A Monte Carlo Comparison of Two Nonparametric Methods for Referendum Contingent Valuation Data

Abstract

A kernel regression estimator, a series-based semi-nonparametic estimator, and a simple logit model are applied to referendum contingent valuation data generated by Monte Carlo. The nonparametric approaches are found to dominate the simple logit model when the estimand is expected willingness to pay conditional on observable variables, with the kernel approach performing best. For estimation of WTP unconditional on covariates, the simple logit model performs almost as well as the best nonparametric alternative.