Skip to main content
PMC home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2020 Oct 1.
Published in final edited form as: Prev Sci. 2019 Oct;20(7):1009–1020. doi: 10.1007/s11121-019-01033-1

Differential Consequences: Racial/Ethnic and Gender Differences in the Enduring Impact of Early Disadvantage on Heavy Drinking in Midlife

Katherine J Karriker-Jaffe 1, Jane Witbrodt 1, Nina Mulia 1
PMCID: PMC6721975  NIHMSID: NIHMS1534270  PMID: 31292812

Abstract

Background:

We use a “chain of risks” model to identify risk factors for prolonged heavy drinking in a nationally-representative US sample followed from adolescence to middle age, focusing on educational mediators and differential consequences of early exposure to family poverty and area-level disadvantage.

Methods:

Using data from the 1979 National Longitudinal Survey of Youth (civilian respondents ages 14–19 at baseline, N=5,781), longitudinal path models assessed racial/ethnic and gender differences in indirect effects of early disadvantage (duration of exposure to family poverty and area-level disadvantage during adolescence) on midlife heavy drinking. Educational mediators were high school academic performance (taking remedial coursework), high school completion, and attaining a college education. Subgroups were based on race/ethnicity (50.7% White, 30.5% Black, 18.8% Hispanic respondents) and gender (49.6% males).

Results:

There was a significant indirect path from family poverty during adolescence to poor high school academic performance, lower educational attainment and more heavy drinking in midlife. For Black respondents, there was an additional direct effect of early area-level disadvantage on greater midlife heavy drinking that was not seen for other groups. The effect of family poverty on reduced high school graduation was stronger for males than females.

Conclusion:

Enduring impacts of family poverty duration during adolescence on educational attainment have consequences for health risk behaviors in midlife. Due to differential exposure to early adversity, intersectoral interventions are needed to reduce disparities in alcohol outcomes and to promote health equity among high-risk populations.

Keywords: alcohol use, racial/ethnic disparities, socioeconomic disadvantage, education

Recent research has highlighted racial/ethnic disparities in heavy drinking and alcohol-related consequences in midlife (Chartier, Vaeth, & Caetano, 2013). In this study, we pursue two goals. First, we identify factors that increase risk for midlife heavy drinking in a nationally-representative, racially-and ethnically-diverse United States (US) sample followed from adolescence to middle age. Second, we explore differential consequences of early exposure to family poverty and area-level disadvantage, emphasizing how pathways from early disadvantage to educational attainment and midlife heavy drinking vary for key population subgroups.

Heavy Drinking over the Lifecourse

Heavy drinking typically declines rapidly from the late 20s and early 30s and into middle adulthood (Maggs & Schulenberg, 2005), but there are important racial/ethnic and gender differences in patterns of heavy drinking over the lifecourse (Mulia et al., 2017). Some evidence suggests Black/African American (hereafter, Black) and Hispanic/Latino/Latina (hereafter, Hispanic) drinkers begin heavy drinking later than White/Caucasian (hereafter, White) drinkers (Williams, Mulia, Karriker-Jaffe, & Lui, 2018), but Black drinkers also are more likely to persist in frequent heavy drinking beyond young adulthood (Mulia, Tam, Bond, Zemore, & Li, 2018) and they have later onset of alcohol use disorder than Whites (Grant et al., 2012). Perhaps due to differences in timing and duration of heavy drinking, there are many racial/ethnic disparities in alcohol-related consequences in middle and later adulthood, such as notably higher rates of alcohol-attributable mortality among Black and Hispanic people, as well as disparities in cirrhosis and other alcohol-related health conditions (Chartier et al., 2013). Interestingly, these disparities exist despite higher population abstinence rates for Black and Hispanic people compared to Whites (Delker, Brown, & Hasin, 2016).

Pathways from Early Disadvantage: A Chain of Risks

To date, the important body of research describing lifecourse drinking largely reflects the developmental psychopathology perspective, giving less attention to potential influences of socioeconomic conditions and associated experiences, such as educational attainment, which also are related to drinking. In addressing our first research goal, we draw from lifecourse epidemiology, and propose a “chain of risks” model (Braveman & Barclay, 2009; Kuh, Ben-Shlomo, Lynch, Hallqvist, & Power, 2003) for examining how early disadvantage contributes to midlife heavy drinking. In this framework, early conditions of socioeconomic disadvantage are linked to subsequent negative experiences, such as difficulty in school, and each successive exposure increases the risk of later adverse outcomes, such as heavy drinking or substance use (Dodge et al., 2009; Sitnick, Shaw, & Hyde, 2014). We focus here on the enduring impacts of early exposure to family poverty and area-level disadvantage on two key educational outcomes: poor academic performance and low educational attainment (Ainsworth, 2002; Crowder & South, 2011; Harding, 2003; Owens, 2010; Rendón, 2014; Wodtke, Elwert, & Harding, 2016). In some cases, poor academic performance includes proficiency below grade level that requires remedial coursework. Such academic problems are an early step on the pathway from family poverty and area-level disadvantage to (under)achievement of important educational milestones such as graduating from high school and attaining a college degree. These later academic achievements have long-term protective health effects (Zimmerman, Woolf, & Haley, 2015).

In relation to adult alcohol use, lower levels of education predict increases in drinking in the early 20s (Casswell, Pledger, & Pratap, 2002; Muthén & Muthén, 2000) and alcohol problems after the mid-20s (Muthén & Muthén, 2000), as well as alcohol use disorders and alcohol-related mortality throughout adulthood (Budhiraja & Landberg, 2016; Crum et al., 2006; Gauffin, Vinnerljung, & Hjern, 2015). Thus, in our chain of risks model, we examine educational outcomes as key mediators by which early disadvantage may be linked with alcohol use later in life (Obradović, Burt, & Masten, 2010; Roisman, Masten, Coatsworth, & Tellegen, 2004).

We include both family poverty and area-level disadvantage as distal predictors of later educational and alcohol outcomes. Growing up poor and in disadvantaged areas not only affects educational outcomes, but it also affects adult health (Kravitz-Wirtz, 2016). Place effects may be independent of impacts of family poverty on both educational and health outcomes (Karriker-Jaffe, Lönn, Cook, Kendler, & Sundquist, 2018; Robert, 1999). Further, education funding and policies are determined at the metropolitan and county levels (Osypuk & Galea, 2007; Zimmerman et al., 2015), rendering these areas highly relevant for understanding pathways from early disadvantage through education to heavy drinking later in life.

Differential Consequences

Our second goal in this study is to evaluate differential consequences of early exposure to family poverty and area-level disadvantage, emphasizing how pathways from early disadvantage to educational attainment and midlife heavy drinking vary for population subgroups defined by race/ethnicity or gender. Our conceptual model is premised on the understanding that race/ethnicity is a social status in the US which is associated with differential opportunities and resources (Williams & Collins, 1995; Williams, Lavizzo-Mourey, & Warren, 1994). Therefore, while there may be core pathways through which education affects heavy drinking and alcohol problems in general, racial/ethnic subgroups may differ in their cumulative exposure to risk factors as well as in the health consequences of that cumulative exposure.

In the US, Black and Hispanic youth are more likely to grow up in disadvantaged areas (Cellini, Signe-Mary, & Ratcliffe, 2008; Crosnoe, 2005), even compared to low-income White youth, in part due to residential racial and economic segregation (Osypuk & Acevedo-Garcia, 2010; Williams, 1999; Williams & Collins, 1995). This has negative implications for the kinds of schools that Black and Hispanic youth are able to attend, resources to support their educational achievement, and their long-term academic performance. Further, disadvantaged youth may be less likely than their more advantaged peers to overcome negative effects of low academic performance on later educational attainment (Albrecht & Albrecht, 2010). Although Blacks and Hispanics overall attain less education than Whites, there also appear to be differences in the benefits of education, including lower income returns (Williams, 1996, 1999) and reduced health benefits (Walsemann, Geronimus, & Gee, 2008) for Blacks as compared to Whites with similar educational attainment. Nonetheless, prior analyses of heavy drinking using the current dataset showed robust protective effects of higher levels of education for women and men of all racial/ethnic backgrounds (Mulia et al., 2017). Here we expand on prior analyses by examining distal predictors of educational attainment by race/ethnicity.

In addition to revealing racial/ethnic differences in drinking patterns in adulthood, prior work with these data show marked gender differences in trajectories of heavy drinking for all racial/ethnic groups (Mulia et al., 2017). A study using Swedish population registry data to examine pathways from early exposure to neighborhood disadvantage to later development of alcohol use disorder showed important gender differences as well: While neighborhood disadvantage was more strongly associated with lower school achievement for young men than for young women, low family socioeconomic status was more strongly associated with lower school achievement for young women than for young men (Karriker-Jaffe et al., 2018). Additionally, reduced social functioning in young adulthood (including lower educational attainment) was more strongly associated with early alcohol use disorder for young men than for young women (Karriker-Jaffe et al., 2018). Here we test a similar model, using longitudinal data from a large, US national survey that allows for testing of gender and racial/ethnic differences.

Study Aims and Hypotheses

Building on our prior work with these and other data sources, we identify pathways from early disadvantage to midlife heavy drinking, emphasizing the role of key educational mediators. Extending prior models, we also test for differential consequences of early exposure to disadvantage, emphasizing how enduring effects of disadvantage on heavy drinking vary by race/ethnicity and gender. We expected there would be an overall indirect effect of early disadvantage on midlife heavy drinking, with both family poverty and area-level disadvantage associated with subsequently less educational success (indicated by remedial coursework during high school, lower rates of high school graduation, and lower rates of college graduation), and with the educational markers being associated with later heavy drinking (see Figure 1). Based on prior research, we also expected enduring effects of early disadvantage would be more pronounced for Blacks and Hispanics compared to Whites, and for men compared to women.

Figure 1.

Figure 1

Open in a new tab

Conceptual model: Key educational pathways from early disadvantage to heavy drinking in middle adulthood

Note. Dashed arrows represent hypothesized moderation effects whereby the relationships would vary by race/ethnicity and/or gender.

Methods

Dataset

Data are from the public access and restricted-use files of the 1979 National Longitudinal Survey of Youth (NLSY), a nationally-representative study of civilian respondents followed for over 30 years (Rothstein, Carr, & Cooksey, 2019). We used data from up to 25 interviews, which were conducted annually through 1994 and then biennially through 2012. The 2012 interview had a response rate of 79% among respondents remaining eligible for follow-up (Rothstein et al., 2019). We excluded the military sample and the oversample of low-socioeconomic status White respondents, because they were not followed into midlife. We focus on remaining respondents who were between ages 14–19 at baseline and who also had information on family poverty and area-level disadvantage during adolescence (N=6,400; 98.6% of eligible 14–19 year-olds interviewed in 1979). Our final analytic sample included 4,226 respondents who self-identified as either White, Black or Hispanic and who had data on baseline covariates and midlife heavy drinking. Given resultant sample sizes and preliminary analyses, analytic subgroups were based on race/ethnicity (50.7% White, 30.5% Black, 18.8% Hispanic) and gender (49.6% males).

Measures

The key outcome was midlife frequency of heavy drinking, defined as the average number of days drinking 6 or more drinks in past month, which was reported at ages 36–52 on 5 surveys conducted between 2002–2012. We also included a measure of frequency of heavy drinking during young adulthood, also defined as the average number of days drinking 6 or more drinks in the past month, which was reported at ages 25–35 on up to 3 surveys (depending on respondents’ age at baseline) conducted between 1988–1994. The response options changed in 2006, so following prior work with these data (Mulia et al., 2017), we used a four-category variable (never, less than once a week, 1–2 times per week, more than 2 times per week) coded using the weighted empirical means of the midpoints for each response category, and limiting the upper value to be consistent across both measures of heavy drinking. Assigned category values were 0, 1.8 (mean of once and 2–3 times per month from pre-2006 response options), 5.1 (mean of 4–5 times and 6–7 times per month), and 11.0 (mean of 8–9 times and 10 or more times per month), and the outcome was treated as a continuous variable. In our analytic sample, the mean frequency of heavy drinking in young adulthood was 1.27 times per month (SD=2.06), with means of 0.66 (SD=1.33) for women and 1.89 (SD=2.45) for men; for midlife heavy drinking it was 0.65 (SD=1.60), with means of 0.30 (SD=0.99) for women and 1.01 (SD=1.98) for men.

Early disadvantage

included two variables averaged across all surveys a respondent completed between ages 14–19, with the number of surveys depending on age at baseline (14 year-olds had data from up to six surveys; 19 year-olds had data from one survey). Family poverty duration was the proportion of time during adolescence (ages 14–19) respondents were exposed to family poverty. Reports of family income were for the prior calendar year, and poverty status was based on family size. In our analytic sample, average family poverty duration was 23.9% (SD=35.9), with averages of 10.4% (SD=24.1) for White respondents, 42.4% (SD=41.5) for Black respondents, and 32.3% (SD=38.3) for Hispanic respondents. Area-level disadvantage was based on proportions of county families living below the poverty level, the unemployed civilian labor force, and female-headed households. These data came from the restricted-use files containing data from the 1980 US Census; information on smaller areas (such as Census tracts) was not available. County-level indicators were averaged into a composite score for each year, and these composites were then averaged across each respondent’s adolescent period (ages 14–19). In our analytic sample, the average area-level disadvantage score was 9.3% (SD=3.4), with averages of 8.2% (SD=2.7) for White respondents, 10.6% (SD=3.5) for Black respondents, and 10.2% (SD=3.7) for Hispanic respondents.

We included three dichotomous educational mediators. High school academic performance was indicated by enrollment in remedial math or English classes during high school. Approximately one-quarter of the sample (22.5%) was enrolled in at least one remedial class during high school, which was more common among minority students (13.8% for White, 34.2% for Black and 34.7% for Hispanic respondents). High school graduation was indicated by completing 12 years of education by age 19. Almost three-quarters (73.2%) had achieved a high school education by age 19, but there were stark racial/ethnic differences (81.9% for White, 67.9% for Black and 57.0% for Hispanic respondents). College education was indicated by completing at least four years of college or university education by age 25, as the majority of 4-year college students complete their degree within 6 years (National Center for Education Statistics). Less than one-fifth (16.9%) had achieved a college education by age 25, again, with large racial/ethnic differences (24.0% for Whites, 10.0% for Blacks and 7.9% for Hispanics).

Sociodemographic controls were age at baseline, gender, race/ethnicity (Black or Hispanic versus White), mother’s and father’s highest level of education (using two dichotomous variables for less than high school and more than high school versus high school graduate; assessed separately for each parent at baseline), whether respondent was living with both parents at baseline, marital status at age 25 (married or single), as well as childbearing and parenting, which included three indicators for having at least one child by age 18, by age 22, and between ages 25–35. We also adjusted for early onset of alcohol use (prior to age 15) and family history of alcohol problems, which was represented by a mean weighted score for all first- and second-degree relatives with alcohol problems (Chartier, Thomas, & Kendler, 2017).

Analysis

Longitudinal path models assessed indirect effects of early disadvantage on midlife heavy drinking, controlling for demographics. Modeling was conducted with Mplus (Muthén & Muthén, 2013), following recommendations of MacKinnon (2008) for assessment of mediated effects using the MODEL INDIRECT sub-command. We used the robust weighted least squares estimator (WLSMV), because the model contains both continuous and categorical variables (MacKinnon, 2008) and this estimator does not assume normally-distributed variables. The final, most parsimonious path model was chosen based on comparisons of nested models using the DIFFTEST procedure (Muthén & Muthén, 2013). For each path in the overall model, control variables that were not statistically significant were trimmed to preserve degrees of freedom. Model fit was assessed using difference testing and fit indices, including the comparative fit index (CFI) and the root mean square error of approximation (RMSEA).

After the full model was specified, we examined subgroup differences by conducting simultaneous multivariate path models with multiple groups analysis, which also used the DIFFTEST procedure. Similar to the overall model, all subgroup models included control variables to address potential confounders of the associations between early disadvantage and subsequent outcomes. Models assessing racial/ethnic differences did not include relationships of the control variables with family poverty or area-level disadvantage, nor the correlation between the two indicators of early disadvantage, due to non-convergence. We were unable to examine differences in the hypothesized relationships for groups defined by both race/ethnicity and gender due to small subsample sizes.

All analyses used weights generated using the NLSY custom weighting program (Bureau of Labor Statistics, 2016), which adjusts for sampling design (including probability of selection, cooperation rates, and oversampling) and includes post-stratification weighting to represent the 1979 Census. These custom weights also account for use of data from multiple surveys.

Results

Early disadvantage was more strongly correlated with educational mediators than with drinking outcomes (Table 1). The educational mediators were modestly correlated with drinking outcomes (r=0.08 to −0.31), with correlations varying somewhat by race/ethnicity and gender.

Table 1.

Correlation matrices

All respondents (weighted N=4,226) (1) (2) (3) (4) (5) (6)
(1) Family poverty (ages 14–19) --
(2) Area-level disadvantage (ages 14–19) 0.289 --
(3) Remedial classes in high school (ages 15–19) 0.281 0.184 --
(4) High school education (by age 19) −0.307 −0.176 −0.544 --
(5) College education (by age 25) −0.294 −0.140 −0.507 0.735 --
(6) Young adult heavy drinking (ages 25–35) 0.004 −0.050 0.104 −0.130 −0.182 --
(7) Midlife heavy drinking (ages 36–52) 0.006 −0.012 0.029 −0.077 −0.168 0.519
White respondents (weighted N=2,395) (1) (2) (3) (4) (5) (6)
(1) Family poverty (ages 14–19) --
(2) Area-level disadvantage (ages 14–19) 0.125 --
(3) Remedial classes in high school (ages 15–19) 0.155 0.010 --
(4) High school education (by age 19) −0.261 −0.154 −0.599 --
(5) College education (by age 25) −0.185 −0.055 −0.518 0.763 --
(6) Young adult heavy drinking (ages 25–35) 0.013 −0.081 0.167 −0.131 −0.186 --
(7) Midlife heavy drinking (ages 36–52) 0.007 −0.026 0.014 −0.076 −0.179 0.538
Black respondents (weighted N=1,091) (1) (2) (3) (4) (5) (6)
(1) Family poverty (ages 14–19) --
(2) Area-level disadvantage (ages 14–19) 0.205 --
(3) Remedial classes in high school (ages 15–19) 0.239 0.069 --
(4) High school education (by age 19) −0.275 −0.048 −0.438 --
(5) College education (by age 25) −0.284 −0.009 −0.426 0.665 --
(6) Young adult heavy drinking (ages 25–35) 0.053 0.016 0.082 −0.206 −0.333 --
(7) Midlife heavy drinking (ages 36–52) 0.078 0.080 0.071 −0.123 −0.235 0.458
Hispanic respondents (weighted N=740) (1) (2) (3) (4) (5) (6)
(1) Family poverty (ages 14–19) --
(2) Area-level disadvantage (ages 14–19) 0.291 --
(3) Remedial classes in high school (ages 15–19) 0.263 0.249 --
(4) High school education (by age 19) −0.277 −0.112 −0.423 --
(5) College education (by age 25) −0.249 −0.181 −0.420 0.658 --
(6) Young adult heavy drinking (ages 25–35) 0.034 0.035 0.061 −0.117 −0.209 --
(7) Midlife heavy drinking (ages 36–52) −0.011 −0.021 0.146 −0.070 −0.185 0.518
Female respondents (weighted N=2,182) (1) (2) (3) (4) (5) (6)
(1) Family poverty (ages 14–19) --
(2) Area-level disadvantage (ages 14–19) 0.276 --
(3) Remedial classes in high school (ages 15–19) 0.319 0.224 --
(4) High school education (by age 19) −0.303 −0.167 −0.542 --
(5) College education (by age 25) −0.220 −0.099 −0.541 0.724 --
(6) Young adult heavy drinking (ages 25–35) 0.004 −0.080 0.068 −0.109 −0.209 --
(7) Midlife heavy drinking (ages 36–52) −0.007 −0.014 −0.002 −0.082 −0.261 0.453
Male respondents (weighted N=2,044) (1) (2) (3) (4) (5) (6)
(1) Family poverty (ages 14–19) --
(2) Area-level disadvantage (ages 14–19) 0.303 --
(3) Remedial classes in high school (ages 15–19) 0.259 0.167 --
(4) High school education (by age 19) −0.320 −0.192 −0.533 -
(5) College education (by age 25) −0.326 −0.188 −0.495 0.747 --
(6) Young adult heavy drinking (ages 25–35) 0.021 −0.020 0.075 −0.114 −0.197 --
(7) Midlife heavy drinking (ages 36–52) 0.031 0.007 0.001 −0.054 − 0.169 0.493
Open in a new tab

Full Sample

Significant coefficients for the full sample are shown in Figure 2 and all pathways, including those for covariates, are included in Table 2. In the full sample, early family poverty was positively associated with poor academic performance during high school and negatively associated with attaining a high school education. Area-level disadvantage was positively associated with attaining a college education. Poor academic performance during high school was negatively associated with attaining a high school education and with college education, as well as with midlife heavy drinking, but it was not associated with heavy drinking during young adulthood. Attaining a high school education was positively associated with attaining a college education, and it was not significantly associated with heavy drinking. Attaining a college education was negatively associated with heavy drinking during young adulthood and midlife.

Figure 2.

Figure 2

Open in a new tab

Standardized coefficients for statistically significant paths for the total sample and key subgroups

Notes. Model for the total sample controls for gender, race/ethnicity, age at baseline, parental education, living with both parents, having children (at age 18, age 22 and between ages 25–35), being married at age 25, onset of drinking by age 15 and family history of problem drinking. See Table 2 for additional coefficients and indicators of model fit for the total sample. Bold arrows indicate the significant indirect effect of family poverty on midlife heavy drinking through the educational mediators. Bold italic text indicates coefficients that vary significantly across racial/ethnic groups or by gender; see Tables 3 and 4 for additional coefficients for each subgroup. T: Total sample; W: White respondents; B: Black respondents; H: Hispanic respondents; F: Female respondents; M: Male respondents; NS=not statistically significant.

Table 2.

Standardized coefficients from overall path model

Family poverty (ages 15–19) Area-level disadvantage (ages 15–19) Remedial classes in high school (ages 15–19) High school education by age 19 Completed 4 years college by age 25 Young adult heavy drinking (ages 25–35) Midlife heavy drinking (ages 36–52)
Age at baseline interview −.030**
Male .361* −.253** .074 .577** .204**
Race/ethnicity a
 Black .552** .574** .292** .249** −.033 −.218** −.074
 Hispanic .313** .488** .400** −.007 −.029 −.150** −.045
Mother’s education b
 Less than high school .328** .230** .148* −.203** −.179**
 More than high school −.079 −.027 −.187* .218** .156*
Father’s education b
 Less than high school .213** .176** .219** −.220** .052
 More than high school .003 .012 −.040 .255** .234**
Lived with both parents −.388** −.033 .053 .354** −.248
Family alcoholism history −.466* .281†
Onset of drinking by age 15 .176**
Children by age 18 −.344**
Married by age 25 −.082† −.211**
Children by age 22 −.876**
Children at ages 25 to 35 −.080*
Family poverty duration c .161** .142** −.064** −.020 .004 .009
Area-level disadvantage .036 −.005 .048* −.027 .023
Remedial classes −.422** −.157** .003 −.065*
High school education .438** −.019 .020
Completed 4 years college −.192** −.088*
Young adult heavy drinking .487**
        R-square .199 .135 .188 .451 .652 .147 .284
Open in a new tab

Fit statistics: estimated degrees of freedom=85, CFI=0.974, RMSEA=0.021;

a

White as referent

b

High school completion as referent

c

association between family poverty and area-level disadvantaged assessed using WITH statement

**

p<0.01

*

p<0.05

†

p<0.10

As shown by bold arrows on Figure 2, there was a significant indirect path from family poverty during adolescence to heavy drinking in midlife that included poor academic performance in high school, lower educational attainment and more heavy drinking in young adulthood (standardized coefficient, std coef=0.002, SE=0.001, p=.002). The residual direct effect of family poverty during adolescence on heavy drinking in midlife was not significant (std coef=0.01, SE=0.02 p=.60). The indirect path from area-level disadvantage during adolescence to heavy drinking in midlife was not significant (std coef=0.001, SE=0.001, p=.23).

Subgroup Models & Multiple Groups Analysis

Racial/ethnic differences.

Coefficients from path models for the racial/ethnic subgroups are shown in Table 3. Paths that varied by race/ethnicity (according to chi-square difference tests) also are shown on Figure 2.

Table 3.

Standardized coefficients, presented by racial/ethnic group

Remedial classes in high school (ages 15–19) High school education by age 19 Completed 4 years college by age 25 Young adult heavy drinking (ages 25–35) Midlife heavy drinking (ages 36–52)
White respondents (N=2363)
Family poverty duration .116** −.058* −.005 −.015 −.010
Area-level disadvantage −.008 −.073* .042 −.069** .011
Remedial classes −.497** −.163* .030 −.156**
High school education .426** −.064 −.028
Completed 4 years college −.137** −.111*
Young adult heavy drinking .517**
         R-square .091 .490 .709 .138 .324
Black respondents (N=1073)
Family poverty duration .182** −.116* −.067 .007 .032
Area-level disadvantage .004 .043 .083 .046 .089**
Remedial classes −.308** −.186† −.078 −.042
High school education .451** −.024 .099
Completed 4 years college −.284** −.190*
Young adult heavy drinking .397**
         R-square .124 .347 .547 .174 .243
Hispanic respondents (N=734)
Family poverty duration .084 −.095* −.001 .001 −.037
Area-level disadvantage .124† .042 −.032 .027 −.057
Remedial classes −.284** −.227* −.137† .105
High school education .464** .107 .075
Completed 4 years college −.374** −.115
Young adult heavy drinking .471**
         R-square .237 .394 .575 .264 .307
Open in a new tab
**

p<0.01

*

p<0.05

†

p<0.10

Notes. Model controls for gender, age at baseline, parental education, living with both parents, having children (at age 18, age 22 and between ages 25–35), being married at age 25, onset of drinking by age 15 and family history of problem drinking. Bold italics indicate coefficients that vary significantly across racial/ethnic groups. Model specifications varied slightly from those used for Table 2 and Table 4. See Analysis section for details.

Multiple groups analysis showed model fit was significantly improved by allowing the associations of area-level disadvantage with the heavy drinking outcomes to vary across groups. Area-level disadvantage was negatively associated with heavy drinking in young adulthood for White respondents only (not statistically significant for either Black or Hispanic respondents), and area-level disadvantage was positively associated with heavy drinking in midlife for Black respondents only (not statistically significant for either White or Hispanic respondents). For White respondents, neither the indirect nor direct effects of area-level disadvantage during adolescence on heavy drinking in midlife were statistically significant. For Black respondents, the direct effect of area-level disadvantage during adolescence on heavy drinking in midlife was statistically significant (std coef=0.09, SE=0.03, p=.002), but the indirect effect through the educational mediators was not. For Hispanic respondents, neither the indirect nor direct effects of area-level disadvantage were significant.

The indirect pathway from family poverty during adolescence to academic performance in high school, educational attainment and heavy drinking in midlife was marginally significant for both White (std coef=0.002, SE=0.001, p=.06) and Black respondents (std coef=0.003, SE=0.002, p=.07). The residual direct effect of family poverty during adolescence on heavy drinking in midlife was not statistically significant for either White or Black respondents, suggesting full mediation for both groups. For Hispanic respondents, neither the indirect nor direct effects of family poverty during adolescence on heavy drinking in midlife were significant.

Gender differences.

Coefficients from path models by gender are shown in Table 4. Chi-square difference tests showed significant gender heterogeneity in several specific paths (also see Figure 2). Family poverty during adolescence was associated with lower odds of high school graduation only for males (not for females). We were unable to test whether the association of area-level disadvantage and college education varied by gender, as the difference test was not estimable. Area-level disadvantage was associated with less heavy drinking in young adulthood for only females, and attaining a college education was associated with less heavy drinking in midlife only for males.

Table 4.

Standardized coefficients, presented by gender

Remedial classes in high school (ages 15–19) High school education by age 19 Completed 4 years college by age 25 Young adult heavy drinking (ages 25–35) Midlife heavy drinking (ages 36–52)
Females (N=2184)
Family poverty duration a .171** −.036 .007 .003 −.016
Area-level disadvantage .074† .005 .077*,b −.053* .033†
Remedial classes −.471** −.227** −.015 −.103†
High school education .311** −.070 −.017
Completed 4 years college −.192** −.108
Young adult heavy drinking .441**
         R-square .177 .470 .660 .101 .224
Males (N=2044)
Family poverty duration a .117** −.094** −.047 .004 .019
Area-level disadvantage .003 −.012 .019 b −.020 .019
Remedial classes −.400** −.119 −.004 −.060
High school education .515** −.012 .060
Completed 4 years college −.216** −.126*
Young adult heavy drinking .479**
         R-square .167 .418 .624 .057 .256
Open in a new tab
**

p<0.01

*

p<0.05

†

p<0.10

a

Association between family poverty and area-level disadvantaged assessed using WITH statement; standardized coefficients=.150 (p<.01) for females and .174 (p<.01) for males.

b

The difference test for whether the association of area-level disadvantage and college education varied by gender was not estimable.

Notes. Model controls for race/ethnicity, age at baseline, parental education, living with both parents, having children (at age 18, age 22 and between ages 25–35), being married at age 25, onset of drinking by age 15 and family history of problem drinking. Bold italics indicate coefficients that vary significantly by gender.

The indirect pathway from family poverty during adolescence to academic performance in high school, educational attainment and heavy drinking in midlife held for both genders. The indirect effect was statistically significant for both women (std coef=0.002, SE=0.001, p=.02) and men (std coef=0.002, SE=0.001, p=.04), and the residual direct effect of family poverty during adolescence on heavy drinking in midlife was not statistically significant, suggesting full mediation for both gender subgroups. Neither the indirect nor direct effects of area-level disadvantage during adolescence on heavy drinking in midlife were statistically significant for either women or men.

Discussion

Building on our prior work showing strong relationships between heavy drinking trajectories and educational attainment (Mulia et al., 2017), we used data from a nationally-representative US sample of adolescents followed into midlife to identify pathways from early disadvantage to later heavy drinking, emphasizing the role of key educational mediators. There was a significant indirect path from family poverty during adolescence to heavy drinking in midlife that included lower academic performance (remedial coursework in high school), lower educational attainment and more heavy drinking in young adulthood. This pathway was statistically significant for both males and females, but it was only marginally significant for White and Black respondents (not significant for Hispanics). Counter to our expectations based on prior literature, the indirect path from area-level disadvantage through the educational mediators to heavy drinking in midlife was not significant for any group. Our findings emphasize the importance of the early family context for later educational success, suggesting that families living in chronic poverty need additional services to support children’s educational development. Fortunately, successful intervention models exist (Hahn et al., 2015; Werner & Smith, 1992). These include programs that provide reading support in the earliest elementary grades; involve youth in a stable, caring relationship with an adult (relative or non-kin), who they view as a mentor; provide social/emotional skills training; and provide case management and linkages for students and their families to build connections with community mental health and financial services. The long-term impacts of such interventions may reach far into midlife.

We tested for differential consequences of early exposure to family poverty and area-level disadvantage, expecting that enduring effects of early disadvantage would be most pronounced for Black and Hispanic respondents compared to White respondents and for men compared to women. We only found partial support for these hypotheses. Specifically, for Black respondents, there was an additional direct effect of early area-level disadvantage on midlife heavy drinking that was not seen for either Hispanic or White respondents. This suggests a differentially negative and enduring effect of early exposure to disadvantaged communities for Black youth, who have a lower likelihood of mobility out of these areas, and thus greater lifecourse exposure to economically distressed environments, even across generations (Sharkey, 2008). Disadvantaged communities often suffer from a proliferation of alcohol outlets (LaVeist & Wallace, 2000), and this type of environment can foster heavy drinking. Future work should explore sources of resilience for low-income communities that may help support Black adolescents as they transition to adulthood.

There also was no evidence of indirect effects of area-level disadvantage through the educational pathways for any of the groups studied here. Although in bivariate analyses area-level disadvantage was negatively correlated with attaining a college education for most subgroups, in the path models, suppression effects are present with regard to educational attainment, as the sign of the coefficient changes (MacKinnon, Krull, & Lockwood, 2000). There are likely to be many pathways from early exposure to disadvantage to midlife health risk behaviors such as heavy drinking, and only one is emphasized here. Other pathways may include social norms (Karriker-Jaffe, Liu, & Kaplan, 2016), as well as psychological, physiological and neuroendocrine mechanisms resulting from—and interacting with (Hussong, Burns, Solis, & Rothenberg, 2013)—prolonged exposure to a variety of stressors that were not captured by our models. Although some studies suggest Black and Hispanic people may have lower health returns on education compared to their White counterparts (Walsemann et al., 2008), we did not find strong evidence that the protective effect of college education on heavy drinking varied systematically by race/ethnicity. Our findings suggest education still is an important protective resource, but future studies could examine alternate pathways to midlife heavy drinking.

Regarding hypothesized gender differences, adolescent family poverty was more strongly linked to reduced likelihood of high school graduation for males than for females. These findings are counter to some studies suggesting that the family environment may be more influential for girls than for boys (Karriker-Jaffe et al., 2018; Kroneman, Loeber, & Hipwell, 2004). Although the adjusted gender differences were not significant, the correlation between family poverty and enrollment in remedial coursework was stronger for high school girls (r=0.32) than for boys (r=0.26), which may indicate that young women are more likely to get support from their schools in response to poor academic performance than their male counterparts, although this speculation deserves further study. However, in this sample, receipt of remedial coursework was associated with significantly reduced likelihood of high school graduation for both females and males (as well as with significantly reduced likelihood of college education for young women), so other support services for high-risk students are needed to reduce long-term impacts of family poverty. Such services might include college programs that provide social support and mentoring by faculty members and older, successful students from similar backgrounds.

Strengths and Limitations

Our study has many notable strengths, such as the large sample size and the long follow-up period spanning adolescence to midlife. There are few longitudinal data resources available for multi-level studies of the long-term impacts of early exposure to disadvantage, and the NLSY data are unique in their comprehensive measures of educational outcomes as well as alcohol use over time. One limitation of these data, however, is a lack of small-area measures of exposure to neighborhood disadvantage. Although counties are important for determining policies and funding for education (Osypuk & Galea, 2007), local communities vary in terms of educational resources and quality. Historically in the US, Black and Hispanic people have been subjected to discrimination, which results in segregation to lower-income areas (Cellini et al., 2008; Crosnoe, 2005; Osypuk & Acevedo-Garcia, 2010; Williams, 1999; Williams & Collins, 1995). In this sample, there were not large differences in exposure to area-level disadvantage during adolescence for White, Black and Hispanic respondents, with average scores ranging from 8% disadvantaged residents in the counties where White adolescents lived, to just over 10% for the counties where Black and Hispanic adolescents lived. This may have contributed to the unexpected null findings for indirect effects of area-level disadvantage on education and heavy drinking. Additionally, we note that our measure of heavy drinking used a threshold of six or more drinks per occasion, which is higher than the recommended daily limits for men (no more than four drinks) and women (no more than three drinks) (National Institute on Alcoholism and Alcohol Abuse, 2009). Thus, we are underrepresenting heavy drinking in this sample, especially for women. Due to the timing of the alcohol consumption measures, we also were unable to account for heavy drinking during adolescence, which may impact educational outcomes. We did include an indicator of early onset (prior to age 15) of drinking; although predictive of heavy drinking in young adulthood, early onset was not related to any educational outcomes. Finally, despite the relatively large number of Black and Hispanic respondents, we did not have sufficient power to assess differential pathways across racial-gender subgroups, and we encourage future research to build upon these results.

Conclusions

Enduring impacts of exposure to family poverty during adolescence on educational attainment have consequences for health risk behaviors in midlife. Due to differential exposure to early adversity, intersectoral interventions are needed to reduce disparities in alcohol outcomes and to promote health equity among high-risk population subgroups.

Acknowledgments

Funding. The current secondary analysis study was supported by the National Institute on Alcohol Abuse and Alcoholism (NIAAA) at the National Institutes of Health (NIH) (R01AA022668, N. Mulia, PI, and P50AA005595, W. Kerr, PI). The research was conducted with restricted access to Bureau of Labor Statistics (BLS) data. The views expressed here do not necessarily reflect the views of the BLS. Content is solely the responsibility of the authors and does not represent official views of the NIH or NIAAA, which had no role in the study design, collection, analysis or interpretation of the data, writing of the manuscript, or the decision to submit the manuscript for publication.

Footnotes

Publisher's Disclaimer: This Author Accepted Manuscript is a PDF file of a an unedited peer-reviewed manuscript that has been accepted for publication but has not been copyedited or corrected. The official version of record that is published in the journal is kept up to date and so may therefore differ from this version.

Disclosure of Potential Conflicts of Interest. The authors declare no conflict of interest.

Ethical Approval. All procedures performed in the National Longitudinal Survey of Youth (NLSY) involving human participants were in accordance with ethical standards described in the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards. The current study is a secondary analysis of NLSY data. The Institutional Review Board of the Public Health Institute reviewed the study protocol and found it to be exempt.

Informed Consent. Informed consent was obtained from all individual participants included in the NLSY study.

References

  1. Ainsworth JW (2002). Why does it take a village? The mediation of neighborhood effects on educational achievement. Social Forces, 81(1), 117–152. [Google Scholar]
  2. Albrecht CM, & Albrecht DE (2010). Social status, adolescent behaviour, and educational attainment. Sociological Spectrum, 31(1), 114–137. [Google Scholar]
  3. Braveman P, & Barclay C (2009). Health disparities beginning in childhood: a life-course perspective. Pediatrics, 124(Suppl. 3), S163–S175. [DOI] [PubMed] [Google Scholar]
  4. Budhiraja M, & Landberg J (2016). Socioeconomic disparities in alcohol-related mortality in Sweden, 1991–2006: A register-based follow-up study. Alcohol Alcohol, 51(3), 307–314. [DOI] [PubMed] [Google Scholar]
  5. Bureau of Labor Statistics. (2016). National Longitudinal Surveys: Custom weighting program documentation [Accessed: 2016-04-20 Archived by WebCite® at http://www.webcitation.org/6guiInSY0]. Washington, DC. [Google Scholar]
  6. Casswell S, Pledger M, & Pratap S (2002). Trajectories of drinking from 18 to 26 years: identification and prediction. Addiction, 97(11), 1427–1437. [DOI] [PubMed] [Google Scholar]
  7. Cellini SR, Signe-Mary M, & Ratcliffe C (2008). The dynamics of poverty in the United States: a review of data, methods, and findings. Journal of Policy Analysis and Management, 27(3), 577–605. [Google Scholar]
  8. Chartier KG, Thomas NS, & Kendler KS (2017). Interrelationship between family history of alcoholism and generational status in the prediction of alcohol dependence in US Hispanics. Psychol Med, 47(1), 137–147. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Chartier KG, Vaeth PAC, & Caetano R (2013). Focus On: Ethnicity and the social and health harms from drinking. Alcohol Research: Current Reviews, 35(2), 229–237. [PMC free article] [PubMed] [Google Scholar]
  10. Crosnoe R (2005). Double disadvantage or signs of resilience? The elementary school contexts of children from Mexican immigrant families. American Educational Research Journal, 42(2), 269–303. [Google Scholar]
  11. Crowder K, & South SJ (2011). Spatial and temporal dimensions of neighborhood effects on high school graduation. Social Science Research, 40(1), 87–106. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Crum RM, Juon HS, Green KM, Robertson J, Fothergill K, & Ensminger M (2006). Educational achievement and early school behavior as predictors of alcohol-use disorders: 35-year follow-up of the Woodlawn Study. J Stud Alcohol, 67(1), 75–85. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Delker E, Brown Q, & Hasin DS (2016). Alcohol consumption in demographic sub-populations: an epidemiologic overview. Alcohol Research: Current Reviews, 38(1), 7–15. [PMC free article] [PubMed] [Google Scholar]
  14. Dodge KA, Malone PS, Lansford JE, Miller S, Pettit GS, & Bates JE (2009). A dynamic cascade model of the development of substance-use onset. Monogr Soc Res Child Dev, 74(3), vii–119. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Gauffin K, Vinnerljung B, & Hjern A (2015). School performance and alcohol-related disorders in early adulthood: a Swedish national cohort study. Int J Epidemiol, 44(3), 919–927. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Grant JD, Vergés A, Jackson KM, Trull TJ, Sher KJ, & Bucholz KK (2012). Age and ethnic differences in the onset, persistence and recurrence of alcohol use disorder. Addiction, 107(4), 756–765. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Hahn RA, Knopf JA, Wilson SJ, Truman BI, Milstein B, Johnson RL, … Community Preventive Services Task Force. (2015). Programs to increase high school completion: a community guide systematic health equity review. American Journal of Preventive Medicine, 48(5), 599–608. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Harding DJ (2003). Counterfactual models of neighborhood effects: the effect of neighborhood poverty on dropping out and teenage pregnancy. American Journal of Sociology, 109(3), 676–719. [Google Scholar]
  19. Hussong AM, Burns AR, Solis JM, & Rothenberg WA (2013). Future directions in the developmental science of addictions. J Clin Child Adolesc Psychol, 42(6), 863–873. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Karriker-Jaffe KJ, Liu H, & Kaplan LM (2016). Understanding associations between neighborhood socioeconomic status and negative consequences of drinking: a moderated mediation analysis. Prevention Science, 17(4), 513–524. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Karriker-Jaffe KJ, Lönn SL, Cook WK, Kendler KS, & Sundquist K (2018). Chains of risk for alcohol use disorder: Mediators of exposure to neighborhood deprivation in early and middle childhood. Health & Place, 50, 16–26. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Kravitz-Wirtz N (2016). Cumulative effects of growing up in separate and unequal neighborhoods on racial disparities in self-rated health in early adulthood. J Health Soc Behav, 57(4), 453–470. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Kroneman L, Loeber R, & Hipwell AE (2004). Is neighborhood context differently related to externalizing problems and delinquency for girls compared with boys? Clin Child Fam Psychol Rev, 7(2), 109–122. [DOI] [PubMed] [Google Scholar]
  24. Kuh D, Ben-Shlomo Y, Lynch J, Hallqvist J, & Power C (2003). Life course epidemiology. Journal of Epidemiology and Community Health, 57(10), 778–783. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. LaVeist TA, & Wallace JM Jr. (2000). Health risk and inequitable distribution of liquor stores in African American neighborhood. Social Science and Medicine, 51(4), 613–617. [DOI] [PubMed] [Google Scholar]
  26. MacKinnon DP (2008). Introduction to statistical mediation analysis New York: Laurence Erlbaum Associates. [Google Scholar]
  27. MacKinnon DP, Krull JL, & Lockwood CM (2000). Equivalence of the mediation, confounding and suppression effect. Prevention Science, 1(4), 173–181. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Maggs JL, & Schulenberg JE (2005). Trajectories of alcohol use during the transition to adulthood. Alcohol Research and Health, 28(4), 195–201. [Google Scholar]
  29. Mulia N, Karriker-Jaffe KJ, Witbrodt J, Bond J, Williams E, & Zemore SE (2017). Racial/ethnic differences in 30-year trajectories of heavy drinking in a nationally representative U.S. sample. Drug and Alcohol Dependence, 170, 133–141. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Mulia N, Tam T, Bond J, Zemore SE, & Li L (2018). Racial/ethnic differences in life-course heavy drinking from adolescence to midlife. Journal of Ethnicity in Substance Abuse, 17(2), 167–186. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Muthén BO, & Muthén LK (2000). The development of heavy drinking and alcohol-related problems from ages 18 to 37 in a U.S. national sample. Journal of Studies on Alcohol, 61(2), 290–300. [DOI] [PubMed] [Google Scholar]
  32. Muthén LK, & Muthén BO (2013). Mplus Version 7.2 Los Angeles, CA: Muthén & Muthén. National Center for Education Statistics; Fast Facts. Graduation rates [Available at: https://nces.ed.gov/fastfacts/display.asp?id=40] [Google Scholar]
  33. National Institute on Alcoholism and Alcohol Abuse (NIAAA). (2009). Rethinking Drinking: Alcohol and your health [Accessed: 2011-03-02 Archived by WebCite® at http://www.webcitation.org/5wtRANlQ7] (pp. 16). Bethesda, MD: NIAAA. [Google Scholar]
  34. Obradović J, Burt KB, & Masten AS (2010). Testing a dual cascade model linking competence and symptoms over 20 years from childhood to adulthood. Journal of Clinical Child and Adolescent Psychology, 39(1), 90–102. [DOI] [PubMed] [Google Scholar]
  35. Osypuk TL, & Acevedo-Garcia D (2010). Beyond individual neighborhoods: a geography of opportunity perspective for understanding for racial/ethnic health disparities. Health Place, 16(6), 1113–1123. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Osypuk TL, & Galea S (2007). What level macro? Choosing appropriate levels to assess how place influences population health. In Galea S (Ed.), Macrosocial Determinants of Population Health (pp. 399–435). New York: Springer. [Google Scholar]
  37. Owens A (2010). Neighborhoods and schools as competing and reinforcing contexts for educational attainment. Sociology of Education, 83(4), 287–311. [Google Scholar]
  38. Rendón M (2014). Drop Out and ‘Disconnected’ Young Adults: Examining the Impact of Neighborhood and School Contexts. Urban Review, 46(2), 169–196. [Google Scholar]
  39. Robert SA (1999). Socioeconomic position and health: The independent contribution of community socioeconomic context. Annual Review of Sociology, 25, 489–516. [Google Scholar]
  40. Roisman GI, Masten AS, Coatsworth JD, & Tellegen A (2004). Salient and emerging developmental tasks in the transition to adulthood. Child Development, 75(1), 123–133. [DOI] [PubMed] [Google Scholar]
  41. Rothstein DS, Carr D, & Cooksey E (2019). Cohort Profile: The National Longitudinal Survey of Youth 1979 (NLSY79). Int J Epidemiol, 48(1), 22–22e. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Sharkey P (2008). The intergenerational transmission of context. The American Journal of Sociology, 113(4), 931–969. [Google Scholar]
  43. Sitnick SL, Shaw DS, & Hyde LW (2014). Precursors of adolescent substance use from early childhood and early adolescence: Testing a developmental cascade model. Development and Psychopathology, 26(1), 125–140. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Walsemann KM, Geronimus AT, & Gee GC (2008). Accumulating disadvantage over the life course: Evidence from a longitudinal study investigating the relationship between educational advantage in youth and health in middle age. Research on Aging, 30(2), 169–199. [Google Scholar]
  45. Werner E, & Smith RS (1992). Overcoming the Odds: High risk children from birth to adulthood New York: Cornell University Press. [Google Scholar]
  46. Williams DR (1996). Race/ethnicity and socioeconomic status: measurement and methodological issues. International Journal of Health Services, 26(3), 483–505. [DOI] [PubMed] [Google Scholar]
  47. Williams DR (1999). Race, socioeconomic status, and health. The added effects of racism and discrimination. Annals of the New York Academy of Sciences, 896, 173–188. [DOI] [PubMed] [Google Scholar]
  48. Williams DR, & Collins C (1995). US socioeconomic and racial differences in health: patterns and explanations. Annual Review of Sociology, 21(1), 349–386. [Google Scholar]
  49. Williams DR, Lavizzo-Mourey R, & Warren RC (1994). The concept of race and health status in America. Public Health Reports, 109(1), 26–41. [PMC free article] [PubMed] [Google Scholar]
  50. Williams E, Mulia N, Karriker-Jaffe KJ, & Lui CK (2018). Changing racial/ethnic disparities in heavy drinking trajectories through young adulthood: a comparative cohort study. Alcoholism: Clinical and Experimental Research, 42(1), 135–143. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Wodtke GT, Elwert F, & Harding DJ (2016). Neighborhood effect heterogeneity by family income and developmental period. American Journal of Sociology, 121(4), 1168–1222. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Zimmerman EB, Woolf SH, & Haley A (2015). Understanding the relationship between education and health: A review of the evidence and an examination of community perspectives. In Kaplan RM, Spittel ML & David DH (Eds.), Population Health: Behavioral and Social Science Insights (pp. 347–384). Rockville, MD: Agency for Healthcare Research and Quality and Office of Behavioral and Social Sciences Research. [Google Scholar]

RESOURCES