A multi-category analysis of consumer shopping behavior
Faculty Advisor: Sunil Gupta
Consumers choose multiple items (across categories) in many situations such as grocery shopping and mail-order purchasing. A fundamental question that arises in these situations is whether the choice of these items is interdependent and, if so, what is the nature of this dependence. In this dissertation, we propose a unique and comprehensive framework to investigate multi-category choice. Under this framework, the three factors that could lead to joint choice are complementarity, co-incidence and heterogeneity. Complementarity refers to factors under managerial control (e.g., price), co-incidence to factors not under managerial control (e.g., store traffic) and heterogeneity to differences in consumer preferences.
In the first part of the dissertation, we model the composition of consumers' “shopping baskets” which result from their purchase incidence decisions during a shopping trip. The composition of the basket is modeled as a function of marketing actions, uncontrollable factors and consumer response heterogeneity. We propose a general model that explicitly allows for the choice of items to be interdependent. The results show the effects of complementarity and coincidence on basket composition and indicate that related category pairs may have a “primary” and a “secondary” category.
In the second part of the dissertation, we examine the role of complementarity in greater detail for purchase incidence and brand choice. We model these two consumer decisions for a pair of categories. The results from this model show the differential impact of each category's pricing and promotion policies on joint choice and the relationship between cross-category brand preferences.
The two models, both based on random utility theory, use different techniques. The first model examines a broader marketing issue using techniques that are new to the marketing literature (hierarchical Bayes models, Markov chain Monte Carlo methods) while the second examines a narrower, deeper issue using established marketing models (nested logit type models).
The insights gleaned from this dissertation are likely to prove useful to retailers and manufacturers trying to optimize pricing and promotion across a portfolio of categories/brands and to database marketers who are interested in cross-selling at the individual household level.