Identifying Preference Heterogeneity in the BLP Model with Micro Data
Abstract:
Micro data helps identify preference heterogeneity in the workhorse “BLP” model of aggregate demand, but the full extent of its identifying content is unknown. We provide flexible identification results with three different types of micro data: (i) cross-sectional data with a single purchase event for each individual; (ii) second-choice data that elicits consumers' next-best option; and (iii) panel data that tracks multiple purchase events for each individual. With cross-sectional data, sufficiently rich demographic-driven variation identifies forms of preference heterogeneity where demographics do not affect preference dispersion. Additional data on second choices significantly reduces – but does not eliminate – the need for demographic-driven variation. Panel data fully identifies more flexible preference heterogeneity, without the need for any demographic-driven variation. Thus, panel data contains the most identifying content; more than cross-sectional and second-choice data combined. We also show that the amount of micro data needed for identification is dramatically reduced when micro and market-level data come from the same markets.