JCPENNY CENTER FOR RETAIL EXCELLENCE

Branding

Retail site location theory recognizes that stores cluster together, implying that retail density affects consumer shopping behavior in a way that enhances retailer performance. Investigations of individual-level shopping behavior, however, have ignored the effects of this clustering (which we call agglomeration). This research models household-level patronage and spending at different store chains as a function of two retail location factors: (1) proximity (i.e., the shopper's travel time to and from the store), and (2) agglomeration (i.e., the effect of shopping at stores in dense retail environments). We further decompose the effects of agglomeration to understand why it affects shopping behavior: Do shoppers plan to shop at multiple stores or do they make unplanned searches among nearby stores based on product price and availability? We also consider agglomeration of different retail formats to determine how locating stores near these formats affects shopping behavior. These effects are measured by estimating a utility-based econometric model on scanner panel data. Store spending is modeled using a Hierarchical Multivariate Tobit specification, with main effects and interactions reflecting retail location factors. Our results show that agglomeration effects actually explain more of household-level shopping behavior than proximity to stores does. We also find that grocery stores are cannibalized by locating near drug and club stores, but not mass merchandisers.

Panel data models have seldom considered quality as a determinant of brand choice, because quality perceptions are not often available from panelists. However, perceived quality is measured for most product categories in different samples. This research augments the choice data in the panel dataset with perceptual data about product quality gathered from a separate source. A Bayesian multinomial choice model is estimated on the panel data with perceived quality modeled as a latent variable. The distribution of perceived quality from the secondary dataset serves as a prior on the latent variable, permitting the perceptual data to inform the choice model. By linking behavioral and perceptual data, the authors are able to compute the returns to perceived quality, and decompose the disposition to buy a brand into quality-based and inertial components.

This research examines the distribution of ranked market shares for consumer packaged goods categories. We find that the ratios of successive ranked market shares (e.g., market share of the second-ranked product to the third-ranked product, the third-ranked product to the fourth-ranked product, etc.) are nearly constant for a given product category. We also find that this ratio varies little across categories, and even less for the same category across markets. Building on this understanding, we use a one-parameter power series to fit ranked market share data. The power series offers a consistently superior fit when compared to the harmonic series commonly used in the social sciences. We are also able to establish a relationship between the power series model and a Markovian switching matrix, thus linking our finding to a plausible underlying behavioral rationale.