Response to cost prompts in stated preference valuation of environmental goods

The stated preference discrete choice experiment, also known as conjoint analysis, is now a standard method for estimating non-use values of natural resources from respondents’ answers to survey questions. A choice experiment (CE) consists of a sequence of choices among several options, each offering various combinations of features together with costs (often expressed as taxes imposed on each household over some number of years). Through their choices respondents are presumed to reveal whether they would accept a given cost in exchange for a better level of a natural resource. These hypothetical “votes” are then fed into econometric models to produce willingness-to-pay (WTP) estimates, both individual and aggregate, which are the ultimate objects of interest in most applications. In the standard random utility model (RUM) widely used for inferring WTP from data from choice experiments, the underlying utility functions are assumed to be logit and uniform throughout the cost distribution. Under these assumptions, the estimated WTPs should be independent of the cost scale, as long as the cost scale is in a range in which some respondents select a do-nothing option and some select a do-something option. Because of the unfamiliar nature of many environmental improvements, and the unfamiliar task of evaluating a non-market good, survey respondents may not know their WTP for environmental amenities or even how to think about their WTP. To perform the choice tasks, the respondents may look for clues in the survey to assist in determining what they think is an appropriate or reasonable WTP. In particular, the costs that are offered in the survey might affect what respondents think they are, or should be, willing to pay. If the effect is small, then it can perhaps be ignored, but if two studies that present different costs but are otherwise identical lead to large differences in willingness-to-pay estimates, it would be inappropriate to assume – without additional analysis – that either set reveals respondents’ prior valuations of the natural resource.