Bayesian inference in modeling recreation demand

Abstract

McFadden's (1981) hypothesis of random utility maximization (RUM) has been a workhorse for researchers in different fields of economics to study household choices among competing goods. The RUM hypothesis speculates that maximization of utility is the driving force behind individual agents decision to choose among available alternatives and thus individual preference distribution is a consequence of choices made by the whole population. This conjectures makes the RUM model appealing to theorists and practitioners alike.The focus of this dissertation is to add to the existing literature on applying RUM models to recreation demand by consistently estimating and evaluating the demand for and welfare derived from recreational sites. The models proposed in this dissertation allows the Random Utility Maximization (RUM) model to be robust to issues such as: limitations in the information available to researchers; relaxing the assumption of constant marginal utility of income and, incorporating model uncertainty in the estimation process.For example, the first paper, looks at re-specifying the RUM model to incorporate unobserved site characteristics by the introduction of alternative specific constants with the ability to recover parameters of the observed site characteristics that is of interest for policy scenario simulations. Ignoring this issues, I will argue, will lead to inconsistent estimates of the parameters that determine the demand for recreational site and consequently lead to mistaken inference and policy recommendations. The second paper relaxes the assumption of constant marginal utility of income. The assumption, though convenient, is a restrictive formulation of individual preferences and choice behavior. a generalized formulation of the Random Utility Maximization (RUM) model that allows the data to dictate the relationship between preferences and income and prices is proposed. The third paper on the other hand proposes a model that draws on the Bayesian paradigm to integrate the variable selection process into the model and reflect the accompanying uncertainty about which is the ``best'' specification used for counter-factual predictions.