Contextualizing Educational Disparities in Health: Variations by Race/ethnicity, Nativity, and County-level Characteristics

Abstract

Educational disparities in health are well-documented; yet, the education-health relationship is inconsistent across racial/ethnic and nativity groups. These inconsistencies may arise from characteristics of the early life environments where individuals attain their education. We evaluate this possibility by investigating: 1) whether educational disparities in cardiometabolic risk vary by race/ethnicity and nativity among White, Black, and Hispanic young adults; 2) the extent to which racial/ethnic/nativity differences in the education-health relationship are contingent on economic, policy, and social characteristics of counties of early life residence; and 3) the county characteristics associated with the best heath at higher levels of education for each racial/ethnic/nativity group. Using data from the National Longitudinal Study of Adolescent to Adult Health, results suggest Black young adults who achieve high levels of education exhibit worse health across a majority of contexts relative to their White and Hispanic counterparts. Additionally, we observe more favorable health at higher levels of education across almost all contexts for Whites. For all other racial/ethnic/nativity groups, the relationship between education and health depends on the characteristics of early life counties. Findings highlight place-based factors that may contribute to the development of racial/ethnic and nativity differences in the education-health relationship among US young adults.

Introduction

Educational attainment is a significant determinant of health (Conti et al. 2010; Hummer and Lariscy 2011; Montez et al. 2019; Zajacova and Lawrence 2018). In recent years, educational disparities in health in the United States (US) have widened due to improvements in health among the most educated and declining life expectancy among the least educated (Case and Deaton 2015; Hayward, Hummer, and Sasson 2015; Zajacova and Lawrence 2018). Given the robust relationship between education and health, education is often conceptualized as the “great equalizer”—a resource that may help all individuals attain a comparable quality of life (Downey et al. 2004; Torche 2011). Yet, accumulating evidence documents inconsistencies in this relationship, particularly by race/ethnicity and nativity. For example, educational attainment has a muted or non-significant association with health among minority groups, especially Black Americans (Assari et al. 2017; Beltrán-Sánchez et al. 2016; Boen 2016; Fuller-Rowell et al. 2015; Kimbro et al. 2008; Williams, Priest, and Anderson 2016). Moreover, evidence that several Hispanic and immigrant subgroups have comparable or better health outcomes than their higher-educated White and native-born counterparts is inconsistent with the idea that more highly educated groups uniformly have better health than less-educated groups (Lariscy, Hummer, and Hayward 2015; Markides and Rote 2015; Ruiz et al. 2016).

Explanations for these heterogenous relationships between education and health by race/ethnicity and nativity generally focus on individual-level factors, such as increased exposure to interpersonal discrimination and psychosocial resources developed in response to the stressors of navigating higher education environments (Destin and Debrosse 2017; Hudson et al. 2013; James et al. 1987; Miller et al. 2016). Selection effects may also explain better health outcomes among foreign-born Hispanic populations relative to their native-born and White counterparts (e.g., Kennedy et al. 2015; Riosmena et al. 2013). Recent work, however, highlights the importance of contextual factors for the education-health relationship by documenting state-level variation in educational disparities in adult health and mortality (Montez and Berkman 2014; Montez, Zajacova, and Hayward 2017; Montez et al. 2019). While important, this line of work does not investigate whether state-level contexts help to explain racial/ethnic differences in the association between educational attainment and health. Also missing from prior research is the consideration of contextual factors across smaller geographic units, such as counties of residence in childhood. Examining early life contexts is critical, as these conditions shape developmental trajectories and patterns of (dis)advantage, and directly or indirectly influence health across adulthood (Friedman et al. 2015; Montez and Hayward 2011, 2014; Warner and Hayward 2006; Zajacova, Walsemann, and Dowd 2015).

This study advances our understanding of the relationship between educational attainment and health by investigating whether higher education is associated with better young adult health across different early life county environments. We ask:

1. Do educational disparities in cardiometabolic risk vary by race/ethnicity and nativity among Black, Hispanic, and White young adults?

2. To what extent are observed differences in the education-health relationship by race/ethnicity and nativity dependent on the economic, policy, and social (dis)advantages of the county in which young adults lived as children/adolescents?

3. In which early-life contexts do racial/ethnic and nativity groups experience the best health at higher levels of education?

Examining the association between education and health across different early life county contexts provides new evidence of factors that may modify the association between higher educational attainment and better health in young adulthood, a life stage that is particularly dynamic and unhealthy for more recent cohorts (Hargrove et al. 2020; Harris 2010; Masters et al. 2018). As such, this study provides unique insight into potential place-based mechanisms that generate differences in the education-health relationship by race/ethnicity and nativity across early adulthood.

Background

Educational Disparities in Health by Race/ethnicity and Nativity

Individuals with higher levels of education experience healthier and longer lives than those with lower levels of education (Sasson and Hayward 2019; Zajacova and Lawrence 2018). This relationship is often attributed to the forms of economic, human, and social capital that generally accompany educational attainment and can be translated into health promoting resources such as income, knowledge, and healthy behaviors (Hayward et al. 2015; Phelan and Link 2015; Zimmerman et al. 2015). Yet, accumulating evidence documents inconsistences in the education-health relationship, especially among racial/ethnic minorities and those not born in the US. For example, the association between education and health tends to be non-significant or weaker for Black adults and to a certain extent, Hispanic adults, relative to Whites across a variety of physical health outcomes, including self-rated health, birth outcomes, body-mass index, disability, and inflammation (Assari et al. 2017; Boen 2016; Colen et al. 2006; Farmer et al. 2020; Fuller-Rowell et al. 2015; Kimbro et al. 2008). Furthermore, racial/ethnic disparities in health exist at every level of educational attainment (Braveman et al. 2010; Williams et al. 2016), and are often widest at the highest levels of education (e.g., Cummings and Jackson 2008; Hargrove 2018). Additionally, studies document that Hispanic and immigrant groups tend to exhibit more favorable health and mortality outcomes than their White and native-born counterparts despite lower average educational attainment (Goldman et al. 2006; Hummer et al. 2007; Markides and Rote 2015; Ruiz et al. 2016).

Collectively, existing evidence suggests a weaker or nonexistent relationship between educational attainment and health for several population groups in the US. These findings, coupled with the fact that the studies that initially established and validated this relationship either exclusively utilized White samples or pooled all racial/ethnic groups, underscore arguments that educational disparities among non-White groups remain an empirical question (Pearson 2008). This is particularly true for young adulthood, as prior work on the education-health relationship tends to focus on middle and older adulthood. While important, results from studies of middle-aged or older adults may be subject to selective mortality bias, as the least educated may have died by middle adulthood. Furthermore, focusing on middle-aged adults does not provide information about when educational disparities in health initially emerge. Better insight into health disparities in young adulthood provide opportunities for interventions that can maximize the benefit of educational attainment across the life course.

Proposed Explanations

A common explanation for the muted or non-significant association between education and health among Black adults is that educational attainment can increase exposure to interpersonal discrimination or other unique race-related stressors, especially as individuals gain access to higher socioeconomic contexts across education, work, and housing domains (Hardaway and McLoyd 2009; Hudson et al. 2012; Hudson et al. 2016; Von Robertson et al. 2016; Wingfield 2010; Wingfield and Wingfield 2014). Moreover, given historical and contemporary barriers to higher education faced by Black and Hispanic populations in the US, those who are able to achieve higher levels of education generally do so at a cost to their physical well-being (Brody et al. 2013; Feagin and McKinney 2005; Gaydosh et al. 2018). These costs stem from the need to sustain more physical, cognitive, emotional, and psychological effort than put forth by Whites (Cole and Omari 2003; Colen et al. 2018; Hudson et al. 2016). Prior work also suggests that educational gradients in health among foreign-born Hispanic populations are shallower compared to their White and US-born Hispanic counterparts. These patterns are often a result of more favorable health outcomes among foreign-born Hispanic adults with low educational attainment relative to individuals of other racial/ethnic groups with similarly low levels of education (e.g., Goldman et al. 2006; Kimbro et al. 2007). Sociocultural factors (e.g., engagement in healthy behaviors, higher levels of social support) and health selection effects in the case of immigrant groups help to explain better health among Hispanic groups with lower levels of education (Abraido-Lanza et al. 2005; Antecol and Bedard 2006; Crimmins et al. 2007; Lariscy et al. 2015).

Finally, inconsistences in the relationship between education and health across groups may stem from the non-equivalence of socioeconomic indicators across race/ethnicity (Williams and Collins 1995). Traditional educational attainment measures do not account for the structural features or characteristics of the communities in which education is achieved. For example, self-reports of highest level of education do not capture differences in how surrounding environments may facilitate or hinder residents’ ability to attain high levels of education and translate education into health promoting resources. Given that place-based resources are often stratified based on the racial/ethnic composition of communities, the contexts wherein racial/ethnic minorities and immigrants begin their educational trajectories likely differ from that of Whites. The non-equivalence of the meanings of education measures, in addition to arguments of sustained effort put forth by minority populations to achieve higher levels of education, point to the importance of studying how different contexts modify the education-health relationship within racial/ethnic and nativity groups. Examining both between and within group comparisons provides unique types of information, helping to better understand the education-health relationship for different population subgroups.

Importance of the County Context

While a long tradition of research documents the importance of place in generating social inequalities in health (Montez, Hayward, and Wolf 2017; O’Brien et al. 2020; Sewell 2016; Woolf and Braveman 2011), scholars have recently highlighted the need for research that evaluates how contextual factors generate and sustain educational disparities in health (Zajacova and Lawrence 2018). Indeed, the contexts in which educational attainment occurs can facilitate or impede individuals’ abilities to convert higher levels of education into health protective resources. Several studies have documented differences in the education-health relationship across states in outcomes such as disability and mortality (Montez et al. 2017; Montez et al. 2019). Less work has considered the role of other geographic entities in shaping educational disparities in health in young adulthood.

An important geographic entity to consider is the county. Counties are tasked with the provision and administration of state services and regulations, making them primarily responsible for meeting the needs and demands of their residents. Services targeting healthcare, affordable housing, air pollution, education, and job opportunities may therefore vary based on the political processes, available resources, and zoning laws in a given county (Marando and Reeve 1991; McLaughlin and Stokes 2002). Thus, counties play an important role in the acquisition and distribution of health-promotive resources. Moreover, given that racial/ethnic minorities disproportionately occupy low socioeconomic strata in the US, they are also more likely to interact with and rely on county and local government structures for social services to meet their basic needs such as housing and healthcare (Herd and Moynihan 2019; Sered and Norton-Hawk 2014). County characteristics may therefore be particularly consequential for the magnitude of the relationship between educational attainment and health among racial/ethnic minorities. While prior research links county-level characteristics to health and mortality (Chambers et al. 2018; Dwyer-Lindgren et al. 2017; Jia et al. 2009; O’Brien et al. 2020), the extent to which educational attainment is associated with better health across different county environments remains unclear.

The Role of County Context across the Life Course

The present study focuses on the role of three specific aspects of early life counties: economic, policy, and social contexts. These characteristics are important structural determinants of health, as they are shaped by systems of inequality such as structural racism, and they reproduce patterns of stratification along racial/ethnic lines (Solar and Irwin 2010). The interplay between county contexts and stratification systems not only influences the distribution of health-related risks and resources, but also the nature of educational attainment processes. For example, the extent to which schooling is characterized by obstacles versus opportunities may be associated with health returns in young adulthood across racial/ethnic and nativity groups. Further, racial/ethnic minorities and immigrants may be particularly sensitive to surrounding economic, political, and social contexts in early life, as historical and contemporary racialized processes (e.g., redlining; voter suppression) shape these place-based characteristics and produce additional disadvantages for minorities, including lack of employment opportunities, exposure to employer discrimination, exposure to air pollution, and decreased funding for schools (Pager and Shepard 2008; Rothstein 2017; Tessum et al. 2021; Viruell-Fuentes, Miranda, and Abdulrahim 2012).

Prospects for future economic opportunity in counties are relevant for health (O’Brien et al. 2020; Venkatarmani et al. 2016). Counties that facilitate upward mobility or make socioeconomic attainment easier—both objectively and perceptually—might provide greater access to educational opportunities, and mitigate the need for, or potentially harmful consequences of, sustained/high-effort coping on health. For example, counties with more economic opportunities may attenuate the need for high-effort coping, resulting in similar relationships between education and health across racial/ethnic/nativity groups. Similarly, in counties where employment prospects are high, the labor market rewards of educational attainment are more salient. Moreover, counties in which there is a more equitable distribution of income may lead to greater access to health protective factors for all residents regardless of educational attainment.

Policy characteristics, reflected in the amount that county governments invest in residents’ education or health, may directly shape opportunities for educational advancement as well as the ability to translate educational attainment into health promoting resources as adults (e.g., Woolf and Braveman 2011). Similar to future opportunity prospects, actual or perceived investment in one’s community may reduce the potentially negative aspects of educational attainment processes, especially for racial/ethnic minorities. Furthermore, policies and economic opportunities in a given county might shape the perceived importance of education for health and well-being (e.g., Montez et al. 2019), therein influencing selection into higher education.

Last, the social aspects of a county are indicated, in part, by the educational attainment of county residents. The degree to which residents are highly educated may not only signal surrounding job opportunities or economic conditions, but also influence the meaning of education for health. The level of education of county residents might influence expectations or aspirations of educational attainment that may or may not be fulfilled. Whether such expectations or aspirations are achieved could lead to unique types of stressors, such as goal-striving stress (e.g., Mouzon et al. 2019) or identity threat (e.g., Geronimus et al. 2016; Pearson 2008), which could weaken the education-health relationship.

A life course perspective helps to describe how and why county contexts experienced in early life may have persistent consequences on the education-health relationship across adulthood. An important pathway is through the accumulation of risks and resources (Ferraro, Schafer, and Wilkinson 2016). Early life environments set trajectories of strain or stress experienced across one’s educational career, which can take a cumulative physiological toll. For example, children living in socioeconomically disadvantaged neighborhoods, particularly children of color, may experience structural barriers to accessing education, limiting their ability to attain high levels of education or secure the income and employment benefits traditionally associated with higher education. Achieving high levels of education in the face of such barriers may come at a cost to energy reserves and bodily systems (Brody et al. 2013; James et al. 1983).

The Present Study

The present study investigates whether and how childhood contexts modify educational disparities in young adult health across and within racial/ethnic and nativity groups. The theories and research outlined above lead to the following hypotheses corresponding with our three research questions:

Hypothesis #1: While higher levels of education will be associated with better health for all racial/ethnic and nativity groups, the education-health relationship will be stronger among White and foreign-born Hispanic young adults relative to Black and US-born Hispanic young adults.

Hypothesis #2: The smallest differences in the education-health relationship by race/ethnicity/nativity will occur in counties characterized by economic, policy, and social advantages.

Hypothesis #3: Across all early life county environments, higher educational attainment will be associated with better health among White young adults. Higher educational attainment will be associated with better health for Black and Hispanic young adults relative to their less educated, same race/ethnic/nativity counterparts only when they are exposed to advantaged county environments in early life.

This study advances prior work in several ways. First, this study examines educational disparities in young adulthood, a critical stage in the life course wherein health trajectories have already begun, and are continuing, to diverge (Hargrove 2018; Harris 2010). This study will provide insight into whether and how the characteristics of places that shape educational attainment in early life are associated with health disparities that persist across the life course. Furthermore, younger cohorts have worse cardiovascular health than older cohorts had at the same age (e.g., Masters et al. 2018; Preston et al. 2018). Thus, this study focuses on an important population whose health may be compromised in ways not experienced in previous cohorts.

Second, we utilize biological indicators of health. While life course theory motivates the study of health as an ongoing process that unfolds as individuals age, it is challenging to measure health in young individuals because chronic diseases are generally not yet manifest, and mortality is rare. Biomarkers of objective physical health risk present an opportunity to capture underlying variation in health in young adult populations (Harris and Schorpp 2018; Harris and McDade 2018). Finally, by examining how educational disparities vary by race/ethnicity, nativity, and county characteristics, results provide insight into factors that may enhance or hinder the association between higher education and better health.

Data and Methods

Sample

We use data from Waves I and IV of the National Longitudinal Study of Adolescent to Adult Health (Add Health), a nationally representative study of US adolescents in grades 7-12 at baseline in 1994-1995 (Wave I; N=20,745). Follow-up data were collected in 1996 (Wave II; N=14,738), 2001-2002 (Wave III; N=15,170), and 2008-2009 (Wave IV; N=15,701). Add Health used a multistage, stratified, school-based, cluster sampling design (Harris et al. 2019). Our analytic sample consists of individuals who participated in Waves I and IV; identify as non-Hispanic White, non-Hispanic Black, or Hispanic; and have valid sampling weights and Wave IV biomarker data. Further exclusions due to missingness on Wave I county measures (98 cases) resulted in a final analytic sample of 13,317.

Measures

Outcome.

Cardiometabolic risk (CMR) is measured at Wave IV, when respondents were age 24-32, and is a summary measure comprised of seven physiological indicators representing biological functioning of the metabolic, inflammatory, and cardiovascular systems: waist circumference, triglycerides, HDL cholesterol, LDL cholesterol, glycosylated hemoglobin, blood pressure, and C-reactive protein (CRP). Each biomarker is dichotomized based on predefined disease risk cut points (Supplemental Table 1) and summed to create a continuous risk score, ranging from 0-7 (Harris et al. 2017).

Predictors of Interest.

Our predictors of interest include race/ethnicity (Wave I), nativity (Wave I), educational attainment in adulthood (Wave IV), and characteristics of the county respondents lived during childhood/adolescence (Wave I). Self-identified race/ethnicity is measured by respondents’ choice of racial/ethnic category: non-Hispanic White; non-Hispanic Black; and Hispanic. Nativity is assessed as whether the respondent was born outside of the US or not. Given that few Black and White respondents were born outside of the US in the Add Health cohort, we only assess differences by nativity among Hispanics. Thus, four racial/ethnic/nativity groups are considered: non-Hispanic US-born Whites (hereafter referred to as White); non-Hispanic US-born Blacks (hereafter referred to as Black); US-born Hispanics; and foreign-born Hispanics. Educational attainment is measured as a dummy variable: 0=less than four-year college; 1=four-year college education or more.

County-level characteristics are taken from the county in which Add Health respondents lived during Wave I (Belsky et al. 2019). Economic features are captured by indicators of future prospects for economic opportunity: absolute mobility, unemployment, and the Gini coefficient. Absolute mobility and the Gini coefficient come from the Equality of Opportunity Project (EOP) (Chetty and Hendren 2018; Chetty et al. 2014). Investigators compiled data from de-identified federal income tax records, providing information on the incomes of more than 40 million children and their parents between 1996 and 2012. Importantly, the “children” in these data are part of the 1980-1982 birth cohort, which overlaps with the Add Health cohort, as most Add Health respondents were born between 1977 and 1982.

Within the EOP data, absolute mobility was calculated as the mean rank (in the national child income distribution) of children whose parents were at the 25th percentile of the national parent income distribution. Higher values of this measure correspond to greater economic opportunity for mobility in one’s county of residence. The Gini coefficient indicates the amount of parental income inequality within counties in the US. Higher values indicate more income inequality. Unemployment is assessed as the proportion of county residents who are unemployed.

The policy features of counties are captured by county per capita direct expenditure on education and health and hospitals. Social features are assessed as the proportion of county residents (aged 25 or older) with low educational attainment (no high school diploma or equivalency). Measures of unemployment, county government expenditures, and educational attainment of residents came from the 1990 Census (Billy et al. 1998).

Controls.

All models control for individual-, family-, and school-level characteristics to address potential confounding of selection into neighborhoods and of the education-health relationship. Individual characteristics include sex (0=male; 1=female); age at Wave IV; and childhood self-rated health at Wave I (range: 1-5, poor-excellent). Family socioeconomic characteristics are captured with measures of parental education, ranging from less than eighth grade (0) to professional training beyond a four-year college or university (9), and household receipt of government assistance (1=yes) in Wave I. We use information from both the parent and in-home questionnaires to construct the family characteristics measures to retain cases.

School-level factors include the percentage of teachers with master’s degrees (range: 0-95%) and an index of school disadvantage comprised of 5 indicators that were aggregated to the school level: the proportion of households receiving welfare; proportion of parents with less than a high school education; proportion of unemployed parents; proportion of single-parent households, and the proportion of non-White students. Items were dichotomized based on respondent inclusion in the top quartile and summed to create a school disadvantage index.

Analytic Strategy

We estimate a series of Poisson regression models to address the research questions. Eq. 1 is the basic form of each model. It estimates the log of the expected count of CMR for individual i as a function of a binary measure of education (CLPS for college plus) and covariates contained in the b₂ vector.

$$\log ({\mu }_{\mathrm{i}})={\mathrm{b}}_0+{\mathrm{b}}_1\text{CLPS}+{\mathrm{b}}_2\text{covariates},$$

where μ_i = E(y_i) is the expected count value of CMR for individual i.

To address our first research question, we regress CMR on interactions between race/ethnicity/nativity and educational attainment. This model evaluates the extent to which the education-health relationship varies by race/ethnicity and nativity. All groups are compared to White adults.

To address our second research question, we stratify the model described above by advantaged county contexts (e.g., high mobility/more investments/low proportions of low-educated residents) and disadvantaged contexts (e.g., low mobility/fewer investments/high proportions of low-educated residents). To determine advantaged vs. disadvantaged county contexts, we dichotomize county measures at race/ethnicity-specific 75^th percentiles. Given the high degree of racial residential segregation that exists in the US (Rothstein 2017), the distributions of county characteristics vary tremendously by race/ethnicity. Consequently, there is not much overlap in the distributions between racial/ethnic groups, especially between Black and White young adults. These patterns may also differentiate definitions of “highly mobile” or “highly educated” counties across groups. Race/ethnicity-specific cut points help to facilitate an investigation of more meaningful categories of early life (dis)advantages for all groups.

Stratifying the models by place allows us to assess whether racial/ethnic/nativity differences in the association between education and health depend on specific characteristics of early life counties. Indeed, place is a fundamental cause of health inequality, shaping opportunities for educational attainment, socioeconomic mobility, and health (Williams et al. 2019). As such, the mechanisms that link education to health in more advantaged counties likely vary from the mechanisms that link education to health in less advantaged counties. White adults who lived in relatively similar county contexts in early life serve as the reference group.

To address our third research question, we estimate the association between educational attainment and CMR within race/ethnicity/nativity groups across different county contexts. We stratify models by race/ethnicity/nativity and county characteristics, and examine educational disparities in CMR. Given stark racialized experiences in the US, it is unlikely that the covariates in the models described above have equivalent consequences for all groups. That is, race/ethnicity (and likely, nativity) exhibit differential associations with multiple variables in our analysis, making it difficult to base conclusions solely on between-group analyses. Stratifying the models by race/ethnicity and nativity in addition to county contexts allows us to identify the types of contexts in which education may have a particularly strong (or weak) association with health for each sociodemographic group. Respondents with less than a college degree who lived in given county context during childhood/adolescence are compared to their higher educated counterparts who lived in counties with similar characteristics.

All analyses use svy commands in Stata 16.1, which adjust the standard errors for clustering at the school level. Svy commands also incorporate sampling weights to account for unequal probability of selection and other design effects. To retain as many cases as possible, we use multiple imputation by chained equations (MICE) on the analytic sample described above. MICE is a flexible imputation technique that produces a series of regression models. These regression models impute missing values conditional upon the other variables in the dataset (Azur et al. 2011). Ten imputed datasets for the total analytic sample, as well as for each racial/ethnic/nativity group, were created to conduct the analyses.

We also conducted a series of robustness checks, including: 1) controlling for the length of time (in years) respondents lived in the residence reported at Wave I; and 2) examining relationships among three categories of educational attainment and health: high school or less (ref. group); more than high school/some college; and college or more. Results from these analyses (available upon request) indicate similar findings and substantive conclusions to those reported here. Moreover, ancillary analyses estimated models using absolute measures of county (dis)advantage rather than race-specific levels, which produced similar results to those presented here. While these ancillary analyses rely on a subset of racial/ethnic minorities—as racialized systems of inequality limit the ability of minorities to access the same spaces as Whites—the results indicate that even the few Black and, to an extent, US-born Hispanic adults who were exposed to the same county conditions and levels of economic, policy, and social (dis)advantages as Whites do not experience comparable levels of CMR at higher levels of education. Given the small proportion of marginalized groups living in the same places as Whites, we employ and present models using race-specific levels of county (dis)advantage to better capture the reality and consequences of racial segregation in the US.

Results

Table 1 summarizes the descriptive characteristics of the analytic sample. Relative to Whites, Black and US-born Hispanic young adults have higher CMR scores, are less likely to have a four-year college degree or more, and tended to live in counties characterized by higher levels of income inequality during childhood and adolescence. Furthermore, Black young adults generally lived in more disadvantaged counties during childhood/adolescence compared to their White counterparts, as indicated by fewer opportunities for upward social mobility, higher proportion of residents who were unemployed or had less than a high school diploma, and less government spending on education. Aside from differences in income inequality, US-born and foreign-born Hispanics tended to live in similar counties in Wave I as Whites. Foreign-born Hispanics had similar levels of CMR as their White counterparts, yet were less likely to have completed a college degree or more.

Table 1.

Weighted Means and Proportions of Study Variables by Race/ethnicity and Nativity

	US-Born Whites	US-Born Blacks	US-Born Hispanics	Foreign-Born Hispanics	Range
Outcome
CMR	1.78 (.03)	2.10* (.04)	2.06* (.06)	1.78 (.07)	0-6
Education
College Degree or More	.33 (.02)	.22* (.03)	.18* (.02)	.22* (.04)	0,1
County Characteristics
Absolute mobility	41.96 (.47)	38.13* (.40)	42.40 (.31)	42.34 (.36)	33 – 992
Prop. residents unemployed	.07 (.00)	.08* (.00)	.07 (.00)	.07 (.00)	.03 - .15
Gini coefficient	.41 (.01)	.51* (.01)	.50* (.02)	.56* (.02)	.22 – 2
Per capita expenditure on education (in dollars)	683.21 (15.19)	634.92* (18.52)	701.16 (18.11)	685.29 (18.75)	2.54 – 2281.68
Per capita expenditure on health (in dollars)	139.34 (15.96)	121.18 (19.43)	173.90 (18.39)	160.79 (21.32)	0 – 839.84
Prop. residents with less than HS	.25 (.01)	.29* (.01)	.23 (.01)	.26 (.02)	.05 - .61
Controls
Female (Wave I)	.49 (.01)	.51 (.01)	.51 (.02)	.49 (.04)	0,1
Age (Wave IV)	28.28 (12)	28.30 (.22)	28.33 (.19)	28.98* (.29)	24 – 34
Self-rated health (Wave I)	2.10 (.02)	2.11 (.02)	2.21* (.03)	2.12 (.06)	1 – 4
Parental education (Wave I)	6.07 (.10)	5.36* (.19)	4..51* (.018)	3.55* (.29)	0 – 9
Household receipt of government assistance (Wave I)	.07 (.01)	.22* (.02)	.19* (.02)	.17* (.02)	0,1
Percentage of teachers with MA or higher (Wave I)	51.58% (2.53)	48.61% (3.58)	49.21% (3.85)	49.11% (5.33)	0 – 95
School disadvantage (Wave I)	.66 (.10)	2.09* (.23)	1.86* (.31)	2.76* (.50)	0 – 5
N	7,920	3,060	1,789	547

^*p<.05 difference from Whites

Note: Range represents documented range among entire analytic sample; standard errors in parentheses

Do educational disparities in cardiometabolic risk (CMR) vary by race/ethnicity and nativity among White, Black, and Hispanic young adults?

Figure 1 displays results from models that estimate an interaction between race/ethnicity/nativity and educational attainment (Appendix Table 1, Column 1). Predicted CMR scores are shown on the y-axis and each racial/ethnic/nativity group is displayed on the x-axis. Results indicate that higher educational attainment is similarly related to lower CMR for US-born Hispanic, foreign-born Hispanic, and White young adults. Higher levels of education, however, are not associated with lower levels of CMR for Black young adults relative to their less educated, Black and high-educated Hispanic and White counterparts. These results provide partial support for Hypothesis 1.

Figure 1.

Predicted CMR scores by education groups and race/ethnicity/nativity. Results from Poisson regression models (Appendix Table 2, Column 1). All models control for age, sex, childhood self-rated health, parental education, household receipt of welfare, percentage of teachers with MA or higher, and school disadvantage.

To what extent are racial/ethnic/nativity differences in the education-CMR relationship contingent upon the economic, policy, and social contexts of childhood counties?

Figure 2 summarizes results from models that are stratified by county characteristics (Appendix Table 1, Columns 2-7). The left column of the figure displays advantaged county contexts and the right column displays disadvantaged county contexts. Each row represents a given county characteristic. Across the 12 different contexts characterized by high or low levels of economic, policy, and social advantages, there are 9 contexts in which the education-health relationship significantly differs for Black adults compared to Whites: low absolute mobility; high and low unemployment; high and low income inequality; low per capita spending on education and health; and high and low proportions of residents with low education. In these contexts, attaining a college degree or more is associated with smaller reductions in CMR for Black young adults relative to their White counterparts with similar levels of education. In counties characterized by high absolute mobility, high per capita spending on education, and high per capita spending on health and hospitals, the association between higher educational attainment and lower CMR is comparable among Black and White young adults (Figure 2, Left Column, Rows 1, 4, 5). Conversely, results suggest the education-CMR relationship is similar for US-born and foreign-born Hispanic young adults relative to Whites regardless of the characteristics of childhood counties. There is one exception for each group. The association between education and CMR is stronger for US-born Hispanics and foreign-born Hispanics relative to Whites when respondents lived in counties characterized by high levels of absolute mobility (Figure 2, Left Column, Row 1) and high levels of income inequality (Figure 2, Right Column, Row 3), respectively, during childhood. Collectively, results provide support for Hypothesis 2.

Figure 2.

Predicted CMR scores by education group, race/ethnicity/nativity, and county context. Results from Poisson regression models stratified by advantaged (high mobility; low unemployment; low income inequality; high spending on education and health; and low proportions of residents with low education) and disadvantaged (low mobility; high unemployment; high income inequality; low spending on education and health; and high proportions of residents with low education) county contexts (Appendix Table 1, Columns 2-7). All models control for age, sex, childhood self-rated health, parental education, household receipt of welfare, percentage of teachers with MA or higher, and school disadvantage.

In which contexts do racial/ethnic and nativity groups experience the highest benefits of educational attainment?

Tables 2a-2d summarize results of educational disparities in CMR within racial/ethnic-nativity subgroups across different county environments. Appendix Figures 1-4 correspond to models from Tables 2a-2d, respectively, and plot predicted values of CMR for each education group within each county context (A=advantaged, D=disadvantaged). Table 2a displays results among Whites. Findings indicate that across almost every type of environment, Whites with higher levels of education experience lower levels of CMR than their lower educated counterparts. There are, however, two exceptions. There is no significant educational disparity in CMR among Whites who lived in counties characterized by high levels of absolute mobility and by high per capita spending on health. Appendix Figure 1 illustrates these results. The predicted CMR score is consistently higher for Whites with no college degree compared to those with a college degree or more in every county context except high mobility counties and counties with high per capita health spending.

Table 2a.

Poisson Regressions of Educational Disparities in CMR among US-Born Whites, by County-level Characteristics (N=7,920)

	Absolute Mobility		Prop. Unemployed		Gini Coefficient		Per Capita Expenditure—Education		Per Capita Expenditure—Health		Prop. Residents with Less than HS Diploma
	High	Low	High	Low	High	Low	High	Low	High	Low	High	Low
Education
College Degree or More	−.046 (.047)	−.200*** (.036)	−.147*** (.034)	−.159*** (.040)	−.210* (.081)	−.146*** (.028)	−.114** (.035)	−.181*** (.040)	−.072 (.045)	−.185*** (.037)	−.181*** (.043)	−.152*** (.039)
Constant	−.529 (.404)	−.024 (.248)	−.739* (.305)	.102 (.270)	−.228 (.399)	−.197 (.247)	−.480 (.392)	−.071 (.254)	.394 (.386)	−.335 (.258)	−.221 (.366)	−.140 (.261)
N	2,002	5,918	2,004	5,916	1,875	6,045	2,489	5,431	2,113	5,807	2.170	5,750

^tp<.10

^*p<.05

^**p<.01

^***p<.001

Note: logged expected count values of CMR are presented; standard errors are in parentheses; all models control for age, sex, childhood self-rated health, parental education, household receipt of welfare, percentage of teachers with MA or higher, and school disadvantage

Table 2d.

Poisson Regressions of Educational Disparities in CMR among Foreign-Born Hispanics, by County-level Characteristics (N=547)

	Absolute Mobility		Prop. Unemployed		Gini Coefficient		Per Capita Expenditure—Education		Per Capita Expenditure—Health		Prop. Residents with Less than HS Diploma
	High	Low	High	Low	High	Low	High	Low	High	Low	High	Low
Education
College Degree or More	−.145 (.272)	−.366** (.133)	−.456*** (.151)	−.1.44 (.156)	−.617*** (.017)	−.220 (.135)	−.203 (.277)	−.324* (.139)	−.323** (.111)	−.199 (.197)	−.414*** (.097)	−.221 (.160)
Constant	−.868 (1.103)	1.065 (.874)	3.037** (1.073)	−.569 (.585)	2.597*** (.478)	.641 (.900)	1.354 (1.579)	.661 (.817)	1.310 (.793)	.069 (.872)	2.113*** (.495)	.009 (.712)
N	101	446	262	285	300	247	130	417	292	255	249	298

^tp<.10

^*p<.05

^**p<.01

^***p<.001

Note: logged expected count values of CMR are presented; standard errors are in parentheses; all models control for age, sex, childhood self-rated health, parental education, household receipt of welfare, percentage of teachers with MA or higher, and school disadvantage

Table 2b displays results among Blacks. In contrast to the pattern for Whites, there are few significant differences in predicted CMR for Black young adults with no college degree compared to those with a college degree or more. College completion is associated with significantly lower CMR only among those who lived in childhood counties with high absolute mobility, fewer unemployed or low education residents, and high per capita spending on health. Appendix Figure 2 indicates that the lowest levels of predicted CMR are observed among Black young adults who completed college and who lived in counties characterized by high levels of upward mobility, low proportions of unemployed residents, and high spending on education.

Table 2b.

Poisson Regressions of Educational Disparities in CMR among US-Born Blacks, by County-level Characteristics (N=3,060)

	Absolute Mobility		Prop. Unemployed		Gini Coefficient		Per Capita Expenditure—Education		Per Capita Expenditure—Health		Prop. Residents with Less than HS Diploma
	High	Low	High	Low	High	Low	High	Low	High	Low	High	Low
Education
College Degree or More	−.235*** (.069)	−.063 (.049)	−.009 (.074)	−.147** (.049)	−.111 (.091)	−.085t (.049)	−.159t (.094)	−.075 (.049)	−.143* (.072)	−.067 (.054)	.040 (.057)	−.157** (.052)
Constant	−.505 (.607)	.128 (.310)	.308 (.419)	−.127 (.349)	.870t (.453)	−.277 (.308)	.550 (.557)	−.123 (.330)	−.576 (.420)	.264 (.343)	−.178 (.472)	.122 (.327)
N	712	2,348	767	2,293	852	2,208	744	2,316	978	2,082	883	2,177

^tp<.10

^*p<.05

^**p<.01

^***p<.001

Note: logged expected count values of CMR are presented; standard errors are in parentheses; all models control for age, sex, childhood self-rated health, parental education, household receipt of welfare, percentage of teachers with MA or higher, and school disadvantage

Similar to Black adults, the education-CMR relationship differs across county contexts for US-born Hispanics (Table 2c). US-born Hispanics who achieve a college degree or more exhibit lower levels of CMR when living in early life county contexts with high absolute mobility, low levels of income inequality, or high per capita spending on education and health. Unlike Black respondents, however, higher education is associated with better health for those growing up in counties with both high and low proportions of residents who are unemployed or who have less than a high school diploma. Predicted CMR values among US-born Hispanics are plotted in Appendix Figure 3. As displayed in this figure, the lowest levels of CMR are observed among US-born Hispanics who achieve a college degree or more and lived in childhood counties characterized by high levels of mobility and high per capita spending on health.

Table 2c.

Poisson Regressions of Educational Disparities in CMR among US-Born Hispanics, by County-level Characteristics (N=1,789)

	Absolute Mobility		Prop. Unemployed		Gini Coefficient		Per Capita Expenditure—Education		Per Capita Expenditure—Health		Prop. Residents with Less than HS Diploma
	High	Low	High	Low	High	Low	High	Low	High	Low	High	Low
Education
College Degree or More	−.396*** (.121)	−.079 (.072)	−.183* (.081)	−.214* (.091)	−.075 (.158)	−.225** (.074)	−.324** (.101)	−.161t (.088)	−.324** (.117)	−.135t (.073)	−.244* (.120)	−.202* (.080)
Constant	.328 (.676)	.375 (.462)	.777 (.850)	.045 (.401)	−.803 (.611)	.471 (.466)	.997 (.762)	.166 (.439)	.201 (.761)	.334 (.487)	1.004 (1.005)	.145 (.433)
N	491	1,298	477	1,312	651	1,138	712	1,077	631	1,158	359	1,430

^tp<.10

^*p<.05

^**p<.01

^***p<.001

Note: logged expected count values of CMR are presented; standard errors are in parentheses; all models control for age, sex, childhood self-rated health, parental education, household receipt of welfare, percentage of teachers with MA or higher, and school disadvantage

The patterns documented among foreign-born Hispanic young adults diverge from those of Blacks and US-born Hispanics. Results indicate that foreign-born Hispanics who achieve a college degree or more exhibit significantly lower levels of CMR relative to their lower educated counterparts only when living in counties characterized by low absolute mobility, high proportions of unemployed and low-educated residents, high income inequality, and low per capita spending on education (Table 2d). One exception is per capita spending on health, in which the educational disparity is significant only among those raised in counties with high per capita spending on health. Appendix Figure 4 indicates that while the lowest levels of CMR occurred in counties characterized by high levels of income inequality and low-educated residents, the predicted CMR scores for foreign-born Hispanics with and without a college degree are quite low, often lower than their highly educated White, Black, and US-born counterparts. Taken together, the above results provide partial support for Hypothesis 3.

Discussion

Education is widely recognized as a powerful determinant of health; yet, the literature to date provides inconsistent evidence as to whether the association between education and health is comparable across racial/ethnic minority and nativity groups. Whether and how characteristics of early life environments differentially influence the educational attainment-health relationship in young adulthood across population groups also remains unclear. This information is needed to advance theories and evidence regarding social stratification and health in contemporary US society. The present study is a first step toward this advancement, examining the extent to which educational differences in health vary across race/ethnicity, nativity, and early life county characteristics among young adults.

Using data from Add Health, findings support three major conclusions. First, consistent with accumulating evidence (Fuller-Rowell et al. 2015; Gaydosh et al. 2018; Kimbro et al. 2008), Black young adults with high levels of education did not exhibit the same levels of CMR as their White and Hispanic counterparts. In the context of past scholarship on the health consequences of social mobility among Black Americans, it is likely that the pathways to higher educational attainment are riddled with institutional barriers and increased exposure to psychosocial and environmental stressors. Such factors may reduce Black young adults’ ability to translate higher education into commensurate levels of income, occupational status, and neighborhood affluence (Adelman 2004; Patillo 2005; South et al. 2016; Williams et al. 2010) as well as over activate the body’s stress response (Brody et al. 2013; James et al. 1983).

Second, there were few contexts in which the association between education and CMR was similar for Black and White young adults, and wherein the association was stronger among Hispanic young adults relative to their White counterparts. Notably, the educational disparity in CMR was similar or even wider for Black and US-born Hispanic young adults than Whites when living in counties characterized by high levels of absolute mobility and high per capita spending on health during childhood. Such contexts may be more likely to provide resources for healthy living for all residents, regardless of eventual educational attainment.

Third, our results indicated that for racial/ethnic minority and foreign-born young adults, the relationship between education and CMR depended on the characteristics of counties in which they lived during childhood/adolescence. Among Whites, however, the educational difference in health was robust, with higher education linked to better health regardless of the context they lived as children/adolescents. It is possible that given their relative advantages in US society, there is less of a need for White Americans to expend psychological resources or employ high effort coping strategies to attain socioeconomic resources (e.g., Pearson 2008), even in the face of contextual disadvantages experienced in early life.

There were, however, two exceptions for Whites. Higher education was not associated with better health among those who lived in counties characterized by high levels of absolute mobility and high per capita spending on health. Comparison of predicted values of CMR across groups indicated that Whites with less than a college degree who lived in high mobility contexts as children/adolescents had the lowest predicted CMR compared to their lower educated, White counterparts in other contexts. Their predicted CMR was even lower than that of Black and US-born Hispanic young adults with high levels of education across almost all contexts. Thus, there may be a potential floor effect of education for Whites who grow up in highly advantaged contexts. For Whites who grew up in counties with high government expenditure on health, further exploration of results suggested that childhood health and school disadvantage accounted for the educational disparity, indicating the importance of early life factors for setting health trajectories in these contexts.

For Blacks and US-born Hispanics, higher education was associated with better physical health primarily among those who lived in advantaged counties as adolescents, particularly in counties that facilitated opportunities for upward social mobility. Conversely, Blacks and US-born Hispanics who lived in counties characterized by limited future opportunity prospects, limited investment in residents, or lower educational attainment of surrounding residents and who went on to attain high levels of education did not exhibit better health than their less educated counterparts. When contrasted to the findings that the education-health relationship in general is weaker for Black young adults, these patterns suggest that living in counties that provide more advantages and support to residents may help facilitate the translation of higher educational attainment into other forms of social, economic, and health capital for all racial/ethnic groups. Moreover, characteristics of early life counties may create specific cultures that shape the development of lifelong psychological and behavioral strategies for dealing with the stresses, strains, or opportunities of educational attainment (e.g., Krieger 2001). Those who grow up in disadvantaged contexts may be more likely to develop coping strategies characterized by resilience and persistence as they strive to acquire higher levels of education in the face of significant structural barriers. While leading to success in terms of educational attainment, such coping strategies may result in cumulative wear and tear on one’s body, particularly for racial/ethnic minorities (Brody et al. 2013; Gaydosh et al. 2018). Positive characteristics of early life counties, however, may alleviate the need for such individual coping strategies, resulting in better health for all residents who attain high levels of education.

For foreign-born Hispanics, results suggest a particularly strong relationship between educational attainment and health when individuals lived in relatively disadvantaged counties as children/adolescents. Such findings suggest that upwardly mobile—attaining high levels of education despite living in lower-resourced counties in early life—is associated with better health. Several possibilities may explain these patterns. First, it is likely that health-protective resources such as higher incomes and access to quality health insurance accompany higher levels of education. Second, perhaps achieving upward mobility, particularly as an immigrant to the US, is linked to social psychological factors that mitigate the harmful effects of stressors on health. Given that Hispanic American immigrants disproportionately occupy lower socioeconomic positions, there may be unique protective or buffering resources (e.g., social support; lower rates of smoking or binge drinking) available to this group within their receiving communities (Kimbro 2009; Osypuk et al. 2009; Viruell-Fuentes et al. 2013; Zhang et al. 2015). Furthermore, prior work suggests that Hispanic Americans living in disadvantaged environments may compare their own circumstances to their lower SES Hispanic counterparts or individuals from their home country who did not migrate (Campbell et al. 2012; Wolff et al. 2010). Such positive comparisons, which influence how individuals perceive, understand, and navigate their objective conditions and realities, may be linked to higher levels of self-esteem, greater sense of accomplishment, weaker perceptions of social devaluation, and lower physiological stress (Adler et al. 2000; Campbell et al. 2012; Cohen et al. 2008; McEwen and Gianaros 2010). Therefore, foreign-born Hispanics who live in disadvantaged counties in early life may experience fewer harmful consequences of stressors associated with attainment processes as a result of social comparison processes.

Findings from this study raise important questions regarding the link between educational attainment and health. First, to what extent are the associations documented here applicable for other indicators of physical, mental, and emotional well-being? Given the distinct pathways linking social factors to physical and mental health outcomes, it is possible that educational attainment is associated with better mental health in the same contexts in which education was not associated with better physical health. Second, what mechanisms might account for these patterns? While not explicitly examined here, we posit that early life counties structure objective conditions that facilitate educational attainment and may shape coping strategies that produce differences in the protective effect of educational attainment on health. Other individual-level (e.g., stressors, social and personal resources) and contextual-level factors (e.g., school and workplace environments), however, also likely play a role. Such factors, included those we posited, should be evaluated in future research.

Third, what additional role might adult contexts have in shaping the link between education and health? Life course accumulation of risks and resources may further differentiate the protective or negative consequences of educational attainment across various contexts. For example, it is possible that living in counties that differ drastically in demographic or socioeconomic characteristics from one’s county of origin may be linked to feelings of isolation, disruption of supportive social networks, or increases in discrimination (Cole and Omari 2003; Hudson et al. 2016), all of which may further attenuate any health benefits associated with educational attainment. Last, given that systems of inequality such as race/ethnicity and nativity do not operate independently of one another (e.g., Collins 2015; López and Gadsden 2016), other important social characteristics such as gender need to be incorporated into this line of research. Accumulating evidence suggests that pathways to health vary by intersections of race/ethnicity, nativity, and gender (Brown 2018; Brown et al. 2016; Hargrove 2018)—thus, it is possible that the relationships among education, place, and health depend on such intersections as well.

Limitations

This study has several limitations. First, small sample sizes precluded our ability to conduct independent analyses of other racial/ethnic and nativity groups, including Asian Americans, Native Americans, foreign-born Blacks and Whites, and ethnic subgroups within the Hispanic population. Further, given the limited sample of foreign-born Hispanics (N=547), results presented for this group should be interpreted with caution as cell sizes are likely small. It will be important for future studies to explore the questions posed here in datasets with larger samples of foreign-born Americans. Second, dichotomizing our county-level predictor variables based on race-specific 75^th percentile cut-points likely reduced some of the variability in our predictors and obscured non-linear relationships between our predictors and outcomes (Alton and Royston 2006). Yet, given the stark patterns of racial and socioeconomic residential segregation in the US (Patillo 2013; Sharkey 2014), as well as our objectives to evaluate the role of advantaged vs. disadvantaged counties, we believe our choice to dichotomize facilitates meaningful comparisons across the county-level contexts under examination in ways that better capture the realities of segregation in US counties.

Third, while we did not have detailed survey data on the types of environments in which young adults lived prior to entering the Add Health study, approximately 95% of respondents remained in the same county between Waves I and II of Add Health when respondents were adolescents. Moreover, ancillary analysis suggested that our results are substantively robust to controlling for length of residence in respondents’ Wave I location. Still, further analysis is needed to directly test whether movement prior to childhood/adolescence shapes educational attainment processes. Relatedly, examination of the characteristics of the schools in which young adults navigate is beyond the scope of the study, as we sought to provide initial documentation of trends in the education-health relationship by race/ethnicity, nativity, and county-level characteristics. Yet, it will be important for future research to examine school environments, as prior work suggests that risk (e.g., interpersonal experiences of discrimination) and protective factors (e.g., membership in social groups) experienced during educational attainment processes are key individual-level mechanisms that may dilute the protective effects of education (Cole and Omari 2003; Griffith, Hurd, and Hussain 2019; Hardaway and McLoyd 2009; Hudson et al. 2013; Von Robertson et al. 2016).

Conclusion

This study advances prior scholarship by documenting educational disparities in health among racial/ethnic-nativity groups who lived in different economic, policy, and social environments in childhood/adolescence. Results suggest the degree to which higher education is associated with better health among young adults is dependent on the intersection of race/ethnicity, nativity, and early life county characteristics. Importantly, counties act as key early life contexts that structure the extent to which educational attainment processes may strain or protect health for diverse population subgroups. By contextualizing educational disparities in health by race/ethnicity and nativity, this study provides insight into the place-based mechanisms that may differentiate the association between higher education and health for various racial/ethnic and nativity groups. Such knowledge deepens our understanding of broader patterns of inequality and may contribute to appropriate interventions to address the differential impact of education on health.