A note on combining on-site samples and supplemental samples in a logit model of recreation demand

In recreation demand studies much of the data is collected via on-site samples. Thus, no data is collected for those not visiting the recreation site. The fact that no information is available for those not visiting the recreation area complicates statistical analyses because the result is an endogenously stratified sample. A supplemental sample contains information on the exogenous variables but no information on the response variable. A supplemental sample can be used, along with an on-site sample, to estimate a probability model for site visitation. In this article, we develop and apply the logit formulation of the model of Lancaster and Imbens (1996). We use an on-site sample from the Sipsey wilderness area in northwest Alabama. For our supplemental sample, we use the Integrated Public Use Microdata Series for corresponding counties in Alabama. Together, these samples are used to estimate factors influencing the probability of visiting the site.