Interval Estimates of Non-Market Resource Values from Referendum Contingent Valuation Surveys

Abstract

Contingent Valuation Methods (CVM) have attracted considerable attention in recent years as researchers have begun to resort to survey methods to elicit individual values for non-market goods (such as environmental resources or public goods). Cummings, Brookshire, and Schulze (1986) and Mitchell and Carson (1989) offer comprehensive assessments of these methods. CVM survey questions come in a variety of forms. One format which has proven to be popular is called the discrete choice (or take-it-or-leave-it, referendum, or closedended) question originally implemented in the literature by Bishop and Heberlein (1979). Since respondents find it very difficult to identify precisely their true point value of access to some resource or public good, open-ended valuation questions can be unreliable or can discourage response. In contrast, most people are familiar with being confronted by a posted for a good, and with deciding whether or not they would buy at that price. This is the strategy behind discrete choice CVM questions. A different hypothetical price is offered to each respondent, and by their yes/ no response to the question of whether they would be willing to pay this price, the researcher gains information about their true underlying point value. One persistent difficulty with the econometric techniques that are currently used most widely to analyze referendum contingent valuation data is the absence of confidence intervals for the ultimate value estimates. Early empirical applications approximate mean valuations using a truncated mean based on a logit model, with the upper limit being the highest bid value in the sample. Unfortunately, the variance of this estimated mean cannot easily be recovered from the numerical integration procedure commonly used to calculate the