Modeling heterogeneous preferences in the valuation of environmental public goodsa An econometric analysis of dichotomous choice contingent valuation survey data

Abstract

Contingent valuation method (henceforth CVM) has become the most extensively used technique to estimate the value of non-market goods. Traditional contingent valuation studies assume that provision of environmental public goods or programs is distinctly utility increasing for everybody. In turn, this assumption has led CVM research to focus on modeling techniques that restrict welfare measures to positive levels. However, by restricting willingness to pay to be non-negative, these studies have ignored that externalities associated to the provision of environmental goods can manifest themselves as benefits for some people and as costs for others. Ignoring these heterogeneous values for the public good can lead to biased welfare measures, which in turn can affect policy decisions.

This Dissertation is composed of an introductory section and three related papers addressing the problem of how to account for the existence of positive and negative values towards environmental public goods. The introductory section discusses the appropriate welfare measures to be considered in contingent valuation when the public good generates benefits to many but also disutility to others. The sequence of three papers discusses different modeling approaches to account for the existence of heterogeneous preferences towards environmental public goods. The first paper presents a double bounded dichotomous choice model of positive and negative preferences. In this model willingness to accept bids were coded as negative willingness to pay bids, which allowed to combine negative and positive values for the good in the estimation process. The double bounded specification was used to estimate net benefits or net losses for the prescribed burning program. In the second paper, the data structure allowed for the identification of negative and positive responses as in the first paper, but also identified the respondents with zero willingness to pay. The different response categories were modeled accordingly using parametric and non-parametric extended spike models. In the third and last paper, heterogeneous preferences were addressed using a random parameter logit (RPL) model, where individual parameters were allowed to vary across people so as to reflect variation in tastes among respondents towards the public good.

The results from all three papers indicate that accounting for heterogeneous preferences yields significantly lower welfare estimates. This information on the variation of preferences and valuation can be relevant to policy makers and natural resource managers who must balance welfare of a diverse population when making decisions.