Identification in Search Models with Social Information

We study the identification of empirical search models when agents access social information, i.e., peers’ choices and experiences. We show that social information changes optimal search and the observable model implications. Consequently, neglecting social information, even in its simplest form, leads to misspecification of the search model. Depending on the dataset, search frictions are underestimated or overestimated. Next, we propose partial identification approaches to recover robust bounds on search cost distributions while imposing minimal assumptions on social information. Stronger assumptions or data requirements help refine these bounds. We evaluate the implications of our findings for quantifying welfare, demand, and counterfactuals.