Improved Estimation of Willingness to Pay in Dichotomous Choice Contingent Valuation Studies

This paper uses recently developed statistical techniques to achieve improved parameter estimates from models of dichotomous choice contingent valuation data. Random effects associated with bid amounts presented to respondents are modelled alongside the variance of individual responses, to give a nested model which more truly reflects the structure of the data being analyzed. Quasi-likelihood methods for estimating parameters in such hierarchical models are discussed, and a simulation method for assessing goodness of fit is demonstrated. However, estimated parameters may still be biased, and hence, a parametric bootstrap technique is presented and compared to a delta method approximation.