Combining multiple revealed and stated preference data sources: a recreation demand application: Daniel J. Phaneuf and Dietrich Earnhart
Introduction In this chapter we provide an example of structurally combining revealed preference (RP) and stated preference (SP) data to address parameter estimation and the valuation of time in a recreation demand context. In particular, we combine observations on actual recreation site visitation with three forms of SP data that include future behavior at baseline conditions, contingent behavior at changed conditions, and contingent pricing (i.e. reporting a choke price). By integrating these data sources into a structural, two-constraint model of recreation demand we demonstrate the payoff to collecting RP and SP data that provide related, but distinctly different, vehicles for quantifying individuals’ preferences. Our application examines recreation visits to Clinton Lake near Lawrence, Kansas, during the summer of 1998. An on-site survey of individual visitors provided information on the number of (actual) trips made to the lake, expected future trips, and changes in future trips resulting from hypothetical changes in the money and time costs of access. The latter are examples of contingent behavior data. The survey also solicited information on how much money and time costs of access would need to increase before the person would stop visiting the lake; the responses constitute our contingent pricing data. The different SP questions provide information on how time and money costs determine different elements of peoples’ overall behavior. Said another way, the contingent behavior data and contingent pricing data serve as natural counterparts. While the contingent behavior data reflects the intensive margin relationship between price (travel cost) and visits along the interior of the demand curve, the contingent pricing data measures the reservation (or choke) price that determines the person’s extensive margin decision to visit the lake or not. Our first contribution in this chapter, therefore, is to provide an empirical framework for exploiting the different theoretical aspects that are relevant for the data-generating process. Our second contribution centers on the opportunity cost of time aspect of our application. While several recent studies combine RP and SP data in recreation studies (e.g. Adamowicz et al. 1994; Cameron, 1992; Layman et al. 1996; Herriges et al. 1999; Huang et al. 1997; Cameron et al. 1996; Adamowicz et al.