ABSTRACT
Welfare evaluation of a policy that alters the resource allocation can often be cast as the problem of calculating mean Willingness to Pay (WTP) in a population of consumers with a distribution of utilities induced by varying tastes and budgets. This interpretation is particularly useful when some of the components of choice at the level of the individual are discrete. This chapter reviews the theory of WTP measurement, and provides easily computed WTP bounds. If preferences satisfy the parallel Engle curve condition of Chipman and Moore, so indirect utility is linear in income with a coefficient that does not vary with consumer, alternative, or allocation, then mean WTP can be characterized as the compensating variation for a representative consumer who has the expected utility function. Absent the parallel condition, compensating variations in the expected utility function do not in general yield mean WTP. The chapter reviews generalized extreme value (GEV) random utility models, which yield an analytic expression for expected utility, and when the parallel Engle curve condition holds, an analytic log sum formula for mean WTP coincides with the compensating variation that keeps expected utility constant. An example shows that when utility is nonlinear in income, so the parallel Engle curve condition does not hold, compensating variation calculated from expected utility can give substantially biased estimates of mean WTP. Finally, the chapter describes a Monte Carlo Markov Chain (MCMC) simulation method that can be used to obtain accurate estimates of mean WTP in GEV random utility models that do not satisfy the parallel Engle curve condition. This procedure is of independent interest as a method for obtaining random draws from a GEV density.
