Applied welfare economics with discrete choice models: implications of theory for empirical specification

The apparatus of the Random Utility Model (RUM) first emerged in the early 1960s, with Marschak (1960) and Block and Marschak (1960) translating models originally developed for discriminant analysis in psychophysics (Thurstone, 1927) to the alternative domain of discrete choice analysis in economics. Whilst some researchers were quick to see its practical potential (e.g. McFadden, 1968, 1975), it was not until the late 1970s and early 1980s that RUM was equipped with a reasonably comprehensive theoretical rationale in terms of the economics of consumption. An important tenet of this rationale was the link between discrete choice and welfare, which established a basis for applying RUM to public policy analysis, and paved the way for the plethora of applications which have been witnessed over the last 30 years. It will be helpful to clarify precisely what we mean by ‘discrete choice’, since Small and Rosen (1981) – which will be referred to as ‘S&R’ in the remainder of this chapter – suggest three alternative rationales, as follows. First, commodities may be available in continuous quantities but only a limited number of varieties. Second, goods may be supplied in discrete units of such magnitude that only a small number of those units are typically consumed (in this case, S&R cite the example of travel mode choice). Third, if the search for the optimal consumption bundle entails a choice between alternative corner solutions, then the problem is reduced to discrete units.