We develop empirical tests for discrimination that use high-frequency evaluations to address the problem of unobserved heterogeneity in a conventional benchmarking test. Our approach to identifying discrimination requires two conditions: (1) the subject pool is time-invariant in a short time...