Is Affirmative Action Responsible for the Achievement Gap Between Black and White Law Students?

Abstract:     
While the Supreme Court upheld some affirmative action programs as constitutional in 2003, the wisdom of affirmative action as a policy decision remains hotly contested. In the law school context, the challenge is to determine how affirmative action policies affect law schools, law students, and the legal profession. This paper takes up one strand of this challenge, estimating how minority students would fare in a world with different affirmative action policies than those currently implemented.

I posit a model of law school performance that controls for entering credentials and allows for a mismatch between student and school (where the student is outmatch by his fellow students). The model also allows for differences in the law school experience for students of different races, which may be the result of discrimination or other differences in the way that law school cultures affect students. The results indicate that, if anything, reverse mismatch boosts the performance of students with low credentials. Using monte carlo simulations of graduation and bar passage with bootstrapped standard errors, I find that removing affirmative action policies decreases the number of new black lawyers each year by 13.4% ± 5.2%. This is in direct conflict with a recent study by Richard Sander that estimates an increase in the number of new black lawyers. Sander, however, assumes that there is no discriminatory effect on law student performance, and therefore confounds discriminatory effects with the mismatch effect in his analysis.

Finally, recognizing that the data upon which I and others rely is imperfect and unable to provide a definitive answer regarding whether the mismatch theory applies in the law school context, I suggest some experimental additions to the data to correct for these problems.

Number of Pages in PDF File: 62

Keywords: affirmative action, simulation, discrimination, mismatch, bootstrap, selection bias

working papers series