Investigating Household State Dependence Effects across Categories

Abstract

Brand choices over time are dictated not just by households' intrinsic brand preferences, but also by marketing variables and state dependence effects. The authors examine brand choice behavior across five product categories as a function of these variables. Choice within a category is modeled with a multinomial probit model. A Bayesian variance components approach is used to model the covariation of household response parameters across categories. The model enables the authors to identify similarities in households' state dependence effects across categories. On the basis of purchase data from a panel of 785 households, they also (1) estimate the correlations between household parameters for the marketing mix and state dependence, (2) study whether state dependence effects diminish over time, (3) investigate whether state dependence for a household is related to its demographic and behavioral characteristics, and ( 4) analyze whether category variables influence state dependence.The results indicate that households display similar state dependence effects across the five categories, with two categories exhibiting significantly lower state dependence levels than the other three. Furthermore, significant correlations exist between state dependence effects and marketing mix sensitivities. A household's level of brand choice inertia diminishes with time elapsed since the previous purchase. State dependence is influenced by shopping behavior variables but not by demographics.

The inter-store mobility of supermarket shoppers

“…We¯nd strong evidence for state dependence in store mobility. This result complements earlier work on brand choice (e.g., Keane 1997) and cross-category purchasing behavior (e.g., Seethuraman et al, 1999). The longer a shopper continues with the current main store, the less likely the shopper will transition away to another main store.…”

State dependence and alternative explanations for consumer inertia

“…However, it is an empirical question as to whether or not state dependence is an important force in our data. With a normal distribution of heterogeneity, a number of authors have documented that positive state dependence is present in consumer packaged goods panel data (see, for example, Seetharaman, Ainslie, and Chintagunta, 1999). However, there is still the possibility that these results are not robust to controls for heterogeneity using a flexible or nonparametric distribution of preferences.…”

“…□Brand‐specific state dependence In the basic utility specification , state dependence is captured by a parameter that is constrained to be identical across brands. Several authors have found the measurement of state dependence to be difficult (see, for example, Keane, 1997; Seetharaman, Ainslie, and Chintagunta, 1999; Erdem and Sun, 2001) even with a one‐component normal model for heterogeneity. The reason for imposing one state dependence parameter could simply be a need for greater efficiency in estimation.…”

“…1 1 See, for example, Keane (1997), Seetharaman, Ainslie, and Chintagunta (1999), and Osborne (2007). Shum (2004) uses a discrete distribution of heterogeneity. …”

An Investigation of Consumer Brand Choice Behavior Across Different Retail Formats

“…The vector Z h contains observable demographic characteristics of the households that affect the household specific response parameters (Seetharaman, Ainslie, & Chintagunta, 1999;Hansen, Singh, & Chintagunta, 2006). The latent term l h accounts for effects of other household-specific characteristics such as time constraints of the households that are not directly observed.…”