Environmental Cost-Benefit Analysis

ANRV357-EG33-14 ARI 15 September 2008 16:26

Environmental Cost-BenetAnalysisGiles Atkinson and Susana MouratoDepartment of Geography and Environment and Grantham Research Institute on ClimateChange and the Environment, London School of Economics and Political Science, London,WC2A 2AE, United Kingdom, email: [email protected]

Annu. Rev. Environ. Resour. 2008. 33:31744

First published online as a Review in Advance onAugust 21, 2008

The Annual Review of Environment and Resourcesis online at environ.annualreviews.org

This articles doi:10.1146/annurev.environ.33.020107.112927

Copyright c 2008 by Annual Reviews.All rights reserved

1543-5938/08/1121-0317$20.00

Key Words

environmental valuation, equity, uncertainty

AbstractEnvironmental cost-benet analysis, or CBA, refers to the economicappraisal of policies and projects that have the deliberate aim of im-proving the provision of environmental services or actions that might af-fect (sometimes adversely) the environment as an indirect consequence.Vital advances have arisen in response to the challenges that environ-mental problems and environmental policy pose for CBA. In this arti-cle, we review a number of these developments. Perhaps most notablythis includes continuing progress in techniques to value environmen-tal changes. Growing experience of these methods has resulted in, onthe one hand, ever greater sophistication in application and, on theother hand, scrutiny regarding their validity and reliability. Distribu-tional concerns have led to a renewal of interest in how appraisals mightthrow light on questions about equity as well as efciency, and there havebeen substantial new insights for discounting costs and benets in thefar-off future. Uncertainty about what is lost when environmental assetsare degraded or depleted has resulted in a number of distinct proposalsalthough precaution is the watchword in each. Just as importantly, thereis a need to understand when CBA is used in practice and why environ-mental decisions are often made in a manner apparently inconsistentwith cost-benet thinking.

317

Click here for quick links to Annual Reviews content online, including:

Other articles in this volume Top cited articles Top downloaded articles Our comprehensive search

FurtherANNUALREVIEWS

Ann

u. R

ev. E

nviro

n. R

esou

rc. 2

008.

33:3

17-3

44. D

ownl

oade

d fro

m a

rjourn

als.an

nualr

eview

s.org

by L

ondo

n Sc

hool

of E

cono

mic

s and

Pol

itica

l Sci

ence

on

10/1

5/08

. For

per

sona

l use

onl

y.


RP: revealedpreference

SP: stated preference

Contents

1. INTRODUCTION . . . . . . . . . . . . . . . . 3182. ENVIRONMENTAL

VALUATION . . . . . . . . . . . . . . . . . . . . . 3182.1. Stated Preference Methods . . . . . 3192.2. Revealed Preference Methods . . 3242.3. Benets Transfer . . . . . . . . . . . . . . . 326

3. EQUITY AND CBA . . . . . . . . . . . . . . . 3284. DISCOUNTING

AND THE FUTURE . . . . . . . . . . . . . 3305. UNCERTAINTY

AND IRREVERSIBILITY. . . . . . . . . 3326. CBA, ITS CRITICS,

AND THE REAL WORLD . . . . . . . 3337. CONCLUSIONS . . . . . . . . . . . . . . . . . . 336

1. INTRODUCTION

Although the principles of cost-benet analy-sis (CBA) remain largely the same, the prac-tice of carrying out appraisals has undergone atransformation over the past two or so decades.1

Nowhere is this more the case than for en-vironmental applications: that is, cost-benetappraisals of policies and projects that havethe deliberate aim of improving the provisionof environmental services or actions that af-fect (sometimes adversely) the environment asan indirect consequence. Current best practicein environmental cost-benet analysis reects, inits turn, an array of conceptual and empirical

1There is an extensive academic literature on CBA, some ofwhich may not use the term cost-benet analysis but insteadrefers to benet-cost analysis, policy evaluation, or projectappraisal. Numerous texts and manuals have appeared, manycovering the general eld of CBA [e.g., Sugden & Williams(1); Just et al. (2); Boardman et al. (3)]. Other contribu-tions have covered detailed procedures for estimating shadowprices (i.e., the true social value of costs of benets as opposedto the distorted market prices we might observe in reality),especially in developing countries [e.g., Ray (4); Londero(5)]. These texts remain highly relevant in environmentalcontexts where, for example, policy distortions remain sub-stantial, such as perhaps in the evaluation of biofuels. Morerecently, texts have appeared specically in the environmen-tal context [e.g., Johansson (6); Hanley & Spash (7); Pearceet al. (8)].

developments primarily in environmental eco-nomics (but also in other elds of economicinquiry such as health economics). It is the ob-jective of this article to review the most impor-tant of these recent developments.

In doing so, we identify what justly can beconsidered to be the core activity in environ-mental CBA: the estimation of monetary val-ues for environmental changes, environmentalvaluation. This has given rise to a prolifera-tion of methods that have sought to uncover,in a variety of ways, the value of environmentalimpacts (and nonmarket impacts more gener-ally). As we discuss, critical scrutiny of thesemethods has also been prominent. Althoughdevelopments in techniques of nonmarket val-uation have been a dening feature of envi-ronmental CBA, a broader array of evolvingissues has also come to the fore in extend-ing economic appraisal to contemporary en-vironmental policy challenges, perhaps mostnotably climate change and biodiversity loss.Some of these issues can be summarized asstemming from distributional concerns with re-gards to how human well-being and wealth aredistributed across generations as well as withingenerations: respectively inter- and intragener-ational equity. Other insights have emerged inresponse to reections on whether the extentand nature of the uncertainty and irreversibilitythat characterize certain environmental prob-lems might require that decision making needsto be weighted more heavily in favor of precau-tion. Finally, there are important issues of howCBA operates in the real world and in relationto its critiques. Just as important are questionsof political economy: What reasons explain whyactual policy design diverges from the optimaldesign of those policies according to CBA?

2. ENVIRONMENTALVALUATION

Two main approaches have been developedto place an economic value on nonmarketgoods and services: revealed preference (RP)and stated preference (SP) methods. Thereare many possible uses for such information.

318 Atkinson Mourato

Ann

u. R

ev. E

nviro

n. R

esou

rc. 2

008.

33:3

17-3

44. D

ownl

oade

d fro

m a

rjourn

als.an

nualr

eview

s.org

by L

ondo

n Sc

hool

of E

cono

mic

s and

Pol

itica

l Sci

ence

on

10/1

5/08

. For

per

sona

l use

onl

y.


Within the context of CBA, the objective is togenerate original (or primary) data on the totaleconomic value that the public places on envi-ronmental changes that arise as a result of somepolicy proposal.2 A complementary approach,that of benets (or value) transfer, draws onthe accumulated ndings within such valuationstudies conducted to date and so applies exist-ing monetary values (or secondary data perhapsafter some adjustment) to new policy contexts.The techniques and recent developments arediscussed in the sections that follow.

2.1. Stated Preference Methods

Stated preference (SP) methods is an umbrellaterm under which are found a range of survey-based methods that use constructed or hypo-thetical markets to elicit preferences for spec-ied policy changes [Bateman et al. (9)]. Byfar the most widely applied SP technique isthe contingent valuation (CV) method. Overthe past 40 years, this technique has been in-creasingly used by academics and policy makersowing to its exibility and ability to estimate theeconomic value of environmental goods thatare not traded in markets. Using an appro-priately designed questionnaire, a hypotheti-cal (or contingent) market is described wherethe good in question (which might be an im-provement in water quality, reduction in a riskto human health, or protection of an ecosys-tem) can be traded. This contingent marketdenes the good itself, the institutional con-text in which it would be provided, and theway it would be nanced. Respondents are then

2It is usual, in environmental economics, to divide this notionof total economic value into use and nonuse (or passive use)values. Use values relate to actual use of the good in question(e.g., a visit to a national park), planned use (a visit planned inthe future), or possible use. Actual and planned uses are fairlyobvious concepts, but possible use could also be importantbecause people may be willing to pay to maintain a good inexistence in order to preserve the option of using it in thefuture. Nonuse value refers to willingness to pay to main-tain some good in existence even though there is no actual,planned, or possible use (e.g., willingness to pay to preservea species as a bequest for future generations or simply for itsown sake).

CV: contingentvaluation

WTA: willingness toaccept

WTP: willingness topay

asked to express their maximum willingnessto pay (WTP) or minimum willingness to ac-cept (WTA) compensation for a hypotheticalchange in the level of provision of the good.The method can be used to estimate all the ben-ets (use and nonuse) associated with a changein the level of provision of a good or service. CVis rmly rooted in the economic theory of indi-vidual choice. For thorough reviews of the CVmethod see Mitchell & Carson (10), Batemanet al. (9), Champ et al. (11), and Alberini & Kahn(12). This asserts that elicited WTP or WTAvalues are monetary measures of respondentspreferences, corresponding to Hicksian welfaremeasures. That is, for a proposal that results inan environmental improvement, the change inwell-being that an individual enjoys can be mea-sured by his or her WTP for, or his or her WTAcompensation to forego, that improvement. Foran environmental deterioration, the change inwell-being is the WTP to avoid that outcomeor WTA compensation to tolerate it. Choosingbetween WTP and WTA, in a given context, ar-guably remains a contentious issue, although inlarge part the choice is determined by whetherpeople affected by a policy have a property rightto the current or the new level of provisionof the good in question. In addition, CV (andSP approaches more generally) combines theseeconomic precepts with insights from psychol-ogy and survey research.

The CV method has been extensively ap-plied in both developed and developing coun-tries and the range of environmental issuesaddressed is wide: water quality, outdoorrecreation, species preservation, forest protec-tion, air quality, visibility, waste management,sanitation improvements, biodiversity, healthimpacts, natural resource damage, and envi-ronmental risk reductions to list but a few.Carsons (13) bibliography contains over 5500published and unpublished CV studies under-taken in just under 100 countries. Much of theimpetus for this body of work was the con-clusions of the special panel appointed by theU.S. National Oceanic and Atmospheric Ad-ministration (NOAA) in 1993 [Arrow et al. (14)]following the Exxon Valdez oil spill in Alaska in

www.annualreviews.org Environmental Cost-Benefit Analysis 319

Ann

u. R

ev. E

nviro

n. R

esou

rc. 2

008.

33:3

17-3

44. D

ownl

oade

d fro

m a

rjourn

als.an

nualr

eview

s.org

by L

ondo

n Sc

hool

of E

cono

mic

s and

Pol

itica

l Sci

ence

on

10/1

5/08

. For

per

sona

l use

onl

y.


CE: choiceexperiment

1989. The panel concluded that, subject to anumber of recommendations, CV studies couldproduce estimates reliable enough to be usedin a (U.S.) judicial process of natural resourcedamage assessment. But very few CV studieshave actually been associated with litigation,as resource equivalency methods (where dam-aged resources are replaced like-for-like per-haps on the basis of physical criteria) becamethe favored approach for damage assessmentsafter 1996 [Smith (15)]. Rather, in the past15 years, CV has mostly been applied to theevaluation of projects and policies and has beenused in over 50 countries by a range of govern-ment agencies and international organizations[Carson (16)].

The rise in popularity of CV has been ac-companied by a very active debate and crit-ical assessment of the merits and limitationsof the technique and its underlying concep-tual framework, mostly within the economicsprofession [e.g., Hausman (17), Diamond &Hausman (18), Portney (19), Hanemann (20),Boyle & Bergstrom (21), Carson et al. (22)]but also beyond [e.g., Fischoff & Furby (23),Schkade & Payne (24), Sagoff (25)]. In fact,according to V. Kerry Smith, Contingent valu-ation has prompted the most serious investiga-tion of individual preferences that has ever beenundertaken in economics (15, p. 46). Some cri-tiques are philosophical in nature and, as such,discuss concepts of value and the appropriate-ness of relying on private economic value sys-tems to inform public policy [see Haddad &Howarth (26)]. But the largest body of liter-ature that has emerged from the CV debatehas focused on technical aspects, constructingrigorous tests of robustness across a variety ofpolicy contexts and investigating and correctingfor the presence of bias or theoretical anoma-lies [Bateman et al. (9), Mitchell & Carson (10);Bateman & Willis (27)].

Nevertheless, the controversy is unlikely todisappear as many critics appear to hold CVup to a criterion of perfection [Boyle (28)].Clearly, this is unrealistic not least becauseperfection does not characterize any empiri-cal methodology or, indeed, actual market deci-

sion. Moreover, taken as a wholeand notwith-standing notable caveatsempirical ndings sofar largely support the validity and reliability ofCV estimatesfrom well-executed studiesofthe value of nonmarket goods. The elaboratemethodology that has earned its place in theenvironmental economists tool-kit today is theculmination of this substantial body of research.That said, a healthy dose of skepticism is im-portant in the application, use, and interpreta-tion of any empirical methodology such as CV[Boyle & Bergstrom (21)]. Hence, the intensescrutiny to which CV has been subjected doespoint to the need for practitioners to follow,in some way, guidelines for best practice. Al-though the NOAA guidelines continue to bea focal point, there are a number of more re-cent guidelines or state-of-art reference pointsuseful to practitioners and to those who are fo-cused on the use of CV in policy evaluation [e.g.,Bateman et al. (9), Champ et al. (11), Alberini& Kahn (12)].

2.1.1. Recent developments

Choice experiments. Choice modeling refersto a family of survey-based methodologies,which has been used extensively in the market-ing and transport literatures to model prefer-ences for bundles of characteristics of goods andto isolate the value of individual product charac-teristics typically supplied in combination withone another. The conceptual microeconomicframework for these analyses lies in Lancasters(29) work, which assumes that the well-beingconsumers derive from goods can be decom-posed into its composing characteristics. Re-cent extensive reviews can be found in Batemanet al. (9), Louviere et al. (30), Bennett & Blamey(31), and Hensher et al. (32). Curiously, theredoes not seem to be a uniform use of termi-nology on these multiattribute valuation tech-niques in the literature. Contingent ranking,contingent choice, conjoint analysis, and choiceexperiments (CEs) are all terms that have beenused to describe similar SP data where choices,ranks, or matches between alternatives areinvolved. Perhaps the most widely used


Ann

u. R

ev. E

nviro

n. R

esou

rc. 2

008.

33:3

17-3

44. D

ownl

oade

d fro

m a

rjourn

als.an

nualr

eview

s.org

by L

ondo

n Sc

hool

of E

cono

mic

s and

Pol

itica

l Sci

ence

on

10/1

5/08

. For

per

sona

l use

onl

y.


variant in environmental economics is the CEtechnique [Bateman et al. (9)]. In a CE sur-vey, respondents are required to choose theirmost preferred out of a set of alternative pol-icy options. For example, in the case of apolicy to improve the quality of some waterbody, such as a river, the attributes might bea boost to the rivers ecology (perhaps as in-dicated by fewer sh deaths), decreased healthrisks to those who are exposed to the watersuch as swimmers and rowers, and increasedvisual amenity. Each alternative choice in aCE is characterized by a number of such at-tributes, which are offered at different levelsacross options. A price or cost variable is typ-ically one of the attributes, and WTP can beindirectly inferred from the choices made. CEcan be seen as a generalization of the dichoto-mous choice CV approach3 [Carson (16)], shar-ing a common theoretical framework as wellas a common basis in terms of econometricanalysis.

CE boasts a number of attractions [Hanleyet al. (33)], but primarily it is the ability to dealwith situations where changes are multidimen-sional, and where trade-offs between these di-mensions are of particular interest, that makesthe strongest case for its inclusion as a tech-nique of environmental valuation. This comesat a cost, however, most notably the cognitivedifculty associated with complex choices be-tween bundles with many attributes and levels.Although task complexity may lead to learning,it may also result in, for example, fatigue ef-fects and the use of heuristics (rules of thumb)to make choices. This, in turn, may increase thevariance of values estimated as well as the occur-rence of random errors in choices and inconsis-tent or irrational responses [Foster & Mourato(34)]. Moreover, CEs are typically much morecomplex to design and to analyze statisti-cally than CV experiments [Bateman et al. (9),Hanley et al. (33)]. And of course, as an SPmethod, CE (and its variants) faces many of the

3This referendum-style method asks respondents whetherthey would be willing to pay or not some specied amount.

criticisms and problems raised with CV. But un-like CV, the CE format has received relativelylittle testing with respect to its vulnerability toempirical anomalies in environmental applica-tions. As a result, a wide range of potential ben-ets suggested for CE methods (vis-a-vis CV)remains little more than informed speculation[Hanley et al. (33)].

Entreaties and hypothetical bias. Afterdecades of theoretical debate, methodologicaladvances and empirical applications of theCV method, the central concerns remainarguably the same: the hypothetical nature ofthe contingent market and the absence of realeconomic commitments (ask a hypotheticalquestion, get a hypothetical answer). So-calledhypothetical biasarising, for example, whenpeoples stated WTP differs from true WTPowing to the hypothetical nature of the goodhas been the subject of much scrutiny. Theconventional wisdom is that individuals arelikely to overstate their economic valuation of agood in a hypothetical context [e.g., Neill et al.(181); although see also Dickie et al. (35)]. Theevidence further suggests that hypotheticalbias (a) increases with the size of elicited values(with the implication that for small hypotheti-cal values common in CV studies, hypotheticalbias may not be a major problem) [Murphy etal. (36)]; (b) is lower for WTP than for WTAformats [List & Gallet (37)]; (c) is higher in thepresence of uncertainty [Champ et al. (38)],unfamiliar policy changes, and less knowledge-able respondents [e.g., List & Gallet (37)]; and(d ) is higher for voluntary contributions (thancompulsory payment mechanisms such astaxes) perhaps because of the free-riding asso-ciated with actual contributions [Murphy et al.(39)].

In recent years, a growing number of stud-ies have begun to explore the use of entreatiessuch as cheap talk to address the problemof hypothetical bias [e.g., Murphy et al. (36),Cummings & Taylor (40), List (41), Poeet al. (42), Lusk (43)]. The term cheap talk(CT) arises from its use in the information, bar-gaining, and game theory literatures where it


Ann

u. R

ev. E

nviro

n. R

esou

rc. 2

008.

33:3

17-3

44. D

ownl

oade

d fro

m a

rjourn

als.an

nualr

eview

s.org

by L

ondo

n Sc

hool

of E

cono

mic

s and

Pol

itica

l Sci

ence

on

10/1

5/08

. For

per

sona

l use

onl

y.


typically refers to the cost-less transmissionof information and signals [Cummings &Taylor (40)]. In the valuation literature, it usu-ally refers to a script, typically quite lengthy,which is added to the valuation scenario in orderto directly draw respondents attention to theproblem of misstating true values as a result ofthe hypothetical setting. Instead, respondentsare asked to focus their responses as if they werein a real-life setting, hence the entreaty natureof this added text. In the rst published study totest CT, Cummings & Taylor (40), in a labora-tory setting, found that a long eight-paragraphscript was effective at reducing hypothetical biasfor three different public goods, with responsesunder the CT treatment being indistinguish-able from responses involving actual payments.Subsequent studies of CT show a number ofpatterns emerging [see, for example, Murphyet al. (39), List (41), Poe et al. (42), Lusk (43)]:(a) despite mixed evidence, on the whole CTappears to be effective at lowering stated val-ues; (b) shorter scripts tend to be less effective;(c) inexperienced consumers are more receptiveto CT scripts than experienced or knowledge-able respondents; (d ) CT appears to be moreeffective at higher payment levels where hy-pothetical bias may be more pronounced; andlastly (e) at least one study, which looked at thepotential for CT to result in overcorrections,suggested no evidence of this effect. Despite itsrelative success at reducing hypothetical bias,the potential of entreaties for dealing with othertypes of CV bias remains unexplored. One ex-ception is a recent study by Atkinson et al. (44),which used an entreaty to address the prob-lem of protest responses in two CV experimentswith encouraging results. More work is neededin this area to assess the usefulness of entreatiesin correcting for other types of bias.

Deliberative group valuation. Scrutiny, em-anating largely from outside the environ-mental economics literature, has evolved intoproposals for more communal and deliberativeapproaches to economic valuation. The ratio-nale is that this may be able to capture and quan-tify collective values that are additional to in-

dividual or household values, typically elicitedin CV or CE, and/or may help to minimizeproblems such as some protest behaviors [Had-dad & Howarth (26)]. Sagoff (45) was one ofthe earlier proponents of the idea of delibera-tive, discursive, jury-like approaches as a wayto clarify public values by group consensus,whereas Prato (46) has proposed community-based collaborative decision-making processes.One practical effect of such proposals has beendeliberative valuation workshops or marketstalls [e.g., Macmillan et al. (47), Alvarez-Farizo et al. (48)] where participants are giventhe opportunity to discuss the individual or col-lective value of the proposed change in a groupcontext. Current evidence on deliberative andgroup approaches to valuation is limited, withonly a small number of studies and relativelysmall sample sizes. For example, Macmillanet al. (47) had 52 people in their workshops,and Alvarez-Farizo et al. (48) had 24 peo-ple. Proponents hope, however, that these ap-proaches might be used in applications involv-ing poorly informed consumers as well as unfa-miliar or complex changes such as biodiversityloss [Macmillan et al. (47)]. Investigating suchclaims through wide-scale testing therefore is arich topic for further research.

Valuing health risks. Environmental policy af-fects human health in a number of ways. First,by reducing environmental risks to lives, itmay save lives, i.e., reduce premature mortality.Second, it may improve the health of those liv-ing with a disease, such as a respiratory illness,and so result in a morbidity benet. Valuingthese mortality and morbidity effects in mone-tary terms can provide extremely useful infor-mation for policy. Past evidence indicates thatthe benets of reducing human health effects(mortality and/or morbidity) often exceed thecosts of pollution control by considerable mar-gins. Over the last 30 years, SP studies (togetherwith RP methods) have been used extensively tocalculate both individual WTP to secure reduc-tions in the risk of death arising from a policyand WTP to avoid particular health outcomes.


Ann

u. R

ev. E

nviro

n. R

esou

rc. 2

008.

33:3

17-3

44. D

ownl

oade

d fro

m a

rjourn

als.an

nualr

eview

s.org

by L

ondo

n Sc

hool

of E

cono

mic

s and

Pol

itica

l Sci

ence

on

10/1

5/08

. For

per

sona

l use

onl

y.


At this time, substantial progress has been madeparticularly in the valuation of mortality risks.

For convenience, WTP for mortality riskreductions is normally expressed in terms ofthe value of statistical life (VOSL). This im-plies dividing the WTP for a given risk reduc-tion by that risk reduction to obtain the VOSL[Pearce et al. (8), Bolt et al. (49), Krupnick (50)].Various countries adopt single (or ranges of )values for the VOSL and then use them in pol-icy appraisal. The U.S. Environmental Protec-tion Agency (EPA), for example, has used formany years a VOSL range of $0.6 million to$13.5 million, with an average of $4.8 million(1990 US$) on the basis of an assessment of theexisting U.S. literature [Robinson (51)]. How-ever, although there is a very large body of re-search on health values for North America andEurope, there is a dearth of evidence for devel-oping countries in general.

Notwithstanding this apparent lack of geo-graphical breadth in the empirical record, con-siderable strides have been made in recent yearsin terms of clarifying both the meaning andsize of the VOSL [see Pearce et al. (8) for anoverview]. Recent research, for example, hasshown that the age of the respondent who isvaluing the risk matters [see Krupnick (52) fora review]. Although age may or may not be rel-evant in valuing risks that are immediate (e.g.,life is threatened in the here and now), in fact,the literature is ambiguous on this; age has beenfound to affect the valuation of future (or la-tent) mortality risks. This is an important nd-ing given that environmental contexts are asso-ciated with both immediate and future healthrisks. That is, the value of a reduction in therisk of death can be very different according towhether a policy affects mostly older individualsalready suffering from some underlying healthcondition (and who thus face a very present riskof dying as a result of exposure to say poor airquality) or younger individuals who might oth-erwise have enjoyed a normal life expectancy (inthe absence of cumulative exposure to poor airquality). Thus, the standard practice of apply-ing the VOSL to value all reductions in mor-tality risk without regard to the age of those

VOSL: value ofstatistical life

who benet may result in erroneous value es-timates. In general terms, a policy that lowersthe general level of exposure to air pollutionshould be evaluated in terms of the (lower thanimmediate VOSL) valuations associated withyounger peoples valuations of future risks plusolder persons valuation of that risk as an im-mediate prospect.

An alternative approach that takes into ac-count the age of persons saved by a particu-lar policy and that may capture the shorter lifeexpectancy phenomenon is the value of a sta-tistical life-year (VSLY) approach. The VSLYis calculated by dividing the value of a statis-tical life by the discounted remaining life ex-pectancy, thereby converting VOSL estimatesinto a value per life-year saved. VSLY can thenbe multiplied by life-years saved, i.e., the re-maining life expectancy, to value the statisti-cal lives of persons of different ages. However,the VSLY approach still assumes that the valueper life-year saved is independent of age and,in addition, is sensitive to the rate used to dis-count the value of future life-years [Krupnick(52)]. There also remains debate about whethervaluing life expectancy changes is a superior ap-proach to valuing reductions in the risk of death.

A rather distinct age-related issue is thatsome environmental risks fall disproportion-ately on the very young. The calculus of WTPseems to break down here because childrenmay have no income to allocate between goods,including risk reduction. They may be ill in-formed about or be unaware of risks and maybe too young to articulate preferences. One im-plication is that adults (or parents) valuationsof the risks on behalf of children need to be es-timated. The literature on which to base suchjudgments is only now coming into existence(53). Preliminary ndings suggest that the re-sulting values of WTP may be higher for adultsvaluing on behalf of children than they are foradults speaking on behalf of themselves.

Valuing ecosystem services. Ecosystem servicesrefer to the wide range of benets that peoplederive from the multitude of resources and pro-cesses that are supplied by natural ecosystems


Ann

u. R

ev. E

nviro

n. R

esou

rc. 2

008.

33:3

17-3

44. D

ownl

oade

d fro

m a

rjourn

als.an

nualr

eview

s.org

by L

ondo

n Sc

hool

of E

cono

mic

s and

Pol

itica

l Sci

ence

on

10/1

5/08

. For

per

sona

l use

onl

y.


(54, 55). The Millennium Ecosystem Assess-ment (55), the most comprehensive survey todate of the state of the planet, indicates thatnearly two thirds of the services provided bynature are in rapid decline. An evolving issue,therefore, is the valuation of ecosystem services[e.g., Daily (54)]. In the case of forests, for ex-ample, advances have been made in measuringthe economic values associated with timber andnontimber products, carbon sequestration andstorage, recreation, and watershed regulation.Progress has been more limited in estimatingthe nonuse values of forests, and a lively de-bate surrounds the value of genetic material inforests for pharmaceutical research [Pearce &Pearce (56)]. Yet, biodiversity and ecosystemservices are perhaps among the most complexenvironmental concepts to dene, measure, andvalue, and there remain major methodologicalissues to be addressed [Daily et al. (57)].

Efforts to estimate the total economic valueof ecosystem change need to be distinguishedfrom efforts to value all ecosystems. It is notsensible to speak of the total value of a typeof ecosystem and even less sensible to speak ofthe total value of all ecosystems. Unfortunately,some of the recent literature on ecosystem valu-ation claims to do just this [e.g., Costanza et al.(58) and, for a critique, Bockstael et al. (59)].Nor is it clear that bottom-up approacheswhereby each type of service is valued separatelyand then the values are added to get some ideaof the total economic value of the ecosystemare capturing the whole value of the ecosystem.Put another way, the value of the system as awhole may be more than the value of the sumof its parts perhaps because of complex ecolog-ical interactions [Arrow et al. (60)]. This is notjust a problem for those approaches that seek toevaluate ecosystems as a whole. For example, asmall economic value for any one service mightsuggest it could be dispensed with, yet its re-moval could reverberate on the other servicesthrough complex changes within the ecosystem.Thus, the bottom-up valuation procedure couldbe misleading as a guide to the social value ofmarginal changes in the provision of ecosystemsas well.

It follows that ecosystem valuation is not astraightforward exercise, and it seems fair to saythat the literature has progressed only a limiteddistance in tackling these issues [Pearce et al.(8)]. Therefore, an important way to extendSP and other valuation methodologies is in de-veloping their suitability to aid decision mak-ing in complex multidimensional areas, suchas in managing the ows of ecosystem services[Alberini & Kahn (12)]. Nevertheless, a consid-erable challenge is presented in translating suchvalues (suited as these are to valuing relativelysmall changes in ecosystem provision) to thetypes of large-scale change that are argued tocharacterize current threats to ecosystems andbiodiversity.

2.2. Revealed Preference Methods

RP methods look at surrogate markets, thatis, analyze or infer preferences for nonmarketgoods as implied by past behavior in an asso-ciated market. These methods seek to quan-tify the market footprint of nonmarket changes.The principal advantage of RP over SP tech-niques is the fact that they are based on ac-tual behavior and hence typically enjoy highercredibility among policy makers. Inevitably, RPmethods have limitations. The rst is the in-ability to estimate nonuse values, as they arebased on market footprints of some form ofuse-related behavior. The second is their in-ability to estimate values for levels of qualitythat have not been experienced and revealedby the market. Nor have the assumptions thatunderpin RP been subject to anything like thescrutiny that has accompanied developments insay CV approaches [Bishop (61)]. The two mainRP techniques are the travel cost method andthe hedonic price method; both are discussedin the following sections.

2.2.1. Travel cost method. The travel cost(TC) method has been used to value spatialnonmarket goods, particularly outdoor loca-tions used for recreational purposes (e.g., parks,woodland, beaches, lakes). Typically, the recre-ational area is an unpriced good. However,


Ann

u. R

ev. E

nviro

n. R

esou

rc. 2

008.

33:3

17-3

44. D

ownl

oade

d fro

m a

rjourn

als.an

nualr

eview

s.org

by L

ondo

n Sc

hool

of E

cono

mic

s and

Pol

itica

l Sci

ence

on

10/1

5/08

. For

per

sona

l use

onl

y.


the basis of the TC approach is the recogni-tion that individuals produce recreational ex-periences through the input of a number offactors, which may, in some way, commandprices. Among these factors are the recreationalarea itself, travel to and from the area, andin some cases, staying overnight at a location.Such information is usually collected throughsurveys carried out at the recreational site andthrough secondary data, although complica-tions abound, such as the treatment of multiplepurpose trips and the estimation of the valueof time. Overviews of the TC method can befound in Habb & McConnell (62) and Parsons(63). Recent treatments using this techniquehave looked at multiple-site TC and, in doingso, analyze visitors choices between a numberof substitute sites, which differ both in location(i.e., some have higher access costs) and sitequalities, for any particular recreational trip.By observing how different visitors choose be-tween sites with different qualities and differentcosts of access, it is possible to use econometrictechniques to estimate how each of the qualityvariables and the cost variable contribute to theutility of a visit.

An example of an application of this tech-nique can be found in Day (64) where amultiple-site TC model was used to estimatethe recreation demand of local residents for fourrenowned games reservesHluhluwe, Um-folozi, Mkuzi, and Italain KwaZulu-Natal,South Africa. The costs of traveling to the dif-ferent game reserves vary for different house-holds living at different distances from the sites.Recreational trips to the game reserves werecharacterized by three dimensions of choice:which reserve to visit (each of the four reservesprovides differing opportunities for viewinggame), how long to stay (short versus long stay),and which type of accommodation to use (ba-sic hut, luxury chalet). Hence, an option wasdened as the choice of a particular type of ac-commodation in a particular reserve for a par-ticular length of time. The three dimensionsof choice facing visitors were modeled using athree-level nested logit model. Data on a ran-dom sample of 1000 visitors were derived from

HP: hedonic price

the KwaZulu-Natal Parks Boards reservationdatabase (e.g., trip duration, accommodationtype, postal code), and accurate measurementsof travel costs and travel times were sought us-ing detailed mapping techniques.

2.2.2. Hedonic price method. The startingpoint for the hedonic price (HP) method is theobservation that the price of a large numberof market goods is a function of a bundle ofcharacteristics. For instance, the price of a caris likely to reect its fuel efciency, safety, andreliability. The HP method uses statistical oreconometric techniques to isolate the implicitprice of each of these characteristics. Recentreviews of the hedonic approach can be foundin Habb & McConnell (62) and Taylor (65).Two types of markets are of particular interestfor the HP method: (a) property markets and(b) labor markets. In the case of the latter, theHP method has also been used to estimate thevalue of avoiding risk of death or injury bylooking for price differentials between wagesin jobs with different exposures to physical risk[Taylor (65), Krupnick (50)]. In the former,hedonic studies of the property market there-fore have been used to identify the value ofnonmarket goods (or bads) affecting houseprices, such as road trafc, aircraft noise, airpollution, water quality, proximity to landllsites, and planning restrictions on open spacesin and around urban areas. The merits of suchvalues will depend on a number of factors,including the existence of a competitiveproperty market and econometric problems.For example, a possible difculty belonging tothe latter problem is multicollinearity. As anillustration, if local air quality and proximityto good schools are highly correlated, then itbecomes all the more difcult to separate outthe specic inuence of each. This HP processitself involves collecting large amounts of dataon prices and characteristics of properties inan area, and applying statistical techniquesto estimate a hedonic price function, re-lating each characteristic of interest to thehouse price (such as those relating to theproperty itself, local environmental quality,


Ann

u. R

ev. E

nviro

n. R

esou

rc. 2

008.

33:3

17-3

44. D

ownl

oade

d fro

m a

rjourn

als.an

nualr

eview

s.org

by L

ondo

n Sc

hool

of E

cono

mic

s and

Pol

itica

l Sci

ence

on

10/1

5/08

. For

per

sona

l use

onl

y.


proximity to local amenities, and publicservices). In the U.K. case, WTP values forreductions in transport-related noiseestimated from a hedonic studyare now usedofcially in road and rail appraisal.

2.2.3. Recent developments

Use of geographical information systems. Avery promising development in the RP litera-ture is the increasing use of geocoded (spatiallyreferenced) data and geographical informationsystems (GISs). Bateman et al. (66) outline thepotential benets of this approach in the con-text of both recreational demand research andhedonic analysis. A GIS offers one means ofobtaining considerable improvements in recre-ational demand modeling via greater accuracyin calculating travel times and description ofavailable substitutes to a particular recreationalresource. To the extent that data availabilitypermits, a GIS presents a way of routinely andcomprehensively tackling such geographical is-sues through detailed mapping and so on [seeDay (64)]. Given the intrinsic spatial relation-ships between properties, a GIS can also be veryusefully applied to improve the estimation ofindependent variables in the HP model (suchas distance between properties and particularlocations or amenities of interest). Moreover,geocoded property value datasets and GIS al-low the use of spatial econometric methods inHP studies [Taylor (65)]. Day et al. (67) illus-trate a recent example of the application of thesetechniques in the context of the valuation oftransport noise externalities using the hedonicapproach.

Combining revealed and stated preferencedata. There is a growing realization that RPand SP information is highly complementaryand can be used in joint estimation of prefer-ences [Louviere et al. (30), Habb & McConnell(62), Cameron (68), Ben-Akiva & Morikawa(69)]. After adequate scaling, RP and SP data re-lating to peoples discrete choices can be pooled(stacked) to allow joint estimation of the pa-rameters associated with the attributes. Swait &

Louviere (70) developed a procedure to accountfor the relative variance of each type of dataand allow joint estimation under a single prefer-ence structure. Combining RP and SP data en-hances the unique strengths of these respectivedata while minimizing their limitations. That is,the approach takes advantage of efciency gainsfrom estimating the same parameters with in-formation from different choices. Examples ofthis approach in the environmental eld includeAdamowicz et al. (71) and Kling (72).

2.3. Benefits Transfer

Although advances in methods that seek togenerate primary data on the value of environ-mental goods and services have been a strik-ing feature of modern CBA, routine use ofvaluation arguably will rely just as much on us-ing secondary data, i.e., the results of existingstudies but applied to new (yet related) policysettingsso-called benets (or value) transfer.This involves taking a unit value of a nonmarketgood estimated in an original or primary studyand using this estimate (perhaps after some ad-justment) to value benets that arise when anew policy is to be implemented. Benets trans-fer is itself the subject of a rapidly growing lit-erature [see, for example, Boyle & Bergstrom(73), Desvousges et al. (74), Navrud & Ready(75)] and, if a valid procedure, has the poten-tial to reduce the need for costly and time-consuming original studies. Indeed, the holygrail of benets transfer is the consolidation oforiginal data on nonmarket values in transferdatabases where values can be taken off theshelf and applied to new policies and projectsas needed [such as the Web-based Environ-mental Valuation Reference Inventory (EVRI),http://www.evri.ca].

This raises considerable challenges, not leastensuring there is an abundance of good qual-ity studies to populate these databases acrossthe array of environmental changes that areof interest to decision makers. Just as impor-tantly, the validity of benets transfer remainsopen to scrutiny. Indeed, a number of contribu-tions have sought to test the accuracy of benet


Ann

u. R

ev. E

nviro

n. R

esou

rc. 2

008.

33:3

17-3

44. D

ownl

oade

d fro

m a

rjourn

als.an

nualr

eview

s.org

by L

ondo

n Sc

hool

of E

cono

mic

s and

Pol

itica

l Sci

ence

on

10/1

5/08

. For

per

sona

l use

onl

y.


transfer exercises. Ready et al. (76), for example,undertook identical CV surveys in ve Euro-pean countries to elicit WTP to avoid healtheffects thought to be associated with air pol-lution. A test of the validity of benets trans-fer was permitted by a comparison of whetherthe WTP values across these countries, forsome identical health outcome, were similar.Similarity is judged using specic (statisticalor other) criteria. Put simply, such tests indi-cate whether transferring a WTP valuein thiscase from one country to anotheris justiable(and how much error is entailed in that trans-fer). Reviews by Brouwer (77) and Rosenberger& Loomis (78) summarize the ndings of anumber of these types of tests for recreationalresources, water quality improvements, andlandscape amenities. Brouwer & Bateman (79)investigate the temporal reliability of transfervalues; that is, at some point, values taken fromolder vintages of original studies are likely tobecome too unreliable to transfer to contempo-rary policy settings. Distilling an overall mes-sage from these tests, however, is not straight-forward. In some cases, transfer error ranges aresmall, and in other cases, these ranges are ex-tremely large indeed. In the case of health valuetransfers across national boundaries, for exam-ple, Barton & Mourato (80) nd large transfererrors for two comparable CV surveys elicit-ing WTP to avoid ill-health symptoms associ-ated with exposure to polluted coastal water inPortugal and Costa Rica. By contrast, Alberiniet al. (81) nd that transferring health valuesfrom the United States to Taiwan provides areasonable approximation of the ndings of anoriginal study conducted in Taiwan.

The evidence to date suggests that there isa need for still more research to secure a bet-ter understanding of when transfers work andwhen they do not as well as to develop meth-ods that might lead to improved transfer accu-racy. Intuitively, more sophisticated transfer ap-proaches might be the answer. Such techniquesseek to control for as many important differ-ences as is possible between an original studysite (or sites) and a new policy site, e.g., relatingto the attributes of the environmental goods as

well as the socioeconomic characteristics of thepopulations at each site. The most ambitiousof these approaches uses meta-analysis, a statis-tical analysis of the results of a (typically) largegroup of WTP studies. Studies have taken placefocusing on urban pollution, recreation, theecological functions of wetlands, values of sta-tistical life, noise, and congestion [see, for exam-ple, chapters in Navrud & Ready (75)]. Anotherdevelopment is the use of GIS in, for exam-ple, the context of recreational value transfers[Lovett et al. (82)]. In practice, however, a num-ber of studies suggest that simpler approaches,e.g., where WTP is adjusted to take accountonly of per capita income differences, performjust as well in terms of transfer accuracy [Al-berini et al. (81), Ready et al. (76)]. Understand-ing more about the relative merits of simpleand sophisticated transfer exercises is of consid-erable practical importance. The implication,however, of focusing mostly on intricate ap-proaches is that benets transferas is the casefor primary valuation methodsmight largelybecome the preserve of the highly trained spe-cialist rather than a tool that can be routinelyused by a broader assortment of practitioners.

Invariably, benets transfer introducesgreater uncertainty into appraisals in that it en-tails additional assumptions and judgments tothose contained in original studies. The keyissue is whether the transfer is still, on bal-ance, informative, a question that was asked in acomparatively early contribution by Brookshire(83). As a practical matter, some degree of im-precision may not matter, and rather what isneeded are rules of thumb relating to the hurdleof accuracy that any transfer must overcome in-stead of a benchmark using overly stringent sta-tistical criteria [Ready et al. (76), Kristofersson& Navrud (84)]. Such pragmatism needs to op-erate within some formal understanding of whatconstitutes best practice. It is surprising thatthere are no generally accepted practical trans-fer protocols to guide practitioners [but seeChamp et al. (11), Brouwer (77), and Navrud& Ready (75)]. Given the potentially centralrole in environmental decision making of bene-t transfers, and its possible problems, it seems


Ann

u. R

ev. E

nviro

n. R

esou

rc. 2

008.

33:3

17-3

44. D

ownl

oade

d fro

m a

rjourn

als.an

nualr

eview

s.org

by L

ondo

n Sc

hool

of E

cono

mic

s and

Pol

itica

l Sci

ence

on

10/1

5/08

. For

per

sona

l use

onl

y.


reasonable to predictas well as to hopethatsuch guidelines will emerge in the near future.That said, the problem surrounding how togeneralize the use of nonmarket or environ-mental values does not start itself with the issueof benet transfer (and its conduct). There isalso the question of whether an adequate andstandardized database of environmental valuesis being established particularly in an academicliterature which up to now has been concernedlargely with methodological developments (A.Randall, personal communication). Fullmentof this empirical task may well lack the glam-our attached to further methodological inves-tigation of, for example, SP approaches but isnonetheless just as crucial (and perhaps moreso) in terms of facilitating policy applicationsthrough benets transfer.

3. EQUITY AND CBA

Although the origins of CBA can be tracedback to middle of the nineteenth century, thebody of modern-day welfare economics, whichunderlies CBA, was established in the 1930sand 1940s [for the history of CBA, see Persky(85)]. In particular, the Kaldor-Hicks compensa-tion principle established the idea of hypotheti-cal compensation as a practical rule for decid-ing on policies and projects in real-life contexts[Hicks (86, 87), Kaldor (88)]. This compensa-tion principle established the rule that bene-ts (gains in human well-being) should exceedcosts (losses in human well-being) for policiesand projects to be approved. In recommend-ing those proposals that confer the largest netbenets, CBA appears to neglect questions ofdistribution and, moreover, views the worth ofa proposal in terms of the outcomes that it gen-erates rather than say other dimensions such asjustice [Randall (89)]. Given that equity and jus-tice concerns often dominate discourse aboutsocial decisions, it has often struck critics as bor-dering on the perverse that CBA has chosen tofocus its attention so squarely on efciency.

Placing the spotlight on efciency, however,does not necessarily reect a judgment that eq-uity, or distribution, does not matter. Rather

it assumes that economic efciency underly-ing the Kaldor-Hicks compensation test canor should be separated out from the issue ofwho gains and loses: the distributional inci-dence of costs and benets. There are a numberof reasons for this separation. In the round, netlosers currently could be net winners for ef-cient projects implemented at some later date,a tenet that Persky (85) calls the classical creed.Perhaps the most prominent view is that, giventhe apparently robust theoretical foundationsof the Kaldor-Hicks test, CBA should be con-ned to maximizing the cake so there is moreto share around according to some morally orpolitically determined rule of distributional al-location. Put another way, equity or distribu-tion matters but not for the purposes of rec-ommending specic projects or policies. Andalthough the emphasis on efciency, on theface of it, might seem one-dimensional, fromthe standpoint of CBAs role as just one inputinto the decision-making process, this single-mindedness could be important. This is becauseone can always rely on the political process rais-ing the equity issue, but not necessarily the ef-ciency issue, and so, in this sense, conventionalCBA redresses the balance [Pearce et al. (8)].

Despite this prevailing view, now classiccontributions to the modern theory of CBA[such as Little & Mirrlees (90), Dasgupta et al.(91), Squire & van der Tak (92)] have all em-braced the notion that ways in which societalwell-being can be described (the underlying so-cial welfare function in CBA) are arbitrarilylarge, and some alternatives may involve givingdifferent weight, on the basis on plausible soci-etal preferences, to different distributional out-comes. Others have come at the problem from asomewhat different angle, noting that, in prac-tice, equity and efciency issues are hard to sep-arate. For example, future generations cannotbe compensated, potentially or otherwise, forclimate change liabilities that we are currentlycreating, or at least the transfers that wouldcomprise these reparations cannot be guaran-teed, given that we have no control on the ac-tions of intervening generations [Lind (93)].Others, in a similar vein, have questioned the


Ann

u. R

ev. E

nviro

n. R

esou

rc. 2

008.

33:3

17-3

44. D

ownl

oade

d fro

m a

rjourn

als.an

nualr

eview

s.org

by L

ondo

n Sc

hool

of E

cono

mic

s and

Pol

itica

l Sci

ence

on

10/1

5/08

. For

per

sona

l use

onl

y.


ease of transfers across international bound-aries [Hepburn (94)]. A further tack points outthat those transfer mechanisms that do existare not costless, and so an empirical questionis, What is the best way of addressing distri-butional concerns (i.e., inefcient transfers orinefcient projects) [Harberger (95)]?

What all these critiques amount to is a chip-ping away at the bedrock of the Kaldor-Hickscompensation test that supports the standardCBA approach. Put another way, if the separa-tion assumption (of efciency and equity) canno longer be sustained, or at least if it is lessclear-cut, then this could serve as a rationaleto consider distributional issues more fully inproject selection or design. It still leaves roomfor debate about what form any distributionalanalysis should take. Kristrom (96) character-izes the options as belonging to a hierarchyranging from relatively simple steps such asidentifying and cataloging how project-relatedcosts and benets are distributed to recalculat-ing a projects net benet using assigning ex-plicit distributional weights to the benets re-ceived and costs incurred by different societalgroups. This gives rise to a revised social de-cision criterion that the project is worthy, onbalance, if the sum of its equity-weightednet benets is at least positive:

i ai NBi 0,

where NBi are net benets (simply, benetsminus costs) and ai are equity weights thatvary across i individuals, households, or soci-etal groups. An important feature of conven-tional CBA is now apparent. It assumes weightsof unity are assigned to the net benets of indi-viduals regardless of who it is that receives a unitof benet or suffers a unit of cost [Dre`ze & Stern(97), Brent (98)]. Starting with Fankhauser et al.(99), there has been a resurgence of interest inequity weights in the literature on the distribu-tion of the burden of climate change damageacross countries [see also Azar (100), Tol (101),Anthoff et al. (102)].

Debate about what form these weightsshould take has surrounded different conceiv-able functions and forms that describe socialwelfare and, more specically, the measure ofinequality aversion (reecting in turn a judg-

ment about how much more, in terms of itsimpact on social welfare, a dollar of climatechange damage suffered by a poor person isworth than a dollar of damage endured by arich person).4 Comprehensive reviews of theliterature on estimates of inequality aversioncan be found in, for example, Pearce & Ulph(103) and Cowell & Gardiner (104). The nd-ings to date suggest that equity weighting couldmake a signicant difference to decision mak-ing using CBA, but the range of plausible valuesthat these weights could take is possibly large.It is, in large part, this ambiguity that makescost-benet analysts reluctant about using eq-uity weights. In other words, the concern is notthat equity weights are irrelevant, but that, as apractical matter, we know too little about whatvalues these weights should take. Most of thesedata thus far have been obtained by studies of re-vealed behavior of individuals when saving andgovernments with regard to redistributive poli-cies combined with thought experiments as towhat seems reasonable. SP methods have beenless commonly used, although a handful of re-lated studies have shown the potential for usingthese methods in the context of environmen-tal equity [Atkinson et al. (105), Scarborough& Bennett (106)]. In the interim, some prac-titioners advocate looking at switching values:that is, by asking, What set of weights wouldbe required to tip the balance between recom-mending that the project go ahead or not goahead [Gramlich (107), Kanninen & Kristrom(108)]? This does not make the controversy ofweights go away but shifts it once more, albeitat a later stage, to the political domain.

4For example, one commonly used way to calculate equityweights can be written: ai = (Y/Yi )e , where Y is average ormean income per capita; Yi is income of the ith individual orhousehold (or group); and crucially e is the elasticity of themarginal utility of income or societys valuation of an incre-ment to that individuals income. Intuitively, this elasticity issaid to reect societys degree of inequality aversion. That is,if e = 0, then each unit of benet is awarded the same weight(i.e., unity) regardless of who it accrues to. Higher values ofe boost the weight a unit (e.g., a dollar) of benet is given forthose with below average income and shrink it for those withabove average income.


Ann

u. R

ev. E

nviro

n. R

esou

rc. 2

008.

33:3

17-3

44. D

ownl

oade

d fro

m a

rjourn

als.an

nualr

eview

s.org

by L

ondo

n Sc

hool

of E

cono

mic

s and

Pol

itica

l Sci

ence

on

10/1

5/08

. For

per

sona

l use

onl

y.


4. DISCOUNTINGAND THE FUTURE

Discounting involves attaching a lower weightto a given unit (say $1) of future benet (or cost)than to an equivalent present unit.5 The weightsare determined by time itself and by a discountrate, which is expressed as a percentage. Dis-counting is justied by the assumption that itis what people do, because they are impatientand the fact that capital is productive (i.e., can beinvested now for some future return), and so ac-cords with the basic value judgment of CBA thatpeople should have what they want and prefer.CBA was primarily formulated for appraisalsof projects whose lifetimes did not stretch fur-ther than the short to medium term [Hepburn(109)]. Environmental problems, however, typ-ically have a longer time horizon and the ef-fect of discounting is often to give very low(and possibly practically zero) weight to eventsin the far-off future. CBA then, by discount-ing, appears to make these problems, such asclimate change, disappear: this has led to adebate over the so-called tyranny of dis-counting [Markandya & Pearce (110)] or thediscounting dilemma [Hepburn (109), Groomet al. (111)].

The response of many economists to this de-bate has been to defend the moral rationale forthe principle of discounting in part by point-ing out the apparent ethical implications of notdiscounting. Much of this defense rests on theimplication of the impoverishment of the cur-rent generation, which is destined ever moreto reduce and postpone its own consumption[Koopmans (112), Olson & Bailey (113), see fora discussion, Groom et al. (111)], although thispoint has nuances [see, for example, Asheim &Buchholz (114)]. Yet the long-standing recog-nition within the economics profession of thepossible limitations of CBA in evaluating pro-

5Discounting implies that the weight, wt , to be attached to again or loss in any future year, t, is less than 1. More specif-ically, the discounting formula is: wt = 1(1+s )t , where s isthe (social) discount rate. This discount factor, wt , thereforeshrinks as t gets larger (i.e., as gains and losses become moredistant).

posals with distant time horizons can be foundin Lind (115) in the context of U.S. energypolicy. Proposals for an environmental dis-count rate, reecting irreversible environmen-tal losses, can be traced back to contributionsfrom Krutilla & Fisher (116). More recently,an edited volume by Portney & Weyant (117)has explored a diverse number of alternatives tothe discounting problem. Emerging from thisrenewed interest in a way around the seemingimpasse is an increasingly inuential view thatthe assumption of a constant discount rate inthe conventional approach should be replaced,in some form, by time-declining discount rates.This call for a radical change in the practiceof discounting comes from a number of differ-ent quarters and includes posing the discount-ing problem as a social choice or moral prob-lem [Chichilnisky (118), Li & Lofgren (119)]or more simply as a reection of individualpreferences toward the future [Henderson &Bateman (120), Frederick et al. (121)]. But per-haps the most prominent rationale has at itsheart uncertainty about future interest rates[Weitzman (122, 123)] or the state of the econ-omy [Gollier (124)].

Although much of this discussion is ratherabstruse, it has a very real implication: the so-cially correct discount rate, to be used by gov-ernments in investment and policy appraisal,is one that declines with time. The effect ofthis is to slow down the increase in the dis-count factor across time (relative to the casewhere the discount rate is constant). In otherwords, the discount rate itself changes withtime, getting smaller as time progresses. Thepractical effect of this recent . . . revolution inthinking about discounting . . . [Pearce et al.(8), p. 186] can be dramatic for representinglong-term issues in CBA. For example, Newell& Pizer (125) conclude that the introductionof time-declining discount rates has the effectof doubling their estimate of the social costs ofcarbon, and Groom et al. (111) show that cer-tain candidate forms of declining discount ratesmight change decisions about U.K. nuclear en-ergy options while others do not. The impli-cations of declining discount rates in a wider


Ann

u. R

ev. E

nviro

n. R

esou

rc. 2

008.

33:3

17-3

44. D

ownl

oade

d fro

m a

rjourn

als.an

nualr

eview

s.org

by L

ondo

n Sc

hool

of E

cono

mic

s and

Pol

itica

l Sci

ence

on

10/1

5/08

. For

per

sona

l use

onl

y.


variety of countries are explored in Hepburnet al. (126).

Time-varying discount rates also have theirown problems, and chief among these is timeinconsistency where plans that are made at onepoint in time are contradicted or reversed bylater behavior [Strotz (127)].6 The indicationsare that certain forms of time-declining dis-count rates solve one problem (the tyranny is-sue) by creating another [Heal (128), Hepburn(129)]. Even so the crucial issues are the rela-tive size of the gains from resolving the tyrannyproblem and the problems from the apparentresulting adoption of time-inconsistent poli-cies. Groom et al. (111), for example, specu-late whether these problems are more signif-icant than policy inconsistencies and reversalselsewhere in the political domain.

The controversy that discounting can causeis nowhere more evident than in the dramaticndings of the Stern Review on the economicsof climate change [Stern (130)] and the ensuingdebate surrounding those conclusions. Much ofthis debate has centered on just one, albeit fo-cal, facet of the review: its ndings on the socialcosts of carbon, which indicate that, in the ab-sence of concerted action, the world risks whatmight amount to a decline of between 5% and20% of its global consumption now and for-ever. Compared to the apparently signicantlylower costs of action to combat climate changealso presented, it is little wonder that the reviewconcluded that the benets of strong, early ac-tion far outweigh the economic costs of not act-ing (130, p. vx). Subsequent debate has focusedon the evidence that underpinned this centralconclusion [Nordhaus (131), Weitzman (132),Dasgupta (133), Maddison (134), Tol & Yohe(135), Mendelsohn (136)], including the role ofthe social discount rate (SDR) and the funda-

6As an illustration, time consistency requires that genera-tion A chooses a policy and that generation B acts in accor-dance with it. Generation B does not revise what generationA planned. If generation As plans are revised by generationB, then generation A will not have optimized its behaviorwhat it intended for generation B will turn out to have beenwrong [Pearce et al. (8)].

SDR: social discountrate

mental reasons for discounting: pure time pref-erence and the utility value of future incrementsin consumption.7 Although this debate seems arather narrow focus, at its heart are fundamentalissues for the use of CBA in this grand setting.

The rst is the increasing recognition thatthe climate change problem is perhaps trulynonmarginal in the sense that the entirety ofdecisions affecting this problem could end upshifting the development path that the worldeconomy is on (with, e.g., business as usualglobal emissions of greenhouse gases possiblyleading to considerably lower future consump-tion levels than now) [Weitzman (132), Hoel& Sterner (137)]. This has signicant implica-tions for the choice of discount rate that mightbe used to evaluate climate policy, and which it-self depends on the path of future consumption.Thus, this basic insight complicates the climatediscounting problem considerably as, for exam-ple, the Stern Review shows not least by movingthe analytical focus away from the cost-benetpractitioners comfort zone of basing the SDRon current market rates.

A second important facet of this discussionabout discounting has turned on whether, in se-lecting component values for the SDR, a de-scriptive approach or a prescriptive approachshould be used [Dietz et al. (138)], a distinctionthat also can be found in (139). Put anotherway, should climate-related investments be ap-praised respectively in the light of informationabout preferences for the future as revealed inactual economic decisions, or is there room forthe analyst to make explicit moral judgmentsabout intergenerational equity [for a discussionof the ways in which ethics and economics areinterwoven in the climate change context, seeDietz et al. (140)]? The Stern Review (130) optsfor the latter approach, for example, in argu-ing that while one factor determining the purerate of time preferencethe (small) risk that

7More specically this discussion focuses on the consumptiondiscount rate (s). This is written as s = + .g, where isthe pure rate of time preference (or utility discount rate), is the elasticity of the marginal utility of consumption, and gis the growth rate of per capita consumption.


Ann

u. R

ev. E

nviro

n. R

esou

rc. 2

008.

33:3

17-3

44. D

ownl

oade

d fro

m a

rjourn

als.an

nualr

eview

s.org

by L

ondo

n Sc

hool

of E

cono

mic

s and

Pol

itica

l Sci

ence

on

10/1

5/08

. For

per

sona

l use

onl

y.


QOV: quasi-optionvalue

future generations will not existis a reasonto discount another factor, pure impatiencepreferring benets now rather than later sim-ply because these come sooneris not. A lowvalue for the pure rate of time preference fol-lows naturally from explicitly taking this ethicalperspective. The substance of Nordhaus (131)and Weitzman (132) is that there is, on the faceof it, very little evidence that this moral rea-soning is reected in peoples actual behaviorand choices, and thus, the empirical evidence isthat the pure rate of time should take a highervalue. Similarly, Dasgupta argues that the cost-benet conclusions follow from . . . viewson intergenerational equity rather than . . .new climatic facts . . . (133, p. 4). Althoughthis claim is disputed by, for example, Dietzet al. (138), it is clear that the discounting debatemust confront profound questions about in-tergenerational equity, to quote Beckerman &Hepburn, on which reasonable minds may dif-fer (141, p. 198). Moreover, in the view of thoseauthors, the tools of environmental CBA have amore proactive role to play in deliberating be-tween seemingly intractable viewpoints by, forexample, the use of SP methods and so on to as-sess social preferences toward different climatechangerelated outcomes [see, for example,Saelon et al. (142)]. Even though fresh perspec-tives on the discounting debate might be worth-while, there is also increasing recognition that ameaningful reection, in cost-benet terms, onthe implication of future but uncertain climatecatastrophes might require additional analyti-cal tools [Weitzman (132)]. It is to these issuesof risk and uncertainty that we now turn.

5. UNCERTAINTYAND IRREVERSIBILITY

Although costs and benets are rarely knownwith certainty, the ip side of thisnamely,uncertaintywhether about a physical impactor its monetary value is a question of extent aswell as its nature. In some cases, a risk prob-ability distribution might be known, yet suchdistributions differ according to their degreeof sophistication. In other cases, those of gen-

uine uncertainty, there is no known probabilitydistribution. The ways in which risk and un-certainty have been integrated into CBA havenot changed much over the years. In practice,this has meant use of sensitivity analysis or thecalculation of expected values (where the deci-sion maker is assumed to be risk neutral) andexpected utilities (where the decision maker isassumed to be risk averse). However, the pres-ence of uncertainty combined with other fea-tures of a decision, e.g., irreversibility about saybiodiversity loss, has given rise to at least two in-teresting developments. These approaches varyin their linkage to conventional economics, butwhat they have in common is the provisionof substance to the notion of a precaution-ary principle in guiding environmental decisionmaking.

The rst development arises from the ex-istence of uncertainty and irreversibilityperhaps because funds committed cannot beuncommitted or because other effects of thepolicy cannot be reversedcombined with thepotential to learn by delaying a decision. In en-vironmental economics, this is mostly known asquasi-option value (QOV), although elsewhere,in nancial economics, it has been called optionvalue or real options [see, for example, Kolstad(143), Dixit & Pindyck (144)]. QOV is not aseparate category of economic value [Freeman(145)]. Instead, it is the difference between thenet benets of making an optimal decision andone that is not optimal because the latter ig-nores the gains that may be made by delayinga decision and learning during the period ofdelay. Thus, a development option, which, forexample, involves the permanent conversion oftropical forestland to agricultural land, must bedebited with the potential forgone costs of notwaiting to learn more about the benets of for-est conservation. How large the gain is fromthis process is essentially an empirical ques-tion, but, in the context of environmental pol-icy, Pindyck (146), for example, speculates thatthe implications could be dramatic. A numberof empirical studies have sought to use the no-tion of QOV to explain WTP/WTA disparities[Zhao & Kling (147)] in plantation forestry [Yap


Ann

u. R

ev. E

nviro

n. R

esou

rc. 2

008.

33:3

17-3

44. D

ownl

oade

d fro

m a

rjourn

als.an

nualr

eview

s.org

by L

ondo

n Sc

hool

of E

cono

mic

s and

Pol

itica

l Sci

ence

on

10/1

5/08

. For

per

sona

l use

onl

y.


(148)] and conservation decisions. For example,in a study of Costa Ricas tropical forests, Bulteet al. (149) show that once uncertainty and irre-versibility are included, the country has too lit-tle forest cover. Nevertheless, in that instance,the authors also conclude that QOV remainsempirically unimportant relative to other con-siderations, such as accounting for the pres-ence of global externalities provided by CostaRicas forests, a nding consistent with anotherstudy that sought to elicit QOV [Albers et al.(150)].

One interpretation of the QOV approach isthat it urges more caution about losing envi-ronmental assets such as ecosystems. Althoughthe approach uses formal economics to dothis, other frameworks have sought to com-bine CBA explicitly with conservation criteria.One idea is a safe minimum standard (SMS)[Ciriacy-Wantrup (151), Bishop (152)]. Whatthe emphasis on SMS does is to reverse the onusof proof, away from assuming that developmentis justied unless the costs to the environmentare shown to be very high, to a presumption thatconservation is the right option unless the sacri-ce that it entails is very high. Farmer & Randall(153) and Randall (154) argue that the appeal ofthe SMS lies in it being an approach that maywell command broad moral consensus for mak-ing decisions. From this perspective, CBA pre-vails, but as a sustainability threat draws closer,the SMS becomes the priority. This thinkingappears to have inuenced conservation policyin the United States (the Endangered SpeciesAct) as well as Europe (the EU Habitats Direc-tive) [Berrens (155), Pearce (156)]. A handfulof contributions, beginning with Barbier et al.(157), have also proposed a cost-benet rule,which is subject to a constraint that natural cap-ital is kept intact overall, perhaps, for example,by requiring that no further degradation or lossof ecosystems should be tolerated. What thisdoes is make the principle of sustainability ap-plicable to a portfolio of projects. Thus, to theextent that any one project degrades or destroysan ecosystem, this must be covered off byimprovements or additions to ecosystems else-where, i.e., so-called shadow or compensating

projects. Roach & Wade (158) provide an em-pirical investigation of this resource compensa-tion or equivalency in the context of habitats.In the United States, the legal notion of publictrust has been used to support such measureswhere if certain resources are damaged then . . . compensation takes the form of in-kindservices . . . [Kopp & Smith (159, p. 2)]. Aswith the SMS, even though resource compensa-tion provides some trade-offs between costs andbenets, it plainly circumscribes cost-benetthinking in a substantial way.

6. CBA, ITS CRITICS,AND THE REAL WORLD

We have noted previously the extensive aca-demic literature on the theory and practice ofCBA. Detailed ofcial guidance on how to carryout CBA is much rarer and tends to be connedto those countries where CBA is part of theprocess of regulatory impact analysis (RIA),(or, sometimes, appraisal or assessment). TheOrganisation for Economic Co-operation andDevelopment (OECD) has issued its ownguidelines on RIA (160) and also maintainsan Inventory of RIA Procedures (161). Inthe United Kingdom, RIA is mandatory forregulations, although guidance on how tovalue costs and benets in monetary termsis given by the U.K. Treasury (162) in TheGreen Book: Appraisal and Evaluation in CentralGovernment. Interestingly, that document alsocontains pioneering ofcial guidance on usinga declining discount rate for projects or policieswhose lifetimes extend beyond 30 years. In theenvironmental policy context in the UnitedStates, CBA is widely used, and indeed, theU.S. EPA has its own extensive guidelines (163)for preparing economic analyses of regulations.The European Commission (EC) is committedto applying some form of cost-benet test to itsDirectives (see, for example, Article 130r of theTreaty on European Union, 1992), yet in prac-tice, formal CBA has until relatively recentlybeen infrequent, especially in the environmen-tal context. For EC regional schemes (struc-tural and cohesion funds), however, a guidance


Ann

u. R

ev. E

nviro

n. R

esou

rc. 2

008.

33:3

17-3

44. D

ownl

oade

d fro

m a

rjourn

als.an

nualr

eview

s.org

by L

ondo

n Sc

hool

of E

cono

mic

s and

Pol

itica

l Sci

ence

on

10/1

5/08

. For

per

sona

l use

onl

y.


document on CBA does exist. This focusesmainly on conventional project appraisalissues but does have a section on valuingenvironmental impacts in money terms, andthis stress on the environment, and nonmarketimpacts more generally, has more recentlybeen expanded (164).

Environmental CBA in developing coun-tries is becoming increasingly applied, partic-ularly as a result of appraisal procedures at theWorld Bank for scrutinizing investments withthe Banks Operational Policy on the EconomicEvaluation of Investment Operations provid-ing the formal impetus [see, for example, Belliet al. (165)]. A rather different array of projects,under the heading of markets for environmen-tal services, also reects this trend in utilizingcost-benet thinking in the Banks environmen-tal investment decisions. Under these arrange-ments, the sufferer (or beneciary of conserva-tion), perhaps through some intermediary, paysa polluter to change his or her behavior. Thereare well-known obstacles to these Coaseanbargains, and moreover, the evidence is thatso far the nancial ows arising from thesedeals are comparatively small [Pearce (166)],notwithstanding some prominent examplessuch as those made under Costa Ricas ForestLaw.

Despite ample evidence that cost-benetthinking is increasingly prominent as an aid tomaking actual environmental decisions, thereare also plenty of signs that the existence ofprocedures cannot be taken as an indicationthat CBA is used. Silva & Pagiola (167), forexample, take a detailed look at how environ-mental valuation methods have been used toappraise more than 100 projects within theWorld Banks environmental portfolio. Takingstock of these ndings brings to mind the glasshalf-full or half-empty analogy. That is, therehas been a signicant increase in the use of envi-ronmental valuation in recent years. Yet, in themajority of projects surveyed, covering energy,transportation, water, and agricultural sectors,no valuation was employed. In a recent reviewof 74 RIAs issued by the U.S. EPA from 19821999, Hahn & Dudley (168) nd that, although

all of these regulations monetized at least somecosts, only about half monetized some benets.Fewer still (about a quarter, on average, acrossthe period) provided a full monetized range ofestimates of benets, even though the numberdoing so increased notably over the sampleperiod. More generally, OECD (161) statesthat despite the desirability of CBA, it is notused in many of its member countries becauseof the difculties of placing money values ona comprehensive range of costs and benets.Whether this reects genuine complexities orprovides a ag of convenience for downgradingthe consideration of benets is another matter.Ignoring CBA, or doing it poorly, has conse-quences, however. Devarajan et al. (169) reporton a study investigating 1200 World Bankprojects in 58 countries nding that the cost ofnot carrying out (formal) economic appraisal ispossibly high, especially if policy distortions arewidespread. Deininger et al. (170) estimate thateach dollar spent on analysis prior to WorldBank project initiation led to an increase of$48 in terms of development impact. In areview of recent European Directives, Pearce(156) nds that few if any of these initiatives wassubject to economic appraisal, and moreover, itis likely that, for the United Kingdom at least,these policies, such as the most recent amend-ments to the EU Bathing Water Directive, areunlikely to pass a cost-benet test. It is encour-aging, therefore, that, at least in the UnitedKingdom case, ofcial cost-benet appraisalsof these Directives have emerged subsequently(see Environmental CBA in Action).

Discussions about the extent of the use ofCBA, as well as the quality of the assessmentsthat are undertaken, segue naturally intoreections on the political economy of CBA.The fact that decisions are often inconsistentwith CBA can be squared with the realitythat, in practice, CBA is only one input to thedecision, and in various circumstances, otherconsiderations trump economic thinking. Italso may be that decision makers have takenan extreme stance on the criticisms of CBA, orits underlying welfare economics foundations,that exist. Gowdy (171) provides a useful and


Ann

u. R

ev. E

nviro

n. R

esou

rc. 2

008.

33:3

17-3

44. D

ownl

oade

d fro

m a

rjourn

als.an

nualr

eview

s.org

by L

ondo

n Sc

hool

of E

cono

mic

s and

Pol

itica

l Sci

ence

on

10/1

5/08

. For

per

sona

l use

onl

y.


thorough summary of many of these criticisms.Debate has focused on, for example, CBAsunderlying value judgment that individu-als preferences should count in any socialdecision-making rule with the environmentproving to be a particularly controversial area,where, for example, claims for the primacy ofcitizens preferences [Sagoff (172, 173)] areprominent. It has also been suggested that CBAis often, in reality, downgraded because it isnot deliberative in the sense of ensuring groupshave inuence over some decision (in additionto having their costs and benets counted ina cost-benet appraisal) [Turner (174)] (seealso Section 2.1.1, above). Others, however,go the other way: To the extent that groups orindividuals are poorly informed about the envi-ronment, there are too many risks to allowinguninformed views to hold sway over decisions.

Though a great deal of the criticism of CBAis based on misunderstanding or overly simplis-tic characterizations, there is also much that hasbeen written that could pose a genuine causefor reection among cost-benet practitioners.Both List (175) and Hanley & Shogren (176)have commented on the discrepancy between,on the one hand, the current esteem in whichcost-benet approaches appear to be held inpublic decision making and, on the other, theemerging evidence within the academic litera-ture on anomalies about individual preferencesthat might diminish the power of CBA as asource of advice (176, p. 27). List (175) showsthat preference anomalies are likely to be a mat-ter of degree and, in part, are determined byexperience, which is often lacking when envi-ronmental goods are not traded directly, a viewshared by Sugden (177). For Hanley & Shogren(176), echoing Randall (89), the difculties thatthese inconsistencies present do nothing toeliminate the basic recommendation that know-ing whether the benets of an action outweighits costs is one critical element of making a gooddecision or not. Moreover, a CBA that can ac-commodate, in a sensible fashion, recent think-ing about the behavioral complexity of thoseindividuals whose preferences cost-benet an-alysts are seeking to unravel might well be one

ENVIRONMENTAL CBA IN ACTION

Although the academic literature on environmental CBA isabound with contributions on theory and methods, it is oftento the gray literature that we must turn for comprehensive andpractical examples of CBA in action. In the United Kingdom,for example, all regulations are subject to a Regulatory Im-pact Assessment (RIA), which, in turn, must consider costsand benets and their distribution. One such example is therecent appraisal of the U.K. implementation of the EuropeanCommission (EC) revised Bathing Water Directive (BWD) (seehttp://www.defra.gov.uk/corporate/consult/bathingwaters/impact-assess.pdf). The objective of this policy is the tighteningof standards governing the quality of recreational waters, suchas those around the U.K. coastline. The RIA itself is a detaileddocument outlining, for example, the physical impact assessment(on bathing water) and its valuation. The latter itself was basedon the ndings of a stated preference study of the willingnessto pay of citizens for improvements in bathing water quality(and other related attributes). Interestingly, part of the RIApublication is a statement, to be signed by the relevant HMGovernment Minister, attesting to the veracity of the appraisaland, furthermore, that with respects to the policy outcome: . . .I am satised . . . that the benets justify the costs.

component of the best of all worlds. Clearly,this raises difculties in that this increases thecomplexity of CBA and decision makers mightbe constrained in obtaining the requisite ex-pertise needed to realize the strengths of thismore nuanced approach to appraisal [Pearceet al. (8)].

Just as likely, and perhaps more so, however,is that government and its constituent decisionmakers, rather than merely maximizing socialwell-being as assumed in CBA textbooks, arefaced with political realities that necessitate thereconciliation of conicting interests, and giv-ing prominence to CBA might not be the bestway of serving those ends [Pearce et al. (8)].One way of conceptualizing this problem is tolook

Documents

Environmental Cost-Benefit Analysis