Sample Selection Bias in the Estimation of Recreation Demand Functions: An Application to Sportfishing

Estimating the value of public actions through behavior implied by the demand for recreation has proved valuable in many settings. Early applications were based on grouped data. However the sharpening of tools of welfare analysis has led economists to exploit individual, rather than grouped, data. The appropriate use of individual observations in estimating recreational demand models is not obvious and depends on the economic theory and its behavioral implications and the statistical characteristics of models. The difficulty in exploiting individual observations in estimating recreational demand functions is that a large portion of a randomly sampled population in a given geographical region is likely to be non-participants in the recreational activity of interest. That is, they are observed to demand zero trips. The qualitative implications of such samples seem to be recognized (Bockstael, Hanemann, and Strand 1986; Mullahy 1987), but there are few (Smith 1988; Shaw 1988) published comparisons of the implications of alternative models.