Following G. Becker's (1993) suggestion that tests for discrimination should attempt to infer whether profits differ for products sold to minorities and nonminorities, this article tests the hypothesis that racial discrimination affects market prices of auto insurance in Missouri. Compared with tests for discrimination in lending markets, the authors' results are less susceptible to bias from omitted variables. Controlling for available demographic and coverage-related factors, they do not find that loss ratios at the zip-code level are negatively related to percent minority population. This finding is inconsistent with the hypothesis that racial discrimination increases premiums relative to expected claim costs for minorities. Copyright 1998 by University of Chicago Press.