Abstract
Background:
We used a relational social-class measure based on property ownership and managerial authority to analyze the longitudinal relationships between class, self-rated health (SRH), and mental illness. To our knowledge, this is the first study using a relational social-class measure to evaluate these relationships longitudinally.
Methods:
Using Panel Study of Income Dynamics data from 1984–2017, we first assigned respondents ages 25–64 to the not-in-the-labor-force (NILF), worker, manager, petit bourgeois (PB), or capitalist classes based on business ownership, managerial authority, and employment status. Next, using Cox models, we estimated the confounder-adjusted associations between two-year-lagged class and incidence of poor/fair SRH and serious mental illness. We also tested whether the associations varied by gender, whether they persisted after more-fully adjusting for traditional socioeconomic-status measures (education and income), and how they changed temporally.
Results:
We identified large inequities in poor/fair SRH. NILFs had the greatest hazard, followed by workers, PBs, managers, and capitalists. We also identified large inequities in serious mental illness; NILFs and workers had the greatest hazard, while capitalists had the lowest. Class inequities in both outcomes lessened but remained considerable after confounder- and socioeconomic-status-adjustment, and we found some evidence that the class-SRH relationship varied by gender, as being NILF was more harmful among men than among women. Additionally, class inequities in the outcomes decreased somewhat over time.
Conclusion:
We identified substantial class inequities in SRH and mental illness. Our findings demonstrate the importance of using relational social-class measures to deepen understanding of the root causes of health inequities.
1. INTRODUCTION
a. Relational social class
Stratificationist (a.k.a. “gradational”) measures of social class based upon socioeconomic status or position dominate health-disparities research.1–3 For example, many epidemiological analyses have documented that U.S. life expectancy has stagnated or declined for the poor and less-educated and risen for the wealthier and more-educated over the last several decades.4 However, because stratificationist measures distinguish classes in terms of intrinsic properties of individuals (e.g., their incomes/educations), they do not reveal how health inequities are shaped by socio-structural relationships (e.g., differences in economic power).1,2 Moreover, since stratificationist measures consider only individual-level factors, they do not reveal the fundamental social processes at the root of the inequities.1,2
In contrast, in epidemiological research, relational measures of social class aim to distinguish classes in terms of the structural causes of health inequities, and to thereby identify social relationships directly affecting health (e.g., those causing stress), and those ultimately causing health inequities (e.g., by producing unequal societal distributions of income and education).2 For example, in Marxist theory, classes are social positions relative to the means of production, and class relations of “domination” and “exploitation”explain substantive inequities.5,6 In capitalism, workers own no productive property and must sell their labor power to capitalists for a wage, while capitalists own productive property, control workers’ labor process (dominating them), and appropriate the value produced by workers’ labor as profits (exploiting them).5–7 Because the material welfare of capitalists and workers are inversely related, class relationships cause material inequities.1 Insofar as these structural relationships shape health outcomes in capitalist societies, relational class measures can be used to identify the fundamental causes of health inequities and to understand temporal and inter-societal patterns of disparities.1,2
b. Neo-Marxist social class
Wright’s neo-Marxist relational theory uses property ownership and managerial authority to distinguish four primary classes:5 1) capitalists own productive property and hire labor, and can live off workers’ labor by appropriating as profit the difference in value between workers’ production and workers’ pay;8 2) the petite bourgeoisie (PB) owns productive property (like capitalists), but PBs do not hire labor and instead must labor themselves (unlike capitalists); 3) managers do not own productive property (unlike capitalists and PBs), but they do enforce company policy and control workers’ labor process (like capitalists), entitling them to substantial income and autonomy; 4) workers, who comprise most of the population,9,10 do not own productive property or enforce company policy; rather, they must sell their labor power and work under terms and conditions set by others. While workers generally lack control over their livelihoods (absent strong unions or political power), managers and PBs occupy ‘contradictory’ class locations, sharing characteristics with workers (a lack of productive property or managerial respectively) and with capitalists (managerial authority or productive property, respectively).
c. Pathways linking health and neo-Marxist social class
Exploitation and domination can cause health inequities.1,11 Capitalists can amass health-promoting assets by increasing profits, which may require intensifying workers’ exploitation and degrading their working conditions, wages, and benefits, thereby exposing workers to occupational hazards and material deprivation.12,13 Moreover, while capitalists may enjoy substantial autonomy and security, workers are dominated and lack control over their livelihoods and the processes and fruits of their labor, which may cause stress, anxiety, and depression.14,15 These health effects may accumulate over the life course, given limited class mobility in the U.S.5
Neo-Marxist theory does not anticipate linear class-health relationships.15,16 For instance, although PBs own productive property, they may lack the resources to compete with capitalists, leading to proletarianization (i.e., falling into the working class)5 and resultant stress and loss of health-promoting assets. Thus, unlike stratificationist approaches, which anticipate linear relationships between socioeconomic status and health, neo-Marxist approaches can explain apparently contradictory class-health relationships, such as the elevated risk of certain mental illnesses among those in the middle classes relative to those in the working or capitalist classes.15
Neo-Marxist theory also does not anticipate identical class-health relationships across racialized groups and genders. Among the working class, white women and people of color, especially women of color, face greater oppression than their non-minoritized counterparts, with deleterious health effects like chronic stress and segregation into low-paying, hazardous jobs.17–19 Moreover, working-class women may be overburdened with domestic labor,20 suggesting a greater vulnerability to ill health than working-class men.18,21 Contrarily, given the gendered division of labor, a heterosexual couple’s economic well-being may depend more heavily on the man’s class position;22 thus, in analyses assessing wage-labor alone, the class-health relationship may appear stronger among men.1
d. Aims
Although prior studies have identified cross-sectional associations between neo-Marxist social class, self-rated health (SRH), and mental illness,1 to our knowledge, none have examined these associations longitudinally. Using Panel Study of Income Dynamics (PSID) data we: 1) estimated the longitudinal, confounder-adjusted associations between class and incidence of poor/fair SRH and serious mental illness, and 2) tested whether the associations varied by gender, whether they persisted after more-fully adjusting for traditional measures of socioeconomic status (education and income), and how they changed temporally.
2. METHODS
a. Study population
The PSID is a panel survey conducted by the University of Michigan.23 In 1968, PSID enrolled a nationally-representative probability sample of US families.23 PSID interviewed these “core” families, as well as subsequent ‘split-off” families (those who moved out of core families to form new, economically-independent families), annually until 1997 and biennially thereafter.23 In PSID, “families” consist of people living in the same household who are related by blood, marriage, or adoption (or of unrelated, permanently co-habiting persons), and who share incomes/expenses.23 Core families and related split-off families constitute “family clans”. Since 1973, most interviews have been conducted via telephone.23 Demographic and socioeconomic questions have been administered since 1968. We used data on family “heads” and heads’ “partners” because non-heads/partners did not have data on all variables of interest.23
b. Class measure
Figure A1 in the Appendix displays how we measured respondents’ social classes at each wave. Capitalists were those who were employed or not in the labor force (NILF, those who identified as “keeping-house”, “retired/disabled”, “student”, or “other” rather than “employed” or “unemployed”) and whose families owned an incorporated business, while PBs were those who were employed or NILF and whose families owned an unincorporated business. Incorporated business owners are three to four times more likely to hire workers than unincorporated business owners and tend to have higher incomes, suggesting the variable distinguishes large and small business owners, a reason the U.S. government uses it for similar purposes.24,25 We included NILF business-owners in the capitalist or PB classes because they may still profit from their businesses even if they do not actively manage them. Meanwhile, managers were those who were employed, whose families did not own a business, and who had an “executive/administrative/managerial” occupation. PSID used 1970 Census occupation codes from 1984–2001, 2000 codes from 2003–2015, and 2010 codes in 2017. We created consistent codes using crosswalks from Autor et al. and NIOSH,26,27 and categorized the resultant occupations into seven categories, one of which was “executive/administrative/managerial”. Workers were those who were employed, whose families did not own a business, and who did not have an “executive/administrative/managerial” occupation. We also included the unemployed in the working class, as many workers cycle between employment and unemployment.9 In our data, the unemployed were eight times as likely to transition into the employed working class than to transition into the next most likely employed class (PBs). Finally, NILFs were those whose families did not own a business and who identified as “keeping-house”, “retired/disabled”, “student”, or “other”. To reduce missingness (~1%), we carried respondents’ class values forwards (and backwards if necessary) for a maximum of one wave.
c. Health measures
From 1984–2017, PSID measured respondents’ SRH using the standard question (“Would you say your health in general is…)”.23 We dichotomized the variable as poor/fair versus good/very-good/excellent because dichotomization improves reliability.28 From 2001–2017, save 2005, PSID administered the Kessler-K6 (K6) to non-proxy respondents.23 The K6 is a six-question scale developed to estimate the prevalence of serious mental illness.29 We dichotomized K6 as <13 versus ≥13, which reliably distinguishes those with and without serious mental illness.29
d. Confounders
Potential confounders included age, race (Black/other/white), gender (female/male), time-varying region of residence (Midwest/Northeast/South/West), time-varying marital status (married or cohabiting/not married or cohabiting), and time-varying education (<HS/HS/some college/≥college).We measured time-varying confounders contemporaneously with the exposure.
e. Sample
Analyses of class and SRH included respondents ages 25–64 to the 1984–2017 waves who did not report baseline poor/fair SRH. Respondents remained in the sample until their last observed wave or the first wave they reported poor/fair SRH, whichever came first. Meanwhile, analyses of class and mental illness included respondents ages 25–64 to the 2001–2017 waves. We assumed that 2005 respondents and proxy respondents did not have K6 values ≥13. Respondents entered our sample at their first K6 measurement and remained in the sample until their last K6 measurement or their first K6 value ≥13, whichever came first. We excluded respondents with baseline K6 values ≥13 and those with fewer than two K6 measurements during follow-up. For both outcomes, we censored respondents who missed a wave of follow-up at their last contiguous wave, and excluded respondents missing any exposure, confounder, or outcome data during follow-up (3%).
f. Primary statistical analyses
To analyze the class-health relationships, we ran Cox proportional hazards models using R’s “survival” package30 to estimate the hazard of each outcome among each class relative to workers’ hazard. In our primary analyses, we used a two-year-lagged time-varying class measure as the exposure (i.e., class measured two-years before the outcome); results from models with unlagged class are in the Appendix. We lagged class to reduce the likelihood of reverse causation (e.g., illness causing respondents to fall into the working class), and because class may take time to affect health. We lagged class two years because the data’s biennial structure post-1997 prevented us from lagging it one year; lagging it more than two years was not possible given our limited sample size. To estimate the total magnitude of class inequities in the outcomes, we first ran minimally-adjusted models that included only gender, age (specified as a 3-knot restricted cubic spline [RCS] using R’s “rms” package31), and year (specified as a 5-knot RCS in the SRH analyses and as a 3-knot RCS in the mental-illness analyses) as covariates. In a second set of models, we added education, race, region, and marital status as covariates. Finally, in a third set of models, we added respondents’ prior exposure (as a four-year lag). We clustered standard errors at the family-clan level. In sensitivity analyses, we addressed possible informative censoring by weighting the Cox models by stabilized inverse-probability-of-censoring weights (IPCW);32 Appendix A4 contains details.
g. Secondary statistical analyses
We conducted several secondary analyses using the two-year lagged Cox models. First, given the gender-class dynamics discussed in the Introduction, and to be consistent with prior research,1 we tested for gender effect-modification of the class-health relationship using class-by-gender interaction terms, the significance of which we evaluated using cluster-robust Wald tests33 calculated via R’s survey package.34 Second, we included two-year-lagged family income –a mediator of the class-health relationship – as an additional covariate in the models to test whether class inequities in health remained after more-fully adjusting for traditional SES measures. Finally, given prior findings of growing socioeconomic inequities in income and health over the study period, we tested whether class inequities in health also grew using class-by-year interaction terms.
3. RESULTS
a. Descriptives
In the lagged SRH analyses, the sample included 18,085 unique respondents with 128,481 observations, while in the lagged K6 analyses, the sample included 8,903 unique respondents with 38,854 observations. At baseline in the lagged SRH analyses, 72% of respondents were workers, 6% were managers, 6% were PB, 3% were capitalists, and 13% were NILF (Table 1).Workers and NILFs tended to have lower incomes and education levels than members of other classes, and were more often Black; NILFs were also more often women. Meanwhile, capitalists tended to have higher incomes and education levels than members of other classes, and were more often white. Additionally, although managers and PBs tended to have similar incomes, PBs tended to be less educated. Finally, among employed respondents, approximately 3% of workers, 4% of managers, 41% of PBs, and 46% of capitalists identified as self-employed.
Table 1.
Demographic and socioeconomic composition of sample at baseline in SRH analyses stratified by the two-year lagged class measure.
| Worker | Manager | Petit bourgeois | Capitalist | NILFa | |
|---|---|---|---|---|---|
| N | 13035 | 1062 | 1156 | 489 | 2343 |
| Income (median; IQR)b | 5.3 [3.1, 8.0] | 7.6 [4.7, 11.2] | 7.1 [4.6, 10.8] | 11.3 [7.2, 16.0] | 4.3 [2.2, 6.8] |
| Age (mean, SD) | 32.2 (8.8) | 34.6 (10.0) | 35.6 (9.8) | 38.8 (10.0) | 33.5 (10.6) |
| Male (%) | 6854 (52.6) | 614 (57.8) | 586 (50.7) | 250 (51.1) | 272 (11.6) |
| Unmarried (%)c | 4208 (32.3) | 274 (25.8) | 155 (13.4) | 42 (8.6) | 480 (20.5) |
| Education (%) | |||||
| <HS | 2317 (17.8) | 69 (6.5) | 132 (11.4) | 22 (4.5) | 596 (25.4) |
| HS | 4720 (36.2) | 249 (23.4) | 411 (35.6) | 129 (26.4) | 969 (41.4) |
| Some college | 3148 (24.2) | 318 (29.9) | 325 (28.1) | 132 (27.0) | 488 (20.8) |
| College+ | 2850 (21.9) | 426 (40.1) | 288 (24.9) | 206 (42.1) | 290 (12.4) |
| Race (%) | |||||
| Black | 4286 (32.9) | 172 (16.2) | 168 (14.5) | 39 (8.0) | 704 (30.0) |
| Other | 939 (7.2) | 52 (4.9) | 71 (6.1) | 29 (5.9) | 184 (7.9) |
| White | 7810 (59.9) | 838 (78.9) | 917 (79.3) | 421 (86.1) | 1455 (62.1) |
| Region (%) | |||||
| Midwest | 3234 (24.8) | 261 (24.6) | 276 (23.9) | 114 (23.3) | 635 (27.1) |
| Northeast | 1931 (14.8) | 196 (18.5) | 179 (15.5) | 123 (25.2) | 329 (14.0) |
| South | 5590 (42.9) | 395 (37.2) | 413 (35.7) | 156 (31.9) | 941 (40.2) |
| West | 2280 (17.5) | 210 (19.8) | 288 (24.9) | 96 (19.6) | 438 (18.7) |
Not in the labor force
Family income in tens of thousands of 2017 dollars. IQR is the interquartile range.
Unmarried or not cohabiting with a partner.
During follow-up, the wave-to-wave probability of transitioning across classes was just 28% (Appendix A2). The most likely class-transitions were from manager-to-worker (32%), NILF-to-worker (29%), and PB-to-worker (23%); the least-likely were from worker-to-capitalist (1%), NILF-to-manager (1%), and NILF-to-capitalist (1%). The probability of a capitalist-to-worker class-transition was just 11%.
b. Class and self-rated health
For the lagged analyses, respondents were followed for a median and maximum of 8 years and 33 years respectively. There were 4,687 events during follow-up, and the probability of survival at the end of follow-up was 82%. In minimally-adjusted models (model 1), class inequities in SRH were large, with managers (HR: 0.50, 95% CI: 0.43, 0.58), PBs (HR: 0.71, 95% CI: 0.63, 0.80), and capitalists (HR: 0.48, 95% CI: 0.39, 0.58) having lower hazards of poor/fair health than workers (Table 2). NILFs had the greatest hazard (HR: 1.24, 95% CI: 1.13, 1.36). The magnitude of the inequities lessened but remained substantial after adjusting for potential confounders (model 2); additionally adjusting for prior exposure did not meaningfully affect the inequities among those in the labor force (model 3). Patterns were similar in unlagged analyses (Appendix A3); however, confounder-adjustment greatly reduced the inequities. IPCW-weighting had little effect (Appendix A4).
Table 2.
Hazard of poor/fair SRH among each class relative to the hazard among workers from Cox proportional hazards models. Class lagged two-years prior to outcome.
| Model 1a | Model 2b | Model 3c | |||||||
|---|---|---|---|---|---|---|---|---|---|
| HR | 95% CI | HR | 95% CI | HR | 95% CI | ||||
| Worker (ref.) | 1.00 | - | - | 1.00 | - | - | 1.00 | - | - |
| Manager | 0.50 | 0.43 | 0.58 | 0.69 | 0.60 | 0.80 | 0.78 | 0.66 | 0.93 |
| Petit bourgeois | 0.71 | 0.63 | 0.80 | 0.92 | 0.82 | 1.03 | 0.89 | 0.77 | 1.02 |
| Capitalist | 0.48 | 0.39 | 0.58 | 0.72 | 0.60 | 0.88 | 0.73 | 0.57 | 0.93 |
| NILF | 1.24 | 1.13 | 1.36 | 1.24 | 1.14 | 1.36 | 1.07 | 0.95 | 1.19 |
Models 1 and 2 used data on 18,085 unique PSID respondents with 128,481
observations; model 3 used data on 16,051 unique PSID respondents with 116,381
observations. Standard errors clustered at family-clan level.
Models adjusted for age, gender, and year.
Models adjusted for age, gender, year, education, race, region, and marital status.
Models adjusted for age, gender, year, education, race, region, marital status, and prior exposure.
c. Class and mental illness
For the lagged analyses, respondents were followed for a median and maximum of 8 years and 16 years respectively. There were 634 events during follow-up, and the probability of survival at the end of follow-up was 96%. In minimally-adjusted models (model 1), class inequities in mental illness were large, with managers (HR: 0.64, 95% CI: 0.41, 0.99) and capitalists (HR: 0.32, 95% CI: 0.16, 0.65) having lower hazards of serious mental illnessthan workers (Table 3). PBs had a similar hazard to workers (HR: 0.99, 95% CI: 0.72, 1.36), while NILFs had the greatest hazard (HR: 2.04, 95% CI: 1.68, 2.47). The magnitude of the inequities lessened but remained substantial after adjusting for potential confounders (model 2), although the estimates’ precision worsened. Adding prior exposure further reduced the inequities (model 3). Patterns were similar in unlagged analyses, although capitalists’ hazard more-closely resembled workers’ hazard after confounder-adjustment (Appendix A3). IPCW-weighting had little effect (Appendix A4).
Table 3.
Hazard of serious mental illness (i.e., K6 ≥ 13) among each class relative to the hazard among workers from Cox proportional hazards models. Class lagged two-years prior to outcome.
| Model 1a | Model 2b | Model 3c | |||||||
|---|---|---|---|---|---|---|---|---|---|
| HR | 95% CI | HR | 95% CI | HR | 95% CI | ||||
| Worker (ref.) | 1.00 | - | - | 1.00 | - | - | 1.00 | - | - |
| Manager | 0.64 | 0.41 | 0.99 | 0.85 | 0.54 | 1.33 | 0.98 | 0.61 | 1.58 |
| Petit bourgeois | 0.99 | 0.72 | 1.36 | 1.21 | 0.88 | 1.67 | 1.25 | 0.87 | 1.80 |
| Capitalist | 0.32 | 0.16 | 0.65 | 0.47 | 0.23 | 0.94 | 0.46 | 0.20 | 1.05 |
| NILF | 2.04 | 1.68 | 2.47 | 1.89 | 1.55 | 2.30 | 1.60 | 1.24 | 2.06 |
Models 1 and 2 used data on 8,903 unique PSID respondents with 38,854
observations; model 3 used data on 8,322 unique PSID respondents with 36,295
observations. Standard errors clustered at family-clan level.
Models adjusted for age, gender, and year.
Models adjusted for age, gender, year, education, race, region, and marital status.
Models adjusted for age, gender, year, education, race, region, marital status, and prior exposure.
d. Secondary analyses
We found some evidence that the class-health relationship varied by gender in the SRH analyses. Across both outcomes, being a NILF tended to be somewhat more harmful among men than among women (i.e., the ratio of male NILFs’ hazard to male workers’ hazard was greater than the ratio of female NILFs’ hazard to female workers’ hazard); additionally, in the K6 analyses, being a capitalist tended to be somewhat less protective among men than among women (Tables 4 and 5). However, the class-by-gender interaction terms were only significant in the SRH analyses (p<0.05).
Table 4.
Hazard of poor/fair SRH among each gender -class group relative to the hazard among male workers from Cox proportional hazards models. Class lagged two-years prior to outcome.
| Model 1a | Model 2b | Model 3c | |||||||
|---|---|---|---|---|---|---|---|---|---|
| HR | 95% CI | HR | 95% CI | HR | 95% CI | ||||
| Male worker (ref.) | 1.00 | - | - | 1.00 | - | - | 1.00 | - | - |
| Male manager | 0.46 | 0.38 | 0.57 | 0.69 | 0.56 | 0.84 | 0.80 | 0.64 | 0.99 |
| Male petit bourgeois | 0.75 | 0.64 | 0.87 | 0.95 | 0.81 | 1.11 | 0.91 | 0.76 | 1.09 |
| Male capitalist | 0.51 | 0.39 | 0.65 | 0.75 | 0.59 | 0.97 | 0.75 | 0.56 | 1.02 |
| Male NILF | 1.60 | 1.31 | 1.95 | 1.56 | 1.28 | 1.90 | 1.28 | 1.02 | 1.61 |
| Female worker | 1.09 | 1.02 | 1.17 | 1.05 | 0.98 | 1.12 | 1.01 | 0.94 | 1.09 |
| Female manager | 0.60 | 0.48 | 0.74 | 0.74 | 0.59 | 0.91 | 0.78 | 0.62 | 0.99 |
| Female petit bourgeois | 0.73 | 0.62 | 0.86 | 0.93 | 0.79 | 1.10 | 0.87 | 0.72 | 1.04 |
| Female capitalist | 0.49 | 0.38 | 0.64 | 0.73 | 0.56 | 0.94 | 0.71 | 0.52 | 0.97 |
| Female NILF | 1.29 | 1.18 | 1.41 | 1.24 | 1.13 | 1.36 | 1.03 | 0.92 | 1.17 |
| Interaction p-valued | <0.01 | <0.01 | <0.05 | ||||||
Models 1 and 2 used data on 18,085 unique PSID respondents with 128,481 observations; model 3 used data on 16,051 unique PSID respondents with 116,381 observations. Standard errors clustered at family-clan level.
Models adjusted for age and year.
Models adjusted for age, year, education, race, region, and marital status.
Models adjusted for age, year, education, race, region, marital status, and prior exposure.
P-value from cluster-robust Wald test of interaction term.
Table 5.
Hazard of serious mental illness (i.e., K6 ≥ 13) among each gender-class group relative to the hazard among male workers from Cox proportional hazards models. Class lagged two-years prior to outcome.
| Model 1a | Model 2b | Model 3c | |||||||
|---|---|---|---|---|---|---|---|---|---|
| HR | 95% CI | HR | 95% CI | HR | 95% CI | ||||
| Male worker (ref.) | 1.00 | - | - | 1.00 | - | - | 1.00 | - | - |
| Male manager | 0.63 | 0.30 | 1.29 | 0.93 | 0.45 | 1.93 | 0.90 | 0.37 | 2.16 |
| Male petit bourgeois | 0.97 | 0.58 | 1.64 | 1.13 | 0.67 | 1.89 | 1.15 | 0.64 | 2.06 |
| Male capitalist | 0.50 | 0.19 | 1.36 | 0.70 | 0.26 | 1.87 | 0.69 | 0.25 | 1.85 |
| Male NILF | 2.98 | 1.97 | 4.50 | 2.53 | 1.67 | 3.84 | 2.20 | 1.38 | 3.50 |
| Female worker | 1.50 | 1.22 | 1.85 | 1.48 | 1.20 | 1.83 | 1.43 | 1.15 | 1.79 |
| Female manager | 0.98 | 0.55 | 1.75 | 1.20 | 0.67 | 2.14 | 1.47 | 0.83 | 2.58 |
| Female petit bourgeois | 1.52 | 1.03 | 2.25 | 1.88 | 1.27 | 2.80 | 1.88 | 1.22 | 2.89 |
| Female capitalist | 0.36 | 0.13 | 0.96 | 0.52 | 0.19 | 1.40 | 0.49 | 0.15 | 1.61 |
| Female NILF | 2.82 | 2.17 | 3.65 | 2.61 | 2.00 | 3.39 | 2.13 | 1.53 | 2.96 |
| Interaction p-valued | 0.41 | 0.45 | 0.23 | ||||||
Models 1 and 2 used data on 8,903 unique PSID respondents with 38,854 observations; model 3 used data on 8,322 unique PSID respondents with 36,295 observations. Standard errors clustered at family-clan level.
Models adjusted for age and year.
Models adjusted for age, year, education, race, region, and marital status.
Models adjusted for age, year, education, race, region, marital status, and prior exposure.
P-value from cluster-robust Wald test of interaction term.
Adjusting for family income slightly attenuated the class inequities (Appendix A5). However, in the SRH analyses, managers’ and capitalists’ hazard remained lower than workers’ hazard, while in the mental-illness analyses, capitalists’ hazard remained lower than workers’ hazard and NILFs’ hazard remained higher than workers’ hazard.
Finally, class inequities in SRH and mental illness lessened somewhat over the study period (Appendix A6), and the Wald tests of the class-by-year interaction terms were significant (p<0.05) in all models.
4. DISCUSSION
a. Summary of results
Applying a neo-Marxist social-class theory, we examined the associations between class, poor/fair SRH, and serious mental illness. We identified substantial class inequities in the hazard of the outcomes, with NILFs typically having the greatest hazard, followed by workers, PBs, managers, and capitalists. The magnitude of these inequities largely did not vary by gender; however, in the SRH analyses, the hazard associated with being NILF was greater among men than among women. The inequities lessened but remained considerable after adjusting for potential confounders and prior exposure, suggesting that the disparate class distribution of socioeconomic and demographic factors, as well as earlier-life class positions, explained part – but not all – of the inequities. That the inequities remained after additionally adjusting for family income – a mediator of the class-health relationship – suggests that class differences in income also cannot explain the inequities, and that class exploitation and domination harm health beyond their effects on traditional SES measures, although formal mediation approaches are needed to more-rigorously elucidate mediating pathways. We note that in addition to income, education could also be a mediator of the class-health relationship rather than a confounder, given limited class mobility in the U.S. and an educational system that reproduces class inequities.2,5,35 We also believe reverse causation cannot totally explain the inequities because: 1) the inequities were generally larger in lagged models than in unlagged models,and 2) limited class mobility means that class often transmits inter-generationally.5 However, given the study’s U.S. context, where illness can cause catastrophic medical expenses and income loss from missed work,36 it is possible that illness impeded certain respondents’ upward mobility or precipitated others’ proletarianisation. Finally, unlike in prior studies of socioeconomic inequities in health and income, we found that class inequities in health have lessened somewhat over the last several decades. However, our trend estimates should be interpreted cautiously because estimated SRH trends can vary considerably across surveys.37
b. Comparison with prior research
To our knowledge, no prior research has analyzed the relationship between neo-Marxist social class and incidence of these outcomes. In a 2015 review article, Muntaner et al. found that in the cross-sectional studies under review, capitalists and managers tended to report better SRH than others, even in studies which adjusted for socioeconomic status, findings that align with ours.1 Later studies report similar findings.10,38,39 In Muntaner et al., studies with mental-illness outcomes are less comparable to our study, since many involved classes we did not analyze, like low-level managers. Comparability aside, however, all found mental-illness inequities across classes, and many found non-linear class-health relationships, which we also found given the elevated hazard of mental illness among PBs. Among studies of SRH or mental illness that tested for gender effect-modification, most found that class inequities were larger among men than among women, which we found in the SRH analyses among those not in the labor force, but not among others. In our study, the lack of effect modification among those in the labor force may be because we used a class measure that incorporated family-level factors; given the gendered division of labor, women’s health may depend more heavily on their partner’s class than men’s health, an issue mitigated by our use of family-level business ownership.
c. Strengths and limitations
Strengths of our analyses included first, a longitudinal design with three decades of follow-up and repeated measures of social class and health. To our knowledge, this is the first study on the topic to employ such a design. Additionally, our access to business-ownership data may have allowed us to more-accurately identify capitalists and PBs than prior studies that relied upon self-employment data, as some individuals who identify as self-employed, such as gig workers, may truly belong to the working class rather than to the capitalist or PB classes.40
Our analyses also had limitations. First, serious-mental-illness’s rarity compromised precision, particularly in the gender-effect-modification analyses. Second, PSID did not ask respondents about their workplace supervisory authority, a question frequently used in neo-Marxist analyses to distinguish classes.1 This has several implications. One, given the lower proportion of managers identified in our study than in prior studies, our worker category likely included some respondents with supervisory authority that would have otherwise been categorized as managers. This misclassification would make managers’ health appear spuriously like workers’ health. Two, we were unable to distinguish low-level managers from high-level managers, preventing us from testing hypotheses about contradictory class locations that have yielded fruitful findings in prior research.1 Finally, although incorporated-business owners are more likely to employ workers than unincorporated-business owners,24,25 our capitalist category likely included some owners who did not employ workers (and vice versa). However, to our knowledge, no other long-running longitudinal surveys have consistent data on supervisory authority and health.
5. CONCLUSION
We identified substantial inequities in incidence of poor/fair SRH and serious mental illness across relational social classes that persisted after adjustment for demographic and socioeconomic indicators, adding to the growing evidence connecting capitalist class relationships to health inequities. Because they produce differences in socioeconomic status, these relationships are at the root of the socioeconomic inequities frequently studied in public-health research. Thus, researchers and practitioners working to rectify health inequities should consider relational social class measures in their work.
Supplementary Material
THUMBNAIL SKETCH.
What is already known on this subject?
Prior studies using relational measures of social class based on property ownership and managerial authority have identified inequities in the prevalence of poor/fair self-rated health and mental illness across classes.
However, nearly all prior studies have been cross-sectional; thus, the relationship between relational social class and incidence of poor/fair self-rated health or mental illness remains uncertain.
What does this study add?
Applying a relational social-class theory, we identified substantial inequities in the hazard of poor/fair self-rated health and serious mental illness across social classes, with workers and those not in the labor force typically having the greatest hazard, followed by the petit bourgeoisie, managers, and capitalists.
These inequities, which varied somewhat by gender in the SRH analyses, persisted after adjustment for confounders, prior exposure, and socioeconomic status.
Acknowledgments
Funding Statement
JEG’s and AH’s work was partly supported by a grant from the National Institute on Aging of the National Institutes of Health (R01AG060011).
Footnotes
Competing Interest
Competing Interest: None to declare.
REFERENCES
- 1.Muntaner C, Ng E, Chung H, Prins SJ. Two decades of neo-Marxist class analysis and health inequalities: a critical reconstruction. Soc Theory Heal. 2015;13(3–4):267–287. doi: 10.1057/sth.2015.17 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Krieger N, Williams DR, Moss NE. Measuring social class in US public health research: concepts, methodologies, and guidelines. Annu Rev Public Health. 1997;18(1):341–378. doi: 10.1146/annurev.publhealth.18.1.341 [DOI] [PubMed] [Google Scholar]
- 3.Lahelma E, Rahkonen O. Class and health in changing societies : the need for novel approaches. J Epidemiol Community Health. 2019:1–2. doi: 10.1136/jech-2019-213325 [DOI] [PubMed] [Google Scholar]
- 4.Bor J, Cohen GH, Galea S. Population health in an era of rising income inequality: USA, 1980–2015. Lancet. 2017;389(10077):1475–1490. doi: 10.1016/S0140-6736(17)30571-8 [DOI] [PubMed] [Google Scholar]
- 5.Wright EO. Class Counts: Comparative Studies in Class Analysis. New York: Cambridge University Press; 1997. [Google Scholar]
- 6.Wright EO. Understanding class: towards an integrated analytical approach. New Left Rev. 2009;60(Nov/Dec):101–116. [Google Scholar]
- 7.Muntaner C, Benach J, Hadden WC, Gimeno D, Benavides FG. A glossary for the social epidemiology of work organisation: part 2 Terms from the sociology of work and organisations. J Epidemiol Community Health. 2006;60(12):1010–1012. doi: 10.1136/jech.2004.032649 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Resnick S, Wolff R. The diversity of class analysis: a critique of Erik Olin Wright and beyond. Crit Sociol. 2003;29(1):7–27. doi: 10.1163/156916303321780537 [DOI] [Google Scholar]
- 9.Braverman H. The structure of the working class and its reserve armies In: Labor and Monopoly Capital. 25th Anniv. New York: Monthly Review Press; 1998:261–278. [Google Scholar]
- 10.Eisenberg-Guyot J, Prins SJ. Relational social class, self-rated health, and mortality in the United States. Int J Heal Serv. 2019;50(1):7–20. doi: 10.1177/0020731419886194 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.McCartney G, Bartley M, Dundas R, et al. Theorising social class and its application to the study of health inequalities. SSM - Popul Heal. 2019;7. doi: 10.1016/j.ssmph.2018.10.015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Foster JB, Clark B. The robbery of nature. Mon Rev. 2018;70(3):1–20. doi: 10.14452/mr070-03-2018-07_1 [DOI] [Google Scholar]
- 13.Engels F. The Condition of the Working Class in England. Oxford: Oxford University Press; 1993. [Google Scholar]
- 14.Borrell C, Muntaner C, Benach J, Artazcoz L. Social class and self-reported health status among men and women: what is the role of work organisation, household material standards and household labour? Soc Sci Med. 2004;58(10):1869–1887. doi: 10.1016/S0277-9536(03)00408-8 [DOI] [PubMed] [Google Scholar]
- 15.Prins SJ, Bates LM, Keyes KM, Muntaner C. Anxious? Depressed? You might be suffering from capitalism: contradictory class locations and the prevalence of depression and anxiety in the USA. Sociol Heal Illn. 2015;37(8):1352–1372. doi: 10.1111/1467-9566.12315 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Muntaner C, Borrell C, Benach J, Pasarín MI, Fernandez E. The associations of social class and social stratification with patterns of general and mental health in a Spanish population. Int J Epidemiol. 2003;32(6):950–958. doi: 10.1093/ije/dyg170 [DOI] [PubMed] [Google Scholar]
- 17.Bailey ZD, Krieger N, Agénor M, Graves J, Linos N, Bassett MT. Structural racism and health inequities in the USA: evidence and interventions. Lancet. 2017;389(10777):1453–1463. doi: 10.1016/S0140-6736(17)30569-X [DOI] [PubMed] [Google Scholar]
- 18.Krieger N, Rowley DL, Herman AA, Avery B, Phillips MT. Racism, sexism, and social class: implications for studies of health, disease, and well-being. Am J Prev Med.1993;9(6):82–122. doi: 10.1016/S0749-3797(18)30666-4 [DOI] [PubMed] [Google Scholar]
- 19.Robinson CJ. Black Marxism: The Making of the Black Radical Tradition. Chapel Hill:University of North Carolina Press; 2005. [Google Scholar]
- 20.Bhattacharya T. Liberating women from “political economy.” Mon Rev. 2020;71(8). [Google Scholar]
- 21.McKetta S, Prins SJ, Platt J, Bates LM, Keyes K. Social sequencing to determine patterns in health and work-family trajectories for U.S. women, 1968–2013. SSM - Popul Heal.2018;6(June):301–308. doi: 10.1016/j.ssmph.2018.10.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Ferguson A, Hennessy R, Mechthild N. Feminist perspectives on class and work In: The Stanford Encyclopedia of Philosophy. Spring; ; 2019https://plato.stanford.edu/archives/spr2019/entries/feminism-class/. [Google Scholar]
- 23.McGonagle KA, Schoeni RF, Sastry N, Freedman VA. The Panel Study of Income Dynamics: overview, recent innovations, and potential for life course research. Longit Life Course Stud. 2012;3(2):1–21. doi: 10.14301/llcs.v3i2.188 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Kochhar R, Parker K, Rohal M. Three-in-Ten U.S. Jobs Are Held by the Self-Employed and the Workers They Hire.; 2015. https://www.pewsocialtrends.org/2015/10/22/three-in-ten-u-s-jobs-are-held-by-the-self-employed-and-the-workers-they-hire/.
- 25.Hipple SF, Hammond LA. Self-Employment in the United States. Washington D.C.; 2016https://www.bls.gov/spotlight/2016/self-employment-in-the-united-states/pdf/self-employment-in-the-united-states.htm.
- 26.Autor D, Dorn D. The growth of low skill service jobs and the polarization of the U.S. labor market. Am Econ Rev. 2013;103(5):1553–1597. doi: 10.1257/aer.103.5.1553 [DOI] [Google Scholar]
- 27.National Institute for Occupational Health and Safety. Industry and occupation coding. https://www.cdc.gov/niosh/topics/coding/nioccsuserdocumentation.html. Published 2019.
- 28.Zajacova A, Dowd JB. Reliability of self-rated health in US adults. Am J Epidemiol.2011;174(8):977–983. doi: 10.1093/aje/kwr204 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Kessler RC, Green JG, Gruber MJ, et al. Screening for serious mental illness in the general population with the K6 screening scale: results from the WHO World Mental Health (WMH) survey initiative. Int J Methods Psychiatr Res. 2010;19:4–22. doi: 10.1002/mpr.310 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Therneau TM. survival. 2019:1–161. https://cran.r-project.org/web/packages/survival/survival.pdf.
- 31.Harrell FE. rms. 2019:1–246. https://cran.r-project.org/web/packages/rms/rms.pdf.
- 32.Cole SR, Hernán MA. Constructing inverse probability weights for marginal structural models. Am J Epidemiol. 2008;168(6):656–664. doi: 10.1093/aje/kwn164 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Kaufman JS, MacLehose RF. Which of these things is not like the others. Cancer.2013;119(24):4216–4222. doi: 10.1002/cncr.28359 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Lumley T. survey. 2019:1–138. https://cran.r-project.org/web/packages/survey/survey.pdf.
- 35.Chetty R, Friedman JN, Saez E, Turner N, Yagan D. Mobility Report Cards: The Role of Colleges in Intergenerational Mobility. Chicago; 2017. http://www.equality-of-opportunity.org/papers/coll_mrc_paper.pdf.
- 36.Himmelstein DU, Thorne D, Warren E, Woolhandler S. Medical bankruptcy in the United States, 2007: results of a national study. Am J Med. 2009;122(8):741–746. doi: 10.1016/j.amjmed.2009.04.012 [DOI] [PubMed] [Google Scholar]
- 37.Salomon JA, Nordhagen S, Oza S, Murray CJ. Are Americans feeling less healthy? The puzzle of trends in self-rated health. Am J Epidemiol. 2009;170(3):343–351. doi: 10.1093/aje/kwp144 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Kong KA, Khang YH, Cho HJ, Jang SM, Jung-Choi K. Neo-Marxian social class inequalities in self-rated health among the employed in South Korea: the role of material, behavioral, psychosocial, and workplace environmental factors. BMC Public Health.2017;17(1):1–12. doi: 10.1186/s12889-017-4269-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Kokkinen L, Muntaner C, Koskinen A, Väänänen A. Occupational class, capitalist class advantage and mortality among working-age men. J Epidemiol Community Health. 2019:1–4. doi: 10.1136/jech-2019-212952 [DOI] [PubMed] [Google Scholar]
- 40.Collins B, Garin A, Jackson E, Dmitri K, Payne M. Is Gig Work Replacing Traditional Employment? Evidence from Two Decades of Tax Returns. Washington D.C.; 2019. https://www.irs.gov/pub/irs-soi/19rpgigworkreplacingtraditionalemployment.pdf.
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.