Modeling Household Purchase Behavior with Logistic Normal Regression

Abstract The successful development of marketing strategies requires the accurate measurement of household preferences and their reaction to variables such as price and advertising. Manufacturers, for example, often offer products at a reduced price for a limited period. One reason for this practice is that it induces households to try the promoted product with the hope of retaining them as permanent customers. The successful implementation of this strategy requires knowledge of the extent of price sensitivity in the population, effective methods of advertising, and the existence of a carry-over effect in the household's evaluation of the product. Logistic regression models are often used to relate household demographics, prices, and advertising variables to household purchase decisions. In this article we extend the standard model to include cross-sectional and serial correlation in household preferences and provide algorithms for estimating the model with random effects. The model is applied to scanner panel data for ketchup purchases, and substantive insights into household preference, brand switching, and autocorrelated purchase behavior are obtained.