Education and Income Better Boost the Health of Non-Hispanic White than Black Americans

Background. Socioeconomic status (SES) indicators such as education and income inuence population health including but not limited to self-rated health (SRH). Based on Minorities' Diminished Returns (MDRs), however, we observe weaker health effects SES indicators for non-Hispanic Blacks compared to non-Hispanic Whites. While such diminished returns of SES resources are shown in nationally representative data and Michigan data, there is still a need to replicate these ndings in other settings. This is particularly relevant because the effects of social SES indicators are not universal but vary across geographic locations. This cross-sectional study tested racial differences in the effects of high educational attainment and income on SRH among Californian adults. Methods. Survey of California Adults on Serious Illness and End-of-Life is a representative survey of Californian adults conducted for the California Health Care Foundation. The independent variables were educational attainment and income. The main outcome was poor/fair SRH. Age, gender, marital status, and employment were control variables (confounders). Racial group membership was the effect modier. Linear regression model was utilized to analyze the data. statistical

As mentioned above, however, most of this data is derived from studies that are conducted in a national sample. That is, we are still unaware of these patterns at local levels. As states and locations may vary in the resources and how resources can operate (67)(68)(69), it is important to conduct local research on whether or not educational attainment, income, and employment generate similar or different health gain for non-Hispanic Blacks than non-Hispanic Whites. The results derived from such studies can potentially inform local policy on the most effective interventions that can equalize health across racial groups, particularly non-Hispanic Blacks and non-Hispanic Whites.
This cross-sectional study used a representative sample of adult residents of California to explore racial variation in the effects of educational attainment and income on SRH. In line with the MDRs literature, we expected weaker effects for non-Hispanic Blacks and non-Hispanic Whites. As we expect MDRs to be related to labor market discrimination and differential availability of jobs in non-Hispanic Black and non-Hispanic White communities, we expected MDRs of education to be partially explained by income (44). That is, one reason education generates less SRH for non-Hispanic Blacks and non-Hispanic Whites is that education generates less income for non-Hispanic Blacks and non-Hispanic Whites.

Design and Settings
Survey of California Adults on Serious Illness and End-of-Life is a cross-sectional representative survey of residents of California in 2019. The study elded June 6 through July 2, 2019. The study was conducted for the California Health Care Foundation. The survey was administered primarily using Ipsos' KnowledgePanel,

Participants and Sampling
This study included 2,588 California adults 18 years old and older. Eligibility were 1) being a California resident, and having an age of 18 or older. Participants were drawn from Ipsos' KnowledgePanel, the rst representative online research panel in the United States. Panel members are randomly recruited through address-based sampling methods. All households are provided with Internet access as well as hardware that might be needed. The survey also increased the number of Black respondents by using supplemental, nonprobability sampling.

Survey weights
Overall, the sample for this survey was designed to target the following numbers of respondents: 1) 800 Californians under 150% federal poverty level, 2) 800 Californians between 150% and 399% federal poverty level, and 3) 800 Californians at 400% federal poverty level and above. After survey data were all collected, cleaned, and processed, design weights were calculated to account for nonresponse as well as strati cation. We applied weights that generates results representative of the California adults.

Process
The survey was conducted in English and Spanish. Participants answered the questions of the survey in their home online.
Income. Income was measured as annual last year income. Income was a 21-level continuous measure. 1) Less than $5,000, 2) $5,000 to $7,499, 3) $7,500 to $9,999, 4) $10,000 to $12,499, 5)  Covariates Sociodemographic Factors. Sociodemographic control variables included gender, age, marital status, and employment. Gender was a dichotomous measure [male = 1, female = 0 (reference group)]. Age was a continuous variable. Marital status was a dichotomous variable: married = 1, other = 0. Employment was a dichotomous variable: Employed = 1, unemployed, not in labor market, searching for job, retired, or disabled. Employed included those who conducted any work for pay or those who were self-employed.

Dependent Variable
Self-Rated Health (SRH). SRH was measured using a single item (70)(71)(72)(73). Participants reported their overall health. Response options included excellent, very good, good, fair, or poor. The Institute of Medicine recommended use of SRH for monitoring the health of the US general population. Similar to the literature, we treated SRH as a dichotomous variable. As such, we combined the poor and fair categories and compared it to other responses. Poor/fair SRH was coded as 1. Poor SRH is high valid, as it independently predicts risk of mortality above and beyond traditional risk factors (70)(71)(72)(73) .

Statistics
We applied SPSS 23.0 (IBM Inc, NY, U.S.) to perform our data analysis. For descriptive statistics, we reported means and proportions (frequencies). For bivariate analysis, we used Chi Square or independent sample t test. For multivariable models, we ran nine linear regression models. In all of these models, educational attainment and income were the independent variables, SRH (poor/fair) was the dependent variables, demographics, marital status, and employment were the control variables, and race was the focal moderator. The rst ve linear regression models were estimated in the total sample that included non-Hispanic Blacks and non-Hispanic Whites. These models rst did not include race by educational attainment or income interaction terms. In the next models, we gradually added race by educational attainment or income interaction terms. Finally, we tested race/speci c models in non-Hispanic Whites and non-Hispanic Blacks. The rst race-speci c model only had education. The second race-speci c model also controlled for income. Adjusted b (b), 95% Con dence Intervals (9% CI), and p-values were reported. P-values equal or less than 0.05 was signi cant.

Results
Descriptive Statistics Table 1 summarizes the descriptive statistics of the participants overall and by race (n = 1627). From all the participants, 714 were non-Hispanic Blacks, 913 were non-Hispanic Whites. Overall Regressions Table 2 provides a summary of the results of our linear regression models that were conducted in the pooled sample in which high educational attainment and income were the independent variables and SRH was the dependent variables. Higher levels of educational attainment and income were associated with better health, above and beyond the effects of covariates. We found an interaction between race and educational attainment and income on SRH. This nding suggested that the boosting effects of high educational attainment and income on health were smaller for non-Hispanic Blacks than non-Hispanic Whites. The protective effect of high educational attainment was partially explained by income. The statistical interaction between race and education was explained by the interaction between race and income suggesting that MDRs of education is possibly due to MDRs of income (Table 2).  Table 3 provides a summary of the results of our race-strati ed linear regression models in which high educational attainment and income were the predictor / independent variables and SRH was the outcome / dependent variable. Higher levels of educational attainment and income were associated with better health, above and beyond the effects of covariates. However, the protective effect of education was smaller for non-Hispanic Blacks than non-Hispanic Whites. Similarly, while both education and income were in the model, both education and income were protective for non-Hispanic Whites, however, only education but not income remained as signi cant for non-Hispanic Blacks (Table 3).

Discussion
The current study showed ve ndings: First, as expected, higher education and income were associated with better SRH of Californian adults, overall. Second, the boosting effect of education and income on SRH were smaller for Non-Hispanic Blacks than Non-Hispanic Whites. Third, MDRs of income partially explained differential effects of educational attainment for Non-Hispanic Blacks.
Our rst nding is a successful replication of what we know about SES, Social determinants of health, and fundamental causes (1-3, 8, 74-78). High education and income prove health, and Californian adults are not an exception. Thus, increasing overall education and income would increase overall health.
This 2nd nding is also a successful replication of what we already knew about MDRs (minorities' diminished returns) in the US and MI. Thus, the same pattern also applies to Californian adults. This nding is in line with the other observations that educational attainment and other SES indicators generate less than expected outcomes for non-Hispanic Blacks (59). Although similar patterns are shown for Hispanics (34,35), in this study we only found MDRs for non-Hispanic Blacks. This observation contributes to the debate whether all minority groups and marginalized groups de ned based on race(59), ethnicity (34,35), or sexual orientation(39) experience MDRs or not. Although all may experience some diminished health gains from the very same SES resources (59), the operant mechanisms of the MDRs may differ from one to another marginalizing identity.
The 3rd nding is proposing that labor market discrimination may have a role in reducing health returns of education for non-Hispanic Blacks. While years of schooling generates less health for non-Hispanic Blacks, this effect is mainly due to income differentials. Previous work has documented MDRs in terms of effects of educational attainment on job quality. For example, highly educated Black people work in high stress jobs, while highly educated White people work in low stress jobs (47). Another study showed that highly educated non-Hispanic Black people work in jobs that are high in second-hand tobacco exposure, while highly educated non-Hispanic White people work in jobs with low second-hand tobacco exposure (45,46). A number of studies have shown that education generates less income in Blacks than Whites (48,49). In another study, controlling for income explained MDRs of education on SRH. These are also indicative of the role of labor market discrimination and differential job availability for Blacks and Whites (44,48,49,79).
All marginalization processes that reduce returns of SES for Blacks include segregation, discrimination, low quality of education, and labor market discrimination. All these factors may have a role in why education gives more health to non-Hispanic Whites than non-Hispanic Blacks. This observation is in line with other economic observations that non-Hispanic Blacks gain less overall bene ts, while non-Hispanic Whites gain the highest levels of tangible outcomes, from the very same social resources (22,23). For example, education and income predicts affect and happiness for non-Hispanic Whites but not non-Hispanic Blacks (49). Similarly, non-Hispanic Blacks with high education stay at risk of poverty, compared to their non-Hispanic White counterparts (44,48).
We only studied MDRs of educational attainment. As shown before, MDRs of income(11), occupation (33), and marital status (80) are well-established. However, the MDRs seem to be more pronounced for educational attainment than other SES indicators such as income, simply because more unfair societal processes can interrupt education than utilization of income. Beyond the effects of education and income on SRH (43,44), affect(49) they also less effectively reduce rate of chronic medical conditions (11) such as asthma (81), depression (82,83), obesity (24,63)and attention de cit hyperactivity disorder (25) for non-Hispanic Blacks compared to non-Hispanic Whites. None of these patterns have been previously shown for Hispanics compared to non-Hispanics. For Hispanics, previous research has shown that family SES generates less protection against substance use (34,36). The contribution of this study is to propose MDRs of access to the health care system as a mechanism behind MDRs of educational attainment for Blacks.

Limitations
Our study has a few methodological and conceptual limitations. The rst limitation is the cross-sectional design of the study. Given the design, we cannot draw any causal inferences between SES indicators and SRH. As SES and health have bidirectional association, and healthy individuals are more likely to climb the social ladder, we can nor rule out reverse causality. Second, the study did not measure contextual, structural, and environmental factors that may explain MDRs of income. Contextual measures such as density of racial groups, availability of SES resources and jobs, availability of resources and services, and types of available jobs and occupations may be responsible for MDRs observed in this study. Other SES indicators such as wealth were not considered in this study. In addition, all variables in this study were at an individual level. Area level SES may be an unmeasured confounder in this study.

Conclusion
The boosting effect of income on health is smaller for non-Hispanic Blacks than non-Hispanic. To undo MDRs-related health disparities, there is a need to go beyond equalizing SES by income-redistribution policies and equalize how racial groups purchasing power and access to goods and services across all social groups. Such efforts may require structural changes such as reducing segregation and social strati cation, and discrimination. We need to reduce inequalities in the life conditions of all social groups who have similar SES resources.

Ethics
The current analysis is exempt from s separate IRB review because we only used fully deidenti ed data. The authors declare no con icts of interest.