Some essential product characteristics may not be readily observable to consumers, for whom a demanding effort is then necessary to determine which available alternative is the best fit for them. Overlooking information acquisition costs may obscure the true sensitivity of purchase behavior to changes in prices and other conditions in the market. I provide a model of rational inattention with additively separable latent heterogeneity in which consumers first make a costly learning decision and then pick the most preferred option accordingly. Given market-level data, additive separability enables me to establish identification results relevant for demand applications with differentiated products. I estimate a parametric specification of my model in an application to the Dutch market for automobiles under the CO2-based vehicle registration tax reform, when car manufacturers faced strong financial incentives to underdeclare fuel consumption.
Stochastic Revealed Equilibrium
Nonparametric testability of general equilibrium models using real-world available data has proved elusive due to computational limitations. Using a revealed preference argument, I show that a class of general equilibrium models for a pure trade economy with heterogeneous static-utility maximizers can be statistically tested using panel data on market prices, individual incomes, and mean endowments. I do an application using household data from the PSID and data on prices from the FRED.