AN EMPIRICAL POOLING APPROACH FOR ESTIMATING MARKETING MIX ELASTICITIES WITH PIMS DATA

The PIMS (Profit Impact of Marketing Strategies) data entail sparse time-series observations for a large number of strategic business units (SBUs). In order to estimate disaggregate marketing mix elasticities of demand, a natural solution is to pool different SBUs. The traditional, a priori approach is to pool together those SBUs which one believes in advance to be very similar with respect to their marketing mix elasticities. We propose an alternative maximum likelihood, latent-pooling method for simultaneously pooling, estimating, and testing linear regression models empirically. This method enables the determination of a ''fuzzy'' pooling scheme, while directly estimating a set of marketing mix elasticities and intertemporal covariances for each pool of SBUs. Our analyses reveal different magnitudes and patterns of marketing mix elasticities for the derived pools. Pool membership is influenced by demand characteristics, business scope, and order of market entry.