Using traditional American, race-conscious university admissions as the example, this paper considers the logic that underpins economists' and statisticians' analysis of affirmative action. The logic includes sufficient statistics, statistical discrimination, and comparisons of distributions. I explain how to measure rigorously the extent of affirmative action that is being practiced. Finally, I analyze numerous race-blind alternatives to race-conscious affirmative action, including plans based on historic differences, plans based on socioeconomic data, plans based on secondary school data, plans based on precise geography, and plans that combine all of the aforementioned approaches. I show the extent to which each alternative sacrifices the merit of a university's admitted class in order to accommodate race-blindness.