Estimating Distributions of Willingness to Pay for Heterogeneous Populations

Abstract

Random parameters logit models have emerged as a standard approach in estimating public preferences for a range of goods and services. However, to simplify calculation of welfare measures, the marginal utility of income is often assumed to be constant across the population. We show that by shifting distributional assumptions from marginal utilities to the welfare measures themselves, a random parameters version of a censored logistic regression model (Cameron 1988) yields distributions of welfare measures more directly and without imposing any unnecessary assumption on the marginal utility of income. We develop the theoretical framework for the model and apply it to contingent choice survey data for siting a noxious facility in Rhode Island. In this application, our proposed model and random parameters logit yield similar mean willingness to pay, but welfare measures are more readily interpretable in the proposed model.