We develop a model of a Parole Board contemplating whether to grant parole release to a prisoner who has finished serving their minimum sentence. The model implies a simple outcome test for racial prejudice robust to the inframarginality problem. Our test involves running simple regressions of whether a prisoner recidivates on the exposure time to the risk of recidivism and its square, using only the sample of prisoners who are granted parole release strictly between their minimum and maximum sentences and separately by race. If the coefficient estimates on the exposure time term differ by race, then there is evidence of racial prejudice against the racial group with the smaller coefficient estimate. We implement our test for prejudice using data from Pennsylvania from January 1996 to December 31, 2001. Although we find racial differences in time served, we find no evidence for racial prejudice on the part of the Parole Board based on our outcome test.