Testing for robustness of dichotomous choice contingent valuation willingness to pay estimates for threatened and endangered wildlife species

Abstract

Economic theory suggests willingness to pay (WTP) should be significantly higher for a comprehensive good than for a subset of that good. This was tested for by using both a split sample design and paired responses for WTP for several endangered fish and wildlife species in the U.S. In the paired response case correlation between the error terms of willingness to pay response models were corrected for by using a bivariate probit model. Surprisingly, the independent split samples passed the scope test but the paired samples did not. Because the results contradict each other, questions of validity for policy implications are raised. However, using either approach, the benefits of maintaining critical habitat for these species exceeds the costs. Economic theory also suggests that willingness to pay for two goods independently offered should remain unchanged when the survey instrument changes slightly. Four survey treatments consisting of a comprehensive good and a subset of that good were used. The surveys alternated in the question ordering and in the embedded good which accompanied the comprehensive good. Sequencing and instrument context effects were tested for by using both a combined and split sample designs. In the combined sample case we found some evidence of sequencing effects in the data containing the first subset good. Likelihood ratio tests indicated that sequencing did not effect scale or location of parameters. In the test for instrument context effects, evidence was found indicating context does effect willingness to pay estimates. The contingent valuation method is the only available method to estimate passive use values. This method is controversial for many reasons, including the techniques used to estimate willingness to pay. This paper demonstrates that while the confidence intervals do not indicate significant differences between various modeling techniques, the mean and median WTP estimates vary relative to each other. This can lead to conflicting interpretations of what the data show. Resource managers need to use caution when interpreting results until an industry standard can be developed for WTP estimates.