Impressive Data, Inconsistent Claims
Curtis Crawford, www.DebatingRacialPreference.org

In the previous posting to the NAS Online Forum, I argued that Professor Richard Sander's sensational claim, that the black students in the Bar Passage Study got lower grades than they would have received without racial preference, is at war with his evidence. On the contrary, black grades were the same or somewhat better than those of whites with similar Academic Index scores.

But grades are only one indication of academic performance. Two other important measures are graduation and passing the bar exam. On these measures, how did blacks compare with equally skilled whites? In both cases, their performance was much worse. Concerning graduation, the BPS data show that for each 60-point interval of the Academic Index, the proportion of blacks not graduating was always larger than the proportion of whites, usually by 25% or more. Concerning the bar, the proportion of blacks failing the first attempt was larger in every Index interval, often by 80-100%. [Tables 5.7, p 441, and 6.2, p 446]

What explains this surprising change in the relationship between skills and performance? Concerning grades, equal black and white skills resulted in nearly equal black and white first-year GPAs. Concerning graduation and the bar, equal black and white skills resulted in much lower black rates. Sander writes as if this were no surprise: "black attrition rates are substantially higher than white rates, simply because racial preferences advance students into schools where they will get low grades." [p 441]

Certainly, black grades were low, much lower, than the grades of whites overall. And this would help explain why black graduation and bar passage rates were lower than white rates overall. But black grades were not lower than the grades of whites with similar Index scores. Thus, taking graduation first, the puzzle is: Since blacks' grades were as good or better, why were their graduation rates much worse than those of whites with equal Index scores? Obviously, it cannot be because black grades were low, since they were no lower than those of the whites with whom they are being compared.

In a regression analysis, Sander finds only two strong predictors of graduation rates: one's first-year GPA, and the selectivity of one's school. The former was three times as powerful as the latter. [Table 5.6, p 439] This finding only adds to the mystery. However accurate a predictor in other situations, first-year GPA totally misfired here. It predicted that black students whose grades were on average equal to those of white students would graduate at an equal rate. In fact, they graduated at a much lower rate. The explanation for this strange effect must lie in another cause.

Sander's other strong predictor is the selectivity of the school one attends. According to his figures, the more selective one's school group, the higher one's chance of graduating. [Table 5.5, p 437] At first glance, this might seem to favor blacks over whites with the same Index scores, since affirmative action places the former in much more selective schools than the latter. But the figures on school selectivity and graduation rates compare blacks with whites overall, not with whites having equal Index scores. Thus, they provide no basis for a prediction, one way or another, about the graduation rates of blacks and whites with similar Index scores. We are left with a major puzzle, unsolved and unacknowledged, at the heart of the analysis.

A sister puzzle is why blacks' first-time bar passage rates are much worse than those of equally skilled whites. As with graduation rates, Sander sees no puzzle:

Blacks with an LSAT-UGPA index score of, say, 600 will tend to end up at much more elite schools than will whites with index scores of 600, but as a result, the blacks will end up with lower law school grades . . . . The net effect will be a markedly lower bar passage rate. [p 445]

This explanation contradicts his own data, which show that black students, whatever the disadvantages of attending "more elite schools," did not "end up with lower law school grades" than whites with similar Index scores. That these black students, nevertheless, had lower bar passage rates than their counterparts cannot be explained by "lower black grades," since their grades were not lower than the grades of whites with similar Index scores. When there is a substantial difference between two groups (e.g., rates of graduation or bar passage), it can't be explained by a factor that is the same or nearly the same in both groups (e.g., distribution of grades).

For law students overall, there is no reason to doubt the high correlation Sander found between grades and bar passage. But however great the correlation, it would be unable to explain why students with nearly the same grades had very different bar passage rates. Other variables were regressed, but their power to solve the puzzle I describe was not tested.

What if the variable hypothesized were not grades per se, but their distance below the class median? The BPS data amply substantiate Sander's contention that black students were boosted by racial preference into much more selective schools than they would otherwise have attended. In this situation, although black grades were not below the grades of whites with equal skills, it is quite plausible that they were much farther below the class median. This, in turn, might well have resulted in a level of discouragement with studying law and being a lawyer that led to lower rates of graduation and bar passage.

If this hypothesis proved true, it might explain why the BPS black students, despite having grades as good or better than those of whites with equal Index scores, had much lower rates of graduation and bar passage. But it would not vindicate Sander's claim that racial preference for black law students caused them to get lower grades than they would have received without it. Such a claim has no more foundation in his data than a claim that pro-black preference caused whites to get lower grades than they would have received without it.

---

Sander also finds (sensationally and paradoxically) that ending affirmative action would increase the annual number of black students passing the bar. This has been vigorously disputed by scholars who argue that Sander's numbers are way off, overestimating the number of blacks who would apply, and greatly overestimating the proportion who would be admitted. I address only one aspect of his finding, but it is fundamental.

The problem documented in this essay is a huge gap in academic skills between most black and white law students who matriculated in 1991. Sander argues that this academic mismatch, caused by preferential admissions, was noxious for black students; and thus that racial preferences in law school admissions do blacks more harm than good.

Sander believes that law school racial preferences in 1991 were so large that without them the number of blacks admitted would have dropped by 50% or more. But he seems to assume that the preferences have now become much smaller. He accepts an estimate that without preferences in 2001, the cut in black admissions would have been only 14%.

If the degree of preference has indeed grown so small, doesn't it follow by Sander's logic that the academic mismatch has also become much smaller, and much less noxious? If both these changes are true, abolishing affirmative action would be a lot easier, but also a lot less helpful to blacks. Unfortunately, we know nothing from the data in this essay about the size of the academic mismatch in 2001 and later, and very little about the amount of current preference.

Sander needs to choose. Either the BPS data are basic to his analysis, his concern, and his prescriptions, or not. If they are, he should let them determine not only his arguments for halting racial preference, but also his estimates as to what would happen to blacks if his proposals were adopted. If, on the other hand, he thinks the mismatch much smaller now, he needs to assemble the factual basis for this judgment, so that the reader can decide whether it warrants the changes proposed. Otherwise, he faces two ways: using first the grim, BPS reality to magnify the evils of racial preference in law school admissions, and then trusting an unexamined, supposedly transformed present to minimize the cost to blacks of ending the preference. This seems more like a sales strategy than social science.

NAS6:46 PM

Impressive Data, Wayward Logic
Curtis Crawford, www.DebatingRacialPreference.org

"Law school racial preferences give blacks fancier degrees, but also systematically lower their GPAs." So concludes a fascinating study by Richard Sander, a law professor at UCLA. Entitled "A Systemic Analysis of Affirmative Action in American Law Schools," it appears in the current issue of the Stanford Law Review (Vol 57: 367-483), and online HERE.

More sharply and fully than anything I have seen, the analysis documents enormous academic gaps, in both skills and performance, between black and white law students. Sander blames the gaps squarely on huge racial preferences in admissions. Due to these preferences, he contends, black students with mediocre skills are placed in schools with white classmates whose skills are far superior. The result: Most black students cluster at the bottom of the class in grades and quality of work.

Sander's primary database is the Law School Admission Council's thoroughgoing Bar Passage Study (BPS), involving 27,000 students who matriculated in 1991. The Study classified 184 American law schools into six groups according to academic selectivity, and followed the students for five years through graduation and examination for the bar.

The BPS data mercilessly confirm the width of the racial gaps in skills and performance -- creating a huge academic mismatch. And they clearly support Sander's view that "Law school racial preferences give blacks fancier degrees." However, they contradict his sensational verdict that the preferences "systematically lower [black] GPAs."

To measure the academic skills of law students, the Bar Passage Study used their LSAT scores and undergraduate grades. Sander collated this information into an Academic Index for each student, weighing their LSATs at 60% and their UGPAs at 40%. You may have thought that the black-white gap measured by such an index would be largest near the top, say, in the fourteen "very elite schools" of Group 1, or in the sixteen "other national schools" of Group 2. If so, you were wrong.

The gaps in Group 3 (50 midrange public schools), Group 4 (50 midrange private schools), and Group 5 (18 low-range private schools) were every bit as large as in the top two groups. In all but Group 2, the gaps were larger than two standard deviations. This means that 50% of the blacks in Group 1 had an Academic Index score below 99% of the whites; in the other groups, 50% of the blacks, respectively, scored below 97%, 99%. 98%, 99%, and 89% of the white students. [The data for this and the two next paragraphs are in Sander's Table 3.2, page 416.]

You may also have thought that if racial preferences had not existed, blacks who attended top schools would still have been eligible, on the merits, for admission one or two tiers down. True for a few, but not for most. On the contrary, the boost implied by the BPS data averages three or four tiers.

Although individual scores within large groups of students vary greatly, the level of one group may be roughly compared with the level of another by taking the median score in each. (The median is the point that half the group scores above, and half, below.) The median Academic Index score for blacks admitted to the fourteen Group 1 schools was 705 (vs. white 875). To find a similar white median score (725), one must descend three tiers to the 50 "midrange private" schools in Group 4. For blacks admitted to the sixteen Group 2 schools, the median Index score was 631 (white 805). A similar white median score (641) is only available four tiers down in the nine "historically minority" schools of Group 6.

These figures indicate a huge mismatch between the vast majority of black students and the academic setting in which they studied law. It seems reasonable to expect that most students would study harder, learn more, and get better grades in classes where their skills are typical, rather than overwhelmingly inferior. Wouldn't black (or any other) students with an Index score of 705 get higher grades in a "midrange private" law school than in a "very elite" school?

Such expectations inform Sander's hypothesis that affirmative action, by boosting blacks into tougher law schools than they would otherwise have attended, caused them to get lower grades than they would otherwise have earned. However, his data not only fail to substantiate this hypothesis; they provide powerful, contrary evidence that he roughly affirms and then ignores. Moreover, he fails to consider aspects of affirmative action that could raise the grades of many beneficiaries.

Some of the contrary evidence appears in Sander's Tables 5.1 and 5.3 [pp 427, 431] . They show an unexpected result: The black-white gap in first-year grades was smaller than the gap in Index scores, in every school group. In other words, despite the putative handicap of affirmative action, black students' first-year academic performance was somewhat better than that of white students with similar academic skills. Thus, in Group 1, 50% of the blacks had a GPA below 95% of the whites. In the other groups, the grades of 50% of the blacks were, respectively, below 93%, 95%, 92%, 93%, and 76% of the white students. In most cases, the black/white difference between index scores and first-year grades is just a few percentage points, but it always favors the black students.

Sander's characterization of his BPS evidence is that the gap in racial performance was "essentially" the same as the gap in racial skills. This concedes that the grades of blacks under affirmative action were no worse than those of equally skilled whites, though it fails to acknowledge that the black grades were somewhat better, and in every school group. Further evidence for his characterization is introduced from a second data base, involving 4000 students at 20 law schools. Controlling for these students' pre-admission LSATs and UGPAs, he finds no statistical difference between black and white first-year law school grades. [Table 5.2, p 428]

By itself, such evidence does not foreclose the possibility that the net effect of racial preference is to depress black grades. If not "handicapped" by such preference, perhaps black law students would have outpaced whites with equal skills even more than in Sander's databases. If he has any evidence to this effect, he should have presented it. I know of none.

It seems more likely that some aspects of affirmative action were unfavorable to black academic performance, while others were favorable. Unfavorable would be the unintended consequence of lower grades at tougher schools. Favorable might be the special treatment-enhanced consideration, extensive tutoring, generous financial aid, and perhaps less rigorous grading-that affirmative action beneficiaries commonly receive.

Sander's hypothesis that affirmative action has the net effect of depressing black law-school grades is contradicted by his data. His causal analysis misled him by stressing a factor that should depress grades, while ignoring factors that could raise them.

NAS12:40 PM