Pathways of Health and Human Capital from Adolescence into Young Adulthood

Pathways of Health and Human Capital from Adolescence into Young Adulthood Abstract Social inequalities in health and human capital are core concerns of sociologists, but little research examines the developmental stage when such inequalities are likely to emerge—the transition to adulthood. With new data and innovative statistical methods, we conceptually develop, and empirically operationalize, pathways of physical health and human capital accumulation from adolescence into young adulthood, using an autoregressive cross-lagged structural equation model. Results reveal that pathways of health and human capital accumulate at differential rates across the transition to adulthood; evidence of cross-lagged effects lends support for both social causation and health selection hypotheses. We then apply this model to assess the presence of social inequality in metabolic syndrome—the leading risk factor of cardiovascular disease in the United States. Findings document social stratification of cardiovascular health that is robust to both observed and unobserved social and health selection mechanisms. We speculate that this social stratification will only increase as this cohort ages. Introduction Social inequality is a fundamental sociological concern. Decades of research have documented the trends, prevalence, and correlates of social stratification (Blau and Duncan 1967; Featherman and Hauser 1978; Grusky 2008; Hout 2007; Hout and DiPrete 2006). However, we still have much to learn about how inequalities develop across the early life course, from adolescence into young adulthood. Most social stratification research focuses on the quantity, quality, and access to human capital. Less attention has been paid to the development of physical health inequalities across the life course (Palloni [2006] excepted) and their relation to human capital development. Using rich longitudinal data, this study conceptually develops, and empirically operationalizes, intra-generational pathways of physical health and human capital accumulation from adolescence into young adulthood. This is the first major contribution of our study. We then gauge the impact of these pathways on cardiovascular health in young adulthood—our study’s second major contribution—as this is highly predictive of health and wealth in old age. Health and Human Capital across the Transition to Adulthood According to Blau and Duncan’s (1967) seminal work, social stratification has its roots in both inter- and intra-generational processes: father’s education and occupation (“social origins”) influences child’s education, and each contributes to child’s occupation (“social destination”). Subsequent conceptualizations affirmed this model (Featherman and Hauser 1978), spawning a generation of social stratification research. A common approach, and the one we take here, is to focus specifically on intra-generational processes, whereby early investments in human capital demonstrate lifelong gains for individuals by maximizing physical health as well as psychological and economic well-being (Aneshensel 1992; Ross and Wu 1996; Sewell and Hauser 1975). A separate literature takes a longitudinal approach, noting a pattern of path dependence in which poor physical health is sequentially contingent from one life stage to the next (O’Rand 2009; Willson, Shuey, and Elder 2007). However, neither literature has examined how human capital and health pathways develop and interact across the transition to adulthood. This agenda is challenging in multiple ways. Concerning human capital, there are conceptual and measurement difficulties in estimating pathways linking income potential across early life stages (Furstenberg 2000; Furstenberg, Brooks-Gunn, and Morgan 1989). Most adolescents live with their parents, are in the midst of their educational trajectories, and do not work full-time; thus, current income, educational level, or occupation are poor proxies for their eventual level of human capital. Similarly, individual income is an unreliable indicator of human capital in young adulthood, when some youth attend college while others move quickly into the labor market. We contribute new conceptual and measurement work in this area by estimating a pathway of human capital accumulation across the transition to adulthood using latent constructs that capture human capital at different life stages. Concerning physical health, much past work either constructs trajectories of health only across adulthood, or builds trajectories of health spanning childhood and adulthood by asking adults to retrospectively report childhood illness and general health (Chen, Yang, and Liu 2010; Haas 2006, 2008; Marmot et al. 2001; Palloni 2006; Warren 2009; Willson, Shuey, and Elder 2007). The latter approach suffers from recall bias, measurement error (Looker 1989), and mortality selection (Pollitt, Rose, and Kaufman 2005), as only individuals surviving to mid- to late adulthood are surveyed. Other studies use longitudinal data to identify health pathways from adolescence through young adulthood (Bauldry et al. 2012) or links between adolescent health and key educational outcomes (Jackson 2009), but employ a single, subjective indicator of physical health (self-reported health). Moreover, these studies face the difficulty of estimating causal effects of social factors on health, aside from key unmeasured factors (Adler, Bush, and Pantell 2012; Goldman 2001; Kawachi, Adler, and Dow 2010; Palloni 2006). All in all, we know relatively little about how early life health habits and behavior manifest within individuals over the early life course. We conceptually develop and estimate pathways of physical health, based on multiple objective and subjective measures of health, across the transition to adulthood. We also address the endogeneity of health (and human capital) by using multiple time-varying instrumental variables to identify the system of equations in our model. In addition, health and human capital development are likely interdependent. In adulthood, poor physical health is associated with lower levels of educational attainment, lower income, fewer hours worked in the labor market, and ultimately less accumulation of wealth over time (Currie and Madrian 1999; Haas 2006, 2008; Jackson 2009), pointing to health as a primary driver of social stratification. In turn, socioeconomic disadvantage is associated with health detriments through greater exposure to stress and poor physical environments (Adler and Ostrove 1999; Adler, Bush, and Pantell 2012)—in neighborhoods and schools, within families, and in the labor market through work in low-wage jobs. Furthermore, these processes are likely deeply intertwined, reflecting a mutual interdependence of health and human capital with strong path-dependent components (O’Rand 2001) that makes it difficult to produce unbiased estimates of either one (Adler, Bush, and Pantell 2012; Kawachi, Adler, and Dow 2010). We model these interdependencies and estimate their causal impacts. Our study takes the additional step of estimating the effects of accumulated human capital and physical health on an important, objective marker of cardiovascular health in young adulthood, metabolic syndrome, that is measured using anthropometric measures and physiological markers of health. Metabolic syndrome is the leading risk factor for adult-onset cardiovascular disease and Type II diabetes—both of which encompass significant morbidity and mortality risk in the United States (Beltrán-Sánchez et al. 2013). Theory and Research on Intra-Generational Development of Health and Human Capital The life course perspective is a useful organizing framework for understanding intra-generational development of human capital and physical health. While acknowledging the import of human agency (Hitlin and Elder 2007), events during childhood, adolescence, and adulthood are cumulative and are connected to broader structural context and social change (Elder 1995; Elder, Johnson, and Crosnoe 2003). The cumulative disadvantage hypothesis articulates how later-life inequalities are anchored in early life experiences (Merton 1968; Ross and Wu 1996). These early life experiences have been broadly defined to include differential exposures to persistent poverty, social support, and stressful neighborhood environments (DiPrete and Eirich 2006). The Importance of the Early Life Course Striking distinctions in economic well-being are drawn between individuals who complete high school and those who do not (Hallinan 1988; Hogan and Astone 1986). High school completion is an important credential for the low-wage labor force and opens many employment opportunities. Early returns to high school success are also evident in college attendance and completion; in turn, those who complete college benefit from more advantageous economic and employment opportunities in adulthood. Thus, adolescence and the transition to adulthood are critically formative for lifelong human capital. Furthermore, as family-, employment-, and schooling-related transitions have become less structured, human agency has played an increasingly important role in the transition to adulthood (Mortimer 1994; Shanahan 2000). This is a time when individuals try on new roles and navigate myriad choices in different realms (labor market, dating market, etc.); behavior during this stage forms the basis for later-observed economic behavior (Macmillan 2006; Zimmer-Gembeck and Mortimer 2006). Social inequalities in adult health are rooted in childhood experiences (Blackwell, Hayward, and Crimmins 2001; Hayward and Gorman 2004; Lynch, Kaplan, and Shema 1997). Childhood socioeconomic status (SES) has persistent effects on adult health outcomes, net of adult SES (Poulton et al. 2002). Identifying early life determinants of adult/old age health are active research questions (Adler, Bush, and Pantell 2012; George 2005; Kawachi, Adler, and Dow 2010). Furthermore, poor health in early life can constrain subsequent socioeconomic attainment and health maintenance (Elman and O’Rand 2007; O’Rand 2001).Yet, most studies rely on retrospective measures of early life events. Studies such as Add Health are advantageous because early life markers of social well-being and health are measured prospectively and, in the case of health, with multiple subjective and objective indicators. Much remains to be learned about the pathways linking early life exposures to adult health outcomes (Power and Hertzman 1997), although a growing literature proposes and tests potential pathways (see, for example, Almond and Currie [2011]; Ben-Shlomo and Kuh [2002]; Hertzman [2006]; Hertzman and Power [2006]; Kuh and Shlomo [2004]). Most such studies, however, have insufficient data describing exposures during adolescence—a stage in which young people assert their independence and begin to manage lifestyle choices and health behaviors that often persist (Harris 2010; Macmillan 2006)—and the transition to adulthood—a life stage dense with critical transitions (e.g., leaving home, attending college, entering full-time employment, entering a marriage or cohabitation, transitioning to parenthood) (Elder 1995; Elder, Johnson, and Crosnoe 2003; Shanahan 2000). These are periods when behavioral and health trajectories gain momentum and inequalities become entrenched. Social Causation, Health Selection, and Mutual Interdependence Physical health and socioeconomic status are tightly linked across the life course. Social causation suggests that early and sustained social and economic hardship leads to self-reinforcing cycles of poor health, including health problems such as obesity (Ferraro and Kelley-Moore 2003), hypertension, heart disease (Dupre 2007, 2008), cognitive decline (Rodgers, Ofstedal, and Herzog 2003), and higher rates of disability and mortality in older adulthood (Lynch, Kaplan, and Shema 1997; Smith and Kington 1997). Health selection posits that poor health decreases labor force participation (for a review, see Currie and Madrian [1999], in turn reducing financial capital and socioeconomic status (Haas 2006). These interrelated processes (social causation and health selection) can induce upward or downward mobility across socioeconomic strata. Past studies have focused on which mechanism dominates. The bulk of evidence and argument suggests social causation (Chandola et al. 2003; Elstad and Krokstad 2003; Mulatu and Schooler 2002), which resonates with Link and Phelan’s (1995) argument that social conditions determine access to health-related resources, thereby producing social inequalities in health. Lower levels of human or financial capital lead to lower access to, or utilization of, healthcare (Ross and Wu 1996), and are associated with lower levels of social support (Thoits 1995) and greater exposure to cumulative stress (McEwen 1998). Social causation processes are also supported by Mirowsky and Ross’s (2003) notion of education as learned effectiveness: education improves health by increasing self-efficacy and problem-solving capacity. Others (Fox 1990; Haas 2006) have argued that both processes are at work (particularly in early life). However, other scholars place emphasis not on social causation or health selection, but rather on the mutual interdependence between health and human capital through which (dis)advantages accumulate (O’Rand 2001). Conceptualizing Pathways of Physical Health and Human Capital across the Transition to Adulthood We now introduce our conceptualization of physical health and human capital across the transition to adulthood. Physical health is a complex construct (Sartorius 2006); thus, multiple indicators are useful and their selection depends on life stage. But generally, key indicators include physical activity (or activities of daily living, among older adults), body composition (e.g., adult body mass index or nutritional intake; infant birth weight), and/or health conditions (e.g., self-reported health, medical examinations, presence of chronic health conditions) (Bircher 2005; Huber et al. 2011). We strive for parsimony and replicability: we choose one indicator with face validity from each domain—level of physical activity, body mass index, and self-reported health. For human capital, we follow past theoretical work in economics that defines investment in human capital as “activities that influence future real income through the imbedding of resources in people” (Becker [1962], p. 9). Thus, we conceptualize human capital as one’s potential for income and seek to identify activities that embed resources within individuals. Of course, these activities will differ by life stage as the key institutions with which individuals interact change dramatically across the transition to adulthood. During adolescence, key resource-embedding activities relate to education. Schools are crucial in the production of training; they transmit knowledge, skills, and resources to youth that influence their potential income (Becker 1962; Coleman 1988). The success of this transmission can be captured, at least in part, by markers of aptitude (e.g., cognitive ability) and school performance (e.g., assigned grades, being on track with age-related peers). These markers are also linked with college attendance—an important factor stratifying future employment opportunities and thereby future income. Pathways through the transition to adulthood are less structured now than in the past (Furstenberg, Rumbaut, and Settersten 2005; Hogan and Astone 1986; Shanahan 2000). Still, most individuals enroll in college or enter the labor market in at least a part-time capacity during the transition to adulthood (Mouw 2005; Sandefur, Eggerling-Boeck, and Park 2005)—decisions that are likely informed by youths’ structural advantage, school performance, and anticipated socioeconomic gains of employment (Mortimer 2003; Mortimer, Staff, and Lee 2005). Both college and early full-time employment involve resource-embedding activities such as gaining credentials or employment. Conversely, remaining idle during this time is defined by exclusion—not working and not attending school (Shanahan 2000). More than a third of idle youth suffer from cognitive impairments (Amato et al. 2008), but idleness outside such impairment can be problematic, as it restricts youths’ investment in future income. Additionally, financial decisions in young adulthood are influential determinants of net worth later in life. Individuals holding low-risk financial assets (e.g., a savings or checking account) in early adulthood accumulate more assets by mid- to late adulthood compared with those who are permanently asset-poor—meaning, never holding a checking account, savings account, bonds or stock, or never owning a home (Keister 2003). More generally, holding such assets within the transition to adulthood likely signals a foundation for longer-term advantageous behaviors (Macmillan 2006; Shanahan 2000; Zimmer-Gembeck and Mortimer 2006). In sum, activities reflecting potential for income during the transition to adulthood include education and employment (versus being idle), cognitive ability, and financial asset activity. By young adulthood, many individuals have exited schooling and have entered the labor force (or engaged in caretaking) (Mouw 2005; Sandefur, Eggerling-Boeck, and Park 2005). At this stage, we conceptualize human capital as past literature focusing on mid- to late adulthood—via employment behavior, educational attainment, and socioeconomic position, whether objective or subjective (Teachman, Paasch, and Carver 1997). Current Study This is the first study to prospectively estimate the development of physical health and human capital pathways from adolescence into young adulthood for a contemporary, nationally representative young adult cohort (Add Health). We use maximum likelihood estimation (MLE) for a structural, multi-equation model to estimate causal pathways of health and human capital development and their cross-lagged effects on young adult cardiovascular health. Throughout, we employ multiple measures of health and human capital at each time point to track developmental change. Our study addresses three questions. First, how does physical health and human capital develop within individuals across the transition to adulthood? Does one life stage seem to be more critical than another? Second, is there evidence of persistent social inequality in adult health, above and beyond underlying selection mechanisms? Third, through what pathways do young adult health inequalities manifest? Data, Measures, and Analytic Strategy Data We use data from the National Longitudinal Study of Adolescent to Adult Health (Add Health), a school-based, nationally representative sample of 20,745 seventh to twelfth graders in 1994–1995. Respondents were re-interviewed in 1996 (Wave II), 2001–2002 (Wave III), and 2008–2009 (Wave IV) (Harris 2010; Harris et al. 2009). Our sample includes male and (non-pregnant) female respondents participating in Waves I, III, and IV (as anthropometric measures of health are not comparable across pregnant and non-pregnant women; n = 12,311). We select native-born respondents (n = 11,499) based on complex differences in health status for foreign-born versus native-born individuals (Hummer et al. 1999; Singh and Miller 2003), particularly in early life (Crosnoe 2006). Respondents with a valid sampling weight are retained to produce nationally representative estimates (n = 9,033). Missing data is minimal (<4 percent for most variables). We use a single imputation procedure in Stata to replace missing data on all independent variables. Results based on listwise deletion and full information maximum likelihood (in Mplus) were largely unchanged. Analyses are weighted and stratified to account for the sampling design. Measures As depicted in figure 1, we constructed a latent variable reflecting poor health for each life stage represented in our data (Poor Health, WI, adolescence; Poor Health WIII, emerging adulthood; and Poor Health WIV, young adulthood) using the same three variables: self-reported health (1 = excellent, 4 = fair or poor), body mass index (range = 11.21 – 80.40),1 and physical inactivity (bouts of physical inactivity in the last 7 days; range [WI] = 0–3; range [Waves III and IV] = 0–5). Figure 1. View largeDownload slide Measurement model of the accumulation of poor health and human capital across the transition to adulthood Note: Path dependence is indicated by paths linking poor health over time and paths linking human capital over time. Social causation is indicated by paths linking human capital to poor health. Health selection is indicated by paths linking poor health to human capital. Social inequality in health net of health selection is indicated by the path linking WIV human capital to metabolic syndrome. Figure 1. View largeDownload slide Measurement model of the accumulation of poor health and human capital across the transition to adulthood Note: Path dependence is indicated by paths linking poor health over time and paths linking human capital over time. Social causation is indicated by paths linking human capital to poor health. Health selection is indicated by paths linking poor health to human capital. Social inequality in health net of health selection is indicated by the path linking WIV human capital to metabolic syndrome. Likewise, we constructed a latent variable reflecting human capital in each life stage (see figure 1) using age- and developmental stage-appropriate measures. In adolescence, human capital is indicated by GPA (range = 1–4), Add Health Picture Vocabulary Test (AH PVT) score—a measure of cognitive ability (range = 14–146), and if the respondent was ever held back a grade (1 = not retained). In emerging adulthood, human capital is indicated by years of schooling (range = 6–22), AH PVT score (range = 7–122), if the respondent was employed or in school (1 = yes), and if the respondent had a checking and/or savings account (1 = yes). In young adulthood, years of schooling (range = 8–26), household income (range = $0–$150,000), employment status (1 = employed), and self-rated socioeconomic status (range = 1–10) characterize human capital.2 Similar to the clinical definition of metabolic syndrome (Grundy et al. 2005),3 a latent variable reflecting metabolic syndrome (shown in figure 1) was constructed using five indicators: elevated waist circumference (≥ 88 cm for women and 102 cm for men), elevated blood pressure (≥ 130 mm Hg systolic blood pressure, ≥ 85 mm Hg diastolic blood pressure, or antihypertensive drug treatment), elevated triglycerides (membership in highest decile of measured triglycerides), reduced high-density lipoprotein cholesterol (HDL-C) (membership in lowest 2 deciles of HDL-C, based on sex), and pre-diabetic value of glycosylated hemoglobin (HbA1c) (> 5.6 percent of hemoglobin molecules are glycosylated). WI control variables include family structure (1 = two biological parent family, 0 = all other family structures), parent education (range = 0–9),4 (logged) household income-needs ratio, race-ethnicity, age, and respondent sex (1 = female). Additionally, we control for WIII residence in urban/non-urban census tract and WIV household composition (reside with parents, married or cohabiting, live alone). Consistent with past research linking structural and economic features of the neighborhood environment to individuals’ health and human capital (e.g., Kawachi and Berkman [2003]), we include multiple, contextual, time-varying instrumental variables for both: subjective indicators of neighborhood safety (if s/he knows most people in their neighborhood, stopped to talk to any neighbors in the past month, usually feels safe in his/her neighborhood), an objective indicator of neighborhood safety (juvenile violent crime rate), indicators of the economic landscape (Bureau of Labor Statistics cost price index,5 per capita income, male unemployment rate, percent of female-headed households with children), and indicators of housing quality (percent of homes with insufficient plumbing, median year housing was built—before or after 1978). Additional instrumental variables of poor health include (Euclidean) distance between the respondent’s home and nearest park, age-specific mortality rate of 15–24-year-olds, and the number of inactivity resources (i.e., movie theaters, arcades) within 1 km of their home. In order to be plausible instruments, these factors must be correlated with health (or human capital) at a given time, but may not predict future human capital or health except through the lagged value. The proposed instruments are plausible on conceptual grounds. Perceived neighborhood safety or Euclidian distance to the nearest park, for example, can regulate one’s ability to engage in physical activity locally, with important implications for overweight status (Gordon-Larsen et al. 2006); however, these factors would not directly impact time2 physical health, except through time1 health behaviors and/or weight status. We also tested the assumptions of instrumental variables. Results based on ivregress post-estimation commands reject the null hypothesis of exogeneity (Wu-Hausman test: p < 0.001 for each instance), while the identification tests fail to reject the null hypothesis of correlation between the error terms in the structural equations and the instrumental variables, thus indicating that the exclusion restrictions are valid (WIII health: χ2 = 6.11, p = 0.41; WIV health: χ2 = 1.18, p = 0.76; WIII human capital: χ2 = 4.24, p = 0.52; WIV human capital: χ2 = 6.19, p = 0.11) (Sargan 1958). Furthermore, the instruments are sufficient in strength (WIII health: F = 21.14; 10 percent value for 2SLS = 10.27; WIV health: F = 10.71; 10 percent value for 2SLS = 9.08; WIII human capital: F = 34.42; 10 percent value for 2SLS = 10.27; WIV human capital: F = 95.12; 10 percent value for 2SLS = 9.08) (Stock, Wright, and Yogo 2002). Analytic Strategy We employ an autoregressive structural equation model (SEM) to estimate pathways of physical health and human capital and cross-lagged effects; we also allow each pathway to have a direct effect on metabolic syndrome. From this base model, we relax a series of parameter constraints and compare five nested models to select a preferred model (see table 1). For example, in the first column, the likelihood ratio test compares model 1 (including none of the instruments, error covariances for latent variables, or cross-lags) to model 2 (including instruments, but not error covariances or cross-lags); the statistical significance of the likelihood ratio test statistic indicates that model 2 is a better fit of the data. This was repeated until all parameter constraints noted were relaxed. The final column indicates that model 5 was the best fit overall. Thus, our preferred model includes cross-lags and error covariances across the system of equations jointly along with model parameters, and uses instrumental variables within the SEM to statistically identify causal effects for health and human capital at each life course stage. This strategy produces consistent estimates of the effect of health (or human capital) at one point in time on health (or human capital) at a subsequent time point, and effectively allows us to address the endogeneity of both health and human capital across time. Table 1. Results from a Nested Model Comparison Test to Select a Preferred Model Model 1: Covariances = 0, cross-lags = 0 Model 2: Covariances = 0, cross-lags = 0, + instruments Model 3: Covariances ≠ 0, cross-lags = 0, + instruments Model 4: Covariances = 0, cross-lags ≠ 0, + instruments Model 5: Covariances ≠ 0, cross-lags ≠ 0, + instruments    Model 1 vs. Model 2  Model 2 vs. Model 3  Model 2 vs. Model 4  Model 2 vs. Model 5  LLx  94,500.11  94,148.71  94,148.71  94,148.71  DFx  76  105  105  105  LLx+1  94,148.71  93,699.98  93,992.82  93,558.29  DFx+1  105  111  109  115  Difference in deviance statistic  351.40  448.73  155.89  590.42  Difference in df  29  6  4  10  LR test  0.000  0.000  0.000  0.000  Best fitting model  Model 2  Model 3  Model 4  Model 5  Model 1: Covariances = 0, cross-lags = 0 Model 2: Covariances = 0, cross-lags = 0, + instruments Model 3: Covariances ≠ 0, cross-lags = 0, + instruments Model 4: Covariances = 0, cross-lags ≠ 0, + instruments Model 5: Covariances ≠ 0, cross-lags ≠ 0, + instruments    Model 1 vs. Model 2  Model 2 vs. Model 3  Model 2 vs. Model 4  Model 2 vs. Model 5  LLx  94,500.11  94,148.71  94,148.71  94,148.71  DFx  76  105  105  105  LLx+1  94,148.71  93,699.98  93,992.82  93,558.29  DFx+1  105  111  109  115  Difference in deviance statistic  351.40  448.73  155.89  590.42  Difference in df  29  6  4  10  LR test  0.000  0.000  0.000  0.000  Best fitting model  Model 2  Model 3  Model 4  Model 5  We estimate this preferred model using maximum likelihood estimation in Mplus 7. MLE produces efficient parameter estimates, assuming error terms follow a multivariate normal distribution. This is a tenuous assumption, however, particularly when the true distribution is non-normal (Guilkey and Lance 2014; Mroz 1999), although very few studies compare how their parameter estimates differ with and without the normality assumption. We assess the extent to which our results are robust to this assumption by comparing MLE results to those produced by a two-stage least squares instrumental variables regression model, estimated in Stata 13.0 using the ivregress command—a statistical specification that does not assume normality but is less efficient. If model results are similar across both specifications, this would suggest that joint normality is not violated and the more efficient estimation (MLE) is preferred. If model results are not similar, joint normality may be violated and the IV results are preferred. We then estimate total, direct, and indirect effects to identify pathways through which health and human capital operate directly and indirectly on metabolic syndrome. Finally, we assess whether or not the endogeneity corrections we employ are necessary. Because the system of equations in our model is jointly estimated using full information MLE, this approach requires IVs for statistical identification. Therefore, to assess whether or not correcting for endogeneity makes a difference, we calculate OLS regression estimates, where we adjust for controls, lagged values of health and human capital, and cross-lagged values, but not for IVs or correlated error terms between latent variables. We refer to this as the uncorrected model. Results Table 2 presents descriptive statistics of all study variables, weighted and adjusted for clustering of the sample design. The mean and variance of self-reported health remained somewhat stable over time, but rose steadily for BMI. Physical inactivity was lowest during high school, increasing within emerging and young adulthood. Average GPA in the current or past academic year was 2.79; 79 percent of respondents never repeated a grade. Average AHPVT score remained similar over time. Eighty-three percent of respondents were not idle in emerging adulthood; many had a checking (71 percent) or savings (63 percent) account. In this life stage, the average level of completed education was 13.1 years; by young adulthood it was 14.2 years. In young adulthood, two-thirds (66 percent) were employed; average household income was $60,125. Over half of sample members were classified with elevated waist circumference (52 percent) and just under half with elevated blood pressure (42 percent). Nearly a third met or surpassed the pre-diabetic cutoff for glycosylated hemoglobin (31 percent). Table 2. Descriptive Statistics   Mean or percent  Std. dev.  Minimum  Maximum  Poor health—Adolescence (Wave I)  Self-reported health  2.14  0.90  1  4  BMI  22.56  4.46  11.21  49.12  Physical inactivity  1.04  0.85  0  3  Poor health—Emerging adulthood (Wave III)  Self-reported health  2.02  0.86  1  4  BMI  26.91  6.44  13.80  73.17  Physical inactivity  2.98  1.57  0  5  Poor health—Young adulthood (Wave IV)  Self-reported health  2.36  0.88  1  4  BMI  29.17  7.62  14.40  80.40  Physical inactivity  2.86  1.50  0  5  Human capital—Adolescence (Wave I)  GPA  2.79  0.72  1  4  Add Health Picture Vocabulary Test  102.47  13.36  14  146  No grade retention  78.96%    0  1  Human capital—Emerging adulthood (Wave III)  Years of completed schooling  13.05  1.97  6  22  Add Health Picture Vocabulary Test  100.56  14.12  7  122  Employed or in school  82.96%    0  1  Have checking account  71.40%    0  1  Have savings account  62.54%    0  1  Human capital—Young adulthood (Wave IV)  Years of completed schooling  14.23  2.59  8  26  Household income  60,125  37,057  0  150,000  Currently employed  66.49%    0  1  Self-rated socioeconomic status  4.95  1.74  1  10  Metabolic syndrome  Elevated waist circumference  51.80%    0  1  Elevated blood pressure  41.84%    0  1  Elevated triglycerides  9.89%    0  1  Reduced HDL cholesterol  16.25%    0  1  Pre-diabetic value, hemoglobin A1c  31.24%    0  1  Sociodemographic controls—Adolescence (Wave I)  Two-parent family structure  56.40%    0  1  Parental education  5.50  2.13  0  9  Household income–needs ratio  2.98  3.00  0.00  97.80   Race-ethnicity  Non-Hispanic White (reference)  72.14%    0  1  Non-Hispanic Black  14.75%    0  1  Hispanic  3.58%    0  1  Non-Hispanic other  9.52%    0  1  Age at Wave I  15.43  1.82  11  21  Sex (1 = female)  51.02%    0  1  Wave IV household compositiona  Live alone  11.44%    0  1  Live with parents  15.85%    0  1  Married or cohabiting  57.50%    0  1  Wave III urbanicity  Living in urban census tract  58.25%    0  1  Instrumental variables          Euclidean distance to nearest park (meters), WI  11,830.92  11,386.65  0  81,562.27  Know most people in neighborhood, WI  74.08%    0  1  Stopped to talk to neighbors in past month, WI  79.23%    0  1  Usually feel safe in neighborhood, WI  91.00%    0  1  Juvenile violent crime rate (per 100,000), WI  48.60  36.47  0  214.94  Bureau of Labor Statistics cost price index, WIII  1.78  0.01  1.75  1.80  Inactivity resources, count, WIII  0.18  0.75  0  26  Median year housing built was before 1978, WIII  74.40%    0  1  Per capita income, WIII  19,493  8,095  2,700  108,600  Age-specific mortality rate (per 1,000), WIII  0.83  0.37  0  3.48  Juvenile violent crime rate (per 100,000), WIV  27.68  21.43  0  90.00  Male unemployment rate, WIV  0.08  0.06  0  0.68  Percent female-headed households with children, WIV  7.90%  5.28  0  49.08  Percent of homes with insufficient plumbing, WIV  1.85%  3.04  0  31.23    Mean or percent  Std. dev.  Minimum  Maximum  Poor health—Adolescence (Wave I)  Self-reported health  2.14  0.90  1  4  BMI  22.56  4.46  11.21  49.12  Physical inactivity  1.04  0.85  0  3  Poor health—Emerging adulthood (Wave III)  Self-reported health  2.02  0.86  1  4  BMI  26.91  6.44  13.80  73.17  Physical inactivity  2.98  1.57  0  5  Poor health—Young adulthood (Wave IV)  Self-reported health  2.36  0.88  1  4  BMI  29.17  7.62  14.40  80.40  Physical inactivity  2.86  1.50  0  5  Human capital—Adolescence (Wave I)  GPA  2.79  0.72  1  4  Add Health Picture Vocabulary Test  102.47  13.36  14  146  No grade retention  78.96%    0  1  Human capital—Emerging adulthood (Wave III)  Years of completed schooling  13.05  1.97  6  22  Add Health Picture Vocabulary Test  100.56  14.12  7  122  Employed or in school  82.96%    0  1  Have checking account  71.40%    0  1  Have savings account  62.54%    0  1  Human capital—Young adulthood (Wave IV)  Years of completed schooling  14.23  2.59  8  26  Household income  60,125  37,057  0  150,000  Currently employed  66.49%    0  1  Self-rated socioeconomic status  4.95  1.74  1  10  Metabolic syndrome  Elevated waist circumference  51.80%    0  1  Elevated blood pressure  41.84%    0  1  Elevated triglycerides  9.89%    0  1  Reduced HDL cholesterol  16.25%    0  1  Pre-diabetic value, hemoglobin A1c  31.24%    0  1  Sociodemographic controls—Adolescence (Wave I)  Two-parent family structure  56.40%    0  1  Parental education  5.50  2.13  0  9  Household income–needs ratio  2.98  3.00  0.00  97.80   Race-ethnicity  Non-Hispanic White (reference)  72.14%    0  1  Non-Hispanic Black  14.75%    0  1  Hispanic  3.58%    0  1  Non-Hispanic other  9.52%    0  1  Age at Wave I  15.43  1.82  11  21  Sex (1 = female)  51.02%    0  1  Wave IV household compositiona  Live alone  11.44%    0  1  Live with parents  15.85%    0  1  Married or cohabiting  57.50%    0  1  Wave III urbanicity  Living in urban census tract  58.25%    0  1  Instrumental variables          Euclidean distance to nearest park (meters), WI  11,830.92  11,386.65  0  81,562.27  Know most people in neighborhood, WI  74.08%    0  1  Stopped to talk to neighbors in past month, WI  79.23%    0  1  Usually feel safe in neighborhood, WI  91.00%    0  1  Juvenile violent crime rate (per 100,000), WI  48.60  36.47  0  214.94  Bureau of Labor Statistics cost price index, WIII  1.78  0.01  1.75  1.80  Inactivity resources, count, WIII  0.18  0.75  0  26  Median year housing built was before 1978, WIII  74.40%    0  1  Per capita income, WIII  19,493  8,095  2,700  108,600  Age-specific mortality rate (per 1,000), WIII  0.83  0.37  0  3.48  Juvenile violent crime rate (per 100,000), WIV  27.68  21.43  0  90.00  Male unemployment rate, WIV  0.08  0.06  0  0.68  Percent female-headed households with children, WIV  7.90%  5.28  0  49.08  Percent of homes with insufficient plumbing, WIV  1.85%  3.04  0  31.23  Note: Sample size is 9,033 men and women. Statistics are weighted and adjusted for clustering. aPercentages do not total 100, as these are not mutually exclusive categories of the same variable. Age-specific mortality rate indicates a smoothed rate across 1999–2001 of the number of deaths among persons age 15–24 per 1,000 residents at the county level. Source: National Longitudinal Study of Adolescent Health. Selected parameters from the SEM are presented in figure 2; all parameters are presented in appendix A.6 Unstandardized (not italicized) and standardized (italicized) beta coefficients are presented with standard errors in parentheses. Figure 2. View largeDownload slide Estimates from the structural equation model of the accumulation of poor health and human capital on metabolic syndrome Note:N = 9,033. Unstandardized (not italicized) and standardized (italicized) beta coefficients with standard errors in parentheses are presented. Covariances of error terms of all latent variables are included but not shown. Additional controls for human capital (WIV) not shown are: household composition (married/cohabiting, live alone, live with parents), urban residence. RMSEA = 0.026; SRMR = 0.050. Figure 2. View largeDownload slide Estimates from the structural equation model of the accumulation of poor health and human capital on metabolic syndrome Note:N = 9,033. Unstandardized (not italicized) and standardized (italicized) beta coefficients with standard errors in parentheses are presented. Covariances of error terms of all latent variables are included but not shown. Additional controls for human capital (WIV) not shown are: household composition (married/cohabiting, live alone, live with parents), urban residence. RMSEA = 0.026; SRMR = 0.050. Our first key question is: how does physical health and human capital develop within individuals across the transition to adulthood? Results demonstrate strong path dependence: all path coefficients linking poor health over time and those linking human capital over time are positive and statistically significant (p < 0.001). Comparing the standardized coefficients, first for poor health, shows that the path coefficient linking poor health in WI and WIII (standardized b = 0.92) is more than twice as large as the path coefficient linking poor health in WIII and WIV (standardized b = 0.40). Thus, path dependence appears to be stronger during later (versus earlier) stages of the transition to adulthood. Results from a Wald test suggest that these two path coefficients cannot be constrained to be equal (PH1 → PH3 ≠ PH3 → PH4; test statistic = 24.46, df = 1, p < 0.001). Substantively, this indicates that poor health appears to develop at a differential rate, rather than a constant rate, as individuals transition into adulthood. Estimates of path dependence for human capital present a different pattern. The standardized path coefficient linking WI and WIII human capital (standardized b = 1.01) is slightly larger than the corollary path coefficient between WIII and WIV (standardized b = 0.99). A Wald test suggests the null hypothesis (HC1 → HC3 = HC3 → HC4) can be rejected (160.47, df = 1, p < 0.001), indicating that the human capital pathway crystalizes quickly in adolescence and is further solidified in young adulthood. Next, we address our second research question: is there evidence of persistent social inequality in adult health, above and beyond underlying selection mechanisms? The key estimate is the effect of young adult human capital on metabolic syndrome. This path coefficient (unstandardized b = −0.02, p < 0.001) indicates that higher levels of young adult human capital are associated with lower risk of metabolic syndrome, net of all other pathways in the model, including the endogeneity corrections for underlying social selection and health selection mechanisms. Thus, we observe salient social stratification of metabolic syndrome in this contemporary cohort. The statistical specification used to produce figure 2 assumed the scaling indicator (elevated waist circumference) for metabolic syndrome was continuous; therefore, the path coefficients linking young adult health and human capital to metabolic syndrome are approximated by a linear probability model. To examine if study results were sensitive to this specification, we replicated the SEM scaling metabolic syndrome by the linear form of waist circumference, triglycerides, and hemoglobin A1c (one at a time). Results presented in appendix B suggest that study findings are extremely robust across different specifications of metabolic syndrome. We also examined the extent to which the normality assumption imposed by MLE was upheld. Results (presented in appendix C) indicate that a two-stage least squares instrumental variable regression specification produces similar, and at times identical, point estimates compared to the MLE. The standard errors produced in the IV model are wider, reflecting the decreased efficiency of this approach relative to MLE. The similarity across specifications provides strong evidence that the normality assumption imposed by MLE is not likely violated in this model. Therefore, the MLE results are preferred due to increased efficiency. We now return to the full model to examine our final research question: through what pathways do young adult health inequalities manifest? We address this in two ways: we examine the cross-lags and then assess total, direct, and indirect effects. As depicted in figure 2, cross-lags linking factors between WI and WIII are not statistically significant. However, the cross-lags between WIII and WIV are statistically significant at the p < 0.001 level. Both are negative, indicating that poor health and human capital are inversely related. The standardized coefficient for the path reflecting social causation is –0.04; that reflecting health selection is –0.05. Results from a Wald test suggest the null hypothesis (HC3 → PH4 = PH3 → HC4) can be rejected (7.161, df = 1, p = 0.008). Thus, we conclude that both social causation and health selection are present during this life stage and appear to differ modestly in strength. Table 3 presents a summary of total, direct, and indirect effects (poor health in panel A, human capital in panel B). The first coefficient shown represents the direct effect of WIV poor health on metabolic syndrome (unstandardized b = 0.42); this also appears in figure 2. The second coefficient represents the total effect of WIII poor health on metabolic syndrome (b = 0.45). (With no direct effect here, the sum of the indirect effects is equal to the total effect.) The two indirect effects are Health3 → Health4 → Metabolic syndrome (b = 0.45), and Health3 → Human Capital4 → Metabolic syndrome (b = 0.002).7 Even though unstandardized betas are presented, the difference exhibited makes the relative association fairly clear. The last set of results in panel A mirror this finding: the total effect of WI poor health on metabolic syndrome (unstandardized b = 0.26) is composed of four indirect effects; the one largest in magnitude does not involve a cross-lag (Health1 → Health3 → Health4 → Metabolic syndrome; b = 0.26). Panel B depicts a very similar pattern of results for human capital. Thus, overall the key causal pathways reflect path dependence, a finding to which we return in the discussion below. Table 3. Total, Direct, and Indirect Effects of Poor Health and Human Capital on Metabolic Syndrome Panel A. Effects of poor health at Waves I, III, and IV  Unstandardized β (SE)  β/SE  Poor health (WIV) → Metabolic syndrome       Total effect = Direct effect  0.42 (0.01)***  35.42  Poor health (WIII) → Metabolic syndrome       Total effect = Indirect effect  0.45 (0.01)***  31.75  Specific indirect paths      Health3 → Health4 → Metabolic syndrome  0.45 (0.01)***  31.89  Health3 → Human capital4 → Metabolic syndrome  0.002 (0.001)*  2.46  Poor health (WI) → Metabolic syndrome      Total effect = Indirect effect  0.26 (0.04)***  6.26  Specific indirect paths      Health1 → Health3 → Health4 → Metabolic syndrome  0.26 (0.04)***  6.23  Health1 → Health3 → Human capital4 → Metabolic syndrome  0.001 (0.001)*  2.37  Health1 → Human capital3 → Health4 → Metabolic syndrome  0.000 (0.000)  1.16  Health1 → Human capital3 → Human capital4 → Metabolic syndrome  0.001 (0.001)  1.13  Panel B. Effects of human capital at Waves I, II, and IV      Human capital (WIV) → Metabolic syndrome       Total effect = Direct effect  −0.02 (0.005)***  −3.76  Human capital (WIII) → Metabolic syndrome       Total effect = Indirect effect  −0.03 (0.006)***  −4.83  Specific indirect paths      Human capital3→ Human capital4 → Metabolic syndrome  −0.02 (0.006)***  −3.61  Human capital3 → Health4 → Metabolic syndrome  −0.01 (0.002)***  −3.84  Human capital (WI) → Metabolic syndrome      Total effect = Indirect effect  −0.12 (0.03)***  −4.57  Specific indirect paths      Human capital1 → Human capital3 → Human capital4 → Metabolic syndrome  −0.08 (0.02)***  −3.59  Human capital1→ Human capital3 → Health4 → Metabolic syndrome  −0.03 (0.01)***  −3.81  Human capital1→ Health3 → Human capital4 → Metabolic syndrome  −0.00 (0.00)  −0.26  Human capital1 → Health3 → Health4 → Metabolic syndrome  −0.01 (0.02)  −0.26  Panel A. Effects of poor health at Waves I, III, and IV  Unstandardized β (SE)  β/SE  Poor health (WIV) → Metabolic syndrome       Total effect = Direct effect  0.42 (0.01)***  35.42  Poor health (WIII) → Metabolic syndrome       Total effect = Indirect effect  0.45 (0.01)***  31.75  Specific indirect paths      Health3 → Health4 → Metabolic syndrome  0.45 (0.01)***  31.89  Health3 → Human capital4 → Metabolic syndrome  0.002 (0.001)*  2.46  Poor health (WI) → Metabolic syndrome      Total effect = Indirect effect  0.26 (0.04)***  6.26  Specific indirect paths      Health1 → Health3 → Health4 → Metabolic syndrome  0.26 (0.04)***  6.23  Health1 → Health3 → Human capital4 → Metabolic syndrome  0.001 (0.001)*  2.37  Health1 → Human capital3 → Health4 → Metabolic syndrome  0.000 (0.000)  1.16  Health1 → Human capital3 → Human capital4 → Metabolic syndrome  0.001 (0.001)  1.13  Panel B. Effects of human capital at Waves I, II, and IV      Human capital (WIV) → Metabolic syndrome       Total effect = Direct effect  −0.02 (0.005)***  −3.76  Human capital (WIII) → Metabolic syndrome       Total effect = Indirect effect  −0.03 (0.006)***  −4.83  Specific indirect paths      Human capital3→ Human capital4 → Metabolic syndrome  −0.02 (0.006)***  −3.61  Human capital3 → Health4 → Metabolic syndrome  −0.01 (0.002)***  −3.84  Human capital (WI) → Metabolic syndrome      Total effect = Indirect effect  −0.12 (0.03)***  −4.57  Specific indirect paths      Human capital1 → Human capital3 → Human capital4 → Metabolic syndrome  −0.08 (0.02)***  −3.59  Human capital1→ Human capital3 → Health4 → Metabolic syndrome  −0.03 (0.01)***  −3.81  Human capital1→ Health3 → Human capital4 → Metabolic syndrome  −0.00 (0.00)  −0.26  Human capital1 → Health3 → Health4 → Metabolic syndrome  −0.01 (0.02)  −0.26  Finally, the supplementary analyses exploring the necessity of correcting for endogeneity show a markedly different pattern of results (see appendix D). Relative to the SEM, the uncorrected OLS model tends to underestimate the path coefficients reflecting path dependence and social inequality in health (net of health selection). In most cases, the uncorrected model estimate is roughly half the magnitude of the SEM estimate. Moreover, the uncorrected model overestimates the cross-lags (in most cases, by more than double) and all four cross-lags are statistically significant—suggesting a starkly different substantive result. In other words, had we not corrected for endogeneity, we would have mistakenly placed less emphasis on the role of path dependence, more emphasis on the role of social causation and health selection, and we would have inaccurately concluded that social causation and health selection were operating throughout each stage across the transition to adulthood. Thus, it is necessary, on both statistical and theoretical grounds, to account for the endogeneity of poor health and human capital and include error covariances. Discussion A vast literature documents social inequalities in health and human capital, but less is known about how these inequalities develop across the early life course. We trace the intra-generational development of these emerging pathways as adolescents transition into adulthood. In addition, little is known about the interdependence of human capital and poor physical health during the transition to adulthood. As we detail below, understanding more about inequality in both domains informs a range of sociology subdisciplines, including social stratification, medical sociology, and health inequalities. Furthermore, no research has yet examined the contributions of intra-generational development of poor health and accumulation of human capital on metabolic syndrome—an important marker for cardiovascular disease and Type II diabetes, two conditions that predict significant morbidity and mortality risks for US adults. A primary contribution of our study is the conceptualization and operationalization of pathways of health and human capital across the transition to adulthood. An empirical test of these pathways showed powerful evidence of path dependence: poor health in adolescence is strongly related to poor health in emerging adulthood, which in turn is related to poor health in young adulthood. The same holds for human capital. This in and of itself is not surprising, but results also show that these pathways develop at differential rates over the transition to adulthood. The health pathway crystalizes more slowly between adolescence (WI) and emerging adulthood (WIII), but more rapidly between emerging and young adulthood. Substantively, this suggests that adolescence is a life stage in which health behaviors, such as food choices and levels of physical activity, are still being formulated, whereas by emerging adulthood these behaviors are more entrenched—a finding consistent with past research (Harris et al. 2006). Conversely, the human capital pathway crystalizes strongly in adolescence and solidifies further in adulthood. This provides evidence of the notion proposed by life course scholars, that human agency, along with structural distributions of opportunity, together play an important role in the transition to adulthood as individuals enact human agency (e.g., open a savings/checking account) and begin to formulate health behaviors, both of which can become the basis of longer-term health and economic behaviors and outcomes in adulthood (Hitlin and Kirkpatrick Johnson 2015; Macmillan 2006; Shanahan 2000; Zimmer-Gembeck and Mortimer 2006). Additionally, this finding buttresses support for Heckman’s (2000) call for greater emphasis to be placed on early intervention programs and policies, as human capital pathways are taking shape well before our study first measures them. Moreover, supplementary analyses showed that failing to account for endogeneity would have led us to inaccurately place less emphasis on path dependency within the model, as estimates from uncorrected OLS models were roughly half the magnitude of SEM estimates (that included endogeneity corrections and error covariances). A second key contribution is strong evidence of emerging social stratification in metabolic syndrome for this cohort. Our study extends past research by showing that the estimated effect of young adult human capital on metabolic syndrome is statistically significant (p < 0.001), and robust to the inclusion of parental socioeconomic status, family background characteristics, and underlying social and health selection mechanisms (which we account for using numerous instrumental variables identifying poor health and human capital at each time point). We speculate that the social stratification of metabolic syndrome signals that the interplay between opportunity structures and human agency is already taking hold in the lives of this cohort, influencing a key marker of adult health. It may also reflect learned effectiveness—in that education, even in young adulthood, provides individuals with knowledge and skills that marshal personal control and enact human agency to produce healthy behaviors and a healthy lifestyle (Mirowsky and Ross 2003). Similarly, higher education provides access to advantageous employment opportunities, relative to low-wage work or unemployment, that accommodate/facilitate healthier lifestyles (Ross and Wu 1996). Based on past studies describing rather large social inequalities in health, we anticipated finding a substantial direct effect of human capital on metabolic syndrome. Instead, we observed a relatively small but statistically significant effect (p < 0.001); we speculate that this stems from two sources. Metabolic syndrome is relatively rare early in life, becoming increasingly common with age (prevalence is 3 percent in children [Friend, Craig, and Turner 2013], 23 percent in adults age 19+ [Beltrán-Sánchez et al. 2013], and 44 percent in adults age 50+ [Alexander et al. 2003]). Therefore, detecting any degree of social stratification in metabolic syndrome within young adulthood, above and beyond endogeneity and selection bias, is a strong signal of things to come. Second, we measured metabolic syndrome objectively, using biomarker data, which are independent of self-reflection bias or social circumstances that are often correlated with subjective measures of health. For this reason, estimated effects of objective health indicators are often smaller in magnitude than those based on subjective indicators (Wu et al. 2013). We suspect that the social stratification we document will only increase as cohort members age. Specifically, mid- to late adulthood is a time when morbidity and mortality risks due to health conditions linked with metabolic syndrome, such as cardiovascular disease, stroke, and Type II diabetes, become more pronounced. It is also a time when these conditions become increasingly concentrated among individuals within lower socioeconomic strata. This is influential, as cardiovascular diseases are among the most expensive health conditions: one study estimated that 17 percent of all medical expenditures each year (totaling $149 billion) can be attributed to cardiovascular diseases alone (Trogdon et al. 2007). As these costs place an undue economic burden on the least advantaged, this will further stratify the health and wealth of this cohort as they age. Recent trends related to the prevalence of metabolic syndrome also suggest increased concentration among minority populations, and particularly among minority women (Beltrán-Sánchez et al. 2013; Mozumdar and Liguori 2011). As such, these trends may reinforce race/ethnic social stratification of adult health. Furthermore, given recent trends pointing to an earlier onset of chronic diseases, such as obesity (Van Cleave, Gortmaker, and Perrin 2010), and rising prevalence of metabolic syndrome across the population (Mozumdar and Liguori 2011), the long-term implications of social inequalities in cardiovascular disease may far exceed what we have observed in the past. Observing these developmental trends now, among a contemporary population at the forefront of the obesity epidemic (Harris 2010), affords time to plan and implement prevention and intervention efforts among vulnerable populations. Our work suggests that interventions would be more impactful during these critical early life stages if they can curb metabolic dysregulation before it manifests in more harmful conditions. The strength of the health and human capital pathways, and the importance of early and sustained investments in human capital, permeate our results. But we also show important (both substantively and statistically) cross-lagged effects between poor health and human capital between WIII and WIV (as respondents move into young adulthood). Thus, our study is the first to pinpoint the stage within the transition to adulthood in which social causation and health selection operate. This is the third contribution of our study. Supplementary analyses showed that failing to account for endogeneity would have led to markedly different substantive conclusions, including inaccurately concluding that social causation and health selection operate throughout the entire span of the transition to adulthood, buttressing the importance of correcting for endogeneity in terms of accurately depicting these cross-lagged associations. The standardized coefficients of the social causation (−0.04) and health selection (−0.05) cross-lags were statistically significant at the p < 0.01 level, and could not be constrained to be equal, according to a Wald test. In contrast, other studies demonstrate stronger effects of social causation for mental health in mid-adulthood (Chandola et al. 2003; Elstad and Krokstad 2003; Mulatu and Schooler 2002) and physical health in late adulthood (Warren 2009). Observing both social causation and health selection processes at work in our study is consistent with past scholarship emphasizing the mutual interdependence between the two that operates in conjunction with strong path-dependent components (O’Rand 2001), and resonates with a developmental perspective identifying the transition to adulthood as one in which youth create, form, and maintain social identities, health habits, and economic preferences simultaneously (Macmillan 2006). In this way, human capital and health in the early life course are substantively distinct from human capital and health in mid- to late adulthood—two life stages in which social causation and health selection signal a distinct underlying process. In mid- to late adulthood, poor health directly impacts labor market participation, earnings, and wealth accumulation (health selection). In turn, participation in the labor market affects health (social causation), partially through employment stability (ensuring stable earnings), access to health insurance through one’s job, or through education itself via learned effectiveness (Mirowsky and Ross 2003). In contrast, aptitude and educational achievement are the important drivers of human capital development in the early stage of the life course (as opposed to labor market activity and income) and this development is both dependent upon and predictive of early life health pathways. In sum, this study provides a general model that traces the development and persistence of health and human capital between adolescence and young adulthood. We focused on measures readily available in other datasets both within (e.g., NLSY97) and outside the United States (e.g., China Health and Nutrition Survey) to encourage future research. Future social stratification studies, for example, could trace the pathways through which ascribed characteristics (race-ethnicity, sex) shape inequalities at each time point, through constraints imposed by social structures and social systems. Other model extensions could chart variations in the accumulation of health and human capital by parental characteristics (education, income, occupation) or by interactions with social structures in childhood (family of origin, neighborhood, school), contributing a broader understanding of the development of the intergenerational transmissions of inequality. Exploring nuances in this model as individuals interact with health institutions or health professionals may contribute to the medical sociology literature. We chose to apply the model to understanding social inequalities in cardiovascular health, but future work could apply the model to any number of social well-being and/or health markers in young adulthood, contributing to a number of sociological subdisciplines. For instance, future research could also address how these pathways operate differentially across geographic location (contributing to urban and rural sociology), or how these pathways are disrupted by engagement in criminal activity (contributing to sociology of crime and deviance) or timing of family formation behavior (contributing to family sociology). We have in mind here the development of the Blau-Duncan model, where solving a measurement issue (operationalizing socioeconomic status) and laying out a causal framework facilitated a generation of comparable research findings. Extensions of this study’s framework across numerous subareas of sociology could enrich our understanding of the longer-term social processes associated with the development of physical health and human capital development. Notes 1 We top-coded weight for nine respondents whose weight exceeded the scale maximum (200 kg/440 lb). Results were unchanged when these individuals were excluded. 2 Pairwise correlations between latent variables:   1  2  3  4  5  6  1. Poor Health–WI  1            2. Poor Health–WIII  0.85  1          3. Poor Health–WIV  0.79  0.93  1        4. Human Capital–WI  −0.15  −0.11  −0.16  1      5. Human Capital–WIII  −0.08  −0.09  −0.14  0.91  1    6. Human Capital–WIV  −0.11  −0.13  −0.16  0.87  0.95  1    1  2  3  4  5  6  1. Poor Health–WI  1            2. Poor Health–WIII  0.85  1          3. Poor Health–WIV  0.79  0.93  1        4. Human Capital–WI  −0.15  −0.11  −0.16  1      5. Human Capital–WIII  −0.08  −0.09  −0.14  0.91  1    6. Human Capital–WIV  −0.11  −0.13  −0.16  0.87  0.95  1  3 Our measure diverges slightly from the clinical definition. Add Health participants were not asked to fast pre-interview. Rather than combine fasting and non-fasting glucose, we use HbA1c, a more stable measure of metabolic dysregulation. We also use membership in the highest decile for triglycerides; the clinical definition is indicated by ≥150 mg/dL or drug treatment for elevated triglycerides. For HDL-C, clinical cutoffs are <50mg/dL for women and <40 mg/dL for men, or drug treatment for reduced HDL-C. The laboratory assaying Add Health specimens for lipids used two different assays—a common occurrence. The two should be directly comparable using a simple algebraic transformation. After extensive data cleaning and checking efforts were performed, only the rank-ordering (by deciles) was released, as it was deemed to be a more reliable measure. Thus, we indicate reduced HDL-C as membership in the lowest category for women (HDL-C = 1; 7.0%) and the lowest two categories for men (HDL-C < 3; 26.2%), based on evidence suggesting that 11.9% of women and 31.4% of men exhibit reduced HDL-C (Carroll, Kit, and Lacher 2012). Results did not change when we defined reduced HDL-C by membership in the lowest two categories for women (HDL-C < 3; 15.6%) and the lowest three categories for men (HDL-C < 4; 37.2%). 4 When this was missing, respondent’s report of parental education was used. 5 The BLS cost price index is only available for urban areas; thus, we control for Wave III residence in an urban census tract. 6 Model fit indices presented indicate acceptable fit (RMSEA = 0.026; SRMR = 0.050). Other model fit indices (CFI, TLI) are below recommended levels. However, residual variances of latent variables are low (range, 0.01–0.17) and R-squared values of latent variables are high (all but one is between 0.50 and 0.97), lending further evidence of adequate model fit. In accordance with recommendations for testing structural equation models (Bollen and Long 1993), we balanced these indications of model fit with indications based on theory and substantive experience when developing our model. 7 Supplementary analyses (not shown) suggested that adding pathways from each WI latent variable to metabolic syndrome did not significantly improve model fit. About the Authors Jennifer B. Kane is an assistant professor in the Department of Sociology at the University of California–Irvine. Her research focuses on social inequality, its emergence over the life course, and the role of the family of origin and the early life environment in shaping social and health disparities. She has published recent work in Demography, Population Research and Policy Review, Journal of Health and Social Behavior, and Journal of Marriage and Family. Kathleen Mullan Harris is the James Haar Distinguished Professor of Sociology at the University of North Carolina–Chapel Hill. Harris’s research focuses on social inequality and health. Harris is Director and PI of the National Longitudinal Study of Adolescent to Adult Health, a longitudinal study of more than 20,000 teens who are being followed into young adulthood. She has published recent work in Demography, Nature, and Proceedings of the National Academy of Sciences. S. Philip Morgan is the Allan Feduccia Distinguished Professor of Sociology at the University of North Carolina–Chapel Hill. Morgan’s research focuses on change and variation in the human family, with special attention to human fertility. His work on fertility in the United States examines fertility levels, fertility timing, and high levels of nonmarital childbearing. Morgan has published recent work in Annual Review of Sociology, Demography, and Population and Development Review. David Guilkey is the Cary C. Boshamer Distinguished Professor of Economics at the University of North Carolina–Chapel Hill. The main focus of his interest is the development and use of estimation methods that can be used to analyze large-survey datasets with limited dependent variables, especially when endogenous right-hand-side variables are present. He has published recent work in Demography, Journal of Applied Econometrics, and Studies in Family Planning. Supplementary Material Supplementary material is available at Social Forces online. References Adler, Nancy, Nicole R. Bush, and Matthew S. Pantell. 2012. “ Rigor, Vigor, and the Study of Health Disparities.” Proceedings of the National Academy of Sciences  109( Supplement 2): 17154– 59. Google Scholar CrossRef Search ADS   Adler, Nancy E., and Joan M. Ostrove. 1999. “ Socioeconomic Status and Health: What We Know and What We Don’t.” Annals of the New York Academy of Sciences  896( 1): 3– 15. Google Scholar CrossRef Search ADS   Alexander, Charles M., Pamela B. Landsman, Steven M. Teutsch, and Steven M. Haffner. 2003. “ NCEP-Defined Metabolic Syndrome, Diabetes, and Prevalence of Coronary Heart Disease among NHANES III Participants Age 50 Years and Older.” Diabetes  52( 5): 1210– 14. Google Scholar CrossRef Search ADS   Almond, Douglas, and Janet Currie. 2011. “ Killing Me Softly: The Fetal Origins Hypothesis.” Journal of Economic Perspectives  25( 3): 153– 72. Google Scholar CrossRef Search ADS   Amato, Paul, Nancy S. Landale, Tara C. Havasevich-Brooks, Alan Booth, David J. Eggebeen, Robert Schoen, and Susan M. McHale. 2008. “ Precursors of Young Women’s Family Formation Pathways.” Journal of Marriage and Family  70( 5): 1271– 86. Google Scholar CrossRef Search ADS   Aneshensel, Carol S. 1992. “ Social Stress: Theory and Research.” Annual Review of Sociology  18: 15– 38. Google Scholar CrossRef Search ADS   Bauldry, Shawn, Michael J. Shanahan, Jason D. Boardman, Richard A. Miech, and Ross Macmillan. 2012. “ A Life Course Model of Self-Rated Health through Adolescence and Young Adulthood.” Social Science & Medicine  75( 7): 1311– 20. Google Scholar CrossRef Search ADS   Becker, Gary S. 1962. “ Investment in Human Capital: A Theoretical Analysis.” Journal of Political Economy  70( 5): 9– 49. Google Scholar CrossRef Search ADS   Beltrán-Sánchez, Hiram, Michael O. Harhay, Meera M. Harhay, and Sean McElligott. 2013. “ Prevalence and Trends of Metabolic Syndrome in the Adult US Population, 1999–2010.” Journal of the American College of Cardiology  62( 8): 697– 703. Google Scholar CrossRef Search ADS   Ben-Shlomo, Y. and D. Kuh. 2002. “ A Life Course Approach to Chronic Disease Epidemiology: Conceptual Models, Empirical Challenges and Interdisciplinary Perspectives.” International Journal of Epidemiology  31( 2): 285– 93. Google Scholar CrossRef Search ADS   Bircher, Johannes. 2005. “ Towards a Dynamic Definition of Health and Disease.” Medicine, Health Care, and Philosophy  8( 3): 335– 41. Google Scholar CrossRef Search ADS   Blackwell, Debra L., Mark D. Hayward, and Eileen M. Crimmins. 2001. “ Does Childhood Health Affect Chronic Morbidity in Later Life?” Social Science & Medicine  52( 8): 1269– 84. Google Scholar CrossRef Search ADS   Blau, Peter M., and Otis Dudley Duncan. 1967. The American Occupational Structure . New York: Free Press. Bollen, Kenneth A., and J. Scott Long. 1993. Testing Structural Equation Models . Newbury Park, CA: Sage Publications. Carroll, M. D., B. K. Kit, and D. A. Lacher. 2012. “Total and High-Density Lipoprotein Cholesterol in Adults: National Health and Nutrition Examination Survey, 2009–2010.” NCHS Data Brief ( 92): 1– 8. Chandola, Tarani, Mel Bartley, Amanda Sacker, Crispin Jenkinson, and Michael Marmot. 2003. “ Health Selection in the Whitehall II Study, UK.” Social Science & Medicine  56( 10): 2059– 72. Google Scholar CrossRef Search ADS   Chen, Feinian, Yang Yang, and Guangya Liu. 2010. “ Social Change and Socioeconomic Disparities in Health over the Life Course in China: A Cohort Analysis.” American Sociological Review  75( 1): 126– 50. Google Scholar CrossRef Search ADS   Coleman, James S. 1988. “ Social Capital in the Creation of Human Capital.” American Journal of Sociology  94: S95– S120. Google Scholar CrossRef Search ADS   Crosnoe, Robert. 2006. “ Health and the Education of Children from Racial/Ethnic Minority and Immigrant Families.” Journal of Health and Social Behavior  47( 1): 77– 93. Google Scholar CrossRef Search ADS   Currie, Janet, and Brigitte C. Madrian. 1999. “ Health, Health Insurance and the Labor Market.” Handbook of Labor Economics  3: 3309– 3416. Google Scholar CrossRef Search ADS   DiPrete, Thomas A., and Gregory M. Eirich. 2006. “ Cumulative Advantage as a Mechanism for Inequality: A Review of Theoretical and Empirical Developments.” Annual Review of Sociology  32: 271– 97. Google Scholar CrossRef Search ADS   Dupre, Matthew E. 2007. “ Educational Differences in Age-Related Patterns of Disease: Reconsidering the Cumulative Disadvantage and Age-as-Leveler Hypotheses.” Journal of Health and Social Behavior  48( 1): 1– 15. Google Scholar CrossRef Search ADS   Dupre, Matthew E. 2008. “ Educational Differences in Health Risks and Illness over the Life Course: A Test of Cumulative Disadvantage Theory.” Social Science Research  37( 4): 1253– 66. Google Scholar CrossRef Search ADS   Elder, Glen H. Jr. 1995. “The Life Course Paradigm: Social Change and Individual Development.” In Examining Lives in Context: Perspectives on the Ecology of Human Development , edited by Phyllis Moen, Glen H. Elder Jr., and Kurt Lüscher, pp. 101– 39. Hyattsville, MD: American Psychological Association. Google Scholar CrossRef Search ADS   Elder, Glen H. Jr., Monica Kirkpatrick Johnson, and Robert Crosnoe. 2003. “The Emergence and Development of Life Course Theory.” In Handbook of the Life Course , edited by Jeylan T. Mortimer and Michael J. Shanahan, pp. 3– 19. New York: Springer. Google Scholar CrossRef Search ADS   Elman, Cheryl, and Angela M. O’Rand. 2007. “ The Effects of Social Origins, Life Events, and Institutional Sorting on Adults’ School Transitions.” Social Science Research  36( 3): 1276– 99. Google Scholar CrossRef Search ADS   Elstad, Jon Ivar, and Steinar Krokstad. 2003. “ Social Causation, Health-Selective Mobility, and the Reproduction of Socioeconomic Health Inequalities over Time: Panel Study of Adult Men.” Social Science & Medicine  57( 8): 1475– 89. Google Scholar CrossRef Search ADS   Featherman, David L., and Robert Mason Hauser. 1978. Opportunity and Change . New York: Academic Press. Ferraro, Kenneth F., and Jessica A. Kelley-Moore. 2003. “ Cumulative Disadvantage and Health: Long-Term Consequences of Obesity?” American Sociological Review  68( 5): 707. Google Scholar CrossRef Search ADS   Fox, John W. 1990. “ Social Class, Mental Illness, and Social Mobility: The Social Selection-Drift Hypothesis for Serious Mental Illness.” Journal of Health and Social Behavior  31( 4): 344– 53. Google Scholar CrossRef Search ADS   Friend, Amanda, Leone Craig, and Steve Turner. 2013. “ The Prevalence of Metabolic Syndrome in Children: A Systematic Review of the Literature.” Metabolic Syndrome and Related Disorders  11( 2): 71– 80. Google Scholar CrossRef Search ADS   Furstenberg, Frank F. 2000. “ The Sociology of Adolescence and Youth in the 1990s: A Critical Commentary.” Journal of Marriage and Family  62( 4): 896– 910. Google Scholar CrossRef Search ADS   Furstenberg, Frank F., Jeanne Brooks-Gunn, and S. Philip Morgan. 1989. Adolescent Mothers in Later Life . New York: Cambridge University Press. Furstenberg, Frank F., Ruben G. Rumbaut, and Richard A. Settersten. 2005. “On the Frontier of Adulthood: Emerging Themes and New Directions.” In On the Frontier of Adulthood: Theory, Research, and Public Policy , edited by Richard A. Settersten, Frank F. Furstenberg, and Ruben G. Rumbaut. Chicago: University of Chicago Press. Google Scholar CrossRef Search ADS   George, Linda K. 2005. “ Socioeconomic Status and Health across the Life Course: Progress and Prospects.” Journals of Gerontology Series B: Psychological Sciences and Social Sciences  60( Special Issue 2): S135– 39. Google Scholar CrossRef Search ADS   Goldman, Noreen. 2001. “ Social Inequalities in Health.” Annals of the New York Academy of Sciences  954( 1): 118– 39. Google Scholar CrossRef Search ADS   Gordon-Larsen, Penny, Melissa C. Nelson, Phil Page, and Barry M. Popkin. 2006. “ Inequality in the Built Environment Underlies Key Health Disparities in Physical Activity and Obesity.” Pediatrics  117( 2): 417– 24. Google Scholar CrossRef Search ADS   Grundy, Scott M., James I. Cleeman, Stephen R. Daniels, Karen A. Donato, Robert H. Eckel, Barry A. Franklin, David J. Gordon, Ronald M. Krauss, Peter J. Savage, and Sidney C. Smith. 2005. “ Diagnosis and Management of the Metabolic Syndrome: An American Heart Association/National Heart, Lung, and Blood Institute Scientific Statement.” Circulation  112( 17): 2735– 52. Google Scholar CrossRef Search ADS   Grusky, D., ed. 2008. Social Stratification: Class, Race, and Gender in Sociological Perspective , 3rd ed. Philadelphia: Westview Press. Guilkey, D. K., and P. M. Lance. 2014. “Program Impact Estimation with Binary Outcome Variables: Monte Carlo Results for Alternative Estimators and Empirical Examples.” In Festschrift in Honor of Peter Schmidt , edited by Robin C. Sickles and William C. Horrace, pp. 5– 46. New York: Springer. Google Scholar CrossRef Search ADS   Haas, S. A. 2006. “ Health Selection and the Process of Social Stratification: The Effect of Childhood Health on Socioeconomic Attainment.” Journal of Health and Social Behavior  47( 4): 339– 54. Google Scholar CrossRef Search ADS   Haas, S. A. 2008. “ Trajectories of Functional Health: The ‘Long Arm’of Childhood Health and Socioeconomic Factors.” Social Science & Medicine  66( 4): 849– 61. Google Scholar CrossRef Search ADS   Hallinan, Maureen T. 1988. “ Equality of Educational Opportunity.” Annual Review of Sociology  14: 249– 68. Google Scholar CrossRef Search ADS   Harris, K. M. 2010. “ An Integrative Approach to Health.” Demography  47( 1): 1– 22. Google Scholar CrossRef Search ADS   Harris, K. M., C. T. Halpern, E. Whitsel, J. Hussey, J. Tabor, P. Entzel, and J. R. Udry. 2009. “The National Longitudinal Study of Adolescent Health: Research Design.” http://www.cpc.unc.edu/projects/addhealth/design. Carolina Population Center, University of North Carolina–Chapel Hill. Harris, Kathleen Mullan, Penny Gordon-Larsen, Kim Chantala, and J. Richard Udry. 2006. “ Longitudinal Trends in Race/Ethnic Disparities in Leading Health Indicators from Adolescence to Young Adulthood.” Archives of Pediatrics & Adolescent Medicine  160( 1): 74– 81. Google Scholar CrossRef Search ADS   Hayward, Mark D., and Bridget K. Gorman. 2004. “ The Long Arm of Childhood: The Influence of Early-Life Social Conditions on Men’s Mortality.” Demography  41( 1): 87– 107. Google Scholar CrossRef Search ADS   Heckman, James J. 2000. “ Policies to Foster Human Capital.” Research in Economics  54( 1): 3– 56. Google Scholar CrossRef Search ADS   Hertzman, Clyde. 2006. “ The Biological Embedding of Early Experience and Its Effects on Health in Adulthood.” Annals of the New York Academy of Sciences  896( 1): 85– 95. Google Scholar CrossRef Search ADS   Hertzman, Clyde, and Chris Power. 2006. “A Life Course Approach to Health and Human Development.” In Healthier Societies: From Analysis to Action , edited by Jody Heymann, Clyde Hertzman, Morris L. Barer, and Robert G. Evans, pp. 83– 106. New York: Oxford University Press. Google Scholar CrossRef Search ADS   Hitlin, Steven, and Glen H. Elder. 2007. “ Time, Self, and the Curiously Abstract Concept of Agency.” Sociological Theory  25( 2): 170– 91. Google Scholar CrossRef Search ADS   Hitlin, Steven, and Monica Kirkpatrick Johnson. 2015. “ Reconceptualizing Agency within the Life Course: The Power of Looking Ahead.” American Journal of Sociology  120( 5): 1429– 72. Google Scholar CrossRef Search ADS   Hogan, Dennis P., and Nan Marie Astone. 1986. “ The Transition to Adulthood.” Annual Review of Sociology  12: 109– 30. Google Scholar CrossRef Search ADS   Hout, Michael. 2007. “ Otis Dudley Duncan’s Major Contributions to the Study of Social Stratification.” Research in Social Stratification and Mobility  25( 2): 109– 18. Google Scholar CrossRef Search ADS   Hout, Michael, and Thomas A. DiPrete. 2006. “ What We Have Learned: RC28’s Contributions to Knowledge about Social Stratification.” Research in Social Stratification and Mobility  24( 1): 1– 20. Google Scholar CrossRef Search ADS   Huber, Machteld, J. André Knottnerus, Lawrence Green, Henriëtte van der Horst, Alejandro R. Jadad, Daan Kromhout, Brian Leonard, Kate Lorig, Maria Isabel Loureiro, and Jos W M van der Meer. 2011. “ How Should We Define Health?” BMJ—British Medical Journal  343( 6): d4163. Google Scholar CrossRef Search ADS   Hummer, Robert A., Richard G. Rogers, Charles B. Nam, and Felicia B. LeClere. 1999. “ Race/Ethnicity, Nativity, and US Adult Mortality.” Social Science Quarterly  80( 1): 136– 53. Jackson, Margot I. 2009. “ Understanding Links between Adolescent Health and Educational Attainment.” Demography  46( 4): 671– 94. Google Scholar CrossRef Search ADS   Kawachi, Ichiro, Nancy E. Adler, and William H. Dow. 2010. “ Money, Schooling, and Health: Mechanisms and Causal Evidence.” Annals of the New York Academy of Sciences  1186( 1): 56– 68. Google Scholar CrossRef Search ADS   Kawachi, Ichiro, and Lisa F. Berkman. 2003. Neighborhoods and Health . New York: Oxford University Press. Google Scholar CrossRef Search ADS   Keister, Lisa A. 2003. “ Religion and Wealth: The Role of Religious Affiliation and Participation in Early Adult Asset Accumulation.” Social Forces  82( 1): 175– 207. Google Scholar CrossRef Search ADS   Kuh, D., and Y. B. Shlomo. 2004. A Life Course Approach to Chronic Diseases Epidemiology , vol. 2. Oxford: Oxford University Press. Google Scholar CrossRef Search ADS   Link, B. G., and J. Phelan. 1995. “ Social Conditions as Fundamental Causes of Disease.” Journal of Health and Social Behavior  ( Extra Issue): 80– 94. Looker, E. Dianne. 1989. “ Accuracy of Proxy Reports of Parental Status Characteristics.” Sociology of Education  62( 4): 257– 76. doi:10.2307/2112830. Google Scholar CrossRef Search ADS   Lynch, John W., George A. Kaplan, and Sarah J. Shema. 1997. “ Cumulative Impact of Sustained Economic Hardship on Physical, Cognitive, Psychological, and Social Functioning.” New England Journal of Medicine  337( 26): 1889– 95. Google Scholar CrossRef Search ADS   Macmillan, Ross. 2006. “‘Constructing Adulthood’: Agency and Subjectivity in the Transition to Adulthood.” Advances in Life Course Research  11: 3– 29. Marmot, M., M. Shipley, E. Brunner, and H. Hemingway. 2001. “ Relative Contribution of Early Life and Adult Socioeconomic Factors to Adult Morbidity in the Whitehall II Study.” Journal of Epidemiology and Community Health  55( 5): 301– 7. Google Scholar CrossRef Search ADS   McEwen, Bruce S. 1998. “ Stress, Adaptation, and Disease: Allostasis and Allostatic Load.” Annals of the New York Academy of Sciences  840( 1): 33– 44. Google Scholar CrossRef Search ADS   Merton, Robert K. 1968. “ The Matthew Effect in Science.” Science (New York, N.Y.)  159( 3810): 56– 63. Google Scholar CrossRef Search ADS   Mirowsky, John, and Catherine E. Ross. 2003. Education, Social Status, and Health . New York: Aldine de Gruyter. Mortimer, Jeylan T. 1994. “Individual Differences as Precursors of Youth Unemployment.” In Youth, Employment, and Society , edited by A. C. Peterson and Jeylan T. Mortimer, pp. 172– 98. New York: Cambridge University Press. Google Scholar CrossRef Search ADS   Mortimer, Jeylan T. 2003. Working and Growing Up in America . Cambridge, MA: Harvard University Press. Mortimer, Jeylan T., Jeremy Staff, and Jennifer C. Lee. 2005. “ Agency and Structure in Educational Attainment and the Transition to Adulthood.” Advances in Life Course Research  10: 131– 53. Google Scholar CrossRef Search ADS   Mouw, Ted. 2005. “Sequences of Early Adult Transitions: How Variable Are They, and Does It Matter?” In On the Frontier of Adulthood: Theory, Research, and Public Policy , edited by Richard Settersten, Frank F. Furstenberg Jr., and Ruben G. Rumbaut, pp. 256– 91. Chicago: University of Chicago Press. Google Scholar CrossRef Search ADS   Mozumdar, Arupendra, and Gary Liguori. 2011. “ Persistent Increase of Prevalence of Metabolic Syndrome among US Adults: NHANES III to NHANES 1999–2006.” Diabetes Care  34( 1): 216– 19. Google Scholar CrossRef Search ADS   Mroz, T. A. 1999. “ Discrete Factor Approximations in Simultaneous Equation Models: Estimating the Impact of a Dummy Endogenous Variable on a Continuous Outcome.” Journal of Econometrics  92( 2): 233– 74. Google Scholar CrossRef Search ADS   Mulatu, Mesfin Samuel, and Carmi Schooler. 2002. “ Causal Connections between Socio-Economic Status and Health: Reciprocal Effects and Mediating Mechanisms.” Journal of Health and Social Behavior  43( 1): 22– 41. Google Scholar CrossRef Search ADS   O’Rand, Angela M. 2001. “Stratification and the Life Course: The Forms of Life-Course Capital and Their Interrelationships.” In Handbook of Aging and the Social Sciences , edited by R. H. Binstock and Linda K. George, pp. 197– 213. San Diego, CA: Academic Press. O’Rand, Angela M. 2009. “Cumulative Processes in the Life Course.” In The Craft of Life Course Research , edited by Glen H. Elder Jr. and Janet Z. Giele, pp. 121– 40. New York: Guilford Press. Palloni, Alberto. 2006. “ Reproducing Inequalities: Luck, Wallets, and the Enduring Effects of Childhood Health.” Demography  43( 4): 587– 615. Google Scholar CrossRef Search ADS   Pollitt, Ricardo A., Kathryn M. Rose, and Jay S. Kaufman. 2005. “ Evaluating the Evidence for Models of Life Course Socioeconomic Factors and Cardiovascular Outcomes: A Systematic Review.” BMC Public Health  5( 1): 7. Google Scholar CrossRef Search ADS   Poulton, Richie, Avshalom Caspi, Barry J. Milne, W. Murray Thomson, Alan Taylor, Malcolm R. Sears, and Terrie E. Moffitt. 2002. “ Association between Children’s Experience of Socioeconomic Disadvantage and Adult Health: A Life-Course Study.” Lancet  360( 9346): 1640– 45. Google Scholar CrossRef Search ADS   Power, C., and C. Hertzman. 1997. “ Social and Biological Pathways Linking Early Life and Adult Disease.” British Medical Bulletin  53( 1): 210– 21. Google Scholar CrossRef Search ADS   Rodgers, Willard L., Mary Beth Ofstedal, and A. Regula Herzog. 2003. “ Trends in Scores on Tests of Cognitive Ability in the Elderly US Population, 1993–2000.” Journals of Gerontology Series B: Psychological Sciences and Social Sciences  58( 6): S338– 46. Google Scholar CrossRef Search ADS   Ross, Catherine E., and Chia-Ling Wu. 1996. “ Education, Age, and the Cumulative Advantage in Health.” Journal of Health and Social Behavior  37( 1): 104– 20. Google Scholar CrossRef Search ADS   Sandefur, Gary D., Jennifer Eggerling-Boeck, and Hyunjoon Park. 2005. “Off to a Good Start? Postsecondary Education and Early Adult Life.” In On the Frontier of Adulthood: Theory, Research, and Public Policy , edited by Richard Settersten, Frank F. Furstenberg Jr., and Ruben G. Rumbaut, pp. 292– 319. Chicago: University of Chicago Press. Google Scholar CrossRef Search ADS   Sargan, John D. 1958. “ The Estimation of Economic Relationships Using Instrumental Variables.” Econometrica: Journal of the Econometric Society  26( 3): 393– 415. Google Scholar CrossRef Search ADS   Sartorius, Norman. 2006. “ The Meanings of Health and Its Promotion.” Croatian Medical Journal  47( 4): 662. Sewell, William H., and Robert M. Hauser. 1975. Education, Occupation, and Earnings. Achievement in the Early Career . New York: Academic Press. Shanahan, Michael J. 2000. “ Pathways to Adulthood in Changing Societies: Variability and Mechanisms in Life Course Perspective.” Annual Review of Sociology  26: 667– 92. Google Scholar CrossRef Search ADS   Singh, Gopal K., and Barry A. Miller. 2003. “ Health, Life Expectancy, and Mortality Patterns among Immigrant Populations in the United States.” Canadian Journal of Public Health—Revue canadienne de santé publique  95( 3): I14– 21. Smith, James P., and Raynard Kington. 1997. “ Demographic and Economic Correlates of Health in Old Age.” Demography  34( 1): 159– 70. Google Scholar CrossRef Search ADS   Stock, James H., Jonathan H. Wright, and Motohiro Yogo. 2002. “ A Survey of Weak Instruments and Weak Identification in Generalized Method of Moments.” Journal of Business & Economic Statistics  20( 4): 518– 29. Google Scholar CrossRef Search ADS   Teachman, Jay D., Kathleen Paasch, and Karen Carver. 1997. “ Social Capital and the Generation of Human Capital.” Social Forces  75( 4): 1343– 59. Google Scholar CrossRef Search ADS   Thoits, Peggy A. 1995. “ Stress, Coping, and Social Support Processes: Where Are We? What Next?” Journal of Health and Social Behavior  ( Extra Issue): 53– 79. Trogdon, Justin G., Eric A. Finkelstein, Isaac A. Nwaise, Florence K. Tangka, and Diane Orenstein. 2007. “ The Economic Burden of Chronic Cardiovascular Disease for Major Insurers.” Health Promotion Practice  8( 3): 234– 42. Google Scholar CrossRef Search ADS   Van Cleave, Jeanne, Steven L. Gortmaker, and James M. Perrin. 2010. “ Dynamics of Obesity and Chronic Health Conditions among Children and Youth.” JAMA: The Journal of the American Medical Association  303( 7): 623– 30. Google Scholar CrossRef Search ADS   Warren, John Robert. 2009. “ Socioeconomic Status and Health across the Life Course: A Test of the Social Causation and Health Selection Hypotheses.” Social Forces  87( 4): 2125– 53. Google Scholar CrossRef Search ADS   Willson, Andrea E., Kim M. Shuey, and Glen H. Elder Jr. 2007. “ Cumulative Advantage Processes as Mechanisms of Inequality in Life Course Health.” American Journal of Sociology  112( 6): 1886– 1924. Google Scholar CrossRef Search ADS   Wu, Shunquan, Rui Wang, Yanfang Zhao, Xiuqiang Ma, Meijing Wu, Xiaoyan Yan, and Jia He. 2013. “ The Relationship between Self-Rated Health and Objective Health Status: A Population-Based Study.” BMC Public Health  13( 1): 320. Google Scholar CrossRef Search ADS   Zimmer-Gembeck, Melanie J., and Jeylan T. Mortimer. 2006. “ Selection Processes and Vocational Development: A Multi-Method Approach.” Advances in Life Course Research  11: 121– 48. Google Scholar CrossRef Search ADS   Author notes This research uses data from Add Health, a program project directed by Kathleen Mullan Harris and designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris at the University of North Carolina–Chapel Hill, and funded by grant P01-HD31921 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, with cooperative funding from 23 other federal agencies and foundations. Special acknowledgment is due to Ronald R. Rindfuss and Barbara Entwisle for assistance on the original design. Information on how to obtain the Add Health data files is available on the Add Health website (http://www.cpc.unc.edu/addhealth). This research received support from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (K99 HD075860, PI: Kane; T32 HD007168, PI: Halpern; P2C HD050924, PI: Morgan). Opinions reflect those of the authors and not necessarily those of the granting agencies. Direct correspondence to Jennifer B. Kane, Department of Sociology, University of California–Irvine, 4171 Social Sciences Plaza A, Irvine, CA 92697; phone: (949) 824-9594; e-mail: jbkane@uci.edu. © The Author 2017. Published by Oxford University Press on behalf of the University of North Carolina at Chapel Hill. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Social Forces Oxford University Press

Pathways of Health and Human Capital from Adolescence into Young Adulthood

Loading next page...
 
/lp/ou_press/pathways-of-health-and-human-capital-from-adolescence-into-young-VPX0QDcpwg
Publisher
Oxford University Press
Copyright
© The Author 2017. Published by Oxford University Press on behalf of the University of North Carolina at Chapel Hill. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
ISSN
0037-7732
eISSN
1534-7605
D.O.I.
10.1093/sf/sox079
Publisher site
See Article on Publisher Site

Abstract

Abstract Social inequalities in health and human capital are core concerns of sociologists, but little research examines the developmental stage when such inequalities are likely to emerge—the transition to adulthood. With new data and innovative statistical methods, we conceptually develop, and empirically operationalize, pathways of physical health and human capital accumulation from adolescence into young adulthood, using an autoregressive cross-lagged structural equation model. Results reveal that pathways of health and human capital accumulate at differential rates across the transition to adulthood; evidence of cross-lagged effects lends support for both social causation and health selection hypotheses. We then apply this model to assess the presence of social inequality in metabolic syndrome—the leading risk factor of cardiovascular disease in the United States. Findings document social stratification of cardiovascular health that is robust to both observed and unobserved social and health selection mechanisms. We speculate that this social stratification will only increase as this cohort ages. Introduction Social inequality is a fundamental sociological concern. Decades of research have documented the trends, prevalence, and correlates of social stratification (Blau and Duncan 1967; Featherman and Hauser 1978; Grusky 2008; Hout 2007; Hout and DiPrete 2006). However, we still have much to learn about how inequalities develop across the early life course, from adolescence into young adulthood. Most social stratification research focuses on the quantity, quality, and access to human capital. Less attention has been paid to the development of physical health inequalities across the life course (Palloni [2006] excepted) and their relation to human capital development. Using rich longitudinal data, this study conceptually develops, and empirically operationalizes, intra-generational pathways of physical health and human capital accumulation from adolescence into young adulthood. This is the first major contribution of our study. We then gauge the impact of these pathways on cardiovascular health in young adulthood—our study’s second major contribution—as this is highly predictive of health and wealth in old age. Health and Human Capital across the Transition to Adulthood According to Blau and Duncan’s (1967) seminal work, social stratification has its roots in both inter- and intra-generational processes: father’s education and occupation (“social origins”) influences child’s education, and each contributes to child’s occupation (“social destination”). Subsequent conceptualizations affirmed this model (Featherman and Hauser 1978), spawning a generation of social stratification research. A common approach, and the one we take here, is to focus specifically on intra-generational processes, whereby early investments in human capital demonstrate lifelong gains for individuals by maximizing physical health as well as psychological and economic well-being (Aneshensel 1992; Ross and Wu 1996; Sewell and Hauser 1975). A separate literature takes a longitudinal approach, noting a pattern of path dependence in which poor physical health is sequentially contingent from one life stage to the next (O’Rand 2009; Willson, Shuey, and Elder 2007). However, neither literature has examined how human capital and health pathways develop and interact across the transition to adulthood. This agenda is challenging in multiple ways. Concerning human capital, there are conceptual and measurement difficulties in estimating pathways linking income potential across early life stages (Furstenberg 2000; Furstenberg, Brooks-Gunn, and Morgan 1989). Most adolescents live with their parents, are in the midst of their educational trajectories, and do not work full-time; thus, current income, educational level, or occupation are poor proxies for their eventual level of human capital. Similarly, individual income is an unreliable indicator of human capital in young adulthood, when some youth attend college while others move quickly into the labor market. We contribute new conceptual and measurement work in this area by estimating a pathway of human capital accumulation across the transition to adulthood using latent constructs that capture human capital at different life stages. Concerning physical health, much past work either constructs trajectories of health only across adulthood, or builds trajectories of health spanning childhood and adulthood by asking adults to retrospectively report childhood illness and general health (Chen, Yang, and Liu 2010; Haas 2006, 2008; Marmot et al. 2001; Palloni 2006; Warren 2009; Willson, Shuey, and Elder 2007). The latter approach suffers from recall bias, measurement error (Looker 1989), and mortality selection (Pollitt, Rose, and Kaufman 2005), as only individuals surviving to mid- to late adulthood are surveyed. Other studies use longitudinal data to identify health pathways from adolescence through young adulthood (Bauldry et al. 2012) or links between adolescent health and key educational outcomes (Jackson 2009), but employ a single, subjective indicator of physical health (self-reported health). Moreover, these studies face the difficulty of estimating causal effects of social factors on health, aside from key unmeasured factors (Adler, Bush, and Pantell 2012; Goldman 2001; Kawachi, Adler, and Dow 2010; Palloni 2006). All in all, we know relatively little about how early life health habits and behavior manifest within individuals over the early life course. We conceptually develop and estimate pathways of physical health, based on multiple objective and subjective measures of health, across the transition to adulthood. We also address the endogeneity of health (and human capital) by using multiple time-varying instrumental variables to identify the system of equations in our model. In addition, health and human capital development are likely interdependent. In adulthood, poor physical health is associated with lower levels of educational attainment, lower income, fewer hours worked in the labor market, and ultimately less accumulation of wealth over time (Currie and Madrian 1999; Haas 2006, 2008; Jackson 2009), pointing to health as a primary driver of social stratification. In turn, socioeconomic disadvantage is associated with health detriments through greater exposure to stress and poor physical environments (Adler and Ostrove 1999; Adler, Bush, and Pantell 2012)—in neighborhoods and schools, within families, and in the labor market through work in low-wage jobs. Furthermore, these processes are likely deeply intertwined, reflecting a mutual interdependence of health and human capital with strong path-dependent components (O’Rand 2001) that makes it difficult to produce unbiased estimates of either one (Adler, Bush, and Pantell 2012; Kawachi, Adler, and Dow 2010). We model these interdependencies and estimate their causal impacts. Our study takes the additional step of estimating the effects of accumulated human capital and physical health on an important, objective marker of cardiovascular health in young adulthood, metabolic syndrome, that is measured using anthropometric measures and physiological markers of health. Metabolic syndrome is the leading risk factor for adult-onset cardiovascular disease and Type II diabetes—both of which encompass significant morbidity and mortality risk in the United States (Beltrán-Sánchez et al. 2013). Theory and Research on Intra-Generational Development of Health and Human Capital The life course perspective is a useful organizing framework for understanding intra-generational development of human capital and physical health. While acknowledging the import of human agency (Hitlin and Elder 2007), events during childhood, adolescence, and adulthood are cumulative and are connected to broader structural context and social change (Elder 1995; Elder, Johnson, and Crosnoe 2003). The cumulative disadvantage hypothesis articulates how later-life inequalities are anchored in early life experiences (Merton 1968; Ross and Wu 1996). These early life experiences have been broadly defined to include differential exposures to persistent poverty, social support, and stressful neighborhood environments (DiPrete and Eirich 2006). The Importance of the Early Life Course Striking distinctions in economic well-being are drawn between individuals who complete high school and those who do not (Hallinan 1988; Hogan and Astone 1986). High school completion is an important credential for the low-wage labor force and opens many employment opportunities. Early returns to high school success are also evident in college attendance and completion; in turn, those who complete college benefit from more advantageous economic and employment opportunities in adulthood. Thus, adolescence and the transition to adulthood are critically formative for lifelong human capital. Furthermore, as family-, employment-, and schooling-related transitions have become less structured, human agency has played an increasingly important role in the transition to adulthood (Mortimer 1994; Shanahan 2000). This is a time when individuals try on new roles and navigate myriad choices in different realms (labor market, dating market, etc.); behavior during this stage forms the basis for later-observed economic behavior (Macmillan 2006; Zimmer-Gembeck and Mortimer 2006). Social inequalities in adult health are rooted in childhood experiences (Blackwell, Hayward, and Crimmins 2001; Hayward and Gorman 2004; Lynch, Kaplan, and Shema 1997). Childhood socioeconomic status (SES) has persistent effects on adult health outcomes, net of adult SES (Poulton et al. 2002). Identifying early life determinants of adult/old age health are active research questions (Adler, Bush, and Pantell 2012; George 2005; Kawachi, Adler, and Dow 2010). Furthermore, poor health in early life can constrain subsequent socioeconomic attainment and health maintenance (Elman and O’Rand 2007; O’Rand 2001).Yet, most studies rely on retrospective measures of early life events. Studies such as Add Health are advantageous because early life markers of social well-being and health are measured prospectively and, in the case of health, with multiple subjective and objective indicators. Much remains to be learned about the pathways linking early life exposures to adult health outcomes (Power and Hertzman 1997), although a growing literature proposes and tests potential pathways (see, for example, Almond and Currie [2011]; Ben-Shlomo and Kuh [2002]; Hertzman [2006]; Hertzman and Power [2006]; Kuh and Shlomo [2004]). Most such studies, however, have insufficient data describing exposures during adolescence—a stage in which young people assert their independence and begin to manage lifestyle choices and health behaviors that often persist (Harris 2010; Macmillan 2006)—and the transition to adulthood—a life stage dense with critical transitions (e.g., leaving home, attending college, entering full-time employment, entering a marriage or cohabitation, transitioning to parenthood) (Elder 1995; Elder, Johnson, and Crosnoe 2003; Shanahan 2000). These are periods when behavioral and health trajectories gain momentum and inequalities become entrenched. Social Causation, Health Selection, and Mutual Interdependence Physical health and socioeconomic status are tightly linked across the life course. Social causation suggests that early and sustained social and economic hardship leads to self-reinforcing cycles of poor health, including health problems such as obesity (Ferraro and Kelley-Moore 2003), hypertension, heart disease (Dupre 2007, 2008), cognitive decline (Rodgers, Ofstedal, and Herzog 2003), and higher rates of disability and mortality in older adulthood (Lynch, Kaplan, and Shema 1997; Smith and Kington 1997). Health selection posits that poor health decreases labor force participation (for a review, see Currie and Madrian [1999], in turn reducing financial capital and socioeconomic status (Haas 2006). These interrelated processes (social causation and health selection) can induce upward or downward mobility across socioeconomic strata. Past studies have focused on which mechanism dominates. The bulk of evidence and argument suggests social causation (Chandola et al. 2003; Elstad and Krokstad 2003; Mulatu and Schooler 2002), which resonates with Link and Phelan’s (1995) argument that social conditions determine access to health-related resources, thereby producing social inequalities in health. Lower levels of human or financial capital lead to lower access to, or utilization of, healthcare (Ross and Wu 1996), and are associated with lower levels of social support (Thoits 1995) and greater exposure to cumulative stress (McEwen 1998). Social causation processes are also supported by Mirowsky and Ross’s (2003) notion of education as learned effectiveness: education improves health by increasing self-efficacy and problem-solving capacity. Others (Fox 1990; Haas 2006) have argued that both processes are at work (particularly in early life). However, other scholars place emphasis not on social causation or health selection, but rather on the mutual interdependence between health and human capital through which (dis)advantages accumulate (O’Rand 2001). Conceptualizing Pathways of Physical Health and Human Capital across the Transition to Adulthood We now introduce our conceptualization of physical health and human capital across the transition to adulthood. Physical health is a complex construct (Sartorius 2006); thus, multiple indicators are useful and their selection depends on life stage. But generally, key indicators include physical activity (or activities of daily living, among older adults), body composition (e.g., adult body mass index or nutritional intake; infant birth weight), and/or health conditions (e.g., self-reported health, medical examinations, presence of chronic health conditions) (Bircher 2005; Huber et al. 2011). We strive for parsimony and replicability: we choose one indicator with face validity from each domain—level of physical activity, body mass index, and self-reported health. For human capital, we follow past theoretical work in economics that defines investment in human capital as “activities that influence future real income through the imbedding of resources in people” (Becker [1962], p. 9). Thus, we conceptualize human capital as one’s potential for income and seek to identify activities that embed resources within individuals. Of course, these activities will differ by life stage as the key institutions with which individuals interact change dramatically across the transition to adulthood. During adolescence, key resource-embedding activities relate to education. Schools are crucial in the production of training; they transmit knowledge, skills, and resources to youth that influence their potential income (Becker 1962; Coleman 1988). The success of this transmission can be captured, at least in part, by markers of aptitude (e.g., cognitive ability) and school performance (e.g., assigned grades, being on track with age-related peers). These markers are also linked with college attendance—an important factor stratifying future employment opportunities and thereby future income. Pathways through the transition to adulthood are less structured now than in the past (Furstenberg, Rumbaut, and Settersten 2005; Hogan and Astone 1986; Shanahan 2000). Still, most individuals enroll in college or enter the labor market in at least a part-time capacity during the transition to adulthood (Mouw 2005; Sandefur, Eggerling-Boeck, and Park 2005)—decisions that are likely informed by youths’ structural advantage, school performance, and anticipated socioeconomic gains of employment (Mortimer 2003; Mortimer, Staff, and Lee 2005). Both college and early full-time employment involve resource-embedding activities such as gaining credentials or employment. Conversely, remaining idle during this time is defined by exclusion—not working and not attending school (Shanahan 2000). More than a third of idle youth suffer from cognitive impairments (Amato et al. 2008), but idleness outside such impairment can be problematic, as it restricts youths’ investment in future income. Additionally, financial decisions in young adulthood are influential determinants of net worth later in life. Individuals holding low-risk financial assets (e.g., a savings or checking account) in early adulthood accumulate more assets by mid- to late adulthood compared with those who are permanently asset-poor—meaning, never holding a checking account, savings account, bonds or stock, or never owning a home (Keister 2003). More generally, holding such assets within the transition to adulthood likely signals a foundation for longer-term advantageous behaviors (Macmillan 2006; Shanahan 2000; Zimmer-Gembeck and Mortimer 2006). In sum, activities reflecting potential for income during the transition to adulthood include education and employment (versus being idle), cognitive ability, and financial asset activity. By young adulthood, many individuals have exited schooling and have entered the labor force (or engaged in caretaking) (Mouw 2005; Sandefur, Eggerling-Boeck, and Park 2005). At this stage, we conceptualize human capital as past literature focusing on mid- to late adulthood—via employment behavior, educational attainment, and socioeconomic position, whether objective or subjective (Teachman, Paasch, and Carver 1997). Current Study This is the first study to prospectively estimate the development of physical health and human capital pathways from adolescence into young adulthood for a contemporary, nationally representative young adult cohort (Add Health). We use maximum likelihood estimation (MLE) for a structural, multi-equation model to estimate causal pathways of health and human capital development and their cross-lagged effects on young adult cardiovascular health. Throughout, we employ multiple measures of health and human capital at each time point to track developmental change. Our study addresses three questions. First, how does physical health and human capital develop within individuals across the transition to adulthood? Does one life stage seem to be more critical than another? Second, is there evidence of persistent social inequality in adult health, above and beyond underlying selection mechanisms? Third, through what pathways do young adult health inequalities manifest? Data, Measures, and Analytic Strategy Data We use data from the National Longitudinal Study of Adolescent to Adult Health (Add Health), a school-based, nationally representative sample of 20,745 seventh to twelfth graders in 1994–1995. Respondents were re-interviewed in 1996 (Wave II), 2001–2002 (Wave III), and 2008–2009 (Wave IV) (Harris 2010; Harris et al. 2009). Our sample includes male and (non-pregnant) female respondents participating in Waves I, III, and IV (as anthropometric measures of health are not comparable across pregnant and non-pregnant women; n = 12,311). We select native-born respondents (n = 11,499) based on complex differences in health status for foreign-born versus native-born individuals (Hummer et al. 1999; Singh and Miller 2003), particularly in early life (Crosnoe 2006). Respondents with a valid sampling weight are retained to produce nationally representative estimates (n = 9,033). Missing data is minimal (<4 percent for most variables). We use a single imputation procedure in Stata to replace missing data on all independent variables. Results based on listwise deletion and full information maximum likelihood (in Mplus) were largely unchanged. Analyses are weighted and stratified to account for the sampling design. Measures As depicted in figure 1, we constructed a latent variable reflecting poor health for each life stage represented in our data (Poor Health, WI, adolescence; Poor Health WIII, emerging adulthood; and Poor Health WIV, young adulthood) using the same three variables: self-reported health (1 = excellent, 4 = fair or poor), body mass index (range = 11.21 – 80.40),1 and physical inactivity (bouts of physical inactivity in the last 7 days; range [WI] = 0–3; range [Waves III and IV] = 0–5). Figure 1. View largeDownload slide Measurement model of the accumulation of poor health and human capital across the transition to adulthood Note: Path dependence is indicated by paths linking poor health over time and paths linking human capital over time. Social causation is indicated by paths linking human capital to poor health. Health selection is indicated by paths linking poor health to human capital. Social inequality in health net of health selection is indicated by the path linking WIV human capital to metabolic syndrome. Figure 1. View largeDownload slide Measurement model of the accumulation of poor health and human capital across the transition to adulthood Note: Path dependence is indicated by paths linking poor health over time and paths linking human capital over time. Social causation is indicated by paths linking human capital to poor health. Health selection is indicated by paths linking poor health to human capital. Social inequality in health net of health selection is indicated by the path linking WIV human capital to metabolic syndrome. Likewise, we constructed a latent variable reflecting human capital in each life stage (see figure 1) using age- and developmental stage-appropriate measures. In adolescence, human capital is indicated by GPA (range = 1–4), Add Health Picture Vocabulary Test (AH PVT) score—a measure of cognitive ability (range = 14–146), and if the respondent was ever held back a grade (1 = not retained). In emerging adulthood, human capital is indicated by years of schooling (range = 6–22), AH PVT score (range = 7–122), if the respondent was employed or in school (1 = yes), and if the respondent had a checking and/or savings account (1 = yes). In young adulthood, years of schooling (range = 8–26), household income (range = $0–$150,000), employment status (1 = employed), and self-rated socioeconomic status (range = 1–10) characterize human capital.2 Similar to the clinical definition of metabolic syndrome (Grundy et al. 2005),3 a latent variable reflecting metabolic syndrome (shown in figure 1) was constructed using five indicators: elevated waist circumference (≥ 88 cm for women and 102 cm for men), elevated blood pressure (≥ 130 mm Hg systolic blood pressure, ≥ 85 mm Hg diastolic blood pressure, or antihypertensive drug treatment), elevated triglycerides (membership in highest decile of measured triglycerides), reduced high-density lipoprotein cholesterol (HDL-C) (membership in lowest 2 deciles of HDL-C, based on sex), and pre-diabetic value of glycosylated hemoglobin (HbA1c) (> 5.6 percent of hemoglobin molecules are glycosylated). WI control variables include family structure (1 = two biological parent family, 0 = all other family structures), parent education (range = 0–9),4 (logged) household income-needs ratio, race-ethnicity, age, and respondent sex (1 = female). Additionally, we control for WIII residence in urban/non-urban census tract and WIV household composition (reside with parents, married or cohabiting, live alone). Consistent with past research linking structural and economic features of the neighborhood environment to individuals’ health and human capital (e.g., Kawachi and Berkman [2003]), we include multiple, contextual, time-varying instrumental variables for both: subjective indicators of neighborhood safety (if s/he knows most people in their neighborhood, stopped to talk to any neighbors in the past month, usually feels safe in his/her neighborhood), an objective indicator of neighborhood safety (juvenile violent crime rate), indicators of the economic landscape (Bureau of Labor Statistics cost price index,5 per capita income, male unemployment rate, percent of female-headed households with children), and indicators of housing quality (percent of homes with insufficient plumbing, median year housing was built—before or after 1978). Additional instrumental variables of poor health include (Euclidean) distance between the respondent’s home and nearest park, age-specific mortality rate of 15–24-year-olds, and the number of inactivity resources (i.e., movie theaters, arcades) within 1 km of their home. In order to be plausible instruments, these factors must be correlated with health (or human capital) at a given time, but may not predict future human capital or health except through the lagged value. The proposed instruments are plausible on conceptual grounds. Perceived neighborhood safety or Euclidian distance to the nearest park, for example, can regulate one’s ability to engage in physical activity locally, with important implications for overweight status (Gordon-Larsen et al. 2006); however, these factors would not directly impact time2 physical health, except through time1 health behaviors and/or weight status. We also tested the assumptions of instrumental variables. Results based on ivregress post-estimation commands reject the null hypothesis of exogeneity (Wu-Hausman test: p < 0.001 for each instance), while the identification tests fail to reject the null hypothesis of correlation between the error terms in the structural equations and the instrumental variables, thus indicating that the exclusion restrictions are valid (WIII health: χ2 = 6.11, p = 0.41; WIV health: χ2 = 1.18, p = 0.76; WIII human capital: χ2 = 4.24, p = 0.52; WIV human capital: χ2 = 6.19, p = 0.11) (Sargan 1958). Furthermore, the instruments are sufficient in strength (WIII health: F = 21.14; 10 percent value for 2SLS = 10.27; WIV health: F = 10.71; 10 percent value for 2SLS = 9.08; WIII human capital: F = 34.42; 10 percent value for 2SLS = 10.27; WIV human capital: F = 95.12; 10 percent value for 2SLS = 9.08) (Stock, Wright, and Yogo 2002). Analytic Strategy We employ an autoregressive structural equation model (SEM) to estimate pathways of physical health and human capital and cross-lagged effects; we also allow each pathway to have a direct effect on metabolic syndrome. From this base model, we relax a series of parameter constraints and compare five nested models to select a preferred model (see table 1). For example, in the first column, the likelihood ratio test compares model 1 (including none of the instruments, error covariances for latent variables, or cross-lags) to model 2 (including instruments, but not error covariances or cross-lags); the statistical significance of the likelihood ratio test statistic indicates that model 2 is a better fit of the data. This was repeated until all parameter constraints noted were relaxed. The final column indicates that model 5 was the best fit overall. Thus, our preferred model includes cross-lags and error covariances across the system of equations jointly along with model parameters, and uses instrumental variables within the SEM to statistically identify causal effects for health and human capital at each life course stage. This strategy produces consistent estimates of the effect of health (or human capital) at one point in time on health (or human capital) at a subsequent time point, and effectively allows us to address the endogeneity of both health and human capital across time. Table 1. Results from a Nested Model Comparison Test to Select a Preferred Model Model 1: Covariances = 0, cross-lags = 0 Model 2: Covariances = 0, cross-lags = 0, + instruments Model 3: Covariances ≠ 0, cross-lags = 0, + instruments Model 4: Covariances = 0, cross-lags ≠ 0, + instruments Model 5: Covariances ≠ 0, cross-lags ≠ 0, + instruments    Model 1 vs. Model 2  Model 2 vs. Model 3  Model 2 vs. Model 4  Model 2 vs. Model 5  LLx  94,500.11  94,148.71  94,148.71  94,148.71  DFx  76  105  105  105  LLx+1  94,148.71  93,699.98  93,992.82  93,558.29  DFx+1  105  111  109  115  Difference in deviance statistic  351.40  448.73  155.89  590.42  Difference in df  29  6  4  10  LR test  0.000  0.000  0.000  0.000  Best fitting model  Model 2  Model 3  Model 4  Model 5  Model 1: Covariances = 0, cross-lags = 0 Model 2: Covariances = 0, cross-lags = 0, + instruments Model 3: Covariances ≠ 0, cross-lags = 0, + instruments Model 4: Covariances = 0, cross-lags ≠ 0, + instruments Model 5: Covariances ≠ 0, cross-lags ≠ 0, + instruments    Model 1 vs. Model 2  Model 2 vs. Model 3  Model 2 vs. Model 4  Model 2 vs. Model 5  LLx  94,500.11  94,148.71  94,148.71  94,148.71  DFx  76  105  105  105  LLx+1  94,148.71  93,699.98  93,992.82  93,558.29  DFx+1  105  111  109  115  Difference in deviance statistic  351.40  448.73  155.89  590.42  Difference in df  29  6  4  10  LR test  0.000  0.000  0.000  0.000  Best fitting model  Model 2  Model 3  Model 4  Model 5  We estimate this preferred model using maximum likelihood estimation in Mplus 7. MLE produces efficient parameter estimates, assuming error terms follow a multivariate normal distribution. This is a tenuous assumption, however, particularly when the true distribution is non-normal (Guilkey and Lance 2014; Mroz 1999), although very few studies compare how their parameter estimates differ with and without the normality assumption. We assess the extent to which our results are robust to this assumption by comparing MLE results to those produced by a two-stage least squares instrumental variables regression model, estimated in Stata 13.0 using the ivregress command—a statistical specification that does not assume normality but is less efficient. If model results are similar across both specifications, this would suggest that joint normality is not violated and the more efficient estimation (MLE) is preferred. If model results are not similar, joint normality may be violated and the IV results are preferred. We then estimate total, direct, and indirect effects to identify pathways through which health and human capital operate directly and indirectly on metabolic syndrome. Finally, we assess whether or not the endogeneity corrections we employ are necessary. Because the system of equations in our model is jointly estimated using full information MLE, this approach requires IVs for statistical identification. Therefore, to assess whether or not correcting for endogeneity makes a difference, we calculate OLS regression estimates, where we adjust for controls, lagged values of health and human capital, and cross-lagged values, but not for IVs or correlated error terms between latent variables. We refer to this as the uncorrected model. Results Table 2 presents descriptive statistics of all study variables, weighted and adjusted for clustering of the sample design. The mean and variance of self-reported health remained somewhat stable over time, but rose steadily for BMI. Physical inactivity was lowest during high school, increasing within emerging and young adulthood. Average GPA in the current or past academic year was 2.79; 79 percent of respondents never repeated a grade. Average AHPVT score remained similar over time. Eighty-three percent of respondents were not idle in emerging adulthood; many had a checking (71 percent) or savings (63 percent) account. In this life stage, the average level of completed education was 13.1 years; by young adulthood it was 14.2 years. In young adulthood, two-thirds (66 percent) were employed; average household income was $60,125. Over half of sample members were classified with elevated waist circumference (52 percent) and just under half with elevated blood pressure (42 percent). Nearly a third met or surpassed the pre-diabetic cutoff for glycosylated hemoglobin (31 percent). Table 2. Descriptive Statistics   Mean or percent  Std. dev.  Minimum  Maximum  Poor health—Adolescence (Wave I)  Self-reported health  2.14  0.90  1  4  BMI  22.56  4.46  11.21  49.12  Physical inactivity  1.04  0.85  0  3  Poor health—Emerging adulthood (Wave III)  Self-reported health  2.02  0.86  1  4  BMI  26.91  6.44  13.80  73.17  Physical inactivity  2.98  1.57  0  5  Poor health—Young adulthood (Wave IV)  Self-reported health  2.36  0.88  1  4  BMI  29.17  7.62  14.40  80.40  Physical inactivity  2.86  1.50  0  5  Human capital—Adolescence (Wave I)  GPA  2.79  0.72  1  4  Add Health Picture Vocabulary Test  102.47  13.36  14  146  No grade retention  78.96%    0  1  Human capital—Emerging adulthood (Wave III)  Years of completed schooling  13.05  1.97  6  22  Add Health Picture Vocabulary Test  100.56  14.12  7  122  Employed or in school  82.96%    0  1  Have checking account  71.40%    0  1  Have savings account  62.54%    0  1  Human capital—Young adulthood (Wave IV)  Years of completed schooling  14.23  2.59  8  26  Household income  60,125  37,057  0  150,000  Currently employed  66.49%    0  1  Self-rated socioeconomic status  4.95  1.74  1  10  Metabolic syndrome  Elevated waist circumference  51.80%    0  1  Elevated blood pressure  41.84%    0  1  Elevated triglycerides  9.89%    0  1  Reduced HDL cholesterol  16.25%    0  1  Pre-diabetic value, hemoglobin A1c  31.24%    0  1  Sociodemographic controls—Adolescence (Wave I)  Two-parent family structure  56.40%    0  1  Parental education  5.50  2.13  0  9  Household income–needs ratio  2.98  3.00  0.00  97.80   Race-ethnicity  Non-Hispanic White (reference)  72.14%    0  1  Non-Hispanic Black  14.75%    0  1  Hispanic  3.58%    0  1  Non-Hispanic other  9.52%    0  1  Age at Wave I  15.43  1.82  11  21  Sex (1 = female)  51.02%    0  1  Wave IV household compositiona  Live alone  11.44%    0  1  Live with parents  15.85%    0  1  Married or cohabiting  57.50%    0  1  Wave III urbanicity  Living in urban census tract  58.25%    0  1  Instrumental variables          Euclidean distance to nearest park (meters), WI  11,830.92  11,386.65  0  81,562.27  Know most people in neighborhood, WI  74.08%    0  1  Stopped to talk to neighbors in past month, WI  79.23%    0  1  Usually feel safe in neighborhood, WI  91.00%    0  1  Juvenile violent crime rate (per 100,000), WI  48.60  36.47  0  214.94  Bureau of Labor Statistics cost price index, WIII  1.78  0.01  1.75  1.80  Inactivity resources, count, WIII  0.18  0.75  0  26  Median year housing built was before 1978, WIII  74.40%    0  1  Per capita income, WIII  19,493  8,095  2,700  108,600  Age-specific mortality rate (per 1,000), WIII  0.83  0.37  0  3.48  Juvenile violent crime rate (per 100,000), WIV  27.68  21.43  0  90.00  Male unemployment rate, WIV  0.08  0.06  0  0.68  Percent female-headed households with children, WIV  7.90%  5.28  0  49.08  Percent of homes with insufficient plumbing, WIV  1.85%  3.04  0  31.23    Mean or percent  Std. dev.  Minimum  Maximum  Poor health—Adolescence (Wave I)  Self-reported health  2.14  0.90  1  4  BMI  22.56  4.46  11.21  49.12  Physical inactivity  1.04  0.85  0  3  Poor health—Emerging adulthood (Wave III)  Self-reported health  2.02  0.86  1  4  BMI  26.91  6.44  13.80  73.17  Physical inactivity  2.98  1.57  0  5  Poor health—Young adulthood (Wave IV)  Self-reported health  2.36  0.88  1  4  BMI  29.17  7.62  14.40  80.40  Physical inactivity  2.86  1.50  0  5  Human capital—Adolescence (Wave I)  GPA  2.79  0.72  1  4  Add Health Picture Vocabulary Test  102.47  13.36  14  146  No grade retention  78.96%    0  1  Human capital—Emerging adulthood (Wave III)  Years of completed schooling  13.05  1.97  6  22  Add Health Picture Vocabulary Test  100.56  14.12  7  122  Employed or in school  82.96%    0  1  Have checking account  71.40%    0  1  Have savings account  62.54%    0  1  Human capital—Young adulthood (Wave IV)  Years of completed schooling  14.23  2.59  8  26  Household income  60,125  37,057  0  150,000  Currently employed  66.49%    0  1  Self-rated socioeconomic status  4.95  1.74  1  10  Metabolic syndrome  Elevated waist circumference  51.80%    0  1  Elevated blood pressure  41.84%    0  1  Elevated triglycerides  9.89%    0  1  Reduced HDL cholesterol  16.25%    0  1  Pre-diabetic value, hemoglobin A1c  31.24%    0  1  Sociodemographic controls—Adolescence (Wave I)  Two-parent family structure  56.40%    0  1  Parental education  5.50  2.13  0  9  Household income–needs ratio  2.98  3.00  0.00  97.80   Race-ethnicity  Non-Hispanic White (reference)  72.14%    0  1  Non-Hispanic Black  14.75%    0  1  Hispanic  3.58%    0  1  Non-Hispanic other  9.52%    0  1  Age at Wave I  15.43  1.82  11  21  Sex (1 = female)  51.02%    0  1  Wave IV household compositiona  Live alone  11.44%    0  1  Live with parents  15.85%    0  1  Married or cohabiting  57.50%    0  1  Wave III urbanicity  Living in urban census tract  58.25%    0  1  Instrumental variables          Euclidean distance to nearest park (meters), WI  11,830.92  11,386.65  0  81,562.27  Know most people in neighborhood, WI  74.08%    0  1  Stopped to talk to neighbors in past month, WI  79.23%    0  1  Usually feel safe in neighborhood, WI  91.00%    0  1  Juvenile violent crime rate (per 100,000), WI  48.60  36.47  0  214.94  Bureau of Labor Statistics cost price index, WIII  1.78  0.01  1.75  1.80  Inactivity resources, count, WIII  0.18  0.75  0  26  Median year housing built was before 1978, WIII  74.40%    0  1  Per capita income, WIII  19,493  8,095  2,700  108,600  Age-specific mortality rate (per 1,000), WIII  0.83  0.37  0  3.48  Juvenile violent crime rate (per 100,000), WIV  27.68  21.43  0  90.00  Male unemployment rate, WIV  0.08  0.06  0  0.68  Percent female-headed households with children, WIV  7.90%  5.28  0  49.08  Percent of homes with insufficient plumbing, WIV  1.85%  3.04  0  31.23  Note: Sample size is 9,033 men and women. Statistics are weighted and adjusted for clustering. aPercentages do not total 100, as these are not mutually exclusive categories of the same variable. Age-specific mortality rate indicates a smoothed rate across 1999–2001 of the number of deaths among persons age 15–24 per 1,000 residents at the county level. Source: National Longitudinal Study of Adolescent Health. Selected parameters from the SEM are presented in figure 2; all parameters are presented in appendix A.6 Unstandardized (not italicized) and standardized (italicized) beta coefficients are presented with standard errors in parentheses. Figure 2. View largeDownload slide Estimates from the structural equation model of the accumulation of poor health and human capital on metabolic syndrome Note:N = 9,033. Unstandardized (not italicized) and standardized (italicized) beta coefficients with standard errors in parentheses are presented. Covariances of error terms of all latent variables are included but not shown. Additional controls for human capital (WIV) not shown are: household composition (married/cohabiting, live alone, live with parents), urban residence. RMSEA = 0.026; SRMR = 0.050. Figure 2. View largeDownload slide Estimates from the structural equation model of the accumulation of poor health and human capital on metabolic syndrome Note:N = 9,033. Unstandardized (not italicized) and standardized (italicized) beta coefficients with standard errors in parentheses are presented. Covariances of error terms of all latent variables are included but not shown. Additional controls for human capital (WIV) not shown are: household composition (married/cohabiting, live alone, live with parents), urban residence. RMSEA = 0.026; SRMR = 0.050. Our first key question is: how does physical health and human capital develop within individuals across the transition to adulthood? Results demonstrate strong path dependence: all path coefficients linking poor health over time and those linking human capital over time are positive and statistically significant (p < 0.001). Comparing the standardized coefficients, first for poor health, shows that the path coefficient linking poor health in WI and WIII (standardized b = 0.92) is more than twice as large as the path coefficient linking poor health in WIII and WIV (standardized b = 0.40). Thus, path dependence appears to be stronger during later (versus earlier) stages of the transition to adulthood. Results from a Wald test suggest that these two path coefficients cannot be constrained to be equal (PH1 → PH3 ≠ PH3 → PH4; test statistic = 24.46, df = 1, p < 0.001). Substantively, this indicates that poor health appears to develop at a differential rate, rather than a constant rate, as individuals transition into adulthood. Estimates of path dependence for human capital present a different pattern. The standardized path coefficient linking WI and WIII human capital (standardized b = 1.01) is slightly larger than the corollary path coefficient between WIII and WIV (standardized b = 0.99). A Wald test suggests the null hypothesis (HC1 → HC3 = HC3 → HC4) can be rejected (160.47, df = 1, p < 0.001), indicating that the human capital pathway crystalizes quickly in adolescence and is further solidified in young adulthood. Next, we address our second research question: is there evidence of persistent social inequality in adult health, above and beyond underlying selection mechanisms? The key estimate is the effect of young adult human capital on metabolic syndrome. This path coefficient (unstandardized b = −0.02, p < 0.001) indicates that higher levels of young adult human capital are associated with lower risk of metabolic syndrome, net of all other pathways in the model, including the endogeneity corrections for underlying social selection and health selection mechanisms. Thus, we observe salient social stratification of metabolic syndrome in this contemporary cohort. The statistical specification used to produce figure 2 assumed the scaling indicator (elevated waist circumference) for metabolic syndrome was continuous; therefore, the path coefficients linking young adult health and human capital to metabolic syndrome are approximated by a linear probability model. To examine if study results were sensitive to this specification, we replicated the SEM scaling metabolic syndrome by the linear form of waist circumference, triglycerides, and hemoglobin A1c (one at a time). Results presented in appendix B suggest that study findings are extremely robust across different specifications of metabolic syndrome. We also examined the extent to which the normality assumption imposed by MLE was upheld. Results (presented in appendix C) indicate that a two-stage least squares instrumental variable regression specification produces similar, and at times identical, point estimates compared to the MLE. The standard errors produced in the IV model are wider, reflecting the decreased efficiency of this approach relative to MLE. The similarity across specifications provides strong evidence that the normality assumption imposed by MLE is not likely violated in this model. Therefore, the MLE results are preferred due to increased efficiency. We now return to the full model to examine our final research question: through what pathways do young adult health inequalities manifest? We address this in two ways: we examine the cross-lags and then assess total, direct, and indirect effects. As depicted in figure 2, cross-lags linking factors between WI and WIII are not statistically significant. However, the cross-lags between WIII and WIV are statistically significant at the p < 0.001 level. Both are negative, indicating that poor health and human capital are inversely related. The standardized coefficient for the path reflecting social causation is –0.04; that reflecting health selection is –0.05. Results from a Wald test suggest the null hypothesis (HC3 → PH4 = PH3 → HC4) can be rejected (7.161, df = 1, p = 0.008). Thus, we conclude that both social causation and health selection are present during this life stage and appear to differ modestly in strength. Table 3 presents a summary of total, direct, and indirect effects (poor health in panel A, human capital in panel B). The first coefficient shown represents the direct effect of WIV poor health on metabolic syndrome (unstandardized b = 0.42); this also appears in figure 2. The second coefficient represents the total effect of WIII poor health on metabolic syndrome (b = 0.45). (With no direct effect here, the sum of the indirect effects is equal to the total effect.) The two indirect effects are Health3 → Health4 → Metabolic syndrome (b = 0.45), and Health3 → Human Capital4 → Metabolic syndrome (b = 0.002).7 Even though unstandardized betas are presented, the difference exhibited makes the relative association fairly clear. The last set of results in panel A mirror this finding: the total effect of WI poor health on metabolic syndrome (unstandardized b = 0.26) is composed of four indirect effects; the one largest in magnitude does not involve a cross-lag (Health1 → Health3 → Health4 → Metabolic syndrome; b = 0.26). Panel B depicts a very similar pattern of results for human capital. Thus, overall the key causal pathways reflect path dependence, a finding to which we return in the discussion below. Table 3. Total, Direct, and Indirect Effects of Poor Health and Human Capital on Metabolic Syndrome Panel A. Effects of poor health at Waves I, III, and IV  Unstandardized β (SE)  β/SE  Poor health (WIV) → Metabolic syndrome       Total effect = Direct effect  0.42 (0.01)***  35.42  Poor health (WIII) → Metabolic syndrome       Total effect = Indirect effect  0.45 (0.01)***  31.75  Specific indirect paths      Health3 → Health4 → Metabolic syndrome  0.45 (0.01)***  31.89  Health3 → Human capital4 → Metabolic syndrome  0.002 (0.001)*  2.46  Poor health (WI) → Metabolic syndrome      Total effect = Indirect effect  0.26 (0.04)***  6.26  Specific indirect paths      Health1 → Health3 → Health4 → Metabolic syndrome  0.26 (0.04)***  6.23  Health1 → Health3 → Human capital4 → Metabolic syndrome  0.001 (0.001)*  2.37  Health1 → Human capital3 → Health4 → Metabolic syndrome  0.000 (0.000)  1.16  Health1 → Human capital3 → Human capital4 → Metabolic syndrome  0.001 (0.001)  1.13  Panel B. Effects of human capital at Waves I, II, and IV      Human capital (WIV) → Metabolic syndrome       Total effect = Direct effect  −0.02 (0.005)***  −3.76  Human capital (WIII) → Metabolic syndrome       Total effect = Indirect effect  −0.03 (0.006)***  −4.83  Specific indirect paths      Human capital3→ Human capital4 → Metabolic syndrome  −0.02 (0.006)***  −3.61  Human capital3 → Health4 → Metabolic syndrome  −0.01 (0.002)***  −3.84  Human capital (WI) → Metabolic syndrome      Total effect = Indirect effect  −0.12 (0.03)***  −4.57  Specific indirect paths      Human capital1 → Human capital3 → Human capital4 → Metabolic syndrome  −0.08 (0.02)***  −3.59  Human capital1→ Human capital3 → Health4 → Metabolic syndrome  −0.03 (0.01)***  −3.81  Human capital1→ Health3 → Human capital4 → Metabolic syndrome  −0.00 (0.00)  −0.26  Human capital1 → Health3 → Health4 → Metabolic syndrome  −0.01 (0.02)  −0.26  Panel A. Effects of poor health at Waves I, III, and IV  Unstandardized β (SE)  β/SE  Poor health (WIV) → Metabolic syndrome       Total effect = Direct effect  0.42 (0.01)***  35.42  Poor health (WIII) → Metabolic syndrome       Total effect = Indirect effect  0.45 (0.01)***  31.75  Specific indirect paths      Health3 → Health4 → Metabolic syndrome  0.45 (0.01)***  31.89  Health3 → Human capital4 → Metabolic syndrome  0.002 (0.001)*  2.46  Poor health (WI) → Metabolic syndrome      Total effect = Indirect effect  0.26 (0.04)***  6.26  Specific indirect paths      Health1 → Health3 → Health4 → Metabolic syndrome  0.26 (0.04)***  6.23  Health1 → Health3 → Human capital4 → Metabolic syndrome  0.001 (0.001)*  2.37  Health1 → Human capital3 → Health4 → Metabolic syndrome  0.000 (0.000)  1.16  Health1 → Human capital3 → Human capital4 → Metabolic syndrome  0.001 (0.001)  1.13  Panel B. Effects of human capital at Waves I, II, and IV      Human capital (WIV) → Metabolic syndrome       Total effect = Direct effect  −0.02 (0.005)***  −3.76  Human capital (WIII) → Metabolic syndrome       Total effect = Indirect effect  −0.03 (0.006)***  −4.83  Specific indirect paths      Human capital3→ Human capital4 → Metabolic syndrome  −0.02 (0.006)***  −3.61  Human capital3 → Health4 → Metabolic syndrome  −0.01 (0.002)***  −3.84  Human capital (WI) → Metabolic syndrome      Total effect = Indirect effect  −0.12 (0.03)***  −4.57  Specific indirect paths      Human capital1 → Human capital3 → Human capital4 → Metabolic syndrome  −0.08 (0.02)***  −3.59  Human capital1→ Human capital3 → Health4 → Metabolic syndrome  −0.03 (0.01)***  −3.81  Human capital1→ Health3 → Human capital4 → Metabolic syndrome  −0.00 (0.00)  −0.26  Human capital1 → Health3 → Health4 → Metabolic syndrome  −0.01 (0.02)  −0.26  Finally, the supplementary analyses exploring the necessity of correcting for endogeneity show a markedly different pattern of results (see appendix D). Relative to the SEM, the uncorrected OLS model tends to underestimate the path coefficients reflecting path dependence and social inequality in health (net of health selection). In most cases, the uncorrected model estimate is roughly half the magnitude of the SEM estimate. Moreover, the uncorrected model overestimates the cross-lags (in most cases, by more than double) and all four cross-lags are statistically significant—suggesting a starkly different substantive result. In other words, had we not corrected for endogeneity, we would have mistakenly placed less emphasis on the role of path dependence, more emphasis on the role of social causation and health selection, and we would have inaccurately concluded that social causation and health selection were operating throughout each stage across the transition to adulthood. Thus, it is necessary, on both statistical and theoretical grounds, to account for the endogeneity of poor health and human capital and include error covariances. Discussion A vast literature documents social inequalities in health and human capital, but less is known about how these inequalities develop across the early life course. We trace the intra-generational development of these emerging pathways as adolescents transition into adulthood. In addition, little is known about the interdependence of human capital and poor physical health during the transition to adulthood. As we detail below, understanding more about inequality in both domains informs a range of sociology subdisciplines, including social stratification, medical sociology, and health inequalities. Furthermore, no research has yet examined the contributions of intra-generational development of poor health and accumulation of human capital on metabolic syndrome—an important marker for cardiovascular disease and Type II diabetes, two conditions that predict significant morbidity and mortality risks for US adults. A primary contribution of our study is the conceptualization and operationalization of pathways of health and human capital across the transition to adulthood. An empirical test of these pathways showed powerful evidence of path dependence: poor health in adolescence is strongly related to poor health in emerging adulthood, which in turn is related to poor health in young adulthood. The same holds for human capital. This in and of itself is not surprising, but results also show that these pathways develop at differential rates over the transition to adulthood. The health pathway crystalizes more slowly between adolescence (WI) and emerging adulthood (WIII), but more rapidly between emerging and young adulthood. Substantively, this suggests that adolescence is a life stage in which health behaviors, such as food choices and levels of physical activity, are still being formulated, whereas by emerging adulthood these behaviors are more entrenched—a finding consistent with past research (Harris et al. 2006). Conversely, the human capital pathway crystalizes strongly in adolescence and solidifies further in adulthood. This provides evidence of the notion proposed by life course scholars, that human agency, along with structural distributions of opportunity, together play an important role in the transition to adulthood as individuals enact human agency (e.g., open a savings/checking account) and begin to formulate health behaviors, both of which can become the basis of longer-term health and economic behaviors and outcomes in adulthood (Hitlin and Kirkpatrick Johnson 2015; Macmillan 2006; Shanahan 2000; Zimmer-Gembeck and Mortimer 2006). Additionally, this finding buttresses support for Heckman’s (2000) call for greater emphasis to be placed on early intervention programs and policies, as human capital pathways are taking shape well before our study first measures them. Moreover, supplementary analyses showed that failing to account for endogeneity would have led us to inaccurately place less emphasis on path dependency within the model, as estimates from uncorrected OLS models were roughly half the magnitude of SEM estimates (that included endogeneity corrections and error covariances). A second key contribution is strong evidence of emerging social stratification in metabolic syndrome for this cohort. Our study extends past research by showing that the estimated effect of young adult human capital on metabolic syndrome is statistically significant (p < 0.001), and robust to the inclusion of parental socioeconomic status, family background characteristics, and underlying social and health selection mechanisms (which we account for using numerous instrumental variables identifying poor health and human capital at each time point). We speculate that the social stratification of metabolic syndrome signals that the interplay between opportunity structures and human agency is already taking hold in the lives of this cohort, influencing a key marker of adult health. It may also reflect learned effectiveness—in that education, even in young adulthood, provides individuals with knowledge and skills that marshal personal control and enact human agency to produce healthy behaviors and a healthy lifestyle (Mirowsky and Ross 2003). Similarly, higher education provides access to advantageous employment opportunities, relative to low-wage work or unemployment, that accommodate/facilitate healthier lifestyles (Ross and Wu 1996). Based on past studies describing rather large social inequalities in health, we anticipated finding a substantial direct effect of human capital on metabolic syndrome. Instead, we observed a relatively small but statistically significant effect (p < 0.001); we speculate that this stems from two sources. Metabolic syndrome is relatively rare early in life, becoming increasingly common with age (prevalence is 3 percent in children [Friend, Craig, and Turner 2013], 23 percent in adults age 19+ [Beltrán-Sánchez et al. 2013], and 44 percent in adults age 50+ [Alexander et al. 2003]). Therefore, detecting any degree of social stratification in metabolic syndrome within young adulthood, above and beyond endogeneity and selection bias, is a strong signal of things to come. Second, we measured metabolic syndrome objectively, using biomarker data, which are independent of self-reflection bias or social circumstances that are often correlated with subjective measures of health. For this reason, estimated effects of objective health indicators are often smaller in magnitude than those based on subjective indicators (Wu et al. 2013). We suspect that the social stratification we document will only increase as cohort members age. Specifically, mid- to late adulthood is a time when morbidity and mortality risks due to health conditions linked with metabolic syndrome, such as cardiovascular disease, stroke, and Type II diabetes, become more pronounced. It is also a time when these conditions become increasingly concentrated among individuals within lower socioeconomic strata. This is influential, as cardiovascular diseases are among the most expensive health conditions: one study estimated that 17 percent of all medical expenditures each year (totaling $149 billion) can be attributed to cardiovascular diseases alone (Trogdon et al. 2007). As these costs place an undue economic burden on the least advantaged, this will further stratify the health and wealth of this cohort as they age. Recent trends related to the prevalence of metabolic syndrome also suggest increased concentration among minority populations, and particularly among minority women (Beltrán-Sánchez et al. 2013; Mozumdar and Liguori 2011). As such, these trends may reinforce race/ethnic social stratification of adult health. Furthermore, given recent trends pointing to an earlier onset of chronic diseases, such as obesity (Van Cleave, Gortmaker, and Perrin 2010), and rising prevalence of metabolic syndrome across the population (Mozumdar and Liguori 2011), the long-term implications of social inequalities in cardiovascular disease may far exceed what we have observed in the past. Observing these developmental trends now, among a contemporary population at the forefront of the obesity epidemic (Harris 2010), affords time to plan and implement prevention and intervention efforts among vulnerable populations. Our work suggests that interventions would be more impactful during these critical early life stages if they can curb metabolic dysregulation before it manifests in more harmful conditions. The strength of the health and human capital pathways, and the importance of early and sustained investments in human capital, permeate our results. But we also show important (both substantively and statistically) cross-lagged effects between poor health and human capital between WIII and WIV (as respondents move into young adulthood). Thus, our study is the first to pinpoint the stage within the transition to adulthood in which social causation and health selection operate. This is the third contribution of our study. Supplementary analyses showed that failing to account for endogeneity would have led to markedly different substantive conclusions, including inaccurately concluding that social causation and health selection operate throughout the entire span of the transition to adulthood, buttressing the importance of correcting for endogeneity in terms of accurately depicting these cross-lagged associations. The standardized coefficients of the social causation (−0.04) and health selection (−0.05) cross-lags were statistically significant at the p < 0.01 level, and could not be constrained to be equal, according to a Wald test. In contrast, other studies demonstrate stronger effects of social causation for mental health in mid-adulthood (Chandola et al. 2003; Elstad and Krokstad 2003; Mulatu and Schooler 2002) and physical health in late adulthood (Warren 2009). Observing both social causation and health selection processes at work in our study is consistent with past scholarship emphasizing the mutual interdependence between the two that operates in conjunction with strong path-dependent components (O’Rand 2001), and resonates with a developmental perspective identifying the transition to adulthood as one in which youth create, form, and maintain social identities, health habits, and economic preferences simultaneously (Macmillan 2006). In this way, human capital and health in the early life course are substantively distinct from human capital and health in mid- to late adulthood—two life stages in which social causation and health selection signal a distinct underlying process. In mid- to late adulthood, poor health directly impacts labor market participation, earnings, and wealth accumulation (health selection). In turn, participation in the labor market affects health (social causation), partially through employment stability (ensuring stable earnings), access to health insurance through one’s job, or through education itself via learned effectiveness (Mirowsky and Ross 2003). In contrast, aptitude and educational achievement are the important drivers of human capital development in the early stage of the life course (as opposed to labor market activity and income) and this development is both dependent upon and predictive of early life health pathways. In sum, this study provides a general model that traces the development and persistence of health and human capital between adolescence and young adulthood. We focused on measures readily available in other datasets both within (e.g., NLSY97) and outside the United States (e.g., China Health and Nutrition Survey) to encourage future research. Future social stratification studies, for example, could trace the pathways through which ascribed characteristics (race-ethnicity, sex) shape inequalities at each time point, through constraints imposed by social structures and social systems. Other model extensions could chart variations in the accumulation of health and human capital by parental characteristics (education, income, occupation) or by interactions with social structures in childhood (family of origin, neighborhood, school), contributing a broader understanding of the development of the intergenerational transmissions of inequality. Exploring nuances in this model as individuals interact with health institutions or health professionals may contribute to the medical sociology literature. We chose to apply the model to understanding social inequalities in cardiovascular health, but future work could apply the model to any number of social well-being and/or health markers in young adulthood, contributing to a number of sociological subdisciplines. For instance, future research could also address how these pathways operate differentially across geographic location (contributing to urban and rural sociology), or how these pathways are disrupted by engagement in criminal activity (contributing to sociology of crime and deviance) or timing of family formation behavior (contributing to family sociology). We have in mind here the development of the Blau-Duncan model, where solving a measurement issue (operationalizing socioeconomic status) and laying out a causal framework facilitated a generation of comparable research findings. Extensions of this study’s framework across numerous subareas of sociology could enrich our understanding of the longer-term social processes associated with the development of physical health and human capital development. Notes 1 We top-coded weight for nine respondents whose weight exceeded the scale maximum (200 kg/440 lb). Results were unchanged when these individuals were excluded. 2 Pairwise correlations between latent variables:   1  2  3  4  5  6  1. Poor Health–WI  1            2. Poor Health–WIII  0.85  1          3. Poor Health–WIV  0.79  0.93  1        4. Human Capital–WI  −0.15  −0.11  −0.16  1      5. Human Capital–WIII  −0.08  −0.09  −0.14  0.91  1    6. Human Capital–WIV  −0.11  −0.13  −0.16  0.87  0.95  1    1  2  3  4  5  6  1. Poor Health–WI  1            2. Poor Health–WIII  0.85  1          3. Poor Health–WIV  0.79  0.93  1        4. Human Capital–WI  −0.15  −0.11  −0.16  1      5. Human Capital–WIII  −0.08  −0.09  −0.14  0.91  1    6. Human Capital–WIV  −0.11  −0.13  −0.16  0.87  0.95  1  3 Our measure diverges slightly from the clinical definition. Add Health participants were not asked to fast pre-interview. Rather than combine fasting and non-fasting glucose, we use HbA1c, a more stable measure of metabolic dysregulation. We also use membership in the highest decile for triglycerides; the clinical definition is indicated by ≥150 mg/dL or drug treatment for elevated triglycerides. For HDL-C, clinical cutoffs are <50mg/dL for women and <40 mg/dL for men, or drug treatment for reduced HDL-C. The laboratory assaying Add Health specimens for lipids used two different assays—a common occurrence. The two should be directly comparable using a simple algebraic transformation. After extensive data cleaning and checking efforts were performed, only the rank-ordering (by deciles) was released, as it was deemed to be a more reliable measure. Thus, we indicate reduced HDL-C as membership in the lowest category for women (HDL-C = 1; 7.0%) and the lowest two categories for men (HDL-C < 3; 26.2%), based on evidence suggesting that 11.9% of women and 31.4% of men exhibit reduced HDL-C (Carroll, Kit, and Lacher 2012). Results did not change when we defined reduced HDL-C by membership in the lowest two categories for women (HDL-C < 3; 15.6%) and the lowest three categories for men (HDL-C < 4; 37.2%). 4 When this was missing, respondent’s report of parental education was used. 5 The BLS cost price index is only available for urban areas; thus, we control for Wave III residence in an urban census tract. 6 Model fit indices presented indicate acceptable fit (RMSEA = 0.026; SRMR = 0.050). Other model fit indices (CFI, TLI) are below recommended levels. However, residual variances of latent variables are low (range, 0.01–0.17) and R-squared values of latent variables are high (all but one is between 0.50 and 0.97), lending further evidence of adequate model fit. In accordance with recommendations for testing structural equation models (Bollen and Long 1993), we balanced these indications of model fit with indications based on theory and substantive experience when developing our model. 7 Supplementary analyses (not shown) suggested that adding pathways from each WI latent variable to metabolic syndrome did not significantly improve model fit. About the Authors Jennifer B. Kane is an assistant professor in the Department of Sociology at the University of California–Irvine. Her research focuses on social inequality, its emergence over the life course, and the role of the family of origin and the early life environment in shaping social and health disparities. She has published recent work in Demography, Population Research and Policy Review, Journal of Health and Social Behavior, and Journal of Marriage and Family. Kathleen Mullan Harris is the James Haar Distinguished Professor of Sociology at the University of North Carolina–Chapel Hill. Harris’s research focuses on social inequality and health. Harris is Director and PI of the National Longitudinal Study of Adolescent to Adult Health, a longitudinal study of more than 20,000 teens who are being followed into young adulthood. She has published recent work in Demography, Nature, and Proceedings of the National Academy of Sciences. S. Philip Morgan is the Allan Feduccia Distinguished Professor of Sociology at the University of North Carolina–Chapel Hill. Morgan’s research focuses on change and variation in the human family, with special attention to human fertility. His work on fertility in the United States examines fertility levels, fertility timing, and high levels of nonmarital childbearing. Morgan has published recent work in Annual Review of Sociology, Demography, and Population and Development Review. David Guilkey is the Cary C. Boshamer Distinguished Professor of Economics at the University of North Carolina–Chapel Hill. The main focus of his interest is the development and use of estimation methods that can be used to analyze large-survey datasets with limited dependent variables, especially when endogenous right-hand-side variables are present. He has published recent work in Demography, Journal of Applied Econometrics, and Studies in Family Planning. Supplementary Material Supplementary material is available at Social Forces online. References Adler, Nancy, Nicole R. Bush, and Matthew S. Pantell. 2012. “ Rigor, Vigor, and the Study of Health Disparities.” Proceedings of the National Academy of Sciences  109( Supplement 2): 17154– 59. Google Scholar CrossRef Search ADS   Adler, Nancy E., and Joan M. Ostrove. 1999. “ Socioeconomic Status and Health: What We Know and What We Don’t.” Annals of the New York Academy of Sciences  896( 1): 3– 15. Google Scholar CrossRef Search ADS   Alexander, Charles M., Pamela B. Landsman, Steven M. Teutsch, and Steven M. Haffner. 2003. “ NCEP-Defined Metabolic Syndrome, Diabetes, and Prevalence of Coronary Heart Disease among NHANES III Participants Age 50 Years and Older.” Diabetes  52( 5): 1210– 14. Google Scholar CrossRef Search ADS   Almond, Douglas, and Janet Currie. 2011. “ Killing Me Softly: The Fetal Origins Hypothesis.” Journal of Economic Perspectives  25( 3): 153– 72. Google Scholar CrossRef Search ADS   Amato, Paul, Nancy S. Landale, Tara C. Havasevich-Brooks, Alan Booth, David J. Eggebeen, Robert Schoen, and Susan M. McHale. 2008. “ Precursors of Young Women’s Family Formation Pathways.” Journal of Marriage and Family  70( 5): 1271– 86. Google Scholar CrossRef Search ADS   Aneshensel, Carol S. 1992. “ Social Stress: Theory and Research.” Annual Review of Sociology  18: 15– 38. Google Scholar CrossRef Search ADS   Bauldry, Shawn, Michael J. Shanahan, Jason D. Boardman, Richard A. Miech, and Ross Macmillan. 2012. “ A Life Course Model of Self-Rated Health through Adolescence and Young Adulthood.” Social Science & Medicine  75( 7): 1311– 20. Google Scholar CrossRef Search ADS   Becker, Gary S. 1962. “ Investment in Human Capital: A Theoretical Analysis.” Journal of Political Economy  70( 5): 9– 49. Google Scholar CrossRef Search ADS   Beltrán-Sánchez, Hiram, Michael O. Harhay, Meera M. Harhay, and Sean McElligott. 2013. “ Prevalence and Trends of Metabolic Syndrome in the Adult US Population, 1999–2010.” Journal of the American College of Cardiology  62( 8): 697– 703. Google Scholar CrossRef Search ADS   Ben-Shlomo, Y. and D. Kuh. 2002. “ A Life Course Approach to Chronic Disease Epidemiology: Conceptual Models, Empirical Challenges and Interdisciplinary Perspectives.” International Journal of Epidemiology  31( 2): 285– 93. Google Scholar CrossRef Search ADS   Bircher, Johannes. 2005. “ Towards a Dynamic Definition of Health and Disease.” Medicine, Health Care, and Philosophy  8( 3): 335– 41. Google Scholar CrossRef Search ADS   Blackwell, Debra L., Mark D. Hayward, and Eileen M. Crimmins. 2001. “ Does Childhood Health Affect Chronic Morbidity in Later Life?” Social Science & Medicine  52( 8): 1269– 84. Google Scholar CrossRef Search ADS   Blau, Peter M., and Otis Dudley Duncan. 1967. The American Occupational Structure . New York: Free Press. Bollen, Kenneth A., and J. Scott Long. 1993. Testing Structural Equation Models . Newbury Park, CA: Sage Publications. Carroll, M. D., B. K. Kit, and D. A. Lacher. 2012. “Total and High-Density Lipoprotein Cholesterol in Adults: National Health and Nutrition Examination Survey, 2009–2010.” NCHS Data Brief ( 92): 1– 8. Chandola, Tarani, Mel Bartley, Amanda Sacker, Crispin Jenkinson, and Michael Marmot. 2003. “ Health Selection in the Whitehall II Study, UK.” Social Science & Medicine  56( 10): 2059– 72. Google Scholar CrossRef Search ADS   Chen, Feinian, Yang Yang, and Guangya Liu. 2010. “ Social Change and Socioeconomic Disparities in Health over the Life Course in China: A Cohort Analysis.” American Sociological Review  75( 1): 126– 50. Google Scholar CrossRef Search ADS   Coleman, James S. 1988. “ Social Capital in the Creation of Human Capital.” American Journal of Sociology  94: S95– S120. Google Scholar CrossRef Search ADS   Crosnoe, Robert. 2006. “ Health and the Education of Children from Racial/Ethnic Minority and Immigrant Families.” Journal of Health and Social Behavior  47( 1): 77– 93. Google Scholar CrossRef Search ADS   Currie, Janet, and Brigitte C. Madrian. 1999. “ Health, Health Insurance and the Labor Market.” Handbook of Labor Economics  3: 3309– 3416. Google Scholar CrossRef Search ADS   DiPrete, Thomas A., and Gregory M. Eirich. 2006. “ Cumulative Advantage as a Mechanism for Inequality: A Review of Theoretical and Empirical Developments.” Annual Review of Sociology  32: 271– 97. Google Scholar CrossRef Search ADS   Dupre, Matthew E. 2007. “ Educational Differences in Age-Related Patterns of Disease: Reconsidering the Cumulative Disadvantage and Age-as-Leveler Hypotheses.” Journal of Health and Social Behavior  48( 1): 1– 15. Google Scholar CrossRef Search ADS   Dupre, Matthew E. 2008. “ Educational Differences in Health Risks and Illness over the Life Course: A Test of Cumulative Disadvantage Theory.” Social Science Research  37( 4): 1253– 66. Google Scholar CrossRef Search ADS   Elder, Glen H. Jr. 1995. “The Life Course Paradigm: Social Change and Individual Development.” In Examining Lives in Context: Perspectives on the Ecology of Human Development , edited by Phyllis Moen, Glen H. Elder Jr., and Kurt Lüscher, pp. 101– 39. Hyattsville, MD: American Psychological Association. Google Scholar CrossRef Search ADS   Elder, Glen H. Jr., Monica Kirkpatrick Johnson, and Robert Crosnoe. 2003. “The Emergence and Development of Life Course Theory.” In Handbook of the Life Course , edited by Jeylan T. Mortimer and Michael J. Shanahan, pp. 3– 19. New York: Springer. Google Scholar CrossRef Search ADS   Elman, Cheryl, and Angela M. O’Rand. 2007. “ The Effects of Social Origins, Life Events, and Institutional Sorting on Adults’ School Transitions.” Social Science Research  36( 3): 1276– 99. Google Scholar CrossRef Search ADS   Elstad, Jon Ivar, and Steinar Krokstad. 2003. “ Social Causation, Health-Selective Mobility, and the Reproduction of Socioeconomic Health Inequalities over Time: Panel Study of Adult Men.” Social Science & Medicine  57( 8): 1475– 89. Google Scholar CrossRef Search ADS   Featherman, David L., and Robert Mason Hauser. 1978. Opportunity and Change . New York: Academic Press. Ferraro, Kenneth F., and Jessica A. Kelley-Moore. 2003. “ Cumulative Disadvantage and Health: Long-Term Consequences of Obesity?” American Sociological Review  68( 5): 707. Google Scholar CrossRef Search ADS   Fox, John W. 1990. “ Social Class, Mental Illness, and Social Mobility: The Social Selection-Drift Hypothesis for Serious Mental Illness.” Journal of Health and Social Behavior  31( 4): 344– 53. Google Scholar CrossRef Search ADS   Friend, Amanda, Leone Craig, and Steve Turner. 2013. “ The Prevalence of Metabolic Syndrome in Children: A Systematic Review of the Literature.” Metabolic Syndrome and Related Disorders  11( 2): 71– 80. Google Scholar CrossRef Search ADS   Furstenberg, Frank F. 2000. “ The Sociology of Adolescence and Youth in the 1990s: A Critical Commentary.” Journal of Marriage and Family  62( 4): 896– 910. Google Scholar CrossRef Search ADS   Furstenberg, Frank F., Jeanne Brooks-Gunn, and S. Philip Morgan. 1989. Adolescent Mothers in Later Life . New York: Cambridge University Press. Furstenberg, Frank F., Ruben G. Rumbaut, and Richard A. Settersten. 2005. “On the Frontier of Adulthood: Emerging Themes and New Directions.” In On the Frontier of Adulthood: Theory, Research, and Public Policy , edited by Richard A. Settersten, Frank F. Furstenberg, and Ruben G. Rumbaut. Chicago: University of Chicago Press. Google Scholar CrossRef Search ADS   George, Linda K. 2005. “ Socioeconomic Status and Health across the Life Course: Progress and Prospects.” Journals of Gerontology Series B: Psychological Sciences and Social Sciences  60( Special Issue 2): S135– 39. Google Scholar CrossRef Search ADS   Goldman, Noreen. 2001. “ Social Inequalities in Health.” Annals of the New York Academy of Sciences  954( 1): 118– 39. Google Scholar CrossRef Search ADS   Gordon-Larsen, Penny, Melissa C. Nelson, Phil Page, and Barry M. Popkin. 2006. “ Inequality in the Built Environment Underlies Key Health Disparities in Physical Activity and Obesity.” Pediatrics  117( 2): 417– 24. Google Scholar CrossRef Search ADS   Grundy, Scott M., James I. Cleeman, Stephen R. Daniels, Karen A. Donato, Robert H. Eckel, Barry A. Franklin, David J. Gordon, Ronald M. Krauss, Peter J. Savage, and Sidney C. Smith. 2005. “ Diagnosis and Management of the Metabolic Syndrome: An American Heart Association/National Heart, Lung, and Blood Institute Scientific Statement.” Circulation  112( 17): 2735– 52. Google Scholar CrossRef Search ADS   Grusky, D., ed. 2008. Social Stratification: Class, Race, and Gender in Sociological Perspective , 3rd ed. Philadelphia: Westview Press. Guilkey, D. K., and P. M. Lance. 2014. “Program Impact Estimation with Binary Outcome Variables: Monte Carlo Results for Alternative Estimators and Empirical Examples.” In Festschrift in Honor of Peter Schmidt , edited by Robin C. Sickles and William C. Horrace, pp. 5– 46. New York: Springer. Google Scholar CrossRef Search ADS   Haas, S. A. 2006. “ Health Selection and the Process of Social Stratification: The Effect of Childhood Health on Socioeconomic Attainment.” Journal of Health and Social Behavior  47( 4): 339– 54. Google Scholar CrossRef Search ADS   Haas, S. A. 2008. “ Trajectories of Functional Health: The ‘Long Arm’of Childhood Health and Socioeconomic Factors.” Social Science & Medicine  66( 4): 849– 61. Google Scholar CrossRef Search ADS   Hallinan, Maureen T. 1988. “ Equality of Educational Opportunity.” Annual Review of Sociology  14: 249– 68. Google Scholar CrossRef Search ADS   Harris, K. M. 2010. “ An Integrative Approach to Health.” Demography  47( 1): 1– 22. Google Scholar CrossRef Search ADS   Harris, K. M., C. T. Halpern, E. Whitsel, J. Hussey, J. Tabor, P. Entzel, and J. R. Udry. 2009. “The National Longitudinal Study of Adolescent Health: Research Design.” http://www.cpc.unc.edu/projects/addhealth/design. Carolina Population Center, University of North Carolina–Chapel Hill. Harris, Kathleen Mullan, Penny Gordon-Larsen, Kim Chantala, and J. Richard Udry. 2006. “ Longitudinal Trends in Race/Ethnic Disparities in Leading Health Indicators from Adolescence to Young Adulthood.” Archives of Pediatrics & Adolescent Medicine  160( 1): 74– 81. Google Scholar CrossRef Search ADS   Hayward, Mark D., and Bridget K. Gorman. 2004. “ The Long Arm of Childhood: The Influence of Early-Life Social Conditions on Men’s Mortality.” Demography  41( 1): 87– 107. Google Scholar CrossRef Search ADS   Heckman, James J. 2000. “ Policies to Foster Human Capital.” Research in Economics  54( 1): 3– 56. Google Scholar CrossRef Search ADS   Hertzman, Clyde. 2006. “ The Biological Embedding of Early Experience and Its Effects on Health in Adulthood.” Annals of the New York Academy of Sciences  896( 1): 85– 95. Google Scholar CrossRef Search ADS   Hertzman, Clyde, and Chris Power. 2006. “A Life Course Approach to Health and Human Development.” In Healthier Societies: From Analysis to Action , edited by Jody Heymann, Clyde Hertzman, Morris L. Barer, and Robert G. Evans, pp. 83– 106. New York: Oxford University Press. Google Scholar CrossRef Search ADS   Hitlin, Steven, and Glen H. Elder. 2007. “ Time, Self, and the Curiously Abstract Concept of Agency.” Sociological Theory  25( 2): 170– 91. Google Scholar CrossRef Search ADS   Hitlin, Steven, and Monica Kirkpatrick Johnson. 2015. “ Reconceptualizing Agency within the Life Course: The Power of Looking Ahead.” American Journal of Sociology  120( 5): 1429– 72. Google Scholar CrossRef Search ADS   Hogan, Dennis P., and Nan Marie Astone. 1986. “ The Transition to Adulthood.” Annual Review of Sociology  12: 109– 30. Google Scholar CrossRef Search ADS   Hout, Michael. 2007. “ Otis Dudley Duncan’s Major Contributions to the Study of Social Stratification.” Research in Social Stratification and Mobility  25( 2): 109– 18. Google Scholar CrossRef Search ADS   Hout, Michael, and Thomas A. DiPrete. 2006. “ What We Have Learned: RC28’s Contributions to Knowledge about Social Stratification.” Research in Social Stratification and Mobility  24( 1): 1– 20. Google Scholar CrossRef Search ADS   Huber, Machteld, J. André Knottnerus, Lawrence Green, Henriëtte van der Horst, Alejandro R. Jadad, Daan Kromhout, Brian Leonard, Kate Lorig, Maria Isabel Loureiro, and Jos W M van der Meer. 2011. “ How Should We Define Health?” BMJ—British Medical Journal  343( 6): d4163. Google Scholar CrossRef Search ADS   Hummer, Robert A., Richard G. Rogers, Charles B. Nam, and Felicia B. LeClere. 1999. “ Race/Ethnicity, Nativity, and US Adult Mortality.” Social Science Quarterly  80( 1): 136– 53. Jackson, Margot I. 2009. “ Understanding Links between Adolescent Health and Educational Attainment.” Demography  46( 4): 671– 94. Google Scholar CrossRef Search ADS   Kawachi, Ichiro, Nancy E. Adler, and William H. Dow. 2010. “ Money, Schooling, and Health: Mechanisms and Causal Evidence.” Annals of the New York Academy of Sciences  1186( 1): 56– 68. Google Scholar CrossRef Search ADS   Kawachi, Ichiro, and Lisa F. Berkman. 2003. Neighborhoods and Health . New York: Oxford University Press. Google Scholar CrossRef Search ADS   Keister, Lisa A. 2003. “ Religion and Wealth: The Role of Religious Affiliation and Participation in Early Adult Asset Accumulation.” Social Forces  82( 1): 175– 207. Google Scholar CrossRef Search ADS   Kuh, D., and Y. B. Shlomo. 2004. A Life Course Approach to Chronic Diseases Epidemiology , vol. 2. Oxford: Oxford University Press. Google Scholar CrossRef Search ADS   Link, B. G., and J. Phelan. 1995. “ Social Conditions as Fundamental Causes of Disease.” Journal of Health and Social Behavior  ( Extra Issue): 80– 94. Looker, E. Dianne. 1989. “ Accuracy of Proxy Reports of Parental Status Characteristics.” Sociology of Education  62( 4): 257– 76. doi:10.2307/2112830. Google Scholar CrossRef Search ADS   Lynch, John W., George A. Kaplan, and Sarah J. Shema. 1997. “ Cumulative Impact of Sustained Economic Hardship on Physical, Cognitive, Psychological, and Social Functioning.” New England Journal of Medicine  337( 26): 1889– 95. Google Scholar CrossRef Search ADS   Macmillan, Ross. 2006. “‘Constructing Adulthood’: Agency and Subjectivity in the Transition to Adulthood.” Advances in Life Course Research  11: 3– 29. Marmot, M., M. Shipley, E. Brunner, and H. Hemingway. 2001. “ Relative Contribution of Early Life and Adult Socioeconomic Factors to Adult Morbidity in the Whitehall II Study.” Journal of Epidemiology and Community Health  55( 5): 301– 7. Google Scholar CrossRef Search ADS   McEwen, Bruce S. 1998. “ Stress, Adaptation, and Disease: Allostasis and Allostatic Load.” Annals of the New York Academy of Sciences  840( 1): 33– 44. Google Scholar CrossRef Search ADS   Merton, Robert K. 1968. “ The Matthew Effect in Science.” Science (New York, N.Y.)  159( 3810): 56– 63. Google Scholar CrossRef Search ADS   Mirowsky, John, and Catherine E. Ross. 2003. Education, Social Status, and Health . New York: Aldine de Gruyter. Mortimer, Jeylan T. 1994. “Individual Differences as Precursors of Youth Unemployment.” In Youth, Employment, and Society , edited by A. C. Peterson and Jeylan T. Mortimer, pp. 172– 98. New York: Cambridge University Press. Google Scholar CrossRef Search ADS   Mortimer, Jeylan T. 2003. Working and Growing Up in America . Cambridge, MA: Harvard University Press. Mortimer, Jeylan T., Jeremy Staff, and Jennifer C. Lee. 2005. “ Agency and Structure in Educational Attainment and the Transition to Adulthood.” Advances in Life Course Research  10: 131– 53. Google Scholar CrossRef Search ADS   Mouw, Ted. 2005. “Sequences of Early Adult Transitions: How Variable Are They, and Does It Matter?” In On the Frontier of Adulthood: Theory, Research, and Public Policy , edited by Richard Settersten, Frank F. Furstenberg Jr., and Ruben G. Rumbaut, pp. 256– 91. Chicago: University of Chicago Press. Google Scholar CrossRef Search ADS   Mozumdar, Arupendra, and Gary Liguori. 2011. “ Persistent Increase of Prevalence of Metabolic Syndrome among US Adults: NHANES III to NHANES 1999–2006.” Diabetes Care  34( 1): 216– 19. Google Scholar CrossRef Search ADS   Mroz, T. A. 1999. “ Discrete Factor Approximations in Simultaneous Equation Models: Estimating the Impact of a Dummy Endogenous Variable on a Continuous Outcome.” Journal of Econometrics  92( 2): 233– 74. Google Scholar CrossRef Search ADS   Mulatu, Mesfin Samuel, and Carmi Schooler. 2002. “ Causal Connections between Socio-Economic Status and Health: Reciprocal Effects and Mediating Mechanisms.” Journal of Health and Social Behavior  43( 1): 22– 41. Google Scholar CrossRef Search ADS   O’Rand, Angela M. 2001. “Stratification and the Life Course: The Forms of Life-Course Capital and Their Interrelationships.” In Handbook of Aging and the Social Sciences , edited by R. H. Binstock and Linda K. George, pp. 197– 213. San Diego, CA: Academic Press. O’Rand, Angela M. 2009. “Cumulative Processes in the Life Course.” In The Craft of Life Course Research , edited by Glen H. Elder Jr. and Janet Z. Giele, pp. 121– 40. New York: Guilford Press. Palloni, Alberto. 2006. “ Reproducing Inequalities: Luck, Wallets, and the Enduring Effects of Childhood Health.” Demography  43( 4): 587– 615. Google Scholar CrossRef Search ADS   Pollitt, Ricardo A., Kathryn M. Rose, and Jay S. Kaufman. 2005. “ Evaluating the Evidence for Models of Life Course Socioeconomic Factors and Cardiovascular Outcomes: A Systematic Review.” BMC Public Health  5( 1): 7. Google Scholar CrossRef Search ADS   Poulton, Richie, Avshalom Caspi, Barry J. Milne, W. Murray Thomson, Alan Taylor, Malcolm R. Sears, and Terrie E. Moffitt. 2002. “ Association between Children’s Experience of Socioeconomic Disadvantage and Adult Health: A Life-Course Study.” Lancet  360( 9346): 1640– 45. Google Scholar CrossRef Search ADS   Power, C., and C. Hertzman. 1997. “ Social and Biological Pathways Linking Early Life and Adult Disease.” British Medical Bulletin  53( 1): 210– 21. Google Scholar CrossRef Search ADS   Rodgers, Willard L., Mary Beth Ofstedal, and A. Regula Herzog. 2003. “ Trends in Scores on Tests of Cognitive Ability in the Elderly US Population, 1993–2000.” Journals of Gerontology Series B: Psychological Sciences and Social Sciences  58( 6): S338– 46. Google Scholar CrossRef Search ADS   Ross, Catherine E., and Chia-Ling Wu. 1996. “ Education, Age, and the Cumulative Advantage in Health.” Journal of Health and Social Behavior  37( 1): 104– 20. Google Scholar CrossRef Search ADS   Sandefur, Gary D., Jennifer Eggerling-Boeck, and Hyunjoon Park. 2005. “Off to a Good Start? Postsecondary Education and Early Adult Life.” In On the Frontier of Adulthood: Theory, Research, and Public Policy , edited by Richard Settersten, Frank F. Furstenberg Jr., and Ruben G. Rumbaut, pp. 292– 319. Chicago: University of Chicago Press. Google Scholar CrossRef Search ADS   Sargan, John D. 1958. “ The Estimation of Economic Relationships Using Instrumental Variables.” Econometrica: Journal of the Econometric Society  26( 3): 393– 415. Google Scholar CrossRef Search ADS   Sartorius, Norman. 2006. “ The Meanings of Health and Its Promotion.” Croatian Medical Journal  47( 4): 662. Sewell, William H., and Robert M. Hauser. 1975. Education, Occupation, and Earnings. Achievement in the Early Career . New York: Academic Press. Shanahan, Michael J. 2000. “ Pathways to Adulthood in Changing Societies: Variability and Mechanisms in Life Course Perspective.” Annual Review of Sociology  26: 667– 92. Google Scholar CrossRef Search ADS   Singh, Gopal K., and Barry A. Miller. 2003. “ Health, Life Expectancy, and Mortality Patterns among Immigrant Populations in the United States.” Canadian Journal of Public Health—Revue canadienne de santé publique  95( 3): I14– 21. Smith, James P., and Raynard Kington. 1997. “ Demographic and Economic Correlates of Health in Old Age.” Demography  34( 1): 159– 70. Google Scholar CrossRef Search ADS   Stock, James H., Jonathan H. Wright, and Motohiro Yogo. 2002. “ A Survey of Weak Instruments and Weak Identification in Generalized Method of Moments.” Journal of Business & Economic Statistics  20( 4): 518– 29. Google Scholar CrossRef Search ADS   Teachman, Jay D., Kathleen Paasch, and Karen Carver. 1997. “ Social Capital and the Generation of Human Capital.” Social Forces  75( 4): 1343– 59. Google Scholar CrossRef Search ADS   Thoits, Peggy A. 1995. “ Stress, Coping, and Social Support Processes: Where Are We? What Next?” Journal of Health and Social Behavior  ( Extra Issue): 53– 79. Trogdon, Justin G., Eric A. Finkelstein, Isaac A. Nwaise, Florence K. Tangka, and Diane Orenstein. 2007. “ The Economic Burden of Chronic Cardiovascular Disease for Major Insurers.” Health Promotion Practice  8( 3): 234– 42. Google Scholar CrossRef Search ADS   Van Cleave, Jeanne, Steven L. Gortmaker, and James M. Perrin. 2010. “ Dynamics of Obesity and Chronic Health Conditions among Children and Youth.” JAMA: The Journal of the American Medical Association  303( 7): 623– 30. Google Scholar CrossRef Search ADS   Warren, John Robert. 2009. “ Socioeconomic Status and Health across the Life Course: A Test of the Social Causation and Health Selection Hypotheses.” Social Forces  87( 4): 2125– 53. Google Scholar CrossRef Search ADS   Willson, Andrea E., Kim M. Shuey, and Glen H. Elder Jr. 2007. “ Cumulative Advantage Processes as Mechanisms of Inequality in Life Course Health.” American Journal of Sociology  112( 6): 1886– 1924. Google Scholar CrossRef Search ADS   Wu, Shunquan, Rui Wang, Yanfang Zhao, Xiuqiang Ma, Meijing Wu, Xiaoyan Yan, and Jia He. 2013. “ The Relationship between Self-Rated Health and Objective Health Status: A Population-Based Study.” BMC Public Health  13( 1): 320. Google Scholar CrossRef Search ADS   Zimmer-Gembeck, Melanie J., and Jeylan T. Mortimer. 2006. “ Selection Processes and Vocational Development: A Multi-Method Approach.” Advances in Life Course Research  11: 121– 48. Google Scholar CrossRef Search ADS   Author notes This research uses data from Add Health, a program project directed by Kathleen Mullan Harris and designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris at the University of North Carolina–Chapel Hill, and funded by grant P01-HD31921 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, with cooperative funding from 23 other federal agencies and foundations. Special acknowledgment is due to Ronald R. Rindfuss and Barbara Entwisle for assistance on the original design. Information on how to obtain the Add Health data files is available on the Add Health website (http://www.cpc.unc.edu/addhealth). This research received support from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (K99 HD075860, PI: Kane; T32 HD007168, PI: Halpern; P2C HD050924, PI: Morgan). Opinions reflect those of the authors and not necessarily those of the granting agencies. Direct correspondence to Jennifer B. Kane, Department of Sociology, University of California–Irvine, 4171 Social Sciences Plaza A, Irvine, CA 92697; phone: (949) 824-9594; e-mail: jbkane@uci.edu. © The Author 2017. Published by Oxford University Press on behalf of the University of North Carolina at Chapel Hill. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.

Journal

Social ForcesOxford University Press

Published: Mar 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off