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