Empirical Analysis of Markets with Differentiated Products: The Characteristics Approach

The aim of this paper is to survey the recent empirical IO literature analysing differentiated products markets. The focus is restricted to the discrete choice approach that is carefully described and some recent applications of the full random coefficients model are analysed. The full random coefficients model is based on a characteristic approach: utility of each consumer is function of a limited number of relevant characteristics of the differentiated products. Allowing for interaction between consumer and product attributes, own, and cross price elasticities are not driven by market shares as in the pure logit framework. The satisfactory substitution patterns, the availability of new sources of data, the possibility of avoiding detailed information on the cost side explain much of the popularity this kind of model is gaining among researchers. The model allows for a wide range of policy analysis: impact of regulatory decisions, consequences of the introduction of taxes and subsidies, effects of mergers, benefits from the introduction of new goods, international trade policy, construction of "ideal" price indexes. The main limitations are its static short run nature, some restrictive functional forms for utility and some computational complexities which do not allow for a widespread application of the method.