On the inference about a willingness-to-pay distribution using contingent valuation data

Although contingent valuation (CV) is one of the main sources of willingness-to-pay (WTP) estimates of environmental goods, little guidance exists regarding parametric approaches to modeling CV data, which could reliably estimate WTP values based on preferences stated in binary choice, payment card, and open-ended questions, among others. Studies that use parametric models of WTP often select a specification from a limited set of commonly used distributions. To improve the reliability of parametric modeling of CV-based welfare estimates, we propose adopting a more flexible approach that considers a broad range of specifications with varying parametric distributions. We demonstrate the advantages of the proposed approach using data sets from two large CV studies: the eutrophication reduction valuation study for the Baltic Sea Action Plan and the Deepwater Horizon natural resource damage assessment. We find that the best parametric specifications that fit the data differ across the two case studies. Moreover, these optimal specifications do not always align with most commonly used distributions. We further observe non-negligible differences in welfare estimates across the specifications. Our results provide tentative evidence that the variation in the estimates is lower when better-fitting specifications are considered. These findings emphasize the need for cautiously identifying the distribution best fitting to the data. Focusing on the best-fitting parametric specifications, we provide alternative WTP value estimates for the two empirical cases studied.