Joint estimation of stated and revealed welfare measures: the conceptual basis: Kenneth E. McConnell
Why we combine stated and revealed preferences The purpose of this chapter is to examine the conceptual foundations for combining models of stated and revealed preferences in empirical applications. Since the advent of stated preference approaches in the 1970s, economists have wondered about the relative virtues of each approach. Comparisons emerged quickly. Bishop and Heberlein (1979) compared contingent valuation measures with real willingness to pay for goose hunting. Brookshire et al. (1982) compared contingent valuation measures with hedonic measures for air quality in Los Angeles. These early comparisons were principally concerned with assessing the validity of stated preference approaches. Since then economists have worked steadily to compare and jointly estimate models of revealed preferences and stated preferences. The motives for jointly estimating revealed and stated preference models have evolved over the years. In the beginning, there was strong objection to the direct questioning approach to valuation, as stated preference approaches were known. Economists tended to believe that only revealed preferences could provide the basis for valuation. The idea that economists could learn anything from direct interviews-so-called survey data-has been debated and discarded several times. Friedman (1953) made the case based on the idea that entrepreneurs don’t have to know what they are doing for firms to maximize profits. Mere survival of firms would guarantee profit maximization. Samuelson (1954) argued persuasively that the incentive to free ride would lead respondents to hide their true valuations of public goods.1 This attitude survived into the 1990s and can be seen in the critique of contingent valuation edited by Hausman (1993). Given this opposition to contingent valuation and other stated preference approaches, much of the early work combining stated and revealed data was motivated by the need to justify the use of stated preferences. The revealed preference approaches would serve as external validity for stated preference results, as Cameron (1992) noted. Carson et al. (1996) assessed a large number of studies, and demonstrated that the tendency was for revealed preference values to exceed stated preference values. In these cases, the maintained hypothesis was that revealed preference approaches provided valid measures of welfare, against which stated preferences could be tested.