Estimating Zero Order Models given Variable Number of Purchase Records per Household*

Assuming a zero order process, the beta-binomial distribution model is often used to describe consumers' brand switching behavior. A variety of methods have been used to estimate the parameters of the beta-binomial distribution, each of which assumes fixed number of purchase occasions across households . Consumer panel data, however, are typically available for a fixed duration of time (say, a year) during which consumers invariably make different number of purchases in the product class. The purpose of this paper is to present a method for estimating zero order models from panel data containing different number of purchase records per household. Empirical data are presented to demonstrate the efficiency of this estimation method.