Testing the Double Jeopardy Loyalty Effect Using Discrete Choice Models

This paper attempts to validate the double jeopardy loyalty effect in a utility framework using a discrete choice approach instead of the Dirichlet model. We specify brand choice and allow for differences in brand loyalty measures across brands in two different product categories. The discrete choice model formulations include a multinomial and a latent class multinomial logit model. Using ACNielsen household scanner panel data to estimate the models, we find that market share leaders enjoy higher purchasing loyalty than do lower market share brands. The results have relevant implications in terms of marketing mix decisions for brand managers.