Parental Educational Attainment and Black-White Adolescents’ Achievement Gap: Blacks’ Diminished Returns

Recent research has documented Minorities’ Diminished Returns (MDRs), defined as weaker protective effects of parental educational attainment and other socioeconomic status (SES) indicators for racial and ethnic minority groups. To explore racial differences in the associations between parental educational attainment and youth educational outcomes among American high schoolers. This cross-sectional study used baseline data from the Education Longitudinal Study (ELS-2002), a nationally representative survey of 10 th grade American youth. This study analyzed 10702 youth who were composed of 2020 (18.9%) non-Hispanic Black and 8682 (81.1%) non-Hispanic White youth. The dependent variables were youth math and reading grades. The independent variable was parental educational attainment. Gender, parental marital status, and school characteristics (% students receiving free lunch, academic risk factors, urban school, public school) were the covariates. Race was the moderating variable. Linear regression was used for data analysis. Overall, higher parental educational attainment was associated with higher math and reading test scores. We found a significant interaction between race (Non-Hispanic Black) and parental education attainment on math and reading test scores, suggesting that the boosting effects of high parental educational attainment on youth educational outcomes might be systemically smaller for Non-Hispanic Black than for Non-Hispanic White youth. While high parental educational attainment promotes educational outcomes for youth, this association is weaker for Non-Hispanic Black youth than non-Hispanic White youth. The diminished returns of parental education are beyond what can be explained by school characteristics that differ between Non-Hispanic Black and non-Hispanic White students. Diminishing returns of parental educational attainment (MDRs) may be an unrecognized source of racial youth disparities. Equalizing SES would not be enough for equalizing outcomes. There is a need for public and economic policies that reduce diminished returns of SES for Black families.


Introduction
Black-White school achievement gap is a major concern in the United States 1 . As academic achievement is an early contributor to later inequalities in life 1 and given that school achievement closely correlates with desired health, developmental, and behavioral outcomes, elimination of racial disparities in school performance is among strategic goals of the US society. This is in part because Black-White achievement gap is believed to be the gateway to future racial health disparities later in life [2][3][4][5][6] .
Worse developmental outcomes of non-Hispanic Black youth compared to Non-Hispanic youth are at least in part attributed to lower socioeconomic status (SES). For example, at least some of the Black-White gap in school achievement is attributed to racial inequalities in parental educational attainment 7,8 . The common belief is that the main reason developmental youth outcomes tend to be worse for non-Hispanic Black youth is lower SES in Black families 9 .
As race closely overlaps with SES indicators (e.g. parental educational attainment) 10,11 and as parental educational attainment has a strong protective influence on academic achievement of youth, researchers have been interested in decomposing the contributions of race from those of SES (e.g. parental educational attainment) in causing Black-White achievement gap [12][13][14] . The results of this line of research are conflicting. While some researchers attribute racial achievement disparities to racial inequalities in SES, other researcher have challenged this traditional assumptions [15][16][17] . For example, some recent research findings suggest that Black-White achievement inequalities sustain at high SES families as well 18 . This is particularly of interest because differential effects of parental educational attainment have been repeatedly shown across domains and outcomes 19,20 . This work suggests that at least some of the Black-White achievement gap in youth is due to diminished returns of SES rather than lack of access to SES resources 19,20 .
MDRs are an overlooked mechanism of racial disparities because this field has traditionally attributes such inequalities to mean differences in SES 19,20 . MDRs, however, suggest that even when racial groups are equal in their mean SES, inequalities in outcomes sustain because of disparities in slope (due to diminished translation of resources to outcomes in Blacks than Whites). Thus, MDRs focus on inequalities that are beyond racial differences in access to SES resources.
Thus, MDRs provide a paradigm shift for understanding the historically overlooked underlying mechanisms behind the racial disparities, particularly those that affect middle class Blacks 19,20 . A major contribution of the MDRs framework is that it provides an explanation for persistence of racial disparities over decades after slavery is over and despite decades of investment made to eliminate inequalities [21][22][23] . MDRs also provide an answer to the questions such as why programs such as head start have had disappointing results in closing the Black-White achievement gap 19,20 .
A literature has been established on MDRs of SES on youth outcomes. Worse than expected educational outcomes are found among Black youth with highly educated and high-income families 24,25,26,27,28 . For example, parental education have shown to better enhance mental health 24 , school performance 25,26 , school attainment 27 , and school bonding 28 for White than Black youth and young adults. These patterns are believed to be systemic and robust as similar patterns (i.e. diminished returns) of parental educational attainment are reported for self-rated health 24 , depression 29,30 , anxiety 31 , obesity 32,33 , asthma 34 , impulse control 35 , attention deficit hyperactivity disorder 36 , health care use 37 , and smoking. 38 With no exception, all these studies have documented worse than expected outcomes for youth from non-Hispanic Black compared to Non-Hispanic White families.
The above research, however, is not conclusive mainly because one remaining dilemma. The main gap in knowledge is that we still do not know to what degree Black-White inequalities in educational quality has a role in explaining MDRs of educational outcomes in Black families. Answering to this research dilemma requires well-controlled studies that can adequately adjust for differences in the school characteristic that Black and White youth tend to attend. For example, we know that Black youth are more likely to attend lowresourced public schools located in urban areas with higher density of low-income Black peers. Contrary to predominantly Black schools, White youth have a higher tendency to attend schools that are located in suburban areas schools with predominantly White students. White youth are also more likely to attend private schools that are known to provide better educational outcomes.
Thus, we still do not know if parental educational attainment generates less tangible outcomes for Black than White youth simply because Black youth have a higher tendency to attend low resourced urban schools which means lower quality schooling. Being able to answer that question is important because it will tell us if eliminating the gap in schooling between White and Black would be enough for equalizing the return of education for Black and White youth. Id the Black-White differences would, however, sustain despite controlling for educational quality, then diminished returns are not merely because of differential school quality but upstream societal process that are beyond educational system (e.g. racism, discrimination, labor market practices, etc.). As mentioned above, we are not aware of any previous studies on diminished returns of parental education on youth outcomes, which has fully controlled for Black-White differences in school characteristics.
Building on our prior research on diminished returns of SES indicators 19,20 and using a nationally representative data that has rich data on school characetristics 39 , this study compares non-Hispanic Black and non-Hispanic White youth for the associations between parental educational attainment and math and reading standard test scores. As explained above, the unique contribution of the current study is that it is one of the firsts to control for educational quality. If MDRs remain after controlling for school characteristics, we would hypothesize that MDRs are probably due to upstream social factors beyond education system (e.g. racism, marginalization, and discrimination) that disproportionately impact daily lives of Blacks.

Design and settings
This cross-sectional study was a secondary analysis of Wave 1 of the Education Longitudinal Study (ELS) 39 . The ELS sample is representative of United States youth at 10 th grade. Funded by the US Department of Education, ELS is a state-of-the-art study of educational outcomes of American youth. Although ELS has enrolled 10 th graders across all race/ethnic groups, we only included 10702 youth who were composed of 2020 (18.9%) non-Hispanic Black and 8682 (81.1%) non-Hispanic White youth. Any student with mixed race, missing data on race, or race other than White or Black was excluded. This included individuals who were Asian American, American Indians/Alaska Natives, mixed race, or unknown race/ethnicity. We also excluded youth who reported Hispanic/Latino ethnicity.

Ethics
All youth who participants in the ELS study provided written assent. Youth parents or guardians also provided written informed consent. The institutional review board (IRB) of the Department of Education approved the ELS study protocol. Given the publicly available data, the current secondary analysis was exempted from a full review according to the rules of National Institute of Health (NIH) as well as Charles Drew University of Medicine and Science.

Sample and sampling
The ELS study's youth samples in Wave 1 were enrolled in the private, public, or Catholic schools in Urban, Suburban, or Rural settings. The ELS study used a multi-stage stratified probability sampling to recruit the participating youth. The analytical sample was 10702 youth.

Study variables
The study variables were as follow: race/ethnicity (moderator), parental educational attainment (predictor/independent variable), youth math and reading test scores (outcomes/ dependent variables), and demographic factors [gender, family income, number of siblings, and family structure] and school characteristics [% students receiving free lunch, academic risk factors, urban school, public school] (covariates). All the study variables were measured at an individual level.
Race.-Race (1 non-Hispanic Black versus 0 non-Hispanic White) was self-identified. Race/ethnicity was operationalized as a dichotomous variable.
Parental Educational Attainment.-Parent educational attainment was a three-level categorical variable: 1) less than high school graduate, 2) high school graduate, 3) college graduate.
Demographic Factors.-Gender, region, and family structure were demographic covariates. Family structure was a dichotomous variable (1 married, 0 unmarried) and calculated based on parents' marital status. Gender was 1=male 0 = female. Region was a nominal variable: Northeast, Midwest, South, and West.
School characteristics.-School characteristics included urban school, catholic school, rural school, public school, and % students receiving free lunch.
Outcomes.-Our dependent variables were standardized test score of math and reading. These variables were transformed to z score which helps comparison of the students and interpretation of the regression coefficients.

Statistical Analysis
We analyzed the ELS Wave 1 data using SPSS 23.0 (IBM Corporation, Armonk, NY). To analyze the ELS data, we needed to adjust for survey weights due to the multi-stage sampling design of the study (clustered stratified sampling). As we adjusted for survey weights, the results are representative of the U.S. youth population. Taylor series linearization was applied to re-estimate the variance of our variables. We had normally distributed outcomes thus could perform linear regression. We also did not find evidence of multicollinearity. Our model passed the assumption of homoscedasticity (e.g., random distribution of error terms). As we had two outcomes, we ran similar models for each outcome. This strategy also helped us with the comparability of MDRs across our two outcomes. We ran two hierarchical linear regression models per outcome, in the pooled sample that included non-Hispanic Whites and non-Hispanic Blacks. The first block of variables only included race/ethnicity, gender, region, and parental marital status. Our second block included educational attainment (high school graduation, college graduation). Our third block included school characteristics. Our fourth block included two interaction terms between educational attainment (high school and college graduation) and race/ ethnicity. From our linear regression models, we reported beta (b), 95% Confidence Intervals (95% CI), and p values. P values less than 0.05 were considered as statistically significant. Table 1 summarizes descriptive statistics for our sample. This study included 10702 American 10 th grader youth. This number was composed of 2020 (18.9%) non-Hispanic Black and 8682 (81.1%) non-Hispanic White youth. Non-Hispanic Black students were more likely to attend urban, public schools with higher % of students receiving free meal. Table 2 presents the mean (SD) of our education outcomes based on the intersections of race/ethnicity and parental educational attainment. This table shows how students' math and reading scores change as a function of educational attainment for non-Hispanic White and non-Hispanic Black youth. Although a significant trend existed in both racial groups, the magnitude of change was larger for non-Hispanic White than non-Hispanic Black youth. Table 3 present the summary of two hierarchical linear regression models in the pooled sample. Based on the man model, race (non-Hispanic Black) and parental educational attainment were associated with the outcome. Model 2 showed a statistical interaction between race and parental educational attainment on youth reading score. This interaction suggests that the boosting effect of high parental educational attainment on youth reading grade is smaller for non-Hispanic Black than for Non-Hispanic White youth. That is non-Hispanic Black youth have reading score even when they have highly educated parents, which is indicative of Blacks' diminished returns of parental education on reading score regardless of school quality. Table 4 present the summary of two hierarchical linear regression models in the pooled sample. In these models, race (non-Hispanic Black) and parental educational attainment were the independent variables and math score was the outcome. Both race and parental educational attainment were associated with math score. Model 2 showed a statistical interaction between race and parental educational attainment on youth math performance. This interaction suggests that the boosting effect of high parental educational attainment on youth math grade is smaller for non-Hispanic Black than for Non-Hispanic White youth. This means that non-Hispanic Black youth on average would have low math grades even when they have highly educated parents. This finding is suggestive of Blacks' diminished returns of parental education on math performance regardless of education/school quality.

Discussion
The current study showed that (a) overall, high parental educational attainment is associated with higher math and reading score in youth, however, (b) these associations are weaker for non-Hispanic Black than for Non-Hispanic White families. That means, youth with highly educated parents would still remain at educational risk, a level of risk which is unexpected and disproportionate to their parental educational attainment.
As MDRs remained despite adjusting for Black-White differences in school characteristics, we hypothesize that a probable cause for the observed diminished returns of parental educational attainment on youth school outcomes and also health outcomes (shown by previous studies) are some upstream social forces not merely lower education quality of Blacks 26 .
In a recent paper 26 , we found higher than expected tobacco dependence, aggression, psychological distress, and chronic diseases, and also worse school performance in Black youth with high parental education. Similar to other studies, 40-42 a plausible conclusion seem to be some upstream and distal social processes that may diminish the effects of parental education for non-White families 26 . Thus, the MDRs of parental education is not all because Black youth attend worse schools and receive lower quality schooling.
Educational inequalities may not be the only reason we see worse outcomes for Black youth in middle class with highly educated parents. Research has shown that risk of asthma 34 [48][49][50] , and Hispanic youth 26 . The universal nature of these patterns points to the hypothesis that the upstream underlying mechanisms such as social stratification, structural racism, and marginalization 26 .
These results have considerable implications. Racial inequalities and disparities are not all due to lower SES of Blacks as inequalities can be also seen in middle class people. Thus, other social mechanisms are at work to cause inequalities across racial groups, even for the families with the highest levels of parental education and human capital.
Bold and innovative policies and public health programs are needed to reduce racial disparities that sustain across SES levels and expand to middle class families. Since some of the inequalities and disparities are shaped by the differential effects of SES, the type of policies that are needed that go beyond exclusively focusing on equal access and also address the broader social processes that place middle class Black families at a relative disadvantage. As these patterns are national and systemic, there is a need for national as well as local policies that specifically equalize the return of family SES. Such policies may reduce inequalities that occur in high SES levels 20,31,33,35,37,43,44,48,51,52 . We need policies and program solutions that equalize highly educated Black families abilities to leverage their educational attainment 19,20 . Some suspect cause of MDRs are labor market practices and preferences 44 . There are strong anti-discrimination laws, however, enforcement of such existing policies may be needed to minimize the exiting diminished returns of educational attainment among Black families. Communities where the majority of residents are Black may benefit from higher and more quality jobs that facilitate translation of educational attainment into tangible real life outcomes 53 . Programs should help highly educated Black parents successfully compete with Whites and secure high paying jobs. At the same time, we may need to minimize the societal and environmental barriers that are common in the everyday life of Black population. At the same time, we should invest in educational programs and investment in urban public schools that Black youth attend. Finally, we need to minimize how Black youth are treated in urban schools 1,54 .
This study had a few methodological limitations. Due to the cross-sectional design of our study, we cannot make any causal inferences. An imbalanced sample size across racial groups prevented us from running race-specific models. This study only included Non-Hispanic Blacks and Non-Hispanic Whites. Other ethnic minorities such as Hispanics, Asians, and Native Americans should be included in future studies. We only studied the differential effect of parental educational attainment. Other family SES indicators such as wealth, income, employment, and area level SES should be studied. This study did not include geocoded data. Thus, educational policies were not included. Despite these limitations, this study still makes an important contribution to the existing literature on MDRs as well as the racial gap in school achievement. Some strengths included a large sample size, a random sample, a representative sample that resulted in generalizable findings to the US, and standardized tests. This was the first study that tested MDRs and also controlled for school characteristics that non-Hispanic Black and non-Hispanic White youth attend.

Conclusion
In the United States, non-Hispanic Black youth remain at a relative disadvantage compared to Non-Hispanic White youth regarding the magnitude of the association between parental educational attainment and their educational outcomes. Such diminished returns of parental education do not seem to be fully due to school differences that non-Hispanic White and non-Hispanic Black youth tend to attend. Assari  Descriptive statistics in the overall sample and by race/ethnicity (10,702