Essays on choice set heterogeneity in demand estimation

The estimation of demand systems is a key activity in empirical economics. The most prominent use of demand estimates relates to the computation of social welfare changes due to modifications in the economic environment. Examples include the introduction of a new tax (Griffith et al. 2010), a change in the regulation of an industry (Crawford & Yurukoglu, 2011), the introduction of a new product (Petrin, 2002), mergers between companies (Nevo, 2000, 2001), the legalization of soft drugs (Jacobi & Sovinsky, 2012). Any sort of welfare exercise, essential for policy making, would require—ideally—perfect knowledge of the economy; in particular individuals’ preferences. In practice, individuals’ preferences cannot be directly observed and have to be estimated. Economic policies based on biased demand estimates may prove rather ineffective or, even worse, counterproductive. In the practice of demand estimation, the empirical literature has prevalently favoured the use of discrete choice models. An individual is represented by a preference relation and, under the assumption of preference maximization, chooses one alternative among a finite collection of them, the choice set. It has been commonly assumed choice set homogeneity among individuals: that is, each decision-maker faces the same set of alternatives. This is a strong assumption that is likely violated in many settings: different individuals involved in apparently similar choice situations do face heterogeneous choice sets. Unfortunately, the very presence of choice set heterogeneity raises new challenges on the way of consistent estimation of consumers’ preferences, i.e., meaningful welfare analyses