border files will not be fully filled. This accountsfor the implicit difference in the I kni to the mask. opposed as this coverage of extent

7.12

As before the majority of omissionswere Clusteredaround the Welsh coast. However, to allow our extrapolation analysis to proceed, such empty squares had to be filled. Inspection of nearby cells for which data was available showed strong spatial trends in all Consequently for type. the of soil empty cells all non-soil variables exception variables with These interpolation inappropriate for filled techniques. were clearly using soil type were interpretable from to and consequently abruptly was tended change not other data which points. Interrogation of the Bartholomew's digital databaseidentified 19 of the 162 5-krn grid being in had areas as urban soil values which surveys soil not been missing squares filled by SSEW 1: 250,000 The the missing were consulting values remaining undertaken. Soils of Wales paper map. All the LandlS data was the interpolated on to aI

km grid and our coast/border

fell delete outside this extent. to squares which outline used With all data now at a common resolution and extent we now had the necessary for in from variables of potential predictor use our regression model and surfaces complete be currently or across all areas, whether planted not, would possible. which extrapolation A final task concerned the extraction of values for all environmental variables for each

SCDB. Ilis in the YC observation 7.3.5: PRINCIPAL

was achieved via a GIS macro command's.

COMPONENTS ANALYSIS

As discussed in our literature review, two approacheshave been adopted for the

(1987a, Worrell b) data. While Worrell YC Malcolm and of and modelling statistical (1991) first Macmillan b) analysis, subjectsexplanatory (1990a, useconventionalregression before (PCA) factors the entering analysis components resultant to principal within variables a decided It be these two that a comparison of approaches was would analysis. a regression data PCA. the made subject of a our was interest so and of Discussionof the PCA approachis given in Johnston(1978), Norusis (1985) and factor is in fact PCA (1994). analysis (Lewis-Beck, 1994) and case of a special Dunteman

interchangeably 'factor" 'component' in following discussions. the terms and the we shall use in essencePCA attemptsto identify pattemsof covarianceso that trendswithin a LJEA, by Andrew Lovett. I to at 111'rhis whom am mostgrateful. written was custom

7.13

by factors, i. large of are summarised variables a smaller number of number e. comparatively it seeksto identify patterns of common variance. For example, in our literature review, we YC between and altitude was actually the product of a range the relation that negative noted A including temperature, interrelated slope, aspect, etc. general elevation, variables of 'height' factor which reflected these interrelations might therefore prove a strong predictor in identifies four PCA: (1985) Norusis steps conducted of tree growth.

1.

A correlation matrix is prepared so that variables which do not appear to be related identified be (suppression-type dataset be the can problems can also to others within identified at this stage). The appropriatenessof PCA can also be assessedat this point.

2.

The number of factors necessaryto adequately represent the dataset is identified. Clearly unless this is substantially less than the number of variables then the exercise

is of little value. interpretable. (rotated) be to them factors transformed make more The may

4.

Factor scoresare computed to indicate how individual observationsperform on each factor. These may then be used as predictors within a regression model. However, before we could start our PCA study we were concerned to first consider

Sitka for both beech be spruce and sites or not. appropriate might analysis single a whether Sitka into data to sites two were planted whether with spruce setsaccording By dividing our faced former more adverse environmental it conditions the that generally beech noted was or for divided details 7.5 Table certain environmental statistics variables summary to the latter. species"'. to site according

19Descripdvestatistics for the full range of environmental variables as used in our best fitting YC models in 5. Appendix detailed beech are for Sitka spruce and

7.14

Table 7.5:

Description of environmental variables for forestry sub-compartments by Sitka (SS spruce; BE = beech) species = Species

Mean

Median

St. dev.

Coef. of Variation

Wselvgr2

Ss BE

323.70 196.83

333 183

102.72 99.90

31.7 50.8

WindIkm2

SS BE

14.890 12.009

14.96 11.89

2.36 2.25

15.9 18.8

Acctemp

Ss BE

1401.2 1591.8

1385.0 1600.0

243.70 240.90

17.4 15.1

Rainfall

Ss BE

1713.6 1386.5

1705.0 1284.0

433.80 423.50

25.3 30.5

Fcapdays

SS BE

313.39 267.29

322.0 258.0

48.27 56.19

15.4 21.0

MdefGra

Ss BE

25.30 57.00

20.0 53.0

25.54 38.20

100.9 67.0

AutMWD

Ss 2E

2.122 16.623

0.0 0.0

9.66 24.23

455.2 145.8

Variable

----

Table 7.5 indicates that, on average,Sitka spruce sites are at higher elevation, colder, is This beech less than their counterparts. not surprising as we would workable and wetter lowland be hardy to to broadleaf confined relatively generally areas while plantations expect been have This Sitka a over wide variety planted of sites. spruce substantial species such as difference in site characteristics suggested that separate rather than common PCA be conducted. should variables investigations of explanatory 7.3.5.2: Defining input variables

in were While most of our environmentalvariables a form amenableto initial PCA, ffsaspgr2) this true not aspect was of our variable a which was within consideration for direction. is This PCA which simply focuseson in compass terms of unsuitable recorded be interpreted V 359' that and would as very different rather values of linear correlations so

7.15

than virtually identical. The solution adopted was to calculate both the sine and cosine of include Cosasp in (Sinasp these the PCA instead. The respectively) and and variables aspect interpreted be in linear transformations to two these terms. allows aspect of combination When an initial attempt was made to undertake PCA using the FACTOR command form 'ill data SPSS-X, the conditioned message of matrix' was encountered a warning of (though results were generated). Further investigation suggestedthat this situation might

reflect either: (e. coefficient of a small variation very g. <0.002%) with variables or ii.

high correlations between a number of the input variables.

Subsequentcalculations suggestedthat the former was unlikely to be a problem (see is be. ironic It latter but that while PCA searchesout 7.5) the almost Table that might well for relationships between variables, if some of these are extremely strong then calculation investigate Pearson To this possibility correlations matrices were problems can arise. beech datasets (see Appendix Sitka both for of environmental variables spruce and calculated 5.2). Inspection of theseresults identified five groupings of correlated variables as follows:

Group 1:

*Acctemp; Growseas; Grazseas

Group 2:

*Rainfall; RetWet; *RetMed; RetDry; *FcapDays;EndWet; *EndMed; EndDry

Group 3:

*MdefGra; MdefCer; MdefSbpt

Group 4:

*AvwatGra; AvwatCer; AvwatPot

Group 5:

AutMWD; *SprMWD

Within each of thesegroups, one or more (dependingupon the degree of correlation) into PCA (marked be * Choice 'input' the to entered above). chosen of variables were

biologicalplausibility of a relationshipwith YC, the,degreeof depended the upon variable data the that the and consequent requirement variables resultant matrix other with correlation All In ill-conditioned. be were these satisfied. to the conditions addition above not should input variableswerealsoidentifiedfor inclusionwithin thePCA less further correlated seven, Windlkm2, Cosasp, Topexlkm, Ds12, Sinasp). Wselvgr2, (Workabil, 7.16

This analysis resulted in a consistent list of predictor variables for both our Sitka AutMWD datasets beech the with single exception of and SprMWD, both of spruce and but it for beech. As important included be to use the spruce not considered was which could AutMWD for the each species, weaker variable was deleted from both same set of variables PCA studies.

7.3.5.3: PCA for Sitka spruce environmental variables i.

Examining the correlation matrLx The first task was to calculate the degree of sampling adequacyfor both individual

This individual be the to shows extent the sample. which entire variables can and variables factors describing by to the the variation of the extent which and other variables explained is by be With Kaiserdataset to the this the respect entire sample given created. can overall Mayer-Olkin (KMO) measure of sampling adequacy. KMO compares the magnitude of If to coefficients. partial correlation partial correlations are coefficients correlation observed be low between KMO high that suggesting correlations then will pairs of variables relatively Conversely by be when partial correlation coefficients are other variables. explained cannot low, KMO is high and communality is high. KMO ranges from 0 (totally inadequate)to I (perfectly adequate)with valuesbelow 0.5 indicating samplesfor which PCA is inappropriate. Calculating KMO for the Sitka spruce dataset gave a value of 0.76 which Kaiser (1974) describes as middling to meritorious. Sampling adequacy for individual variables was (ATC) matrix (see Appendix 5.2 inspection the correlation of anti-image through confirmed for details).

ii.

Component extraction Here linear combinations of the variables are formed. The first principal component

for largest in data. The be the the amount of that accounts variance factor) which (or will is for lesser first. We factor and amount of uncorrelated the a variation with accounts second

in factors but be defining to the the this of variables up number sample would can carry on Therefore to the need consider we amount exercise. of variationexplained pointless a rather draw devise determine to we the line with respect will factor some rule where by each and factors input to of can reduce which we our number variables. The most to the minimum factors is to and all standardise variables with a meanof zeroandvariance approach common 7.17

of one. This will mean that the total standardisedvariance of the samplewill be equal to the here input 15. The total amount of standardisedvariance explained by of variables, number any one factor (known as its eigenvalue) can then be compared to the total standardised

varianceof the sampleand the percentagevarianceexplainedcalculated. Factors which have eigenvaluesof less than I perforni less well than simple variables (which are constrained to have a standardisedvariance of 1) and so this is commonly used

factors below which are discarded. In this casethe first five factorsall as a cut-off point for 76.9% of the total variance in the sample. this account criteria and satisfy iii.

Improving interpretability: factor rotation Interpretation of the factors may be achievedby calculating a 'factor matrix' detailing

loading' between 'component each factor and each variable. the correlation coefficient or This is then 'rotated' using the 'varimax' method of Kaiser (1958) to minimise the number factor high loading interpretability having thereby the each on enhancing a of of variables for our rotated factor matrix. loadings details 7.6 Table factor. component each Table 7.6: Rotated factor matrix: Sitka spruce sub-compartments

Acctemp Rainfall RetMed EndMed FcapDays MdefGra AvwatGra Workabil SprMWD WseIvgr2 Dsl2 Topexlkm Windlkm2 Cosasp Sinasp

Factor 1

Factor 2

Factor 3

Factor 4

Factor 5

-0.34 0.92

0.15 0.20

-0.28 0.13

-0.59 0.08

-0.94 0.91 0.94

-0.14 0.20 0.15 0.11

-0.18 0.17 0.17

-0.09 0.13 0.06

-0.38 -0.06 0.05

-0.14 0.88 0.87 0.12 0.51 0.06 -0.07 0.36

-0.19 -0.04 -0.03 0.15 0.41 0.06 0.02 0.18

-0.09 0.16 .

-0.10 0.84

-0.77 0.16 0.19 -0.51 0.16 0.10 0.21 0.00 -0.03 0.07

-0.04 -0.10 0.27 -0.38 0.73 0.81 -0.78 0.13 0.05

7.18

-0.05 0.05 -0.22 -0.04 -0.06 -0.16 0.38 0.31 0.04 0.23 0.81 -0.22

Inspection of the PCA factors detailed in Table 7.6 indicated that they were relatively follows: interpret to as easy

FactorNo.

Label

1

. Soil wetness/rainfall

2

Steeperslopes/low windiness

3

Waterlogging/workability/high elevations

4

Cold/sine aspect

5

Cosine aspect/elevation

The 'communality' or proportion of variance in each input variable which is 'explained' by the five factors" was also calculated. This indicated that the only variable is (communality 0.39), is sprMWD explained poorly all other variables relatively = which having a reasonableproportion of variance explained by our five factors (mean communality 0.80). = Calculating factor scores

iv.

The factor scorecoefficient matrix was calculatedvia the regressionmethod described indicate (which Factor the position of each observation (here (1985)21 scores by Norusis . factors) in then were the rotated calculated the extracted, on normal sub-compartment) each Sitka both for beech factor matrices). 5.2 our spruce (Appendix and examples gives manner by factor scoresobtained this processcan then be entered directly into our The site specific YC regression model as the environmental explanatory variables.

7.3.5.4: PCA for beechenvironmental variables for the

beech identical sub-compartments the to PCA The was to that used procedure applied be in brief. presented will results Sitka spruce sites and so

Examining the correlation matrix The KMO measureof sampling adequacywas calculated to be 0.77, a figure similar loadings. factor 2&Fhecommunality is the sum of the squared 21ThiSis the default method in SPSS-X.

7.19

to that for Sitka spruce. Inspection of the AIC matrix for beech. Generally these are indicated that sampling adequacy for individual variables was generally as desired for a individual for Avwatgra and Workabil were rather lower PCA the values although successful than for Sitka spruce (see Appendix 5.2 for details).

ii.

Component extraction

As beforefive factorssatisfiedour criteria for extraction. improving interpretability: factor rotation A rotated factor matrix was calculated as before and is detailed in Table 7.7. Table 7.7: Rotated factor matrix: beech sub-compartments

FF--

Factor 2

Factor 3

Factor 4

Factor 5

-0.44 0.94

-0.60 0.03

0.06 0.02

-0.38 0.19

-0.01 0.03

-0.96 0.94 0.96

-0.09 0.08 0.10

-0.15 0.29 0.16

-0.85 -0.03 0.10

-0.35 0.02 0.01

-0.03 0.04 0.05 0.05 0.95 0.92

-0.14 0.03

-0.02 0.04 0.02 0.12 0.02

-0.74 0.16 0.19

-0.22 0.89 0.07

-0-07 0.14 0.16 0.77

-0.14 0.17 0.06 0.04

Factor I

Acctemp Rainfall RetMed EndMed FcapDays MdefGra AvwatGra Workabil SprMVVD Wselvgr2 DslI

0.51 0.11

TopexI km Windlkm2 Cosasp Sinasp

-0.15 -0.19

-0.06 0.04 -0.00

-0.03 0.04

-0.14 0.83 0.17 0.14

1

-0.13 -0.11

0.65 -0.42 0.39 0.24

follows: factors as We can interpret theserotated Label No. Factor I

Soil wetness/rainfall

2

High elevation/cold/windiness

3

Waterlogging(workability

4

Steep slopesAowwindiness Aspect 7.20

-0.05 -0.05 -0.65 0.77

Communality coefficients were calculated. These were relatively high for all input 0.60 having (mean communality = 0.81). under values variables, none iv.

Calculating factor scores Factor scoreswere calculated as discussedpreviously.

7.4

YIELD MODELS FOR SITKA SPRUCE

In this section we present details for the various regression models estimated for details YC. Further Sitka regarding the regressionmodels estimatedas spruce prediction of descriptions base data in the and correlation matrices of are given accompanying well as Appendix 5.3. Three types of model were fitted.

These varied according to whether the

by: described (i) data; (ii) factors for a site were raw of our characteristics environmental PCA; (iii) a mixture of these two (ensuring that raw variables retained in the model were not significantly

Clearly factors). these latter mixed models are invalid correlated with retained

if the site characteristic being described by a particular factor is also being explained by a raw data variable.

For example Factor

1, which represents (for our Sitka spruce sub-

be included the could not within and rainfall same model as the wetness soil compartment) However, Rainfall. to test whether some site characteristics might we wished raw variable be better described by factors while, within the same model, other uncorrelated characteristics initial Our dataset for Sitka by described be raw variables. spruce contained optimally could data key YC deleted for these was missing and so or other sites were which sites a number of is This far larger 6082 datas. initial than the sites. et of leave any of studies complete an to demonstrates literature in the one of principal and advantages review of our our considered based to analyses common database upon small site surveys. more large approach compared Our regressions analyses followed the approach set out by Lewis-Beck (1980) and identification initial An the (1982). concerned objective of an appropriate functional Achen indicated linear These better that a for model performed than models. marginally form our is both interpretable and typical of form forms that such a easily given and, other standard choice. a sensible this seemed studies, other Initial comparison across the factor only, variable only and mixed model types

7.21

suggested that their, was little difference in the degree of explanation afforded by these various approachesbut that the mixed model performed marginally better than the others and is reported as Model 7.1. Inspection of this model shows that the large sample size has permitted the identification of a large number of highly significant predictors many of which With to respect to the environmental characteristicsof sites we expectations. conform prior increasing YC falls that rainfall (Rainfalo, elevation (Wselvgr2) and cosine with can see low (Factor 2). 5) (Factor windiness and rises with aspect Model 7.1: Initial regressionmodel for Sitka spruce (mixed model) Coef

Predictor

Stdev

0.2482 0.00008489 0.0003906 0.03586 0.03365 0.06729 0.7437 0.0003260 0.002682 0.08576 0.06524 0.07692 0.2783 0.1808 0.007079 0.07685 0.5983

17.0792

Constant Rainfall WseIvgr2 Factor2 Factor5 Soi123 Soill Area Plantyr I st Rot mixCrop Park Ancient Uncleared Unprod Reserve Serni-nat

-0.00177733 -0.0070769 0.07469 -0.16595 0.89814 -4.9538 0.0037050 0.030379 1.52753 . -0.21314 0.91121 1.1777 2A639 -0.076776 -0.36615 -4.5487 s

t-ratio

R-sq= 40.9%

2.297

68.83 -20.94 -18.12 2.08

P

0.000 0.000 0.000 0.038 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.000 0.000 0.000 0.000 0.000 0.000

-4.93 13.35 -6.66 11.36 11.33 -17.81 -3.27 11.85 4.23 13.63 -10.85 -4.76 -7.60 R-sq(adj) 40.7%

Analysisof Variance Source Regression Error Total

DF

ss

ms

Fp

16 6062 6078

22122.7 31978.7 54101A

1382.7 5-3

262.10

0.000

Becauseof its discrete nature, soil type is consideredas a seriesof dummy variables, is YC by statistically significant. significantly proved elevated planting on which two of brown earth or podzol soils (SoI123,which is a simple combinadon Of S0112and S0113)and

by lowland lithornorphs depressed (soil). Both resultsconformto on planting significantly prior expectations. 7.22

The model also highlights the importance of silvicultural factors. The positive interesting (Area) is the the of plantation size and, to our knowledge, has relationship with indicate before. This identified formally been to that trees which are part of seem would not large plantations are more likely to thrive than those in small areas. This might be because large standsprovide advantagesin terms of the easeof adopting speciesspecific management regimes, or becausesuch standstend to condition their environment to their own advantage (for example, by reducing competition from both flora and fauna). Conversely this latter factor may be one of the pressuresmitigating againstsmaller stands. The strong and positive influence of the time variable (plantyr) is confirmed. This is usually explained as reflecting improvements in silvicultural methods such as the introduction of ploughing and fertilisers in improvements the genetic stock. and/or Two further silvicultural factors are identified. Trees planted on ground which hasnot been previously used for afforestation (Ist Rot) perform relatively worse than those planted in successiverotations. This may be becausesecond rotation trees have on average been in first (although low than those the the rotation relatively correlation recently planted more be inherit indicates the that trees story) this of or all second rotation may not a with plantyr forbears. Trees less base from to their also seem perform well when nutrient enriched soil in (MixCrop) finding than in monoculture, a plantation which suggests species grown a mixed benefit latter. be the timber associated the with amenity productivity cost a of there may that Next, a number of site factors which arise from the interaction of environmental identified. is YC higher in significantly were management practice and characteristics The (park), careful reflect more a may silvicultural result which management. parkland areas (ancient) boosts in tree ancient previously woodland areas were which that planting result Ist impact Rot. A further interesting be the the of and to corollary of rather growth seems ' identifies in by implied is trees the which uncleared growing areas boost to growth variable have but been by It been have not windblow yet cleared. affected seems previously which in from immediate (and that their trees actually profit windblow neighbours that the surviving boosting However, their to thus access nutrients. while growth rate removed competitors) are lack from the the the that ensuing of cover events, such raises probability benefit may fall themselves. to victim windblow subsequently survivors will 22Acounterexplanation,given by a seniorFC official who shall be nameless,is that this effect may also YC in tables. the effors aris,Cout of

7.23

Finally, three negative environmental/managementfactors are identified. Plantations land (unprod) higher, of unproductive not surprisingly perform relatively worse amounts with than otherwise similar others. Sub-compartmentswhich fall within the boundaries of lower do YC, (reserve) also exhibit relatively as areaswhich are allowed conservation areas to remain as semi-natural habitat (semi-nat); results which may reflect the application of less intensive silvicultural techniquesin such areas. Conversations with a number of forestry experts' suggestedthat model fit might be improved by omitting those standswhere YC measurementshad been taken relatively soon is difficult in YC The the early years of a rotation of particularly assessment after planting. likely have higher is hypothesis to that observations are therefore such variance than and our hypothesis (sage) To from this test survey a age stands. variable mature more taken those (plantyr) YC (survyr) data from and survey the year previously planting year was calculated described. Sub-compartmentswere iteratively removedfrom the datasetand on each iteration Model 7.1 was re-estimated. Figure 7.1 illustrates the resulting impact upon the fit of the (Appendix A5.3 truncation (R2-adj) of survey this age reports precise progressive of model values). Close inspection of Figure 7.1 confirms the expected (although small) increase in Omitting fit early age are removed. all observations at a with very surveyed stands as model leaves less ten still a reasonable assumption which us with than seems years a survey age of 24 from All scratch and the no three model variants were re-estimated 5168 observations. in interpretable Model 7.2. found as the results reported clearly to most factor model provide We also use this model to provide an interesting asideregarding the effect of aspectupon tree Sinasp in Cosasp including by is the the and ibis model. variable achieved growth.

FC Douglas Whiteman Macmillan of the MLURI. Adrian the Quine Chris of and included 2Mese and 2413ywhich we mean the full procedure for entering variables into the model was repeated. This was best describes be the untruncateddatasetwill also that the set of variables which sure cannot we as necessary less than ten of age a survey years are omitted. with stands all be optimal when

7.24

Figure 7.1: The impact of omitting standssurveyed at different ages

V

CL ,c

5

10

15

20

25

30

Survey age below which all observations

35

40

45

50

are omitted

Comparison of Model 7.2 with Model 7.1 shows that the omission of sites with in improvement degree but in the overall noticeable small a of explanation. sage<10 results factors has PCA allowed some new environmental variables to enter the The removal of all increases (Topexlkrn) does YC. As that shelter so as geomorphic see can model and we in the model to assessaspecteffects. Cosasp included Sinasp deliberately have and stated, we it is likely interpretable that, as a result of how variables as a pair only As these variables are them of one may model, appear a regression statistically within variation explain if we adopt a conventional 5% confidence test then neither of these However, significant25. it is Nevertheless, do have that clear we significant. to relax such not appear aspect variables be having does before to a by aspect significant appear effect. much too test a due it the to variation that two these aspect so may absorb appearsthatthereis little for 251ntaitively of one be However, the entered separately variables would meaningless. to the other explain.

7.25

Model 7.2: YC model for Sitka spruce after omitting standswith survey age <10 years

Stdev

Coef

Predictor

p ratio

Constant Rainfall Wselvgr2 Topexlkm Sinasp Cosasp Soil23 Soill Area Plantyr 1st Rot MixCrop Park Ancient Uncleared Unprod Reserve Serni-nat s

-0.00176521 -0.0084288 0.025931 0.7872 -0.6841 0.82527 -4.8614 0.0038847 0.050639 -1.7636 -0.28948 0.86170 0.9345 2.4261 -0.086657 -0.44077 -4.6318 2.306

Analysis of Variance DF Source Regression Error Total

0.2697 0.00009584 0.0003633 0.006818 0.4540 0.45792 0.07273 0.7504 0.0003639 0.003230 0.1005 0.06928 0.08295 0.2985 0.1821 0.007912 0.08421 0.7299

16.6333

17 5150 5167

R-sq = 42.1%

61.66

0.000 0.000 0.000 0.000 0.083 0.137 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.002 0.000 0.000 0.000 0.000

-18.42 -23.20 3.80 1.73 -1.49 11.35 -6.48 10.67 15.68 -17.56 -4.18 10.39 3.13 13.32 -10.95 -5.23 -6.35 R-sq(adj) 41.9%

SS

NIS

Fp

19921.2 27394.0 47315.1

1171.8 5.3

220.30

0.000

if we temporarily accept that some weak aspecteffect is occurring then we can use

is. illustrates Figure 7.2 in 7.2 Model this to this see what predicted given the coefficients Malcolm Worrell (1990b) from their study that and of with impact and compares our result in Britain. sites northern on upland Sitka spruce gowing of

7.26

Figure 7.2: The effect of aspectupon YC

Aspect curve for Sitka spruce in Wales (upland and lowland) Change in YC = 0.79 sinO-0.68

cos. 0

0

4 ect curve for Sitka spruce inspIlor thorn Britain (all upland) Change in YC = 1.61 sinO+0.29 cosO (from Worrell and Malcolm, 1990b)

0 0A

cis

-2-

L-

45

L-

90

"L 135

-L 180

oo,

225

270

315

360

Angle of aspect 0 (") The comparison of our results with those of Worrell and Malcolm (1990b) proves interesting. The magnitude of aspect impacts is slightly higher in the latter study, a result in is the given relatively more adverse conditions surprising of upland not areas which feature is in However, direction the Britain. the most striking the subtle shift of northern Worrell Malcolm between YC is two these that studies. and report most aspect effects facing This highest depressed slopes. eastern sites and on complete negation west on severely from increased be due the south to might seem to the solar radiation which of any effect has impact Considering the wind prevailing westerly upon such which sites. clearly powerful has here that the effect to aspect shifted the south see can round we results own our it is facing do best. Wales in It to sites which appear south east that would so somewhat Wales less that the of mean conditions adverse southern solar energy the relatively that seem by Nevertheless it is the is cancelled out prevailing west wind. the still completely not effect facing faces outperform site a south easterly one which makes south that which wind effect of

west. Returningto considerFigure7.1, while theredoesappearto be an increaseof fit from in that at a young age their sub-compartments surveyed surveyed prime are site omitting dramatic fall is in fit there a comparatively which occurs when we predicted, relatively well in YC sub-compartments to examining only which surveying occurred very confine ourselves 7.27

many years after planting. This does not seemto be a product of the smaller sample size of such analysesas we are still consideringmany hundredsof sites (indeed, as samplesize falls, the relatively large number of predictors in the model would tend to inflate goodnessof fit statistics)26. Two reasonsmay in part account for this effect, both of which arise from the observation that, as we restrict ourselves to older survey age, we are in turn restricting improved First, to stands. silvicultural methods, now applied to virtually all older ourselves in less have been applied a uniform manner to such older stands. New new stands,may well techniques may not have been simultaneously adopted for all plantations but rather tried on be, The these. result would as observed,that theseolder standsare more variable a subset of Secondly, it be that recordsregarding planting age are relatively less may than younger ones. function As YC is if becomes for then this a of plantation age stands. older uncertain reliable YC increase. Comparison estimates of will the with our subsequentanalysis variability so be in beech that there this argument to which suggests may some merit sub-compartments of we shall return. Whatever the reasonit seemsthat omission of those standswith relatively old survey improve fit A further is likely the to of our model. sensitivity analysis suggestedthat ages 36 in fit for survey age above an with years resulted optimal site of our models omission 4307 in As before leaving some sub-compartments us with our sample. models while still for better describing the to possibility of allow new explanatory variables afresh rebuilt were before As levels dataset. the aspect somewhat exhibited variables suspect of this revised from final these omitted models. accordingly significance and were All three model types were estimated. Model 7.3 reports results from our model factors. PCA While is interest describes this characteristics via environmental site of which it is by both to outperformed prior expectations conform our no-factor and all relationships This is interesting as each other. well as perfon'ned equally which an and mixed models finding suggesting that the PCA approach used by Macmillan (1991) may not provide any improvement by the conventional regression over more widespread models used significant b) Malcolm (1990a, in Worrell his more recent work, b), (1987a, and also, and worrell 1995,1996)". Dutch, Macmillan (Tyler, and Macmillan

2'lndeedin AppendixA5.3 theseriesof truncationsis extendeduntil thiseffectstartsto increaseRI statistics. 27Thisstudy concernsspeciesother than thoseunderinvestigationand is consequentlyomittedfrom our literature review.

7.28

Model 7.3:

Optimal PCA factor model for Sitka spruce: observations with sage<10 or sage>36 omitted Stdev

Coef

Predictor

t-

p

ratio

s

-0.70932 0.29481 -0.92229 -0.23857 -0.40778 0.0441 -4.2384 0.0036537 0.049234 -2.0853 -0.26907 0.80303 0.8805 2.7353 -0.086739 -0.42987 -4.3591 2.372

Analysis of Variance DF Source Regression Error Total

0.3090 0.04135 0.04177 0.06664 0.03667 0.03685 0.1366 0.9869 0.0003872 0.004954 0.1117 0.07848 0.09635 0.3171 0.2329 0.008315 0.09636 0.7831

11.8800

Constant Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 Soi123 Soill Area Plantyr Ist Rot MixCrop Park Ancient Uncleared Unprod Reserve Semi-nat

17 4289 4306

R-sq = 40.4%

38.45 -17.15 7.06 -13.84 -6.51 -11.07 0.32 -4.29 9.44 9.94 -18.67 -3.43 8.33 2.78 11.75 -10.43 -4.46 -5.57 R-sq(adj) .

SS

NIS

Fp

16342.51 24131.05 40473.56

961.32 5.63

170.86

0.000 0.000 0.000 0.000 0.000 0.000 0.747 0.000 0.000 0.000 0.000 0.001 0.000 0.006 0.000 0.000 0.000 0.000 40.1%

0.000

Given the very similar performanceof our no-factor and mixed models, the former is is Model 7.41. The interpretation list. of predictor for as reported and of optimal ease preferred in between lack be before found this to change of model as and specification variables was to overall validity. some added weight gives truncation options

7.29

Model 7.4:

Best fit YC model for Sitka spruce: no PCA factors used, observationswith sage<10 or sage>36omitted Coef

Predictor

Stdev

t-

p

ratio

s

-0.0016700 -0.0087750 0.024262 0.80489 -4.8827 0.0039518 0.049890 -1.9280 -0.30832 0.94769 0.9266 2.6411 -0.085426 -0.43395 -5.1415 2.319

Analysis of Variance DF Source Regression Error Total

0.3487 0.0001067 0.0003933 0.007592 0.08046 0.9660 0.0003788 0.004838 0.1093 0.07670 0.09385 0.3089 0.2276 0.008143 0.09452 0.7644

16.7097

Constant Rainfall Wselvgr2 Topexlkm Soi123 Soill. Area Plantyr I st Rot MixCrop Park Ancient Uncleared Unprod Reserve Semi-nat

15 4291 4306

R-sq = 43.0%

47.92 -15.65 -22.31 3.20 10.00 -5.05 10.43 10.31 -17.64 -4.02 10.10 3.00 11.61 -10.49 -4.59 -6.73 R-sq(adj)

0.000 0.000 0.000 0.001 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.003 0.000 0.000 0.000 0.000 42.8%

SS

ms

F.

p

17391.3 23082.2 40473.6

1159.4 5.4

215.54

0.000

For the purposesof extrapolationAppendix A5.3 gives descriptive statisticsfor all the best in The fit for all appropriateness of using our models. model such variables explanatory YC for 4307 by the actual comparing predicted with observations assessed extrapolation was in our revised dataset. Results of this analysis are presentedin Table 7.8 which shows that division YC". YC actual of one are within 76.5% of predictions

29, nis is a higherdegreeof accuracythanthatachievedby thethematicmapperapproachof Gemmell(1995) , 25% 75% Here have of actual were within growth of prcdicfions 75% rate. that roughly we over reports who in 40% 20% no predictions excess of of actual, with of actual. of predictionswithin

7.30

mi

C) C14

rn r-

0

CD (> r-

C) 000--.

mm

4 cn cn 0

ci

cn m le

00

00 rA

Gn

0 \o

Q E! gaq

c)

C)

c9 -4

r-4

-4

V's 10 cl) C,4 C) 00 iý - , It'l C'4 rn

C>

vl kn

nt

a

>4 cn 0

r_

-4

cli

rn

vli

r-

Igt cq

10

CD CD CD 0

CD

CO) ce

00

(D CD m

". i

Ici LL,

0C

=U JD

e ýo oo 0N

-,

9 Gn

-

8

cm ch M ce

CA cqi

rA

rA

0

0

CD rA

20 u :

=

4

EU f

8

xt %M00 c> (11

u rA

7.5

YIELD MODELS FOR BEECH The analysis of YC for beech sub-compartmentsfollowed the same methodology

Sitka Consequently investigation in sites. spruce only brief discussionsof of adopted our being here detailed in Appendix 5.3. results again presented with presented are methodology Following the deletion of sites for which key data was missing (giving us a dataset investigations initial linear 766 the again confirmed suitability of a observations), of functional form for our model. However, now a no-factor model provided the best initial fit

in Model 7.5. data as reported to the Model 7.5: Initial regession model: beech Stdev

Coef

Predictor

t-

p

ratio Constant Rainfall Wselvgr2 Avwatgra Plantyr Historic Monument NpAonbSa OthESA ForPark National FCconst Soil2

-0.0002490 -0.0043064 0.003182 0.008443 0.5229 -0.9295 0.4978 -0.4987 -0.3877 1.0305 -0.6026 0.2423

R-sq = 22.2%

s=1.363 Analysis of Variance DF Source Regression Error Total

0.5600 0.0001686 0.0005302 0.002302 0.002452 0.1067 0.6180 0.1444 0.2998 0.1894 0.3223 0.1468 0.1323

5.5089

12 753 765

ms

ss 399.763 1398.070 1797.833

7.32

33.314 1.857

9.84

0.000 0.140 0.000 0.167 0.001 0.000 0.133 0.001 0.097 0.041 0.001 0.000 0.067

-1.48 -8.12 1.38 3.44 4.90 -1.50 3.45 -1.66 -2.05 3.20 -4.10 1.83

R-sq(adj) = 21.0%

Fp 17.94

0.000

The explanatory variablesincluded in Model 7.5 are similar to thoseconsideredwithin interpretation is before. While some of these Sitka their so and as models spruce our it felt that this model provided an adequatebase to was weak, rather variables are clearly basis increasing impact the sub-compartments; on the of survey age. This of omitting analyse analysis was undertaken as before and results are illustrated in Figure 7.3 which for from Sitka results our analysis reproduces of sprucesub-compartments. purposes comparative Figure 7.3:

Impact upon model fit of omitting sites at successivesurvey age: beech and Sitka spruce

50

40

35

30

25

0

20

I

10 05

10

15

20

25

30

Survey age below which all observations

7.33

35

40

are omitted

45

so

55

in assessing Figure 7.3 an immediate point is the relatively lower degree of fit beech by of growth. This is very likely to be a product of the relatively our models exhibited beech (as the opposed to Sitka spruce) dependentvariable discussedin restricted range of Section 7.2.1. However, both curves initially rise (albeit slowly), peak and then eventually decline. Considering the curve for beech,the increase in fit from about sage=20 is probably due to the exclusion of standssurveyedat an early age. Note that this upward trend is much longer lasting than for our Sitka spruceanalysis indicating, as expected,that it is much more difficult to assessthe YC of a beech stand at say sage=10 than a Sitka spruce stand. Here is low by fit observations only sage achieved omitting all sites with excluding the optimal lower for Sitka This truncation (this at sage<10 spruce). an optimal with compares sage<38 best fit. 7.6 for 359 the dataset model provided which observations of gave a Model 7.6: Optimal (no-factor) model for beech: sites with sage<38 omitted Predictor

Constant Rainfall Wselvgr2 Avwatgra Plantyr Historic Monument NpAonbSa OthESAt ForPark National FCcons Soil2

Stdev

Coef

0.7357 0.0002479 0.0007218 0.003648 0.003044 0.1535 0.9340 0.2317 0.4753 0.2602 0.5096 0.2238 0.1863

4.7663 -0.0001754 -0.0043157 0.003301 0.013391 0.4699 -0.0937 0.6353 -0.0556 -0.4153 0.4156 -0.3452 0.2145

R-sq = 27.9%

s=1.258

t-ratio

p

6.48

0.000 0.480 0.000 0.366 0.000 0.002 0.920 0.006 0.027 0.111 0.415 0.124 0.250

-0.71 -5.98 0.90 4.40 3.06 -0.10 2.74 -2.22 -1.60 0.82 -1.54 1.15

R-sq(adj)

25.4%

Analysis of Variance Source Regression Error Total

DF 12 346 358

NIS

SS

17.643 1.583

211.712 547.798 759-510

7.34

Fp 11.14

0.000

Figure 7.3 also shows (as observed in our Sitka spruce data) that the degree of falls by models as we consider stands with relatively high sage, here afforded explanation 50 in about years seem to raise variance substantially. As previously of excess values be likely to this connected to such stands being consequentlyquite old at seems postulated introduction Uneven of advancesin silviculture may in part account the time of surveying. for the increase in variance here. Furthermore it may be that planting date is less certain in likely be beech is This to a problem more with sub-compartmentsthan with these stands. Sitka spruce as the latter were almost all originally planted by the FC, who generally keep in (and techniques silvicultural new apply a more uniform manner), while may good records by been have for beech a variety of planted private agents may which complete stands older be Given importance the available. not may of accurateage records planting and accurate into higher YC in translate may such uncertainty well variance calculating measurements within such stands. Given this we felt justified in additionally omitting those stands with high sage. A fit >49 that the sage would optimise of omission of our model. sensitivity analysis suggested Given 205 dataset the extent of the omission observations. of some This gave an effective begun to again afresh so as redefine was an appropriate analysis regression of observations, failed When Here to the enter variables model. many using our variables. explanatory set of Factor 2 describing of characteristics sites the only environmental to proved PCA approach is Model 7.7. as reported which to model enter our adequately significant beech: for sage<38 with sites and sage>49 omitted factor-only Best model 7.7: Model

Predictor Constant Factor 2 Plantyr2 AONBINSA OthESA

Stdev

Coef

t-ratio,

1.854 0.08458 0.01278 0.2719 0.4941

-5.227 -0.35371 0.08038 0.4614 -1.5826

s=1.266

0.005 0.000 0.000 0.091 0.002

-2.82 -4.18 6.29 1.70 -3.20

R-sq = 35.6%

p

R-sq(adj) = 34.3%

Analysis of Variance Source Regression Error Total

DF 4 200 204

ms

SS

44.285 1.602

177.140 320.303 497.444

7.35

IP

p

27.65

0.000

A no-factor alternative was also estimated and is reported as Model 7.8. Model 7.8: Optimal (no-factor) model for beech: sites with sage<38 and sage>49 omitted Coef

Predictor Constant Wselvgr2 Plantyr AONB/NSA OthESA

Stdev 1.923 0.0009149 0.01279 0.2710 0.4969

-4.428 -0.0038638 0.07995 0.4751 -1.4812

R-sq = 35.7%

S=1.265

t-ratio

p

-2.30 -4.22 6.25 1.75

0.022 0.000 0.000 0.081 0.003

-2.98 R-sq(adj) = 34.4%

Analysis of Variance Source Regression Error Total

DF

SS

4 200 204

177.649 319.794 497.444

ms 44.412 1.599

Fp 27.78

0.000

Models 7.7 and 7.8 are extremely similar both in terms of their degreeof explanation in Factor I Model is 7.7 explanatory variables; of essentially the effect of their choice and data is Wselvgr2 in the Model 7.8. raw environmental variable which used elevation Consequently we cannot have a mixed model for beech. Given its easeof interpretation we for YC in beech 7.8 Model optimal model predicting our as sub-compartments. prefer

An interestingsupplementaryanalysisconcernsthe considerationof aspecteffects. investigated been fit had best these building model In and rejected as insignificant. up our Nevertheless it is interesting to seeif the logical relationship between aspecteffects for Sitka had Wales implications Britain for in previously any noted and northern aspecteffects spruce Sinaspand Cosaspwere therefore added into our The Wales. in aspect beech variables upon best fit model which was then re-instated to produce Model 7.9.

7.36

Model 7.9: Including aspecteffects within our preferred beech model

Coef

Predictor Constant Wselvgr2 Sinasp Cosasp Plantyr AONB/NSA OthESA

Stdev 1.921 0.0009141 0.1274 0.1242 0.01278 0.2703 0.5007

-4.375 -0.0037821 0.1203 -0.1905 0.07952 0.4856 -1.4455

R-sq = 36.7%

s=1.261

t-ratio

p

-2.28 -4.14 0.94

0.024 0.000 0.346 0.127 0.000 0.074 0.004

-1.53 6.22 1.80 -2.89

R-sq(adj) = 34.8%

Analysis of Variance Source Regression Error Total

DF

SS

6 198 204

182.734 314.710 497.444

ms 30.456 1.589

Fp 19.16

0.000

As can be seen, both of the aspect variables are of very low significance. This of itself is interesting as aspect was clearly significant in the study conducted by Worrell and Malcolm (1990b) and on the edge of statistical significance in our Sitka spruce study. Similarly, consideration of coefficient estimates shows that the absolute magnitude of in Worrell largest Malcolm less in Sitka the and study, sizeable was effects our predicted here. Inspection of summary statistics given at the end of this smallest and spruce study While Worrell Malcolm these the explanation of all results. a consistent us and gives section in Britain, Sitka of northern areas sites upland or only spruce analysis study considered in less harsh lowland Wales. Furthermore both the climate sites of and upland considers

Sitka beech for descriptive beech and spruce our studies that statistics shows of comparison lower altitudesthan thoseof Sitka sprucesites. So it is generally planted at significantly depends impact lowland tree of aspect upon growth upon the altitude that such on that seems insignificant have be on upland sites aspect can while tree a major effect may upon this sites by implied 7.4 Model 7.9 the the Figure aspect curvc results of superimposes on to growth. in Sitka for described the Britain (from Worrell uplands of spruce northern previously those in the uplandsand lowlandsof Wales(from our models). 1990b) and Malcolm, and

7.37

Figure 7.4: Aspect effects for Sitka spruce and beech in differing locations Aspect curve for Sitka spruce in Wales (upland and lowland) Change in YC = 0.79 sino-0.68 cosO Aspect curve for beeCh in Wales (mainly lowland) Change in YC = 0.12 sin 0 -0.19 cos

0 cm c

Aspect curve for Sitka spruce in Northern Britain (all upland)

Change in YC = 1.61 sin 8+0.29 cosO (from Worrell and Malcolm, 1990b) 45

90

135

180

225 Angle of aspect 0 (a)

270

315

360

inspection of Figure 7.4 tells a clear and coherent story. In the upland areas of intensity be the to Britain the of prevailing westerly causes aspect wind a major northern in that trees determining tree relatively shelteredcast facing sites perform growth such factor facing west. The radiative energy advantageof south facing better those than significantly In impact Welsh by the the is prevailing wind. our of study negated of completely slopes both lowland Here both the magnitude and sites. and upland Sitka spruce we consider is Furthermore, impact in the the reduced. aspect reduction the of of significance statistical because lower both (induced are considering we sites at altitude wind the prevailing of power Britain) Wales to that the northern relative means less solar of conditions arduous the and 'sites be detected is as our aspect effect can now now southerly of advantage energy is facing This (rather trend than sites. continued east) when east we south at maximised is Here again beech substantially altitude reduced suchthat sub-compartments. our consider is the of aspect effect significance markedly reduced. statistical and magnitude the absolute impact in the of the prevailing westerly wind meansthat the solar Furthermore, the reduction boosted rind that the aspect curve is further facing that such we south of cnergy advantage facing for south-south-east. sites for beech sites now peaks illustrating to theseaspecteffects. Here the 7.5 Figure shows an alternative approach

is dotted by is directly the circle which centred given for uponthe compass basis comparison 7.38

in illustrates the absenceof any aspecteffect with points around the This the situation axes. impact YC. The a of zero aspect this upon showing results of Worrell and circle perimeter of Malcolm (1990b) are representedby the dashedline circle which is centred,a considerable facing the relatively positive aspect effect of east sites and the showing towards east way off displacement The impact this sites. extent of shows the magnitude of westerly the negative by in just I tree this a case raises growth maximum of over rný which of this aspect effect ha7l yf 1. The thick solid line circle representsour results for Sitka spruce in upland and lowland Wales. Here the displacementis a little less extreme, being most positive in the in Finally line the the thinner north west. solid most negative circle south east quadrant and in beech lowland Wales. Here from growing mainly areas of of analysis our shows results

be displaced the is to most effect shows positive aspect and on sites the circle only slightly facing south-south-east. Figure 7.5: Comparison of aspecteffects between Wales and upland northern Britain

N

E

w

S 7.39

Finally we can attempt to assessthe validity of our best fit model (7.8) by comparing in YC at all sub-compartments our final datasetwith YC as predicted by our model. actual Table 7.9 details results from this comparison.

Table 7.9:

Comparing actual with predicted YC for our best fit YC model of beech (cell contents are counts)

PredictedYC Actual YC 2 4 6 8 10 ALL

468 0 9 7 0 0 16

ALL 1 29 66 29 0 125

0 2 20 37 5 64

Predicted YC compared to actual YC

1 40 93 66 5 205

Percentageof total sample

Prediction is two classestoo high

1.5

Prediction is one class too high

23.9

Predicted YC equals actual YC

54.6

Prediction is one class too low

20.0

Prediction is two classestoo low

0.0 I

Considerationof Table7.9 showsthat98.5%of YC predictionsarewithin onedivision is a considerablyhigherrate of correctpredictionthan that achievedby This YC. of actual for dependent for the the range given although Sitka restricted model variable spruce our be therefore is treatedwith a little caution. and should surprising not perhaps beech this is the this the apparent validity warning, of model accepting Nevertheless,even encouraging.

7.40

7.6

MAPPING YIELD CLASS We have now estimated,for both of the tree speciesconsidered,two YC models, one

including PCA factorexplanatoryvariablesandthe otherwithout.For our Sitka sprucedataset model 7.3 providesthe best fitting PCA basedmodel while 7.4 gives a slightly better fit without using PCA factors.Similarly for our beechdataset,model 7.7 gives the bestPCA basedpredictionsof yield while model7.8 providesa slightly betterfit without usingPCA factors.Thesefour modelsareusedto provideestimatesfor the GIS imagesof YC presented below. and analysed 7.6.1: PRODUCING GIS IMAGES OF YIELD CLASS In this context, an image is simply a spatially referenced depiction of a dataset GIS be displayed by then the can upon a screenor printed as required. To which produced data GIS image for YC the the on all predictor requires variables a all the grid points produce (the 'coverage') for which we want to predict, in this case the entire land area of Wales. if Sitka VAR (7.4) fitting best model as an example, we can see that this is spruce the take we The is in by of explanatory and a variables. number constant a constant essencea predicted has identical (here 16.709) for its in values right which all grid points. The first own coverage in is Rainfall for have this the predictor model which we variable a full coverage explanatory from the LandIS database.We can therefore build up our GIS predicted YC map by telling Rainfall being image by its the coverage multiplied to a new calculate coefficient the software is by GIS 1drisi Using performed use of the Scalar command. this the operation (-0.00167). for by be the that image constant then with combined The resultant use of the Overlay can images in these two its overlays to produce a third command, which as name suggests, effect in Subsequent first the by model. two elements these being YC as predicted predictors are being images created by multiplying the incorporated in a similar manner with separate Scalar its by the command incorporating using then coefficient and values variables coverage Overlay command. YC the into image the using map the resultant first to construct component score images based PCA modelswe need When using the first by creating images Z-score Wales. This achieved was of each of covering the whole of

7.41

the variables considered in the PCA" and then using the component score coefficients beech for Sitka to produce images of each factor. These were then spruce and calculated treated as were the explanatory variables discussedabove. In all the models a number of the predictor variables are related to management(e.g. Area), policy (e.g. reserve) or when the site was planted (e.g. plantyr). These are not by holding fixed treated them are so where variables spatial at certain values (i. e. specifically in The these certain of the and varying a analysis. sensitivity constant) variables as per MixCrop, ancient, unprod, reserve, park, uncleard and semi-nat are all dummies for infrequently occurring, unusual sites and were consequentlyheld at zero (their median value) for all images. Similarly the variable Area was held at its median value of 33 ha for Sitka Given low for beech ha 10 the very sites. value of the coefficient on this spruce sites and descriptive (see in Appendix its 5) the statistics range given small relatively and variable here. However, justified did for this the not the not seem case was analysis sensitivity full Ist here. Rot analyses sensitivity were conducted and and variables plantyr SITKA SPRUCE FOR IMAGES YIELD TIMBER GIS 7.6.2 We produced images basedon both our best non-PCA and PCA basedyield models. impact from 0 (being the the of changing plantyr variable considered also Further to this we Sitka Commission Forestry 75 to (being in plant started to the spruce) base the year which 75 Sitka i. about thereby years day, commenced ago) planting arguably spruce e. the present both For initially these of that analyses period'o. we over progress reflecting technological both first these However, trees time i. of at 1, periods. rotation Rot Ist hold e. examining = in Therefore their second rotation. Sitka now day are plantations we also spruce many present (i. 75. Rot This Ist rotation) letting when second e. plantyr = =0 test the effect of in 6 images being Table resulted differing created. assumptions and models combination of labelling simple images system which we adopt and provides a 7.10 details these subsequently.

for taken from this operation were the variable values for deviations necessary 29'rhemeans and standard be different somewhat These from those for the entirety will (both species). all the forestry sub-compartments is liable discrepancy be to dataset, forestry any minor. the of the size but given of Wales this of effect. interpretations discussion possible of 3OSee previous our

7.42

Table 7.10: Sitka spruce GIS timber yield class images created: image labels Model type

plantyr-0 lst Rot=1

plantyr= 75 Ist Rot--1

plantyr=75 1st Rot--O

No PCA factors used (model 7.4)

SSIVAR

SS2VAR

SS3VAR

P actors used (model 7.3)

SSIFAC

SS2FAC

SS3FAC

Images were produced using the procedure outlined in section 7.6.1. Figure 7.6 illustrates the predicted YC image createdfrom model 7.4 (no PCA factors used) with plantyr Ist Rot day) (replanting (present 75 and =0 on a previously planted site) i. e. image = SS3VAR. Inspection of figure 7.6 clearly shows the very strong influence which environmental have YCinfluences The our predictions upon of of lower altitude, better soil characteristics high YC. to The less-excessive combine produce rainfall pattern of lower YC produced and by higher elevations is particularly noticeable with the mountain ranges of Snowdonia, the Brecon Beacons Cambrians Less the clearly and picked out. extreme upland areassuch mid YC Mountains Preseli produce values which lie betweentheseextremes.Also clearly the and lying is to the east of the Carnbrianswhich results the excess rain-shadow adverse noticeable

YC valuesstretchingin somecasesup to (andacross) in large areasof relatively depressed The border. English adverseeffect of sandyandestuarinesoils upongrowth can also be the depressed low but in significantly areas of yield at placessuchas the tip of seen the small Pembrey, Peninsula the southemmostpart of Anglesey and the nearby and Gower the LandudnopeninsulaO'.

Nnterestingly both Pcmbrey and Newborough (Anglesey) are the sites of large forests, underlining the point land. to the marginal most confined that forests are often

7.43

Figure 7.6:

Image SS3VAR: predicted yield class from our optimal (no factor) model of Sitka spruce growth (assuming plantyr = 75; lst Rot = 0).

Class Yield Spruce Sitka Predicted Model Variable from (m'/ha/year) M<=

10

16 0

12

18

14

22() 7.44

lp

2t0 30

40

I -. 1 300 000

50

kni

Figure 7.7 reproducesimage SS3FAC, which uses the same assumptionsregarding Plantyr and lst Rot as Figure 7.6, but employs our best fitting factor based model (7.3) of YC. While the general pattern of YC predictions is similar between our factor-based(figure 7.7) and no-factor models (figure 7.6), some interesting differences can be detected.Figure 7.7 illustrates a smaller range of YC values than does figure 7.6 (compare estimates for Pembroke, the Lleyn Peninsula,Anglesey and the North Wales coast where figure 7.6 records Snowdonia figure 7.7; high than areas such as and also compare upland values many more is lower Another difference figure 7.6 Beacon Brecon values). noticeable records where the is figure "blocky" 7.6. This because is than 7.7 figure arises more of the considerably that formers reliance upon PCA factors dominated by 5krri' resolution variables such as those linked to water availability, while the latter is driven by variables such as elevation which is lkmý grid. a on recorded These difference excepted,imagesSS3VAR (figure 7.6) and SS3FAC (figure 7.7) give YC falls However, YC predicted systematically alter when we predictions. similar reasonably Ist Rot. Table 7.16 details YC for predicted and all our plantyr our assumptions regarding Sitka spruce images showing the extent of this decline. While our YC images seemhighly plausible (and we would defend them as such for in figures 7.6 indicate 7.7 do 7.11 Wales), table and and a weakness our of the majority for YC to conditions their extreme environmental such to predict ability regard with models fails fitting (SS3VAR) Our best to tops. predict any sites model for mountain example, as, if tops However, the they trees at very of YC6mountains were planted clearly less than of low YC. Similarly best only very fail produce our model at to or would survive well might 20, dataset indicated in few have YC our of yet to excess a cases does not predict any cells be YC lower 24. We to the high overpredicting at therefore appear extreme being as YC as of tail. the upper at and underpredicting

7.45

Figure 7.7:

Image SS3FAC: predicted yield class from our best fitting factor based model 0). 75; Ist (assuming Rot Sitka plantyr growth = = spruce of

Class Yield Spruce Sitka Predicted I Model PCA from (m'/ha/year) 12 14

18 0 10 EheIý6.

20

16

I:

7.46

20

3p

40 50 ýi=--Zm: j

13 00 000

kni

u

I

I

I

#

" -< 4

t-.: -4

) c1r En C/)

"

t-

00 CN 140 -

C4

6 06 C14 kill

cli -*

1*0

1

tn

't

C. 4

tn

, I

I

<

C) en 00

00 1-4

0

00 "D -4

00

00

Cý 't

- cnI

W) Cý

0 cf) Cý

r1l:

ON cl

0

til)

0

--4

tr;

I I Cq 00

t00 oo o ý, C-4

ý6 1-4 eq

'o en

C*4

C4 in

00 tfl 0%

Co

C4

t't

C' j C'n En



LLI

( 14

C14

e

-4 cn

N C4

cq C-4 in in

'IT

en cq

a, 8

C-4 [ý,

t -, C7,

00 00

(q

0 %,

'.0

kn

-

C5 C)

C4 Cn

cn

C14



t W4:

M

t-

-- !

tr;

M

C'i vi C14 Cý C-4 I--*

IT

141;

> r1l

en C/)

1.4

cn

*10

LLI

W)

wl

en

en

-4

r-

ýo

W, ,t

r4

C14 W*l

IRt

C7%

cc7ns

Zý 00 C-4 r- m - (2N

Ci C4 N'i

10 9

ttý .4 't -

C14 00

C,4

C/) C/) tn

LL,

qd, tn tn



00

-4

"

C14 tn

04

d

<

CA U)

(4. al Lu

C)

C.4 4 C, Cý

00 r-4

ý,o 10 0ý 0

cn 00 cn --4 CIR ttl

C4

Cq

. JD

Cf)

PA M U cn

C4 'IR 00 cf )

C-4 't

'D

ro- wl s C14 11,

V)

C4

_4

00

r-

00 -4

u cq

ýc

00

clq

00

0

Cd 0

rA

z

C14

Three factors seempertinent in explaining this. Firstly, we are predicting averageYC This Ikm' will tend to remove any extremes and therefore gives some square. grid over a less discussed in Secondly, Appendix 5, in creating findings. and positively, as to support our fully digital to capture the upper extremes of altitude. unable our elevation model we were This means that we are under-representingelevation at the tops of mountains and therefore is Thirdly, little YC there these as points. relatively at planting at the over-estimating low YC in the observations are under-represented relatively so resultant extremes of altitude FCs sub-compartmentdatabaseresulting in a lesserability of statistical models basedon such data to estimate accurately for such locations. However, while these are problems, the actual in 7.8 degree table that the suggests of over and reported versus predicted comparison is tails the not overly serious. underprediction at 7.63: GIS TIMBER YIELD IMAGES FOR BEECH As before, we produced images based on both our best non-PCA and PCA based yield

impact Ist Rot the the Further of changing plantyr this considered and to again we models. Sitka there In the unlike our spruce analysis was variable, no plantyr the of case variables. Thus have date beech in commenced. although we a planting at which distinct year which in dataset (some 162 to the the oldest record purely years ago) this corresponds plantyr =0 it date. Accordingly decided initial to was planting adopt a to actual some than rather here and our sensitivity analysis examined two values: plantyr different strategy somewhat date); 162 (the both and plantyr the planting and median (which mean = 144 equalled = beech few in dataset ne sub-compartments were day). comparatively not showed present being Ist Rot held not performed, at this a value of was first analysis so and rotation their factor differing The images. and non-factor models and of beech plantyr combination I for all images.Table 7.12 details these images and provides beech YC different four values yielded labels as before.

7.48

Table 7.12: Beech GIS timber yield class images created: image labels

Model type

plantyr--144

Ist Rot=I

plantyr=162

Ist Rot=I

No PCA factors used (model 7.8)

BEIVAR

BE2VAR

PCA factors used (model 7.7)

BEIFAC

BE2FAC

Images were produced using the procedure outlined in section 7.6.1. Figure 7.8 illustrates the predicted YC image createdfrom our best fit beechmodel 7.8 (no PCA factors day) Ist Rot 162 (present (first i. image BE2VAR. and rotation) plantyr =I e. = with used) As expected the general pattern of YC predictions observed for our Sitka spruce images is repeated in our beech images with high elevation and poor soils being associated YC its both However, lower YC. lower of the values and absolute range are much than with

before. This is again as expectedand reflects the restricted range of beechYC values recorded in the sub-compartmentdatabase.Our commentsregarding theseand other limitations to these Sitka images. discussion for the spruce of our predictions are as As for our Sitka spruce analysis, the general pattern of predicted YC for beech is being images FAC images (with between blocky again somewhat more consistent reasonably here. However, further 7.13 are reproduced VAR table maps and'so no equivalents) than their in 7.12. detailed images four table from YC the results presents

7.49

Figure 7.8:

Image BE2VAR: predicted yield class from our optimal (no factor) model of beech growth (assuming plantyr = 162; lst Rot = 1).

Class Yield Beecb Predicted (m'/ha/year) from Variable Model o

<= 6

10

30

40

50 -1

I:

7.50

20

1 300 000

km

Table 7.13: Predicted timber yield class from various beech maps'. Yield Class

%

Freq!

Mean

6.25

s.d.

0.80

Notes:

Freq

0.005 OA09 9.580 50.756 39.250

1 84 1970 10437 8071

3 4 5 6 7 8 9

BE2VAR

BEIVAR

-

%

17 421 7003 12925 197

0.083 2.047 34.056 62.856 0.958

BEIFAC Freq

BE2FAC %

14 1725 13251 5573 -

Freq

0.068 8.389 64.440 27.102 -

%

208 6775 13580

7.69

6.19

7.63

0.78

0.76

0.70

1.012 32.948 66.041

1. For key to images see table 7.12 2. Each map consists of 20563 lkm' land cells.

With our predicted YC images for Sitka.spruce and beech defined we can take the BE2VAR images (SS3VAR them to respectively) and and these use produce of optimal of

timber value.

7.7

VALUING

TIMBER YIELD

In chapter 6 we producedtables of NPV and annuity equivalentsfor Sitka spruceand YC full discount and at of a range various across rates (details in beech timber values Appendix 4). We can now use those results to convert our optimal predicted YC images to those yields. detailing of the equivalent monetary maps

SPRUCE SITKA VAýUE: TIMBER QF MAPS 7.7.1: its NPV timber have annuity equivalent. Each of these and We value, two measuresof in discount following the and rate at been various analysis we shall calculated have 3% discount 1.5%, 6%; 6% four the these: rates and exponential and a of concentrate on images 8 Sitka have We timber discount therefore spruce value rate. which we hyperbolic labels for details 7.14 Table provides subsequent these and referral. to create. wish

7.51

Table 7.14: Sitka spruce GIS timber value images created: image labels Discount rate'

Value measure

1.5%

3%

6%

6% hyperbolic

NP

SSltNPV

SS3tNPV

SS6tNPV

SS61itN

SSltANN

SS3tANN

SS6tANN

SS6HtANN

ity

Note: 1. All discount rates are exponential unless otherwise stated. 7.7.1.1: Estimating equations to convert from yield class to values A simple method to relate the YC images to their value equivalents was to use the data linear 4 in Appendix to a source of estimate as equationsrelating NPV and tables given for discount YC the rates considered. to various annuity values All timber values are considerably influenced by the planting grants and subsidy in 6 As there chapter shown are a multitude of possible scenarios, applicable. schemes be Consideration trees which might planted. schemes under subsidy of and planting grants following impractically the analysis make cumbersome and would all these permutations have in following is for Accordingly taken the the we case which most general our complex. benefit Community the grassland upon unimproved planting without of study area, namely from financial Deviations Supplement. the resulting measurescan be calculated Woodland

from the tablesreportedin chapter6 andappendix4. Within this general casewe have two rates of grant payable depending upon whether for disadvantaged/specially disadvantaged (DA/SDA) the areas rate at or grants are paid details linear equations linking Sitka spruce NPV sums for DA/SDA 7.15 Table otherwise.

details 7.16 for discount table YC results rates while an equivalent various across to areas non-disadvantaged area.

7.52

Table 7.15:

NPV of timber from an optimal rotation of Sitka spruce: linear predictive (various YC discount rates). the as single explanatory variable with equations For disadvantagedand severely disadvantagedareas.

Discount rate 1.5% 3% 6% 6% hyperbolic

Table 7.16:

Intercept (t-value)

Slope (t-value)

W (adj)

-3645.4 (-31.96)

996.621 (140.34)

100.0

-3013.7 (-16.80)

570.20 (51.06)

99.7

-1540.2 (-9.12)

209.02 (19.88)

97.8

-2037.6 (-12.57)

558.78 (55.37)

99.7

NPV of timber from an optimal rotation of Sitka spruce: linear predictive (various discount YC the single explanatory variable rates). as equations with For non-disadvantagedareas.

Discount rate 1.5% 3% 6% 6% hyperbolic

Intercept (t-value)

Slope (t-value)

R' (adj)

-4204.9 (-36.88)

996.670 (140.39)

100.0

-3540.9 (-19.74)

570.20 (51.07)

99.7

-2008.0 (-11.89)

209.01 (19.87)

97.8

-2518.6 (-15.53)

558.74 (55.34)

99.7

Sitka link YC to to spruce annuity conducted values A similar analysis was also

Sitka linking for linear details 7.17 spruce annuity equivalents equations Table estimates. 7.18 details discount for table while YC rates, results nonacrossvarious DA/SDA areasto disadvantagedareas.

7.53

Table 7.17:

Timber annuity equivalent of a perpetual series of optimal rotation of Sitka YC linear equations with as the single explanatory variable predictive spruce: (various discount rates). For disadvantagedand severely disadvantagedareas.

1.5% 3% 6% 6% hyperbolic

Table 7.18:

R2 (adj)

-104.183 (-34.55)

25.3951 (135.28)

100.0

-119.003 (-15.90)

21.4204 (45.97)

99.6

-104.24 (-8.37)

13.8902 (17.91)

97.3

-172.51 -14.98)

44.0728 (61.45)

99.8

Timber annuity equivalent of a perpetual series of optimal rotation of Sitka YC linear as the single explanatory variable with equations predictive spruce: (various discount rates). For non-disadvantagedareas.

Discount rate 1.5% 3% 6% ----------6% 1 hyperbolic

Slope (t-value)

Intercept (t-value)_

Discount rate

1

R2

Intercept (t-value)

Slope (t-value)

-116.398 (-38.55)

25.3135 (134.67)

100.0

-136.324 (-17.88)

21.3151 (44.90)

99.6

-132.22 (-iO. 67)

13.7573 (17.83)

97.2

207.35 . (-17.74)

43.9472 (60.38)

99.8

1

7.54

1

(adj)

7.7.1.2 Maps of timber NPV: Sitka spruce Given that the majority of Wales qualifies for DA/SDA rates of subsidy we shall use 32 images NPV maps for Sitka spruce timber value were following in the these rates . image YC (SS3VAR) by linear by the optimal relevant our equation as multiplying produced detailed in table 7.15. This was achieved using the Scalar command discussedpreviously. This operation was repeated for each of the four discount rates considered to produce the images defined in the upper row of table 7.14. Table 7.19 details results from this analysis. Table 7.19 clearly shows both the range of NPV sums which are implied by our YC discount As discount impact these. rate upon exponential the of varying rates predictions and increase so the absolutevalue of NPV, its range and consequentlyvariance,decline markedly. Switching to hyperbolic discounting increasesthesemeasuresof NPV substantially as shown. Given our discussion of discount rates in chapter 6 we are less interested in the 6%

is because it including Treasury this the mainly current relevant rate. exponential rate, Furthermorewe recognisethat the resistancewhich the economicsprofessionhas towards be discounting to unlikely an approach given too much weight. such makes hyperbolic Consequentlywe prefer to concentrateon out 1.5%and 3% ratesand choosethe latter to by NPV distribution the aboveanalysisas shownin figure sums estimated of illustrate the 7.9. The distribution of NPV sums shown in figure 7.9 strongly reflects that of the YC Consequently before. (figure 7.6). is based it comments are as our image upon which Sitka timber spruce Maps annuity: of 7.7.1.3 in detailed NPV table 7.19 were prepared.'niis was the sums Annuity equivalents of

Sitka YC (7.4) Scalar optimal our spruce relating the command now model againachievedvia four in 7.17 (DA/SDA to the table linear areas), produce annuity given equations the through from Results lower 7.14. detailed in in table this the of row exercise described are images table 7.20.

in future is hope to preparea DA/SDA boundary image to work. address we "An obvious extension, which Wales. However, to areas of the time of writing, permission all at define applicable map single a to this and use is Crown Copyright) had been requestedbut not granted. (which image to use such an

7.55

Table 7.19:

NPV sums for Sitka sprucetimber GIS images at various discount rates (Ma, 1990)

NPV (f/ha)

SSILNPV Freql

-500A 0:499 500:999 1000:1499 1500:1999 2000:2499 2500:2999 3000:3499 3500:3999 4000:4499 4500:4999 5000:5499 5500:5999 6000:6499 6500:7000 7000:7499 7500:7999 8000:8499 8500:8999 9000.9499 9500:9999 loooo. 10499 10500:10999 ii ooo:11499 11500:11999 12000:12499 12500:12999 13000:13499 13500.13999 14000:14499 14500:15000 15000.15499 15500:15999 i 6ooo:16499 16500:16999 Mean s.d.

Notes: I. 2.

%

1

0.005

4 5 10 11 8 17 23 62 80 207 352 525 649 739 826 1112 1194 1595 1820 2162 2225 2605 2600 1561 168 2

0.019 0.024 0.048 0.053 0.039 0.083 0.112 0.302 0.389 1.007 1.712 2.553 3.156 3.594 4.017 5.408 5.807 7.757 8.851 10.514 10.820 12.668 12.644 7.591 0.817 0.010

13362.45 1938.29

SS3LNPV Freq

1 2 8 20 24 48 163 514 1019 1307 1757 2556 3380 4055 4534 1173 2 -

%

0.005 0.010 0.039 0.097 0.117 0.233 0.793 2.500 4.956 6.356 8.544 12.430 16.437 19.720 22.049 5.704 0.010

SS6LNPV Freq

%

1 31 187 2232 5786 11208 1118 -

0.005 0.151 0.909 10.854 28.138 54.506 5.437 -

SS6HLNPV Freq

%

1 0.005 4 0.019 13 0.063 16 0.078 30 0.146 81 0.394 239 1.162 711 3A58 1139 5.539 1480 7.197 2073 10.081 2927 14.234 3919 19.059 4447 21.626 3358 16.330 125 0.608

-

6707.30 1189.19

2023.25 2

438.32

7488.72 1167.57

Froma totalof 20563WrOlandcells. Estimated(not calculated dueto the GIS assigningzerovaluesto non-land for in thecalculationof themean). cells;thisproblemis adjusted

7.56

Figure 7.9:

Image SS3tNPV: predicted timber NPV sums for Sitka spruce (based on yield class image SS3VAR; optimal no-factor model 7.4). Discount rate = 3% (f/ha, 1990)

for Sitka Spruce Value Net Present Timber 0 ý

k III

I -. 1

7.57

Table 7.20: Annuity values for Sitka spruce timber at various discount rates (f/ha, 1990) Annuity value (f/ha) -25:-1 0:24 25:49 50:74 75:99 100:124 125:149 150:174 175:199 200:224 225:249 250:274 275:299 300:324 325:349 350:374 375:399 400:424 425:449 450:474 475:499 500.524 525:549 550:574 575:599 600:624 625:649 650:674 675:699 700:724 Mean s.d.

Notes: I2.

SSRANN

SSRANN

%

Frcq'

1 2 15 18 34 115 411 1044 1460 1994 3010 4172 4837 3380 70 -

Freq

0.005 0.010 0.073 0.088 0.165 0.559 2.000 5.077 7.100 9.697 14.638 20.289 23.523 16.437 0.340

-

%

3 16 22 60 263 993 1682 2413 3962 5175 5626 348 -

0.015 0.079 0.107 0.292 1.279 4.829 8.180 11.735 19.268 25.167 27.360 1.692 -

SS6tANN Freq 1 21 53 479 2183 4068 7318 6434 6 -

% 0.005 0.102 0.258 2.329 10.616 19.783 35.588 31.289 0.029

-

328.84

246.18

132.57

54.17

47.61

30.101

SS6HLANN Freq

1 2 5 10 13 8 22 29 78 136 312 546 730 812 966 1230 1551 1865 2326 2539 2897 2946 1447 92

%

0.005 0.010 0.024 0.048 0.063 0.039 0.107 0.141 0.379 0.661 1.517 2.655 3.550 3.949 4.698 5.982 7.543 9.070 11.312 12.347 14.088 14.327 7.037 0.447

578.86 86A4

From a total of 20563 IkO land cells. Estimated (not calculated due to the GIS assigning zero values to non-land cells; this problem is adjusted for in the calculation of the mean).

As with our NPV analysis, table 7.20 clearly shows that increasing the discount rate For range and of annuity variance value, sums. the absolute comparative purposes reduces image SSRANN. 7.10 figure reproduces

and

Figure 7.10 again reflects the broad distribution pattern observed in previous images between NPV the surns and their annuity equivalents. relationship underscores 7.58

Figure 7.10:

Image SS3tANN: predicted timber annuity equivalents for Sitka spruce (based on yield class image SS3VAR; optimal no-factor model 7.4). Discount rate 3% Wha, 1990)

Timber Annuity Value for Saka Spruce (f/ha, 3% Discount Rate) < 150 150 - 199 IM

250-299

kill

>= 300 I -. 13 00 000

200-249

7.59

7.7.2 MAPS OF TIMBER VALUE: BEECH As before we calculateNPV and annuity equivalentsfor our four discount rates. Table 7.21 details the 8 beech timber value images created from such an analysis. Table 7.21: Beech GIS timber value images created: image labels. Discount rate' Value measure

1.5%

3%

6%

6% hyperbolic

NP

BEltNPV

BE3tNPV

BE6tNPV

Annuity

BEItANN

BE3tANN

BE6tANN

BE6HtNPV NBE6HtAN

Note: 1. All discount rates are exponential unless otherwise stated. from to Estimating yield class to values convert 7.7.2.1: equations As before linear equations were estimated to related our Beech YC images to their from Appendix 4 Data taken assuming planting on unimproved was equivalents. value details 7.22 Woodland Table linking Community supplement. equations grassland without beech NPV sums for DA/SDA areas to YC across various discount rates while table 7.23 for areas. non-disadvantaged results reports Table 7.22:

NPV of timber from an optimal rotation of beech: linear predictive equations (various discount YC the variable explanatory single rates). For as with 'disadvantagedand severely disadvantagedareas. R2

Intercept (t-value)

Slope (t-value)

-1513.9 (-3.09)

749.95 (11.26)

97.7

3%

-260.0 (-1.35)

349-50 (9.63)

96.8

6%

455.90 (5.89)

63.30 (6.01)

92.1

-1024.8 (-2.65)

624.90 (11.89)

9-Y0

Discount rate 1.5%

6% hyperbolic

7.60

(adj)

Table 7.23:

NPV of timber from an optimal rotation of beech: linear predictive equations with YC as the single explanatory variable (various discount rates). For nondisadvantagedareas.

Discount rate 1.5% 3%6% 6% hyperbolic

R2

Intercept (t-value)

Slope (t-value)

-2299.9 (-4.70)

749.95 (11.26)

97.7

-1096.7 (-4.10)

349.35 (9.60)

96.8

-160.20 (-2.07)

63.10 (5.98)

92.1

-1679.4 (-4.36)

624.70 (11.92)

97.9

(adj)

A similar analysis was also conductedto link beech annuity values to YC estimates. Table 7.24 details linear equations linking annuities to YC across various discount rates for DA/SDA areas, while table 7.25 details results for non-disadvantagedareas.

Timber annuity equivalent of a perpetual series of optimal rotation of beech: linear predictive equationswith YC as the single explanatory variable (various discount rates). For disadvantagedand severely disadvantagedareas.

Table 7.24:

F[-iscount

Intercept (t-value)

Slope (t-value)

R' (adj)

-29.832 (-3.20)

14.416 (11.36)

97.7

3%

-12.813 (-1.48)

11.327 (9.64)

96.8

6%

27.032 (5.63)

4.009 (6.13)

92.4

-76.02 (-2.76)

44.553 (11.88)

97.9

rate 1.5%

6% hyperbolic

7.61

Table 7.25:

Timber annuity equivalent of a perpetual series of optimal rotation of beech: linear predictive equationswith YC as the single explanatory variable (various discount rates). For non-disadvantagedareas.

Diicount rate

Intercept (t-value)

Slope (t-value)

R2 (adj)

-44.445 (-4.73)

14.373 (11.23)

97.7

-35.687 (-4.10)

11.246 (9.50)

96.7

-10.143 (-2.10)

3.9165 (5.97)

92.0

-121.30 (4.36)

44.444 (11.74)

97.9

1.5% 3%. 6% 6% yperbolic

7.7.2.2 Maps of timber NPV: beech

As before we assumeDA/SDA rates for the following analysis. NPV images were Table Sitka 7.26 details for four beech spruce analysis. the our timber per results as produced NPV images defined in the upper row of table 7.21.

Table 7.26:

IFI-v . PTV (f/ha)

500: 999 1000:1499 1500:1999 2000:2499 2500: 2999 3000:3999 4000:4999 mean

s.d.

Notes: I-

NPV sums for beechtimber GIS imagesat various discount rates (f/ha, 1990) BEANPV

BEItNPV %

Frcq' 10 97 5410 15046

0.049 0.472 26.307 73.165

Freq

BE6tNPV %

Freq

10 1281 14524 4748

0.049 6.229 70.626 23.088

-

-

20563

%

BE6HLNPV Freq

%

27 332 13440 6764

0.131 1.615 65.355 32.891

100.000

4250.78

2326.53

942.49

1 3778.66

494.83

331.31

317.49

426.95

From a total of 20563 lkný land cells.

7.62

Analysis of table 7.26 shows a similar pattern of NPV to those observed for Sitka longer in delay However, the a of as result returns and their lower growth rate, the spruce. beech for NPVs level timber are considerably below those observed for Sitka absolute of

SS3tNPV image (figure figure 7.9), To 7.11. the with printed above comparison spruce. allow reproduces image BE3tNPV. Figure 7.11 shows the now familiar pattern of values corresponding closely to the Comments before. therefore sites. of are as environmental characteristics 7.7.2.3 Maps of timber annuity: beech Annuity equivalentswere preparedasbefore. Resultsfor all four of the imagesdefined in the lower row of table 7.26 are given in table 7.32.

Table 7.27: 7m -ýj -y ý nuit ha)

F[7 j 40:49 50:59 60:69 70:79 80:89 90.99 100:149 150:199 200:249 250:310

Annuity equivalentsfor beechtimber GIS imagesvarious discount rates (Vha, 1990) BE3tANN

BERANN %

Freql 20 179 1798 6253 8960 3353

%

Freq

0.097 0.870 8.744 30.409 43.573 16.306

20 327 4756 10841 4619

0.097 1.590 23.129 52.721 22A63

BE6tANN Frcq

%

37 16203 4323

Frcq

%

0.180 78.797 21.023

-

1 0.005 173 0.841 4962 24.131 15427 75.023

-

-

BE6HLANN

-

mean

80.98

74.25

57.8

266.45

s.d.

12.97

12.09

11.52

26.97

Notes: 1.

the

From a total of 20563 IkO land cells.

image BE3tANN. The showsclearly 7.12 figure reproduces For comparativepurposes, Other before. are as comments values. of expected pattern

7.63

Figure 7.11:

Image BE3tNPV: predicted timber NPV sums for beech (based on yield class image BE2VAR; optimal no-factor model 7.8). Discount rate = 3% (f/ha, 1990)

Timber Net PresentValue for Beech (F-/ha,3% Discount Rate) 2250 -2499

< 1750 1750-1999 2000-2249

M

2500-2749

I () ý-Nrw ý

I: 7.64

f) -1

10

40

1 3()o ooo

K

ii

Figure 7.12:

Image BE3tANN: predicted timber annuity values for beech (based on yield class image BE2VAR; optimal no-factor model 7.8). Discount rate = 3% (f/ha, 1990)

Timber Annuity Value for Beech (F-/ha,3% Discount Rate) < 59

70 -79

60 -69

80 -89

7.65

()

I ()

I:

()

30

4o

13 00 ()() ()

()

km

7.8

CONCLUSIONS We have estimated yield class models for Sitka spruce and beech basedin part upon

GIS datasets from Wales. This has allowed us drawn the covering entire extent of variables to use those models to produce predicted yield maps for both species for the entire Principality. We have then used thesemaps in conjunction with our previous work on timber In happy NPV equivalent and annuity maps. general are we reasonably to with produce values this analysis. However, we would mention at least one point of caution regarding the YC fit in The data by developed the this the standards models study. quite well methodology literature. Furthermore, linking in YC NPV the the to equations and of models reported fit If this were not the case the possibility exists that well. also clearly annuity equivalents is Ibis first in those the these with at second. might multiply a point to models the of errors be wary of in any wider application of such a methodology. Accepting that such a possible problem does not seem to be present here, the timber the comparison a common unit with recreation value maps permit produced maps value in Given that most woodland recreation occurs productive woodlands produced previously. it seems reasonableto assumethat these values may be additive. We now turn our attention to the last forest value we shall consider in our analysis: carbon sequestration.

7.66

REFERENCES Achen, C.H. (1982) Interpreting and using regression. Quantitative Applications in the Social SciencesNo. 29. Sage, London. Bradley, R.I. and Knox, J.W. (1995) The prediction of irrigation demandsfor England and Wales. Proceedings of the GIS Research UK 1995 Conference,5-7 April 1995. Department of Surveying, University of Newcastle upon Tyne, Association for Geographical Information. Avery, B.W. (1980) Soil classification for England and Wales (higher categories). Technical Monograph No. 14, Soil Survey of England and Wales, Rothamsted,Harpenden,Herts. Bateman, I.J., Lovett, A. A. and Brainard, J.S. (forthcoming) Applying Environmental Economics through Geographical Information Systems: Cost-BenefitAnalysis of Converting Agricultural Land to Forestry. Manchester University Press. Blakey-Smith, J., Miller, D. and Quine. C. (1994) An appraisal of the susceptibility of forestry to windblow GIS Research UK 1994 Conference, 11-13 April 1994. Department of GIS. Proceedings the of using Geography, University of Leicester, Association of Geographical Information. Blyth, J.F. and MacLeod, D.A. (1981) Sitka spruce (Picea sitchensis) in north-cast Scotland: 1. Relationships between site factors and growth. Forestry, 54:41-62. Busby, R.JN. (1974) Forest site yield guide to upland Britain. Forestry CommissionForest Record 97, HMSO, London. Blyth, J.F. (1974) Land capability assessmentfor forestry in north-east Scotland. Unpublished PhD Thesis, University of Aberdeen. Dunteman, G.H. (1994) Principal componentsanalysis,in: Lewis-Beck, M. S. (cd) Factor Analysis and Related Techniques(International Handbooksof Quantitative Applications in the Social Sciences,volume5). Sage Publications and Toppan Publishing, London. Gale, M. F. and Anderson, A. B. (1984) High elevation planting in Galloway. Scottish Forestry, 38(l): 3-15. Gernmell, F.M. (1995) Effects of forest cover, terrain, and scale on timber volume estimation with thematic Mountain Remote Sensing in Rocky Environment, 51: 291-305. data site, a mapper Grace. J. (1977) Plant Responsesto Wind, Academic Press,London, 204pp. Hart. C.E. (1991) Practical Forestry, 3rd ed. Alan Sutton Publishing, Stroud. Jarvis, N. J. and Mullins, C.E. (1987) Modelling the effects of drought on the growth of Sitka spruce in Scotland. Forestry, 60: 13-30. Johnston, R.J. (1978) Multivariate Statistical Analysis in Geography. Longman, London. Jones, R.J.A. and Thomasson, A. J. (1985) An agroclimatic databack for England and Wales, Technical Monography No. 16, Soil Survey of England and Wales, Harpcnden. Kaiser, H. F. (1958) The varimax criterion for analytic rotation in factor analysis. Psychometrika,23: 187-200. Kaiser, H. F. (1974) Little Jiffy: Mark IV. Educational and Psychological Measurement,34: 111-117. S. (1981) Top height growth curves for Sitka sprucein Northern Ireland. Forestry, Savill, P. J. D. and Kilpatrick, 54:63-73. Lewis-Beck, M. S. (1980) Applied regression:an introduction. Quantitative Applications in the Social Sciences No. 22, Sage, London. Lewis-Beck, M. S. (1994) Editor's introduction, in: Lewis-Beck, M. S. (ed.) Factor Analysis and Related Techniques (International Handbooksof Quantitative Applications in the Social Sciences:Volume5, Sage Publications and Toppan Publishing, London. Sitka Predicting (1991) C. the D. class general of spruce on better quality land in Scotland. yield Macmillan, Forestry, 64(4):359-372. Growth Sitka Spruce. Unpublished PhD Thesis, University of Site Factors (1970) C. D. the of and Malcolm, Edinburgh. in form high (1972) Yield Studholme, W. P. C. D. and elevation stands of Sitka spruce and Malcolm, and European larch in Scotland. Scottish Forestry, 26(4):296-308. The level (1973) Yield Class of Sitka spruce. Scottish the GJ. of on effect altitude, above sea Mayhead, Forestry, 27:231-237. information Geographical (1996) systems and agricultural economics, Journal ol Aricullural A. Moxey, Economics, 47(l): 115-116. Statistics Advanced Guide, SPSS-X: McGraw-Hill, New York. (1985) J. M. Norusis, in Quantitative British forestry. Forestry, 43:45-56. some practical (1970) site assessment: applications G. page,

7.67

Rudeforth, C. C., Hartnup, R., Lea, J.W., Thompson, T. R. E. and Wright, P.S. (1984) Soils and their use in Wales, Bulletin No. H, Soil Survey of England and Wales, Harpenden. Soil Survey of England and Wales (1983) Legendfor the 1.250,000 Soil Map of England and Wales. Soil Survey of England and Wales, Rothamsted, Harpenden, Herts. Tranquillini, W. (1979) Physiological Ecology of the Alpine Treeline - Tree Eidstence at High Attitudes with Special Reference to the European Alps, Springer-Veriag, Berlin. Tyler, A. L., Macmillan, D. C. and Dutch, J. (1995) Models to predict the general yield class of Douglas fir, Japanese larch and Scots pine on better quality land in Scotland, mimeo. Macauley Land Use Research Institute. Tyler, A. L., Macmillan, D. C. and Dutch, J. (1996) Models to predict the General Yield Class of Douglas fir, Japanese larch and Scots pine on better quality land in Scotland, Forestry, 69(l): 13-24. Worrell, R. (1987a) Geographical variation in Sitka spruce productivity and its dependence on environmental factors. PhD Thesis, Department of Forestry and Natural Resources, University of Edinburgh. Worrell, R. (1987b) Predicting the productivity of Sitka spruce on upland sites in northern Britain. Forestry Commission Bulletin 72, HMSO, London. Worrell, R. and Malcolm, D. C. (1990a) Productivity of Sitka spruce in northern Britain: 1. The effects of 63(2): 105-118. Forestry, climate. and elevation Worrell, R. and Malcolm, D. C. (1990b) Productivity of Sitka spruce in northern Britain: 2. Prediction from site factors. Forestry, 63(2): 119-128.

7.68

Chapter 8:

8.1

Modelling and Valuing Carbon Sequestration in Trees, Timber Products Soils Forest and

INTRODUCTION

The global processof industrialisation which has grown so rapidly over the past two in led detectable increases in has, to the concentration of more recent years, centuries insulating greenhouse gases (GHGs).

These have coincided with elevations in global

GHG for foreseeable the to continue rising with emissions are expected temperatureswhich fC by temperature than Best that surface air future. global will rise more estimatessuggest day less from (Houghton in ZT by 2050 than the 1990 a century present between and and 1992)1. The Raper, Wigley 1992; consequencesof such climatic change are and et al., 1993; (Parry Warr Smith 1993). highly but adverse and potentially uncertain The advent of the global warming debatehas raised interest in the potential for using dioxide (Sedjo, 1989; concentrations of carbon atmospheric forestry as a way of reducing in largest 1991a), Nordhaus, the terms the 1990; absolute provides gas which Myers, insulation. In effect such papers add a new category to the timber, to global contribution benefits However, traditional of woodland; namely carbon sequestration. other recreation and is benefit straightforward. not this of assessment An initial and daunting problem concernsthe valuation of sequesteredcarbon. This literature. A debate heated the number of articles economics within has been a subject of for failing to gasp the complexity of climatic processeswhich heavily been criticised, have in detail in 6 literature Appendix defend We the some and review underlie global warming. both being Sam Fankhauser as more sophisticated and of work valuation the recent our use of brief A than models change climate preceding work. realistic based upon significantly more 8.2. Section is debate the start of at presented the of review

of carbonsequestration Our review of literaturealsoconsidersthe physicalprocesses liberation back timber products, and forest eventual within storage carbon soils, in trees and

in detail Appendix in is 6.1 literature some reviewed 'The.global warming 21tshouldbe stressedthat this can only be a marginalstopgapmeasureproviding temporaryrelief in the (1993) Nowak As in 10 emphasises, planting trees emissions. million per annurn reductions necessary wake of US during less 101o then that emissions of period. 5o sequester will years for the n,, xt

8.1

to the atmosphere,for carbon storagewithin treesis only a transitory processand total storage can only grow while the volume of timber increases. Neverthelessthe potential for expanding forest areas (heightened in the EC by surplusesof agricultural land) means that forests do before breathing space policy and technological change can addressthe root provide a vital cause of global warmine. Section 8.3 presentsa brief overview of our researchmethodology. This is applied in Section 8.4 to the modelling of carbon flux in both Sitka spruce trees and their products, beech. 8.5 The impacts Section this to analysis extends of afforestation upon soil while in Section 8.6, Section 8.7 presentsresults from the above levels considered while are carbon analyses.

Finally, Section 8.8 applies GIS techniques to the production of carbon

sequestrationpotential maps and correspondingevaluation maps.

8.2

LITERATURE

REVIEW

This section opens by considering the ongoing debate concerning the valuation of The their section then moves to consider three aspectsof and storage. carbon emissions its in liberation; the storage trees; afforestation: of carbon via post-felling sequestration carbon flux. impact carbon afforestation upon soil of and the 8.2.1

THE SHADOW PRICE OF CARBON EMISSIONS

While a number of studieshave examined the costs of fixing carbon via afforestation its benefits. For our purposesthe most interesting have few to quantify attempted relatively damage-avoided If to approach a valuation. those adopting accurate,estimates of these are

be directly incorporated by may are prices shadow which methods such within the produced framework our study. underpins wider which cost-benefit The pioneering work on the shadow price of C02 emissions is that of Nordhaus

discount 3% Using a rate he calculatessocial a very simplemodelandassuming (1991b,c). C These $7.3/tonne provoked a numberof critical'responses estimates of emitted. of costs

of suchpolicy change(whichseemsquitepossible),forestryextendstheperiod 3AItematively,in theabsence human before full the may enjoy consequences hit race the homel of global w arming which race g Of , degree the of carbonsequestration to be an alternativeto policy change. provide necessary cannot Afforestation

8.2

(Ayres and Walter, 1991;4 Daily et al., 1991; Grubb, 1992) the most perceptive of which (Cline, 1992a) highlights the simple linear structure of the underlying model implying both a constant level of C02 emissions' and constant shadow price through time.

In subsequentwork Nordhaus(1992a,b) addressesmany of thesecriticisms. His Dynamic Integrated Climate Economy (DICE) model usesoptimal economic growth analysis in combination with a climate model which feeds climate changesback into the economy as damages. Ile resulting carbon shadow prices are similar to his earlier estimates($5.3AC in 1995 rising to $10AC in 2025). However, Nordhaus' results have again been criticised by Cline (1992b) who suggeststhat the parameter values used result in an underestimation of true costs. A similar model, utilizing a more detailed economy component, is used by Peck and Teisberg (1992a,b). Their'Carbon EmissionTrajectory Assessment'(CETA) model produces from $10/tC $22/tC in in 1990 2030. to ranging of carbon the price shadow estimates of Given that the CETA model is structurally similar to DICE, the main reason explaining differences in the shadowprice estimatesproducedappearsto be discrepanciesin assumptions damages. carbon regarding important recent contributions to the shadowpricing debateare provided by the papers introduce fully These b, 1995). b, 1994a, (1993a, stochastic, a greenhouse Fankhauser of highly the non-linear and uncertain aspects of the damages model, explicitly recognising key parameters as random by is incorporated Uncertainty all modelling climate process. future emissions; The examining: atmospheric modules of consists model variableS6. damage; forcing; temperature annual costs rise; of sea-level rise radiative concentration; discounting. protection; and

(1991b, Nordhaus Walter Ayres the ironic c) estimatesas too low given that criticise that 41Lis somewhat and $5-10/ton C02($18-37/tC) (Walter Ayres, between damage and at costs emissions they assess in an earlier Paper different his Nordhaus they assumptions to to apply model produce a of In critique their subsequent 1990). However, linear Walter, 199 1). (Ayres C the $30-35/tonne the given problems of simple and darnageestimate of (Fankhauser, be 1993b, in caution treated extreme with that, must shows estimates such Nordhaus model, it first Nordhaus the model, also contains a mathematical error). of the simplicity to addition in 1990 9-14 from GtC 7.4 GtC to by 2025 (IPCC, 1992). to C02 rise are predicted sAnnual emissions linear. first-order not clearly Climate processesare bounds best (using distributions the and guessestimate) are generally assumed upper/lowcr 6Heretriangular 10% bounds lower best a unknown modest range of were the around and guess was upper where although by Fankhauser. the of ongoing research subject are -1bese assumptions used.

8.3

Fankhauser(1994b) addressesthe discounting problem in a more detailed mannerthan other shadow pricing assessmentof carbon. Considering the literature on the subject, he sets the pure rate of time preference(p) as a random variable with upper and lower bounds of 3% and 0% respectively and with a best guess (mode) value of 0.5%. Similarly the income is defined as a random variable with upper and lower bounds of 0.5 (co) elasticity of utility best 1.5 guess(mode) value of 1. T'his random variable discounting and a respectively and captures the uncertainty regarding theseparameters. Furthermore, if we recall our discussion low in 6, discount rate resulting,from such a choice of parameter Chapter discounting the of defensible as a reflection of social preference regarding the much more values seems impacts, does than the comparatively high 3% rate used in the warming global of assessment However, to allow comparability with these other studies above. other studies reviewed Fankhauseralso conducts a conventional discounting sensitivity analysis using values of p 0 and 0.03 with co=I throughout. The Fankhauser(1994b) model differs therefore from its predecessorsin at least three important aspects: it models climate feedback mechanismsin a more detailed and realistic manner, it use'sexpected (means)rather than best guess (mode) values; it employs a discount rate sensitivity analysis.

Table 8.1 contrastsresultsfrom Fankhauser's(1994b)randomvariablediscounting discussed For damage latter best C02 those the costs with previously. only a of guess model

(mode) value is reported while, emphasisingthe importanceof damagedistributions, Fankhauserreports expected(mean)valuesas well as 5% and 95% percentiles,standard deviation and skewedness. Given factors (i) to (iii) above, the discrepancy between Fankhauser's results and those of other studieS7are to be expected.

71gnoringAyres and Walter (1991) for reasonsgiven previously.

8.4

Table 8.1: The social costs of C02 emissions ($/tC): comparison across studies

Study

Measure

1991-2000 2001-2010

1

2011-2020

2021-2030

Nordhaus' (1991ab)

Best guess (mode)

7.3 (0.3-65.9)

Ayres and Walter' (1991)

Best guess (mode)

30-35

Nordhaus' (1992a)

Best guess (mode)

5.3

6.8

8.6'

10.0

Peck and Teisberg' (1992b)

Best guess (mode)

10-122

12-142

14-182

18-2V (3A-57.6)

Fankhauser (1994b)'

Expected (mean) 5th percentile 95th percentile standarddev. skewedness

20.3 6.2 45.2 14.3 2.5

22.8 7.4 52.9 16.0 2.5

25.3 8.3 58.4 17.5 2.5

27.8 9.2 64.2 19.0 2.4

Notes: Figures in brackets denote confidence intervals. I Discount rate = 3% for all studies except Fankhauser(1994b). 2 Figures measuredfrom graph as reported in Fankhauser(1993b). 3 Random variable discounting: p= (0,0.005,0.03); CO=(0.5,1,1.5).

Results from Fankhauser'sdiscount rate sensitivity analysis are given for two time be seen, using a common time preference rate of 3% the As 8.2. Table in can periods (1992a) Nordhaus Arguably (1994b) Fankhauser are quite comparable. this and of estimates (ii) (i) differences above that and are not particularly be significant. evidence taken as could

discount in fact it the that the of choice rate calculatingdamage However,moresurely reflects is long issue Global discounting importance. term is a very and warming estimates of prime Given large. Fankhauseros the this, assumptions underpinning effects are consequently defensible. more approachseem

8.5

Table 8.2: The social costs of C02 emissions ($/tC): discount rate sensitivity analysis Value ($/LC) Discountingassumption

Statistic

Random case

mean ($/tC)

p= (0,0.005,0.03)

5th percentile

1991-2000 20.3

62

2021-2030 .

27.8

9.2

0) = (0.5,1,1.5)

95th percentile standarddev. skewedness

45.2 14.3 2.5

64.2 19.0 2.4

Low discounting P0

mean ($/tC) 5th percentile 95th percentile standarddev. skewedness

48.8 27.6 80.1 15.6 0.9

62.9 34.9 104.6 22.4 1.3

High discounting p=0.03 CO=1

mean ($/tC) 5th percentile 95th percentile standarddev. skewedness

5.5 3.7 7.6 1.2 0.5

8.3 5.3 12.0 2.1 0.8

Source: Fankhauser (1994b)

In conclusion, the debate regarding the valuation of carbon emissions is still in its Nevertheless ongoing". much very the physical science underpinning and years early has the sophistication of ensuing and analysis models progressedmarkedly in recent economic Fankhauser be in The both to the appears of on cutting work edge of these respects years. from his firmest feel that the model estimates provide contemporary basis for our we and wider valuation work.

8.2.2 CARBON STORAGE IN TREES' 8.2.2.1 Calculating carbon storage Roughly 50% of the woody biomassof a tree is carbon,thereforegrowingnew trees fixes carbon over the lifetime of those trees. 77hequantity of carbon stored by a particular follows": be as calculated tree can $Seeour discussion of equity issues in Appendix 6.1. 9Sedjo et al., (1995) review die literature concerning the economics of storing carbon in trees. IOThe following description draws upon conversations during 1994 and 1995 with Robert Matthews, Forestry Commission's ResearchStation, Alice Holt Lodge, surrey. the at officer mensuration

8.6

The Forestry Commission produces yield models (Edwards and Christie 1981) quantifying cumulative timber production (in m`/ha) adjusted for:

a) species; b) growth rate (measuredas YQ c) spacingof trees; d) thinning regime. These models record the merchantablevolume of timber per hectare at varying ages from planting. Merchantable volume is defined as "Vol/ha: the overbark volume, in cubic metresper hectare, of the live trees. In conifers, all timber on the main stem which has an is included. least broadleaves, 7 In diameter the measurementlimit of cm at overbark is either to 7 cm, or to the point at which no main stem is distinguishable, whichever Christie, 1981). This (Edwards definition first" and means that merchantable comes less be than overall woody volume (which is more relevant significantly may volume for the assessmentof carbon storage),particularly in young trees. Consequently the be inflated by to the ratio of total woody volume estimate volume needs merchantable This initially be very high (technically -) and ratio will volume". to merchantable fall rapidly as the tree grows until an asymptotic equilibrium is attained. Figure 8.1 illustrates an example of such a multiplier for Sitka spruce (YC12). Broadleaved have, higher beech declining at will all ages, ratio as values to an such species 1.8-2.0. of about asymptotic equilibrium The total woody volume can now be related to the corresponding oven dry biomass level (the dry weight; DW) by referenceto the density (nominal specific gravity; SG) higher in broadleaves is SG Adger than conifers. generally the wood. and Brown of (1994) report SG = 0.34 for Sitka spruce and SG = 0.60 for beech. iv)

Thompson and Matthews (1989a) note that variance in the "Proportions of cellulose,

hemicellulosesandlignin" (ibid) betweendiffering speciesleadsto differencesin the However, is Matthews (1993)t2 biomass figures carbon. which of reports proportion beech. for both 49% Sitka just spruce and over of

"Corbyn, CrockfordandSavill (1988)givedetailsregardingthebranchwood componentof totaltreevolume. "Note that this referencerefers to the paperby GeorgeMatthews(1993), all subsequentreferencesto by Robert Matthews. to the (1993) paper refer Matthews

8.7

Figure 8.1:

Changein ratio of total/merchantablevolume with age for YC12 Sitka spruce (2 m spacing, intermediate thinning). 3.5

&

2-5

w 2.0

I. S

C3 1.0 I0

10

20

30

40

50

60

70

80

90

loo

AGE, YOM-

Source: Matthews (1991)

Calculating tree carbon storage: example 40-year old YC12 Sitka spruce, 2 m. spacing, intermediate thinning O/ha in 40) 399 (cumulative Merchantable volume production year = Total/merchantable volume ratio @ 40 years = 2.0 798 Total rný/ha volume woody = -. . Nominal specific gravity = 0.33 263 biomass Total rný = -. .

Total carbon= 0.42 * biomass= 110tC/ha The aboveexampleis basedupon the cumulativemerchantablevolumefor year40. for (after date (in both that thinnings thinning) to the and all year includes maincrop This is different to the early work of Matthews (1991) 40). This approach 25,30,35 and years 1.5 for total/merchantable ratio of ignores constant thinnings and uses a all years". all who 57 fixing carbon tC/ha at year 40. to reduces estimated Using such assumptions

in (1991) Pearce draw on this early work. reported 13NOte the estimates that

8.8

Because of uncertainties surrounding the total/merchantable volume curve (such a curve has not to date beenpublished for broadleaves)rather than attempt to calculate carbon storage at different points over the rotation, we rely principally upon the work of Matthews (1993) with respect to conifers and Dewar and Cannell (1992) with respect to broadleaves. These have distinct advantagesover earlier references in that they supply at least some information regarding the shape of carbon storage functions and incorporate up-to-date knowledge regarding sequestrationin trees. Neverthelessa considerableamount of analysis was necessaryin order to produce models which predicted acrossYC (the above references hold this factor constant) and provide the necessaryinformation for economic analysis. In order to construct such a flexible model we need to first consider the variety of factors which living carbon the wood. storage of within affect 8.2.2.2 Factors affecting tree carbon storage Physical factors affecting tree carbon storageare as follows:

(i)

factors and related class yield species management regime

These factors are now consideredin turn.

(i)

Yield classand relatedfactors

Tree carbonstorageis directlyrelatedto growthrateandso increasesover time from in Chapter 7 illustrated YC S-shaped discusses which the curves also the per as planting YC/carbon (1991) Cape Cannell YC. produce determinants and storagecurvesfor of specific Sitka spruce as illustrated in Figure 8.2. Studying Figure 8.2 we can see that, while the volume growth curve is sigmoidal, YC between line and increment mean annual (MAI) is relationship straight roughly a there 14 of carbon sequestration .

S"v'cultural'st ForestrY Principal (1993), the "t Tbompson Donald Commission'sAlice "'Conversations with is line relationship ar-ccptable. a straight that such confirtned station, Lodge Holt research

8.9

Figure 8.2:

tc/ha

200

Volume (M/ha), biomass (t/ha) and carbon sequestration(tCjha) against tree age (years) for YC 8,16 and 24 Sitka spruce

t/ha

'001

M3/ha

2-Om spacing (Intermediate thinning)

YC 24

t

400-

M AI MAI 150-

YC 16 3-3

1000300-

2-2

100-

YC8

20050050-

0

1.1 tC/ha/yr 100-

06-

0

20

60

40

L 80

:Foo Age (years)

Notes: MAI = mean annual increment (tC/ha/yr) Source: Cannell and Cape (1991)

The growth/sequestrationcurves shown in Figure 8.2 only cover the period during Clearly, is felled, growing. plantation once much of the carbon locked up which a particular in a specific rotation will be liberated back to the atmosphere via decomposition, Indeed becomes the combustion. or wastage, end usage the crucial factor manufacturing determining the rate of carbon liberation (seesubsequentdiscussions). However, if replanting begin fix to trees again carbon. will the new occurs then An interesting long term factor in tree carbon storageanalysis concernsthe possibility impacts The feedbacks. precise of global warming upon tree growth is of global warming Two difficult important: effects increasing seem to particularly C02 predict. extremely levels; and climatic change. In a review of existing literature Eamus and Jarvis (1989) report that studies have found C02 increased of to enhance rates of tree growth concentrations were found that this effect the of were Whilst disagreeing quite extent varied. of estimates not although with Cape (1991) point out that the studies reviewed were Cannell findings, and all of short such 8.10

duration (less than 12 months) and that a long term adaption process whereby growth rates is feasible. levels Nevertheless,most studies(e.g. Waggoner, physiologically to return present 1983; D'Arrigo et al., 1987) do point to some positive relationship betweenC02 and growth Heath (1995) In et al., experiments, reported that increasing concentrationlevels recent rate. levels 250 by C02 ambient of about 350 Ppm, resulted in a 23% increase in over pprn of increase for 25% beech Murray for (1995) impact and oak. the et al., examine rate growth C02 by beech 300 in Sitka of raising and ppm conjunction with varying nutrient spruce upon levels, concluding that C02 elevation may have little impact at low nutrient levels but that at high nutrient levels such C02 elevation may raise biomass weight by about 35%. Cannell and Cape (1991) examine the impact of a potential, climate change induced IC increase in UK temperatureconcluding that this will generally raise tree growth rates. However, they also stressthat "less confidence may be put in the prediction that trees already locations benefit from further (p. in 23). will and westerly southerly warming" mild growing Although a wider variety of species may become viable the authors highlight possible including damage from during dry to roots global warming arising summers. negative effects is damage increased for also noted. rain acid The potential Given these uncertainties,global warming feedbacksare not incorporated within our it likely balance On flux that such seems omissions tend to model. will carbon subsequent However, long in face term sequestration. carbon the in of underestimate small some result data of we paucity prefer to this relative and adopt uncertainty considerable of such

conservativestance. Species between One species. substantially reason for this, Carbon sequestrationrates vary higher (SG) is broadleaves specific gravity briefly the of have mentioned, already we which including Sitka spruceand for SG details species 8.3 selected Table to conifers. compared as beech.'s

in Lavcrs (1969). given are IsFurtherdetailsregardingSG for a varietyof species

8.11

Table 8.3: Specific gravity by species Species

Specific Gravity

Sitka spruce Corsican pine Birch Oak Beech

0.33 0.40 0.53 0.56 0.56

As a result of differences in SG, two speciesgrowing at the same YC may well be fixing quite different levels of carbon. Also, as differing species have differing optimal felling ages (see Chapter 6), so the S-shapedgrowth curve for living wood will return to zero in its differing time. Dewar and Cannell (1992) illustrate this at points path and restart divergence for two specieswhich are assumedto be replanted after felling to produce the illustrated functions in Figure 8.3. tree carbon storage saw-toothed Figure 8.3: Tree carbon storagefor two species 4)

150 10050O-r 0

50

100

150

zuu

idbu

300

years since planting

Oak YC4 Poplar YC12 ,

Source: Basedon CannellandDewar (1992) be length in species will difference across particularly important cycle The temporal Furthermore, felling as discounted Optimal date values. is carbon storage consider when we 6), it Chapter (see be to discount needs YC both rate and modelled as such itself a function of

8.12

" within our carbon storage analysis.

(iii)

Management regime Modem silvicultural practices have conflicting implications for tree carbon storage.

As noted in our YC model (Chapter 7) modem intensive plantations produce higher growth forest it is believed that this raises natural systems and and than generally extensive rates carbon storage: "Moving from natural forest management to plantation-based strategies increasescarbon fixation as foresters can plant trees of a type and in such a way as to maximise the rate of timber production" Thompson and Matthews (1989a, p. 19) There are some drawbacksof plantation style managementtechniques. One problem from directly forest management,particularly imply does certain emissions is that the move be lower likely than those associatedwith agricultural However, to felling. these are during less than those associatedwith the manufacturing "several land use and orders of magnitude" 1993). (Matthews, of wood products A second, more important, effect arises from crop thinning, a technique typical of (1993) Matthews 'unmanaged' compares newly planted plantations. commercially managed finding latter fix that the less plantations significantly managed woods with commercially his Thompson work (thus with earlier former do quoted above). contradicting the than carbon long thinning which of affects term carbon result a as difference primarily arises This does in being Firstly, thinning result remaining trees in while of two a ways. sequestration in lower is biomass resulting reduced a significantly stems per of the number girth, greater be brief lifetime to to tend put 8.4). Secondly, thinnings end uses with hectare (see Figure dates (to turn subsequently). we which short carbon-release

in function is turn a YC NPV Strictly function Of fact which is in of and date r. speaking felling a i6optirnal felling to be influence date benefits allowed (through impacts should storage carbon upon net of the Inonefisation that and programming this the would be necessary of only the valid with complexity However, given NpV). an extension forestry such was Given values, not undertaken. to the private opposed as respectto social be will small; error any a result in sequestration, confirmed carbon of a recent values estimated subsequently (1995). Kooten et al., analysisby van

8.13

Figure 8.4: Simulated carbon storage by thinned and unthinned (YC12) Sitka spruce +

2C

It

8 lo(

40

20

a 10

20

30

so

40

do

?o

to

to

YEAR

Source: Matthews (1992) Does this imply that all standsshould be left unmanagedand that we should abandon There YC thinning-based are two reasonswhy this is probably not models? consideration of in Firstly, the absenceof carbon storage subsidy payments, both private a wise move. incentive have EC to adjust silvicultural practice to increase carbon the no producers and by Matthews (1993) 17 is his reason provided who second extends storage .A analysis to in terms of reduced emissions where commercially produced savings the potential consider Using information for from is sources. carbon Keighley to existing substitute used wood (1983) regarding the burning efficiencies of coal and oil as opposed to spruce wood, Matthews shows that, providing the wood is burned as a direct substitute for fossil fuel, then leaving is forest fuel for to the forest preferable "harvesting the unmanaged" (p.6). from forests the both a pragmatic and correct option therefore seems Consideration of thinned theoretical standpoint. 8.23 CARBON LIBERATION

FROM WOOD PRODUCTS

begins be fixed its liberated back to the is felled to store carbon Once a tree if the is Wood This quickly fuel, C02. left quite occur may as used as to atmosphcre "Note thatthis differs from thecaseof non-markctwoodlandrecreationwherCe . Privat, operatorscanreceive corresponding duty and both has grant EC of provision a the aid. and payments subsidy

8.14

decompose (e.g. small trimmings), or used for short term purposes. The carbon liberation rates resultant from these various end-usescan vary substantially. For example, 77hompson and Matthews (1989a) compareconventionally grown YC16 Corsican pine with short rotation coppice (SRQ Poplar plantations, noting that the latter fixes significantly more carbon per annum than the former. However, because SRC is generally used as fuel, its long term is lower significantly than that of Corsican pine which is typically rate sequestration average used for more enduring products". A rigorous examination of the impact of end use upon carbon fixing is given in Thompson and Matthews (1989b). Results were obtained for a variety of species,those for Sitka spruce being graphically summarisedin Figure 8.5'. Figure 8.5: Longevity of Sitka spruce timber when put to different uses

cis

1.0

1.0

0.5

0.5

1.0

1..

0.5

PUJN=

PARTICLZDOAM

0.0

0.0

ce

0

20

60

40

80

0.0 0

20

40

60

.

80

0.5

0.5

0.0

0.0

20

40

60

80

0

20

40

60

80

60

80

OE, >n

MEO"

0

20

1.0

1.0

1.0

0

40

60

PALLrr

PAC9AGtNG

0.0

0

80

0.5

9

20

40

60

80

1.0

1.0 ýGgel

0.5

0.5

MINING

" 0.0

0.0

0.0

0

20

40

ý7-OTHER

60

80

0

20

40

60

80

20

40

YEARS FROM HARVESTING

Matthews (1989b) Thompson and Source:

important highlight (1993) Matthews (1992) consequence an I'Marland andMarland and of suchexamples: fossil fuels, further for high-carbon a fuel existing benefitwill accrue. is and substitutes net as used timber where, because in of uncertainties regarding likely analysis our an such assumpLion have We adopted substitution not fuelling systems to non-timber mean that that commitments capital assume In any conversion we effect rates. low. be very rate will for Poplar liberation (1992) curves Cannell and Oak. carbon reportproduct 19Dcwarand

8.15

Thompson and Matthews (1989a) also report mode and 95%

carbon liberation periods

for a variety of products and speciesas detailed in Table 8.4. Table 8.4:

Mode and 95% carbon liberation periods (years) for various timber products and species Years to specified carbon loss' Product

Wastoark, /fu& Pulpwood Particle board MDP3 Pallet and packaging Fencing Construction and engineering Mining Other

Sitka spruce

Corsican pine

Mode

95%

Mode

1 1 15 20 1 15 70 40 15

8 5 40 80 4 30 150 n/a

1 1 15 n/a 2 20 100 40

95%-8 5 40 n/a 5 40 200 n/a

Oak

Birch

Mode

95%

Mode

95%

2 1 15 n/a 2 40 150 40

18 5 40 n/a 5 80 300 n/a

1 1 15 n/a 1 40 5 10

6 5 40 n/a 5 80 40 n/a 20

-in

Mode = that year in which the largest annual carbon loss occurs. 95% = that year during which only 5% of carbon remains sequestered. 2 For hardwoods observations are for waste wood only (i. e. excludes bark and fuel). I MDF = medium density fibreboard.

Source: Adapted from Thompson and Matthews (1989a) Given the findings of Figure 8.5 and Table 8.4 it is clear that end use has a major influence upon plantation averagecarbon storagelevels. Indeed Matthews (1995) cites this being determinant signiflcantly storage, of overall carbon stronger than factors the major as 8.5 itemises Table for Forestry regime. management end silvicultural uses as such Comrnission timber in 1991. Thompson draws 8.5 Table final upon and Matthews (1989b) to The column of follows: longevity their as to catcgorise end uses according S= short emission times: waste/bark/fuel; pulpwood; pallet and packaging MDF; fencing; board; M= medium emission times: particle other L=

long cmission times: construction and engineering; mining

indicates MY 8.5 that Table rou g 30% of Pres nt UK Following this classifIcation 20% uses; end time to is to emission short medium term; consigned and production

8.16

approximately 50% to long term end uses". Table 8.5: Forestry Commission timber end use (1991) End use

Volume (Mimon, M, under bark)

Softwood sawn logs (mainly construction) Hardwood sawn logs (construction and furniture) Pit props (mining) Particleboard Fibreboard Paper/cardboard Other industrial wood Fuel Bark

2.9 0.6 0.02 1.2 0.02 1.1 0.2 0.2 0.9

Total. (underbark + bark)

7.1

Notes: IL=

Source:

% of total volume

41.1 8.5 0.003 17.0 0.003 15.5 2.8 2.8 11.5

95% carbon liberation (years from felling)

classt (see text)

150 300 200 40 80 5 30 5 5

L L L M M S M S S

Emission

long; M= medium; S= short emission times (see text).

Compiled from Thompson and Matthews (1989b); Cannell and Cape (1991); Forestry Commission (1992); Whiteman (1993) pers. comm.

9.2.4 CARBON FLUX IN SOILS 8.2.4.1 Determinants of soil carbon levels All soils contain a certain natural level of carbon. Ibis generally consistsof decaying (usually (SOM) less a small amount although than 5%) is held as soil matter soils organic 1988). On (Jenkinson, factors influence soils a number of uncultivated natural organisms Soil including: temperature; lignin texture; moisture; and the soil content content of the carbon lowland 1987). In (Parton the areas quantity et al., cover and type of organic plant natural is, in long dead tissue the balanced the plant to by soil as run, returned the material (Jenkinson, C02 1988). Such soils are SOM water and and release of decomposition of 14owever, drained balance. are poorly in which soils frequently and carbon therefore in upland areas) exhibit very slow decomposition rateS2'. Where (typically waterlogged (Askew formed is decomposition peat deposition 1985). et Such al., exceeds soils organic levels (although SOM limit upon average levels can be have no predetermined upper

level of (1992), the manufacturing emissions JISS13ciated by Matthews is issue, 2oAfurther considered with intensive for products high such capital as These paper and low for sawn wood. relatively are differing end usesetc.

between soil moisture-deficit and carbon (1995) relation 21Harrisonet al., report a strong negative content. (1975). Edwards See also

8.17

calculated) and consequentlymay have very high carbon contents (Adger et al., 1992). On cultivated soils a variety of additional factors may influence soil carbon levels including:

tillage regime;

crop selection;

addition of fertilizer and organic matter,

irrigation; and residue treatments2'(Parton et al., 1987). The transition from uncultivated to intensive arable land, particularly where bare fallow rotation systems are used, is commonly associatedwith very significant losses in SOM. The majority of a soils carbon is held near the surface and repeatedtillage exposesthe SOM to the atmosphereincreasing decomposition rates significantly above natural levels (Jenkinson, 1988). Tiessen et al., (1982) reports a 35% loss in carbon levels over a 70-year period as a result of switching Jenkinson (1988) into reports a similar loss over roughly 30 years for croppingý3. grassland into switched grassland established various arable crops, lossesbeing greatest an area of old being land ploughed with no crop cover regularly was sown. where The growth of intensive agriculture worldwide during the twentieth century has led in levels. The depletions soil carbon extent of these depletions has provided a to massive C02 emissions: of global major source f1soil carbon losses have been a primary anthropogenic source of carbon dioxide, second only to fossil fuel combustion in contributing to historical increasesof global carbon dioxide concentrations"

Postet aL (1990) Concern regarding the global impact of soil carbon loss has recently led to the Agency (EPA), Protection US Environmental by of the BIOME project; a the instigation, degree "the be to agroecosystems which initiative technically can examining research basis, sequester and carbon, to conserve thereby reducing the sustainable a on managed, 1991). (Barnwell et in aL, C02 the atmosphere" of accumulation 8.2.4.2 Afforestation and soil carbon been done had little on the long tenn effects of work Until recently relatively

is burned. stubble not 22Forexample, whether or losses from 35% leguminous reduced Use crops loam to 18% (Tiessenet al., 1982). of 23Clayand silt soils.

8.18

in important levels An UK24. is the carbon soil early exception upon provided afforestation by the work of Jenkinson (1971,1988) who examined two areaswhich had been arable for for being 80 before to to and allowed abandoned revert woodland some years. many years This natural afforestation resulted in very considerableincreasesin soil carbon as detailed in

Table 8.6. Table 8.6: Soil carbon increasesover an 80-year period from natural afforestation Site

Initial soil carbon level(tC/ha)

Final soil carbon level(tC/ha)

Increase over 80 years (tC/ha)

Broadbalk Geescroft

60 61

110 81

50 20

1

Source: Jenkinson (1971) Matthews (1993), in his model of Sitka spruce forest carbon budgets, combines the (1975) in Whitehead formulating Wilson (199 1) his Jenkinson al., that et and of with work of is have been intensive Here to flux assumed under soil previously predictionsI5. soil carbon initial, 30 in is This tC/ha. soil carbon of pre-afforestation, content an cropping resulting 70 200 tC/ha some to years after planting and reach a approximately assumed to rise Similar by Sampson 100 (1992) in a tC/ha. results are reported of maximum subsequent long increases 50 US term soil carbon equilibrium exhibit of about which sites study of two from afforestation. tC/ha arising Dewar conditions, and Cannell (1992) and management soil In a study using similar for hardwoods Matthews to those similar which are of curves storage report soil carbon here. is However, species significant effect there a particularly that riot (1993) suggesting

impacts flux. have soil carbon upon factors substantial very can other The major determinants of soil carbon change under afforestation are soil type and

Professor Steven McGrath Jenkinson David Experimental Professor and at Rothamsted 2-Tonversations with into lack this research area. the contemporary apparent of (1993) confirmed Station 25Afurther assumption,thatclearfelling will not reducesoil carbonprovidingreplantingoccurswithin one Edwards (1993) (1983). to Ross-Todd the However, Matthews of by work reference and with is year, also made during decline SOM (1995) first 15 yearsfollowing that the Harrison may suggests by et al., recent work 60 begins SOM taking to to slowly anything to again up years rise to return equilibrium. which after replanting (1994). Brown Seealso Adger and

8.19

From these we can estimate present carbon levels and predict long term equilibrium levels under afforestation. McGrath and Loveland (1992) estimateorganic carbon concentrations (%) for eight soil types as detailed in Table 8.7. prior usage.

Table 8.7: Soil carbon levels (%) for various soil types' Major sail groups

Lithomorphic soils Pelosols Brown soils Podzolic soils Surface-water gley soils Ground-water gley soils Man-made soils peat 9615 All soils

No. of samples

Minimum

Lower hinge

25Lh percentile

50th percentile

75th. percentile

Upper hinge

I

Maxiý 11

397 262 2116 488 1409

0.2 1.0 0.1 0.8 0.6

0.2 1.0 0.1 0.8 0.6

2.7 1.9 1.8 4.0 2.5

4.5 2.8 2.8 6.0 3.8

8.4 4.4 4.1 10.5 5.8

16.9 7.8 7.5 19.9 10.6

61 5 . 19 1 . 22 4 . 53 3 . 58 3 .

614

0.8

0.8

2.5

4.0

6.8

13.2

54.5

138 204 5666

0.2 <12.0 0.1

0.2 <12,0 0.1

2.1 28.6 2.3

3.3 46.2 3.6

5.o 50.3 5.9

8.9 65.9 113

3.10 65.9 65.9

Note: I Calculated on a dry soil basis.

Source: McGrath and Loveland (1992)

The proportional carbon estimatesgiven in Table 8.7 can now be related to absolute levels. However, this will vary according to land use. Adger et al., (1992) carbon storage levels for soil carbon a variety of soils and land uses as detailed in Table equilibrium report 8.8. The work of Adger et al., (1992) gives us further information regarding the soil carbon implications of agriculture to forestry conversionsacrossa variety of soil types. However, is incomplete and so the equilibrium levels quoted in somewhat possibilities of the matrix Adger et aL, (1992) were combinedwith information gatheredin conversationswith Professor David Jenkinson (Rothamsted),Dr Robert Sheil (University of Newcastle upon Tyne), and professor Steven McGrath (Rothamsted)to produce estimates of the full range of changes This types. soil through of various afforestation occur analysis was extended to which could becauseof varying rainfall lowland both areas which, and upland and land use, may consider

26Thr,SSLRCLandISsystemprovidesthe bestsourceof soil type datafor Enslandand Wales. Land use 5 km Furthermore, Soil PrOP'O'ty, data may be obtainedfrom the ITEXERC database. nutrient and elements data (1992) the Loveland supporLing in McGrath Lh'se although and MIIPSwas not available mapsare provided CORINE land include daLl

for this study. Alternative approaches by Cruikshank et al., (1995).

use of the

8.20

cover

base (EU, 1992) as employed

exhibit significantly different rates of soil carbon accumulation. Table 8.9 presentsresults from this analysis.

Table 8.8: Equilibrium soil carbonlevelsfor varioussoils underdifferent land uses

Land use

Additionsto soil (tC/ha)

Broadleaved woodland Coniferous woodland Mixed woodland Upland heath Upland smooth grass Upland coarse grass Blanket bog Bracken Lowland rough grass Lowland heath Crops Market garden Improved grass Rough pasture Neglected grassland Built up§ Urban open spaces§ Transport§ Mineral workings§ Derclict§

Non-harvested biomass(tC/ha)

0-5.0t 04.0t 0-4.5t 0.9 2.0 1.3 0.7 1.5 2.1 1.0 2.7 1.5 3.9 1.4 2.1 0.4 1.2 0.4 0.4 0.8

0-164 0-95 0-129 2.4 2.0 3.2 3.2 1.6 2.4 1.6 0.0 0.0 1.6 2.9 2.4 1.2 4.0 1.0 0.8 2.0

Soil type

Equilibrium soil carbon(tC/ha)

G SHP GSH Sz GSH HPS P Sz G z BO B GB HSP GS BGP GBP

170 450 250 200 180 400 1200* 200 120 80 60 50 90 350 120 10 70 70 90 120

t Excluding final harvest waste. t No upper limit. § Not in primary land use sector. Soil types (from Avery. 1980): G, stagnoglcy; S, hurnic stagnopodsol; H, hurnic glcy; P, peat; 7, podsol; 13,brown earth.

Source: Adger et al., (1992)

Inspection of Table 8.9 shows that afforestation is generally synonymous with long

levels increases in that these and increases storage carbon soil areliable to be somewhat term intensive due lowland of more in to the prevalence larger sites prior agriculturalland uses. The one clear exceptionto this trend ariseswhereplanting occurson previouslyunplanted levels high Here Of soil prior carbon the by extremely are substantially reduced soils. peat (Harrison 1995; et Davidson al., tree processes" growth and andGrieve, 1995). the planting

far (although more extreme) to the loss Of lowland SOM through intensive in is vThis process similar nature 1990). (Post et al., previously noted agriculture

8.21

Table 8.9:

Post-afforestation changes in equilibrium: soil carbon storage levels for various soils previously under grass (tC/ha): upland and lowland sites

Peat Humic gley Podzol Brown earths Hurni stagno podzol Stagnogley Notes:

Lowlandsites

Upland sites

Soil type Under grass

Under trees

Change

Under grass

Under trees

Change

1200 180-400 200-400 n/a 180-400 170-400

450 250-450 250-450 n/a 250-450 170-450

(750) 50-70 50 n/a 50-70 0-50

n/a 180-350 100-200 100-120 120-350 100-120

rkla 180-450 100-450 100-250 120-450 100-450

n/a 0-100 0-250 0-130 0-100 0-330

.

1. Use prior to afforestation is assumedto be long establishedagricultural pasture (dairy, cattle or sheep).

this type at altitude. common not soil applicable; not n/a = Bracketsindicatenegativeamounts.

Source: see text.

Given the impact of discounting upon our subsequentvaluations of carbon flows, the important. is is The function flux that consensus clearly general carbon the soil shape of in following initial high declines is flux the years planting and relatively marginal soil carbon (Cannell Milne, 1995). Robert period extended some and over equilibrium to reach smoothly 95% in the 1994) that of net roughly change soil carbon will suggests Shiel (pers. comm., (1993) Matthews Dewar Cannell (1992) Both and 200 and planting. of years within occur have Combining negative exponential which shapes. illustrate total soil carbon storagecurves both total to and information model marginal soil carbon storage us allows of these pieces

curves.

8.3

METHODOLOGY

the in amount of is quantify carbon trees,soilsand and stored to assess our objective discussed Ilis the values unit using exercise this storage previously. then value and products, initiated by flux is both fact afforestation the carbon by that complex is complicated the by non-peaty and trees soils, fiberaflon from and carbon (involving carbon sequestration occurs over and The a soils) very peaty extended period. felling and waste products, 8.22

function flux benefits (sequestration) that carbon means overall an of and costs complexity (liberation) occur at various points in time. Furthermore, temporal considerationsmean that be here. issue discounting pertinent also will the

Choice of appropriate modelling

initiated if flux by is to the are accurately assess carbon therefore we crucial methodology differing is (1995) Matthews indeed the that adoption of methodologies argues afforestation, in he factor that paper. the of estimates variety which reviews explaining the prime One apparently straightforward solution to theseproblems is to use long term average details 8.10 first Table (per total average and rotation marginal net sequestrationestimates. for in trees, and soils a variety of speciesand yield classes. products annum) carbon storage Table 8.10: Carbon storagecharacteristicsof different forest types in Britain FF_ Forest type (yield class:

Rotation length

m3ftWyr)

(years)

Sitka spruce (24) Sitka spruce (20) Sitka spruce (16) Sitka spruce (12) Sitka spruce (8) poplar (12) willow coppice 0 Nolhofagus (16) Scots pine (10) Lodgepole pine (8) Bocch woodland (6) Oak woodland (4)

47 51 55 59 65 26 8 28 71 62 92 95

Long-term average amount of carbon in trees and products (equilibrium

storage)

Long-term average amount of carbon in trees, products, litter and forest soil

(tC/ha)

(equilibrium storage) K/ha)

98 94 86 74 61 102 22 57 79 63 85 67

211 208 192 167 146 212 93 179 178 155 200 154

Net annual carbon flux over the first rotation (rate of storage) (tCAWyr)

4.4 4.1 3.6 3.0 2.4 7.3 5.9 4.6 2.7 2.5 2.4 1.8

intermediate thinning. Sitka to for to subject data stands refer spruce -rbe Note:

(1992) Cannell Dewar and Source:

While the information given in Table 8.10 provides an indication of the magnitude of for crude measures are and quantities unsuitable economic average -only storage, carbon in Table detailed 8.10 in rates the would storage in result marginal particular analysis. bene f, if This is because storage to carbon its. estimate they used are overstatement substantial is implying high in first that storage the year of planting carbon as the rotation across constant

8.23

Given discounting is it that the practice of mid-rotation. places a weight of I upon at say as in lower benefits the of planting year and progressively weights upon those received net in the average carbon of substantial use storage result quantities will received subsequently, first from Furthermore, the sequestration. carbon of as the value net present overestimatesof date felling, from liberated be the begins of net marginal sequestrationrates will to rotation in lower. be first. Indeed, in the they be as substantially the rotation will second the same not A superior approachto modelling net carbon storagein trees and products is adopted by Pearce (1991-1994) who provides the only published UK study of forestry sequestration by Thompson (1991) Matthews date. contemporary and unpublished and work to values Alice Holt Lodge Research Station estimatesmoving Commission's Forestry Matthews at the

illustrated in line 8.6. by Figure the as solid for rotations across averages totalcarbonstorage in Sitka YC16 trees Moving total storage and spruce products: 8.6: carbon average Figure 180-1

16

w cc

140TOTAL ACCUMULATED CARBON

;5 120UJ 0- loo w z z 0

0 ca cc

A

A AVERAGE CARBON

-

cl)

A

IN A FIXED FORM

80 -

Z

60-

C.)

SECOND ROTATION

FIRST C ROTATION

THIRD ROTATION

40-

20. ....... v-

0

2'0

40

60

80 Yr=ARS

(1991) Pearce Source:

8.24

100

120

............ 1ý0

--------

1ý0

- --------

180

Pearce implicitly assumes that the moving average total carbon curve gives a reasonableapproximation of the total carbon storage curve and models this as the negative exponential given in equation (8.1).

TCF

=M

(I - e-g')

where TCF

total carbon storageat t

M 9

equilibrium total carbon storage (M =F at t rate of carbon storage (per annum)

t

year (0 to o-)

For evaluation purposeswe are interestedin marginal (annual) carbon storage which is simply the differential of (8.1) as shown in equation (8.2) marginal carbon storage =9= dt

Mg. -It

(8.2)

While this representsa considerable improvement over the use of simple averages discussed above, it still has some of the samedrawbacks. Figure 8.7 illustrates our point by felling (where first to is the prior rotation upon solely all carbon stored as concentrating line dashed in total the carbon The storage curve estimated by shows panel upper trees). dashed line in (8.1) the the lower panel shows the (1991) while Pearce using equation from (8.2). This lower equation curve storage carbon marginal curve shows corresponding implies it that is marginal carbon in flaw as storage an approach such at a maximum in the A is declines realistic more by thereafter. model presented and the solid planting the year of follows in total carbon that storage the panel, upper lines which show, a sigmoidal curve

lower the of curve domed panel. implying the marginalcarbonstorage

8.25

Figure 8.7: Total and marginal tree carbon storage curves in the first rotation

Total Carbon storage

Years from planting

t

Marginal Carbon storage

t Years from planting

Pearce (1991) This study

implicit in the Pearceapproachwill be The overestimateof marginalcarbonstorage discounting tend to by emphasise importance will which the the of effects relative exacerbated be it Nevertheless that Pearce noted the should years. approachis a early these of by improvement averages the simple of and, use upon combining tree and considerable function form, straightforward a relatively employing and avoids the product carbon turn. now to we which approach complexitiesof our own Our own approachis to separate out thetree,productandsoil elementsof carbonflux flux In total carbon individually. curves case from each are estimated each which and model Where these are (as increments positive in obtained. the are caseof tree annual marginal 8.26

benefit they to are soils) monetised produce values. Conversely where non-peaty carbon and is liberated (as in forest the case of peaty soils carbon and products, within stored previously felling include waste) these produce monetary costs. which we

Comparisonof thesebenefitsand costsyields a streamof undiscountednet benefits for desired frame. discounted be to time sums any net present value produce then which can Unlike our timber yield valuations, we no longer have a simple replication of the first in liberation Ongoing the earlier stored rotations and approachof a postof carbon rotation. benefits the that mean net sequestration of successive equilibrium afforestation soil carbon NPV for Because first is higher in decline this, the tirne2. of than rotation over rotations based This the of annuity use supports equivalents upon continual subsequent rotations. do have Given felling. to that a not repeated our overall carbon pattern we after replanting in Chapter 6 Brealey function be to function, used needs revised. and the annuity storage Myers (1984) give formulae relating present values to annuities which we can rewrite as per discounting: to (8.3) exponential which applies equation NPV

Annuity (exponential)

[r

(8.3)

is by formula (8.4): discounting given equation annuity our For hyperbolic NPVP&Pdwty_ It1-I r(I +rt)]

Annuity (hyperbolic)

8.4

(8.4)

TREES IN STORAGE CARBON MODELLING

SPRUCE LIVE WOOD SITKA IN STORAGE 8.4.1 CARBON

benefits in marginal theearly carbon over-estimating of In orderto avoidtheproblems by Pearce (1991) total carbon storage curve used exponential the negative a rotation, of years S-shaped curve characteristic growth favour the of all unthinnedcrops. in of is rejected This does time to over moves land equilibrium. a new storage of course carbon use 211nessencethe overall decision in discontinued be as such a would result to the a return then precan replanting that not mean storage. level carbon of afforestation

8.27

Figure 8.4 illustrates such a function but also shows the importanceof allowing for the impact of thinning upon carbon storage. In thinned crops the total carbon storagecurve is non-linear, following the unthinned S-shapedgrowth curve up to the date of first thinning (7IDl) after

which a significantly shallowerpath is followed up to the felling date (F). Figure 8.8 illustrates these curves for various yield classesof Sitka spruce. Figure 8.8: Total carbon storagecurves for unthinned and thinned Sitka spruce (r = 5%) tc

24 ,

6

Years from planting

...........

(and prior to TD 1) = total carbon stored in unthinned live wood (uTWCS) total carbon stored in thinned live wood (tTWCS)

involved in complexities the illustrates of 8.8 some modelling tree carbon Figure Cannell From Cape (1991. and Figure 8.2) species. a single see we within can even storage , linear in across However, manner yield for a class. rises any storage given carbon that see

8.28

yield class,the impact of thinning upon carbon storageis substantial. This impact is triggered by TD1. However, both this and F are, as shown in Chapter 6, functions of yield class and discount rate (the latter being held constant in Figure 8.8). Carbon storage modelling therefore needs to reflect this complex interaction of diverse factors. We start the modelling process by considering the S-shapedcurve which is total carbon storage in unthinned live wood (uTWCS). '17hiscan be modelled as the cubic given in equation (8.5):

uTWCSiYC

":

PljYC

'i'

P2iYCt

+

Plyc

t2

+p

3 4iyCt

(8.5)

where: i

species(for Sitka spruce i= SS; for beech i= BE)

YC

(for i= 1) 4,6,8 .... 26 0,1,2 (t from plandng = years P3 >0;

A priori we would expect P,= 0; %>0; and < 0. In order to estimate YC12 were taken from Matthews (1992,1993). Thisdata Sitka for data (8.5) spruce equation P4

is based upon a superior total/merchantablevolume function than that used in Matthews (1991) based. Pearce Initial investigations confirmed are the of estimates (1991) upon which (8.5) based P, on equation gave model estimates statistical of which were that an optimal (as from different per expectations)and so this element was dropped from zero insignificantly

(8.6). is as equation final reported which model our I

UTWCSSS.

k2

99.9% =

12,, --

0.43727t+ 0.10747t'- 0.00102670 (-29.21) (28.09) (4.40) 81 n=

(8.6)

Figuresin bracketsare t-statistics.

tree of the growth patterns, the model reported predictability given Not surprisingly,

All data well. parameter fits the extremely (8.6) estimatesare very highly in equation have in coefficients and expected 0.000 cases) (p all < signsandmagnitudes. significant As classes. yield across to generalise notedin our literaturereview, We now need 8.29

Cannell. and Cape (1991) show that carbon storage varies linearly across YC. We can therefore derive a speciesspecific YC adjustmentfactor Aiyc which allows us to adjust from the YC of our baseline data (YC12) to any other Sitka spruceYC. Using the data given in Cannell and Cape (1991) we derive the Sitka spruce adjustment factor given in equation (8.7)29:

Ass,yc = 0.08333 YC

(8.7)

A generalisedfunction for uTWCSiyc for i= SS and any YC can then be derived as (8.8): equation per

uTWCSss,

Ass. UTWCSSS. = 12 yc* yc

(8.8)

These functions will continue to rise until t=F (the felling date). However, as noted, F is a complex function of both the discount rate (r) and YC. This relationship was investigated using YC/discount rate analysis of optimal felling dates reported in Chapter 6. in is detailed fit best equation (8.9): Our resultant model Fss,yc = 114.43 - 997.3 r+ 7167 r2 - 2.8657 YC + 0.05919 yC2 (3.62) (-9.21) (32.67) (-6.25) (5.79)

R2= 96.6%

or

n= 39

(8.9)

Figuresin bracketsare t-statistics.

Equation (8.9) fits the data extremely well with all parameterssignificant at p=0.001 F declines that (expressed It previously noted with better. r as shows as a decimal in

YC, in the significance (8.9) clear the (8.9)) of quadratic terms although and shows equation relationship. is straight-line simple a not that this We can now begin to move from unthinned to thinned crops. To do this we first need

the in Examination models Edwards yield Christie TDI. of given (1981) and to estimate F TD1, YC. between Table and 8.11 data for this the relationship reports clear a shows detailed by Edwards Christie Sitka (1981). models and spruce thinned relevant

2NOte that when

YC = 12 thenAss,12-=

IA

8.30

Table 8.11:

Date of first thinning (TDI) for Sitka spruce yield models (2 m spacing: no delay in thinning; r=0.05 throughout)

YC

Year of first thinning' (TD1)

Optimal felling year' (F)

Ratio (TD1/F)

6 8 10 12 14 16 18 20 22 24

33 29 26 24 22 21 20 19 18 18

68 67 64 58 54 51 50 50 49 48

0.485 0.433 0.406 0.414 0.407 0.412 0.400 0.380 0.367 0.375

Sources:

1. = Edwards and Christie (1981) 2. = from Chapter 6, this study

Inspecting table 8.11 shows that TDI falls as both F and YC increase. One simple is first TDIT function YC to the this model ratio as a relationship of capturing method of (8.10): in equation as shown RATIOTDIss, yc = 0.48149 - 0.0049061 YC (-5.27) (32.21)

(8.10)

where: RATIOTD I= Ratio of TD I to F W= 77.7%

n=9

Figures in brackets are t-statistics

While the small samplesizeusedin equation(8.10)somewhatreducesthe degreeof highly further data is individual t-statistics significant are very and, as no explanation, be for TD1 YC then can calculated any approach. given a reasonable available,this seems (8.10) in date by felling (8.11): equation as shown equation the corresponding by rnultiplying * * YQ] Fssyc (0.004906 [0.4815 TDiss,yc = -

8.31

(8.11)

As shown in figure 8.8, once thinning commences total tree carbon storage falls by function. from below Using data Matthews that predicted our uTWCS progressively (1991,1992,1993) we can measurethis proportion as the Thinning Factor (TF) detailed in

the final column of table 8.12.

Table 8.12: Thinning factor (TDF) for Sitka spruceYC12

Years after date of first thinning (t* t-TD I)

L

0 5 10 15 20 30 40 50 60

Total unthinned tree carbon storage (tC/ha) (uTWCS)

50 67 84 109 133 169 192 206 211

Total thinned tree carbon storage (tC/ha) (tTwCS)

Reduction in total potential tree carbon storage arising from thinning (tC/ha)

50 55 61 71 82 95 107 116 120

0 12 23 38 51 74 86 90 91

Thinning factor ITF-UTWCS Twcs. I

1.00 0.83 0.73 0.65 0.62 0.56 0.56 0.56 0.56

(1991,1992,1993) in Matthews data based Source: on

be by TFss log well predicted could the that investigation natural showed Statistical fitting best TFss. details (8.12) model our Equation of t-TDI. t* = t* where of (8.12)

TFss = 1.000 - 0.1158 Int* (37.90) (-13.41) R2 = 96.3%

n=8

Figures in brackets are t-statistics

TF TDO constrain before we to t= 1. (i. equal <0 e. t* Note that where

8.32

We are now able to calculate total live wood tree carbon storage for thinned stands of Sitka spruce (tTWCSssyc) as per equation (8.13).

t-MCSss, yc = uTWCSssyc * TFss

(8.13)

The function shown in equation (8.13) grows in each year from planting to felling is follow to assumed within one year and the function returns to zero after which replanting and restarts its growth path. Given that the model detailed in equations(8.13) (and subsequentlyin equation (8.21)) is discontinuous it cannot readily be differentiated. Consequentlymarginal carbon storage (8.13) (and by (8.21)) iteratively for each year in our time equation solving was calculated series and calculating the annual addition as storage". 8.4.2 CARBON

STORAGE

IN BEECH LIVE WOOD

The modelling of carbon storagein beech live wood followed the methodology used for Sitka spruce and will therefore be only briefly described. Information regarding for its is beech in than sparser widespreadconiferous cousin, so much somewhat sequestration for is based (YC4) in the Dewar estimates oak upon given analysis that our and Cannell so for beechgiven in Edwards and Christie (198 1). YC4 by the model (1992) adjusted consulting George Matthews (1993) findings the of This exercise reinforced who suggeststhat, within YC bands, carbon storage for oak and beech will be similar. Using this approach, S-shaped built for storage curve carbon uTWCS,,, the unthinned were up on use observations 4 ,. (8.14): in is equation reported in the estimated model which I 0.2414 t+0.030752 = uTWCS13E, 4

(2.17)

99.9%

n= 26

t2_0.00014252

(8.14)

(-13.24)

(13.73)

Figures in brackets are t-statistics

following felling path the growth of that restarting was not recorded as a fall in 3OCarewas taken to ensure function by the is liberation relating captured felling All to carbon waste and timber tree carbon storage. products.

8.33

As with Sitka spruce,the model of total carbon storagein unthinned beech live wood fits the data very well. All parameterestimatesare highly significant (p < 0.05 for t and p e) e have (the latter differing for 0.000 expected signs and coefficients magnitudes < and and

logically from thoseof our Sitka sprucemodel). As before we can calculate an adjustmentfactor (Al.yc) to allow comparison between YC as per equation (8.15)31. A13Eyc = 0.25 YC

(8.15)

A generalisedfunction for uTWCSi.yc for i= BE and any YC can then be derived as (8-16). per equation

UIVCSBE.

YC -=

ABE.

YC

*

UTWCSBE.

(8.16)

4

We now estimate F for beech as a function of r and YC using the data reported in Chapter 6. Our best fit model is reported as equation (8.17). FBF-YC

=

173.86 - 1901.4 r+ 8870.8 r' - 5.387 YC + 0.2500 YC' (-2.25) (11-99) (1.47) (20.78) (-18.07)

R2 = 97.8%

n= 31

(8.17)

Figures in brackets are t-statistics

Equation(8.17)fits thedatavery well andreconfirmstherelationshipsnotedregarding better All data. the significant at p<0.05 or are with estimates exception Sitka spruce the is in itself insignificant While is has this the term YC2 p=0.152. retained term which the of it because improvement in yields a slight model previous our for both comparison with fit. model

adjusted is before. CID1) Table 8.13 details the first estimated as also thinning The year of lack be in YC for As British beech the of variation seen, can for this analysis. data available. observations the of number reduces considerably

31NOte that when YC =4 then ABF,4 'ý 1*0,

8.34

Table 8.13:

YC

Date of first thinning for beech yield models (1.2 m spacing, no delay in thinning, r=0.05 throughout) Year of firýt thinning (TDI)

Ratio (TD I/F)

81 75 71 69