This article offers an economic analysis of color-blind alternatives to conventional affirmative action policies in higher education, focusing on efficiency issues. When the distribution of applicants' traits is fixed (i.e., in the short-run) color blindness leads colleges to shift weight from academic traits that predict performance to social traits that proxy for race. Using data on matriculates at several selective colleges and universities, we estimate that the short-run efficiency cost of "blind" relative to "sighted" affirmative action is comparable to the cost colleges would incur were they to ignore standardized test scores when deciding on admissions. We then build a model of applicant competition with endogenous effort in order to study long-run incentive effects. We show that, compared to the sighted alternative, color-blind affirmative action is inefficient because it flattens the function mapping effort into a probability of admission in the model's equilibrium.