Abstract
Racism is embedded in society, and higher education is an important structure for patterning economic and health outcomes. Historically Black Colleges and Universities (HBCUs) were founded on antiracism while predominantly White institutions (PWIs) were often founded on white supremacy. This contrast provides an opportunity to study the association between structural racism and health among Black Americans. We used the National Longitudinal Study of Adolescent to Adult Health (Add Health) to estimate the long-term causal effect of attending an HBCU (vs. PWI) on depressive symptoms among Black students in the United States from 1994–2018. While we found no overall association with attending an HBCU (vs. PWI) on depressive symptoms, we found that this association varied by baseline mental health and region, and across time. For example, among those who attended high school outside of the South, HBCU attendance was protective against depressive symptoms 7 years later, and the association was strongest for those with higher baseline depressive symptoms. We recommend equitable state and federal funding for HBCUs, and that PWIs implement and evaluate antiracist policies to improve mental health of Black students.
Abbreviations
- Add Health
National Longitudinal Study of Adolescent to Adult Health
- CES-D
Center for Epidemiologic Studies Depression Scale
- HBCU
Historically Black College or University
- PWI
predominantly White institution
Mental health is an essential aspect of well-being. It affects physical health, feelings, social interactions, and choices (1). Depressive disorders are a major contributor to diminished health and are associated with suicide, which is a leading cause of death for people aged 15–29 years (2). Adolescence and early adulthood are critical periods in the life course for mental health because most adult mental health disorders originated in their youth (3).
Racial inequities in mental health are persistent and widespread, with Black Americans experiencing worse depressive symptoms (4) and psychological distress (5) compared with White Americans. Among people with psychiatric diagnoses, disorders are more persistent for Black individuals (6). However, White people exhibit higher rates of mental disorder diagnoses including major depression, a paradox that is still being explored (7). Racial health disparities are well-documented; however, understanding their causes requires investigating the role of racism.
Racism is a larger system of oppression that operates at multiple levels, including cultural racism (valuing White culture above others (8)) and structural racism (society’s unequal allocation of opportunities, resources, and power according to race (9, 10)). Education is a strong driver of racial inequities, including racial health inequities (11). Racial education disparities are large, with Black and American Indian students half as likely to graduate within 4 years compared with White and Asian students (12). People who graduate from college have better outcomes across multiple domains (13), but little is known about the impact on health of structural racism within educational institutions. Articles in leading higher education journals tend to ascribe student-level reasons for racial disparities, with only 16 of 255 articles about racial disparities engaging with the concept of racism thoroughly enough to use “racism” or “racist” more than twice (14).
Historically Black Colleges and Universities (HBCUs) were established when other colleges and universities prohibited Black students from attending (15). The Higher Education Act of 1965 defines an HBCU as a nationally accredited college or university established prior to 1964 whose principal mission was/is educating Black Americans (16). Postsecondary schools that were established primarily for White students are referred to in this paper as predominantly White institutions (PWIs). Until the 1950s and 1960s, almost all Black college graduates in the United States attended HBCUs (17). Today, there are 101 private and public HBCUs in the United States, mostly located in the South and East, and they still play an integral part in this country’s college landscape (15). Of all bachelor’s degrees awarded to Black students in the United States in 2020, HBCUs awarded 13% of them (18).
In these analyses, we theorize that Black students at HBCUs are exposed to less structural racism during college than Black students at PWIs. By comparing experiences of Black Americans at HBCUs and at PWIs, we aim to estimate the effects of structural antiracism vs. structural racism in higher education, and its impact on mental health.
Conceptual framework
Figure 1 shows our conceptual framework, which illustrates our analytical strategy as well as our theoretical considerations. It is arranged in a temporal order, and depicts factors that precede college, moving through college (at either an HBCU or a PWI), then after college, to then affect mental health among Black people who attend college. The factors are grouped into categories to show the levels—individual, family, school, and place-based (19)—that link HBCU or PWI attendance and mental health.
Conceptual framework for the relationship between college type (Historically Black Colleges and Universities (HBCUs) vs. predominantly White institutions (PWIs)) and later depressive symptoms.
A range of factors influence enrollment in an HBCU, and these can be conceptualized as confounding or selection factors if they differ from those preceding enrollment in PWIs. Such precollege factors are also an important part of college students’ outcomes and are depicted in the first box on the left in Figure 1, “precollege confounders and effect modifiers.” These are variables we conceptualized first as confounders. For instance, Black students at HBCUs begin college with lower high-school grades, lower standardized test scores, and parents with less education and lower incomes compared with Black students at PWIs (20). This also builds on the literature that links individual-level factors—such as race, gender, and lower socioeconomic status—to higher mental health risk (2, 21). We then further conceptualized and modeled some of these variables as potential effect modifiers.
Black students at HBCUs may experience less structural racism than Black students at PWIs through a variety of mechanisms. The second box, “college exposure,” depicts the HBCU college experience compared with that at a PWI and factors associated with them, such as the social environment, curriculum, faculty support, and campus resources. Black students at PWIs get less social support and experience more discrimination than Black students at HBCUs (22). This could be important for later mental health because perceived racism is associated with psychological distress among Black Americans (23). HBCUs have more African-American studies courses, providing a perspective other than White people’s, which is often the higher education lens (24). In addition, graduates from HBCUs have higher salaries after controlling for precollege factors (25). Still, however protective HBCUs’ environments may be, they exist within a larger white supremacist society and may therefore be deficient in some ways. For example, PWIs have access to more funding (26) at the state and federal level as well as in private endowments, which can have an impact on students’ experiences (i.e., infrastructure and student services) (27).
The third gray box, “postcollege mediators,” conceptualizes effects of college (PWI or HBCU) that might lead to differences in mental health. At the individual level, Black students who attend HBCUs have different scores on measures of racial identity compared with Black students who attend PWIs (28). Differences in racial identity might have implications for mental health. For example, Black college students for whom being Black is a central part of their identity, and who believe negative stereotypes about Black people, experience more psychological distress (29).
Social and economic pathways between a college education and health are also hypothesized, including income, occupational prestige, employment (13), social networks, insurance coverage (30), health knowledge, social status, and job quality (31). All of these pathways could be altered by HBCU vs. PWI attendance. For instance, HBCUs might provide social capital—a network of trusted relationships that help a community thrive—for Black Americans in a way that PWIs do not (32). At the same time, it is possible that there is a penalty in the job market for being identifiable as a Black applicant by listing one’s postsecondary school, which could compound across one’s career.
Across all the pathways in this conceptual framework, depicted in the arrow at the bottom, are structural racism, cultural racism, and interpersonal racism, emphasizing that racism is present in Black students’ lives regardless of the type of college they attend (10).
The model is informed by previous models (31, 33) and by intersectionality (34), which posits that marginalization resulting from oppression related to aspects of a person’s identity—such as race, gender, socioeconomic status, etc.—intersect to produce unique and synergistic forms of interlocking advantage/disadvantage. One way we incorporated that idea into our analysis was in considering that the relationship between HBCU attendance and depressive symptoms may vary by a variety of baseline characteristics (potential effect modifiers). We also drew upon critical race theory, which centers the experiences of marginalized groups while emphasizing the pervasiveness of racism (35). According to critical race theory, within-group analyses (e.g., among Black Americans) subvert the frame of whiteness as being the norm and can generate new information about structural inequities (36).
HBCUs and health
Although HBCUs are major players in postsecondary education in the United States, there is has been little research on how HBCU attendance is associated with mental health compared with PWI attendance. Cross-sectional surveys found better health for Black students enrolled at HBCUs, such as less drinking and fewer mental health conditions (37), better body image (38), and more social support (39). One longitudinal analysis found that Black students at HBCUs had a reduced risk of metabolic syndrome 7 years later (40). However, there is no research on the long-term mental health effects of attending an HBCU. Using the National Longitudinal Study of Adolescent to Adult Health (Add Health), this study tested whether attending an HBCU was associated with long-term depressive symptoms among Black students, compared with Black students who attended PWIs, and whether the association varied by baseline characteristics.
METHODS
Data
This study uses the National Longitudinal Study of Adolescent to Adult Health—Add Health—which began in 1994–1995 as a nationally representative multistage sample of seventh to twelfth graders (41). High schools were sampled based on size, region, racial composition, school type, and urban/rural setting. From 80 high schools and 52 middle schools, students were randomly selected with oversampled groups, including Black students with highly educated parents. The first wave had 20,745 participants, and the most recent (fifth) wave occurred in 2016–2018, when participants were 32–42 years old.
Our analyses include Add Health participants identifying as Black, in college at wave III (ages 18–26), who participated in survey waves III, IV, and V. In the unweighted Add Health sample, 23% of participants identified as Black at wave 1 (n = 4,807), and of those, 21% (n = 842) were enrolled in college at wave III when HBCU status was measured. An additional 317 were lost to attrition in waves IV or V, 9 did not complete all the Center for Epidemiologic Studies Depression Scale (CES-D) items, and 28 participants in the remaining sample did not have weights for waves I through IV, and were therefore not included in the analyses. This left a final analytical sample of 488 participants.
Measurements
Enrollment at an HBCU.
In wave III (2001–2002, ages 18–26), if participants reported the college’s name, Add Health linked it to the US Department of Education’s Integrated Postsecondary Education Data System (IPEDS) data. This allowed Add Health to link institutional-level variables, including whether the school was classified as an HBCU. We denote this variable as “college type,” or “attending an HBCU,” and we refer to non-HBCU schools as PWIs.
Depressive symptoms: waves III, IV, and V.
The outcome is the sum of 3 items from the CES-D. We used the 3 items available across waves and used by Add Health investigators for longitudinal analyses (42). The items ask: How often was each of the following true during the past week: 1) you felt like you could not shake off the blues even with help from your family and friends; 2) you felt depressed; 3) you felt sad. Responses are coded 0: less than 1 day; 1: 1–2 days; 2: 3–4 days; and 3: 5–7 days. The responses are summed (range, 0–9), with higher scores indicating more depressive symptoms.
Baseline covariates and potential effect modifiers: wave I.
In wave I, participants were asked “What is your race? You may give more than one answer.” Anyone who responded “Black or African American” was eligible to be included in this analysis. Every other racial/ethnic group had 5 or fewer participants who attended an HBCU at wave III, and was therefore excluded. Participants self-reported age, sex, overall health, depressive symptoms, grades, church attendance, and whether they were born in the United States. The female household head reported parental education and household income. School administrators reported average class size and proportion of teachers with a master’s degree. The high school’s region is from the Quality Education Database. The participants’ baseline addresses were linked to 1990 Decennial Census–based measures of the census tract, including the percent Black, percent below poverty, and percent without a high-school degree. Grade point average was the average of reported grades for math, science, English, and history. Add Health coded the reported grades as A = 1, B = 2, C = 3, D or lower = 4.
We used the conceptual framework to guide our selection of covariates. We included a baseline covariate in all regression models if it was a significant predictor (P < 0.2) of depressive symptoms at any of the 3 waves; if a variable was not a significant predictor of depressive symptoms, then it could not confound the association (43). The final set of baseline covariates was CES-D, sex, self-rated health, parent education, church attendance, grade point average, proportion of high-school teachers who have a master’s degree, region, and proportion of census tract who were in poverty, without a high-school degree, or Black.
Data analysis
We examined distributions of the variables overall and by HBCU vs. PWI. To test for selection into an HBCU (vs. PWI), we performed statistical tests of difference of baseline covariates by college type. We used linear regression to test whether attending an HBCU (vs. PWI) at wave III (ages 18–26 years) was associated with higher or lower depressive symptoms during college, 7 years later, and 15 years later. We adjusted all regressions for the complex multistage, stratified, oversampled, longitudinal design (weights specific to outcome wave, and strata variable for high school region), baseline depressive symptoms, and baseline covariates selected because they preceded college enrollment and were conceptually and statistically associated with college type and depressive symptoms.
Effect modification testing
We tested whether the relationship between HBCU (vs. PWI) attendance and depressive symptoms differed by baseline covariates, including depressive symptoms, sex, parent education, proportion of high school teachers with a master’s degree, household income, region, and percentage of the neighborhood who were Black, without a high school degree, or below poverty. Finally, to examine whether the results could be explained by differential rates of degree attainment, we tested bachelor’s degree as an effect modifier. We tested statistically for multiplicative effect modification by modeling interactions between HBCU attendance and each baseline variable—one model for each interaction using a product term. For binary and continuous variables, we interpreted the P value of the interaction term to determine significance at P < 0.05. For categorical variables, we ran an overall test to determine the significance at P < 0.05.
The sample of 488 participants was used in all analyses. We used survey weights recommended by Add Health (44), which account for attrition and the complex sample design. We performed multiple imputation in order to reduce bias from missing data at baseline. All analyses were performed using Stata, version 15.1 (StataCorp LLC, College Station, Texas).
Missing data imputation
Add Health recommends imputing missing items only for participants who completed surveys (44). Therefore, we focused on imputing baseline missing data, which is where the most items were missing. Before missing data imputation, most of the baseline covariates had less than 5% missing values, with the exceptions of household income (24%), self-reported grades for grade-point average calculation (30%), and being born in the United States (21%).
The missing data imputation model used for all model results and for Figure 2 used the “ice” command in Stata 15.1 and created 5 imputed data sets. All baseline covariates listed in Web Table 1 (available at https://doi.org/10.1093/aje/kwac199) were included as predictors for all the covariates. The weight for wave V was used in the imputation model because the analytical sample is defined by who was present at wave V. Web Table 1 shows descriptives and missingness for all variables before multiple imputation, and Web Table 2 shows descriptives after multiple imputation.
Baseline characteristics with statistically significantly differences (P < 0.05) according to college type (historically Black vs. predominantly White), National Longitudinal Study of Adolescent to Adult Health, United States, 1994–2002. Proportions are survey weighted but not adjusted for any covariates. See Web Table 1 for all baseline characteristics according to college type. HBCU, Historically Black College or University; PWI, predominantly White institution; HS, high school.
Survey weights
Add Health instructs analysts to use the “subpopulation” commands rather than restricting the data set to our analytical sample, in order to produce correct standard errors. However, the subpopulation commands were not compatible with multiple imputation and survey commands in Stata. In preliminary analyses, we compared 1) using the subpopulation commands with 2) restricting the data, and the coefficients and standard errors were identical. Therefore, we restricted the data in final models rather than use subpopulation commands.
While we originally proposed a longitudinal hypothesis with repeated outcomes, the complexity of doing so while also applying the survey weights and the multiple imputation was not supported by Stata. The weights, which are recommended by Add Health and account for sampling and nonresponse, and the imputation, which corrects for bias from important missing baseline covariates, were too important to leave out. Therefore, we decided to forgo the longitudinal model and instead model the outcome at each wave in its own separate linear regression model, and use the same sample in each of the 3 models to make them comparable across waves.
RESULTS
Of the 488 Black participants in our sample, 24% attended an HBCU and 76% attended a PWI in 2001–2002. The baseline covariates at wave I (seventh to twelfth grade) that predicted selection of Black college-going students’ enrollment at an HBCU (vs. PWI) were geography-related variables (see Figure 2). For example, Black students who enrolled at HBCUs (vs. PWIs) were more likely living in the South at baseline (vs. West, Midwest, Northeast; P < 0.0001) and more likely to come from neighborhoods that had a higher proportion of Black residents (P = 0.0006), higher proportion of residents without a high-school degree (P = 0.04), and higher proportion of residents living in poverty (P = 0.0008). We saw no selection (e.g., no differences) across college type for the rest of the variables—see Web Table 3.
Adjusted for covariates, attending an HBCU (compared with a PWI) for Black students was not associated with CES-D scores at any of the 3 time points measured (see Table 1). However, there were 2 significant interactions with HBCU enrollment on depressive symptoms in wave IV, 7 years after college enrollment (Figure 3). First, there was a significant interaction of depressive symptoms at baseline and HBCU attendance on depressive symptoms in wave IV (7 years after college enrollment). Second, there was a significant interaction of HBCU attendance and the region where the student attended high school on depressive symptoms.
Linear Regression Results for the Main Effect of Attending a Historically Black College or University vs. Attending a Predominantly White Institution on Mean Depressive Symptoms Among Black College Students in the National Longitudinal Study of Adolescent to Adult Health, United States, 1994–2018
| Wave of Depressive Symptoms Outcomea | Years | Age, years | Coefficientb | SE | P Value | 95% CI |
|---|---|---|---|---|---|---|
| Wave IIIc | 2001–2002 | 18–26 | −0.042 | 0.181 | 0.815 | −0.403, 0.318 |
| Wave IV | 2008 | 24–32 | 0.134 | 0.206 | 0.517 | −0.276, 0.544 |
| Wave V | 2016–2018 | 32–42 | 0.354 | 0.233 | 0.132 | −0.109, 0.818 |
| Wave of Depressive Symptoms Outcomea | Years | Age, years | Coefficientb | SE | P Value | 95% CI |
|---|---|---|---|---|---|---|
| Wave IIIc | 2001–2002 | 18–26 | −0.042 | 0.181 | 0.815 | −0.403, 0.318 |
| Wave IV | 2008 | 24–32 | 0.134 | 0.206 | 0.517 | −0.276, 0.544 |
| Wave V | 2016–2018 | 32–42 | 0.354 | 0.233 | 0.132 | −0.109, 0.818 |
Abbreviations: CES-D, Center for Epidemiologic Studies Depression scale; CI, confidence interval; SE, standard error.
aThree-item CES-D.
bAdjusted for baseline covariates including CES-D score, sex, self-rated health, parent education, church attendance, grade point average, proportion of high-school teachers who have a master’s degree, region, and proportion of census tract in poverty, proportion of census tract without a high-school degree, and proportion of census tract Black.
cAttendance at Historically Black College or University or predominantly White institution was modeled at wave III.
Linear Regression Results for the Main Effect of Attending a Historically Black College or University vs. Attending a Predominantly White Institution on Mean Depressive Symptoms Among Black College Students in the National Longitudinal Study of Adolescent to Adult Health, United States, 1994–2018
| Wave of Depressive Symptoms Outcomea | Years | Age, years | Coefficientb | SE | P Value | 95% CI |
|---|---|---|---|---|---|---|
| Wave IIIc | 2001–2002 | 18–26 | −0.042 | 0.181 | 0.815 | −0.403, 0.318 |
| Wave IV | 2008 | 24–32 | 0.134 | 0.206 | 0.517 | −0.276, 0.544 |
| Wave V | 2016–2018 | 32–42 | 0.354 | 0.233 | 0.132 | −0.109, 0.818 |
| Wave of Depressive Symptoms Outcomea | Years | Age, years | Coefficientb | SE | P Value | 95% CI |
|---|---|---|---|---|---|---|
| Wave IIIc | 2001–2002 | 18–26 | −0.042 | 0.181 | 0.815 | −0.403, 0.318 |
| Wave IV | 2008 | 24–32 | 0.134 | 0.206 | 0.517 | −0.276, 0.544 |
| Wave V | 2016–2018 | 32–42 | 0.354 | 0.233 | 0.132 | −0.109, 0.818 |
Abbreviations: CES-D, Center for Epidemiologic Studies Depression scale; CI, confidence interval; SE, standard error.
aThree-item CES-D.
bAdjusted for baseline covariates including CES-D score, sex, self-rated health, parent education, church attendance, grade point average, proportion of high-school teachers who have a master’s degree, region, and proportion of census tract in poverty, proportion of census tract without a high-school degree, and proportion of census tract Black.
cAttendance at Historically Black College or University or predominantly White institution was modeled at wave III.
Predicted mean of Center for Epidemiologic Studies Depression scale (CES-D) depressive symptoms in Black students 7 years after enrollment in college (wave IV), according to enrollment in a Historically Black College or University (HBCU) vs. a predominantly White institution (PWI), modified by baseline CES-D depressive symptoms and participant’s high school region, National Longitudinal Study of Adolescent to Adult Health, United States, 1994–2008. A) High school (HS) in the South; B) HS not in the South. CES-D modeled continuously in the linear regression model and graphed based on postestimation regression commands. Two interactions are included: HBCU × region and HBCU × baseline CES-D. Adjusted for baseline covariates including CES-D score, sex, self-rated health, parent education, church attendance, grade point average, proportion of HS teachers who have a master’s degree, region, proportion of census tract in poverty, proportion of census tract without a HS degree, and proportion of census tract Black.
The association between HBCU (vs. PWI) and depressive symptoms differed for subgroups. For example, Black students who exhibited lower depressive symptoms in middle and high school were not affected by enrollment in an HBCU vs. PWI. However, for students with higher baseline depressive symptoms, there was a protective association between attending an HBCU and later depressive symptoms, when participants were age 24 or older. The beneficial association of attending an HBCU on later mental health among people with high baseline depressive symptoms was especially pronounced for students who attended a high school that was not in the South (i.e., a high school in the Northeast, West, or Midwest). Figure 3 shows the linear regression–predicted probabilities of depressive symptoms for a linear regression model with each of these 2-way interactions (HBCU × baseline depression, HBCU × region) in a single model. The 3-way interaction was not significant and therefore not included in the final model. Full regression results for the main effect models and the interaction models at each wave are presented in Web Tables 3 and 4.
At wave IV (ages 24–32), there was a weak interaction between obtaining a bachelor’s degree and attending an HBCU for depressive symptoms (ages 24–32) (P = 0.15), in which students who attended an HBCU, and did not finish their degree, experienced higher depressive symptoms compared with students who graduated. For Black students who attended a PWI, depressive symptoms were the same whether they graduated or not. At wave V (ages 32–42), there was not an interaction between a bachelor’s degree and HBCU status on depressive symptoms (P = 0.99) (Table 2).
Linear Regression Results for the Interactions Between Baseline Covariates and Attending a Historically Black College or University vs. Attending a Predominantly White Institution on Mean Depressive Symptoms Among Black College Students in the National Longitudinal Study of Adolescent to Adult Health, United States, 1994–2018
| Interaction Between Baseline Coefficients and HBCU Status | Years | Age, years | Interaction Coefficienta | SE | P Value | Group Test P Value |
|---|---|---|---|---|---|---|
| Wave III | 2001–2002 | 18–26 | ||||
| Baseline depressive symptoms × HBCU | −0.03 | 0.15 | 0.85 | |||
| Wave IV | 2008 | 24–32 | ||||
| Baseline depressive symptoms × HBCU | −0.27 | 0.11 | 0.01 | |||
| Male × HBCU | 0.15 | 0.43 | 0.72 | |||
| Parent education × HBCU (referent: college) | 0.44 | |||||
| Less than HS | 1.16 | 0.92 | 0.24 | |||
| HS degree | 0.32 | 0.46 | 0.50 | |||
| Proportion of neighborhood below poverty × HBCU | −0.15 | 1.36 | 0.92 | |||
| Proportion of neighborhood Black × HBCU | −0.03 | 0.71 | 0.96 | |||
| Proportion of neighborhood with no HS degree × HBCU | −0.15 | 1.62 | 0.93 | |||
| Percent of school’s teachers with a master’s degree (0–100) × HBCU | 0.01 | 0.01 | 0.49 | |||
| Household income (in thousands) × HBCU | 0.00 | 0.01 | 0.54 | |||
| Region × HBCU (referent: South) | 0.004 | |||||
| West | 0.52 | 0.51 | 0.32 | |||
| Midwest | −0.92 | 0.45 | 0.05 | |||
| Northeast | −1.48 | 0.66 | 0.03 | |||
| Bachelor’s degree at wave 4 × HBCU | −0.69 | 0.47 | 0.15 | |||
| Wave V | 2016–2018 | 32–42 | ||||
| Baseline depressive symptoms × HBCU | 0.14 | 0.11 | 0.20 | |||
| Male × HBCU | −0.54 | 0.31 | 0.08 | |||
| Parent education × HBCU (referent: college) | 0.09 | |||||
| Less than HS | −1.12 | 0.36 | 0.01 | |||
| HS degree | −0.08 | 0.23 | 0.72 | |||
| Proportion of neighborhood below poverty × HBCU | −0.03 | 1.28 | 0.98 | |||
| Proportion of neighborhood Black × HBCU | −0.30 | 0.74 | 0.68 | |||
| Proportion of neighborhood with no HS degree × HBCU | −0.17 | 1.16 | 0.88 | |||
| Percent of school’s teachers with a master’s degree (0 to 100) × HBCU | 0.004 | 0.008 | 0.64 | |||
| Household income (in thousands) × HBCU | −0.01 | 0.01 | 0.31 | |||
| Region × HBCU (referent: South) | 0.007 | |||||
| West | 0.19 | 0.28 | 0.50 | |||
| Midwest | 0.64 | 0.32 | 0.05 | |||
| Northeast | 0.32 | 0.56 | 0.57 | |||
| Bachelor’s degree at wave V × HBCU | 0.004 | 0.29 | 0.99 | |||
| Interaction Between Baseline Coefficients and HBCU Status | Years | Age, years | Interaction Coefficienta | SE | P Value | Group Test P Value |
|---|---|---|---|---|---|---|
| Wave III | 2001–2002 | 18–26 | ||||
| Baseline depressive symptoms × HBCU | −0.03 | 0.15 | 0.85 | |||
| Wave IV | 2008 | 24–32 | ||||
| Baseline depressive symptoms × HBCU | −0.27 | 0.11 | 0.01 | |||
| Male × HBCU | 0.15 | 0.43 | 0.72 | |||
| Parent education × HBCU (referent: college) | 0.44 | |||||
| Less than HS | 1.16 | 0.92 | 0.24 | |||
| HS degree | 0.32 | 0.46 | 0.50 | |||
| Proportion of neighborhood below poverty × HBCU | −0.15 | 1.36 | 0.92 | |||
| Proportion of neighborhood Black × HBCU | −0.03 | 0.71 | 0.96 | |||
| Proportion of neighborhood with no HS degree × HBCU | −0.15 | 1.62 | 0.93 | |||
| Percent of school’s teachers with a master’s degree (0–100) × HBCU | 0.01 | 0.01 | 0.49 | |||
| Household income (in thousands) × HBCU | 0.00 | 0.01 | 0.54 | |||
| Region × HBCU (referent: South) | 0.004 | |||||
| West | 0.52 | 0.51 | 0.32 | |||
| Midwest | −0.92 | 0.45 | 0.05 | |||
| Northeast | −1.48 | 0.66 | 0.03 | |||
| Bachelor’s degree at wave 4 × HBCU | −0.69 | 0.47 | 0.15 | |||
| Wave V | 2016–2018 | 32–42 | ||||
| Baseline depressive symptoms × HBCU | 0.14 | 0.11 | 0.20 | |||
| Male × HBCU | −0.54 | 0.31 | 0.08 | |||
| Parent education × HBCU (referent: college) | 0.09 | |||||
| Less than HS | −1.12 | 0.36 | 0.01 | |||
| HS degree | −0.08 | 0.23 | 0.72 | |||
| Proportion of neighborhood below poverty × HBCU | −0.03 | 1.28 | 0.98 | |||
| Proportion of neighborhood Black × HBCU | −0.30 | 0.74 | 0.68 | |||
| Proportion of neighborhood with no HS degree × HBCU | −0.17 | 1.16 | 0.88 | |||
| Percent of school’s teachers with a master’s degree (0 to 100) × HBCU | 0.004 | 0.008 | 0.64 | |||
| Household income (in thousands) × HBCU | −0.01 | 0.01 | 0.31 | |||
| Region × HBCU (referent: South) | 0.007 | |||||
| West | 0.19 | 0.28 | 0.50 | |||
| Midwest | 0.64 | 0.32 | 0.05 | |||
| Northeast | 0.32 | 0.56 | 0.57 | |||
| Bachelor’s degree at wave V × HBCU | 0.004 | 0.29 | 0.99 | |||
Abbreviations: CES-D, Center for Epidemiologic Studies Depression scale; HBCU, Historically Black College or University; HS, high school; SE, standard error.
aAdjusted for baseline covariates including CES-D score, sex, self-rated health, parent education, church attendance, grade point average, proportion of HS teachers who have a master’s degree, region, proportion of census tract in poverty, proportion of census tract without a HS degree, and proportion of census tract Black.
Linear Regression Results for the Interactions Between Baseline Covariates and Attending a Historically Black College or University vs. Attending a Predominantly White Institution on Mean Depressive Symptoms Among Black College Students in the National Longitudinal Study of Adolescent to Adult Health, United States, 1994–2018
| Interaction Between Baseline Coefficients and HBCU Status | Years | Age, years | Interaction Coefficienta | SE | P Value | Group Test P Value |
|---|---|---|---|---|---|---|
| Wave III | 2001–2002 | 18–26 | ||||
| Baseline depressive symptoms × HBCU | −0.03 | 0.15 | 0.85 | |||
| Wave IV | 2008 | 24–32 | ||||
| Baseline depressive symptoms × HBCU | −0.27 | 0.11 | 0.01 | |||
| Male × HBCU | 0.15 | 0.43 | 0.72 | |||
| Parent education × HBCU (referent: college) | 0.44 | |||||
| Less than HS | 1.16 | 0.92 | 0.24 | |||
| HS degree | 0.32 | 0.46 | 0.50 | |||
| Proportion of neighborhood below poverty × HBCU | −0.15 | 1.36 | 0.92 | |||
| Proportion of neighborhood Black × HBCU | −0.03 | 0.71 | 0.96 | |||
| Proportion of neighborhood with no HS degree × HBCU | −0.15 | 1.62 | 0.93 | |||
| Percent of school’s teachers with a master’s degree (0–100) × HBCU | 0.01 | 0.01 | 0.49 | |||
| Household income (in thousands) × HBCU | 0.00 | 0.01 | 0.54 | |||
| Region × HBCU (referent: South) | 0.004 | |||||
| West | 0.52 | 0.51 | 0.32 | |||
| Midwest | −0.92 | 0.45 | 0.05 | |||
| Northeast | −1.48 | 0.66 | 0.03 | |||
| Bachelor’s degree at wave 4 × HBCU | −0.69 | 0.47 | 0.15 | |||
| Wave V | 2016–2018 | 32–42 | ||||
| Baseline depressive symptoms × HBCU | 0.14 | 0.11 | 0.20 | |||
| Male × HBCU | −0.54 | 0.31 | 0.08 | |||
| Parent education × HBCU (referent: college) | 0.09 | |||||
| Less than HS | −1.12 | 0.36 | 0.01 | |||
| HS degree | −0.08 | 0.23 | 0.72 | |||
| Proportion of neighborhood below poverty × HBCU | −0.03 | 1.28 | 0.98 | |||
| Proportion of neighborhood Black × HBCU | −0.30 | 0.74 | 0.68 | |||
| Proportion of neighborhood with no HS degree × HBCU | −0.17 | 1.16 | 0.88 | |||
| Percent of school’s teachers with a master’s degree (0 to 100) × HBCU | 0.004 | 0.008 | 0.64 | |||
| Household income (in thousands) × HBCU | −0.01 | 0.01 | 0.31 | |||
| Region × HBCU (referent: South) | 0.007 | |||||
| West | 0.19 | 0.28 | 0.50 | |||
| Midwest | 0.64 | 0.32 | 0.05 | |||
| Northeast | 0.32 | 0.56 | 0.57 | |||
| Bachelor’s degree at wave V × HBCU | 0.004 | 0.29 | 0.99 | |||
| Interaction Between Baseline Coefficients and HBCU Status | Years | Age, years | Interaction Coefficienta | SE | P Value | Group Test P Value |
|---|---|---|---|---|---|---|
| Wave III | 2001–2002 | 18–26 | ||||
| Baseline depressive symptoms × HBCU | −0.03 | 0.15 | 0.85 | |||
| Wave IV | 2008 | 24–32 | ||||
| Baseline depressive symptoms × HBCU | −0.27 | 0.11 | 0.01 | |||
| Male × HBCU | 0.15 | 0.43 | 0.72 | |||
| Parent education × HBCU (referent: college) | 0.44 | |||||
| Less than HS | 1.16 | 0.92 | 0.24 | |||
| HS degree | 0.32 | 0.46 | 0.50 | |||
| Proportion of neighborhood below poverty × HBCU | −0.15 | 1.36 | 0.92 | |||
| Proportion of neighborhood Black × HBCU | −0.03 | 0.71 | 0.96 | |||
| Proportion of neighborhood with no HS degree × HBCU | −0.15 | 1.62 | 0.93 | |||
| Percent of school’s teachers with a master’s degree (0–100) × HBCU | 0.01 | 0.01 | 0.49 | |||
| Household income (in thousands) × HBCU | 0.00 | 0.01 | 0.54 | |||
| Region × HBCU (referent: South) | 0.004 | |||||
| West | 0.52 | 0.51 | 0.32 | |||
| Midwest | −0.92 | 0.45 | 0.05 | |||
| Northeast | −1.48 | 0.66 | 0.03 | |||
| Bachelor’s degree at wave 4 × HBCU | −0.69 | 0.47 | 0.15 | |||
| Wave V | 2016–2018 | 32–42 | ||||
| Baseline depressive symptoms × HBCU | 0.14 | 0.11 | 0.20 | |||
| Male × HBCU | −0.54 | 0.31 | 0.08 | |||
| Parent education × HBCU (referent: college) | 0.09 | |||||
| Less than HS | −1.12 | 0.36 | 0.01 | |||
| HS degree | −0.08 | 0.23 | 0.72 | |||
| Proportion of neighborhood below poverty × HBCU | −0.03 | 1.28 | 0.98 | |||
| Proportion of neighborhood Black × HBCU | −0.30 | 0.74 | 0.68 | |||
| Proportion of neighborhood with no HS degree × HBCU | −0.17 | 1.16 | 0.88 | |||
| Percent of school’s teachers with a master’s degree (0 to 100) × HBCU | 0.004 | 0.008 | 0.64 | |||
| Household income (in thousands) × HBCU | −0.01 | 0.01 | 0.31 | |||
| Region × HBCU (referent: South) | 0.007 | |||||
| West | 0.19 | 0.28 | 0.50 | |||
| Midwest | 0.64 | 0.32 | 0.05 | |||
| Northeast | 0.32 | 0.56 | 0.57 | |||
| Bachelor’s degree at wave V × HBCU | 0.004 | 0.29 | 0.99 | |||
Abbreviations: CES-D, Center for Epidemiologic Studies Depression scale; HBCU, Historically Black College or University; HS, high school; SE, standard error.
aAdjusted for baseline covariates including CES-D score, sex, self-rated health, parent education, church attendance, grade point average, proportion of HS teachers who have a master’s degree, region, proportion of census tract in poverty, proportion of census tract without a HS degree, and proportion of census tract Black.
None of the other effect modifiers we tested had significant interactions at P < 0.05, including sex, parent education, household income, proportion of teachers with a master’s degree, or proportion of neighborhood in poverty, Black, or with less than high school degree (Table 2).
DISCUSSION
We found that for Black students who had higher depressive symptoms at baseline, attending an HBCU was more beneficial than attending a PWI in terms of depressive symptoms 7 years later. This association appears to be stronger for students who attended a high school that was outside of the South, although the 3-way interaction was not statistically significant. For someone with a baseline depressive symptom score of 4, HBCU enrollment led to 1.5 less points on the depressive symptoms scale compared with PWI enrollment. It is worth noting that for Black students who attend HBCUs, frequency of depressive symptoms 7 years later is low even for those who had high depressive symptoms in high school. The other potential effect modifiers we tested were not significant, although an HBCU participant sample of n = 113 may have been underpowered to detect those associations. Future research should continue to consider intersectional identities and how, within Black students, there is still important heterogeneity and differential positioning to power. These findings provide evidence that the HBCU environment can have a positive impact on mental health for Black students in the long term. As the conceptual framework (Figure 1) shows, there are many conceivable mechanisms between HBCU attendance and depressive symptoms. HBCUs may provide a less racist environment that buffers against short-term depressive symptoms for those who are vulnerable. Students of color who are discriminated against by peers more in the first year of college are more likely to be depressed in their fourth year of college (45). In addition, Black students at HBCUs report better overall satisfaction with their education (27), less interpersonal stress, and more social support (39). Add Health did not measure discrimination in wave III when HBCU status was attained, but future research can delve further into mechanisms between HBCU enrollment and long-term health outcomes.
These results reinforce that racism is a complex social phenomenon and there is much we do not understand about how it operates. Still, there are many reasons to address racial educational inequity, and several implications of our study for policy.
Policy recommendations
Allocate public funds equitably to HBCUs.
Our analyses showed that HBCUs positively affect the short-term mental health of students at high risk for depressive symptoms and produce more graduates, despite enrolling students with lower grade point averages from less advantaged schools and neighborhoods (Web Table 1). Remarkably, HBCUs produce these positive results for Black students even though they have been consistently underfunded (46). For instance, in April 2021 Maryland settled a federal lawsuit that awarded $577 million to their 4 HBCUs for underfunding them compared with PWIs (47). Even as HBCUs enroll student populations that are more diverse, they will continue to be well-positioned to produce positive results such as improved mental health for Black students.
About half of HBCUs are public institutions (16) and deserve public support proportional to the work they do to support Black students. One way to direct funds might be increased recruiting efforts outside of the South. While most HBCU students come from the South, we saw a trend toward a larger mental health benefit for students from other regions. HBCUs might also benefit from more funding for financial aid because HBCUs enroll students with lower socioeconomic status on average (20). It is also important to acknowledge that—due to the enormous Black-White wealth gap in the United States (48)—the endowments many PWIs receive from alumni cannot be equally achieved by HBCU alumni.
Institutions of higher education should fund and evaluate programs to support Black students.
PWIs may learn from HBCUs about structurally supporting Black students. For example, PWIs could incorporate aspects of HBCUs such as increased Black faculty mentors, safe within-race spaces for Black students, and a curriculum that decenters the dominant White perspective. Interventions on these different aspects of the college environment require dedicated funding and institutional buy-in from the top, as well as evaluations to determine their effectiveness. The National Association of Diversity Officers in Higher Education’s antiracism framework includes 10 priority areas, as well as suggestions for specific areas to focus efforts (49).
Colleges and universities must take steps to improve mental health services on campuses for all students. A nationwide study indicated that over 30% of undergraduates screened positive for depression, anxiety, thoughts of suicide, or self-injury (50). Recently increased demand for mental health services among college students has led to shorter counseling sessions, long waitlists, and more referrals to off-campus providers (51). Institutions must address this lack of access to mental health services on campuses.
Conclusions
This study showed that for Black students with more depressive symptoms in high school, whether they attended an HBCU or a PWI was associated with their depressive symptoms 7 years later. Early adulthood is an important time in the life course for shaping mental health, and educational systems are important contexts for transmitting structural racism, or promoting antiracism, which may have long term effects on mental health. Understanding and unpacking the impact higher education may have on the mental health of Black Americans is important for addressing pervasive racial disparities in population health.
ACKNOWLEDGMENTS
Author affiliations: Division of Epidemiology and Community Health, School of Public Health, University of Minnesota, Minneapolis, Minnesota, United States (Naomi Harada Thyden, Jaime Slaughter-Acey, Rachel Widome, Theresa L. Osypuk); Minnesota Population Center, University of Minnesota, Minneapolis, Minnesota, United States (Naomi Harada Thyden, John Robert Warren, Theresa L. Osypuk); Community Health Sciences, Center of Excellence in Maternal and Child Health, School of Public Health, University of Illinois, Chicago, Chicago, Illinois, United States (Naomi Harada Thyden); Division of Health Policy and Management, University of Minnesota, Minneapolis, Minnesota, United States (Cydney McGuire); Paul H. O’Neill School of Public and Environmental Affairs, Indiana University, Bloomington, Indiana, United States (Cydney McGuire); and Department of Sociology, University of Minnesota, Minneapolis, Minnesota, United States (John Robert Warren).
We appreciate support provided by the Eunice Kennedy Shriver National Institute for Child Health and Human Development of the US National Institutes for Health, for supporting N.H.T.’s training (grant T32HD095134, MPIs: J.R.W., T.L.O.), and the University of Minnesota’s Minnesota Population Center (grant P2CHD041023) and the National Institute on Aging for supporting the University of Minnesota’s Life Course Center (grant P30AG066613). This study was supported by a Hawley Student Research Award from the University of Minnesota, Division of Epidemiology and Community Health (PI: N.H.T.). This research uses data from Add Health, funded by grant P01 HD31921 (Harris) from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD), with cooperative funding from 23 other federal agencies and foundations. Add Health is currently directed by Robert A. Hummer and funded by the National Institute on Aging cooperative agreements U01 AG071448 (Hummer) and U01AG071450 (Aiello and Hummer) at the University of North Carolina at Chapel Hill. Add Health was designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris at the University of North Carolina at Chapel Hill.
Data are protected by a restricted access data agreement with Add Health.
Data from this study were presented at the 2021 Interdisciplinary Association for Population Health Science (online), October 19–21, 2021.
Conflict of interest: none declared.
