Heterogeneous consumption experience, product choice and product substitution

This dissertation examines and models the demand for product variety. The demand for product variety is a result of varying consumer tastes. The diversity of consumer tastes arises from two general sources. First, each consumer has his own idiosyncratic preference that differs from the others; second, his preference might vary over time. This dissertation proposes an innovative way to capture the impact of these inter-individual and intra-individual preference diversities on the demand for a brand (brand choice) and for a product (SKU choice). Specifically, we use past purchase history in an innovative way to model the time-varying idiosyncratic preference of a consumer. One of our key contributions in this thesis is developing accurate choice models with less parameter.; Our investigation begins with the question of "effective variety" of a brand for consumers. We develop three consumer-specific brand width measures (product variety in a brand) using a product tree structure and past purchase history. These measures are superior to the traditional brand width measure (number of products) in explaining brand choice. Hence, these measures are better at capturing the "effective variety" of a brand. Our empirical findings conform to the conventional wisdom that higher brand width leads to higher brand share.; Next, we extend previous SKU choice models (e.g. Fader and Hardie (1996)) by harnessing additional information from the purchase history. Previous choice models reinforce the idiosyncratic preference of the chosen products; our models also reinforce that of the unchosen products. Specifically, the amount of reinforcement a product received depends on the amount of experience the consumer has with the product. We also differentiate the responsiveness to marketing activities by the amount of experience. We found that, for familiar products, consumers are less responsive to price change while more responsive to display and advertising.; Lastly, we propose a framework to utilize the consumer choice models in deriving optimal product lines. We suggest several possible extensions to this framework.