The Unequal Benefits of Upward Mobility

Lauren Gaydosh1, Kathleen Mullan Harris1, Kristen Schorpp1,

Sara McLanahan2, Edith Chen3, and Greg Miller3

Background

The socioeconomic gradient in health in the United States is persistent, and increasing over time.1–5 Individuals in the top 1% of the income distribution can expect to live 10-15 years longer than those in the bottom 1%.4 Individuals with a college degree can also expect to outlive their less educated counterparts by about a decade.5 However, higher socioeconomic status (SES) is not uniformly beneficial for all individuals; there is evidence that the education payoff is not as steep for minority groups compared to non-Hispanic whites,5,6 and may in some cases actually be associated with worse health.7 Furthermore, it is not only achieved SES in adulthood that shapes health; childhood SES is strongly predictive of childhood health and endures across the life course.8,9

There is emerging evidence that the interaction between childhood and adult SES may actually have important implications for health. In a set of papers by Brody, Chen, Miller and colleagues, the authors document a pattern that they refer to as “skin-deep resilience”, wherein rural from severely disadvantaged backgrounds who manage to achieve upward mobility demonstrate lower rates of psychosocial problems, but poorer physiological health compared to African Americans from similar backgrounds who remain disadvantaged.7,10,11 Two studies suggest that this relationship may exist in other populations and subgroups beyond rural African Americans. Research from the Dunedin birth cohort in New

1 Carolina Population Center, University of North Carolina at Chapel Hill 2 Office of Population Research, Princeton University 3 Department of Psychology, Northwestern University DRAFT – PLEASE DO NOT CITE OR CIRCULATE - 1

Zealand finds that upwardly mobile individuals have similar or worse health compared to stably low status individuals.12 A study among American adults documents a greater prevalence of chronic conditions among individuals with some college education compared to those with only a high school education, but this study did not explore the role of upward mobility.13 In this project we examine the health consequences of individual trajectories of upward socioeconomic mobility in a nationally representative sample of American adults.

Upward mobility may be associated with multiple stressors that could have health consequences. First, the sustained level of effort, self-control, and single-mindedness required to achieve upward mobility from a highly disadvantaged setting may be stressful and physiologically taxing.7,10,11 In African Americans, such high-effort coping strategies have been referred to as “John Henryism”.14 There is mixed evidence as to how sustained vigilance relates to health outcomes, with variation by physiological and psychological outcomes,15–17 as well as by gender.18 Second, when individuals from disadvantaged backgrounds achieve upward mobility, it is likely that the higher SES environment in which they find themselves differs greatly from their social environment of origin.19,20 Such incongruence may lead to feelings of isolation, and experiences of .15,21–23 Upward mobility may also weaken or sever social ties, leading to a lack of social support, which is predictive of poor health.24,25 Third, both sustained effort and social isolation may lead to greater levels of perceived , which has been demonstrated to affect health.26,27 Upwardly mobile individuals may also feel that their achieved position is tenuous,28,29 and perceived stress may reflect perceptions of threat. Finally, stress associated with upward mobility may lead to unhealthy coping behaviors that vary by gender.30,31 Indeed, Jackson and colleagues find that the coping behaviors (smoking, drinking, poor diet) employed by black individuals may serve to buffer against psychological disorders

DRAFT – PLEASE DO NOT CITE OR CIRCULATE - 2 while creating physiological health costs.32 These findings underscore the need to examine physiological, psychological, and cognitive health separately.

We investigate the health consequences of upward mobility using nationally representative integrated demographic, social, contextual and biological data. We hypothesize that upward mobility is physiologically taxing, resulting in poorer physiological health. We extend this research by examining other race/ethnic groups that have not been included in prior research.7,10,15 To our knowledge, ours is the first study to test “skin-deep resilience” in a nationally representative, racially and ethnically diverse sample.

Understanding the health risks associated with upward mobility will provide a more nuanced understanding of the relationship between SES and health. Documenting the health consequences of social mobility in early adulthood provides a foundation from which to understand different aging trajectories for those from disadvantaged backgrounds that begin during the transition to adulthood. Furthermore, documenting and explaining the physiological health costs associated with upward mobility can inform the development of interventions to reduce stress and avoid the presumptive biological costs associated with upward mobility, thereby encouraging healthy aging.

Data and Methods

Data - We rely on data from the National Longitudinal Study of Adolescent to Adult Health

(Add Health). Add Health is an ongoing national longitudinal study of the social, behavioral, and biological linkages in health and developmental trajectories from early adolescence into adulthood. The data are representative of American adolescents in grades 7-12 in 1994-1995.

The initial sample included 20,745 adolescents aged 12-20; since the start of the study,

DRAFT – PLEASE DO NOT CITE OR CIRCULATE - 3 participants have been interviewed in home at four data collection waves. At Wave IV in 2008-

2009, respondents were aged 24-32 (n=15,701, 80.3% response rate) and asked to participate in biological specimen collection (over 95% provided specimens, almost 15,000). These data are particularly well-suited to address the specific aims described here, as they include detailed family, contextual, health and biological measures. Furthermore, the Add Health sample is diverse, with an oversample of particular ethnicities/races enabling an exploration of heterogeneity of the influence of upward mobility. Finally, the physiological biomarkers of health risk examined here are predictive of health before disease is manifest, permitting a window for potential aging interventions.

Childhood Disadvantage – To measure childhood disadvantage, we construct a count of 27 binary indicators that capture cumulative exposure to household, school, and neighborhood disadvantage over childhood and/or during adolescence (Wave I). Household disadvantage indicators include a binary indicator of cumulative family instability across childhood and adolescence, low parent education (less than high school), the bottom quartile of household income, household welfare receipt in the past month, parent unemployment, and parent-reported difficulty paying bills.

Neighborhood disadvantage indicators were taken from the 1990 U.S. Census to best approximate neighborhood conditions during Wave I of the Add Health study. Neighborhood disadvantage measures include the tract-level proportion of households receiving welfare, proportion of unemployed adults, proportion of households below poverty line, proportion of adults with less than a high school education, proportion female headed households, proportion black residents, proportion vacant homes, and the county-level infant mortality rate and violent

DRAFT – PLEASE DO NOT CITE OR CIRCULATE - 4 crime rate. Each item was recoded so those residing in neighborhoods at the top quartile of the distribution were coded as disadvantaged. Parental reports of neighborhood conditions or problems were also included in the index, including parent-reported problems with neighborhood litter, neighborhood drug use, and desire to move away from the neighborhood.

Finally, indicators of school disadvantage at Wave I included school-level aggregated measures of the proportion of households receiving welfare, the proportion of unemployed parents, the proportion of single-parent households, and the proportion of parents with less than a high school education. All items were recoded as binary indicators, with the top quartile coded as disadvantaged. School disadvantage was also captured using Wave I school administrator reports of grade retention, the school dropout rate, class sizes, the proportion teachers with a Master’s degree, and daily school attendance. Consistent with other items in the index, school administrator items were recoded as binary indicators, with the top quartile of grade retention, dropout rate, and class size coded as disadvantaged, and the bottom quartile of teachers with an

MA and daily school attendance coded as disadvantaged.

We sum all of the indicators to create a score ranging from 0 to 25 with a mean of 5.7.

We standardize the score, so that the coefficients associated with the disadvantage index can be interpreted as the change in health risk associated with a one standard deviation increase in disadvantage.

Adult Achievement - We measure adult achievement with an indicator for whether the respondent has completed any college education by Wave IV (age 25-34).

DRAFT – PLEASE DO NOT CITE OR CIRCULATE - 5

Adult Health Outcomes - We analyze the adult physical health of Add Health respondents using four measures derived from biomarker assessments taken at the Wave IV interview. We include four measures because it is plausible that the physiological consequences of achievement may manifestly differently across multiple biological systems. We analyze high-sensitivity C-reactive protein (hsCRP) as a measure inflammation, as a measure of cardiovascular health risk, and body mass index (BMI) and metabolic syndrome as indicators of metabolic risk.

We use a continuous measure of hsCRP, transforming the values by taking the natural logarithm to normalize the distribution. Individuals are considered hypertensive if their measured meets clinical thresholds, they report ever having been diagnosed with hypertension, or report taking anti-hypertensives in the medication inventory (~25% of the sample). Body mass index is taken from measured height and weight at the time of interview.

Finally, metabolic syndrome is a composite measure of five biomarkers - blood pressure, blood glucose, high-density lipoprotein (HDL) cholesterol, triglycerides, and waist circumference. An individual’s value for each biomarker is classified as high risk if it exceeds the guidelines established by the National Cholesterol Education Program Expert Panel. If an individual has three or more high risk values, they are classified as having metabolic syndrome (~10%). Add

Health data collection procedures and biomarker validation are available elsewhere (Hussey et al. 2015; Nguyen et al. 2011; Whitsel et al. 2012).

Mechanisms - Sustained vigilance is measured using measures of academic striving at Wave I, and a perseverance personality scale measured at Wave IV. We test the role of social isolation using several measures, including perceived discrimination and social support, measured at

Wave IV. We also construct a measure of dissimilarity between the school attended in

DRAFT – PLEASE DO NOT CITE OR CIRCULATE - 6 adolescence and college attended in young adulthood, using rich data on school demographics such as size and racial composition. Similar measures are constructed for neighborhoods of residence in childhood and adulthood. Respondent’s perception of stress is measured, using the

Cohen perceived stress scale measured at Wave IV. Finally, health behaviors are measured by smoking, alcohol consumption, drug use, diet, and exercise, measured at intervening Wave III when respondents were aged 18-26.

Methods – We use regression analysis with appropriate specification depending on the outcome under investigation to determine the association between childhood disadvantage, adult achievement, and health in adulthood. CRP and BMI are modeled continuously using linear regression. Hypertension and metabolic syndrome are modeled using logistic regression. All results use weights to adjust for the sampling strategy of Add Health.

Preliminary Findings

In Table 1, we examine the association between childhood disadvantage, college attendance, and each health risk measure. The interaction between the disadvantage index and college attendance tests for whether the benefit of college attendance varies by level of childhood disadvantage. In the full Add Health sample, we find that college attendance is negatively associated with health risk across all measures; on average, individuals who attend any college have better health than those who do not attend college. Conversely, the disadvantage index is positively associated with health risk across all measures; adults from disadvantaged childhood backgrounds have worse health on average than adults from more advantaged backgrounds. However, the returns to

DRAFT – PLEASE DO NOT CITE OR CIRCULATE - 7 college attendance are lower for individuals from disadvantaged backgrounds; the relationship is statistically significant for BMI and metabolic syndrome, and marginally significant for CRP.

In Table 2, we restrict the analysis to black respondents. Similar to the results for the full sample, we find that, at increasing levels of childhood disadvantage, college attendance is associated with greater health risk. The interaction between childhood disadvantage and adult achievement is statistically significant for CPR and BMI, and marginally significant for metabolic syndrome. The crossover between disadvantage and achievement can be seen more clearly in Figure 1. Adults from highly disadvantaged backgrounds who attend some college have higher levels of inflammation, higher BMI, and are more likely to have metabolic syndrome.

We present the results for the white subsample in Table 3; the findings are mixed, and there is support for a significant interaction between disadvantage and achievement for metabolic syndrome (marginally significant). The crossover between those who attend college and those who do not can be seen in Figure 2. The coefficients are in the opposite direction for CRP and hypertension.

To briefly summarize, these preliminary analyses provide suggestive evidence that upward mobility is not uniformly beneficial in terms of health risk, particularly among black respondents. In future analyses, we will test the potential mechanisms outlined above – sustained vigilance, social isolation, discrimination, and social support.

DRAFT – PLEASE DO NOT CITE OR CIRCULATE - 8

Table 1. Adult Health Risk, Childhood Disadvantage, and Adult Achievement - Add Health Full Sample lnCRP BMI Hypertension Metabolic Syndrome Age 0.02+ 0.02+ 0.15* 0.15* 0.06** 0.06** 0.07* 0.07* (0.01) (0.01) (0.06) (0.06) (0.02) (0.02) (0.03) (0.03) Male -0.57*** -0.57*** -0.30 -0.30 0.73*** 0.73*** 0.36*** 0.36*** (0.04) (0.04) (0.23) (0.22) (0.07) (0.07) (0.07) (0.07) College attendance -0.29*** -0.25*** -1.68*** -1.44*** -0.22** -0.25*** -0.55*** -0.47*** (0.04) (0.04) (0.23) (0.24) (0.07) (0.07) (0.12) (0.11) Disadvantage index (DI) 0.06* 0.04 0.62*** 0.50*** 0.12*** 0.13*** 0.20*** 0.15*** (0.02) (0.02) (0.13) (0.14) (0.03) (0.03) (0.04) (0.04) DI x College attendance 0.10+ 0.61* -0.10 0.32** (0.05) (0.26) (0.08) (0.11) n 8633 9578 9435 8300 *** p<0.001, ** p<0.01, * p<0.05, + p<0.10. Standard errors in parentheses

DRAFT – PLEASE DO NOT CITE OR CIRCULATE - 9

Table 2. Adult Health Risk, Childhood Disadvantage, and Adult Achievement - Add Health Black Subsample lnCRP BMI Hypertension Metabolic Syndrome Age 0.02 0.01 0.02 0.00 0.09** 0.09** 0.05 0.04 (0.03) (0.03) (0.13) (0.13) (0.03) (0.03) (0.05) (0.05) Male -0.68*** -0.69*** -3.22*** -3.27*** 0.26+ 0.26+ -0.20 -0.22 (0.11) (0.11) (0.54) (0.53) (0.14) (0.13) (0.18) (0.18) College attendance -0.02 -0.21+ -0.34 -1.20+ -0.45** -0.55** -0.13 -0.40 (0.10) (0.11) (0.48) (0.60) (0.15) (0.19) (0.18) (0.26) Disadvantage index (DI) 0.05 -0.02 0.20 -0.10 -0.02 -0.05 0.10 0.02 (0.05) (0.05) (0.28) (0.34) (0.06) (0.07) (0.09) (0.11) DI x College attendance 0.29** 1.37* 0.16 0.37+ (0.10) (0.61) (0.15) (0.21) n 1830 2094 2059 1722 *** p<0.001, ** p<0.01, * p<0.05, + p<0.10. Standard errors in parentheses

DRAFT – PLEASE DO NOT CITE OR CIRCULATE - 10

Table 3. Adult Health Risk, Childhood Disadvantage, and Adult Achievement - Add Health White Subsample lnCRP BMI Hypertension Metabolic Syndrome Age 0.02+ 0.02 0.18* 0.18* 0.07* 0.06* 0.09* 0.09* (0.01) (0.01) (0.08) (0.08) (0.03) (0.03) (0.04) (0.04) Male -0.54*** -0.54*** 0.09 0.09 0.86*** 0.86*** 0.58*** 0.58*** (0.04) (0.04) (0.23) (0.23) (0.09) (0.09) (0.11) (0.11) College attendance -0.31*** -0.34*** -1.81*** -1.77*** -0.20* -0.27* -0.66*** -0.44** (0.05) (0.07) (0.26) (0.34) (0.08) (0.12) (0.15) (0.16) Disadvantage index (DI) 0.06+ 0.06+ 0.51* 0.50* 0.15** 0.16** 0.12+ 0.08 (0.03) (0.04) (0.20) (0.21) (0.06) (0.05) (0.07) (0.08) DI x College attendance -0.05 0.05 -0.11 0.37+ (0.09) (0.43) (0.15) (0.21) n 5062 5510 5469 4886 *** p<0.001, ** p<0.01, * p<0.05, + p<0.10. Standard errors in parentheses

DRAFT – PLEASE DO NOT CITE OR CIRCULATE - 11

DRAFT – PLEASE DO NOT CITE OR CIRCULATE - 12

DRAFT – PLEASE DO NOT CITE OR CIRCULATE - 13

References 1. Elo, I. & Preston, S. Educational differentials in mortality: United States, 1979–1985. Soc. Sci. Med. (1996). 2. Williams, D. Socioeconomic differentials in health: A review and redirection. Soc. Psychol. Q. (1990). 3. Kitagawa, E. & Hauser, P. Differential mortality in the United States: A study in socioeconomic . (1973). 4. Chetty, R. et al. The Association Between Income and Life Expectancy in the United States, 2001-2014. JAMA (2016). doi:10.1001/jama.2016.4226 5. Hummer, R. A. & Hernandez, E. M. The Effect of Educational Attainment on Adult Mortality in the United States. Popul. Bull. 68, 1–16 (2013). 6. Goldman, N., Kimbro, R. T., Turra, C. M. & Pebley, A. R. Socioeconomic gradients in health for white and Mexican-origin populations. Am. J. Public Health 96, 2186–2193 (2006). 7. Chen, E., Miller, G. E., Brody, G. H. & Lei, M. Neighborhood Poverty, College Attendance, and Diverging Profiles of Substance Use and Allostatic Load in Rural African American Youth. Clin. Psychol. Sci. a J. Assoc. Psychol. Sci. 3, 675–685 (2015). 8. Case, A., Lubotsky, D. & Paxson, C. Economic status and health in childhood: The origins of the gradient. Am. Econ. Rev. 92, 1308–1334 (2002). 9. Miller, G. E. & Chen, E. The biological residue of childhood poverty. Child Dev. Perspect. 7, 67–73 (2013). 10. Miller, G. E., Yu, T., Chen, E. & Brody, G. H. Self-control forecasts better psychosocial outcomes but faster epigenetic aging in low-SES youth. Proc. Natl. Acad. Sci. 112, 201505063 (2015). 11. Brody, G. H. et al. Is resilience only skin deep?: rural African Americans’ socioeconomic status-related risk and competence in preadolescence and psychological adjustment and allostatic load at age 19. Psychol. Sci. 24, 1285–93 (2013). 12. Poulton, R. et al. Association between children’s experience of socioeconomic disadvantage and adult health: a life-course study. Lancet (London, England) 360, 1640–5 (2002). 13. Zajacova, A., Rogers, R. & Johnson-Lawrence, V. Glitch in the gradient: Additional education does not uniformly equal better health. Soc. Sci. Med. (2012). 14. James, S., Hartnett, S. & Kalsbeek, W. John Henryism and blood pressure differences among black men. J. Behav. Med. (1983). 15. Hudson, D., Neighbors, H. W., Geronimus, A. T. & Jackson, J. S. Racial Discrimination, John Henryism, and Depression Among African Americans. J. Black Psychol. 0095798414, (2015). 16. Bonham, V., Sellers, S. L. & Neighbors, H. W. John Henryism and self-reported physical health among high-socioeconomic status African American men. Am. J. Public Health 94, 737–738 (2004). 17. Bronder, E. C., Speight, S. L., Witherspoon, K. M. & Thomas, A. J. John Henryism, depression, and perceived social support in Black women. J. Black Psychol. 40, 115–137 (2014). 18. Dressler, W. W., Oths, K. S. & Gravlee, C. C. Race and ethnicity in public health research: models to explain health disparities. Annu. Rev. Anthr. 34, 231–252 (2005).

DRAFT – PLEASE DO NOT CITE OR CIRCULATE - 14

19. Reay, D., Crozier, G. & Clayton, J. ‘Strangers in paradise’? Working-class students in elite universities. Sociology (2009). 20. Jetten, J., Iyer, A., Tsivrikos, D. & Young, B. M. When is individual mobility costly? The role of economic and social identity factors. Eur. J. Soc. Psychol. 38, 866–879 (2008). 21. Cole, E. R. & Omari, S. R. Race, Class and the Dilemmas of Upward Mobility for African Americans. J. Soc. Issues 59, 785–802 (2003). 22. Geronimus, A. T., Hicken, M., Keene, D. & Bound, J. ‘Weathering’ and age patterns of allostatic load scores among blacks and whites in the United States. J. Inf. 96, 826–833 (2006). 23. McClure, H. et al. Discrimination-Related Stress, Blood Pressure and Epstein-Barr Virus Antibodies Among Latin American Immigrants in Oregon, Us. J. Biosoc. Sci. 42, 433–461 (2010). 24. Berkman, L. & Glass, T. Social integration, social networks, social support, and health. Soc. Epidemiol. (2000). 25. Yang, Y. C. et al. Social relationships and physiological determinants of longevity across the human life span. Proc. Natl. Acad. Sci. 113, 201511085 (2016). 26. McEwen, B. S. Brain on stress: how the social environment gets under the skin. Proc. Natl. Acad. Sci. U. S. A. 109 Suppl , 17180–5 (2012). 27. Vasunilashorn, S., Lynch, S. M., Glei, D. a, Weinstein, M. & Goldman, N. Exposure to Stressors and Trajectories of Perceived Stress Among Older Adults. J. Gerontol. B. Psychol. Sci. Soc. Sci. 70, 1–9 (2014). 28. McBrier, D. & Wilson, G. Going down? Race and downward occupational mobility for white-collar workers in the 1990s. Work Occup. (2004). 29. Pattillo, M. Black picket fences: Privilege and peril among the black middle class. (2013). 30. Elliott, G. & Eisdorfer, C. Stress and human health. (1982). 31. Skalamera, J. & Hummer, R. Educational attainment and the clustering of health-related behavior among US young adults. Prev. Med. (Baltim). (2015). 32. Jackson, J., Knight, K. M. & Rafferty, J. A. Race and unhealthy behaviors: chronic stress, the HPA axis, and physical and mental health disparities over the life course. Am. J. Public Health 100, 933–939 (2010). 33. Miech, R., Pampel, F., Kim, J. & Rogers, R. G. The Enduring Association between Education and Mortality: The Role of Widening and Narrowing Disparities. Am. Sociol. Rev. 76, 913–934 (2011). 34. Smith, J. Unraveling the SES: Health Connection on JSTOR. Popul. Dev. Rev. 30, 108– 132 (2004). 35. Kiecolt-Glaser, J. K., McGuire, L., Robles, T. F. & Glaser, R. : psychological influences on immune function and health. J. Consult. Clin. Psychol. 70, 537–547 (2002). 36. Feinstein, L., Ferrando-Martinez, S. & Aiello, A. E. Variation in a Novel Measure of Immune Status by Sociodemographic Characteristics and Health Status: Findings from the Detroit Neighborhood Health Study. in Annual Meeting of the Population Association of America (2016). 37. Franceschi, C. et al. Inflamm‐aging: an evolutionary perspective on immunosenescence. Ann. N. Y. Acad. Sci. 908, 244–254 (2000). 38. Austin, P. C. An Introduction to Propensity Score Methods for Reducing the Effects of Confounding in Observational Studies. Multivariate Behav. Res. 46, 399–424 (2011).

DRAFT – PLEASE DO NOT CITE OR CIRCULATE - 15

39. Freedman, D. A. & Berk, R. A. Weighting regressions by propensity scores. Eval. Rev. 32, 392–409 (2008).

DRAFT – PLEASE DO NOT CITE OR CIRCULATE - 16