A semi-parametric estimator for revealed and stated preference data—An application to recreational beach visitation

We present a semi-parametric approach for jointly estimating revealed and stated preference recreation demand models. The discrete factor method (DFM) allows for correlation across demand equations and incorporates unobserved heterogeneity. Our model is a generalized negative binomial with random effects; the random effect is composed of a discrete representation of unobserved heterogeneity and a factor loading that translates the heterogeneity measure into a demand effect. Our empirical application is to beach recreation demand in North Carolina. Statistical evidence supports our DFM specification, which imposes less restriction on model dispersion and incorporates unobserved heterogeneity in a flexible manner. Elasticity estimates are smaller than those derived from models with parametric specifications for unobserved heterogeneity, and welfare estimates are slightly larger (and less precise). While parametric models clearly dominate if the specification of unobserved heterogeneity is correct, the semi-parametric DFM provides a flexible alternative in cases where mis-specification is a potential problem.