Race, the Supreme Court, and Social Science Evidence
by David J. Armor - June 15, 2009
New civil rights cases being considered by the Supreme Court bring to mind the recent Seattle and Jefferson County cases, in which Justice Kennedy cast the fifth and deciding vote in striking down racial diversity plans in these two school districts. His opinion managed to side with both the liberal and the conservative wings on different but critical aspects of the decision. This commentary offers an interpretation of social science evidence in these cases that supports Justice Kennedy's novel opinion.
Once again the Supreme Court is reviewing major civil rights cases in the areas of voting rights, employment, and education. It is likely that Justice Kennedy will play a central role, as he did in the most recent civil rights case in education. Justice Kennedy cast the 5th and deciding vote in the Seattle and Jefferson County school cases, which disapproved their desegregation plans because of excessive reliance on race and racial classifications.1 Since social science evidence is introduced in these new cases, this decision may be instructive for anticipating how the Court and Justice Kennedy might evaluate and utilize this evidence .
For some observers, Justice Kennedys vote in Seattle was puzzling, because in one part of his opinion he sided with the four liberal Justices, and in another part he sided with the four conservative Justices. He agreed with the liberal faction that school desegregation constitutes a compelling government purpose, thereby permitting school boards to pursue racial diversity policies. Significantly, he did not endorse their rationale, which relied partly on social science evidence about the educational and social benefits of racial diversity. Rather, Kennedys reasons were based on broader considerations, including Browns original objective of equal educational opportunity. He stated clearly that the problems of de facto segregation and racial isolation were problems that school boards could and should legitimately address.
But he also agreed with the conservative faction that these plans were not narrowly tailored to meet this compelling purpose. Again, he did not endorse all of the rationale offered by the plurality, and in particular he expressed concerns that Chief Justice Roberts opinion might convey the message that race can never be considered in designing diversity policies, a position he found too extreme. Justice Kennedy concluded that a school boards compelling interest in remedying de facto racial isolation did not justify policies that required classifying all students by race and using that information to assign students to schools, which he states is among the most pernicious actions our government can undertake.2 He would not have a problem if school boards used race in a more general way, such as deciding on where to build schools, adjusting attendance zones, and the like.
Significantly, none of Kennedys views about compelling purpose or narrow tailoring relied on social science evidence; he did not even mention the extensive evidence on benefits produced by the parties and discussed in great detail in many amicus briefs. The liberal faction relied heavily on this evidence to support their views about compelling purpose, and they believed the evidence was strong enough to support the racial classifications of all students to attain a particular degree of racial diversity in schools. The Roberts opinion asserted that evidence on benefits was moot because of their conclusions about narrow tailoring, leaving Justice Thomas as the only member of the plurality to address the social science evidence. To the extent it was relevant, Justice Thomas judged it as inconclusive.
Given these conflicting views, can social science evidence play a role in future civil rights decisions by the Supreme Court? With respect to the benefits of racial diversity, I offer one interpretation of social science evidence that is consistent with Justice Kennedys complex views about racial diversity. In fact, this interpretation offers support for his view that racial diversity can be a compelling government goal, but not one that justifies using racial classifications to make decisions about individuals. That is, current research on racial diversity does suggest modest educational and social benefits in certain settings, but the magnitude of these benefits is not large enough to justify a policy of that classifies all students by race and then treats them differentially based on their race.
The evidence on educational and social benefits of racial diversity is usually assessed by comparing various outcomes for persons in settings (e.g., schools) that are either racially isolated or racially diverse, and by demonstrating that minority students in the diverse settings have better outcomes, on average, than those who are racially or ethnically isolated. While definitions of diversity and social outcomes differ from one study to another, most social science studies attempt to calculate benefits in terms of a summary measure called a standardized effect. A standardized effect translates different types of outcomes into a common metric based on the standard deviation of an outcome. 3 In the case of school desegregation, the benefits of diversity often differ from one school district to another, but across many different studies the effect is usually no less than .1 standard deviation, but it is rarely greater than .3 standard deviations.
It is important to stress that these effects represent the outcomes for the average student in a diverse setting and the average student in an isolated setting. Moreover, one might find positive outcomes in one setting such as a particular school district, but one might not find the same benefit in another school district, because there may be large variations in outcomes from one setting to another.
Using Charlotte Mecklenburg as an illustrative example, a longitudinal analysis of North Carolina achievement test data finds that, between 2001 and 2006, desegregated black students scored a total of about .2 standard deviations higher than segregated black students over a five year period as they progressed from third grade to eighth grade (controlling for poverty and parents education).4 In other outcome areas, such as race relations, desegregation often shows similar levels of benefits for both black and white students. This is the sort of evidence that both Seattle and Jefferson County assembled to argue that desegregation met the compelling purpose requirement, and this evidence was sufficient to convince four Justices of the Supreme Court.
Assuming that such quantitative outcomes are approximately normally distributed, which is true for the Charlotte Mecklenburg achievement scores, an effect of .2 standard deviations means that three-fifths of desegregated black students score higher than the average segregated black student. But it also means that two-fifths of segregated black students score higher than the average desegregated black student! The reason that so many segregated black students score higher than the average desegregated black student is that an effect of .2 standard deviations is quite small, and it allows for a large overlap in the distribution of test scores. In order for 90 percent of desegregated students to score higher than the average segregated student, the desegregation effect would have to exceed one standard deviation. To my knowledge, an effect of this size has never been documented in a study of racial diversity outcomes.
I would argue that this result for Charlotte Mecklenburg is relevant to Justice Kennedys opinions about compelling purpose and narrow tailoring. The fact that there is a modest average achievement benefit effect in this case could support a finding that the Charlotte Mecklenburg would have a compelling interest in fostering school desegregation. However, classifying all students by race and assigning them to schools based on their race implies that any or all desegregated students will have better outcomes than any or all segregated students, and yet this is not what the statistical evidence actually proves. Statistical evidence deals with averages; it implies only that some number of desegregated students will perform better than some number of segregated students, but we do not know which ones. Although Justice Kennedy did not rely on an argument of this type, it might reflect some of his underlying concerns. Statistical averages do not describe the full range of outcomes for all individuals who are affected by a governmental policy.
It remains to be seen, of course, whether this type of statistical reasoning is adopted by any judicial body, much less the Supreme Court. Of course, even if Justice Kennedy or other members of the Court accepted the results of this analysis, they still might not believe that social science evidence is a proper basis for deciding constitutional questions. In addition, the social science evidence before the Court in the current civil rights cases is somewhat different than the evidence examined here. The employment case concerns the validity of standardized job skill tests, and the education case concerns the effect of school expenditures on student achievement. Nonetheless, in the Seattle case, social science evidence was considered relevant by five of the nine Justices, which gives it greater weight than might it might have had several decades ago. As social science research becomes more rigorous and more relevant to legal questions, and as courts become more adept at grappling with its intricacies, it could have growing importance in civil rights cases.
1. Parents Involved in Community Schools v. Seattle, No. 05-908, U.S. June 28, 2007
2. Ibid, Kennedy Opinion, p. 15
3. A standard deviation is a statistical measure of variation; for a distribution that is normal or bell shaped, about two-thirds of the observations fall within plus or minus one standard deviation.
4. David J. Armor and Stephanie Duck O'Neill, After Seattle: In Search of Narrowly Tailored School Desegregation Plans, Teachers College Record (2009, in press)