Does affirmative action reduce the number of black lawyers?

• Author: Ayres, Ian
• Position: Response to Richard H. Sander, Stanford Law Review, vol. 57, p. 367, November 2004

INTRODUCTION I. ESTIMATING THE IMPACT OF AFFIRMATIVE ACTION ON THE NUMBER OF BLACK LAWYERS A. The "White Median Tier" B. The Impact of Eliminating Affirmative Action II. TESTING THE MISMATCH HYPOTHESIS A. The Relative Tier Analysis B. Analysis of Students Who Were Admitted to Their "First Choice" C. Stereotype Threat and Lift III. MAXIMIZING THE PROBABILITY THAT BLACK STUDENTS WILL BECOME LAWYERS IV. HIGH-RISK STUDENT ANALYSIS CONCLUSION INTRODUCTION

Richard Sander's study of affirmative action at U.S. law schools highlights a real and serious problem: the average black law student's grades are startlingly low. (1) With the exception of traditionally black law schools (where blacks still make up 43.8% of the student body), the median black law school grade point average is at the 6.7th percentile of white law students. (2) This means that only 6.7% of whites have lower grades than 50% of blacks. One finds a similar result at the other end of the distribution--as only 7.5% of blacks have grades that are higher than the white median. (3)

Given these low grades, it should not be surprising that black students are less likely to graduate from law school and less likely to pass the bar. In fact, in the LSAC data, (4) 83.2% of whites graduated and passed the bar within five years of entering law school, while only 57.5% of blacks entering law schools became lawyers. Sander has made an important contribution by simply bringing national attention to the racial disparities in law school grades and bar passage rates. This represents another instance of the pervasive pattern of observed black-white disparities in health, occupational, and educational outcomes. (5)

However, beyond merely identifying another set of racial disparities, Sander has gone further by claiming that affirmative action is the cause of the black-white gap in law school grades: "virtually all of the black-white gap ... seems attributable to preferences: virtually none of it seems attributable to race or correlates of race (such as income)." (6)

His core idea, based on the "academic mismatch hypothesis," is compelling in its simplicity: Because blacks tend to have systematically lower entering credentials than the median (white) student, black students learn less than they might have if they had attended schools at which they were better matched, and thus they should be expected to earn lower law school grades and to graduate and pass the bar at lower rates.

Sander claims that the mismatch effect caused by affirmative action is so significant that it actually reduces the overall number of black lawyers. While estimating that affirmative action causes 14.1% more blacks to enter law school, he concludes that the lower graduation rates and bar passage rates of mismatched black students on net reduce the number of black lawyers by 7.9% (relative to the number that would be produced in a system without affirmative action). (7)

While the mismatch hypothesis is plausible, this response refutes the claim that affirmative action has reduced the number of black lawyers. We find no persuasive evidence that current levels of affirmative action have reduced the probability that black law students will become lawyers. We estimate that the elimination of affirmative action would reduce the number of lawyers. Indeed, some of our results suggest an equally plausible "reverse mismatch effect," where the probability of black law students becoming lawyers would be maximized under a system involving an affirmative action program with larger racial preferences than those presently in place. We emphasize, however, that we do not view these estimates as definitive, as they are derived within the simple tier-index score framework offered by Sander. We put them forward to underscore our conclusion that, even within his framework, there is not persuasive evidence indicating that affirmative action is responsible for lowering the number of black attorneys.

However, consonant with Sander's analysis, we do find that attending law school is a risky proposition for some black law students. While Sander argues that most of this risk is attributable to the mismatching effect of affirmative action, we estimate that affirmative action on net reduces (but by no means eliminates) the risk for black law students. Somewhat similar to Sander's analysis, we consider whether there may be a group of black students who in a different equilibrium would not attend law school. But while Sander envisions a world with reduced law school demand for black law students occasioned by the elimination of affirmative action, we consider a world with reduced black law student supply occasioned by better information about the true risk of the undertaking.

This Response is organized into four Parts. The first estimates the impact of various degrees of affirmative action on the expected number of black lawyers. The second explores the mismatch hypothesis in detail. The third identifies the type of school that maximizes the probability that black students (with particular entering credentials) will become lawyers. And the fourth identifies the set of students for whom becoming a lawyer is a more uncertain outcome.

1. ESTIMATING THE IMPACT OF AFFIRMATIVE ACTION ON THE NUMBER OF BLACK LAWYERS

   Sander uses regression analysis to try to establish a core proposition: the probability at matriculation that a student will become a lawyer is dominantly determined by the student's entering credentials (LSAT and undergraduate grade point average) relative to other students at her school. This proposition derives from two regressions suggesting that (1) a student's relative entering credentials determine the student's law school grades, (8) and (2) a student's entering credentials and law school grades determine the probability that the student will pass the bar. (9) But since the student's relative entering credentials determine her law school grades, the probability of passing the bar--especially judged at the moment of admission--is, for Sander, determined by her relative entering credentials.

   An even more audacious claim of Sander is that, after controlling for a student's relative entering credentials, the probability at the moment of entering law school that a student will become a lawyer is not importantly determined by the student's race or any other factors. (10) Sander's argument seems to be that if you know a student's LSAT and undergraduate GPA relative to those of her classmates at the moment she enters law school, you can make the best prediction possible (at that point) about her chances of ultimately graduating and becoming a lawyer. For the most part, our response will accept Sander's claim that other factors do not influence the probability of becoming a lawyer (we are, however, extremely skeptical of that claim (11)).

   But, importantly, Sander does not use these regressions to estimate his bottom line figure that eliminating affirmative action would reduce the number of black lawyers by 7.9%. (12) Instead, he merely groups existing blacks from the LSAC data set into twenty-six distinct ranges of entering credentials and then calculates how many black lawyers there would be if these black law students became lawyers at the same rate as whites with the same entering credentials. Thus, for example, there were 106 black students in the LSAC data that had entering credential indexes with values between 620 and 640. (13) Sander assumes that in a world without affirmative action, 75.6% of these black students would become lawyers--because this is the same rate at which white law students with these entering credentials became lawyers.

   This approach is consistent with Sander's core claims outlined above. In a world without affirmative action, black students would have the same entering credentials as their white peers at particular schools. Since relative entering credentials are the sole statistically significant determinants of the probability that a student will become a lawyer, Sander doesn't need to control for any other fact. But if other factors (including a student's own race or family income or the race of her classmates (14)) determine the probability that a student will become a lawyer, then this approach would not be accurate because it fails to account for any factor other than the student's index score. (15)

   1. The "White Median Tier"

      A casual reader of Sander's article might think that black and white students with the same index score would almost never attend law schools in the same tier: almost all whites within one of these narrow index ranges would go to law schools in one quality tier and blacks with the same credentials would almost always go to law schools in a higher-ranked quality tier. But as shown in Figure 1, it turns out that there is a substantial overlap in the law school tiers attended by blacks and whites with the same entering credentials.

      [FIGURE 1 OMITTED]

      Figure 1 shows the proportion of white and black students with the same entering credentials (using Sander's same twenty-point index ranges) that go to particular law school tiers relative to the tier attended by the median white student with that index score. (16) We retained the tier numbering found in Sander's data set--with elite schools having a tier number of 6 and the least selective law schools (historically black schools) having a tier number of 1. One can see that there is a substantial overlap in the quality tiers of law schools attended by white and black law students. Forty-four percent of whites attend a school in the white median tier, but 26% of black students with the same entering credentials attend a school in this same tier. The figure shows that blacks are more likely than whites to attend quality tiers that are above the white median tier. But there still is substantial overlap: 31% of blacks attend a school one tier higher than the white median tier, but 19% of whites with the same entering...