Does the Negro Need Separate Schools?A Retrospective Analysis of the Racial Composition of Schools and Black Adult Academic and Economic Success
W.E.B. Du Bois asserted that black students are better served by attending predominantly black schools than hostile integrated schools in a context of racial discrimination. The conventional assumption is that black students benefit educationally by attending schools with more white peers, which have access to greater resources. However, the theory of the functionality of discrimination advances the idea that black students may face greater discrimination in school settings with numerous white peers as a result of a competitive process and white appropriation of preferred resources. Using the National Survey of Black Americans, we find evidence of a nonmonotonic relationship between high school racial composition and years of schooling completed, high school graduation, likelihood of being employed, and likelihood of owning a home. We conclude, contrary to conventional belief, that it is not unambiguously the case that black students gain from attending schools with more white peers.
stratification economics, desegregation, high school graduation, high school dropouts, racial composition
Does attending a school with a greater presence of white children improve the academic performance of black children? Do black children need to be in a school with white children to learn? The conventional wisdom has it that attending school with white peers benefits black students on a wide array of front, including their academic outcomes (Orfield and Lee 2005; [End Page 166] Hanushek, Kain, and Rivkin 2009; Johnson and Nazaryan 2019). The conventional wisdom prevails despite conflicting evidence and the concerns expressed by no less a luminary than W.E.B. Du Bois (1935) about the physical, emotional, and intellectual maltreatment black students might face when they attend desegregated schools, rather than schools where virtually all of their peers are black.1
Moreover, to the extent that academic gains for black students are associated with attending desegregated or integrated schools, existing research is not clear about the mechanism producing those gains. Does a sheer benefit accrue to black students from having contact with white students? Or are the putative gains attributable to integrated schools having better resources in terms of the quantity and quality of teachers per student, superior equipment, books and materials, classrooms, and curricula?
Indeed, would it be possible to achieve at least the same level of academic gains for black students if predominantly black schools had been upgraded to a comparable level of resources as predominantly white schools? In short, could the perverse policy of "Deluxe Jim Crow" applied to schooling actually have produced similar academic gains for black students as school desegregation, if in fact, "separate but equal" were really realized (Cobb 1952; Thomas 2011)?2
The latter question has immense policy relevance. The combination of a tendency toward the resegregation of schools at the facility level coupled with the tenacious nexus between sustained residential segregation and school composition (Clotfelter 2004) means that the process of school integration has not been completed and shows signs of regressing (Fiela 2013; McNeal 2009; Orfield and Lee 2007). Furthermore, evidence suggests that many integrated schools implement racially stratified curricula, where black students are overwhelmingly "tracked" into a low achieving curriculum and the high achieving curriculum remains overwhelmingly the domain of white students (Oakes 2005; Tyson, Darity, and Castellino 2005; Darity and Jolla 2010; Diette 2012).
Therefore, an interesting question is whether efforts to close the racial schooling gap should prioritize desegregation or improved quality of facilities, curriculum, and instruction for black students wherever they attend school. If a more diverse racial composition of schools does not reliably produce an intrinsic academic benefit for black students, it may be appropriate to prioritize the improvement of existing schools rather than relocating students for greater racial balance without adequate attention to how curricular and instructional resources are distributed within more racially balanced schools.
To date, research on the link between educational attainment for black students and the racial composition of the schools they attend has for the most part focused on the racial makeup of high schools. The convention (the conventional hypothesis) in this literature is to assume that as the high school black students attend becomes more white, the educational attainment of black students rises at a monotonic rate.
We extend the literature in two ways using data from the National Survey of Black Americans. First, we investigate whether the association between black students' educational attainment—both years of schooling completed and the probability of graduating from high school—and the share of their high school that is white is monotonic. Second, we investigate the association between the racial composition of the high school blacks attend and two important [End Page 167] measures of subsequent financial well-being.
We find evidence at odds with the common view that educational attainment for blacks increases monotonically with the percentage of their high school peers who are white. Our findings do reveal that relative to blacks who attend a high school that is predominantly black, those attending a high school that is predominantly white are more likely to graduate and to complete more years of schooling. However, our results also reveal that black students attending a racially balanced high school, one with approximately equal shares of black and white students—which we also refer to as mixed or integrated schools—complete less schooling or no more schooling relative to those enrolled in predominantly black high schools and predominantly white high schools. These findings are particularly striking given potential concern about selection bias—that parents with the wherewithal will select into the schools with more resources and a larger share of white peers—which would bias the results in favor of integrated and predominantly white schools.
That black students who attend racially balanced high schools do not complete more schooling than black students who attend predominantly black high schools calls into question whether white peers, in and of themselves, necessarily have a direct and positive impact on black students' academic achievement. This is consistent with the hypothesis advanced in stratification economics that blacks will face more discrimination in racially balanced schools than if they attend predominantly black or predominantly white schools (Darity 2001, 2005; Darity, Hamilton, and Stewart 2015; Hamilton 2017).
The relationship between employment and homeownership and the share of white high school peers for black students is also non-monotonic. Thus our findings outside the educational sphere further demonstrate that black students do not consistently benefit from attending more-white high schools.
The following section presents a conceptual discussion of the potential link between the racial composition of schools that black students attend and their educational and economic outcomes.
In a 1935 essay, during the interim between the Plessy and the Brown decisions,3 W.E.B. Du Bois anticipated the hostility that black youth might experience in integrated schools. He also expressed an awareness that the production of knowledge is a complex process involving both technical and human inputs. A proper education, he asserted, required nurturing or a "sympathetic touch between teacher and pupil" (1935, 328), as well as a knowledgeable educator with appropriate infrastructure or facilities. Further, he advocated for curriculum content relevant and appropriate to the life experiences of black students.4
In addition, Du Bois was a proponent of learning outside the classroom and engagement in practical activities to provide opportunities to apply and refine knowledge and to foster excitement about the life of the mind. He understood that peers have skills they can share with classmates that promote learning. Moreover, he observed that administrators, teachers, and students shape a culture that plays a substantive role in the educational process by influencing students' norms, motivations, and aspirations.
His primary policy concern was not the racial composition of schools per se, but instead that black youths receive quality education:
It is simply calling a spade a spade. It is saying in plain English: that a separate Negro school, where children are treated like human beings, trained by teachers of their own race, who know what it means to be black in the year of salvation 1935, is infinitely better than making our boys and girls doormats to be spit and trampled upon and lied to by ignorant social climbers, whose sole claim to superiority is ability to kick "niggers" when [End Page 168] they are down. I say, too, that certain studies and discipline necessary to Negroes can seldom be found in white schools.(1935)
In the post–Civil War era, black and white students were taught in almost exclusively racially segregated schools. The question Du Bois wrestled with was whether black students could receive a strong education in desegregated schools given the hostility they would likely face from white students. In integrated schools, he noted, blacks were tolerated at best, but not embraced, and thus wondered whether black youths would receive a proper education in such an environment.5 He concluded that black students "will have separate schools or [they] will not be educated" (Du Bois 1935, 330). In Du Bois's view, to endure a poor education simply because schools are integrated "is a costly if not fatal mistake" (330). Given these circumstances, he was a fierce critic of school desegregation even as white society continued its discriminatory treatment toward blacks and excluded curricula essential to the black historical and current experience.6
His critics argued that supporting segregated schools would be giving up or backpedaling on the principle of equal access to quality public education. Du Bois recognized, though, that "the Negro needs neither segregated schools nor mixed schools. What he needs is Education" (1935, 355). For Du Bois, "a mixed school with poor and unsympathetic teachers with hostile public opinion … is bad," and a segregated school with "inadequate equipment, poor salaries, and wretched housing is equally bad" (for a discussion, see Hamilton 2014).
Du Bois (1970) made clear that his opposition was to discrimination in the provision of schooling, not desegregated schooling per se. Thus he opted to focus his attention on securing better funding for predominantly black schools so they could offer better services to their students.
Du Bois's objective of closing the racial school resource gap—ironically an objective shared with Booker T. Washington, with whom he often wrangled—was supported by Julius Rosenwald, president of Sears from 1908 to 1922 and a leading philanthropist of the early twentieth century (Ascoli 2006). Through the Rosenwald Foundation, he financed the construction of five thousand schools between 1917 and 1932 to serve black children exclusively. The Rosenwald Foundation also supported a number of NAACP initiatives, so Rosenwald knew Du Bois's position on race and schooling. More than one- third of the black children in the rural South during this period were educated in Rosenwald Schools, which were located in the eleven states of the former Confederacy as well as Oklahoma, Missouri, Kentucky, and Maryland.
Grants from the Rosenwald Foundation to facilitate construction of the Rosenwald Schools required matching monetary contributions from the black students' parents, amounting in total to $4.7 million, as well as parental contributions of labor and land. The idea was that this would ensure deep parental commitment to the goal of better educational facilities for their children. Indeed, the parents' devotion to the importance of education for their sons and daughters was so pronounced that in addition to raising the money to complement the foundation's donations they paid local taxes that went to support the public schools their children could not attend.
Rosenwald Schools, operating under conditions that necessarily made them all-black schools, were highly successful. Although rural black schools rarely if ever had the same monetary resources as the white schools, [End Page 169] Daniel Aaronson and Bhashkar Mazumdar (2011) report that the improved educational opportunities associated with the Rosenwald Schools led to a 40 percent closure of the gaps in educational achievement between blacks and whites for the relevant cohorts.
Drawing on census data and World War II records, they "find significant effects on school attendance, literacy, years of schooling, cognitive test scores, and northern migration [with] gains are highest in the most disadvantaged counties, suggesting that schooling treatments have the largest impact among those with limited access to education" (2011, 821). Because the Rosenwald Schools were not directly under the control of local all-white school boards, black parents could exercise authority over hiring teachers and curriculum content. This also played a role in the positive effect of attending a Rosenwald School. Thus, experience with the Rosenwald Schools seems to confirm Du Bois's belief that improved quality of education, regardless of the demographic composition of the school, is what matters the most.
Ultimately, Du Bois concluded that caring teachers, supportive peers, and exposure to the truth about black history would be more beneficial to black youths than education in "hostile" integrated schools.7 Thus what we label the Du Bois hypothesis is the premise that black students are better served by attending predominantly black high schools, relative to integrated schools, as long as society engages in racial discrimination. The Du Bois hypothesis posits that white peers, as a result of their ability to influence the distribution of resources and opportunities within a school and their desire for preferred educational and economic outcomes, have a negative effect on educational attainment of black students.
STRATIFICATION ECONOMICS, FUNCTIONAL ROLE OF DISCRIMINATION THEORY, AND SCHOOL INTEGRATION
Stratification economics asserts that discrimination functions as a mechanism dominant groups use to maintain their relative social hierarchical position over subaltern groups. A similar perspective was advanced by both Herbert Blumer (1958), who characterized race prejudice as a sense of group position, and the Nobel laureate W. Arthur Lewis (1985), who deployed the concepts of competing and noncompeting groups.8
Stratification economics assigns a functional role to discrimination (Darity 2001; Hamilton 2017). Discriminatory practices by the dominant group intensify when potential competition over valued resources and access to opportunities between the two groups is greater. Thus, whites would not bother to discriminate against blacks to a great extent in the absence of perceived gain or if blacks are not perceived as threatening their position of privilege.
In the context of schooling, in a racially balanced environment, white students and their parents are likely to consider black students as serious competitors for relative access to resources. Examples include gaining teacher attention, [End Page 170] seats in desirable classes, and positions of status in the school involving leadership, athletics, and academics. Access to these resources and positions are thought to lead to positive economic returns. Competition for these advantages may lead to hostility and discrimination that undermines black educational achievement. However, when white students have only a handful of black peers, they will see their position of privilege as less threatened and engage in relatively less discrimination against their black classmates.
Thus, our application of the functionality of discrimination hypothesis makes two claims. First, it predicts that in racially balanced schools, the competitive threat to white students for preferred academic advantages posed by black students will be enough to foster a hostile treatment effect that offsets, and possibly dominates, the resource effect. Second, it posits that in predominantly white schools, the more limited threat to white students for preferred positions will not lead the hostile treatment effect to dominate the resource effect. Thus our functionality of discrimination hypothesis predicts that black students who attend racially balanced schools may not acquire more schooling and may have a lower level of educational attainment relative to those who attend either a predominantly black school or a predominantly white school.
PREDICTIONG THE LINK BETWEEN RACIAL COMPOSITION ATTAINMENT
We consider three hypotheses regarding the impact on black student's educational attainment of the share of their high school peers who are white: the conventional hypothesis, the Du Bois hypothesis, and the functionality of discrimination hypothesis. These hypotheses are generated by different perspectives about the combined impact of differences in both high school resources and discrimination against black students in school settings that differ in terms of the racial composition of their peers. Each of the three hypotheses embeds the assumption that school resources are greater when the white share of students in a school is larger and these resources will enhance black student educational attainment.9 Moreover, they each acknowledge that racial discrimination undermines learning and may prevent black students from accessing school resources. They differ, however, on the extent of discrimination that prevails when the racial composition of school peers differs.
The conventional hypothesis assumes that the benefits from additional resources always outweigh any negative effects of discrimination (which may include denied access to these resources) or, in some cases, does not acknowledge the potential presence of negative effects of desegregation. Thus the standard view is that a higher share of white peers is associated with greater educational achievement among black students. This produces a positive relation between more white peers in high school and black students' educational attainment.
However, the Du Bois hypothesis, set out earlier, suggests that the negative effects of discrimination in the form of poor treatment from white students, teachers, and administrators will be larger than the benefits of additional school resources. It implies that black students complete less schooling as the proportion of their white peers increases, which yields a negative relation.
The functionality of discrimination hypothesis posits that discrimination against black students is much stronger at racially balanced schools than at predominantly white schools—where poor treatment of black students is expected to be modest.10 Recall that school resources [End Page 171] rise as the share of white peers increases. Thus, the functionality of discrimination hypothesis we advance asserts that, relative to their peers who attend predominantly black schools, black students attending racially balanced schools will complete equivalent or less schooling than their black peers (discrimination effect is equivalent to or dominates resource effect respectively). In operationalizing the functionality of discrimination hypothesis, we contend that black students at schools with a high proportion of white students will finish more schooling (resource effect dominates a diminished discrimination effect).
DATA, MEASUREMENT, AND DESCRIPTIVE STATISTICS
We analyze data from the National Survey of Black Americans (NSBA) (Jackson and Gurin 1996). The NSBA was designed to collect information on the economic conditions of black Americans, including their educational attainment and health status. The initial wave was carried out on a nationally representative group of black respondents age eighteen and older between 1979 and 1980. The data were gathered in face-to-face interviews with respondents conducted in the privacy of their homes. Three subsequent waves of data collection were conducted eight, nine, and twelve years after the initial interview. Because of extensive sample attrition, our analysis uses variables drawn only from the first wave.
A desirable feature of the data is that the survey collected retrospective respondent information on the level of education completed along with the racial composition of the high school they attended.11 NSBA respondents were given five response options for the question relating to school racial composition: all blacks, mostly blacks, about half blacks, mostly whites, and almost all whites.12 This information allows us to parse individuals who experienced different high school racial environments.13 Thus we are able to investigate whether the relationship between the racial composition of the high schools black students attended and their educational attainment is systemic.
In a well-known study, Michael Boozer, Alan Krueger, and Shari Wolkon (1992) used data drawn from the NSBA to evaluate whether the racial composition of the high school black youths attended influenced years of education completed. They converted the categorical responses provided by NSBA respondents on the racial structure of the high school they attended into a proportional measure of the share of peers who were black. They did so by assigning numerical values to the Likert-score responses: 1 (all blacks), 0.75 (mostly blacks),0.50 (about half blacks), 0.25 (mostly whites), and 0.1 (almost all whites). Using this continuous measure, they estimated a model of educational attainment under the presumption of a direct and linear relationship between the share of black classmates in high school and the educational attainment of blacks.
We use the five possible responses to identify three types of high schools: white (almost all white or mostly white), racially balanced [End Page 172]
(about half black and half white), and black (mostly black or all black).
In addition, we construct a measure of the share of high school peers who are white, called percent white, following the Boozer, Krueger, and Wolkon methodology. To test for a non-monotonic relationship between educational outcomes and the racial composition of the high school peers, we generate percent white- squared and test a quadratic specification.
The categorical approach to measuring school racial composition allows us to compare schooling outcomes for students attending schools that are highly racially stratified relative to each other and schools that are racially balanced—rather than the effect of a one-unit increase in the share of students who are black.
In our empirical work, we explore the link between the racial composition of the high school attended and educational attainment, using both of these methods (continuous and categorical measures of high school racial composition) to account for the racial makeup of schools. For conciseness, we discuss the results of the categorical model in the body of this article and the continuous findings in the footnotes.
We adopt virtually the same approach in selecting the sample to analyze as Boozer, Krueger, and Wolkon did (1992). We restrict the sample to persons between the ages of twenty- five and sixty-five, who have completed at least nine years of education, and were born in the United States. Therefore, our analysis sample includes 1,121 respondents.14 Following Boozer, Krueger, and Wolkon, our analysis of the association between educational attainment and the racial composition of high school peers uses this sample. In addition, we also estimate a model specification that allows this effect to vary by gender; we find no statistical difference between females and males at conventional levels.
Table 1 provides educational attainment information for the entire sample and for the sample separated into the three categories of the racial composition of the high school attended. Educational attainment is measured in two ways: years of schooling completed and whether the respondent graduated from high school.
Respondents completed an average of 12.6 [End Page 173] years of education, and 77 percent completed high school. Most—approximately two-thirds of the sample—attended a black high school. Those who attended a white high school were the most likely to graduate (87 percent) and attain the highest average years of education (13.2 years), and both of these differ statistically from the educational outcomes at black high schools. Those who attended a racially balanced high school had the lowest average years of education (12.3). Only 74 percent and 76 percent of those black students who attended a black or a racially balanced high school respectively graduated.
We estimate three specifications of equation (1) using ordinary least squares (OLS). Each specification includes the indicator variables for white high school and racially balanced high school while black high school is the reference category.
The dependent variable Schooling indicates the number of years of formal education completed, as in the Boozer, Krueger, and Wolkon study (1992). We also estimate an OLS linear probability model with Schooling as a binary outcome for whether the individual graduated from high school. In model 1, X includes gender, age, age squared, geographic location, and an indicator for educational context: whether the respondent attended high school after the 1964 landmark Civil Rights Act.15
Model 2 also uses equation (1) and adds additional controls for parent education level and family characteristics as a youth. These controls include number of children in their home and their gender, whether the respondent grew up with both biological parents, an indicator of family religiosity, and an indicator that accounts for whether the neighborhood where the respondent grew up was all or mostly black. This last control is included to better isolate the effect on educational attainment of the racial composition of the respondent's high school from the racial composition of the neighborhood in which they lived. Moreover, this model contains an indicator of whether the respondent has light skin shade, a personal characteristic, to account for possible differences in treatment due to colorism (see, for example, Goldsmith, Hamilton, and Darity 2007; Diette et al. 2015).16
Finally, we estimate model 3 by interacting the indicator for gender with the type of racial composition of the high school. This specification allows us to evaluate possible differences in the impact of high school peer racial composition on educational attainment between males and females.17 [End Page 174]
Table 2 presents OLS results for models 1 through 3 for both years of school completed and high school graduate as the dependent variables. We also present chi-squared tests to evaluate the relative impact of attending a racially balanced high school versus a white high school for the sample as a whole. In addition, results for model 3 include test statistics that indicate the relative impact for men and women [End Page 175] separately of attending each type of high school based on peer racial structure.
Our findings for model 1, using years of education as the dependent variable, suggest that attending a predominantly white high school relative to a predominantly black high school is associated with the completion of 0.431 more years and 0.828 years of school relative to a racially balanced school. However, we also find that attending a racially balanced high school is associated with a reduction of 0.397 years of schooling relative to attending a black high school.
In column 2, model 2 reveals that black students who attend a racially balanced high school complete significantly less schooling than those at either a black or white high school. Our findings reveal that those attending racially balanced high schools complete a half year less of education than those attending a black high school and three-quarters of a year less than those at a white high school. Moreover, black students who attend a white high school complete a quarter of a year more of schooling than those who attend a black high school, but this result is not statistically significant.
Model 3 (column 3), interacting female with our categorical measures, reflects the same pattern for both black men and women; the impact of attending a racially balanced high school relative to both black and white high schools is significantly negative and that of attending a white high school relative to a black high school is negligibly positive. Thus model 2, pooling the data across gender groups, is our preferred model specification.
These results are suggestive of a U-shaped negative and convex relationship between black student years of school completed and school racial composition, moving from black to racially balanced to white high schools. We find that the same nonmonotonic pattern between high school peer racial composition and educational attainment exists for both men and women.18
We extend our analysis by assessing the relationship between high school racial composition and a second measure of schooling often of primary concern to policymakers: whether a person graduates from high school. Graduating from high school is an important gauge of educational attainment, especially given the cohort we examine and the substantive wage penalty those who are not high school graduates face.19 The effects of school racial composition on high school graduation for the pooled sample (model 2) are presented in column 5. Black students who attend a white high school are about 9 percentage points more likely to graduate from high school than their counterparts who attend a black high school, and 11 percentage points more likely to graduate from high school than those who attended a racially balanced high school. These results are statistically significant at the 1 percent level. Our estimate of the likelihood of graduating high school for blacks attending a racially balanced high school is negative relative to those who attend a black high school, but this finding is not significant. This same nonmonotonic pattern can be observed for both men and women in column 6; the distinction between the likelihood of graduating from a black or racially balanced high school is not significant, and the likelihood of graduating from a white high school (relative to both of the other types of high school) is significantly higher.20 [End Page 176]
Discussion of Core Model Findings
We find evidence of a nonmonotonic relationship between high school racial composition and both measures of educational attainment. As predicted by the Du Bois hypothesis, black students who attend integrated—racially balanced—schools complete fewer years of education than those who attended high schools with peers who are predominantly black. Similarly, we find that the likelihood of graduating from high school is lower, though not statistically significant, for black students who attend racially balanced high schools relative to those who went to a high school where virtually all of their classmates were black. These findings contradict the conventional hypothesis, which predicts greater educational attainment for black students simply from attending more integrated schools.
Our findings also reveal that black students complete more years of schooling and have a greater likelihood of graduating from high school if they attend a largely white high school instead of a more racially balanced high school. These findings match the conventional hypothesis' assertion that the resource effect outweighs the discrimination effect, and more white peers in high school positively affect black students. Our findings add an empirical nuance to the Du Bois hypothesis that measures the conditions by which the school resource effect may dominate or be dominated by the school discrimination effect associated with a white-black competitive process for school resources. The results are consistent with our contention that predominantly white schools are less racially discriminatory than racially balanced schools in which a critical mass of black students are perceived as more of a competitive threat to white students for preferred resources. Thus, neither the conventional hypothesis nor our characterization of the Du Bois hypothesis fully captures the effects on educational attainment of a typical black student attending high schools that differ across the racial-peer spectrum from a predominantly black high school to a racially balanced high school and a predominantly white high school.
Our evidence is consistent with the functionality of discrimination hypothesis. This idea postulates that the intensity of discrimination that black students face will be greater in more racially balanced high schools than in predominantly black and predominantly white high schools. Essentially, for black students attending a racially balanced high school relative to attending a predominantly black school, the discrimination effect dominates the resource effect and results in fewer years of school completed. The likelihood of graduation is smaller as well but not a statistically significant difference.
However, black students' years of schooling completed and high school graduation are both positively affected by attending a predominantly white high school relative to attending a racially balanced high school (and low resourced predominantly black schools). In this case, the benefits of school resources outweigh the negative effect of discrimination. Thus the pattern of a nonmonotonic relationship between the share of white peers in high school and educational attainment for black students [End Page 177] we find is consistent with the functionality of discrimination hypothesis.
Our finding that Schooling for blacks is not greater at racially balanced high schools, where resources are presumably higher, is striking. This outcome is unlikely to be the result of omitted variable bias or parental selection.21 It is possible that black students attending schools with more white peers than the average black student in our sample (see table 1) have families that especially value education. This may lead them to make greater parental investments in the education of their offspring. To the extent that we do not have information in the NSBA to account for such factors, which are also positively correlated with educational attainment, this leads to an upward, or positive, bias in our estimates of the relationship between attending a racially balanced school and Schooling. Thus, our findings of a significantly negative relationship between a racially balanced high school and educational outcomes are all the more powerful.
Possible Mediating Factors
We further explore the generality of the non-monotonic pattern observed by examining whether the association between racial composition of school types and educational attainment is influenced by select individual and family characteristics. We examine four characteristics that may result in a black student being more or less affected by discrimination in high school.
We assess whether the link between high school racial composition and educational outcomes depends on the respondent's: mother's educational attainment, neighborhood racial composition, educational context, and skin shade. To evaluate whether these factors mediate this link, we use equation (1) to sequentially interact four variables—mother has twelve or more years of education, student grew up in an all or mostly black neighborhood, student attended high school after 1965, and student has light skin—with racially balanced high school and white high school.22 Our analysis is conducted using model 2, which contains our full set of controls. Our results for years of schooling completed and high school graduation are reported in tables 3 and 4.
Table 3 reports the results with years of education as the dependent variable. Each column reveals that for both the reference and the interaction groups, black students who attend a racially balanced high school complete fewer years of education than their counterparts who attend either black or white high schools.
Of the sixteen reported results for these comparisons to racially balanced high school, all have a negative coefficient and thirteen are statistically significant at conventional levels. In addition, our estimates reveal a positive association between attending a white high school [End Page 178]
[End Page 179]
[End Page 180] (relative to a black high school) and acquiring additional schooling; two of the eight reported results for these findings are statistically significant. This pattern reflects the same U-shaped relationship observed in table 3 between educational attainment and high school peer racial structure. This pattern holds across all four of the possible mediating factors.
Recall from our findings reported in table 2 that the probability of graduating from high school is statistically equivalent for those attending racially balanced and black high schools. Table 4 shows that the same pattern holds for both the reference and interaction groups who attend a racially balanced or black high school. In addition, table 4 supports our findings in table 2 for black students who attend a white high school: they are more likely to graduate than their peers in either racially balanced or black high schools. These findings are significant for eleven of the sixteen reported results. This nonmonotonic pattern holds for both the reference and the interaction groups within each of the four family and individual characteristics. Thus our finding of a non-monotonic relationship between high school peer racial composition and educational outcomes is robust to a range of personal and family characteristics.23
HIGH SCHOOL RACIAL COMPOSITION: LINKAGES TO EMPLOYMENT AND HOMEOWNERSHIP
How should we interpret our evidence that black students who attend black schools complete more years of schooling than black students who attend racially balanced high schools? In addition, no evidence indicates that black students are more likely to graduate from a racially balanced high school than from a black high school. Is it possible that black high schools offer a significantly lower quality of education and a greater ease of graduation than racially balanced high schools? If this were the case, one could hypothesize that those who attended schools with fewer white peers would be at a distinct disadvantage in both labor and asset-ownership markets.
Thus, to evaluate the long-term impact of attending high schools with a particular racial makeup, we explore the link between the racial structure of high school attended and both subsequent patterns of employment and home-ownership. We use both the continuous and discrete measures of high school peer racial composition.24 We select these outcomes because employment affects earnings and home-ownership is an important form of wealth accumulation.
Therefore, to shed light on the possible association between Economic Status—employment and homeownership—and the racial composition of the high school attended we utilize an OLS linear probability model to estimate equation (2). This equation is similar to equation (1), which uses a discrete categorization of high school type, respectively, but with employment and homeownership as the dependent variables. Our estimates of equation (2) are conducted using the more complete set of controls described earlier (model 2).
Our analysis is conducted on subsamples of the data based on gender due to structural differences in the determinants of employment [End Page 181]
and asset ownership. Table 5 is a summary table that presents our findings on the connection between the racial composition of the high school and Employment Status (columns 1 and 2) along with Home Ownership (columns 3 and 4) for both black women and men, respectively. Table 5 reveals that black men and women who attended black high schools are no less likely to be Employed than those who attended racially balanced high schools. Our evidence reveals that, relative to those attending more integrated or racially balanced schools, neither black men nor women are punished in terms of employment by attending a black high school.25
Similarly, both black men and women who attended a racially balanced high school are estimated to have a smaller likelihood of home-ownership than those who attended a black high school. However, this is statistically significant for black women.26 Our findings suggest that attending a black high school does not lower quality relative to attending either more integrated or racially balanced schools based [End Page 182] on the connection between these types of schools and subsequent employment and homeownership.
Our findings suggest that in the years following Brown v. Board of Education, the Civil Rights Act, and the subsequent efforts to improve opportunities for black students, desegregation did not work as a cure-all to address racial disparity in educational attainment. The results we present, based on national representative data of black Americans, suggest that the relationship between educational attainment for blacks and the share of their high school peers who are white is U-shaped. This contrasts with the conventional underlying presumption that as high schools whiten in terms of the racial composition of students that educational attainment for black students improves monotonically.
We find a U-shaped negative and convex relationship between school integration and black student years of schooling completed. In addition, we find a nonmonotonic relationship between school integration and high school graduation, where the probability of graduating only differs significantly for black students who attend white high schools. These results, in totality, lend significant support to the functionality of discrimination hypothesis. We find these results despite potential selection bias that would likely favor racially balanced schools. The potential for selection bias does temper our confidence in the positive findings for predominantly white schools.
Indeed, the problem is compounded by the fact that in different communities different routes were taken toward school desegregation. For example, long before the Brown decision, Cincinnati, Ohio, produced integrated schools by giving parents the option of selecting whether their children would go to predominantly black or predominantly white schools. Unsurprisingly, only black parents exercised the option of sending their children to schools of both types (Crowley 1932).
After Brown, some southern communities adopted a parental choice policy putatively to meet the terms of Supreme Court's decision. One example is Drew, Mississippi, where initially only one black family, the Carters, opted to have their children desegregate the all-white schools. As more and more black families followed the Carters' lead, white families withdrew their children from the public schools altogether in favor of all-white private academies, leading to restoration of segregation (Intolerable Burden 2003). The sequential character of desegregation under parental choice suggests that one must account for selectivity in family attributes associated with the process.
In addition, our findings provide an important note of caution to researchers whose empirical models of educational attainment do not account for the nonlinear association between racial composition of peers and schooling. A similar note of caution applies to education activists looking to racial desegregation of schools as a panacea for addressing racial differences in educational attainment.
This is especially important given that any contemporary efforts to desegregate our increasingly segregated schools are more likely to shift black students from black high schools to racially balanced high schools instead of from black to white high schools. In light of our findings, greater integration on a real-world scale as a policy in and of itself is unlikely to improve outcomes for these black students and may actually be harmful.
Today, racially balanced schools appear to be a comparatively poor environment for black student education in terms of opportunities and achievement for two major reasons (Gleibermann 2017; Lewis and Diamond 2015). First, evidence indicates that white students are more likely to be selected into advanced and gifted programs than black students with objectively similar academic records (Grissom and Redding 2016). This results in fewer black student enrolled in these paths to success. Given the importance of being with peers with common experiences and histories, an initial round of assignments made based on race may create a social incentive for black students with strong academic records to opt out of such opportunities (Francis and Darity 2020), leading to poorer educational outcomes for blacks attending racially balanced schools.
Second, evidence suggests that in racially [End Page 183] balanced schools black students, especially males, are more likely to be singled out for detention and discipline (Lindsay and Hart 2017), undermining both educational achievement and future career prospects. Indeed, these conditions may undergird the lack of any major relative improvement in relative black-white academic achievement after more than fifty years of school desegregation (Camera 2016).
Our work suggests that if the goal is to ensure that black students are well educated, then it is time to move away from policies that focus exclusively on school integration. Students thrive when they are educated in resource-rich environments with relevant curriculum and are treated with dignity. Moreover, excellence for all is promoted with relevant curriculum; detracking or at least eliminating the discriminatory aspects of tracking that exclude black students; and eliminating the discriminatory treatment of black students in terms of in-school punishment, suspension, and expulsion. These adjustments—in curriculum, school policies, and management—are feasible and essential for black students to be treated fairly and, regardless of the racial composition of the school, to ensure that the school system has their best interests at heart. We conclude with Du Bois's (1935, 335) prophetic eighty-year-old cautionary observation:
Theoretically, the Negro needs neither segregated schools nor mixed schools. What he needs is Education. What he must remember is that there is no magic, either in mixed schools or in segregated schools. A mixed school with poor and unsympathetic teachers, with hostile public opinion, and no teaching of truth concerning black folk, is bad. A segregated school with ignorant placeholders, inadequate equipment, poor salaries, and wretched housing, is equally bad. Other things being equal, the mixed school is the broader, more natural basis for the education of all youth. It gives wider contacts; it inspires greater self-confidence; and suppresses the inferiority complex. But other things seldom are equal, and in that case, Sympathy, Knowledge, and the Truth, outweigh all that the mixed school can offer.
Timothy M. Diette is professor of economics and senior advisor to the president for strategic analysis at Washington and Lee University, United States.
Darrick Hamilton is Henry Cohen Professor of Economics and Urban Policy and founding director of the Institute for the Study of Race, Stratification, and Political Economy at the New School, United States.
Arthur H. Goldsmith is Jackson T. Stephens Professor of Economics at the Williams School of Commerce, Economics, and Politics at Washington and Lee University, United States.
William A. Darity Jr. is Samuel DuBois Cook Professor of Public Policy, African and African American Studies, and Economics at Duke University, United States.
1. A number of researchers have found that black students complete more years of school, and hence are more likely to finish high school, when they attend schools with a higher white enrollment share (Crain 1970; Boozer, Krueger, and Wolkon 1992; Reber 2010; Guryan 2004; Johnson 2011). However, other studies find no link between educational attainment and more integrated schooling (Rivkin 2000; Bifulco, Fletcher, and Ross 2009). Studies exploring the link between standardized test scores of black students and the proportion of the school they attended that is black are mixed also (for a recent review of this literature, see Card and Rothstein 2007).
2. The concept of Deluxe Jim Crow refers to an imagined regime where separate black and white health-care facilities had similar resources (see Thomas 2011). Correspondingly, Deluxe Jim Crow in the educational sphere refers to no gap in school resources across racially separate schools.
3. Plessy v. Ferguson, 163 U.S. 537 (1896); Brown v. Board of Education of Topeka, 347 U.S. 483 (1954).
4. For instance, Du Bois recollects that he arrived at "Atlanta University to teach history in 1897, without the slightest idea from my Harvard tuition that Negroes ever had any history" (1935, 335).
5. The legal scholar Davison Douglas (2005) offers convincing evidence that during the first half of the twentieth century both racially segregated schooling and a hostile reception for black students in desegregated schools also existed in the northern part of the United States. He documents how most northern states enacted laws prohibiting school segregation following the Civil War, yet in many northern communities—especially those in states that bordered southern states—local school boards acted in defiance of the law and maintained sharply segregated schools.
6. Of course, this view was not shared universally in the black community, including the leadership of National Association for the Advancement of Colored People (NAACP), which he helped found in 1909, an organization at odds with his position on this topic.
7. Many of the factors that concerned Du Bois in the early part of the twentieth century remain intact: better funding for schools that are predominantly white (Condron and Roscigno 2003), a level of racial discomfort borne of limited contact, and the persistence of negative stereotypes. Indeed, Derrick Bell (2005, 1062) looking back on the fiftieth anniversary of the Brown decision, noted that "desegregated schools adopted tracking mechanisms that placed most blacks on non-academic tracks. Black children were disproportionately disciplined, and there was little consideration given to black cultural interests." Given these discrepancies, researchers show that strong differences in the experiences of white and black students are still evident today (Diette 2012; Lucas 1999).
8. Lewis describes how members from a dominant group render individuals from subaltern groups as noncompeting by using their dominant position to limit access to market-rewarded credentials, such as educational degrees, to subaltern groups to maintain their social hierarchical position. He also describes two other strategies that dominant groups deploy when subaltern groups are able to overcome structural barriers and acquire competitive credentials. First, dominant groups can alter the credentialing criteria to favor their own attributes (changing the rules in the middle of a game). In addition, they may simply discriminate by limiting rewards to competitive, or "credentialed" members of the subaltern group (wage discrimination).
9. Today a 13 percent gap in per pupil expenditure for black and white students remains (Morgan and Amerikaner 2018). The gap in expenditures required to equalize services or expected outcomes may be larger, however (Bifulco 2005).
10. When schools were first desegregated with only a small number of black children, such as the Little Rock Nine, the students likely faced severe discrimination. However, after the early phase of desegregation is complete poor treatment of black children in such a setting is expected to be small. Thus, the functionality of discrimination theory we specify applies to a situation where black children are enrolled in mostly white schools that have been integrated long enough for white members of the school community to adapt to desegregation.
11. Respondents who attended a number of different high schools were asked to answer the question for the high school they primarily attended.
12. The share of public school students of Latino ancestry in U.S. high schools in 1972, the furthest year back that the government published such data, was 5.6 percent. The share rose modestly to 6.3 percent in 1979. Thus, NSBA respondents indicating they attended a "mixed" school probably were attending a school that was largely white and black. By 2012, a much larger share of public school students were Latino (23.1 percent of high school students).
13. It is possible that self-reported measures of childhood environmental context are inaccurate. The potential imprecision could be due to many factors including poor recollection, which could worsen with aging, and retrospection bias (framing responses to explain life outcomes). However, a body of scientific literature has emerged suggesting that memories of emotionally salient childhood experiences—such as the racial composition of the high school attended—are clear and reliable (that is, do not substantially change with aging) and largely independent of gender and mood at survey baseline (Corso et al. 2008; Yancura and Aldwin 2009). Thus many scholars now view measures of emotive life-events based on retrospective self-reports as reliable for analysis (Corso et al. 2008; Yancura and Aldwin 2009; Hardt and Rutter 2004; Dube et al. 2004).
14. Our analysis sample contains nineteen more observations than the Boozer, Kreuger, and Wolkon study because their study limited the sample to individuals who completed at least ten years of schooling, whereas we included those who completed nine or more years. Because the respondents ranged in age from twenty-five to sixty-five around 1980, the sample analyzes the high school experiences in the time period from the 1930s to the early 1970s.
15. The Civil Rights Act provided legislative funding mechanisms to enforce school desegregation.
16. Portable skin reflectometers are commonly used in anthropological and biomedical research to obtain reliable quantitative skin tone measurements (see, for example, Jablonski and Chaplin 2000; Keil et al. 1992). However, NSBA interviewers graded respondents on a salient phenotypical dimension, skin shade, using a Likert scale. Prior to conducting interviews, the orientation of NSBA interviewers included training to establish consistency in the coding of respondent skin shade, and reduce the possibility of measurement error. The interviewers used five categories (very dark, dark, medium, light, and very light) when coding to describe the complexion of blacks who participated in the survey. We collapsed the data into two categories: light (which includes very light and light), and dark (containing persons rated as having medium, very dark, and dark complexion). We also estimate the model using the continuous measure of the high school racial composition to replicate the Boozer, Kreuger, and Wolkon study. The only difference is the manner in which geographic location is measured. Boozer, Krueger, and Wolkon use state fixed effects, whereas we use eight regional controls for residence as a youth to allow for greater statistical power. We then add a percentage white squared term to test whether the results of the Boozer, Krueger, and Wolkon study are sensitive to allowing a nonmonotonic relationship between percentage white and schooling.
17. Further, this permits us to assess the validity of pooling the data across gender groups. We acknowledge the potential selection bias due to nonrandom assignment of students to schools and the school's associated racial composition. If parents hold the common view that schools with a higher percentage of white students are likely
18. The interaction terms (female*racially balanced high school and female*white high school) are both statistically insignificant, indicating that the association between attending either of these types of high schools, relative to attending a black high school, is equivalent for men and women.
19. Those who graduate from high school relative to those who complete a GED or drop out have seen their earnings premium increase from 20 percent in 1980 to about 25 percent since 2000 (Murnane 2013; Heckman and LaFontaine 2010; Autor, Katz, and Kearney 2008).
20. When we estimate model 1 with the continuous variable for the school racial composition percent white (but not including a squared term), our findings are very similar in magnitude to the results of Boozer, Kreuger, and Wolkon (1992, table 8). However, when we add additional controls in model 2, the link between percent white and years of school completed becomes substantially smaller and statistically insignificant.
We evaluate whether the relation between Schooling and high school racial composition is nonmonotonic, by including measures for both percent white and percent white-squared in model 2. We find that both of these results are statistically significant at the 5 percent confidence level, with percent white and percent white-squared having negative and positive relationships respectively. We perform a chi-squared test of joint significance of the linear and squared terms, which is positive and statistically significant, which indicates that the racial composition of high school is nonlinearly related to years of school completed. This highlights a U-shaped negative and convex relationship between school integration and black student years of schooling.
Our results for the interaction terms female*percent white and female*percent white-squared are both statistically insignificant at conventional levels, which shows that males and females are not differentially affected by an increase in the share of white high school peers. Using virtually the same data as Boozer, Krueger, and Wolkon (1992) but a different model specification dictated by the functional role of discrimination theory, we offer evidence that runs counter to their claim that school integration is necessarily good for black educational attainment measured by years of schooling completed.
21. Quasi-experimental studies of the link between the racial composition of the schools attended by black youths and their level of educational attainment may suffer from selection bias. For instance, in our case, parents who most highly value education—and thus provide their children more assistance and resources towards education—may be drawn to locate in communities where their children are more likely to attend racially integrated schools because of the greater school resources available. This would lead to the accumulation of more schooling for students who attend racially balanced schools, making it less likely to find, as we do, that black children in racially balanced schools complete less schooling. Thus, we are more confident in our finding that attending school with white peers does not advance educational attainment.
In a cleverly designed study, Rucker Johnson and Alexander Nazaryan (2019) exploit variations in the timing of court-ordered desegregation plans at the municipal or district level to evaluate the effect of school integration on educational attainment. They find that black students who attended schools in districts that underwent desegregation earlier—those who attended integrated school for a longer portion of their education—complete more schooling. This finding is at odds with ours. However, it is possible that court-ordered desegregation plans faced less resistance and thus took place earlier, in communities where black students were likely to face less hostility in school and hence have better educational outcomes. Thus, in their study, selection may contribute to their findings, making it challenging to interpret their results confidently.
22. The reference groups are mother has fewer than twelve years of education, student lives in an integrated neighborhood, student attended high school before 1965, and student has dark skin.
23. Moreover, none of the interacted terms in tables 3 or 4 (interaction*racially balanced high school and interaction*white high school) are statistically significant. This indicates that differences are not discernible for each of the four pairs of reference and interaction variables in the association between attending a racially balanced or a white high school, relative to attending a black high school, on Schooling.
24. Jomills Braddock and James McPartland (1989) analyzed the long-term effects for black students of attending either segregated or integrated schools on occupational status and racial composition of coworkers using data from the National Longitudinal Surveys Youth Cohort. They find that young black adults who attended majority black high schools are less likely both to have white coworkers and to hold a white-collar job than those who attended majority white high schools. Our study is the first we are aware of to examine the association between the racial structure of the high school attended by black males and females and their subsequent employment status and home ownership.
25. The results for the continuous measure of school racial composition are similar. A black person educated in a black high school is no more or less likely to be employed than a similarly situated black person who attends either type of integrated school for women and men.
26. For our sample, 86 percent of females and 87 percent of males who attended a black high school lived in a predominantly black neighborhood as youths. It is possible that the price of housing is lower in communities with a higher share of black high school students, which may explain this result for homeownership by blacks. However, blacks who attended racially balanced schools also mostly lived in black neighborhoods during their youth, 66 percent of females and 61 percent of males. The continuous measure of school racial composition shows statistically significant relationships. Both women and men are more likely to own a home if they attended a black high school rather than a more integrated school.