Elsevier

Ecological Economics

Volume 60, Issue 2, 1 December 2006, Pages 450-460
Ecological Economics

The aggregation of environmental benefit values: Welfare measures, distance decay and total WTP

https://doi.org/10.1016/j.ecolecon.2006.04.003Get rights and content

Abstract

We review the literature regarding the aggregation of benefit value estimates for non-market goods. Two case studies are presented through which we develop an approach to aggregation which applies the spatial analytic capabilities of a geographical information system to combine geo-referenced physical, census and survey data to estimate a spatially sensitive valuation function. These case studies show that the common reliance upon political rather than economic jurisdictions and the use of sample mean values within the aggregation process are liable to lead to significant errors in resultant values. We also highlight the fact that for resources with use values then we should expect overall values to reduce with increasing distance from such sites, but that changes in the choice of welfare measure will determine whether such ‘distance decay’ is to be expected within values stated by those who are presently non-users. The paper concludes by providing recommendations for future improvements to the methodology.

Introduction

Cost benefit analysis requires the aggregation of individuals' benefits in order to compare these with the total costs of a project or policy. Because the methods for measuring non-market benefit values are based on analyses of individual behaviour, there is a problem in knowing how changes in a resource will affect aggregate values. This will depend on both the benefits per person and the population of beneficiaries. It is thus necessary to identify the extent of the market and how this varies with changes in the good. In a manner reminiscent of the The Tale of Horse and Rabbit Stew, the extent of the market may well be more important in determining aggregate values than any changes related to the precision of the estimates of per-person values. This paper builds upon earlier work by Smith (1993) and Loomis (2000) and others to argue that differences in measures of individual values are the haggled-over rabbits in the stew, whereas the aggregation process itself is the horse which can so pervasively dominate the outcome.

The issue of defining the extent of the market for a public good essentially concerns the question of how broadly should individuals' marginal benefit schedules be vertically summed. Whether this should be confined to those living in the close vicinity of the good, or extended across the region, country or even further afield, has implications for the appropriate level of government financing and provision. This issue becomes further complicated when considering the distinction between user and non-user values.

In considering the extent of the market, it is useful to distinguish between the political jurisdiction, concerning some administrative area, and the economic jurisdiction incorporating all those who hold economic values regarding a project. From an economic cost–benefit point of view, the notion of accounting stance (or who has standing) defines the relevant jurisdiction for including benefits and costs. Now in order to achieve efficiency in resource allocation, it is necessary that the accounting stance should be such that it captures all Pareto-relevant impacts. Where political and economic jurisdictions coincide, there is said to be fiscal equivalence (Olsen, 1969, Cornes and Sandler, 1996). However, it seems likely that such equivalence will be the exception rather than the rule. Simple logic suggests that, where some resource site generates any element of use value, the density of such users will be higher near to that site. Furthermore, as users typically hold higher values than non-users, then we would expect average values to decay with increasing distance from that site. It seems likely that the resultant economic jurisdiction will often (but, as discussed subsequently, not always) fail to coincide with any political jurisdiction. Although there is an extensive literature examining variability in estimates of mean WTP (see, for example, Hanemann and Kanninen, 1997, Kerr, 1996, Ready and Hu, 1995), nevertheless Smith (1993) suggests that such variability will have far less impact on estimates of total WTP than will changes in the extent of the market arising from the use of different approaches to aggregation. He provides an example related to the estimation of natural resource damages, in which the differences in plaintiff's and defendant's estimates of per unit values were small but total damage assessment differences were huge as a result of each parties assumptions regarding the appropriate jurisdictions used to define the extent of the market.

The highest profile UK application of a political jurisdiction approach to aggregation occurred in 1998 when the Environment Agency (EA) refused a water company application to abstract water from the River Kennet in southern England (ENDS, 1998, Moran, 1999). The EA approach followed recommendations from FWR (1996) to use water company operations boundaries as the relevant area for aggregation. Total benefit value was then calculated by multiplying the population within this jurisdiction by a sample mean WTP estimate transferred from a contingent valuation (CV) study of the River Darrent (Willis and Garrod, 1995). Aside from concerns regarding differences in the river types, commentators criticised the use of a sample mean taken from one area and applied without adjustment to an entirely different area which not only had differing geographic size but also contained a much larger population with very different socioeconomic characteristics (Moran, 1999, Bateman et al., 2000). These academic reservations were reflected in judicial review through which the EA ruling was overturned after an appeal in which the plaintiff attacked both the valuation and aggregation procedure employed.

Now in principle the use of a sample mean WTP value need not necessarily lead to biased estimates of aggregate values. If a representative2 sample is drawn from the entire economic jurisdiction (or indeed some larger sample area), then multiplying the sample mean WTP by the population of the sampled area should give an accurate estimate of aggregate values in that sampled area. Indeed even if the sample area is a subset of the economic jurisdiction, then the sample mean approach to aggregation still gives unbiased estimates of total value within that subset area. However, problems may well arise where a mean value from some subset area is used as the basis for estimating aggregate values for the entire economic jurisdiction (or some non-coincident political jurisdiction, as in the River Kennet case). There are two sources of error here. First, as noted above, underlying values for changes to some spatially confined resource are likely to decay with increasing distance from that resource (an issue we discuss in greater detail with respect to both users and non-users, in Section 2 of this paper). Failure to account for this distance decay will lead to error if this mean value is used to aggregate values for a larger (or indeed smaller) area. This problem is exacerbated by common practice where valuation studies indeed do fail to sample entire economic jurisdictions, but instead focus survey effort upon areas around the resource in question where values are highest. However, this is where underlying values are highest. The application of consequent sample means in an unadjusted manner to some large (typically political) jurisdiction seems liable to generate over-estimates of aggregate values. A second problem arises when, as seems likely, the probability of responding to a valuation survey is positively related to underlying values. In such cases, self-selection bias (Heckman, 1974) is likely to increase with distance from the resource. Thus, while underlying values should exhibit distance decay, sampled values will understate this as individuals with relatively higher values self-select themselves into the sample at a greater rate than those with lower values. Aggregation procedures need to recognise and address both of these problems.

Clearly, given unlimited survey resources, one would ideally sample from the entire feasible economic jurisdiction. But even here we face the problem that, a-priori, the extent of this area is unknown. In this paper, we argue that, given these challenges, a superior approach to aggregation is to use survey data to adjust for self-selection bias, capture underlying distance decay and define the economic jurisdiction. This, we argue, is best achieved through the identification and estimation of a spatially sensitive valuation function. Such an approach explicitly addresses self-selection and incorporates distance decay relationships into defining the limits of the economic jurisdiction while allowing for variability in the socioeconomic characteristics of the encompassed population within the aggregation process. We use this approach to generate estimates of aggregate WTP and contrast this with estimates based upon both the use of a political jurisdiction and reliance upon sample means. The impact of using different aggregation approaches is then contrasted with the variability induced by uncertainty in the estimate of mean WTP to examine Smith's contention that such variability will be overwhelmed by aggregation impacts.

The following section considers expectations regarding the underlying decay of values across space, highlighting important interactions between choice of welfare measure and the expected response of users and non-users of a site. We utilise these expectations to guide a review of existing evidence regarding such distance decay. This highlights a divergence in expectations depending on whether we are estimating an equivalent loss measure (in which respondents are typically asked to state their WTP to preserve some resource such that the future level of resource quality stays the same as its present quality) or a compensating surplus measure (in which respondents are typically asked to state their WTP for some improvement in the resource such that the future level of resource quality exceeds the present quality). This divergence is reflected in the empirical part of this paper. Section 3 considers self-selection bias and reports an aggregation exercise for an equivalent loss measure using sample mean and valuation function approaches applied to a case where political and economic jurisdictions coincide. In contrast, Section 4 aggregates a compensating surplus measure and presents aggregate values estimated across differing political and economic jurisdictions as calculated using both sample mean and valuation function approaches. Through these applications, we develop a methodology based upon the spatial analytic capabilities of a geographic information system (GIS). Section 5 concludes by discussing results from both our studies, embedding these within the previously reviewed literature, before closing by giving some pointers toward ways in which this methodology might usefully be extended.

Section snippets

Expectations and existing evidence regarding distance decay in underlying values

We can begin by considering the spatial distribution of values for some open access, public good resource at its present level of quality. For the sake of simplicity, we can identify two types of individual: users of a resource and non-users. Users will hold use and option-use values and may well hold non-use values. Non-users may hold option and non-use values or have negligible WTP.3

Case study I: aggregating an equivalent loss measure: WTP to preserve a National Park wetland

The first case study is based on data taken from a postal CV survey of WTP for schemes designed to preserve the Norfolk Broads National Park in Eastern England from the threat of saline flooding (i.e. an equivalent loss measure). Approximately 1000 questionnaires were sent to households across Great Britain and a GIS was used to calculate distance from each address to the Norfolk Broads. Following two reminders, a sample of 310 completed questionnaires was collected. This relatively modest

Case study II: aggregating a compensating surplus measure: WTP to improve urban river water quality

The second case study extends the GIS-based valuation function approach to aggregation developed in the prior analysis. Data is taken from a face-to-face CV survey of WTP for water quality improvements to the River Tame (i.e. a compensating surplus measure), for which a sample of 675 responses was gathered in interviews conducted at the respondents home address. The river rises on the west of the City of Birmingham, in Central England, flows eastward through the city and turns north till it

Conclusion

This paper has considered some of the various factors which can influence the calculation of aggregate WTP estimates. Our analysis confirms the findings of Smith (1993) and Loomis (2000) that the choice of whether to aggregate across a politically defined or economic jurisdiction can have a very substantial impact upon estimates of aggregate value. Similarly, we have shown that the use of simple approaches such as aggregation via sample means can severely bias such estimates, and is likely to

Acknowledgements

Institutional affiliation for all authors is the Centre for Social and Economic Research on the Global Environment (CSERGE) or the Centre for Environmental Risk (CER), School of Environmental Sciences, University of East Anglia, Norwich, NR4 7TJ. This research was supported by the Catchment Hydrology, Resources, Economics and Management (ChREAM) project, which is funded by the ESRC, BBSRC and NERC, Rural Economy and Land Use (RELU) programme, the Economics for the Environment Consultancy Ltd.

References (41)

  • R.C. Bishop et al.

    The economic value of Illinois Beach State Nature Preserve, Final Report to the Illinois Department of Conservation

    (1985)
  • R.T. Carson et al.

    Incentive and informational properties of preference questions

  • A.C. Case

    Spatial patterns in household demand

    Econometrica

    (1991)
  • R. Cornes et al.

    The Theory of Externalities, Public Goods and Club Goods

    (1996)
  • C.V. Deutsch et al.

    GSLIB: Geostatistical Software Library and User's Guide

    (1992)
  • ENDS

    Water abstraction decision deals savage blow to cost–benefit analysis

    ENDS

    (1998)
  • FWR

    Benefits assessments manual for surface water quality improvements, Foundation for Water Research

    (1996)
  • Georgiou, S., Bateman, I., Cole, M., Hadley, D., 2000. Contingent ranking and valuation of river water quality...
  • W.H. Greene

    Econometric Analysis

    (1990)
  • M. Hanemann et al.

    The statistical analysis of discrete response data

  • Cited by (350)

    View all citing articles on Scopus

    Submitted to Environmental Benefits Transfer: Methods, Applications and New Directions, Special Issue of Ecological Economics, Edited by John Hoehn and Matthew A. Wilson.

    View full text