Beyond the Endogeneity Bias: The Effect of Unmeasured Brand Characteristics on Household-Level Brand Choice Models

Abstract

We investigate the role of potential weekly brand-specific characteristics that influence consumer choices, but are unobserved or unmeasurable by the researcher. We use an empirical approach, based on the estimation methods used for standard random coefficients logit models, to account for the presence of such unobserved attributes. Using household scanner panel data, we find evidence that ignoring such time-varying latent (to the researcher) characteristics can lead to two types of problems. First, consistent with previous literature, we find that these unobserved characteristics may lead to biased estimates of the mean price response parameters. This argument is based on a form of price endogeneity. If marketing managers set prices based on consumer willingness to pay, then the observed prices will likely be correlated with the latent (to the researcher) brand characteristics. We resolve this problem by using an instrumental variables procedure. Our findings suggest that simply ignoring these attributes may also lead to larger estimates of the variance in the heterogeneity distribution of preferences and price sensitivities across households. This could overstate the benefits from marketing activities such as household-level targeting. We resolve the problem by using weekly brand intercepts, embedded in a random coefficients brand choice model, to control for weekly brand-specific characteristics, while accounting for household heterogeneity. Overall, our results extend the finding on the endogeneity bias from the mean of the heterogeneity distribution (i.e., the price effect) to include the variance of that distribution.brand choice, choice models, econometric models, targeting, endogeneity, instrumental variables

Intertemporal price discrimination with forward-looking consumers: Application to the US market for console video-games

“…This approach is not suitable in my context for three reasons: (a) my primary goal is normative, to recommend optimal prices to firms, which would not be possible if restrictions from the optimal pricing policy are imposed in estimation; (b) the density of prices implied by optimal profit maximization behavior requires computation of the full dynamic pricing equilibrium for every guess of the parameter vector, which hugely increases the computational burden of the estimator; and (c) as has been pointed out in the literature (e.g. Chintagunta et al 2005), imposing restrictions from the wrong pricing policy could potentially to bias estimated demand parameters.…”

“…The approach is termed limited information because the density of prices helps handle the correlation induced by ξ, but does not provide any additional information about the demand parameters. The technique is analogous to the approach of Villas-Boas and Winer (1999) and Yang et al (2003;models 5/10 in Table 2) for household-data, and to the parametric control function approach of Petrin and Train 2004 (see also the discussion in Chintagunta et al 2005). The empirical strategy I adopt is as follows.…”

Pass-through timing

“…So we follow other researchers (e.g., Chintagunta, 2002;Chintagunta et al, 2005) in using wholesale prices as an instrument. BLP (1995) consider the average of product characteristics of competing products as instruments.…”

Demand Estimation with Availability Variation

“…They include a product-location random effect with product-specific variance to account for local tastes, and find that local variation in demand plays an important role in determining optimal assortments when local assortment size is restricted (e.g., because of stocking costs). The second application, Chintagunta, Dubé, and Goh (2005), investigates the importance of unobserved components of demand for margarine sales in Denver, CO. The authors use household-level panel data to compare random-coefficient models estimated (via ML) with and without an error term that is the product-time analogue of the product-location unobservable in the assortment models.…”