Assessing user benefits with discrete choice models: Implications of specification errors under random taste heterogeneity

Over the past decade, significant advances have been made in the capacity of choice models to accommodate flexible patterns of taste heterogeneity and substitution. However, relatively little research has been done on the implication of these more advanced model structures for the computation of economic welfare and consequent benefit measures. This paper investigates the extent that degradation in model specification affects the welfare changes attributed to a transport policy measure. In particular, the focus is on the random taste heterogeneity, and simulated data are used to analyze how model estimation and assumptions about the specification to taste heterogeneity influence the ability to account for it in the user's benefits. The results indicate that there can be circumstances in which the most commonly used taste distributions cannot provide an adequate representation of the underlying distribution of tastes and that misspecification can have unpredictable effects on benefit estimates. The implications of these findings are discussed.