Home Articles Reader Opinion Editorial Book Reviews Discussion Writers Guide About TCRecord
transparent 13

After Seattle: Social Science Research and Narrowly Tailored School Desegregation Plans

by David J. Armor & Stephanie Duck O'Neill - 2010

Background/Context: In most judicial and social science debates about school desegregation, it is usually assumed that educational and social benefits, if significant, help establish a compelling purpose for school desegregation plans. Less thought has been given to whether the degree of benefits should factor into whether a plan is narrowly tailored according to Supreme Court standards. Hence, many were surprised when Justice Kennedy argued in the Seattle case that school desegregation meets the compelling purpose requirement but that desegregation plans that classified and assigned all students on the basis of race violated the narrow tailoring requirement.

Purpose/Objective/Research Question/Focus of the Study: The essay offers a social science rationale for Justice Kennedy’s views about narrow tailoring issues and suggests several approaches to desegregation plans that may meet narrow tailoring requirements in the future.

Setting: For illustrative purposes, the study uses student achievement data from Charlotte-Mecklenburg, North Carolina, between 2001 and 2006. The data were obtained from the North Carolina Education Research Data Center at Duke University.

Population/Participants/Subjects: From a larger North Carolina data set, a cohort was drawn consisting of about 4,050 Black students from the Charlotte-Mecklenburg district who began third grade in 2000–2001 and completed the eighth grade in 2005–2006.

Research Design: This essay is primarily an analytic essay with a brief quantitative analysis of student achievement data.

Data Collection and Analysis: The quantitative analysis of achievement data utilized longitudinal methods, tracking socioeconomic-status-adjusted achievement trends for Black students over time.

Findings/Results: A modest relationship between school desegregation and Black student achievement (0.2 standard deviations over 5 years) masks a substantial overlap in test scores for individual students in desegregated versus segregated schools. That is, many Black students in segregated schools score higher than the typical desegregated Black student, and vice versa.

Conclusions/Recommendations: Using evidence from within social science, the authors find that the current design of most mandatory desegregation plans does not fit within the current Court’s (especially Justice Kennedy’s) view of narrowly tailored plans. However, there are several ways in which school boards and other policy makers can promote school desegregation: considering race when drawing attendance zones, building new schools, closing schools, and locating magnet schools in strategic areas. Further, magnet schools might meet narrow tailoring requirements if race is only one of several factors considered during the application and administration process.

Many observers anticipated that the Supreme Court would clarify the future of school desegregation when it agreed to review the Seattle and Jefferson County cases. Like many other Supreme Court decisions on civil rights issues, however, a highly divided Court issued opinions that settled very few questions (Parents Involved in Community Schools v. Seattle School District, 2007, hereafter PICS). Although the Court found the Seattle and Jefferson County desegregation plans unconstitutional by a 5–4 vote, even the majority failed to agree on several fundamental issues, and it is likely that more decisions will be required to resolve the school desegregation problem.

Although many commentators described the decision as a rightward shift of the Roberts Court, in fact the legal situation is far more complex. Indeed, the decision reflects multiple disagreements among the justices, reminiscent of other 14th Amendment decisions such as Milliken v. Bradley (1974) and Bakke v. University of California (1978). It cannot be characterized simply as more conservative justices prevailing over more liberal justices.

The major disagreements on the Court concern the two-pronged “strict scrutiny” test required whenever the government uses a racial classification—in this case, assigning students to schools according to race. There are significant disagreements among the nine justices on both parts of the test: whether school desegregation represents a compelling governmental purpose, and whether a given desegregation plan is narrowly tailored to meet that purpose.

From a practical standpoint, narrow tailoring is the more important issue, at least for the present. Although five justices now agree that school desegregation can be a compelling interest in K–12 school systems, a different majority of five justices found that these plans were not narrowly tailored to meet this purpose. Justice Kennedy’s opinion is critical, because he was the fifth vote for both of these majorities. He stated unequivocally that even though school desegregation is a compelling purpose, classifying and assigning students to schools according to race is permitted only under exceptional conditions.

The Kennedy opinion thus creates both a legal and a policy conundrum. First, how can one conclude that school desegregation is a compelling government interest and, at the same time, conclude that students cannot be assigned to schools according to their race? Historically, racial assignment of students to schools has been the essential feature of a school desegregation plan. Second, if a school board declares that school desegregation is a compelling interest, how will it accomplish school desegregation if it cannot use race?

This essay addresses both of these questions. First, we examine the argument and rationale for Justice Kennedy’s seemingly inconsistent opinion. Although he does not base either part of his opinion on social science evidence, we believe the evidence on educational benefits, properly interpreted, may offer a basis for reconciling his views about compelling purpose and narrow tailoring. Second, we believe that this social science evidence does suggest various school desegregation methods that are more likely to survive Kennedy’s narrow tailoring requirements. We explore the feasibility of voluntary desegregation plans through which individual students and families can decide about the value of potential desegregation benefits. Our goal is to help schools, courts, and policy makers fashion desegregation plans that are both legal and feasible.


The majority opinion written by Chief Justice Roberts found that the Seattle and Jefferson County desegregation plans were not narrowly tailored to the compelling purpose of educational benefits. Because of this finding, the Roberts majority said they did not need to address the compelling-purpose issue. The dissenting opinion written by Justice Breyer agreed with the two school boards that desegregation in their systems satisfied compelling purposes (based mainly on social science evidence about educational benefits) and that their desegregation plans were narrowly tailored to meet those purposes. The separate opinion written by Justice Kennedy underscored key differences with both the Roberts majority opinion and the Breyer dissent.

Justice Kennedy agreed with his majority colleagues on their narrow tailoring conclusion; namely, these plans violate the standards for using race as set forth in the University of Michigan decisions (Gratz v. Bollinger, 2003, and Grutter v.  Bollinger, 2003). However, he disagreed with their failure to address compelling purpose. He stated that “diversity, depending on its meaning and definition, is a compelling educational goal a school district may pursue” (PICS, Kennedy, 2-3). At the same time, he made it clear that the type of desegregation plans adopted by Seattle and Jefferson County were not narrowly tailored.

Significantly, Kennedy did not base his compelling purpose conclusion on social science evidence. He believes that Brown established the goal of equal educational opportunity; he said, “If school authorities are concerned that the student-body compositions of certain schools interfere with the objective of offering an equal educational opportunity to all of their students, they are free to devise race-conscious measures to address the problem in a general way”(PICS, Kennedy, 7). One advantage of this approach is that, unlike the dissent, his conclusion is not subject to varying interpretations of social science research on educational benefits. The downside, of course, is that options for desegregation that satisfy his narrow tailoring requirements may be unduly restricted.

Justice Kennedy’s narrow tailoring views depend in part on the legal distinction between de jure and de facto segregation, and he strongly disagrees with Breyer’s rejection of a principle that has been central to every Supreme Court decision on desegregation dating back to Swann. Kennedy believes the de jure requirement should serve as a limitation of governmental authority for any type of racial classification:

Reduction of an individual to an assigned racial identity for differential treatment is among the most pernicious actions our government can undertake. . . . Where there has been de jure segregation, there is a cognizable legal wrong, and the courts and legislatures have broad power to remedy it. The remedy, though, was limited in time and limited to the wrong. . . . The limitation of this power to instances where there has been de jure segregation serves to confine the nature, extent, and duration of governmental reliance on individual racial classifications. (PICS, Kennedy, 15)

In contrast, the Breyer dissent states that educational benefits should trump the distinction between de jure and de facto desegregation.

The compelling interest at issue here . . . includes an effort to create school environments that provide better educational opportunities for all children; it includes an effort to help create citizens better prepared to know, to understand, and to work with people of all races and backgrounds, thereby furthering the kind of democratic government our Constitution foresees. (PICS, 42)

In Breyer’s opinion, the evidence on educational benefits is strong and pervasive, thereby justifying assigning students to schools according to their race.

Thus the Court is divided on the role of social science evidence in these school cases. The Roberts’ plurality did not address whether educational benefits created a compelling interest, because the plans were not narrowly tailored. A concurring opinion by Justice Thomas did review the evidence, concluding that it was equivocal and not strong enough to justify a compelling purpose. The Breyer minority found the evidence strong enough to justify race-based school assignment as both compelling and narrowly tailored. Kennedy is the only justice who believes that school desegregation is compelling enough to consider race but not compelling enough to justify racial assignments, and it appears that he arrived at this conclusion without evaluating the social science evidence.

We agree with Wilkinson that the position of Kennedy in Seattle resembles Powell in Bakke, the only justice between two strongly disagreeing factions who believes that race should be a factor in a government policy but not a basis for classifying all students (Wilkinson, 2007). Clearly, this is not a final opinion on these school desegregation issues; the full Supreme Court must eventually move toward one faction or the other, although it took 25 years for the Court to resolve the Bakke split!1

When the Court does return to the issues raised in the Seattle and Louisville cases, can social science evidence on educational benefits help resolve this impasse? This article takes the position that it might be able to help, although the evidence has to be assembled and interpreted differently than it has been to date. In particular, we believe that the existing research on educational benefits is consistent with Kennedy’s view that desegregation constitutes a compelling interest but does not justify racial assignment of all students to schools. The next section explains how we arrive at this conclusion, as well as other ways in which social science research on educational benefits can be relevant to the narrow tailoring issue.


The debate over social science research in school desegregation cases dates back to the Brown decision itself, which cited social science evidence of psychological harm in its famous “footnote 11” (Brown v. Board of Education, 1954). Although most legal scholars do not believe that Brown relied on this evidence to find a constitutional violation, social science evidence on the “harm and benefit thesis” continued to be introduced in desegregation cases, particularly when the constitutionality of de facto segregation was being debated (Goodman, 1972). If de facto segregation is harmful to minority students, so the argument goes, then it should also be declared unconstitutional.  

Social science evidence on the causes of the Black-White achievement gap has also been introduced in many unitary status hearings to help decide whether vestiges of former de jure segregation still exist (Armor, 1995). The Supreme Court addressed this issue explicitly in the Kansas City case, saying in effect that the achievement gap is relevant in unitary status hearings only if adverse effects on achievement were found in the original Court decision, and then only that part of the gap caused by the constitutional violation (Missouri v. Jenkins, 1995).2  

Of course, the introduction of social science evidence by the Seattle and Louisville school boards had a different goal, which was to establish that school desegregation serves a compelling interest, and four justices believe they succeeded. Given that the narrowly tailored use of race is supposed to further those interests, and only the compelling interests, can this same evidence be used to establish the parameters of narrowly tailored desegregation plans?

Our approach is motivated by two broad requirements of a narrowly tailored policy: (1) that it is shaped and limited by the compelling interest found (in this case, educational benefits) and (2) that the effectiveness of race-neutral alternatives be explored before adopting a race-based solution. These two requirements suggest three different uses of social science evidence. First, we discuss the issue of average desegregation benefits versus variation in benefits for individual students. Second, we will consider evidence that compares the effects of racial diversity with the effects of other types of diversity. Finally, we will examine the extent to which benefits depend on particular types of desegregation plans, including voluntary versus mandatory, magnet schools, and the level of desegregation (i.e., the proportion of minorities).

The discussion of social science evidence in each of these areas is not intended to be a comprehensive research review of any particular issue. This is primarily a concept paper, and our goal is to illustrate each issue with appropriate examples of studies rather than trying to arrive at a consensus about what research says about any particular issue. Indeed, under the narrow tailoring approach we propose, a school district seeking desegregation should tailor a plan according to evidence based on demonstrable conditions and circumstances in that district.


Of the three social science topics reviewed here, the variability of benefits is perhaps most central to the relationship between the compelling interest of educational benefits and narrowly tailored policies that advance that interest. The social science evidence that Seattle and Louisville used to show a compelling interest is fundamentally statistical in nature. That is, their evidence purports to show that, on average, Black students in desegregated schools have higher achievement, improved racial attitudes, better long-term outcomes, and so forth, than students in racially isolated schools.

Assuming for the moment that this is a fair assessment of their research and that it applies to both school districts, this conclusion does not mean that all Black students in desegregated schools, or even a large majority of students, will experience the benefit. Social science research aims to establish a causal effect by comparing the average outcome of persons in one condition (desegregation) with the average outcome of persons in another condition (segregation), controlling for other factors that might also contribute to the outcome.3 If the Seattle and Louisville research was sound, they could legitimately conclude that a typical group of desegregated Black students will experience better educational outcomes than a comparable group of nondesegregated students. However, they cannot assume that all students in the desegregated group will have better outcomes than students in the nondesegregated group.

This is a critical point given the requirements of narrow tailoring. The process of classifying all students according to their race and assigning them to desegregated schools (or restricting their choices of schools based on their race) makes the presumption that all students will experience benefits. The social science evidence only says that there will be an average benefit across groups of students, and typically the research does not identify which students will experience the benefit. Moreover, depending on the size of the average effect, it may not be possible to conclude that even a large majority will benefit.

Illustration Using Charlotte Mecklenburg

To flesh out our argument in more detail, we illustrate the variability-of-benefits issue using achievement data from Charlotte Mecklenburg, North Carolina, a large school district that has figured prominently in desegregation law (e.g., see Swann v. Charlotte-Mecklenburg Board of Education, 1971). The Charlotte Mecklenburg school district has been studied extensively with respect to numerous desegregation issues, including the effects of segregation and desegregation on achievement. At least two studies have found statistically significant negative effects of segregation on academic achievement, although the size of the effects were modest (Godwin, Leland, Baxter, & Southworth, 2006; Mickelson, 2001, 2003).4 Another study found significant negative effects of desegregation for Black students bused to predominantly White schools compared with Black students who were not bused (Becoats, 2001), and a fourth study found no significant effects for Black students one way or the other (Armor, 2002).

The data for our illustration are taken from a larger statewide study of North Carolina achievement that used statewide achievement data maintained by the North Carolina Education Research Data Center at Duke University (Armor & Duck, 2007b). The larger study utilized sophisticated longitudinal regression models developed for a similar study using statewide Texas achievement (Hanushek, Kain, & Rivkin, 2002/2004; Hanushek & Rivkin, 2006). The data presented here are just a subset of those data that track student achievement trends between 2000 and 2005.

Figure 1 shows longitudinal changes in math scores for the Charlotte Mecklenburg Black students as they progress from Grade 3 in 2000 to Grade 8 in 2005. We emphasize that these are actual test scores; at this point, they have not been adjusted for student background factors such as poverty or parents’ education.5 A district might assemble data like this to establish a compelling purpose for desegregation.

For this discussion, because there are only a few Black students in predominantly White schools (0%–25%), we will consider desegregated schools to be those with 25%–50% Black students—which is reasonable considering that the district is nearly 50% Black overall. We will consider segregated schools to be those that are predominantly Black (75%–100%). The test scores have been standardized to have a mean of 100 and a standard deviation of 10.

We see that by the time these Black students attain eighth grade, there is a difference of 4 points between those in segregated schools and those in desegregated schools, representing a relatively large potential “effect” of 0.4 standard deviations. However, note that these two groups start out about 2 points apart in the third grade, reflecting the fact that Black students in segregated schools also have lower socioeconomic status (SES)—a well-known cause of lower achievement. So the potential net effect of attending segregated schools is reduced to 2 points (0.2 standard deviations) if we assess the change in test scores from Grade 3 to Grade 8.

Figure 1. Longitudinal math scores for Charlotte-Mecklenberg Black students, by school

click to enlarge

A better way to do this analysis is to adjust math scores for socioeconomic status, thereby removing the effect of SES, and then track the changes in adjusted math scores. This analysis is shown in Figure 2, which results in much smaller differences among the trend lines. Now the difference between segregated and desegregated eighth-grade students is just 2 points, the same as the net effect of changes in the unadjusted math scores. We also note that the difference between Black students in desegregated schools and majority-Black schools (50%–75%) is less than 1 point.

This difference of 2 points between the math scores of segregated and desegregated Black eighth graders would be considered fairly large when compared with the effects of other educational interventions. Despite this magnitude, this difference does not mean that most desegregated students are outperforming most segregated students. Figure 3 shows the distribution of the adjusted math scores for eighth-grade Black students in desegregated and segregated schools, both of which follow a normal distribution. Although the average of students in desegregated schools is 2 points higher than their average in segregated schools, there is considerable overlap when it comes to individual students. Indeed, one can see, visually, that many segregated Black students are outperforming desegregated Black students.

Figure 2. SES-adjusted math scores for Charlotte-Mecklenberg Black students, by school

click to enlarge

Given the properties of a normal distribution, a difference of 2 points (or 0.2 standard deviations) between the two groups means that approximately 42% of the Black students in segregated schools are scoring above the average for desegregated schools, or more than 97 points. Likewise, 42% of the Black students in desegregated schools are scoring below the average for segregated schools, or less than 95 points. 6

Figure 3. Distribution of Black Students’ Eighth-Grade Math Scores in Desegregated Schools Compared With Predominantly Black Schools, Charlotte Mecklenburg, 2005

click to enlarge

Some might argue that narrow tailoring should not require that every desegregated student outperform every segregated student; that would be too extreme an interpretation. However, to justify classifying all students by race, it would seem reasonable to expect that a large majority of desegregated students should benefit. For this to happen, the group difference would have to be much larger than what is observed here. For example, the average math score difference between the two groups would have to be about 1.3 standard deviations in order for 90% of the desegregated students to score higher than the average segregated students. To our knowledge, there is no research on the effects of desegregation that finds effects of this magnitude.  

Moreover, this would not end the analysis of achievement benefits for Charlotte Mecklenburg Black students from the standpoint of narrow tailoring. In Figure 1, of the 4,000 Black students in this cohort, 60% are attending majority Black schools (51%–75% Black); a little more than one fourth are attending desegregated schools, and only 10% are attending predominantly Black schools. If the school district wanted to racially balance all its schools, all schools would be approximately 50% Black; therefore, the net benefit for all Black students would be smaller than 2 points because the difference between majority-Black and desegregated schools is only 0.6 of a point.

So far, nothing has been said about achievement benefits for White students. Generally speaking, most desegregation research finds no significant achievement benefit for White students, but no adverse effect either (Brief of 553 Social Scientists, 2006, App. 17). In the case of Charlotte Mecklenburg, our longitudinal data show that SES-adjusted scores of White students in majority-Black schools average about 100 and do not change much over time.7 Scores for White students in desegregated schools start out at 101 in Grade 3 and increase a little more than a point by Grade 8. So if all schools were racially balanced and majority Black, White scores might be expected to decline slightly. The most important point from the standpoint of narrow tailoring, however, is that the educational benefit of increased academic achievement is limited to Black students, thereby adding to the narrow tailoring problem for policies that classify all students according to race.

Other Educational and Social Outcomes

Most school districts are unlikely to have their own studies of race relations or racial attitudes, although they might be able to conduct some special studies prior to developing a desegregation plan. For illustrative purposes, we will use a national study. One of the most widely cited studies of race relations and the reduction of prejudice shows that intergroup contact reduces prejudice of all participants by about 0.2 of a standard deviation on average (Pettigrew & Tropp, 2006).8 This potential benefit is about the same as the SES-adjusted achievement benefit for Charlotte Mecklenburg Black students, and it presumably applies to White students and minority students. Assuming the measure of prejudice is normally distributed, a 0.2 standard deviations difference means that just 58% of desegregated students are expected to experience reduced prejudice as compared with the average segregated student, whereas 42% of segregated students would develop less prejudice than the average desegregated student.

It is possible, of course, that some desegregated Black students who do not improve in achievement might improve on race relations, and vice versa. Because studies that make these sort of comparisons are rare, it is unlikely that hard evidence for such an outcome would be available for a school district trying to make this case. The main point we want to make here is the same as that made for achievement: The existence of a modest average desegregation benefit in race relations does not guarantee that all desegregated students will experience that benefit.

Some argue that longer term effects of desegregation, such as educational and occupational outcomes, are more important than the relatively short-term studies of achievement (Wells & Crain, 1994). Some of the major studies of long-term educational outcomes (e.g., college attendance and total years of education) show that, like the achievement literature, the effects of desegregation are quite small and often not statistically significant (Boozer, Krueger, & Wolkon, 1992; Braddock & McPartland, 1982; Crain, 1989; Crain & Mahard, 1978; Eckland, 1979). It is unlikely, however, that any of these long-term effects are larger than the effects for achievement and race relations already discussed; therefore, this evidence has the same problem with respect to narrow tailoring: Average benefits are not the same as individual benefits.

From the standpoint of narrow tailoring, showing that desegregated students have better educational and social outcomes than segregated students on average does not mean that all, or even a large fraction, will benefit from a desegregated school. This creates a narrow tailoring problem for policies that classify all students by race and then assign them to desegregated schools, because such a process assumes that all students thus classified and assigned will benefit from the desegregated school.

For example, most traditional racial balance plans classify all students by race, and then students are assigned to a desegregated school based on their race (within some permissible percentage range). Controlled choice is a variation on this approach; parents and students may choose among desegregated schools first, and then mandatory assignments are made only for those who did not choose a school or if no space is available at their chosen schools. If such a plan is to be narrowly tailored to the compelling purpose of educational and social benefits, then the presumption is that all (or at least most) students would benefit from being assigned to a desegregated school. However, the statistical evidence available from most social science research does not warrant such a conclusion about individual students.

One might protest, of course, that social science is a statistical science, so it is unreasonable to require that all students benefit before adopting a racial balance plan; a positive (and significant) average effect should suffice. This argument reveals a potential disconnect between a social science definition of benefit and a rigorous application of the legal concept of narrow tailoring. If a school board cannot ensure that every student classified by race and assigned to a desegregated school will benefit, then a court might reasonably ask, Why are all students being classified by race and assigned to desegregated schools? As stated by the majority in Adarand, the Equal Protection Clause “protect[s] people, not groups” (Adarand Constructors v. Pena, 1995; emphasis in original)

For the sake of discussion, assume that the Supreme Court adopted a more flexible narrow tailoring requirement so that, for a systemwide mandatory plan like Louisville, the school district only has to show that a large majority of students assigned to desegregated schools would benefit. That is, two thirds or three fourths of the students must experience significant benefits compared with students in racially isolated schools to justify racial classification of all students. If a measure of the benefit is normally distributed (e.g., achievement test scores), then it can be demonstrated that the effect of desegregation has to be +.5 standard deviations for two thirds of desegregated students to benefit compared with the average segregated student.9

It is hard to see how a court would conclude that a systemwide racial balance plan is narrowly tailored if the achievement benefit accrues to slightly more than half of the Black students and to none of the Whites in the system. Although proponents of school desegregation might point to other benefits for White students, such as improved race relations and reduced racial prejudice, it is by no means clear that social gains for White students compensate for the absence of academic achievement benefits, particularly for White parents who perceive substantial burdens from mandatory assignments away from the schools they want to attend.

In summary, the modest impact of desegregation on educational and social outcomes appears to erect serious barriers to school boards that wish to pursue systemwide racial balance policies. For school districts whose experiences are similar to the data shown here for Charlotte Mecklenburg, the effects are not strong enough to argue that most students will benefit from the desegregation experience. Accordingly, extensive systemwide racial balance plans would not be narrowly tailored to the compelling interest of creating education benefits for most students. However, that there are positive effects for some students might be sufficient for pursuing other kinds of desegregation plans, as discussed next.


Another consideration for narrowly tailored plans is the manner in which students are assigned to schools. Because Justice Kennedy is concerned about systemwide plans that classify and assign all students by race, we need to evaluate the effectiveness of alternative plans with respect to educational and social benefits, including the degree of desegregation itself. Although there are many specific desegregation techniques, desegregation plans can be generally classified as one of three types: mandatory, voluntary with magnets, and controlled choice.

Following the 1971 Swann decision, most school systems with a history of de jure segregation adopted, or were ordered to adopt, systemwide mandatory plans that achieved racial balance (to some specified degree) in all schools. In larger districts with substantial residential segregation, these plans usually required extensive transportation; hence, these plans came to be know as “mandatory busing.”

Extensive “White flight” following the adoption of mandatory busing plans in many cities led some courts to approve “voluntary busing” methods using magnet schools and “majority to minority” (m-to-m) transfers (Rossell & Armor, 1996). These plans begin with assignment of students to geographic (neighborhood) attendance zones with boundaries drawn to maximize integration while maintaining contiguous zones. Magnet schools are located strategically to draw an integrated student body, whereas m-to-m transfers can integrate predominantly White schools in outer regions. Voluntary busing plans were never as numerous as mandatory plans, but in some cases, they were more effective than mandatory plans in maintaining stable desegregation levels .

A third type of desegregation plan that became popular in the late 1980s is controlled choice (Alves & Willie, 1987). Under controlled choice, no student has a default school assignment; instead, all parents rank the schools they want their children to attend, and then assignments are made to maximize parents’ choices subject to racial balance and capacity constraints. Usually only a fraction of parents have to be mandatorily assigned—often those who do not respond to the choice survey or who move into a district after schools are full.

Although all these plans classify all students by race, some have elements more compatible with race-neutral school assignments. For example, voluntary magnet plans start with geographic school assignment. This step is race neutral for individual students, but racial demography might be considered when drawing the attendance zones, as suggested by Justice Kennedy. The issue becomes how to create magnet programs and voluntary transfers that do not require consideration of every student’s race. In controlled choice plans, the initial school choice process can also be race neutral; it becomes race conscious only when parental choices are restricted because of their race. Just how these two approaches might be modified to meet the narrow tailoring requirement will be discussed in the next section.  

The next question is whether the type of desegregation plan makes a difference for educational and social benefits. School systems that maintain mandatory racial balance plans (like Jefferson County) often assume that this is the only way to benefit from desegregation, but the research literature does not necessarily support this view. There are only a few studies on whether educational benefits vary by type of plan, and these studies do not support the notion that mandatory plans produce more benefits than voluntary plans.

With respect to achievement, two meta-analyses examined results for voluntary versus mandatory plans. One review concluded that mandatory and voluntary plans “show approximately equal achievement gains” (Crain & Mahard, 1982, p. 33). The second review, looking at a smaller number of more rigorously designed studies, found that “mandatory programs were not associated with reading gains but that voluntary programs were” (Cook, 1984, p. 28). The review did not conclude, however, that this was a causal association, noting that there were too few studies to rule out other possible explanations.

With respect to reduction of racial prejudice, a review by St. John (1975) compared changes in racial prejudice across voluntary and mandatory (or involuntary) plans. For White students, voluntary plans led to reduction of prejudice in three studies and mixed or no effect in two studies; no studies showed increase in prejudice. For involuntary plans, she found three studies with positive results, six with negative results, and five with mixed or no effects.10 For Black students, voluntary plans led to reduced prejudice in three studies, increased prejudice in five studies, and mixed or no effect in four studies. Involuntary plans for Black students led to increased prejudice in two studies and reduced prejudice in one study. The results are clearly mixed, leading St. John to conclude that “the immediate effect of desegregation on interracial attitudes is sometimes positive but often negative. Thus White racism is frequently aggravated by mixed schooling” (p. 119).

Unfortunately, more recent studies of race relations have not classified studies according to the type of desegregation plan. A 1995 review came to a conclusion similar to St. John’s, stating that “there is no guarantee that desegregation will promote positive intergroup behavior,” and that “the evidence on the impact of desegregation on intergroup relations is generally held to be inconclusive and inconsistent” (Schofield, 1995, pp. 610–611). The most recent and most comprehensive meta-analysis of interracial contact, which did find significant positive effects, did not separate out school desegregation from other kinds of interracial contact (Pettigrew & Tropp, 2006). Thus, the study that found the largest positive effect does not shed any light on variation of results according to type of desegregation plan. There is no reason, of course, that future research on desegregation benefits could not do a better job in looking at the effects of different types of desegregation plans.

Admittedly, the data in these reviews are limited, and none of the study authors felt that they could come to a definitive conclusion about the effects of mandatory versus voluntary desegregation plans. One can say, however, that there is little hard evidence supporting the notion that mandatory desegregation plans generate more educational and social benefits than voluntary plans.

Finally, there is the issue of the social and political costs of mandatory versus voluntary plans that should enter into the narrow tailoring calculus. White opposition to mandatory plans and White flight have been widely documented (Clotfelter, 2004; Rossell & Armor, 2002). During the early years of court-ordered desegregation, White opposition to mandatory busing was dismissed as a manifestation of White racial prejudice against Blacks. There is little question that such prejudice existed, particularly in the South during the 10–15 years following Brown, and it undoubtedly motivated much of the early opposition to mandatory busing plans.

As time passed, however, and as the goal of racial integration became increasingly accepted, White parents’ opposition to mandatory busing did not diminish even as they accepted minority students bused into their neighborhood schools. Most minority and White parents, everything else being equal, would prefer to attend a neighborhood school, particularly during the elementary grades. In many school districts, more Black than White parents do not believe their neighborhood school is offering as good an education as majority-White schools. This leads to more Blacks than Whites being willing to transfer out of their neighborhood school, and it also leads to greater acceptance of mandatory plans on the part of Black parents (Armor, 1995; Rossell & Armor, 2002).

This is not to say that all White parents prefer their neighborhood school. Both White and minority parents can become disaffected with their home school, and they are willing to choose schools farther away that offer better programs for their children. This is why charter schools and magnet schools have become very popular where they are offered. This underscores the point that some parents are willing to choose a school other than their neighborhood school, which means that school choice is also a viable option in a desegregation plan.

Mandatory assignment by race, however, continues to be a problem for many parents who do not think that race is a proper basis for assigning children to schools. For parents who are happy with their neighborhood schools or who are willing to attend another school of their own choosing, a mandatory assignment to a school they did not choose is seen as harmful to their children’s interest, and this is true regardless of their race or ethnicity.

Whether it is harmful objectively is beside the point; the perception will lead to unhappy parents, some fraction of whom will avoid the reassignment by moving out of the district or transferring to private schools. Opposition to mandatory assignment by race is definitely a cost that must be considered when designing a narrowly tailored desegregation plan, particularly when the academic achievement benefits are modest for Black parents and nonexistent for White parents.


One of the most important requirements of narrow tailoring is that school boards must examine alternatives to race-based desegregation that would accomplish the same purpose. This implies that, before using race to assign students to schools, race-neutral alternatives should be evaluated and determined to be less effective than race-based methods in producing the educational and social benefits sought by a school board.

The most obvious alternative to the use of race is socioeconomic status, particularly poverty, which is assessed by virtually all school districts to determine eligibility for free lunch and Title 1 programs. Some educational policy experts have advocated using socioeconomic status, or poverty, instead of race to produce economic desegregation of schools (Kahlenberg, 2001, 2006). Kahlenberg noted that some major school districts such as San Francisco and Wake County, North Carolina, have adopted socioeconomic integration in place of racial desegregation.

What does social science evidence say about the effectiveness of alternative definitions of diversity? There is some quantitative research on the effects of socioeconomic integration, although it is not nearly as extensive as its racial counterpart. Studies that simultaneously compare the effects of racial diversity with socioeconomic diversity are even rarer.

One of the more comprehensive national studies used the National Education Longitudinal Survey of 1988 (NELS) to assess the effects of socioeconomic versus racial integration on academic achievement of secondary school students (Rumberger & Palardy, 2005). The study found that when both factors were controlled simultaneously, socioeconomic composition of schools had a significant effect on achievement but racial composition did not. An interesting finding was that effects for White students were nearly as large as effects for minority students—a finding that is not replicated in most studies of racial composition alone.

A study using the 2003 National Assessment of Educational Progress (NAEP) data compared the effects of racial composition with the effects of economic composition as measured by the percentage of students eligible for free lunch, which is a surrogate for poverty status (Armor & Watkins, 2006). The results for eighth-grade Black math scores resembled the NELS study. The effect of school percent Black on eighth-grade Black math scores is .11 standard deviations. When math scores are regressed simultaneously on school percent Black and school percent free lunch, the effect for school percent Black falls to .05, whereas the effect of school poverty is nearly twice that, at .09 standard deviations. For reading scores, when percent Black and free lunch are controlled simultaneously, the racial composition effect drops to 0, whereas the poverty effect is also .09.

For the Charlotte Mecklenberg data shown in Figure 1, a similar regression analysis was carried out. The effect of school percent Black on Black math achievement is 0.1 standard deviations, and the effect of school percent poverty is 0.12 standard deviations. When both are entered into the regression simultaneously, the effect of school percent Black drops to 0, and the effect of school poverty remains at .11 standard deviations. Results are similar for Black reading scores.

Although these three studies are not intended to represent a comprehensive review of the literature, they do support the notion that the effect of racial desegregation can be substantially weaker (or nonexistent) when compared with the effect of economic desegregation on academic achievement. Of course, these are very modest average effects, and they have the same limitation regarding the proportion of students who would benefit from the experience. The difference is that the 14th Amendment does not restrict school boards from using economic status to assign students to schools, so economic integration plans are not subject to the strict scrutiny rule. A school board that believes the educational benefits of economic desegregation are worthwhile can adopt an economic integration plan without demonstrating that the plan is narrowly tailored to serve a compelling interest.

There is very little research about the impact of economic integration on other social outcomes such as race relations or prejudice. Kahlenberg (2006) suggested that these benefits accrue to the extent that economic integration also produces racial integration, because race and poverty are highly correlated. Although Wake County’s economic integration plan has led to considerable racial desegregation, according to Kahlenberg, the economic integration plans in San Francisco and Charlotte Mecklenburg have not had as much success.

Given that the several studies document that greater or equal educational benefits from economic integration, as compared with racial integration, school boards that are seeking the modest educational benefits of racial integration should consider the economic integration approach as a legally viable alternative. Indeed, based on the evidence reviewed here, it seems reasonable to conclude that under the narrow tailoring rule, a school board should have the burden of demonstrating that their goals of enhanced achievement can be attained by racial integration and not by economic integration

The problem with economic desegregation is its failure to target students who have been the long-term victims of racial discrimination, which is why the NAACP and many other civil rights advocates are not enthusiastic about this type of desegregation plan (Peoples, 2002). Economic desegregation treats poor White students the same as poor Black students, even though the former have not been subjected to the same kind of prejudice and disadvantage as the latter. Moreover, working-class Blacks above the poverty line who could benefit from desegregation programs might not be eligible. In addition, a systemwide mandatory economic integration plan might generate the same kind of opposition from middle-class parents as mandatory race-based busing programs. Given the modest benefits discussed, school boards might be reluctant to initiate such plans unless there is substantial support from the community.


Armed with information about the magnitude of educational and social effects, different ways to define diversity, and the effectiveness of alternative plans, school boards can consider options for promoting school desegregation that might pass the narrow tailoring test.

First, school boards should rule out all systemwide mandatory and controlled choice plans that classify students by race and assign them to schools to maintain some degree of racial balance. The current Supreme Court majority does not consider such plans to be narrowly tailored, even with the more optimistic view of the evidence held by the minority. A more careful assessment of the evidence on benefits, such as that presented in our Charlotte Mecklenburg example, might reinforce the majority view if and when the Court undertakes a review of the benefit issue.

Second, a school board can adopt an economic desegregation plan at any time without any concern about legal challenge, so long as the plan does not use poverty as a surrogate for race. For example, if poverty is used to classify students only in predominantly Black neighborhoods, that might be challenged as using poverty as a surrogate for race and subject the plan to a strict scrutiny review.

Third, it seems clear from Justice Kennedy’s current opinion that a school district can foster desegregated schools by considering race and taking race into account when drawing attendance zones, deciding where to build new schools, deciding which schools to close, and locating magnet schools in strategic areas that might draw an integrated student body. The question is whether desegregation experiences can be extended to individual students. Can voluntary transfer programs and magnet schools be used to increase desegregation at some schools?

With respect to voluntary transfers, a school board could adopt an at-risk transfer program that would offer transportation-paid transfers to students who are below the poverty line, below an achievement test threshold (e.g., the state proficiency standard), or both. These students could be eligible to transfer to designated receiving schools with low poverty levels and high levels of proficiency. As long as the poverty and proficiency levels are determined independent of race, there is no reason why such a transfer program should raise constitutional issues.

With respect to magnet schools, one way to foster racially integrated magnet programs is to have special (and popular) themes with eligibility requirements based on academic performance. Because of the correlation between race and achievement test scores, an academic eligibility requirement would tend to create integrated student bodies in the magnet program. For example, if the eligibility requirement for a number of magnet programs is above-average test scores, then these race-neutral magnet programs should have a reasonable level of racial integration. In many cases, such “honors” academies would be popular with middle-class parents, and they might be in high demand in many school districts. Of course, they would offer desegregation only to minority students who had above-average achievement.

Is there any chance for considering race in a strictly voluntary magnet plan? If a magnet program is strictly voluntary, and if it determines eligibility according to multiple student characteristics such as test scores, poverty, and academic interests, it seems to us that race might be one of the several student characteristics considered. Such an eligibility and selection process would seem consistent with the way race is used according to Grutter: as one of several factors rather than the only factor. If the program is voluntary, this would not require racial classification of all students in a school district. It might even be possible to define and control the composition of the magnet school with ranges for each of the student characteristics, although this would be tricky if it led to exclusion of students according to their race.

These various types of desegregation programs would give opportunities to many minority students to attend an integrated school, and it would also offer integration and academic opportunities to many minorities who have higher (or lower) test scores or who are below the poverty line. Although such programs would not desegregate an entire school system, they offer desegregation opportunities to those who want that experience. Given the very modest educational and social benefits documented for desegregation, and given that no existing research can precisely identify those students most likely to benefit, it seems reasonable and prudent to let those who want the desegregation experience to have it rather than forcing it on those who are content with attending their regular neighborhood (or choice) school.


1. The Bakke opinion was rendered in 1978, but a five-vote majority on the use of race in college admissions did not occur until the Grutter decision in 2003.

2. A unitary status hearing takes place when a school district petitions to be discharged from federal court supervision for prior constitutional violations.

3. The most rigorous causal inference is when other factors are controlled via a randomized experiment, but other statistical procedures are accepted in the research literature.

4. In the Mickelson study, Black and White students were not analyzed separately, so the differential effects of predominantly Black schools on Black student achievement cannot be discerned. In the Godwin et al. (2006) study, Black students were separated out, but the school composition variable was socioeconomic status (percent free/reduced lunch) rather than race; moreover, a statistically significant negative effect occurred for White students but not Black students.

5. It is known that Black students in predominantly Black schools usually have lower socioeconomic characteristics, so if adjustments were made for these factors, the test scores would converge to some extent.

6. These calculations are based on the properties of a normal distribution: for a z-value of .24, 60% of the distribution lies before the z-value and 40% after.

7. There are only 50 White students in predominantly Black schools, so this group is dropped from our analysis.

8. It should be noted that this study does not estimate benefits specifically for racial desegregation of schools; rather, the effect applies to intergroup contact among all types of groups (race, gender, age, and so on), all settings, and all ages. The effect size for interracial groups and the effect for adolescents is .21.

9. In one widely cited recent study of Texas achievement by Hanushek and Rivkin (2006) and Hanushek et al. (2002/2004), larger statewide effects of desegregation have been reported as compared with the effect reported here for Charlotte Mecklenburg. These very complex analyses appear to be highly sensitive to model specification, and new research by Armor and Duck (2007a, 2007b) failed to replicate the Texas results using data for North Carolina, South Carolina, and the national Early Childhood Longitudinal Study.

10. These involuntary plans include schools receiving minority students from open enrollment, so the racial contact was not selected voluntarily by the white students.


Adarand Constructors v. Pena, 515 U.S. 200 (1995).

Alves, M. J., & Willie, C. V. (1987). Controlled-choice assignments: A new and more effective approach to school desegregation. The Urban Review, 19, 67–88.

Armor, D. J. (1995). Forced justice. New York: Oxford University Press.

Armor, D. J. (2002). Desegregation and academic achievement. In C. R. Rossell, D. J. Armor, & H. J. Walberg (Eds.), School desegregation in the 21st century (pp. 291–322). Westport, CT: Praeger.

Armor, D. J., & Duck, S. (2007a, February 8). Unravelling Black peer effects on Black test scores (Working Paper). University of Arkansas Education Working Papers Archive, Fayetteville.

Armor, D. J., & Duck, S. (2007b, May 28). The effect of Black peers on Black test scores. George Mason University, School of Public Policy, Fairfax, VA.

Armor, D. J., & Watkins, S. J. (2006, July 28). School segregation and Black achievement: New evidence from the 2003 NAEP. In The benefits of racial and ethnic diversity in elementary and secondary education (pp. 28–49). Washington, DC: U.S. Commission on Civil Rights.

Bakke v. University of California, 438 U.S. 265 (1978).

Becoats, E. (2001). Busing and academic achievement of Black students. Unpublished doctoral dissertation, College of Education, University of North Carolina, Charlotte.

Boozer, M. A., Krueger, A. B., & Wolkon, S. (1992). Race and school quality since Brown v. Board of Education (Brookings Institution Papers on Economic Activity). Microeconomics, 269–338.

Braddock, J. H., II, & McPartland, J. M. (1982). Assessing school desegregation effects: New directions in research. Research in Sociology of Education and Socialization, 3, 259–282.

Brown v. Board of Education, 347 U.S. 483 (1954).

Clotfelter, C. T. (2004) After Brown: The rise and retreat of school desegregation. Princeton, NJ: Princeton University Press.

Cook, T. D. (1984). What have children learned academically from school integration? An examination of the meta-analytic evidence. In T. D. Cook and others, School desegregation and Black achievement (pp. 6–42). Washington, DC: National Institute of Education

Crain, R. L., and others. (1989, January). Finding niches: Desegregated students sixteen years later. Unpublished manuscript, Teachers College, Columbia University, New York.

Crain, R. L., & Mahard, R. E. (1978). School racial compositions and Black college attendance and achievement test performance. Sociology of Education, 51, 81–101.

Crain, R. L., & Mahard, R. E. (1982) Desegregation plans that raise Black achievement: A review of the research (Report No. N-1844-NIE). Santa Monica: Rand

Eckland, B. K. (1979). School racial composition and college attendance revisited. Sociology of Education, 52, 122–125.

Godwin, R. K., Leland, S. M., Baxter, A. D., & Southworth, S. (2006). Sinking Swann: Public school choice and the resegregation of Charlotte’s public schools. Review of Policy Research, 23, 983–997.

Goodman, F. I. (1972). De facto school segregation: A constitutional and empirical analysis. California Law Review, 60, 275–284.

Gratz v. Bollinger, 539 U.S. 244 (2003).

Grutter v. Bollinger, 539 U.S. 306 (2003).

Hanushek, E. A., Kain, J. F., & Rivkin, S. G. (2002). New evidence about Brown v. Board of Education: The complex effects of school racial composition on achievement (Working Paper No. 8741). Cambridge, MA: National Bureau of Economic Research. (Revised February 2004)

Hanushek, E. A., & Rivkin, S. G. (2006, October). School quality and the Black-White achievement gap (Working Paper No. 12651). Cambridge, MA: National Bureau Of Economic Research.

Kahlenberg, R. D. (2001) All together now: Creating middle-class schools through public school choice. Washington, DC: Brookings Institution Press.

Kahlenberg, R. D. (2006). A new way on school integration. Washington, DC: The Century Foundation.

Mickelson, R. A. (2001). Subverting Swann: First- and second-generation segregation in the Charlotte-Mecklenburg Schools. American Educational Research Journal, 38, 215–252.

Mickelson, R. A. (2003). The academic consequences of desegregation and segregation: Evidence from the Charlotte-Mecklenburg Schools. North Carolina Law Review, 81,  1513–1562.

Milliken v. Bradley, 418 U.S. 717 (1974).

Missouri v. Jenkins, 515 U.S. 70 (1995).

Parents Involved in Community Schools v. Seattle School District, No. 05-908 U.S. June 28, 2007 (Seattle).

Peoples, B. (2002, March/April). Desegregation plan to focus on income rather than race. The New Crisis. Retrieved from http://findarticles.com/p/articles/mi_qa3812/is_200203/ai_n9068133/

Pettigrew, T. F., & Tropp, L. R. (2006). A meta-analytic test of intergroup contact theory. Journal of Personality and Social Psychology, 90, 751–783.

Rossell, C. R., & Armor, D. J. (1996). The effectiveness of school desegregation plans, 1968–1991. American Politics Quarterly, 24, 267–302.

Rossell, C. R., & Armor, D. J. (2002). Attitudes on race and desegregation. In C. R. Rossell, D. J. Armor, & H. J. Walberg (Eds.), School desegregation in the 21st century (pp. 291–322). Westport, CT: Praeger.

Rumberger, R. W., & Palardy, G. J. (2005). Does segregation still matter? The impact of student composition on achievement in high school. Teachers College Record, 107, 1999–2045.

St. John, N. (1975). School desegregation: Outcomes for children. New York: Wiley.

Schofield, J. W. (1995). Review of research on school desegregation’s impact on elementary and secondary students. In J. A. Banks & C. A. M. Banks (Eds.), Handbook of research on multicultural education (pp. 597–616). New York: Macmillan.

Brief of 553 Social Scientists as Amicus Curia in Support of Respondents, Parents Involved in Community Schools v. Seattle School District, No. 05-908 (2006).

Swann v. Charlotte-Mecklenburg Board of Education, 402 U.S. 1 (1971).

Wells, A. S., & Crain, R. L. (1994). Perpetuation theory and the long-term effects of school desegregation. Review of Educational Research, 64, 531–556.

Wilkinson, H. J., III (2007). The Seattle and Louisville school cases: There is no other way. Harvard Law Review, 121, 158–183.

Cite This Article as: Teachers College Record Volume 112 Number 6, 2010, p. 1705-1728
https://www.tcrecord.org ID Number: 15696, Date Accessed: 5/27/2022 1:44:05 AM

Purchase Reprint Rights for this article or review
Article Tools
Related Articles

Related Discussion
Post a Comment | Read All

About the Author
  • David Armor
    George Mason University
    E-mail Author
    DAVID J. ARMOR is a professor of public policy in the School of Public Policy, George Mason University, Fairfax, Virginia. He writes and conducts research in the areas of education, school desegregation, cognitive development, and military manpower. He is author of several books, including Maximizing Achievement, Forced Justice: School Desegregation and the Law, and Competition in Education: A Case Study of Interdistrict Choice.
  • Stephanie O'Neill
    George Mason University
    STEPHANIE DUCK O'NEILL is currently a graduate student at George Mason University. Her recent publications include "Better Understanding Teacher Compensation" with William Fowler, presented to the American Education Finance Association Conference, April 2008, and "Evaluating Peer Effects on Student Achievement in South Carolina" with David Armor, presented to the Association for Public Policy Analysis and Management Fall Research Conference, November 2007.
Member Center
In Print
This Month's Issue