Exact and approximate welfare measures in the repeated random utility model with an application to valuing Great Lakes fish

Non-market values for changes in environmental quality at recreational angling sites are commonly approximated by, or translated into, units such as value per trip or value per fish. In this research the appropriateness of such approximations is investigated, both theoretically and empirically, using a repeated random utility model. The properties of the exact seasonal welfare measure for the model are elaborated, and some key relationships are established between the exact welfare measure and several alternative seasonal measures and approximate measures (bounds) that have appeared in the literature. In contrast to the usual concavity assumptions regarding environmental benefits, discrete-choice welfare measures for changes in site quality are shown to be convex when the underlying site utility functions are linear.;A repeated random utility model of recreational angling in Michigan is used to estimate the value of changes in lake trout catch rates at various Great Lakes. Exact seasonal welfare measures are compared to the bounds and some value-per-fish approximations. The information on the value of lake trout can aid lake management agencies in their fish stocking decisions and in setting predator control targets.;Estimated benefits of changes in fish stocks at individual Great Lakes were largely independent of the quality conditions at substitute lakes. Consequently, for multidimensional policies, the valuation sequence had little effect on the site-specific values and the sum of these values. The minimal embedding effects are explained by the spatial and physical properties of the model and the types of policies considered. This result provides strong support for treating benefits at each Great Lake as independent of one another for fisheries management purposes.;The empirical results exhibited substantial convexity in the discrete-choice welfare measures. In some cases, gains were more than twice as large as losses for comparable increases and decreases in quality. This raises serious questions about the reliability of translating welfare measures into site-quality units such as values per fish. However, for the empirical model considered here, the bounds using values per trip (baseline trips times per-trip welfare measures) closely approximated the exact seasonal measures.