| This dissertation addresses both how consumers' past purchases affect current purchase choices and how accessibility of the past purchases to firms can lead to discriminatory pricing. The relationship between past purchases and current choices can be driven by either the causal relationship between the past and future, termed state dependence, or heterogeneity in consumer tastes. To isolate the state dependence, I estimate demand using a set of random coefficients that allow a distribution of individuals' tastes to persist throughout the panel of purchase data. The demand estimation also incorporates a dynamic programming problem that accounts for forward-looking behavior by consumers. The fully specified demand model can therefore characterize both the relationship between the past and the present, as well as the present and the future. Elasticity estimates of this model demonstrate that consumers in an empirical example substitute both backward and forward in time. These purchase carryovers, together with the heterogeneous tastes that can be indicated by past purchases represent demand variation both for a given consumer over time and across consumers. The purchase history may therefore represent a reasonable basis for firms to charge discriminatory prices. I estimate the effects of firms conditioning prices on the past purchases of their customers by performing counterfactual simulations on the estimated demand model. The counterfactuals suggest that purchase history data can result in discriminatory prices that increase the profits of firms. |
