Article Text

Do wealth disparities contribute to health disparities within racial/ethnic groups?
1. Craig Evan Pollack1,
2. Catherine Cubbin2,
3. Ayesha Sania3,
4. Mark Hayward2,
5. Donna Vallone4,
6. Brian Flaherty5,
7. Paula A Braveman6
1. 1Johns Hopkins School of Medicine and Bloomberg School of Public Health, Baltimore, Maryland, USA
2. 2Population Research Center, University of Texas at Austin, Austin, Texas, USA
3. 3Harvard University, Boston, Massachusetts, USA
4. 4American Legacy Foundation, Washington, DC, USA
5. 5University of Washington, Seattle, Washington, USA
6. 6University of California, San Francisco, California, USA
1. Correspondence to Dr Craig Evan Pollack, Johns Hopkins School of Medicine and Bloomberg School of Public Health, 2024 E. Monument Street, Rm 2-521, Baltimore, MD 21287, USA; cpollac2{at}jhmi.edu

Abstract

Background Though wide disparities in wealth have been documented across racial/ethnic groups, it is largely unknown whether differences in wealth are associated with health disparities within racial/ethnic groups.

Methods Data from the Survey of Consumer Finances (2004, ages 25–64) and the Health and Retirement Survey (2004, ages 50+), containing a wide range of assets and debts variables, were used to calculate net worth (a standard measure of wealth). Among non-Hispanic black, Hispanic and non-Hispanic white populations, we tested whether wealth was associated with self-reported poor/fair health status after accounting for income and education.

Results Except among the younger Hispanic population, net worth was significantly associated with poor/fair health status within each racial/ethnic group in both data sets. Adding net worth attenuated the association between education and poor/fair health (in all racial/ethnic groups) and between income and poor/fair health (except among older Hispanics).

Conclusions The results add to the literature indicating the importance of including measures of wealth in health research for what they may reveal about disparities not only between but also within different racial/ethnic groups.

• Health Status
• Inequalities
• Self-Rated Health
• Social Class
• Social Inequalities

Introduction

Wealth, defined as a household's accumulated financial assets minus debts, reflects financial status over time and has been linked with a wide variety of health outcomes in population-based samples.1–5 Wealth can buffer the health-damaging effects of temporary income loss and allow individuals to afford greater healthcare resources.6 Additionally, wealth, as with other measures of socioeconomic status (SES), is hypothesised to affect health through mechanisms that are not strictly monetary in nature; wealth may be considered as a marker of power and prestige, may be associated with norms, attitudes and behaviours, and is related to subjective social status.2 ,7 ,8 However, wealth may reflect unique facets of socioeconomic status and may be an especially important marker of socioeconomic status overall2 ,9 and among individuals in which income, education and wealth are discongruent with one another, for example, among retirees and the unemployed.10–12

Furthermore, wealth differentials between racial/ethnic groups are large: in 2009, the average wealth for non-Hispanic white households in the USA was $113 149 compared to$6325 for Hispanic households and $5677 for non-Hispanic black households.13 Differences in wealth have been documented between racial/ethnic groups which have similar levels of income2 ,14 and wealth disparities have increased over recent years.13 Although household wealth has decreased in the population overall due to the economic recession and the corresponding loss of housing value (which tends to be a family's largest asset), it has decreased much more among people of colour who have experienced higher rates of unemployment, greater loss of home value and higher rates of foreclosure.13 ,15 Prior research has demonstrated that wealth is associated with diverse indicators of health and partially accounts for racial/ethnic disparities in health and healthcare after ‘controlling’ for more widely used socioeconomic indicators such as income, educational attainment and occupation.1 ,2 ,12 ,16–20 Far less information has been available on socioeconomic gradients (and on gradients in wealth in particular) in health within racial/ethnic groups, because wealth is generally more difficult to measure than income and education. Two large population-based US studies have revealed, within racial/ethnic groups and across diverse health indicators, that health incrementally improved with rising income and educational attainment.21 ,22 This graded association between health and income/education has been observed repeatedly among non-Hispanic blacks and non-Hispanic whites (hereafter, ‘blacks’ and ‘whites’) and less consistently among Hispanics.21 A handful of studies have examined health differences within racial/ethnic groups by wealth;1 ,20 ,23–27 they have studied only black and white23–25 or white and non-white20 ,26 ,27 populations. These studies have generally found a weaker association between wealth and health (mortality,20 self-rated health23 ,24 ,26 and depression25) among non-white compared to white populations. This study aimed to confirm or refute the limited, existing evidence on the role of wealth in health within black and white populations, and to provide previously unavailable information on the wealth–health relationship among Hispanics. This study examined the relationship between wealth and self-rated health within the three largest racial/ethnic groups in the USA, using information from detailed wealth questions in two large nationally representative surveys. The two data sets—which focus on adults of different age ranges—allow us to further examine whether there are differences in the association between wealth and health by age range. Methods Data sources Data are used from the Survey of Consumer Finances (SCF) and the Health and Retirement Survey (HRS), both from 2004. These surveys were chosen due to their detailed wealth measures, inclusion of self-rated health and because they represent two broad age groups. Further details about the sample design and methodology are available for the SCF (http://www.federalreserve.gov/pubs/oss/oss2/method.html) and HRS.28 Head of household respondents aged 25–64 (SCF) and aged 50 and over (HRS) who were identified as black, Hispanic or white were included (N=3310 for the SCF and N=11 847 for the HRS) in the analytic samples. The SCF defines the head of the household as the man in mixed-sex households or the older individual in same-sex households. The HRS analytic sample did not include persons residing in nursing homes or those who were not the financial respondent (the person designated to answer household-level financial questions). Variables We used net worth as our measure of wealth. Net worth—a frequently used wealth measure in health research1—is defined as the sum of all assets minus the sum of debts. It may take on a negative value if the amount of debt exceeds assets. Questions were asked about the following assets in the SCF: checking accounts, savings accounts, money market accounts, call accounts, certificates of deposit, savings bonds, mutual funds, stocks, bonds, retirement funds, life insurance, other managed accounts, other financial assets, vehicles, primary residence, other residential real estate, non-residential real estate, businesses and other non-financial assets. Debts included mortgage on primary residence, other residential property debt, other lines of credit, credit card balances, other debt and other instalment loans. SCF researchers imputed all missing data using a multiple imputation procedure yielding five values for each missing value to approximate the distribution of the missing data. Similarly, in the HRS, assets included checking/savings/money market accounts, certificates of deposit/saving bonds, mutual funds/stocks, bonds, retirement funds, vehicles, primary residence, other residential real estate, non-residential real estate, businesses and other non-financial assets/other savings. Debts consisted of mortgage on primary and secondary residence, other residential property debt and other debt/credit card balance. Bracketing was employed in HRS. In this approach, subjects who were unable or unwilling to provide the exact monetary value for a particular asset/debt were asked whether the value was more or less than an unfolding range of values, and responses to the brackets were used to impute missing wealth data.29 The dependent variable is self-rated health status, measured on a four-point (SCF) or five-point (HRS) Likert scale and dichotomised as fair or poor versus better health status. Age, age-squared (to allow for potential non-linear effects), gender, marital status (married or partnered, previously married or never married) and family size (number of adults and children in a financially interdependent household) were included as covariates in the analyses. Census region (Northeast, Midwest, South or West) was also included in the HRS analyses but was unavailable in the public-use SCF data set. Census regions were included to help adjust for regional variation in income and wealth. Educational attainment was classified into four categories: less than high school, high school graduate or passed the General Educational Development (GED) examinations (high school equivalent), some college or college graduate and above. Annual pretax household income from all sources was determined and was log transformed. For both data sets, missing data were imputed.30 ,31 In SCF, five data sets with imputed values were employed, and the repeated-imputation inference technique was used to estimate SEs, which incorporates the variability in the data due to missing data.32 Analysis All analyses were stratified by racial/ethnic group. Wealth measures were categorised into four roughly equal groups with cut-points depending on the distribution within each racial/ethnic group (table 1). Racial/ethnic group-specific cut-points were chosen in order to examine the association of wealth and health within racial ethnic groups. A small number of black and Hispanic adults would have been included in the highest wealth quartile if population-based cut-points had been employed, which would have resulted in unstable estimates. For example, in the HRS, only 5.5% of the sample of black women and men (N=104) and 8.5% of the sample of Hispanic women and men (N=92) would be categorised into the highest wealth quartile based on the entire sample. Table 1 Cut-points for net worth quartiles by race/ethnicity, Survey of Consumer Finances (SCF) and Health and Retirement Survey (HRS), 2004 Our first model was an unadjusted logistic regression model between net worth and poor self-rated health. The second model added age, age-squared, gender, marital status, household size, region (HRS only), educational attainment and income as covariates. Additionally, we ran models that excluded net worth in order to determine whether its inclusion changed the association between education/income and health. In sensitivity analyses, in order to make our results more comparable across racial/ethnic groups, we reclassified wealth among white respondents using the same cut-points used among black participants. The institutional review board at the University of Texas at Austin designated this study as exempt. Results Table 2 shows the demographic and socioeconomic characteristics of participants in the two data sets, stratified by racial/ethnic group. The higher percentage of men among the SCF Hispanic and white respondents is explained by the convention of selecting men as the household head in coupled households in that data source. Higher proportions of whites were currently married and lower proportions were never married compared to black or Hispanic adults. Table 2 Demographic and sociodemographic characteristics by racial/ethnic group, Survey of Consumer Finances (SCF) and Health and Retirement Survey (HRS), 2004 In both data sets, there were striking racial/ethnic disparities in levels of education, income and net worth. For example, in the SCF, median net worth was six and a half times greater among whites compared to blacks, and over eight and a half times greater compared to Hispanics; and whites had higher levels of median assets as well as debts. Similarly, in the HRS, whites had the highest levels of net worth and assets, though total median debt among the three groups was similar. The differences in net worth between racial/ethnic groups appear largely due to lower rates of asset ownership among blacks and Hispanics along with lower median values among those who had these assets (table 3). For example, fewer blacks and Hispanics owned homes in the SCF (49% and 47%, respectively) compared to 77% of whites. Among homeowners, the home median value was$113 000 among blacks, $148 000 among Hispanics and$175 000 among whites. Blacks and Hispanics reported higher rates of fair or poor health compared to whites.

Table 3

Prevalence and median values (\$)* of assets and debts, Survey of Consumer Finances (SCF) and Health and Retirement Survey (HRS), 2004

Table 4 shows the unadjusted analysis and full regression models for self-rated health for each racial/ethnic group. With the exception of the Hispanic population in the SCF, the unadjusted results show a strong and consistent association between lower net worth and poor self-rated health in both samples and across racial/ethnic groups. The adjusted model controls for age, age-squared, gender, marital status, family size, region (HRS), educational attainment and income. In these models, white respondents with the lowest wealth had over five times the odds of fair/poor health compared to those in the highest wealth category. In comparison, there was a lower OR for black respondents in the lowest wealth category compared to the highest (OR 3.75), though with wider CIs (95% CI 1.50 to 9.36). For white and black respondents, there was an apparent stepwise gradient (with incremental improvement in health as wealth increased) in the association between net worth and self-rated health; however, there was no clear pattern in the association between wealth and health among Hispanics in the SCF sample. As shown in online supplementary appendix table 1, the association between education and poor health was attenuated in models that included net worth, and the association between income and poor health was no longer significant among blacks after including net worth.

Table 4

Unadjusted (model 1) and adjusted (model 2) ORs and 95% CIs by race/ethnicity group for fair/poor health status, Survey of Consumer Finances (SCF) and Health and Retirement Survey (HRS), 2004

Similarly, in the HRS, an apparent stepwise gradient between net worth and fair/poor health was generally observed among white, black and Hispanic adults, with the exception of a higher OR for black individuals in quartile 3 than those in quartile 2. Again, the relationships between both income and education in relation to self-rated health in the HRS tended to be significantly attenuated after adding net worth. Among Hispanics, however, income was only significant in the model that included net worth (see online supplementary appendix table 1).

In sensitivity analyses in which we re-classified wealth among white respondents using the same cut-points used among black participants (see online supplementary appendix table 2), we continued to observe an association between net worth and self-rated health in both data sets. Though the point estimates of the OR among whites were of a greater magnitude than among black and Hispanic respondents, the CIs were overlapping.

Discussion

Within younger and older black and white populations and among older Hispanics, being in the lowest quartile of wealth was associated with 3–5 times higher odds of reporting fair or poor health status compared with those in the highest quartiles of wealth, after controlling for income and education. In contrast, we did not find a significant association between wealth and self-rated health among younger Hispanics. The lack of association among young Hispanics was an unexpected finding. Overall, the results reinforce the importance of examining wealth as a potential contributor to health within racial/ethnic groups.

Across different racial/ethnic groups and populations of varying ages, the total amount of wealth as well as its distribution among different specific assets varied widely.13 ,14 In contrast, the amount of debts appeared to be more consistent across racial/ethnic groups, a finding that has been observed in studies using HRS and other data sets.13 ,16 ,33 Yet significant associations with self-rated health were observed within younger and older black and white populations and older Hispanic adults. The results suggest that the distributions of wealth relative to other members of the same racial/ethnic group may be an important factor in health disparities. Within groups, wealth may affect health through a wide variety of mechanisms including differences in access to goods and services, beliefs about health promotion and behaviours such as diet and exercise.21 Wealth may also function through different mechanisms according to age;10 for example, younger adults may be more likely to be in the process of wealth accumulation, whereas older adults may be more likely to spend down their savings. While our results suggest the potential importance of wealth to health disparities across broad age ranges, we are unable to assess whether potential mechanisms differ across these age groups.

In contrast to the findings among black and white and older Hispanic populations, we did not find a significant association between wealth and health among younger Hispanics. There are multiple, potentially overlapping explanations for this finding. First, there are different distributions of wealth between racial/ethnic groups. However, the relatively restricted wealth distributions and cut-points chosen for creating quartiles were similar between younger blacks and Hispanic populations. Second, among the younger age groups, Hispanics appeared more likely to own vehicles and less likely to have a checking and savings account and retirement funds compared to the other groups. Though it is plausible that specific assets/debts may be differently associated with health, the potential monetary and non-monetary mechanisms through which each of these would function is uncertain. Third, it may be that the heterogeneity among the Hispanic population (ie, combining persons of different origins and acculturation levels) may have masked clear patterns for some subgroups. Fourth, there may also be protective factors among Hispanics that blunt the effects of SES differences, including wealth differences; Hispanics generally exhibit better levels of health than their disadvantaged socioeconomic status would predict, often referred to as the ‘Hispanic paradox’.34 Finally, there may be differential reporting of self-rated health among Hispanics compared to the other racial/ethnic groups.35 ,36

The associations between more widely used measures of SES—income and education—and self-rated health were generally attenuated after adding wealth; yet each appeared to have an ‘independent’ effect on health, at least for some groups. Interventions may seek to improve different facets of socioeconomic status. For example, ‘bank-on’ programmes providing free banking and checking accounts for low-income individuals are an attempt to build wealth through savings.37 The impact of these programmes both on wealth generation and on health should be evaluated. Similarly, programmes that encourage home ownership are another intervention designed to increase wealth; however, these programmes have come under increasing and necessary scrutiny with the recent foreclosure crisis.38 In contrast, other programmes and policies attempt to influence income or education, such as the various taxation policies, minimum wage programmes geared toward increasing high school completion rates, and student loans and grants for college education. Documenting the effects of such policies and programmes on subsequent socioeconomic indicators, and testing whether those effects have an influence on health, is an important public health priority requiring longitudinal designs.

This study has several limitations. First, this was a cross-sectional study from which no causal inferences can be drawn. For some individuals, poor health may have led to lower levels of wealth and not the other way around. Second, we examined only self-rated health; other health measures may have different relationships with wealth. Self-rated health has been associated with all-cause mortality and a wide range of health outcomes.39 Third, as noted above, reports of self-rated health may vary across racial/ethnic groups and by income/education.35 ,36 ,40 Differential reporting across racial/ethnic groups should not affect within-group comparisons; however, the lower reliability of self-rated health among those with lower education levels may bias within-group findings. Fourth, using racial/ethnic group-specific cut-points for our wealth measure allowed us to observe a wealth gradient within groups and afforded greater power to see differences within each population. Because the quartiles represent different amounts and distributions of wealth within each racial/ethnic group, the approach prevents us from examining the strength of the association and relative magnitudes across groups. Similarly, because different cut-points were chosen in the two data sets and the Census region was included in HRS, caution should be used when comparing the ORs of net worth between younger and older participants of the same racial/ethnic group. Fifth, we do not have information on other indicators of socioeconomic status, such as neighbourhood socioeconomic characteristics, occupation, educational quality or childhood socioeconomic status, which may be important contributors to health and health disparities.2 ,41 Sixth, wealth differentials between racial/ethnic groups have increased over time, highlighting the potential importance of wealth as a contributor to health disparities.13 Our results using 2004 data present an important baseline from which to continue assessing within-group wealth and health disparities.

This study also has a number of significant strengths. We were able to examine wealth disparities within the three largest racial/ethnic groups in the USA. In addition, our findings are based on samples of two different age groups from two nationally representative data sets, both including detailed wealth measures reflecting a gold standard of wealth measurement.

Conclusion

Previous studies have found dramatically different wealth distributions across racial/ethnic groups, and accounting for wealth has reduced observed racial/ethnic differences in health outcomes. We extend this work by showing that wealth differentials within racial/ethnic groups: (1) are associated with self-rated health after accounting for income and education; (2) tend to attenuate the association between self-rated health and more widely used indicators of socioeconomic status and (3) are generally consistent across younger and older black and white populations and among older Hispanics population. The results confirm the importance of measuring wealth in health studies that examine the role of socioeconomic status within racial/ethnic groups.

What is already known on this subject

• Prior research suggests that accounting for wealth—a household's accumulated financial assets minus debts—helps attenuate racial/ethnic disparities in diverse health outcomes. However, there has been limited research into whether gradients in wealth within racial/ethnic groups are associated with health outcomes.

• We found that wealth differentials within racial/ethnic groups were significantly associated with differences in self-rated health among younger non-Hispanic black and white populations and among older Hispanic populations. We did not observe a significant association among younger Hispanic adults. These findings suggest the importance of measuring wealth when examining health disparities within racial/ethnic groups.

Acknowledgments

We wish to thank Tara Powell, Rose Pulliam and Erika Delanoy for their assistance with the preparation of the article.

• Supplementary Data

This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Files in this Data Supplement:

Footnotes

• Contributors CC and CEP designed the study and directed its implementation, including quality assurance and control, and drafted the paper. AS conducted statistical analyses and provided revision of the manuscript. PAB, DV, BF and MH each assisted with the design of the study and provided a critical revision of the manuscript.

• Funding This work was supported by the American Legacy Foundation (UTA08-E667). Dr Pollack's salary is supported by a career development award from the National Cancer Institute and Office of Behavioral and Social Sciences Research (1K07CA151910-01A1). The authors also wish to acknowledge institutional support from the Population Research Center and the Center for Social Work Research at the University of Texas at Austin.

• Competing interests None.

• Ethics approval University of Texas, Austin.

• Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.