Amicus Brief on School Integration

I've read through the amicus brief signed by 500+ social scientists in the Supreme Court's upcoming school integration case (which will decide whether Seattle can consider a student's race in determining whether that student can transfer to another school). The brief broadly supports both the goal of school integration and the alleged necessity of a Seattle-type plan to achieve integration.

I was interested in some of the claims that the brief makes (mostly in a lengthy Appendix that purportedly summarizes research on school desegregation).

At App. 13-14, the brief makes this point:

Reviews of early desegregation research lead to the conclusion that school desegregation has a modest positive impact on the achievement of African-American students.⁴⁰ Specifically, desegregation appears to have a positive impact on reading achievement, but there appears to be little or no effect on math scores.

In footnote 39, the brief further claims: "Most school reforms have little or no effect on improving students’ outcomes . . . . Thus, the modest impact that desegregation has had on student achievement relative to these other reforms is substantial."

But if you look at the source that is listed first in footnote 40, the evidence doesn't seem so compelling. That source is: Thomas D. Cook, “What Have Black Children Gained Academically From School Integration?: Examination of the Meta-Analytic Evidence,” School Desegregation and Black Achievement, ed. Thomas D. Cook et al. (Washington, D.C.: Department of Education, May 1984).

In that article, Thomas Cook actually found "that reading effects [were] positive but quite small and not educationally significant in all but a few studies." Id. at 60. Cook then went on to say, "the studies . . . tell us nothing about whether segregation created the Black-White achievement gap, but they do tell us that [integration] by itself will not close it to any important degree." Id.

How big was this "quite small and not educationally significant" improvement in reading achievement? Between two and six weeks worth of instruction. That's all that Cook could point to as the academic benefit of desegregation. And that's what the social scientists' amicus brief is now telling the Supreme Court amounts to a "substantial" benefit compared to other education reforms.

Indeed, if you look at the other articles published in the same volume that Thomas Cook edited -- several of which the brief also cites in footnotes 40, 41, and 43 -- the picture is even less convincing. Here's a review article from the time that summarizes the various findings:

The volume reveals a remarkable convergence about the fundamental question.

Armor decided that "the conclusion is inescapable: the very best studies available demonstrate no significant and consistent effects of desegregation on black achievement." Walberg concluded that "school desegregation does not appear to prove promising in the size or consistency of its effect on learning of black students." Stephan decided that ". . . [T]he magnitude of these effects translates into rather trivial increase of about twenty points on the typical SAT." Wortman found a "two-month gain or benefit for desegregated students." Cook decided that all the analyses taken together justified four conclusions: "(1) desegregation does not decrease the achievement of black children; (2) it probably does not increase math achievement; (3) it probably raises reading scores; and (4) the increase in reading scores is somewhere between .06 and .16 standard deviation units or about two and six weeks."

* * *

Cook and Stephan pointed out that in the area of reading, desegregation was associated with negative effects on black student achievement in programs of mandatory desegregation (four studies) but generally positive effects in programs of voluntary desegregation (fifteen studies). Armor and Miller observed that when rigorous studies show positive effects, it is likely that such gains are not due to racial mixing but rather to any changes in educational programs that may have occured simultaneously with disegregation. . . . Stephan pointed out, among other things, that blacks in desegregated schools may have more anxiety with regard to achievement because of negative comparisons of themselves with white students; and Miller pointed out, among other things, that while white students seem to accept black students who are academically equal, white acceptance does not cause black academic achievement.

The amicus brief also cites another paper by Rita Mahard and Robert Crain, which also appears in the 1984 book. That Mahard/Crain paper is supposed to support this conclusion: "The impact of desegregation on achievement varies by context, appearing somewhat stronger for younger students."

But the amicus brief does not mention the fact that Mahard and Crain's conclusion here was criticized by other researchers in that same 1984 volume:

Cook, Armor, Wortman and Walberg all commented on and criticized the validity of the Crain and Mahard contention that desegregation can have a significant effect on black student achievement provided that it begins at the first grade or before. The four agreed that many of the studies upon which Crain and Mahard based this contention had serious methodological flaws, the most important of which usually had to do with the fact that the pre-tests were given to different groups and the fact that pre-and post-tests measured different things, thus making genuine comparisons of the test results impossible. Both Armor and Cook pointed out that when the methodologically-weak studies were eliminated from Crain and Mahard's analysis, their conclusions were roughly equivalent to the conclusions of the panel.

Overall, it seems fair to say that the brief gives an overly-rosy depiction of the very academic research that it cites on the question whether desegregation improves black academic achievement.

Also, I think it's incomplete, at best, for the brief to suggest, as if it's beyond debate, that a reading gain of "somewhere between .06 and .16 standard deviation units" is "substantial" compared to other school reforms. Paul Peterson and William Howell found, for example, that in a New York City voucher program, "African Americans, and only African Americans, posted significant and positive test score gains associated with attending a private school that in year three ranged from one quarter to two fifths of a standard deviation, depending upon the model estimated." In other words, if this study is correct, the effect of vouchers could be 2 to 4 times as great as the effect of desegregation.

UPDATE: One more recent study that the amicus brief could have, but didn't, cite is by David Card and Jesse Rothstein. (Yes, the same David Card who has written with Alan Krueger on the minimum wage.) Card and Rothstein find that:

Our empirical analysis leads to two main conclusions. First, although black relative test scores are negatively correlated with measures of school segregation, when controls are added for neighborhood segregation the effects of school segregation become uniformly small and statistically insignificant. Specifications that instrument either segregation measure confirm this pattern. Second, neighborhood segregation has robust negative impacts on black relative test scores, though these also fall in size and significance once controls are added for the relative exposure of black and white children to neighboring families with differing income, education, and marital status. Thus, as suggested by Wilson (1987), race may not be the primary source of neighborhood segregation effects: rather, racial segregation may proxy for relative exposure to economically successful neighbors.

* * *

The finding that school segregation has no effect on black relative achievement once
controls are added for neighborhood segregation leads us to consider three potential mechanisms that might confound the true effect of school desegregation: unobserved differences in school quality, unobserved differences in schoolmate characteristics, and within-school segregation. Using data from the CCD and the Schools and Staffing Survey, we conclude that there is no strong relationship between school segregation and observable indicators of the relative quality of the schools attended by black students. To evaluate the potential influence of unobserved schoolmate characteristics we use school-level data on the fraction of students who participate in free or reduced price lunch programs. We find that school segregation is highly correlated with black-white relative exposure to low-income schoolmates. To the extent that poor schoolmates lower achievement, however, this pattern would tend to reinforce any causal effect of school segregation, and thus cannot explain our finding of no effect.

Finally, we study within-school segregation trends using data on the relative participation of black and white students in honors and advanced placement (AP) classes. Holding constant the level of neighborhood segregation, we find that the black-white gap in honors and AP participation is wider in cities with more racially integrated schools. This pattern is consistent with claims that ability tracking and related programs offset the integrative effects of between-school desegregation efforts, and may help to explain why differences in school segregation do not appear to influence black relative achievement.

Also note that in summing up their conclusions on per-pupil spending, Card and Rothstein later say, "It appears from these columns that there is little relationship between segregation and spending . . . ."