Both theoretical and experimental research by economists and social psychologists suggests biases in how people draw inferences about the preferences and characteristics of others. Eyster and Rabin (2005) review extensive experimental evidence from a broad array of settings that suggests people do not fully take into account how other people’s actions depend on their private information and, thereby, systematically underestimate the relationship between others’ actions and types. Using application and admission data from two liberal arts colleges with optional SAT I policies, this paper is the first to use non-experimental data to estimate the extent to which players underestimate this relationship. This policy provides applicants with a choice of whether to voluntarily disclose their SATI scores to the college. Our empirical estimates are consistent with the experimental research – colleges underestimate the relationship between an applicant’s action (not submitting) and type (actual SATI score). In practice, the underestimate has very little effect on the student body composition because the colleges have a great deal of other information to base their acceptance decision and to infer SATI scores. In other private information environments with less public information, a player would likely incur greater costs associated with incorrectly inferring the other players’ private information based on their actions.