Tariff Choice with Consumer Learning and Switching Costs

Consumers choosing flat-rate contracts tend to have insufficient usage to warrant the cost, particularly for new products. We propose and estimate a Bayesian learning model of tariff and usage choice that explains this "flat-rate bias'' without relying on behavioral misjudgments or tariff-specific preferences. For new products, consumers are uncertain about both their utility relative to the population mean and the mean itself. This latter uncertainty inflates prior variances, which leads consumers to weight their private signals more heavily when updating beliefs. Posteriors are unbiased across products. For a given product, however, the unknown mean yields a "winner's curse'': consumers with high posteriors tend to overestimate their utility. These consumers choose fixed-rate tariffs and lower their usage as they correct their beliefs. The flat-rate bias arises when switching costs deter them from changing tariffs. Using the estimated model, we find tariff menus are ineffective screening devices for price discrimination by an online grocer. Predicted revenues increase by 20 percent when the per-use tariff is dropped, since more consumers choose and stay with the flat rate