The authors derive n-brand solutions for several first-order Markov models of consumer variety seeking, among them those of Givon, Lattin and McAlister and several variants. Such solutions allow a comparative static analysis of long-run market shares relative to changes in the models’ parameters: variety-seeking intensity, brand preference and degree of feature sharing. Along with a simple Multinomial and a fully general first-order Markov model, these models are calibrated for consumers in a behavioral experiment. Such fits allow the nested models to be compared through a likelihood ratio test, and the non-nested ones to be compared through Hauser’s U2 measure. These tests indicate that the Lattin-McAlister model performs arguably better than all but the general Markov model.
