Abstract
The double-bounded contingent valuation method (DBCVM) has long been criticized as an inappropriate estimation method because of respondents’ inconsistent behavior during a course of DBCVM survey. Many experimental studies have reported that two subsequent answers are not made independently, but the second answer depends upon the first. There are many kinds of hypotheses, each of which contends a different pattern of dependence of the answers. In this study, it will be tested which hypothesis is more appropriate in explaining the dependence of the answers. For the test, the DBCVM under uncertainty is newly developed here in which some uncertainties associated with the CVM are recognized and the risk premium plays a criterion for respondents to answer.