Abstract
Using time-varying, prospectively measured income in a nationally representative sample of Baby-Boomer men (the National Longitudinal Survey of Youth – 1979 [NLSY79]), we identify eight group-based trajectories of income between ages 25–49 and use multinomial treatment models to describe the associations between group-based income trajectories and mental and physical health at midlife. We find remarkable rigidity in income trajectories: less than 25% of our sample experiences significant upward or downward mobility between the ages of 25 to 49 and most who move remain or move into poverty. Men’s physical and mental health at age fifty is strongly associated with their income trajectories, and some upwardly mobile men achieve the same physical and mental health as the highest earning men after adjusting for selection. The worse physical and mental health of men on other income trajectories is largely attributable to their early life disadvantages, health behaviors, and cumulative work experiences.
Decades of research have demonstrated a relationship between socioeconomic status (SES) and health at middle age (Elo 2009; Haas 2007; Hayward and Gorman 2004; Link and Phelan 1995). But there remains vigorous debate about which measures of SES—income, wealth, occupation, education—best capture the relationship between SES and health (see for example Adler et al. 2008; Kawachi, Adler, and Dow 2010). Because income often varies from year to year, both education and wealth are often the preferred markers of SES when evaluating the SES-health relationship (Bond Huie et al. 2003). Yet income may tell us important information about the role of liquid financial resources for health over the life course. Moreover, accounting for income’s variability over time may provide a more accurate depiction of people’s changing lives than cumulative or fixed measures like wealth and education, as economic stability has become less commonplace today (Kalleberg 2009). A focus on income trajectories – both how much men earn each year and how these earnings change with time – would allow for an evaluation of how upwardly mobile men, downwardly mobile men, and men at different levels of relative income stability fare in their midlife mental and physical health. Thus, this paper’s goal is to identify and estimate men’s trajectories of income over time, to explore the relationships between men’s income trajectories and mental and physical health at middle-age, to take into account many of the factors that place men onto these income trajectories, and to identify the mechanisms linking income trajectories to health at midlife.
Such an investigation is also important because the relationships between income and physical and mental health likely reflect gendered expectations that define masculinity in the United States (Townsend 2002). Men continue to face strong gender expectations that they will be the family breadwinners during their working years (Coltrane 2004; Townsend 2002). Yet while the family wage—through which one worker, presumed to be the husband, could support an entire family—has long held cultural importance in the United States (May 1987), jobs that provided such stability were scarce even when the family wage was first introduced and have grown less common with time (Kalleberg 2009; May 1987). Thus, men’s consistent presence in the paid workforce across their lives, the primacy of paid work for men’s identities, and the downward trends in workers’ likelihood to maintain stable employment and a “family wage” make men particularly well-suited for a study of the relationships between income trajectories and health at midlife.
Although, to our knowledge, no prior research has identified group-based pathways of men’s longitudinal income trajectories to examine their relationships with midlife health, the income-health gradient is well established for both mental and physical health (Elo 2009; Elo and Preston 1996; Schnittker 2004). For example, higher levels of income are protective when one becomes ill (Smith 2007), and functional limitations are greater at middle-age when incomes are lower (Zimmer and House 2003). Higher levels of income are also protective of a decrease in functional ability as well as an increase in chronic conditions (Herd, Goesling, and House 2007). More clarification is needed on the degree to which income affects health both contemporaneously (providing access to resources at a single point in time) and cumulatively, through the accumulated socioeconomic advantages or disadvantages reflected in men’s earnings from prior years (Deaton 2002; Elo 2009). If we were to find evidence that upwardly mobile men achieve the same health at midlife as men who stably maintain the highest earnings over the same period, this would suggest men benefit both from conforming to the consistent and stable male breadwinner norm and from attaining higher earnings later in life. In other words, there may be health premiums for men who achieve high incomes by midlife, even if these men first enter the workforce relatively disadvantaged. By the same logic, there may also be health penalties for men who experience stagnant income or income that declines as men age.
Health selection complicates any analysis of how income influences health. The relationship between health and income is reciprocal and multifaceted: poor mental and physical health early in life restricts socioeconomic attainment later on, often due to truncated education and poorer employment prospects (Elo 2009; Haas 2006; Haas, Glymour, and Berkman 2011; Miech 1999). Income itself is causally associated with health through direct and indirect access to resources and absolute and relative social status (Adler et al., 2008; Link and Phelan, 1995). Our focus is on the causal relationships between income trajectories and health, adjusting for unequal selection into different income trajectories during early adulthood. We add to existing literature by identifying the number, direction, and shape of men’s group-based income trajectories over time and by evaluating how group-based income trajectories are associated with men’s mental and physical health at midlife. Using the National Longitudinal Survey of Youth – 1979 (NLSY79), we investigate the following research questions: What are men’s group-based income trajectories across their twenties, thirties, and forties? Net of early life health, work-family, and socioeconomic characteristics that select men into these income trajectories, what are the relationships between men’s income trajectories and their mental and physical health at age fifty? Finally, through what mechanisms do long-term pathways of income influence health at middle age?
Life Course and Income Pathways
Understanding the relationships between men’s income trajectories and health at middle age is essential because the greatest socioeconomic variation in long-term, chronic health problems emerges during this life course stage (Adler and Stewart 2010; House 1994; House, Lantz, and Herd 2005). A life course perspective would lead us to anticipate that trajectories of earnings leading up to midlife may provide important insight into the relationships between income and health overlooked by prior studies focusing on income measured at a single point in time or across two time points. A life course framework suggests that health and income should be traced as men enter the labor market and face opportunities and constraints in maintaining their employment and income as they age (Elder, Johnson, and Crosnoe 2003). Following income across the life course allows life course scholars to demonstrate the ways that initially small income differences when men are young may accumulate over time, leading to widening financial and health disparities at midlife. Men may also have expectations that their incomes will increase continuously as they age; deviations from this expectation may be associated with worse health.
Current research would lead us to anticipate that some men achieve and maintain relatively high income status across the life course, while others remain mired in relatively low income status across the life course (Kalleberg 2009). Job loss and unemployment can cause irreparable downward income trajectories (Mishel et al. 2012; Stevens 1997), while some men see their fortunes improve across adulthood with increasing income as they complete schooling, are promoted, or advance upwards along their career trajectory (Moen and Roehling 2005). But the exact scope of group-based income trajectories is yet unknown. Thus, our first aim is to identify group based developmental trajectories of men’s income between ages 25 and 49.
Income, Income Trajectories, and Mental and Physical Health
Prior research documenting the physical health benefits of higher incomes often draw on a fundamental cause perspective (Link and Phelan 1995; Phelan, Link, and Tehranifar 2010), which argues, in part, that better social conditions cause better health across numerous health outcomes, across a range of mechanisms, and as old diseases are cured and new ones emerge vis-à-vis access to monetary and nonmonetary resources that are unequally allocated according to social status. Those with higher incomes can use their monetary resources to engage in preventative health care, purchase better health insurance, save money to respond to health shocks, and purchase more technologically advanced care, resulting in wider socioeconomic health disparities for more preventable or better-understood health conditions (Phelan et al. 2004; Rubin, Clouston, and Link 2014). The more advantaged also have greater access to nonmonetary resources that are associated with better health, including greater health knowledge, stronger social support networks, the ability to avoid unhealthy activities and engage in health-promoting behaviors, and higher-quality neighborhoods and housing. The benefits of high status also extend to personal psychosocial resources: higher socioeconomic status is also associated with higher levels of life satisfaction, greater self-efficacy, and lower levels of stress and depression (Ross and Wu 1995; Thoits 2010). Fundamental cause scholars argue those with high status can activate these resources as needed to treat illness, prevent future diseases, and gain knowledge regarding health treatment and health care (Link and Phelan 1995; Phelan and Link 2005).
The combination of monetary, personal, and structural resources enjoyed by those with higher SES may help to explain why the health benefits of high status accumulate with time. Yet relatively little research has examined fundamental cause in the context of changing socioeconomic standing over the life course, as much of the recent work drawing on the fundamental cause perspective uses county-level measures of socioeconomic status to study aggregate geographic disparities (Rubin et al. 2014; Saldana-Ruiz et al. 2013) or uses education, and not income, as the sole indicator of social status (Landes 2017; Masters, Link, and Phelan 2015). Studies that do take a long view of social standing use childhood socioeconomic resources to predict later life health (Ferraro, Schafer, and Wilkinson 2015), draw on community studies that combine men and women (Johnson-Lawrence, Galea, and Kaplan 2015), measure dichotomous indicators of economic hardship over time (Willson and Shuey 2016), or compare health disparities across cohorts as causes of death and disease change with time (Masters et al. 2015). While there has been scant longitudinal research using a fundamental cause perspective with income as the primary indicator of social standing, longitudinal investigations of trends in career progressions suggest that upward occupational mobility is critically important to later health (Cundiff et al. 2017; Pavalko, Elder, and Clipp 1993; Smith and Frank 2005; Tiffin, Pearce, and Parker 2005). It is likely that the long-term relationship between income trajectories and health play out in two ways: one, the cumulative gains over time (see Prus (2007) for a discussion of cumulative benefits to health and Amick et al. (2002) for a discussion of cumulative effects of work environments) and two, the fulfillment of social expectations of upward progression in careers (Pavalko et al. 1993; Smith and Frank 2005). In their investigation of men’s career progression over the course of their lives, Pavalko et al. (1993) found that men had shorter lifespans when they either experienced job change without career advancement or experienced early promotions but no additional progress later in their careers. Furthermore, men who did not achieve work equal to their educational attainment saw a decline in their self-rated health over a four-year period (Smith and Frank 2005).
Combined, this body of research suggests that relative changes in occupational status and the accompanying changes in income have lasting effects on men’s health. This raises the question of whether long-term trajectories of income work the same way: do workers whose income remains low, or those whose income stagnates at a medium or low level, or those who experience income declines have poorer health in middle-age than those with rising incomes, or those whose income is comparatively high and remains high across the life course? Group-based trajectories offer a novel test of fundamental cause: Examined as trajectories of income, long-term high incomes ensure stability in access to the direct and indirect health-promoting resources, leading us to expect that those with the highest stable incomes will report better health than those with lower levels of stable income or downward income trajectories. In contrast, rising tides may lift all boats—or at least may lead to better health at middle-age.
But do these different income trajectories also explain differences in mental health? The better mental health observed among those with higher income has historically been attributed to stress differences across socioeconomic status (Thoits 2010). Stress and mental health gradients, much like the physical health gradient, are unequally distributed such that those with the lowest socioeconomic status report the highest levels of stress and the fewest personal resources to cope with these stressors (Turner and Avison 2003; Thoits 2010; Turner and Lloyd 1999; Turner 1995). Some of the specific domains where status differences in mental health are most acute, particularly for men, relate to employment status, job stability, and job characteristics. Temporary and unstable employment, low job satisfaction, and perceived job insecurity are consistently associated with worse mental health as measured by depressive symptoms, burnout, anxiety disorders, antidepressant use, and fatigue (Benach et al. 2014; Burgard and Seelye 2017; Faragher, Cass, and Cooper 2005). Jobs that are characterized by repetitive work, low control, and few opportunities for advancement are similarly associated with higher job stress and lower job satisfaction, undermining mental health (Tausig and Fenwick 2011). Men who experienced downward occupational mobility over their life course report poorer self-rated mental health (Tiffin et al. 2005). Those who had unstable employment and later achieved stable employment found greater mental health benefits when compared to those who remained unstably employed and those whose work trajectories further soured (Virtanen et al. 2005). Taken together, fundamental cause research on physical health and social stress research on mental health, lead us to our first hypothesis:
H1: We anticipate that men who earn the highest stable incomes will report better mental and physical health at midlife than men with stable middle, stable low (H1a), or declining incomes (H1b).
Very little existing research tests for the health benefits of upward mobility (see Pavalko et al. 1993), but the research on the health benefits of creative and challenging work, as well as the well documented negative effects of bad work (see Burgard and Lin (2013) for a review), lead us to expect that increasing incomes are likely to indicate jobs that are better for physical and mental health, and that declining or stagnant incomes may indicate jobs that are (at least relative to high earners) more harmful for physical and mental health. This leads to our second hypothesis:
H2: Upwardly mobile men (men whose incomes rise with time) will report mental and physical health similar to that of the stable highest earning men.
Early Life Predictors
A life course framework further suggests that men’s access to steady and well-paying jobs is constrained by broader social and structural forces that shape both their income and their health (Elder et al. 2003). Thus, it is important to account for early life-course predictors of income trajectories over time, including information about family of origin; a respondent’s own demographic, education, and occupational characteristics; and details of the local labor market context, as these factors may all shape men’s health as well as their ability to earn a high and a steady income in later years.
Respondent Characteristics:
Educational attainment is perhaps the most important predictor of both income and health (Elo 2009; Schnittker 2004). On average, those with higher education levels have higher incomes and are more likely to have better health in later life (Elo 2009). But one’s own educational attainment is related to one’s cognitive skill (Haas and Fosse 2008) and one’s parents’ education (albeit in a U-shaped curve, see Torche 2011). Those who experienced early poverty are at higher risk of poverty in adulthood (Musick and Mare 2004, 2006). Occupation in young adulthood is also a key form of social stratification, through which individuals gain access to income as well as social status and standing, as well as other employer-provided benefits, such as health insurance (Krueger and Burgard 2011). Yet even while some working conditions provide income and/or health benefits, some occupations increase stress, while others increase the likelihood that workers will be injured or even may die on the job (Nelson et al. 2005; Schieman, Glavin, and Milkie 2009; Tausig 1999). Thus distinguishing between different types of occupations early in men’s careers may prove important for both income and health. Race-ethnicity and nativity shape both men’s occupational attainment as well as their employment status and their health, independent of their income (Bertrand and Mullainathan 2004; Kalleberg 2011; Pager, Bonikowski, and Western 2009). Those with work-limiting health conditions early in adulthood may end up in jobs with poorer working conditions, which may further deteriorate their health (Haas 2006). Men’s young adult marital status and number of residential children may also play a role in both their incomes and their health later in life (Carr and Springer 2010; Nomaguchi, Milkie, and Bianchi 2005).
Local Labor Market Context:
As Alwin and McCammon (2003) note, life course studies are often deliberately nested in cohorts so that the respondents will have experienced similar periods of history, but there remains great variation in their lives, including, we argue, where they live and the labor market conditions there. Earnings are likely to be at least partially influenced by the type of employment available when men make early attempts to find a full-time job (McCall 2001). Wages may be depressed when men live in areas with higher unemployment rates and experiencing unemployment may depress their future wage potential (McCall 2001; Mishel et al. 2012). Area union rates indicate the likelihood that the labor market has “good jobs” with better wages and better benefits that may influence long-term earnings, health, and marital prospects (Schneider and Reich 2014). Those living in rural areas may also face higher rates of nonemployment and poverty (Findeis and Jensen 1998; Slack and Jensen 2002).
Mechanisms linking income pathways and health
Just as it is important to account for the factors that may lead men to earn different incomes across their adult lives, it is also important to identify some of the mechanisms that may help to account for the relationships between group-based income trajectories and health at midlife. Scholars continue to note that growing inequalities at work appear tied to growing differences in family formation by socioeconomic status (Cherlin 2014; Silva 2013). Thus, differences in pathways of income may grow in tandem with differences in family formation, particularly in the likelihood to be married and have children, as well as differences in the number of children and the age of childbearing (Cherlin 2014; McLanahan and Percheski 2008; Sweeney and Phillips 2004). Men’s health behaviors and health status are also reflective of their work experiences and early life context and may help to explain the poorer health of lower earning men; men with lower status jobs may be more likely to smoke, drink, or report other behaviors that both reflect their lower status and further damage their health at midlife (Krueger and Chang 2008). Finally, rising incomes may lead to a cumulative earnings advantage over time, but they also likely signal an entrance into “good jobs” that have less risk of unemployment, higher job satisfaction, and better benefits (Kalleberg 2009). Conversely, the jobs that pay the least are likely less satisfying and less secure (Clawson and Gerstel 2014).
The Cumulative Benefits of Family and Household Resources:
The relationship between work and men’s health may also be tied to the financial resources and social support offered by family, as men who are married experience better mental and physical health than those who are not, and men who are higher earners are more likely to be married (Carr and Springer, 2010; Williams, Frech, and Carlson, 2010). The well-established wealth premiums associated with marriage provide additional financial stability as well as a resource for responding to health shocks (Lee and Kim 2003; Vespa and Painter 2011) Household size at age 50 is a limited indicator of household financial needs and poverty risk, depending both on how many children men have had and also on the age at which they start bearing children—both of which are tied to SES (Lanjouw and Ravallion 1995; Nau, Dwyer, and Hodson 2015; Sweeney 2002; Sweeney and Raley 2014).
The Cumulative Benefits of Health Behaviors and Health Status:
Health behaviors are also linked to both mental and physical health and structured by income: Poor mental health and high stress are strongly related to engagement in behaviors that aid in stress-coping, but are reflective of distress and harm physical health in the long-term (Jackson, Knight, and Rafferty 2010; Krueger and Chang 2008). For example, smoking and heavy drinking provide short term relief from stress among those with low status, but these same behaviors, along with physical inactivity, are associated with both mental distress and worsening physical health over time (Jackson et al. 2010; Krueger and Chang 2008; Prince et al. 2007). Shorter sleep duration is associated with job insecurity and financial strain, and can lead to greater inflammation and impaired immunity (Faraut et al. 2012). This means that we would expect those with lower incomes to have worse health behaviors than their higher income peers, exacerbating any initial differences in mental and physical health attributable to economic differences. Health insurance can help to maintain good health and treat new or ongoing conditions. Finally, the relationship between mental and physical health – that poorer mental health can cause poorer physical health and vice versa – is well established in prior research (Farmer and Ferraro 1997; Prince et al. 2007).
Job Benefits and Strains:
High and stable income pathways may be associated with cumulative financial and job resources that are associated with better mental and physical health, including access to important employer benefits including a retirement plan, and access to full-time hours (see Kalleberg 2011). “Bad jobs” often come both with poor pay, few prospects to raise one’s income, and with little to no job stability nor benefits (Kalleberg 2011). Unemployment also exerts a long-term impact on men’s mental and physical health as well as their long-term earnings even after adjusting for current income (Mishel et al. 2012; Stevens 1997; Strully 2009). Low levels of job satisfaction similarly undermine mental health and physical health (Faragher et al. 2005). This leads to our final hypothesis:
H3: The cumulative health benefits associated with members of the highest group-based income trajectory will be partially attributable to family resources, health behaviors, and job benefits and strains.
DATA AND METHODS
Data and sample
We used the National Longitudinal Survey of Youth – 1979 to evaluate relationships between group-based income trajectories and health at midlife. The NLSY79 began as a nationally representative sample of 12,686 14–21 year olds in 1979 who were interviewed yearly through 1994 and every other year through the most recently released wave of data from 2014 (the survey is ongoing). We first limited our sample to the 6,403 NLSY79 men because of earnings, wage, and workforce participation differences across gender; gender differences in expectations to work; and the potential resulting relationships between income pathways and health. Men who were part of oversampled groups that were excluded from the NLSY after 1984 were not included here as they could not report an age-50 health module (n=638 men from the military oversample, n=731 from the poor white male oversample). Men who were not interviewed at age 50 because they were deceased (n=343), in the military (n=2) or incarcerated (n=28) were also excluded. We also excluded men who attrited prior to reporting an age fifty health module (n=936) or who refused to answer age 50 health questions (n=39), as well as men who didn’t know or refused to disclose their income at any wave (n=25), men whose income was zero at every wave (n=29), men who could not be matched with the Geocode data (n=6) and men who were unable to work at age 25 (n=21). At each wave, we excluded person-year observations where men earned zero income due to unemployment, disability, or time out of the workforce.1 Our final sample included 3,605 men, contributing a total of 37,545 person-years of information about their income between the ages of 25–49 during calendar years 1982–2014. Health was measured at the NLSY age 50 health module, during calendar years 2008–2014. Men in our sample contributed an average of 11.20 out of a possible 13 waves of data.
Measures
Time-varying income was constructed at each wave from respondents’ reports of personal income from all jobs over the last calendar year, which we adjusted to 2014 dollars and logged to correct for positive skew. We restructured the time-varying income data to compare men’s income at different ages rather than calendar years. In other words, rather than all men reporting income from 1982–2014 and then adjusting for the age variation in the sample, we restructured the data by age so that men born in 1957 reported “age 25” income in 1982, men born in 1958 reported “age 25” income in 1983, and so on, until men born in 1964 reported “age 49” income in 2014, the most recent survey wave used in our group-based trajectory models. We used income measures for men at or near ages 25, 27, 29, 31, 33, 35, 37, 39, 41, 43, 45, 47, and 49 for a maximum of thirteen observations per respondent.
Health at age 50 was evaluated using the Short-Form 12 (SF-12), version one. The twelve questions (listed in Appendix 1) use a proprietary method to assign separate scores for both mental health and physical health with scores ranging from 0–100 (National Longitudinal Surveys n.d.).2 For both mental and physical health, a score of “50” corresponds to the average mental or physical health of a US adult. Higher scores indicated better health (Brazier et al. 1992; Gandek et al. 1998). The index evaluated respondents’ reports of pain, mobility, self-rated health, emotional distress, feelings of calm, and whether emotional problems get in the way of work or family life.
Selection equation variables, measured at or prior to the start of income trajectories at age 25, have established associations with income attainment or financial hardship in adulthood (see for example Ferraro, Schafer, and Wilkinson, (2015)) and were used to adjust for selection into group-based income trajectories while testing Hypotheses 1 and 2. These variables included race-ethnicity, US nativity, cognitive ability (using the AFQT-80 revised scale), maternal education, poverty status in young adulthood, whether the respondent lived with both biological parents at 14, educational attainment by age 25, occupation at age 25 (including categories for students and those out of the workforce), county-level unemployment rates during young adulthood (calculated using the NLSY Geocoded data merged with historic Bureau of Labor Statistics data), percent of employees in the respondent’s state working in a union job when the respondent was 25 (Hirsch, Macpherson, and Vroman 2001), whether the respondent lived in a rural area, whether the respondent’s health limited his ability to work in young adulthood, marital status at age 25, and number of residential children at 25.
Variables predicting health, included in the age 50 health equations following adjustments for selection into group-based income trajectories, adjusted for early-life predictors of later-life health and later-life mechanisms that may explain differences in health across income trajectories. Early-life predictors were measured at or prior to age 25, and included many but not all of the variables in the selection equation, including race-ethnicity, US nativity, educational attainment at age 25, cognitive ability, family of origin socioeconomic variables, and whether health limited the ability to work. Mechanisms added at or about age fifty reflected (a) family resources, (b) health behaviors and health status, and (c) job benefits and job strains. Family resources included household net worth (logged), marital status at age 50, and household size at age 50. Health behaviors and health status included whether the respondent had health insurance, Body Mass Index (BMI), usual hours of sleep on a weekday, whether the respondent smoked, binge drank (and frequency of binge drinking), or was physically inactive. SF-12 mental health score was also included in models predicting physical health, and SF-12 physical health score was included in models predicting mental health. Job benefits and strains included whether the respondent experienced any unemployment between ages 25–49 and if so, total length of unemployment in weeks (standardized with a mean of zero and a standard deviation of one), average hours worked while employed between ages 25–49, whether the respondent had access to a retirement plan other than Social Security Insurance (as a measure of fringe benefits)3, and job satisfaction at the respondent’s current or most recent job. Descriptive statistics are shown in Table 1.
Table 1.
Descriptive statistics for all variables in selection and health equations, N=3,605
| Respondent characteristics | Mean | Std. Dev. | Min | Max |
|---|---|---|---|---|
| Physical health, age 50 | 50.1 | 9.4 | 11.2 | 67.3 |
| Mental health, age 50 | 53.9 | 8.1 | 12.9 | 68.6 |
| Non-Latino, non-Black ab | 50.0% | |||
| -Black | 30.9% | |||
| -Latino | 19.1% | |||
| Non-US native ab | 6.7% | |||
| AFQT-R percentile, 1980ab | 38.3 | 30.4 | 1 | 99.0 |
| Years of education, age 25 (centered at 12)ab | −.64 | 3.8 | −12 | 8 |
| Skilled labor, age 25a | 16.1% | |||
| -Professional | 10.2% | |||
| -Sales | 16.0% | |||
| -Service | 11.1% | |||
| -Farm | 3.0% | |||
| -Machine labor | 8.6% | |||
| -Transportation | 6.2% | |||
| -Unskilled labor | 7.7% | |||
| -Military | 6.6% | |||
| -Student | 2.5% | |||
| -Out of work | 12.0% | |||
| Number of residential children, age 25a | .38 | .76 | 0 | 5 |
| Never-married, age 25a | 58.0% | |||
| -Married | 35.3% | |||
| -Divorced or widowed | 6.5% | |||
| Health limits work ab | 3.4% | |||
| Family of originab | ||||
| -Mother did not graduate HS | 42.3% | |||
| -Lived below poverty line | 17.5% | |||
| -Did not live with both parents | 31.2% | |||
| Local contexta | ||||
| County-level unemployment | 19.3% | 4.9% | 4.5% | 43% |
| % union workers in state, age 25 | 17.9% | 7.8% | 0% | 44.6% |
| Lives in rural area, age 25 | 18.4% | |||
| Mechanisms, age 50b | ||||
| Never-married, age 50 | 17.8% | |||
| -Married | 57.2% | |||
| -Divorced | 20.7% | |||
| -Widowed | 4.2% | |||
| Household size | 2.6 | 1.4 | 1 | 10 |
| Net worth (logged) | 14.6 | .32 | 0 | 15.6 |
| Body Mass Index (centered at 25) | 4.1 | 5.2 | −13.4 | 44.1 |
| R is a smoker | 27.2% | |||
| Did not binge drink in last month | 79.5% | |||
| -Binge drank <1x/week in last month | 8.9% | |||
| -Binge drank ≥1x/week in last month | 11.6% | |||
| Sleep hours on weekdays (centered at 7) | −.43 | 1.4 | −7 | 5 |
| Unable to/never engages in physical activity for >10 min | 26.9% | |||
| Lacks health insurance | 18.4% | |||
| Experienced any unemployment, ages 25–49 | 71.9% | |||
| Total weeks unemployed, ages 25–49 | 33.2 | 49.8 | 0 | 442 |
| Weeks unemployed X Any unemployment, Standardized |
0 | 1 | −.89 | 7.5 |
| Average hours employed per week, ages 25–49 (centered at 40) | 5.5 | 8.0 | −38 | 61 |
| Employer provides access to retirement plan | 63.2% | |||
| Likes current or most recent job very much | 46.4% | |||
| -Likes current/most recent job fairly well | 44.8% | |||
| -Dislikes somewhat | 6.3% | |||
| -Dislikes very much | 2.5% |
= variables in selection equation predicting group-based income trajectory
= variable in second-stage equation predicting health at age 50
Methods
Group-based developmental trajectories, a type of finite mixture modeling used to identify groups of individuals following similar age or time-graded trajectories of change, were used to identify trajectories of income stability, growth, or loss that emerged over time across our sample of men while they were employed in the paid workforce and earned income between ages 25–49. Researchers using this method specify the number of trajectories (we compare model fit across 4 to 14 trajectory models), the shape of each trajectory (our final model specifies a cubic relationship between logged income and age for all eight trajectories)4, and the distribution of the variable of interest (here, logged income takes a censored normal distribution) (Nagin 2005: 28):
Latent variable is individual i’s potential logged income at age t, are parameters that determine the shape of trajectory j when age and logged income have a cubic relationship, and is an error term. Each trajectory is estimated using a unique set of parameters.
We drew on Nagin (2005) and Jones and Nagin (2013) to estimate and identify the number and shape of men’s group-based income trajectories, using logged and time-varying prospective reports of men’s personal income. We compared measures of model fit (person and person-year Bayesian Information Criterion (BIC) statistics as well as Akaike Information Criterion (AIC) statistics) for four to fourteen group-based pathways of income, identifying the eight group model as the most parsimonious and theoretically-grounded depiction of men’s income trajectories (see Appendix 2 for measures of model fit). These pathways represent eight trajectories of income growth, stability, or loss over time among the NLSY79 men in our sample, in logged 2014 dollars and across ages 25–49.
Following this, we employed multinomial treatment models (Deb 2009; Deb and Trivedi 2006) to estimate the relationships between group-based income trajectories and mental or physical health at age fifty. Multinomial treatment models are two-stage regression models that first predict individuals’ membership in one of the group-based income trajectories using observed variables previously found to be associated with selection into the categorical “treatment.” From this first stage equation, lambda coefficients are calculated and added to a second stage equation predicting mental or physical health. This method produces results that predict the relationship between group-based income trajectories and age fifty health, adjusting for unequal selection into group-based income trajectories.
The first stage equation (1) predicts group-based income trajectory membership dj, with j representing one of seven group based income trajectories relative to the reference group, for individual i using observed exogenous covariates zi and term lij denoting unobserved characteristics affecting both income trajectory and age fifty health (Deb and Trivedi 2006).
| (1) |
The second-stage expected outcome equation (2) for individual i, integrating the first-stage selection model is as follows (Deb and Trivedi 2006):
| (2) |
SF-12 mental or physical health score yi is a function of the respondent’s group-based income trajectory membership γj, observed exogenous covariates xi, and parameters associated with latent factors affecting selection into income trajectory λj. Missing data were imputed using the ice command in Stata 14. Health scores and time-varying income were not imputed in the models shown (but see Appendix 4 for results using imputed income). Mental and physical health were modeled separately.
RESULTS
Group-based trajectories of income
Men’s group-based income trajectories are depicted in Figure 1. We identified groups of men with incomes that were stably high, stably low, increasing with time, and declining with time. Specifically, we identified three groups of men with relatively stable incomes over time (Low, Middle, and Highest income), two groups of men whose incomes declined, but at different ages (Low to Very Low – early and Low to Very Low – late), and three groups of men whose incomes started low and rose with time (Very Low to Low, Lowest to Middle, Lowest to Very Low). Importantly, the men who experienced upward mobility reported either Very Low or Lowest incomes at age 25 and most moved up to Low incomes and none moved higher than the Middle earners. Figure 1 graphs the eight groups and their logged income (adjusted to 2014 dollars) across ages 25–49. Table 2 includes descriptive statistics at the group level, including APP or average probability of individuals’ placement onto the correct income trajectory, percent of respondents per category, and median values for starting (age 25) and ending (age 49) incomes for each category in 2014 dollars. Average probability of correct placement (APP) for these eight groups was high, ranging from .833 for those with Low incomes to .935 for those with Low to Very Low –early income.
Figure 1.
Men’s group-based trajectories of personal income, ages 25–49
See tiff file
Table 2.
Characteristics of men’s eight group-based income trajectories from ages 25–49
| Percent | APP | Median income, age 25 | Median income, age 49 | |
|---|---|---|---|---|
| 1. Lowest to very low | 2.5% | .931 | $1,351 | $5,150 |
| 2. Very low to low | 5.5% | .869 | $8,106 | $30,055 |
| 3. Lowest to middle | 5.2% | .888 | $2,220 | $51,280 |
| 4. Low to very low early | 1.9% | .935 | $13,415 | $4,320 |
| 5. Low to very low late | 7.9% | .860 | $16,569 | $7,665 |
| 6. Low | 23% | .833 | $21,000 | $32,960 |
| 7. Middle | 37% | .860 | $32,960 | $56,625 |
| 8. Highest | 17.1% | .908 | $46,100 | $116,640 |
Note: Incomes are presented in 2014 dollars. APP = Average probability of individuals’ placement onto the correct group-based trajectory of income.
Our choice of the eight-group model was related to indices of model fit as well as the substantive value of the additional groups. For example, the seven group model differs by only 1% in fit when compared to the eight group model (see Appendix A2). However, the eight group model adds a new and theoretically meaningful category – men whose incomes rise, but only from a Very low to Low level from ages 25–49. Including this group allowed us to evaluate the midlife health of men who do experience upward mobility, yet fail to reach even the Middle income category. Adding a ninth group to the 8-group model improved model fit by less than 1%, did not add a substantively meaningful group, and resulted in four distinct groups comprising less than 5% of our sample each, thus limiting generalizability and our ability to detect statistically significant differences across groups and adding little to our overall findings.
Multinomial treatment results
Multinomial treatment models predicted mental and physical health at age 50 net of observed characteristics that select men into group-based income trajectories. Table 3 includes results from the first-stage equation in our multinomial treatment models, a multinomial logistic regression predicting entry onto each group-based income trajectory relative to the “Highest” income trajectory. Most variables in the multinomial treatment model played a significant role in placing men onto long-term income trajectories, including race-ethnicity, US nativity, cognitive ability, educational attainment, occupation at age 25, marital status, family of origin characteristics (living with two parents, maternal education, family poverty status), and local context (including county unemployment rate, presence of union workers in the respondent’s state, and living in a rural area).
Table 3.
Multinomial logistic regressions of entry onto group-based income trajectories, taken from first-stage of multinomial treatment models of physical health (Model 1 of Table 4)
| Lowest to very low | Very low to low | Lowest to middle | Low to very low early | Low to very low late | Low | Middle | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Coef. | p | Coef. | p | Coef. | p | Coef. | p | Coef. | p | Coef. | p | Coef. | p | |
| -Black [ref = non Latino, non-Black] | .99 | * | .04 | .53 | .57 | .43 | .13 | −.19 | ||||||
| -Latino [ref = non Latino, non-Black] | −.08 | −.20 | .09 | .00 | −.23 | −.05 | −.24 | |||||||
| Non-US native | −2.55 | * | −.53 | −1.01 | * | −1.45 | * | −.39 | −.80 | ** | −.46 | |||
| AFQT-R percentile, 1980 | −.05 | *** | −.04 | *** | −.02 | *** | −.05 | *** | −.04 | *** | −.04 | *** | −.02 | *** |
| Years of education, age 25 | −.11 | ** | −.10 | *** | −.06 | −.06 | −.08 | ** | −.07 | ** | −.03 | |||
| Occupation, age 25 [ref = skilled labor] | ||||||||||||||
| -Professional | −1.62 | −1.02 | −1.20 | * | −40.2 | *** | −2.92 | *** | −1.05 | *** | −.61 | ** | ||
| -Sales | −1.83 | * | −.54 | −.34 | −1.19 | −.86 | * | −.43 | −.34 | |||||
| -Service | .45 | .56 | .87 | .70 | .85 | * | .95 | ** | .51 | |||||
| -Farm | 1.42 | 1.30 | .63 | 1.88 | * | .46 | 1.24 | * | .28 | |||||
| -Machine labor | −.62 | .49 | .43 | .27 | .63 | .57 | .16 | |||||||
| -Transportation | −.73 | .22 | .21 | −.03 | .51 | .81 | * | .73 | * | |||||
| -Unskilled labor | −.19 | .94 | * | 1.01 | * | −.07 | .72 | .73 | * | .38 | ||||
| -Military | 1.57 | * | 2.88 | *** | 3.38 | *** | 1.05 | .72 | 1.01 | ** | .55 | |||
| -Student | .60 | .77 | 1.25 | * | .18 | −.59 | −.48 | −.85 | * | |||||
| -Out of workforce | 1.51 | ** | 2.29 | *** | 1.80 | *** | 1.42 | ** | 1.38 | *** | 1.21 | *** | .20 | |
| Health limits work [ref = no limits] | 1.65 | ** | .67 | .55 | .80 | .73 | .08 | −.24 | ||||||
| Number of residential children, age 25a | −.18 | .16 | .19 | −.23 | .08 | .06 | .06 | |||||||
| -Married, age 25 [ref = never married] | −2.06 | *** | −1.09 | *** | −1.89 | *** | −.98 | * | −1.39 | *** | −.85 | *** | −.26 | |
| -Divorced or widowed [ref=never married] | −.09 | .12 | .19 | .88 | .28 | .15 | .09 | |||||||
| Mother not HS grad | .57 | .63 | ** | .35 | .84 | * | .54 | ** | .86 | *** | .37 | * | ||
| Below poverty line | 2.24 | *** | 1.77 | *** | 1.08 | *** | 1.31 | *** | 1.00 | *** | .50 | * | −.07 | |
| Did not live with both parents | .95 | ** | .53 | * | .33 | .67 | * | .67 | ** | .61 | *** | .34 | * | |
| County-level unemployment | −2.06 | −1.51 | −5.72 | * | −3.60 | −1.75 | −3.81 | * | −2.16 | |||||
| % union workers in state, age 25 | 1.70 | −4.27 | ** | 1.12 | .82 | −3.38 | ** | −3.90 | *** | −1.33 | ||||
| Lives in rural area, age 25 | −.01 | .60 | * | .58 | .20 | .81 | ** | .67 | ** | .38 | * | |||
| Constant | −1.30 | .24 | −.18 | −.77 | 1.05 | 2.73 | *** | 2.79 | *** | |||||
Note: Reference category is steady highest income.
p<.05,
p<.01,
p<.001, two tailed hypothesis tests.
Physical health
Model 1 of Table 4 estimated the relationships between income trajectories and health after adjusting for unequal selection into group-based income trajectories. Model 1 additionally controlled for early life characteristics that are likely to influence health via cumulative advantages and disadvantages that may accrue or persist with time (e.g. Ferraro et al. 2015; Hayward and Gorman 2004). Net of the selection (described above and shown in Table 3) and early life characteristics, six of the seven income categories reported worse physical health at age fifty compared with those in the Highest income category, providing support for Hypotheses 1a and 1b. Our test for the health benefits of upward mobility (H2) – that men who were upwardly mobile and earned higher incomes later in adulthood may experience health similar to that of men who are steady Highest earners – was partially supported. One of the upwardly mobile groups – the Lowest to middle earners – reported similar health at age 50 to the highest earners net of selection, and controlling for early life characteristics. Early life characteristics were also associated with midlife health: after adjusting for other early life predictors, black men reported better physical health than non-Latino, non black men, as did non-US native men compared with US natives, and men with higher cognitive skill. Low maternal education and work-limiting health conditions in young adulthood were each associated with significantly worse health at age fifty, adjusting for other model variables.
Table 4:
Physical health SF-12 scores, multinomial treatment results
| Model 1 | Model 2 | Model 3 | Model 4 | |||||
|---|---|---|---|---|---|---|---|---|
| Group-based income trajectory (ref = Highest) | Coef. | p | Coef. | p | Coef. | p | Coef. | p |
| 1. Lowest to very low | −7.56 | *** | −7.17 | *** | −5.57 | −4.69 | * | |
| 2. Very low to low | −3.86 | *** | −3.36 | *** | −2.30 | −1.68 | ||
| 3. Lowest to middle | −1.46 | −1.16 | −.54 | −.13 | ||||
| 4. Low to very low early | −5.58 | ** | −5.00 | ** | −4.44 | −3.73 | ||
| 5. Low to very low late | −7.54 | *** | −6.88 | *** | −5.43 | *** | −4.57 | *** |
| 6. Low | −4.66 | *** | −4.19 | *** | −3.61 | −3.01 | ** | |
| 7. Middle | −1.35 | * | −1.12 | −1.16 | −1.01 | |||
| Early life characteristics/ Demographics | ||||||||
| -Black (ref = non-Black, non-Latino) | 1.29 | ** | 1.40 | ** | 1.49 | ** | 1.53 | *** |
| -Latino (ref = non-Black, non-Latino) | .57 | .64 | .70 | .63 | ||||
| Non-US native | 1.59 | * | 1.52 | * | 1.44 | 1.44 | * | |
| AFQT-R | .03 | *** | .03 | *** | .02 | .02 | ** | |
| Years of education (centered at 12) | .05 | .04 | .04 | .04 | ||||
| Health limits work | −2.59 | ** | −2.53 | * | −1.88 | −1.97 | * | |
| Mother not HS grad | −1.09 | ** | −1.11 | ** | −.86 | * | −.82 | * |
| Below poverty line | −.10 | −.13 | −.09 | .05 | ||||
| Did not live with both parents | .50 | .60 | .60 | .58 | ||||
| Family and household mechanisms, age 50 | ||||||||
| -Married (ref = never-married) | .54 | .57 | .47 | |||||
| -Divorced (ref = never-married) | −.90 | −.63 | −.68 | |||||
| -Widowed (ref = never-married) | −1.38 | −.94 | −1.03 | |||||
| Household size | −.02 | −.05 | −.05 | |||||
| Net worth (logged) | 1.10 | .42 | .31 | |||||
| Health behavior and health status mechanisms, age 50 | ||||||||
| Body Mass Index (centered at 25) | −.23 | *** | −.23 | *** | ||||
| R is a smoker | −1.36 | ** | −1.20 | ** | ||||
| Binge drank <1x/week in last month (ref = none in last month) | 1.65 | ** | 1.53 | *** | ||||
| Binge drank ≥1x/week in last month (ref = none in last month) | 1.16 | ** | 1.24 | ** | ||||
| Sleep hours on weekdays (centered at 7) |
1.10 | *** | 1.10 | *** | ||||
| Unable to/never engages in physical activity for >10 min | −4.50 | *** | −4.50 | *** | ||||
| Lacks health insurance | 1.35 | ** | .13 | *** | ||||
| Mental health SF-12 score | .15 | *** | −.23 | *** | ||||
| Job benefits and strains mechanisms, age 50 | ||||||||
| Experienced unemployment, ages 25–49 | −1.13 | *** | ||||||
| Experienced unemployment X Total weeks unemployed, ages 25–49 (standardized) | −.33 | |||||||
| Average hours employed per week, ages 25–49 (centered at 40) | −.01 | |||||||
| Employer provides access to retirement plan | .86 | * | ||||||
| Likes current/most recent job fairly well (ref = likes very much) | −.55 | |||||||
| Dislikes somewhat (ref = likes very much) | −.85 | |||||||
| Dislikes very much (ref = likes very much) | −3.04 | * | ||||||
| Constant | 51.53 | *** | 33.37 | *** | 39.34 | *** | 41.93 | *** |
| ln(∂) | 2.15 | *** | 2.14 | *** | 2.08 | *** | 2.08 | *** |
| Λ(Lowest to very low) | −.66 | −.62 | .08 | .07 | ||||
| Λ(Very low to low) | .65 | .63 | .30 | .21 | ||||
| Λ(Lowest to middle) | −1.58 | * | −1.64 | ** | −1.08 | −1.02 | ||
| Λ(Low to Very low early) | −1.52 | * | −1.52 | * | −.49 | −.42 | ||
| Λ(Low to Very low late) | −.02 | −.11 | −.03 | −.03 | ||||
| Λ(Low) | 1.06 | 1.06 | 1.41 | 1.31 | ||||
| Λ(Middle) | −.63 | −.62 | .09 | .20 | ||||
| ∂ | 8.57 | 8.53 | 7.91 | 8.00 | ||||
Note: Reference group is steady highest income.
p<.05,
p<.01,
p<.001, two tailed hypothesis tests.
First stage (selection) regression results not shown.
Model 2 added family and household resources at age fifty that may help to identify some of the resources associated with better health among those in the Highest earning group. This first set of mechanisms included marital status at age fifty, household size, and family net worth. None of these variables achieved statistical significance, which did not support the hypothesis that social and economic support from families would partially explain the better health of those with the Highest incomes. However, adjusting for these differences across groups reduced the coefficient for the Middle earners to nonsignificance, suggesting that worse physical health for this group is partially attributable to their different access to family and household resources. Model 3 added health behaviors and health status, including body mass index (BMI), respondent smoking and drinking behaviors, sleep, physical inactivity, health insurance, and mental health score at age 50. Each of these variables was independently and significantly associated with physical health at age fifty. This includes lacking health insurance, which was associated with better health at age fifty. Further, after adjusting for health behaviors and health status, the upwardly mobile Lowest to very low and Very low to low, the downwardly mobile Low to very low – early, and the stable Low earners no longer reported worse health at age 50 relative to the Highest earners. The worse health of these groups appears to be largely attributable to their health behaviors and status at midlife compared with those in the Highest income trajectory.
Our final series of mechanisms considered the cumulative and recent work experiences that men reported by age fifty. Average hours employed while employed between ages 25–49, cumulative weeks spent unemployed between ages 25–49 (conditional on reporting any unemployment during years where men earned an income), availability of fringe benefits such as a retirement plan, and job satisfaction at current or most recent job were added in Model 4. Experiencing any unemployment was associated with worse physical health, but the interaction between any unemployment and length of unemployment indicated that there was no added health penalty or benefit for those experiencing longer or shorter cumulative unemployment. The availability of fringe benefits – specifically, a retirement plan other than Social Security Insurance (SSI) -- was associated with significantly better health at age 50. Finally, those who very much disliked their current or most recent job reported significantly worse physical health at age 50. The associations between group-based income trajectories and health also changed between Model 3 and Model 4. Controlling for cumulative job experiences, health behaviors and health status, and family characteristics, groups with Lowest to very low, Low to very low – late, and Low incomes reported significantly worse health than those with the Highest income.
Mental health
Mental health across group-based income trajectories is estimated in Table 5. Controlling for early life characteristics and selection into group-based income trajectories, men with stable Middle incomes reported mental health that was equivalent to the Highest earners, which did not support Hypothesis 1a. Downwardly mobile groups reported worse mental health, supporting Hypothesis 1b. One upwardly mobile group – the Lowest to middle earners – reported health similar to that of the Highest earners, supporting Hypothesis 2. Thus, for mental health, we see some evidence that income stability—at least at the Middle status—appears to be associated with better mental health, as does upward mobility that results in middle class status. Early life characteristics associated with mental health at age fifty showed that black and Latino men reported better mental health than non-Latino, non-black men, and men with work-limiting health conditions in young adulthood also reported worse mental health, adjusting for other model variables.
Table 5:
Mental health SF-12 scores, multinomial treatment results
| Model 1 | Model 2 | Model 3 | Model 4 | |||||
|---|---|---|---|---|---|---|---|---|
| Group-based income trajectory (ref = Highest) | ||||||||
| 1. Lowest to very low | −8.12 | *** | −7.78 | *** | −6.40 | *** | −5.20 | ** |
| 2. Very low to low | −2.43 | ** | −2.02 | * | −1.31 | −.61 | ||
| 3. Lowest to middle | −.95 | −.69 | .01 | .29 | ||||
| 4. Low to very low early | −4.93 | ** | −4.37 | ** | −2.94 | * | −2.19 | |
| 5. Low to very low late | −4.33 | *** | −3.75 | *** | −2.33 | ** | −1.52 | |
| 6. Low | −1.58 | ** | −1.21 | * | −.20 | .18 | ||
| 7. Middle | .53 | .73 | 1.14 | * | 1.23 | * | ||
| Early life characteristics/ Demographics | ||||||||
| -Black (ref = non-Black, non-Latino) | 1.11 | ** | 1.25 | ** | 1.16 | ** | 1.24 | *** |
| -Latino (ref = non-Black, non-Latino) | .89 | * | .93 | * | .83 | * | .77 | * |
| Non-US native | −.81 | −.89 | −1.05 | −1.09 | * | |||
| AFQT-R | .01 | .00 | .00 | .00 | ||||
| Years of education (centered at 12) | −.06 | −.07 | * | −.09 | ** | −.10 | ** | |
| Health limits work | −2.40 | ** | −2.33 | ** | −1.90 | * | −2.36 | ** |
| Mother not HS grad | −.12 | −.13 | .13 | .22 | ||||
| Below poverty line | −.73 | −.76 | −.66 | −.48 | ||||
| Did not live with both parents | −.01 | .08 | .04 | .01 | ||||
| Family and household mechanisms, age 50 | ||||||||
| -Married (ref = never-married) | .48 | .35 | .22 | |||||
| -Divorced (ref = never-married) | −.92 | −.69 | −.84 | |||||
| -Widowed (ref = never-married) | −1.76 | * | −1.40 | −1.41 | ||||
| Household size | .10 | .08 | .05 | |||||
| Wealth (net worth), logged | .34 | −.09 | −.18 | |||||
| Health behavior and health status mechanisms, age 50 | ||||||||
| Body Mass Index (centered at 25) | −.05 | −.05 | ||||||
| R is a smoker | −.86 | * | −.78 | * | ||||
| Binge drank <1x/week in last month (ref = none in last month) | −.48 | −.56 | ||||||
| Binge drank ≥1x/week in last month (ref = none in last month) | −.36 | −.19 | ||||||
| Sleep hours on weekdays (centered at 7) | 1.00 | *** | .99 | *** | ||||
| Unable to/never engages in physical activity for >10 min | −1.22 | *** | −1.24 | *** | ||||
| Lacks health insurance | −.25 | −.17 | ||||||
| Physical health SF-12 score | .15 | .12 | *** | .11 | *** | |||
| Job benefits and strains mechanisms, age 50 | ||||||||
| Experienced unemployment, ages 25–49 | −.23 | |||||||
| Experiences unemployment X Total weeks unemployed, ages 25–49 (standardized) | −.37 | |||||||
| Average hours employed per week, ages 25–49 (centered at 40) | .04 | * | ||||||
| Employer provides access to retirement plan | .04 | |||||||
| Likes current/most recent job fairly well (ref = likes very much) | −1.14 | *** | ||||||
| Dislikes somewhat (ref = likes very much) | −4.45 | *** | ||||||
| Dislikes very much (ref = likes very much) | −4.62 | *** | ||||||
| Constant | 54.43 | *** | 48.88 | *** | 49.84 | *** | 52.39 | *** |
| ln(∂) | 2.05 | *** | 2.04 | *** | 2.00 | *** | 1.99 | *** |
| Λ(Lowest to very low) | .01 | .02 | .04 | .11 | ||||
| Λ(Very low to low) | .38 | .38 | .33 | .34 | ||||
| Λ(Lowest to middle) | −.68 | −.75 | −.82 | −.60 | ||||
| Λ(Low to Very low early) | −.81 | −.84 | −.79 | −.53 | ||||
| Λ(Low to Very low late) | −.05 | −.09 | −.16 | −.25 | ||||
| Λ(Low) | .19 | .23 | −.03 | .02 | ||||
| Λ(Middle) | −1.06 | −1.09 | −1.20 | * | −.96 | |||
| ∂ | 7.77 | 7.72 | 7.38 | 7.33 | ||||
Note: First stage (selection) regression results not shown. Reference group is steady highest income.
p<.05,
p<.01,
p<.001, two tailed hypothesis tests.
Family and household resources were added in Model 2 and did not alter the relationships between group-based trajectories of income and mental health at age 50. Net worth and household size were not associated with mental health. The widowed reported worse mental health at age 50 than the married. In Model 3, health behaviors and health status were added. Smokers and the physically inactive reported significantly worse mental health, while those who reported more sleep and those with better physical health reported significantly higher mental health scores. Adjusting for health behaviors and health status, a suppression effect is revealed –Middle income earners with similar health behaviors and health status to the Highest earners reported significantly higher SF-12 scores at age fifty than the Highest earners. Further, adjusting for these variables reduced the upwardly mobile earners in both Very low to Low and Low coefficients to nonsignificance.
In Model 4, job related mechanisms were also associated with mental health. While unemployment and unemployment length were not associated with mental health, men averaging more hours of work per week experienced better mental health, and current or most recent job satisfaction was strongly associated with mental health. Adjusting for these variables reduced the downwardly mobile Very Low to low early and late groups to nonsignificance, suggesting that the worse mental health of this group is largely attributable to their cumulative differences in family and household resources, health behaviors and health status, and job benefits and strains.
Sensitivity tests
To test the robustness of our findings, we conducted a number of sensitivity tests, presented in Appendices 3 and 4 (only Model 4 results are shown, full results available upon request). First, we tested whether results were robust when imputing income and when excluding those with work-limiting health conditions as young adults. The results are quite consistent: the only exceptions are the Very low to low earners, who reported worse physical health and mental health than the Highest earners after adjusting for all Model 4 variables when imputing income and when excluding those with work-limiting health conditions. Additionally, the Low to very low – early earners reported worse physical health in the models that excluded those with work-limiting health conditions, and the Low to very low – late earners reported worse mental health in a model that imputed men’s income across ages 25–49.
We also tested whether results were robust to seven or nine group-based trajectories of income (see Appendix 3). Using seven group-based income trajectories, we identified many of the same groups found in the eight-group model, including stable Highest, Middle, and Low incomes, as well as the two previously identified Low to Very Low groups (“early” and “late”). Two of the previously identified upwardly mobile groups are also found in the seven group model (Lowest to middle and Lowest to Very Low). The only group not found is the Very Low to Low group. The seven group model showed that nearly all of the groups identified in our eight-group model were stable across an alternative specification of men’s income. The results using seven group-based income trajectories are virtually identical to those found using the eight groups when predicting physical health. For mental health, the seven-group model indicates that Low to very low – early earners report worse mental health, and those with Middle incomes report mental health similar to that of the Highest earners adjusting for all variables in Model 4 of Table 5.
The nine-group model differed from the eight group model primarily because four of the groups became exceptionally small – the Lowest to very low, Lowest to low, Low to very low – early, and Very low to high groups each comprised less than five percent of our total sample, limiting our ability to detect statistically significant differences. Only the Lowest to very low earners reported worse physical health than the Highest earners after controlling for all Model 4 variables, all other groups were not statistically significant from the Highest earners. For mental health, we see similar findings to the eight group model with the exception of the Low to very low – late earners, who reported worse mental health than the Highest earners adjusting for all Model 4 variables.
DISCUSSION
Understanding the relationship between SES and health has long been of interest to scholars of public health (Elo 2009; Hayward and Gorman 2004; Kawachi, Adler, and Dow 2010) and refining measures of socioeconomic status continues to be central to scholars of stratification (Weeden and Grusky 2005; Chan and Goldthorpe 2007; Gangl 2005), but rarely are the two literatures combined. We make several important contributions to these existing literatures. First, our study is the first to examine how men’s group-based income trajectories over time are associated with mental and physical health. We find men’s physical health is directly related to an ability to maximize earnings over time or improve earnings to Middle class status over time, while men had similar mental health outcomes across the Highest, the Middle, and the Lowest to middle groups, suggesting both stable as well as rising incomes can be beneficial to mental health. Second, our findings support the use of income, especially income trajectories measured over time, in future studies examining the relationships between SES and health. Unlike studies that rely on men’s educational attainment, our focus on income trajectories allow us to capture men’s time-varying differences in social status and relative prestige. If economic insecurity continues to expand, as has been suggested by Kalleberg (2009), income trajectories may become an even more important measure of the relationship between SES and health.
Third, we find remarkable rigidity in the income groups that had formed by the time the men were 25—less than 25% of our sample experience significant upward or downward mobility between the ages of 25 to 49 and most who moved remained in or moved into poverty. While prior work on men’s upward mobility has indicated US workers enjoy gains in income across their 20s and 30s, but income declines as men enter into their forties (Gangl 2005), our findings show far greater heterogeneity than can be detailed through a cross-sectional examination of income by age. We demonstrate men do not follow a single age-graded pathway of mobility over time. Instead, men experienced eight distinct income trajectories across their twenties, thirties, and forties, and these trajectories had implications for both mental and physical health at midlife. Some men experienced significant upward mobility, others remained stable Low, Middle, or Highest earners, and still others experienced a loss in earnings over time.
The modal category was those with stable middle incomes (37%) and it was also fairly common to have a stable low income (23%) or a stable highest (17.1%) income. Notably, 13.2% of men experienced upward income mobility between the ages of 25 and 49, yet no upwardly mobile groups moved into the highest income range—they moved into either very low (5.5%), low (2.5%), or middle (5.2%) income by the age of 50. These are men reporting incomes ranging from $1,351 to $8,106 at age 25 and incomes of $5,150 (very low), $30,055 (low), and $51,280 (middle) at age 49. While about five percent made notable gains and achieved middle-class status, others remained near or below the poverty line. There were no groups that experienced meteoric rises that would have brought them level to the highest earners with a median income of over $116,000 annually. Another 10 percent experienced downward mobility during this time period and, around the age of 50, these men reported median incomes between four and seven thousand dollars annually, suggesting a plunge into deep poverty. Thus long-term trajectories of income do vary, placing men at different levels of status both relative to their past earning years and relative to their peers on other trajectories, but most men experienced remarkable consistency during the 25 year period. These findings demonstrate the importance of a life course perspective, which investigates whether and how incomes changes over time, constrained by early experiences and operating in concert with family, macro-level, and occupational contexts.
Our findings provide support for the importance of maintaining a high income over time for physical health, as well as evidence that a downward trend in income is bad for one’s overall health—both mental and physical. Both of these findings extend prior research studying SES-health relationships over time (Cundiff et al. 2017; House et al. 2005; Tiffin et al. 2005) and lend support to our argument that measuring income over time is an important test of the relationship between SES and health. Studies using cross-sectional measures of income or other measures of socioeconomic standing such as education or occupation have not previously shown how rare upward mobility is, or how important income stability is (particularly at Middle or Highest levels) for well-being. In a climate of greater economic insecurity for men – particularly working class men – each of these findings indicate that working-class and working-poor men are at significant risk of poorer mental and physical health at midlife.
But some of our hypotheses were only partially supported, including our expectation that upward mobility would provide strong health benefits. This is surprising, but should be noted in the context of the upward trends we described, as none of the groups included men whose income rose to Highest levels by age 50 and those who moved from Lowest to middle incomes did experience physical and mental health parity with their highest earning peers. Surprisingly, our mental health findings indicated that men who experienced either a stable Middle or Highest income trajectories reported similar mental health at age 50, despite the fact that the highest earners were earning roughly double the income, on average, of the middle earners. This suggests men’s mental health may not be as closely tied to earning a high income as it is to securing a steady good income over time. Gender theories on the importance of breadwinning in men’s lives combined with prior research on the income and health gradient may help us understand these trends, as the inability to fulfill breadwinning expectations may strain men’s self-esteem and well-being (Townsend 2002). Men’s ability to earn and sustain a family wage may directly influence his mental health via less exposure to unemployment, a more reliable paycheck, and higher job satisfaction - and also indirectly through feelings of self-esteem, autonomy, or self-efficacy, as men meet (or fail to meet) cultural ideals of masculinity related to breadwinning (Link, Lennon, and Dohrenwend 1993; Townsend 2002). Further research is needed to confirm this connection.
We also found that men’s early life characteristics helped to place them on income trajectories and continued to play an independent role in influencing health at midlife. Men with disadvantages early in life such as lower cognitive ability, lower educational attainment, non-professional occupational experience, and disadvantaged families of origin were less likely to achieve high or upwardly mobile income trajectories and were more likely to report trajectories associated with income losses over time. However, even after adjusting for the unequal likelihood of men to achieve higher incomes, income trajectories continued to be associated with midlife health, and early life disadvantages continued to play a role in predicting midlife health net of their role in income trajectories, though this held true more for physical health rather than mental health. We add to a broader body of literature showing the intertwined long-term impacts of early life disadvantages on both employment prospects and well-being (Ferraro et al. 2015; Hayward and Gorman 2004).
Some of the differences in income trajectories could be explained by our age 50 mechanisms, which were grouped under family and household resources, health status and health behaviors, and job benefits and strains. It was unexpected that, net of other variables in the model, neither mental nor physical health at age 50 was associated with marriage, which prior research identifies as providing numerous health benefits including better access to health insurance, spousal monitoring of health behaviors, and emotional and financial support (Umberson and Montez 2010; Waite 1995; Williams et al. 2010). It is possible that our selection model – which adjusted for marital status and helped place men on all but the middle earnings trajectory in young adulthood – prevented us from observing an independent role for the effect of marriage on health by midlife, as we have already adjusted for many of the variables associated with selection into marriage.
The age 50 health behaviors, mental health status, and unemployment experiences differed across income trajectories and appeared to provide some of the significant physical health benefits of high earnings. All of the health behaviors we included were independently associated with physical health and played critical roles in explaining the poorer physical health of downwardly mobile and lower-income earners. Surprisingly, health insurance was associated with poorer physical health, which may be explained by men’s lower likelihood to seek medical care or visit a doctor despite having fewer barriers to purchasing insurance than women, which may lead men to purchase insurance only after becoming ill (Bird and Rieker 1999; Perry et al. 2008). Future research should explore this finding further. Fringe benefits such as access to a retirement plan help men to plan for the future and also benefited physical health. The health consequences of unemployment, which we also observe, are well-known and include a greater risk of mortality and suicide, worse health behaviors (including diet, alcohol use, and sleep), and a greater risk for onset of chronic conditions (Garcy and Vågerö 2012; Kalousova and Burgard 2014; Leeflang, Klein-Hesselink, and Spruit 1992; Modrek and Cullen 2013). Including cumulative or longitudinal measures of income, work hours, and unemployment is a novel addition to prior literature and shows us that the life course of income growth or loss should not be considered without the inclusion of unemployment, as both exert independent effects on health. Once again, our findings point, albeit somewhat indirectly, to the importance of occupational characteristics that are tied to the growing number of “bad jobs” (Kalleberg 2009).
Our mechanisms operated somewhat differently when predicting mental health. Men reported worse mental health when they were less physically healthy, smokers, physically inactive, working fewer hours, and less satisfied with their current or previous jobs, but adjusting for these behaviors did not help to explain why lower earners reported worse mental health. Our findings also suggest that increased job satisfaction may be both a reward for work and one with consequences for one’s mental health, and supports other studies identifying the significant role of job satisfaction for health (Faragher et al. 2005). Although the relationship between income and satisfaction is less well-established, some qualitative evidence suggests that “good jobs” bring higher levels of job satisfaction (Judge et al. 2010; Nelson and Smith 1999).
We also found that black and Latino men reported higher mental health scores than non-black, non-Latino men adjusting for other model covariates. In part, these findings may not be surprising: black and Latino men are less likely to achieve the same socioeconomic status as white men, so a statistical model that compares men across income trajectories and controls for the many financial and structural burdens that non-white men face may not reveal a health penalty related to race. At the same time, there is some precedent for these findings. Jackson et al. (2010) found black adults are better able than white adults to use unhealthy coping strategies as a buffer against psychological distress (but at a cost to physical health), and Williams et al. (2007) found African American respondents reported lower lifetime and 12-month incidence of depression despite greater depression severity and chronicity for African Americans. Among lower-income Latino men, rates of severe psychological distress are lower than for non-Latino whites (Centers for Disease Control 2016, Table 46), but at higher levels of income Latino men report greater incidence of distress than whites. These additional studies indicate that our findings should be interpreted with caution. The processes underlying mental health may operate differently by race in ways we are unable to address in our analyses.
Limitations
Our study has a number of limitations. First, we exclude person-year observations when men are out of the labor force for the entire calendar year, as we are not always able to distinguish the reasons for being out of work. Although this strategy is preferred because it prevents men’s group-based income trajectories from being pulled downward by unequally distributed observations of zero income, it also prevents us from estimating the absolute income differences across all men in the sample. Second, we are unable to account for events that occur between ages 25 and 49, including changes in family size or occupation, spousal employment, and macro-level economic factors. All of these undoubtedly affect men’s well-being, but we cannot control for them in our models. Third, we are also limited by attrition between ages 25 and 49. Supplemental analyses (available on request) show that men who attrit during this time do not differ significantly on race, family of origin poverty status, occupational experiences at age 25, or education. Nonetheless, the men who attrit may have differed in their health profiles in ways that we cannot adjust for in our models. Fourth, the NLSY79 does not measure health at every wave and so we are only able to investigate relationships between income trajectories and health at age fifty. Ideally, we would be able to measure both health and income prospectively to examine the bidirectional relationship between socioeconomic status and health. However, the SF-12 is collected only at ages forty and fifty. We cannot estimate the role of reverse causation – that changes in health may lead to changes in income. Fifth and finally, gender may be associated with differential item functioning in the SF-12 (Fleishman and Lawrence 2003). Compared with women, men may underreport their health concerns, particularly those related to mental health. Although there is no reason to believe that our across-group comparisons would be affected by reporting bias, it may be the case that we are overestimating men’s overall mental and physical health.
Some have contended that our method itself – group-based developmental trajectories – may not produce accurate and consistent results regarding the assignment of respondents onto pathways or the number and shape of the pathways themselves (Warren et al. 2015). The present analysis is able to address many of the most significant concerns regarding the validity of results obtained using group-based developmental trajectories. First, our sensitivity analyses show that our main conclusions are robust to alternative specifications of men’s income trajectories, with similar findings for men across seven and eight group-based income trajectories and when imputing income or excluding those with work-limiting health conditions. Second, our analysis avoids many of the pitfalls associated with group-based developmental trajectories: the waves of the NLSY79 data are at closely spaced intervals, which is associated with more precise estimates of who is assigned to a trajectory and whether the analysis accurately captures the “true” number of trajectories (Warren et al. 2015). Additionally, we found a high probability of correct placement onto trajectories, as noted in Table 2. Nagin (2005) recommends that the APP measure at least 0.7 for each group; our APP scores average 0.866. A limitation that remains valid is that we may underestimate the true number of trajectories of income. Model convergence becomes more difficult as groups approach 1% or less of the total sample; consequently, we cannot conduct sensitivity tests for ten or more group-based trajectories of income, and even our supplemental analysis using nine group-based trajectories yields only limited results because of small cell sizes.
A remaining limitation of the paper is that the NLSY79 cohort is currently still in midlife. We cannot comment on how these advantages may continue to cumulate or may stabilize or even flatten with age. Examining the PSID data, Willson et al (2007) found evidence that cumulative disadvantage accrues over time and that much (but perhaps not all) of the apparent flattening effects of aging can be attributed to selection effects. Our findings point to the importance of the midlife health behaviors and job benefits and strains as important mechanisms through which these advantages are accrued. While some of the job benefits/strains may not last past retirement (such as job satisfaction), given the very low incomes of many in the study, some may find themselves unable to retire and many of the health behaviors likely will have lasting implications into old age.
Conclusion
While the relationship between SES and health has been well-established, the best measures to evaluate these relationships remain highly contested. Our research shows that group-based income trajectories do, as we anticipated, provide an important assessment of the differing roles of trajectories of earnings for mental versus physical health at midlife. We find evidence that the income pathways matter differently for mental and physical health. For physical health, it appears to matter most to be either one of the Highest earners or to experience substantial upward mobility from Very low to middle income. For mental health, either upward mobility to a Middle income, remaining stably Middle, or in the Highest trajectory appears beneficial. This is significant, because it suggests policies need not focus on making everyone high earners nor on addressing the mental health needs of the vast group of Middle earners. Instead, much of our policy attention could turn to those in Low, Very low, and Lowest earning groups with particular attention to those who have experienced downward mobility or who have moved up only to remain in poverty. Indeed our selection models point to the importance of unions and to the lasting effects of wages, suggesting that strong unionization efforts and campaigns to raise the minimum wage could have health implications for the population into middle age.
Moreover, while gender and work is often studied through a comparison of men and women or with a focus on women alone, it is important to remember that men, too, are gendered and have gendered obligations in society. Our research suggests that some men were better able to meet their obligation of breadwinning and that being able to do so (or failing to do so) impacts both mental and physical health. Further research is necessary to connect the relationship between gendered breadwinning norms, income trajectories, and mental health. Future research is also necessary to examine the relationship between women’s income trajectories and their physical and mental health at mid-life. While there is much focus on the relationship between women’s paid work and their health (see Frech and Damaske (2012) for discussion), recent research suggests that employment and income may work separately (Schnittker 2007), so asking these questions may reveal additional important linkages in our understanding of SES and health. Looking forward, our research raises urgent concerns as men’s ability to earn and maintain a breadwinning wage has diminished over the last thirty years (Boushey 2016) and it is unclear if this trend will reverse itself or if economic precarity will continue to rise. Should men’s economic conditions continue their downward trends, our research suggests this may have lasting negative implications for men’s physical and mental health, as those who cannot earn high or rising incomes will have poorer physical health and those whose income trajectories are low or decline over time will experience poorer mental health.
Supplementary Material
Acknowledgments
The authors thank Jessica Halliday Hardie, Steven Haas, Erin Kelly, Richard Petts, Jarron Saint Onge, Ranjan Shrestha, Mark Tausig, and members of the Working paper group at the School of Labor and Employment Studies at Penn State for their valuable comments. Research reported in this publication was supported by grants from the National Institutes of Health (R03HD088806) and (P2CHD04102). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This research was conducted with restricted access to Bureau of Labor Statistics (BLS) data. The views expressed here do not necessarily reflect the views of the BLS.
Footnotes
Excluding person-year observations where men did not earn any income prevents trajectories from being unevenly pulled downward by the unequal distribution of zero income occurrences across trajectories. For example, men with low incomes across their working years were also more likely to have person-year observations with zero income compared with their higher earning peers. Because the years that some men spent out of the workforce were likely to negatively affect their earnings upon reemployment, these years out of the workforce are indirectly reflected and accounted for in our models.
Although high mental health is associated with higher physical health (and vice versa), the SF-12 mental health and physical health composite scores are only weakly correlated in our sample at 0.25.
Access to employer provided medical benefits, dental benefits, retirement benefits, sick pay, a flexible schedule, and life insurance were collinear. We controlled for retirement as a proxy for the availability of fringe benefits.
Because the number and shape of men’s group-based income trajectories was unknown, a cubic form was used to allow for flexibility in identifying these trajectories. Additional analyses suggest that although this is not the most parsimonious model (three of the trajectories take a quadratic form); however, the results here are consistent with the more parsimonious model.
Contributor Information
Adrianne Frech, University of Missouri.
Sarah Damaske, The Pennsylvania State University.
References
- Adler Nancy, Singh-Manoux Archana, Schwartz Joseph, Stewart Judith, Matthews Karen, and Marmot Michael G.. 2008. “Social Status and Health: A Comparison of British Civil Servants in Whitehall-II with European- and African-Americans in CARDIA.” Social Science & Medicine 66(5):1034–45. [DOI] [PubMed] [Google Scholar]
- Adler Nancy E. and Stewart Judith. 2010. “Health Disparities across the Lifespan: Meaning, Methods, and Mechanisms.” Annals of the New York Academy of Sciences 1186(1):5–23. [DOI] [PubMed] [Google Scholar]
- Alwin Duane F. and McCammon Ryan J.. 2003. “Generations, Cohorts, and Social Change.” Pp. 23–49 in Handbook of the Life Course, Handbooks of Sociology and Social Research. Springer, Boston, MA. [Google Scholar]
- Amick Benjamin C., McDonough Peggy, Chang Hong, Rogers William H., Pieper Carl F., and Duncan Greg. 2002. “Relationship between All-Cause Mortality and Cumulative Working Life Course Psychosocial and Physical Exposures in the United States Labor Market from 1968 to 1992.” Psychosomatic Medicine 64(3):370–81. [DOI] [PubMed] [Google Scholar]
- Benach J., Vives A., Amable M., Vanroelen C., Tarafa G., and Muntaner C.. 2014. “Precarious Employment: Understanding an Emerging Social Determinant of Health.” Annual Review of Public Health 35(1):229–53. [DOI] [PubMed] [Google Scholar]
- Bertrand Marianne and Mullainathan Sendhil. 2004. “Are Emily and Greg More Employable Than Lakisha and Jamal? A Field Experiment on Labor Market Discrimination.” The American Economic Review 94(4):991–1013. [Google Scholar]
- Bird Chloe E. and Rieker Patricia P.. 1999. “Gender Matters: An Integrated Model for Understanding Men’s and Women’s Health.” Social Science & Medicine 48:745–55. [DOI] [PubMed] [Google Scholar]
- Huie Bond, Stephanie A., Krueger Patrick M., Rogers Richard G., and Hummer Robert A.. 2003. “Wealth, Race, and Mortality.” Social Science Quarterly 84(3):667–84. [Google Scholar]
- Boushey Heather. 2016. Finding Time: The Economics of Work-Life Conflict. Cambridge, Massachusetts: Harvard University Press. [Google Scholar]
- Brazier JE et al. 1992. “Validating the SF-36 Health Survey Questionnaire: New Outcome Measure for Primary Care.” BMJ 305(6846):160–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Burgard Sarah A. and Lin Katherine Y.. 2013. “Bad Jobs, Bad Health? How Work and Working Conditions Contribute to Health Disparities.” American Behavioral Scientist 57(8):1105–27. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Burgard Sarah A. and Seelye Sarah. 2017. “Histories of Perceived Job Insecurity and Psychological Distress among Older U.S. Adults.” Society and Mental Health 7(1):21–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Carr Deborah and Springer Kristen W.. 2010. “Advances in Families and Health Research in the 21st Century.” Journal of Marriage and Family 72(3):743–761. [Google Scholar]
- Centers for Disease Control. 2016. Health, United States, 2015: With Special Feature on Racial and Ethnic Health Disparities. Hyattsville, MD. [PubMed] [Google Scholar]
- Chan Tak Wing and Goldthorpe John H.. 2007. “Class and Status: The Conceptual Distinction and Its Empirical Relevance.” American Sociological Review 72(4):512–32. [Google Scholar]
- Cherlin Andrew J. 2014. Labor’s Love Lost: The Rise and Fall of the Working-Class Family in America. New York: Russell Sage Foundation. [Google Scholar]
- Clawson Dan and Gerstel Naomi. 2014. Unequal Time: Gender, Class, and Family in Employment Schedules. New York: Russell Sage Foundation. [Google Scholar]
- Coltrane Scott. 2004. “Elite Careers and Family Commitment: It’s (Still) about Gender.” The ANNALS of the American Academy of Political and Social Science 596(1):214–20. [Google Scholar]
- Cundiff Jenny M., Boylan Jennifer Morozink, Pardini Dustin A., and Matthews Karen A.. 2017. “Moving Up Matters: Socioeconomic Mobility Prospectively Predicts Better Physical Health.” Health Psychology 36(6):609–17. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Deaton Angus. 2002. “Policy Implications Of The Gradient Of Health And Wealth.” Health Affairs 21(2):13–30. [DOI] [PubMed] [Google Scholar]
- Deb Partha. 2009. MTREATREG: Stata Module to Fits Models with Multinomial Treatments and Continuous, Count and Binary Outcomes Using Maximum Simulated Likelihood. [Google Scholar]
- Deb Partha and Trivedi Pravin K.. 2006. “Specification and Simulated Likelihood Estimation of a Non-Normal Treatment-Outcome Model with Selection: Application to Health Care Utilization.” The Econometrics Journal 9(2):307–31. [Google Scholar]
- Elder Glen H. Jr, Johnson Monica Kirkpatrick, and Crosnoe Robert. 2003. “The Emergence and Development of Life Course Theory.” Pp. 3–19 in Handbook of the Life Course, Handbooks of Sociology and Social Research, edited by Mortimer JT and Shanahan MJ. Springer US. [Google Scholar]
- Elo Irma T. 2009. “Social Class Differentials in Health and Mortality: Patterns and Explanations in Comparative Perspective.” Annual Review of Sociology 35(1):553–72. [Google Scholar]
- Elo Irma T. and Preston Samuel H.. 1996. “Educational Differentials in Mortality: United States, 1979–1985.” Social Science & Medicine 42(1):47–57. [DOI] [PubMed] [Google Scholar]
- Faragher EB, Cass M., and Cooper CL. 2005. “The Relationship between Job Satisfaction and Health: A Meta-Analysis.” Occupational and Environmental Medicine 62(2):105–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Faraut Brice, Boudjeltia Karim Zouaoui, Vanhamme Luc, and Kerkhofs Myriam. 2012. “Immune, Inflammatory and Cardiovascular Consequences of Sleep Restriction and Recovery.” Sleep Medicine Reviews 16(2):137–49. [DOI] [PubMed] [Google Scholar]
- Farmer Melissa M. and Ferraro Kenneth F.. 1997. “Distress and Perceived Health: Mechanisms of Health Decline.” Journal of Health and Social Behavior 39(September):298–311. [PubMed] [Google Scholar]
- Ferraro Kenneth F., Schafer Markus H., and Wilkinson Lindsay R.. 2015. “Childhood Disadvantage and Health Problems in Middle and Later Life Early Imprints on Physical Health?” American Sociological Review 81(1): 107–133. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Findeis Jill L. and Jensen Leif. 1998. “Employment Opportunities in Rural Areas: Implications for Poverty in a Changing Policy Environment.” American Journal of Agricultural Economics 80(5):1000–1007. [Google Scholar]
- Fleishman John A. and Lawrence William F.. 2003. “Demographic Variation in SF-12 Scores: True Differences or Differential Item Functioning?” Medical Care 41(7 Suppl):III75–86. [DOI] [PubMed] [Google Scholar]
- Frech Adrianne and Damaske Sarah. 2012. “The Relationships between Mothers’ Work Pathways and Physical and Mental Health.” Journal of Health and Social Behavior 53(4):396–412. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Gandek Barbara et al. 1998. “Cross-Validation of Item Selection and Scoring for the SF-12 Health Survey in Nine Countries: Results from the IQOLA Project.” Journal of Clinical Epidemiology 51(11):1171–78. [DOI] [PubMed] [Google Scholar]
- Gangl Markus. 2005. “Income Inequality, Permanent Incomes, and Income Dynamics: Comparing Europe to the United State.” Work and Occupations 32(2):140–62. [Google Scholar]
- Garcy Anthony M. and Vågerö Denny. 2012. “The Length of Unemployment Predicts Mortality, Differently in Men and Women, and by Cause of Death: A Six Year Mortality Follow-up of the Swedish 1992–1996 Recession.” Social Science & Medicine 74(12):1911–20. [DOI] [PubMed] [Google Scholar]
- Haas Steven. 2007. “The Long-Term Effects of Poor Childhood Health: An Assessment and Application of Retrospective Reports.” Demography 44(1):113–35. [DOI] [PubMed] [Google Scholar]
- Haas Steven 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. [DOI] [PubMed] [Google Scholar]
- Haas Steven A. and Fosse Nathan Edward. 2008. “Health and the Educational Attainment of Adolescents: Evidence from the NLSY97.” Journal of Health and Social Behavior 49(2):178–92. [DOI] [PubMed] [Google Scholar]
- Haas Steven A., Glymour M. Maria, and Berkman Lisa F.. 2011. “Childhood Health and Labor Market Inequality over the Life Course.” Journal of Health and Social Behavior 52(3):298–313. [DOI] [PubMed] [Google Scholar]
- Hayward Mark D. and Gorman Bridget K.. 2004. “The Long Arm of Childhood: The Influence of Early-Life Social Conditions on Men’s Mortality.” Demography 41(1):87–107. [DOI] [PubMed] [Google Scholar]
- Herd Pamela, Goesling Brian, and House James S.. 2007. “Socioeconomic Position and Health: The Differential Effects of Education versus Income on the Onset versus Progression of Health Problems.” Journal of Health and Social Behavior 48(3):223–38. [DOI] [PubMed] [Google Scholar]
- Hirsch Barry T., Macpherson David A., and Vroman Wayne G.. 2001. “Estimates of Union Density by State.” Monthly Labor Review 124(7): 51–55. Data available at http://unionstats.gsu.edu/State_Union_Membership_Density_1964-2017.xlsx [Google Scholar]
- House James S. 1994. “The Social Stratification of Aging and Health.” Journal of Health and Social Behavior 35(September):213–34. [PubMed] [Google Scholar]
- House James S., Lantz Paula M., and Herd Pamela. 2005. “Continuity and Change in the Social Stratification of Aging and Health Over the Life Course: Evidence From a Nationally Representative Longitudinal Study From 1986 to 2001/2002 (Americans’ Changing Lives Study).” The Journals of Gerontology Series B: Psychological Sciences and Social Sciences 60(Special Issue 2):S15–26. [DOI] [PubMed] [Google Scholar]
- Jackson James S., Knight Katherine M., and Rafferty Jane A.. 2010. “Race and Unhealthy Behaviors: Chronic Stress, the HPA Axis, and Physical and Mental Health Disparities Over the Life Course.” American Journal of Public Health 100(5):933–39. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Johnson-Lawrence Vicki, Galea Sandro, and Kaplan George. 2015. “Cumulative Socioeconomic Disadvantage and Cardiovascular Disease Mortality in the Alameda County Study 1965–2000.” Annals of Epidemiology 25(2):65–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Jones Bobby L. and Nagin Daniel S.. 2013. “A Note on a Stata Plugin for Estimating Group-Based Trajectory Models.” Sociological Methods & Research 42(4):608–13. [Google Scholar]
- Judge Timothy A., Piccolo Ronald F., Podsakoff Nathan P., Shaw John C., and Rich Bruce L.. 2010. “The Relationship between Pay and Job Satisfaction: A Meta-Analysis of the Literature.” Journal of Vocational Behavior 77(2):157–67. [Google Scholar]
- Kalousova Lucie and Burgard Sarah A.. 2014. “Unemployment, Measured and Perceived Decline of Economic Resources: Contrasting Three Measures of Recessionary Hardships and Their Implications for Adopting Negative Health Behaviors.” Social Science & Medicine 106:28–34. [DOI] [PubMed] [Google Scholar]
- Kalleberg Arne L. 2011. Good Jobs, Bad Jobs: The Rise of Polarized and Precarious Employment Systems in the United States, 1970s-2000s. Russell Sage Foundation. [Google Scholar]
- Kalleberg Arne L. 2009. “Precarious Work, Insecure Workers: Employment Relations in Transition.” American Sociological Review 74(1):1–22. [Google Scholar]
- Kawachi Ichiro, Adler Nancy E., and Dow William H.. 2010. “Money, Schooling, and Health: Mechanisms and Causal Evidence.” Annals of the New York Academy of Sciences 1186(1):56–68. [DOI] [PubMed] [Google Scholar]
- Krueger Patrick M. and Burgard Sarah A.. 2011. “Work, Occupation, Income, and Mortality.” Pp. 263–88 in International Handbook of Adult Mortality, International Handbooks of Population, edited by Rogers RG and Crimmins EM. Springer; Netherlands. [Google Scholar]
- Krueger Patrick M. and Chang Virginia W.. 2008. “Being Poor and Coping With Stress: Health Behaviors and the Risk of Death.” American Journal of Public Health 98(5):889–96. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Landes Scott D. 2017. “The Association between Education and Mortality for Adults with Intellectual Disability.” Journal of Health and Social Behavior 58(1):70–85. [DOI] [PubMed] [Google Scholar]
- Lanjouw Peter and Ravallion Martin. 1995. “Poverty and Household Size.” The Economic Journal 105(433):1415–34. [Google Scholar]
- Lee Jinkook and Kim Hyungsoo. 2003. “An Examination of the Impact of Health on Wealth Depletion in Elderly Individuals.” The Journals of Gerontology Series B: Psychological Sciences and Social Sciences 58(2):S120–26. [DOI] [PubMed] [Google Scholar]
- Leeflang RLI, Klein-Hesselink DJ, and Spruit IP. 1992. “Health Effects of Unemployment--II. Men and Women.” Social Science & Medicine 34(4):351–63. [DOI] [PubMed] [Google Scholar]
- Link Bruce G., Lennon Mary Clare, and Dohrenwend Bruce P.. 1993. “Socioeconomic Status and Depression: The Role of Occupations Involving Direction, Control, and Planning.” American Journal of Sociology 98(6):1351–87. [Google Scholar]
- Link Bruce G. and Phelan Jo C.. 1995. “Social Conditions as Fundamental Causes of Disease.” Journal of Health and Social Behavior 35(Extra Issue):80–94. [PubMed] [Google Scholar]
- Masters Ryan K., Link Bruce G., and Phelan Jo C.. 2015. “Trends in Education Gradients of ‘Preventable’ Mortality: A Test of Fundamental Cause Theory.” Social Science & Medicine 127:19–28. [DOI] [PMC free article] [PubMed] [Google Scholar]
- May Martha. 1987. “The Historical Problem of the Family Wage.” in Families and Work. Philadelphia: Temple University Press. [Google Scholar]
- McCall Leslie. 2001. Complex Inequality: Gender, Class, and Race in the New Economy. New York: Routledge. [Google Scholar]
- McLanahan Sara and Percheski Christine. 2008. “Family Structure and the Reproduction of Inequalities.” Annual Review of Sociology 34(1):257–76. [Google Scholar]
- Miech Richard A. 1999. “Low Socioeconomic Status and Mental Disorders: A Longitudinal Study of Selection and Causation during Young Adulthood.” American Journal of Sociology 4(January):1096–1131. [Google Scholar]
- Mishel Lawrence, Bivens Josh, Gould Elise, and Shierholz Heidi. 2012. The State of Working America, 12th Edition. Cornell University Press. [Google Scholar]
- Modrek Sepideh and Cullen Mark R.. 2013. “Health Consequences of the ‘Great Recession’ on the Employed: Evidence from an Industrial Cohort in Aluminum Manufacturing.” Social Science & Medicine 92:105–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Moen Phyllis and Roehling Patricia. 2005. The Career Mystique: Cracks in the American Dream. Rowman & Littlefield. [Google Scholar]
- Musick Kelly and Mare Robert D.. 2004. “Family Structure, Intergenerational Mobility, and the Reproduction of Poverty: Evidence for Increasing Polarization?” Demography 41(4):629–48. [DOI] [PubMed] [Google Scholar]
- Musick Kelly and Mare Robert D.. 2006. “Recent Trends in the Inheritance of Poverty and Family Structure.” Social Science Research 35(2):471–99. [Google Scholar]
- Nagin Daniel. 2005. Group-Based Modeling of Development. Harvard University Press. [Google Scholar]
- National Longitudinal Surveys. n.d. “Appendix 19: SF-12 Health Scale Scoring.” Retrieved August 1, 2018 (https://www.nlsinfo.org/content/cohorts/nlsy79/other-documentation/codebook-supplement/nlsy79-appendix-19-sf-12-health-scale).
- Nau Michael, Dwyer Rachel E., and Hodson Randy. 2015. “Can’t Afford a Baby? Debt and Young Americans.” Research in Social Stratification and Mobility 42:114–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Nelson Deborah Imel, Concha-Barrientos Marisol, Driscoll Timothy, Steenland Kyle, and Fingerhut Marilyn. 2005. “The Global Burden of Selected Occupational Diseases and Injury Risks: Methodology and Summary.” American Journal of Industrial Medicine 48(6):400–418. [DOI] [PubMed] [Google Scholar]
- Nelson MK and Smith J.. 1999. Working Hard and Making Do: Surviving in Small Town America. Berkeley: University of California Press. [Google Scholar]
- Nomaguchi Kei M., Milkie Melissa A., and Bianchi Suzanne M.. 2005. “Time Strains and Psychological Well-Being: Do Dual-Earner Mothers and Fathers Differ?” Journal of Family Issues 26(6):756–92. [Google Scholar]
- Pager Devah, Bonikowski Bart, and Western Bruce. 2009. “Discrimination in a Low-Wage Labor Market: A Field Experiment.” American Sociological Review 74:777–99. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Pavalko Eliza K., Elder Glen H., and Clipp Elizabeth C.. 1993. “Worklives and Longevity: Insights from a Life Course Perspective.” Journal of Health and Social Behavior 34(4):363–80. [PubMed] [Google Scholar]
- Perry Megan, Williams Robert L., Wallerstein Nina, and Waitzkin Howard. 2008. “Social Capital and Health Care Experiences Among Low-Income Individuals.” American Journal of Public Health 98(2):330–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Phelan Jo C. and Link Bruce G.. 2005. “Controlling Disease and Creating Disparities: A Fundamental Cause Perspective.” The Journals of Gerontology: Series B 60(Special Issue 2):S27–33. [DOI] [PubMed] [Google Scholar]
- Phelan Jo C., Link Bruce G., Diez-Roux Ana, Kawachi Ichiro, and Levin Bruce. 2004. “‘Fundamental Causes’ of Social Inequalities in Mortality: A Test of the Theory.” Journal of Health and Social Behavior 45(3):265–85. [DOI] [PubMed] [Google Scholar]
- Phelan Jo C., Link Bruce G., and Tehranifar Parisa. 2010. “Social Conditions as Fundamental Causes of Health Inequalities: Theory, Evidence, and Policy Implications.” Journal of Health and Social Behavior 51(suppl):S28–40. [DOI] [PubMed] [Google Scholar]
- Prince Martin, Patel Vikram, Saxena Shekhar, Maj Mario, Maselko Joanna, Phillips Michael R., and Rahman Atif. 2007. “No Health without Mental Health.” The Lancet 370(9590):859–77. [DOI] [PubMed] [Google Scholar]
- Prus Steven G. 2007. “Age, SES, and Health: A Population Level Analysis of Health Inequalities over the Lifecourse.” Sociology of Health & Illness 29(2):275–96. [DOI] [PubMed] [Google Scholar]
- Ross Catherine E. and Wu Chia-Ling. 1995. “The Links between Education and Health.” American Sociological Review 60(October):719–45. [Google Scholar]
- Rubin Marcie S., Clouston Sean, and Link Bruce G.. 2014. “A Fundamental Cause Approach to the Study of Disparities in Lung Cancer and Pancreatic Cancer Mortality in the United States.” Social Science & Medicine 100:54–61. [DOI] [PubMed] [Google Scholar]
- Saldana-Ruiz Nallely, Clouston Sean AP, Rubin Marcie S., Colen Cynthia G., and Link Bruce G.. 2013. “Fundamental Causes of Colorectal Cancer Mortality in the United States: Understanding the Importance of Socioeconomic Status in Creating Inequality in Mortality.” American Journal of Public Health 103(1):99–104. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Schieman Scott, Glavin Paul, and Milkie Melissa A.. 2009. “When Work Interferes with Life: Work-Nonwork Interference and the Influence of Work-Related Demands and Resources.” American Sociological Review 74(6):966–88. [Google Scholar]
- Schneider Daniel and Reich Adam. 2014. “Marrying Ain’t Hard When You Got A Union Card? Labor Union Membership and First Marriage.” Social Problems 61(4):625–43. [Google Scholar]
- Schnittker Jason. 2004. “Education and the Changing Shape of the Income Gradient in Health.” Journal of Health and Social Behavior 45(3):286–305. [DOI] [PubMed] [Google Scholar]
- Schnittker Jason. 2007. “Working More and Feeling Better: Women’s Health, Employment, and Family Life, 19742004.” American Sociological Review 72:221–38. [Google Scholar]
- Silva Jennifer M. 2013. Coming Up Short: Working-Class Adulthood in an Age of Uncertainty. New York: Oxford University Press. [Google Scholar]
- Slack Tim and Jensen Leif. 2002. “Race, Ethnicity, and Underemployment in Nonmetropolitan America: A 30-Year Profile.” Rural Sociology 67(2):208–33. [Google Scholar]
- Smith James P. 2007. “The Impact of Socioeconomic Status on Health over the Life-Course.” Journal of Human Resources XLII(4):739–64. [Google Scholar]
- Smith Peter and Frank John. 2005. “When Aspirations and Achievements Don’t Meet. A Longitudinal Examination of the Differential Effect of Education and Occupational Attainment on Declines in Self-Rated Health among Canadian Labour Force Participants.” International Journal of Epidemiology 34(4):827–34. [DOI] [PubMed] [Google Scholar]
- Stevens Ann Huff. 1997. “Persistent Effects of Job Displacement: The Importance of Multiple Job Losses.” Journal of Labor Economics 15(1, Part 1):165–88. [Google Scholar]
- Strully Kate W. 2009. “Job Loss and Health in the U.S. Labor Market.” Demography 46(2):221–46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Sweeney Megan M. 2002. “Two Decades of Family Change: The Shifting Economic Foundations of Marriage.” American Sociological Review 67(1):132–47. [Google Scholar]
- Sweeney Megan M. and Raley R. Kelly. 2014. “Race, Ethnicity, and the Changing Context of Childbearing in the United States.” Annual Review of Sociology 40(1):539–58. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Sweeney Megan and Phillips Julie A.. 2004. “Understanding Racial Differences in Marital Disruption: Recent Trends and Explanations.” Journal of Marriage and Family 66(August):639–50. [Google Scholar]
- Tausig Mark. 1999. “Work and Mental Health.” Pp. 255–74 in Handbook of the Sociology of Mental Health, Handbooks of Sociology and Social Research, edited by Aneshensel CS and Phelan JC. Springer US. [Google Scholar]
- Tausig Mark and Fenwick Rudy. 2011. “Occupational Determinants of Job Stress: Socioeconomic Status and Segmented Labor Markets.” Pp. 79–109 in Work and Mental Health in Social Context, Social Disparities in Health and Health Care. Springer, New York, NY. [Google Scholar]
- Thoits Peggy A. 2010. “Stress and Health Major Findings and Policy Implications.” Journal of Health and Social Behavior 51(1 suppl):S41–53. [DOI] [PubMed] [Google Scholar]
- Tiffin Paul A., Pearce Mark S., and Parker Louise. 2005. “Social Mobility over the Lifecourse and Self Reported Mental Health at Age 50: Prospective Cohort Study.” Journal of Epidemiology & Community Health 59(10):870–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Townsend Nicholas. 2002. Package Deal: Marriage, Work And Fatherhood In Men’s Lives. Philadelphia: Temple University Press. [Google Scholar]
- Torche Florencia. 2011. “Is a College Degree Still the Great Equalizer? Intergenerational Mobility across Levels of Schooling in the United States.” American Journal of Sociology 117(3):763–807. [Google Scholar]
- Turner R. Jay. 1995. “The Epidemiology of Social Stress.” American Sociological Review 60(February):104–25. [Google Scholar]
- Turner R. Jay and Avison William R.. 2003. “Status Variations in Stress Exposure: Implications for the Interpretation of Research on Race, Socioeconomic Status, and Gender.” Journal of Health and Social Behavior 44(4):488–505. [PubMed] [Google Scholar]
- Turner R. Jay and Lloyd Donald A.. 1999. “The Stress Process and the Social Distribution of Depression.” Journal of Health and Social Behavior 40(4):371–404. [PubMed] [Google Scholar]
- Umberson Debra and Montez Jennifer Karas. 2010. “Social Relationships and Health A Flashpoint for Health Policy.” Journal of Health and Social Behavior 51(1 suppl):S54–66. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Vespa Jonathan and Painter Matthew. 2011. “Cohabitation History, Marriage, and Wealth Accumulation.” Demography 48(3):983–1004. [DOI] [PubMed] [Google Scholar]
- Marianna Virtanen, Kivimäki Mika, Joensuu Matti, Virtanen Pekka, Elovainio Marko, and Vahtera Jussi. 2005. “Temporary Employment and Health: A Review.” International Journal of Epidemiology 34(3):610–22. [DOI] [PubMed] [Google Scholar]
- Waite Linda J. 1995. “Does Marriage Matter?” Demography 32(4):483–507. [PubMed] [Google Scholar]
- Warren John Robert, Luo Liying, Halpern-Manners Andrew, Raymo James M., and Palloni Alberto. 2015. “Do Different Methods for Modeling Age-Graded Trajectories Yield Consistent and Valid Results?” American Journal of Sociology 120(6):1809–56. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Weeden Kim A. and Grusky David B.. 2005. “The Case for a New Class Map.” American Journal of Sociology 111(1):141–212. [Google Scholar]
- Williams David R., González Hector M., Neighbors Harold, Nesse Randolph, Abelson Jamie M., Sweetman Julie, and Jackson James S.. 2007. “Prevalence and Distribution of Major Depressive Disorder in African Americans, Caribbean Blacks, and Non-Hispanic Whites: Results From the National Survey of American Life.” Archives of General Psychiatry 64(3):305–15. [DOI] [PubMed] [Google Scholar]
- Williams Kristi, Frech Adrianne, and Carlson Daniel L.. 2010. “Marital Status and Mental Health.” Pp. 306–20 in A Handbook for the Study of Mental health, Second Edition: Social contexts, theories, and systems. Eds. Scheid Teresa L. and Brown Tony N.. New York: Cambridge University Press. [Google Scholar]
- Willson Andrea E. and Shuey Kim M.. 2016. “Life Course Pathways of Economic Hardship and Mobility and Midlife Trajectories of Health.” Journal of Health and Social Behavior 57(3):407–22. [DOI] [PubMed] [Google Scholar]
- Zimmer Zachary and House James S.. 2003. “Education, Income, and Functional Limitation Transitions among American Adults: Contrasting Onset and Progression.” International Journal of Epidemiology 32(6):1089–97. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.