Valuing nonmarket goods and services : a test for bias in the iterated referendum-contingent valuation format

Graduation date: 1994Contingent valuation has been used to elicit values\ud for changes in the quantity and quality of nonmarket\ud amenities. Yet hypothetically trading dollars for\ud environmental improvements is not a familiar activity for\ud most survey respondents. A variety of response effects\ud associated with survey collected data can potentially lead\ud to biased contingent valuation welfare estimates.\ud A discussion concerning how respondent values are\ud formulated and reported provides the basis for building a\ud behavioral model of decision making. Individuals are\ud presumed to act like Bayesians, incorporating information\ud where available to update their prior outlooks. This\ud behavioral model is linked to the iterated referendum-contingent\ud valuation format to test for the presence of a\ud type of information bias.\ud The iterated referendum format poses successive\ud valuation questions to survey participants with the second\ud question based on the response to the first. Although this\ud format has been shown to yield an increase in precision\ud over the single question referendum, the validity of the\ud iterated referendum has not been empirically verified. The\ud central research question is whether the initial bid\ud provides respondents with a stimulus to revise or update\ud their valuations.\ud A test for the constancy of respondents' distribution\ud of willingness-to-pay indicates that updating is occurring\ud in two large data sets. Welfare benefits measures are\ud estimated for the single question and for two forms of the\ud iterated referendum format. Nonparametric tests indicate\ud that the behavioral model differs statistically from the\ud two other valuation formats. The iterated referendum\ud welfare estimate derived from the Bayesian behavioral\ud model yields larger and less precise welfare benefits\ud estimates than the standard iterated referendum for the\ud two data sets. The Bayesian structure of the iterated\ud referendum provides a conceptually valid valuation format.\ud It is both a statistical improvement and a more realistic\ud model of decision making