Identification in Search Models with Social Information

Abstract

We theoretically study how social information affects agents' search behavior and the resulting observable outcomes that identify search models. We generalize canonical empirical search models by allowing a share of agents in the population to observe some peers' choices. Social information changes optimal search. First, we show that neglecting social information leads to non-identification of search cost distributions under various standard datasets. Whether search costs are under or overestimated depends on the dataset. Next, we propose remedies---such as data requirements, offline estimation techniques, exogenous variations, and partial identification approaches---that restore identification and consistent estimation.

Keywords: Search and Learning, Social Information, Identification, Networks, Robustness.

JEL Classification: C1, C5, C8, D1, D6, D8.