Two-Stage Estimation to Control for Unobservables in a Recreation Demand Model with Unvisited Sites

Abstract

The role of unobserved site attributes is a growing concern in recreation demand modeling. One solution in random utility models (RUMs) involves separating estimation into two stages, where the RUM is estimated with alternative specific constants (ASCs) in the first stage, and the estimated ASCs are regressed on the observed site attributes in the second stage. Prior work estimates the second stage with ordinary least squares (OLS) and two-stage least squares regression. We present an application with censored regression in the second stage. We show OLS produces inconsistent parameters when there are unvisited sites with no estimable ASCs and that censored regression avoids this problem. (JEL Q21, Q51)