On aggregation bias in structural demand models

Consumer demand analysis attracts considerable attention. It remains an open question, however, whether estimating demand with aggregate data is reliable when disaggregate store-level data is given. Demand models may produce biased results when applied to data aggregated across stores with different pricing strategies. In this study, the graphical model is used to investigate the following question: Do we find the same structure when we fit causal models on sub-groupings of stores, as we find when we fit models on aggregate data from all stores?