WRDS Library [Home]
Digital Library Publications Videos Card Catalog

WWRC 91-19
Preference Learning and Contingent Valuation Methods


Standard rhetoric in the guidelines for implementing the contingent valuation method (CVM) insists that respondents be "familiar" with the decision problem that the investigator poses' Familiarity does not breed contempt; it improves the reliability of responses. Reliability is properly said to be enhanced by investigator clarification of what is to be valued (the contingent commodity) and the means by which it is to be valued (the exchange medium). In spite of the intuitive appeal of clarification, only Hoehn and Randall (1987) have investigated its analytical implications for CVM survey designs. They refer to it as the value formulation problem.

Hoehn and Randall (1987) suggest that the value formulation problem originates in "the time and resource constraints of the CVM context..." (p. 229). Repetition and review constraints inhibit the investigator's abilities to communicate complex commodity attributes, thus making what is being valued appear ambiguous to respondents. Moreover, even in the absence of ambiguities about what is being valued, the respondent's time and decision resources inhibit his value formulation process. Hoehn and Randall (1987) model value formulation by comparing the results obtained with a standard consumer optimisation problem to those produced when elements of the posited constraint system are ambiguous or when the respondent must engage in a "time-constrained search and decision process" (p. 231) that results in incomplete optimisation. They demonstrate that both imperfect communication and incomplete optimisation cause the Hicksian compensating measure of value to be weakly less than the same measure formulated under ideal situations. They also show that this compensating measure is non-decreasing with time or with other resources devoted to respondent decision-making.

Throughout their demonstration, Hoehn and Randall (1987, p. 230) presume "...that the respondent ... knows his/her initial level of well being." In short, they assume that the respondent's utility function is static and invariant and that he suffers no doubts about the utility associated with any particular lottery or outcome. They thus follow Stigler and Becker (1977) in proposing that it is more useful and valid to treat individual preferences as constant and to seek a constraint system explanation for any observed changes in commodity demands. Such changes can arise only from shifts in the shadow prices of the individual's resources or changes in household technologies. They cannot arise from the individual's lack of knowledge about how a particular commodity enters his utility function or from changes in his preference ordering (March, 1978; Cohen and Axelrod, 1984).

In spite of the CVM guidelines on familiarity, numerous applications involve nonmarketed environmental commodities with which respondents are unfamiliar. Frequently, respondents plausibly have had little or no actual experience with the commodity, e.g. atmospheric visibility in the Grand Canyon or acid mine drainage in Colorado mountain streams. In other cases, respondents may daily experience the commodity, e.g. trace metal exposures. However, they may view efforts to influence these experiences as futile and have therefore devoted little effort to understanding how the commodity affects their well-being. In order to forge values for such commodities, they must first explore their preference orderings.

In the next section, we construct a model in which an individual who is unconstrained by time and decision resources must form conjectures and accumulate experience about the effect that a well-specified and clearly communicated commodity has on his well-being. We presume that the individual does not know everything about his preferences, not that he has changing tastes. We develop two propositions: (1) if one does not know his preference for a commodity, his willingness-to-pay for a given quantity will be weakly greater than when he has better preference knowledge; and (2) an initial lack of knowledge about one's preferences causes willingness-to-pay to be non-increasing with time or with the application of other decision-making resources. We also argue that although the sequence of beliefs about own-preferences converges to a limit belief, incomplete preference learning may nevertheless prove optimal. A third section reports empirical tests of our two basic propositions. The results from a series of experiments involving a pair of hypothetical and nonhypothetical markets do not refute either proposition. A concluding section summarises our results, and discusses their implications for the reliability of CVM.

Water Resources Publications List
Water Resources Data System Library | Water Resources Data System Homepage