“Speed of Replacement”: Modeling Brand Loyalty Using Last-Move Data

Abstract

The authors study brand loyalties for durable goods using automobile survey data that are peculiarly censored and track only elapsed times since transitions but not the transition times themselves. This censoring problem is typical of commercially available durable goods survey data. However, little attention has been paid to such “last-move” data in the statistics or marketing literature on the analysis of transition times. The authors propose a multistate, continuous-time, nonstationary Markov model with a parsimonious brand loyalty structure and also propose an estimation approach to recover the parameters of the proposed model using the automobile survey data. The proposed model fits observed brand choice outcomes even better than a model with a fully unrestricted (and, therefore, highly parameterized) transition structure. The authors also obtain several substantive findings. For example, Chrysler is significantly “weaker” than General Motors and Ford insofar as it has the lowest brand loyalty during the study period. The authors illustrate the managerial implications by predicting time-varying market shares of brands in periods subsequent to the period of analysis.