A Model for Observation, Structural, and Household Heterogeneity in Panel Data

Standard methods of understanding customer behavior in marketing allow for differences in sensitivity across consumers, but often assume that the sensitivity of a particular individual is ®xed through time. In many situations, this assumption may not be valid. Both the importance of variables, and the manner that they are combined to form an overall measure of value for an offer, can change. In this paper we propose an approach of modeling a customer's purchase history that allows identi®cation of when these aspects of customer behavior are likely to change. This information is useful, for example, in planning when particular customers will be most likely to respond to an offer. Our approach nests common methods of dealing with individual differences, and allows for the introduction of covariates associated with changes in customer behavior. We illustrate our model with data from a national sample of credit card usage and adoption.