Investigating retailer product category pricing from household scanner panel data

A growing body of industrial organization literature has directed attention to using the household level logit model, aggregated to the firm level, as the firm's demand function in order to study markets with differentiated products. Such models, however, have ignored the role played by the channel intermediaries, i.e., the retailer, in setting retail prices. In this paper, we study optimal retailer product category pricing policies where the retailer's objective is to maximize total category profits, and show how household scanner panel data can be used to determine these optimal prices. The empirical results obtained from analyzing the AC Nielsen yogurt data base reveals that the retailer's category profit maximizing markups are brand specific and simplified pricing rules such as an across-the-board constant markup can be suboptimal. We also find that there are significant interactions between the optimal prices and the retailer's decision to promote brands using feature advertisements. Additionally, we find that promoting different brands of yogurt are not equally attractive, and the retailer can be better off negotiating for different types of trade promotions (marginal cost reduction vs. payment of fixed sum) with the different manufacturers. Finally, by comparing these optimal retailer pricing rules to those obtained based on aggregate store-level sales data, we show that for purposes of normative analyses, using the household-level logit model aggregated to the brand level is theoretically and/or empirically superior to certain familiar aggregate specifications such as the constant-elasticity and linear demand models.