A test of 'residual discrimination' in prison commitment rates and time served is performed. A large and statistically significant racially discriminatory gap in time served is found. Black Federal offenders would serve 5.6 fewer months in prison if treated the same as white offenders. The residual racial gap in punishment, however, would be deemed 'efficient' if its elimination would increase black crime rates. A conceptual test is performed that shows in fact the residual gap in inefficient. There are similarly found to be racial differentials in preprison employment disadvantage, although elimination of these differentials would have but a small narrowing effect on the racial gap in post-prison recidivism. Tabular data and 24 bibliographic listings are provided. (Author abstract modified)