You are not currently logged in.
Access JSTOR through your library or other institution:
Variety Seeking, Purchase Timing, and the "Lightning Bolt" Brand Choice Model
Pradeep K. Chintagunta
Vol. 45, No. 4 (Apr., 1999), pp. 486-498
Published by: INFORMS
Stable URL: http://www.jstor.org/stable/2634819
Page Count: 13
Preview not available
The "Lightning Bolt" (LB) model provides a comprehensive framework for accommodating the effects of habit persistence, unobserved heterogeneity, and state dependence on household brand choice behavior. This paper presents a discrete, dynamic brand-choice model that belongs to the LB class of models. We propose a method for incorporating the effects of variety seeking into the LB model formulation. The proposed formulation explicitly links brand choice and purchase timing behavior via the effect of state dependence. This state-dependence based linkage between brand choice and purchase timing comes about due to the attribute satiation notion associated with household variety seeking behavior. Empirical implementation of the model specification requires recognizing that interpurchase times, like brand choices, also depend upon marketing variables and household characteristics. A hazard model is specified to capture this relationship. Empirical results are presented using household-level data for two different products. Our results reveal the following. (1) State dependence effects decline over time. Hence, house-holds' brand switching and repeat purchase probabilities vary over time, independent of any variation in marketing mix activities. (2) Accounting for the effects of marketing variables on interpurchase times may be important when empirically estimating the proposed model. While the nature of substantive implications is largely unchanged, the magnitudes of the different effects are nevertheless affected. (3) We find evidence for purchase reinforcement (and its effects declining over time), but no evidence for attribute satiation based variety seeking behavior. (4) A comparison of the model's predictive ability with those from two extant specifications reveals that the proposed model outperforms rival specifications.
Management Science © 1999 INFORMS