Preference aggregation and repeat buying in households

Marketing researchers have long used brand switching analyses and Markov transition matrices to gain insights into managerial problems. Almost without exception, this work makes (inappropriate) inferences about individual consumers by analyzing household-level data. This paper presents a procedure based on the distribution of run lengths in household level panel data that allows more insights into the choice behavior of the individuals in the household. We test these procedures in a large simulation study by attempting to recover the underlying (known) structure of the process generating a string of panel data. Finally, we use the procedure to classify the purchase behavior, with respect to powdered soft drinks, of a set of households in a panel. Our results show that marketing scientists have the potential to learn and test more hypotheses about the individuals in a household by examining the distribution of run lengths.