Barnes, K Y --- "Is Affirmative Action Responsible for the Achievement Gap between Black and White Law Students" [2007] LegEdDig 49; (2007) 15(3) Legal Education Digest 35

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

K Y Barnes

[2007] LegEdDig 49; (2007) 15(3) Legal Education Digest 35

101 NW U LRev, Fall 2007, pp 1759–1808

In Grutter v. Bollinger, the Supreme Court upheld some affirmative action programs in legal education as constitutional. The wisdom of affirmative action as a policy decision, however, remains highly contested. The challenge is to determine how affirmative action policies affect law schools, law students, and the legal profession. Several scholars have provided nuanced descriptions of the effects of affirmative action on students and graduates at specific elite institutions. These studies demonstrate that the affirmative action policies at these schools are generally consistent with the stated policy goals of affirmative action. Namely, these studies find that students value diversity in the classroom, that black and white graduates have equally successful careers after law school, and that, in some situations, black graduates give back more to the community than white graduates do.

In a recent article in the Stanford Law Review, Professor Richard Sander focused on one such datum: Half of black law students are in the bottom tenth of their class. Based upon his analysis, Sander predicts a 7.9 per cent gain in the number of black lawyers absent affirmative action. According to Sander, this counterintuitive result is the product of the ‘mismatch hypothesis.’ The mismatch hypothesis posits that, as a result of current affirmative action policy, blacks systematically are admitted to and attend law schools for which they are under-qualified.

There are three potential explanations for the gap in achievement between black and white students as measured by law school grades. First, the mismatch theory suggests that blacks are outmatched by their classmates, disengage from the learning process, and receive poor grades as a result. Second, it may be that the law school atmosphere creates substantial barriers to high achievement for black students. Third, black law students may simply be performing to their potential, as measured by their lower incoming credentials. Under this third theory, black law students have not disengaged from the learning process and would not learn more at a less demanding school, nor do they face special barriers to achievement.

If correct, each of these theories would require a different calculus for the appropriate response. First, the mismatch theory implies that law schools are doing a disservice to black students by admitting them when they will likely fail, but would have done better elsewhere. One straightforward policy change on this basis would be to refuse to admit such under-qualified students by removing the ‘boost’ that affirmative action provides black applicants in the admissions process. Another policy change would be to recognise that it is not the under-qualification, but the disengagement with learning that makes mismatch a bad deal for black students. Schools could therefore implement academic outreach programs to keep the students engaged. Second, the barriers theory suggests that law school culture must change to remove race-based barriers to achievement. By race-based barrier theory, I do not intend to invoke solely a description of intentional discrimination. Instead, I mean to describe the myriad ways that law school atmosphere can be uninviting to black law students, whether intentionally or not. Finally, if neither mismatch nor the race-based barriers theory has a significant impact on black law student achievement, then schools must determine as a normative matter whether the appropriate response is to change nothing. One could argue that while not directly hurting black law students, affirmative action perpetuates the stereotype that black students are less able than their white classmates.

Schools could also provide additional support services for struggling students, or simply acknowledge that while stereotype reinforcement might be a small effect, the pedagogical and societal advantages of diversity are more important.

In order to clarify issues that often can be muddled in the context of affirmative action, I first describe the mismatch theory in more detail. The mismatch theory is that students who are outmatched in class disengage from the learning process and, in turn, do not learn as much as those students with the same credentials who matriculate to schools at which they are not outmatched. Note, however, that mismatch has nothing to do with race: All students who matriculate to a law school where their classmates outmatch them are vulnerable to disengaging and falling behind. Sander’s critics implicitly discuss mismatch as if it is a problem only for black students because of affirmative action, but affirmative action and the mismatch theory are not necessarily intertwined. Any preference based on something other than student credentials can potentially lead to mismatch. Indeed, mismatch theory may apply to the very bottom group of law students in any class, no matter the details of the admissions policies.

Ironically, considering all the ink spilled over testing the mismatch theory more appropriately than Sander does, it is only Sander’s analysis that tests mismatch without using race. Sander does not, however, test the interaction of credentials and schools. Instead, he tests only whether the two variables separately affect student performance, implicitly assuming that all students who matriculate to a particular school get the same benefit of higher performance, regardless of credentials. While Sander does not test the mismatch theory by using race, he implicitly assumes that only minority students can be mismatched, and thereby also falls prey to the red herring that the mismatch theory somehow relates to race. In addition, Sander’s analysis is seriously flawed because he assumes, without adequate empirical support, that discriminatory law school culture does not influence black achievement in law schools.

While the mismatch theory is unrelated to race, the race-based barriers theory is explicitly related to race. The phrase ‘race-based barriers,’ as I use it, is a somewhat fluid concept that incorporates many types of behavior. I use this phrase to signal that the specific cultures of individual law schools are quite different and, therefore, treatment that differs across race (intentionally or not) may also be quite different across schools. Race-based barriers include: hostile learning environment, direct discrimination in outcomes, and other issues like stereotype threat. Most of these race-based barriers are school-specific and are based upon the individual law school’s culture. However, other barriers, such as stereotype threat, implicate a broader societal stereotype and therefore generally are not school specific.

Because race-based barriers may be general or school-specific, statistical testing of discrimination should allow for both possibilities. Statistical tests can accomplish this by testing race school interactions for school-specific race-based differences, as well as race main effects (which measure the average performance gaps between racial groups across all schools) for more general race-based differences. Unfortunately, without data on the cultures of specific schools or students’ experiences during law school, one cannot differentiate between the different types of barriers listed above. The question one can ask is this: After controlling for possible mismatch and student credentials, does race predict performance? Put this way, it is clear that the test cannot determine whether race-based barriers exist; it can only demonstrate whether race is an important predictor of performance. If race is an important predictor of performance, then race-based barriers might be the cause.

An alternative explanation for race as a predictor of performance is that unmeasured student credentials differ across race. If this is the case, these unmeasured credentials will be confounded with race, making the variable of race appear to be an important predictor of performance when, in fact, it is the unmeasured student credentials that are important predictors. Because the data available focus primarily on pre-law school admissions and post-law school outcomes, they provide little information about student experiences during law school. Thus, only a rough statistical test of the race-based barriers theory is possible, even under the assumption that unmeasured student credentials are not strongly correlated with race.

In order to develop a reasonable measure for the rank of school attended, which testing the mismatch theory requires, I am forced to follow earlier researchers by using what I have named ‘school type’ as a proxy for rank. I recognise that the results have limited applicability as a consequence. The school types, based on the six clusters in which LSAC researchers grouped the schools, are best described as Small Top 30 law schools, Large Top 30 law schools, Mid-range Public law schools, Mid-range Private law schools, Lower Ranked law schools, and Historically Black law schools. I present them here in the rank order used by Sander and his critics, which is based upon the average LSAT score for each cluster. Given these descriptions, there appears to be significant overlap in the ranks of schools between the top two clusters (Small Top 30 and Large Top 30 law schools) and the mid-range clusters (Mid-range Public and Mid-range Private law schools). Because such overlap can mask the effects of mismatch, I collapse the school types into four: Top 30 Schools, Mid-range Schools, Low-range Schools, and Historically Black Schools.

Second, the data do not contain information about all possible student credentials. The two objective variables included in the LSAC data are LSAT score and UGPA.

Flexibility when using LSAT and UGPA to measure credentials is important because it allows LSAT and UGPA to affect performance in nonlinear ways.

Thus, I achieve flexibility without assuming that LSAT and UGPA have a specific linear relationship to student credentials.

As the LSAT is explicitly set on a nonlinear curve and many undergraduate institutions do not grade linearly, this is necessary to model student credentials accurately.

Thus, while somewhat problematic, I have devised the three variables - race, school type, and credentials - that I will use to test the mismatch and race-based barriers theories and how these theories affect various performance measures. I focus on three performance measures: bar passage, graduation, and obtaining a well-paying first job after law school. Although these performance measures are by no means exhaustive, they provide a snapshot of potential performance measures, are readily available in the data, and measure performance on goals that many law students share.

The results show little support for the mismatch theory. For the mismatch theory to be correct, the change in graduation rate should decline as one compares similarly credentialed students at higher ranked schools. Thus, according to the mismatch theory, for students at any given credential level, Historically Black Schools should have the highest graduation rate and should exhibit a large positive change in graduation rate when compared to the baseline graduation rate of white students at Mid-range Schools. Low-range Schools should have a smaller but still positive difference in graduation rates from the baseline. Finally, Top 30 Schools should have a negative difference in graduation rates from the baseline.

None of the comparison groups — with credentials at the fifth, tenth, twenty-fifth, or fiftieth percentiles — demonstrate this pattern.

The results indicate that race does matter, likely because different types of schools have different law school cultures that affect the learning experience of minority students. In particular, for all school types except Top 30 Schools, black students with the same credentials as white students are less likely to graduate.

The results also show that minority students attending Historically Black Schools are less likely to graduate than white students. In general, the race-based differences are smaller than the differences between school types, suggesting that the school type at which a student matriculates is more important to success than the race of the student.

Moving to the results for bar passage, the overall test of the school type credentials is statistically significant. Table 2A provides the results for each pairing, and provides strong evidence that law school is a risky proposition for law students with low-percentile credentials, assuming that those students hope to pass the bar exam.

Table 2A also demonstrates a consistent reverse mismatch trend: the lower ranked the school type, the lower the probability that a student will pass the bar exam. No matter what a student’s credentials, students matriculating to Top 30 Schools are most likely to pass the bar, while students at Historically Black Schools are least likely.

Turning to Table 2B and the race-based barriers theory, the statistical test again finds that race and race school type interactions are significant. As with graduation rates, Historically Black Schools demonstrate the largest difference between white and minority student bar passage rates, with minority students having a bar passage probability of 7.1 to 26 percentage points lower than white students. These drops in bar passage rates are very different across races, making it clear that law school cultures affect different minority groups differently. In contrast, for Mid-range Schools, the drop in bar passage rates for minority students is more consistent across races (from 5.8 to 8.6 percentage points). Once again, minority students attending Top 30 Schools are the least likely of all minority students to experience a racial barrier to bar passage. As with graduation status, the race-based differences are smaller than the school type differences on average.

Finally, I test the mismatch and race-based barriers theories with respect to obtaining a well-paying job.

This result provides strong support for Wilkins’s argument that affirmative action provides access to elite jobs for minority students, particularly at the most elite schools, where the probability of obtaining a well-paying job is largest by far.

For each given set of credentials, black students have higher graduation rates and bar passage rates at higher ranked schools. These effects are consistent across different levels of credentials and for every school type.

In sum, the three counterfactual models presented in this paper refute Sander’s mismatch theory that students are more successful at lower ranked schools. In fact, the three models suggest the opposite: Matriculating to higher ranked schools provides significant benefits. The results regarding race-based barriers are more mixed. In some cases, there seems to be at least the suggestion of some sort of race-based barriers; the results for Historically Black Schools are particularly stark. Mid-range Schools also have consistently lower bar passage and graduation rates for minority students. Top 30 Schools have fewer differences, except with respect to well-paying jobs, where they provide a significant boost for black students in particular, but also to a lesser extent for Hispanic and Asian students.

In this Part, I move beyond current practice to investigate what might happen under alternative policies of affirmative action.

Before proceeding to the analysis, it is important to state my assumptions. I assume that the models of law student performance I estimate in Part I still hold. This is equivalent to assuming that any mismatch or race-based barriers problems that are currently at play will not change. These assumptions do not perfectly mirror the more complicated real world changes that would occur if the alternative affirmative action policies were implemented, but they provide conservative estimates of the results that would stem from each potential policy. The projected effects on black students will accordingly be underestimated.

In his article, Sander argues that affirmative action triggers mismatch, which in turn causes black students to perform poorly. As the graduation and bar passage inquiries in Part I demonstrate, one’s level of achievement in and after law school depends in part on the type of school attended. Sander’s argument therefore sounds quite logical: Affirmative action changes the type of school black students attend; LSAT scores, UGPA values, and school type predict grades; and grades predict bar passage, so affirmative action, by changing the type of school for black students, affects bar passage.

As a technical matter, this logic does not hold. In a series of unlinked regression models, if A predicts B, and B predicts C, one cannot conclude, as a matter of logic, that A must predict C. Neither Sander nor his critics test directly what would happen absent affirmative action. Relying on Sander’s assumptions that admission and matriculation patterns would not change in an alternative world without affirmative action, Ayres and Brooks calculate that changing admissions policies to eliminate affirmative action would result in a loss of 12.7 per cent or 9.4 per cent of new black lawyers each year, depending on the model used. Chambers and his co-authors perform a calculation allowing for more realistic and complex changes in application and admission patterns. This calculation suggests that eliminating affirmative action would result in a loss of 30-40 per cent of new black lawyers each year. Sander estimates that removing affirmative action would increase the number of new black lawyers by 7.9 per cent. None of these researchers provide a confidence interval or other measure of the variability of their estimates. Therefore, each fails to provide a test to determine whether the number of new black lawyers would be systematically different with and without affirmative action.

Instead of making the assumption that ending affirmative action will make black students’ law school experiences identical to that of white students, I assume that black and white student performance is predicted by the same models I report in Part I. Students’ performance will change if school type, a significant predictor of performance, changes because of the interaction between school type and credentials. However, my assumption is that the general prediction mechanism, as estimated by the performance models, does not change. Race-based barriers and mismatch (or reverse mismatch) will therefore continue to predict performance to the same extent as under the current affirmative action regime.

I investigate four different treatments: the current affirmative action policy; a policy of no affirmative action; a less expansive policy of affirmative action that I deem ‘affirmative action light,’ which provides about half the boost that current affirmative action plans do; and finally, a more aggressive policy of affirmative action that I deem ‘affirmative action plus,’ which provides about double the boost that current affirmative action policies do.

The first result to note is that the two more intensive affirmative action policies produce more black lawyers than a policy of no affirmative action. Overall, there would be a drop in the number of new black lawyers from the 845 lawyers produced by the current system to 732 if affirmative action were no longer an element of law schools’ admission decisions.

Another interesting result is that boosting the school type at which successful black applicants matriculate (using the affirmative action plus simulation) does not have a significant effect on the number of new black lawyers.

This basic pattern of results is replicated in the simulations of graduation and well-paying job status. With current affirmative action policy, 1492 black students, or 81.9 per cent of black matriculants, graduate from law school within five years of entering. Without affirmative action, the number of graduates is drastically reduced to 1152. However, because the number of matriculants decreases significantly, the percentage who graduate goes up to 84.1 per cent. Overall, the loss of 337 black graduates is statistically significant at the 0.01 level, and represents a 22.6 per cent drop in the yearly number of new black graduates, with a 95 per cent confidence interval of 22.6 per cent +/- 2.4 per cent. Again, the results for affirmative action light lie somewhere in between these two policies. Under affirmative action light, the annual number of black graduates is significantly lower than under the current affirmative action policy and significantly better than under no affirmative action policy. Affirmative action plus is not significantly different from the current affirmative action policy in terms of the number of new black law graduates, which suggests that admitting the same black matriculants to school types of higher rank does not significantly affect the number of black law school graduates.

Moving to no affirmative action would drop the number of black graduates who obtain well-paying jobs from 297 to 228 — a decline of 23 per cent. This decrease is statistically significant at the 0.01 level, with a 95 per cent confidence interval of 23.0 per cent +/- 19.6 per cent. Once again, affirmative action plus does not have a statistically significant effect. Although affirmative action plus provides an increase of almost 10 per cent in the number of black graduates obtaining well-paying jobs, the standard errors are large enough that this increase is not statistically significant.

How can one reconcile these results with the very real data that suggest blacks’ law school grades are significantly worse than those of whites? My results focus on outcomes — graduation, bar passage, and first jobs — rather than law school grades. Although black law students receive lower grades on average, it appears that this single marker of achievement is inadequate to comprehensively describe black law student performance. At the very least, my results give some reason to doubt Sander’s regression that purports to demonstrate that law school grades are the primary determinant of bar passage. At least with respect to the three performance measures I investigate, my results demonstrate that the properly modeled data do not support the inference that affirmative action triggers mismatch for black law students. Moreover, the results demonstrate a real cost to eliminating affirmative action. Ending affirmative action would lead to 13.4 per cent fewer new black lawyers, 22.6 per cent fewer new black law graduates, and 23 per cent fewer black law graduates with well-paying jobs. The results presented here are not definitive because they suffer from the same data limitations as the studies of Sander and his critics, but they provide strong evidence that affirmative action has significant benefits and that the evidence of negative consequences Sander provides is highly suspect.

While researchers may point to other problems with the LSAC data, these three problems are the primary stumbling blocks to accurate assessments of both the mismatch and race-based barriers theories of student performance.

The challenge is to determine what in law school culture helps students, particularly minority students, thrive. The first step in this determination is to map law school culture: How do students, faculty, and staff describe the environment, and how do they experience law school? One ongoing project, the Educational Diversity Project, is surveying a large number of individuals at different law schools to investigate how diversity in law schools affects students. This is a very important step in determining what individuals subjectively believe affects performance, but it would also be advantageous to take the descriptions of law school atmosphere and correlate them with student performance. If one controlled for student credentials at the same time, a researcher could determine what about law school culture helps minority students succeed or, perhaps more importantly, fail. Although looking at both student credentials and law school culture would greatly expand the scope of the project, it would provide significant benefits by helping to determine why minority students’, and particularly black students’, achievements in law school lag behind their white classmates’ achievements.

This Essay attempts to clear away the confusion surrounding the mismatch theory and its relationship to affirmative action policies. Despite much confusion to the contrary, I clarify that race is unrelated to the mismatch theory, and describe appropriate tests of the mismatch theory that flow from that argument.