UC San Diego

This research explores the intersection of econometric theory and consumer choice applications. Consumer choice panel data often exhibit persistence in choices which could be explained by unobservable heterogeneity across consumers, like brand preferences, or structural state dependence, like habit formation and brand loyalty. A semiparametric method for identifying and estimating structural parameters in a binary choice model with structural state dependence in the form of a lagged choice variable is presented. The method requires the availability of auxiliary data that satisfy a conditional exogeneity assumption the additional data must adequately explain any systematic relationship between observable and unobservable components of the model. However, it is not necessary to specify the functional form of the relationship. The distribution of the error term is also left unspecified, and certain types of serial correlation of the errors are accommodated. A constructive two-step estimation procedure is proposed. The method is applied to consumer choice data using the IRI Academic Dataset. For a variety of datasets that may be available to marketing researchers, data that may satisfy the assumptions required for the new method is suggested. This discussion highlights specific applications where using the method described above can be helpful in disentangling structural state dependence from unobservable heterogeneity. Simulations show that the semiparametric method estimates structural state dependence better than the usual techniques. A brand choice application using data from the milk product category indicates that standard techniques may overestimate structural state dependence