Handling unobserved site characteristics in random utility models of recreation demand

Abstract

This paper presents a new random utility model of recreation demand that addresses the unobserved characteristics of recreation sites. Most random utility recreation demand models implicitly assume that all relevant site characteristics are observed by the researcher or allow for the possibility of unobserved characteristics in a restrictive manner. Unlike traditional approaches, the proposed model avoids the bias unobserved site characteristics can cause in welfare estimates and the travel cost parameter. Monte Carlo simulations motivate the proposed model by showing that it provides more efficient parameter estimates. In contrast, existing methods yield less efficient estimates and biased standard errors that overstate precision. An empirical application to recreational fishing in Wisconsin illustrates the potential importance of this modeling innovation. In this application, controlling for unobserved characteristics is important for a range of model specifications.