INTRODUCTION

Racial/ethnic and socioeconomic differences
in health have persisted in the United States.
Recent mortality data provide a glimpse of
health status in the United States. The ageadjusted
death rate in 2018 was highest among African
Americans (892.6 per 100,000 population), followed by
American Indian and Alaska Natives (AIAN) (790.8 per
100,000 population), Whites (748.7 deaths per 100,000
population), Native Hawaiian and Pacific Islanders
(NHOPI) (675.7 per 100,000 population), Latinos
(524.1 per 100,000 population), and Asians (381.2 per
100,000 population). The age-adjusted mortality rate
was higher for males than for females in each racial/
ethnic group. The highest age-adjusted mortality
rate was highest among African American males
(1,102.8 per 100,000 population) (Murphy et al., 2021).
In 2019, life expectancy at birth was 78.8 years in the
United States. The Asian population had the highest
life expectancy at birth (85.6 years), with an advantage
of 3.7 years over the Latino population (81.9 years), 6.8
years over the White population (78.8 years), 10.8 years
over the African American population (74.8 years), and
13.8 years over the American Indian and Alaska Native
population (AIAN) (71.8 years) (Aries E and Xu JQ, 2022). Life expectancy at
birth varies significantly by gender and race/ethnicity.
In 2019, Asian females had the highest life expectancy at
birth (87.4 years), followed by Latinas (84.4 years), Asian
males (83.5 years), White females (81.3 years), Latino
males (79.1 years), Black/African American females
(78.1 years), White males (76.3 years), AIAN females
(75.0 years), African American males (71.3 years, and
AIAN males (68.6 years) (Aries E and Xu JQ, 2022).
There is significant variation across racial/ethnic
groups in the top 10 leading causes of death in 2018.
Heart disease was the first leading cause of death for
White, African American, AIAN, and Native Hawaiian
and Other Pacific Islander (NHOPI) populations,
but it was the second leading cause for Asian and
Latino populations. Cancer was the first leading
cause of death for Asian and Latino populations,
but it was the second leading cause for White,
African American, AIAN, and NHOPI populations.
Unintentional injuries, stroke, diabetes, and kidney
disease also ranked among the top 10 causes of
death for each racial/ethnic group (Heron, 2021).
What has been consistent for much research regarding
the relationship between race/ethnicity and health
is that African Americans have higher mortality and
poorer health status than does any other groups, as do
Native Americans. Mortality
rates are also higher for
Whites than for Latinos or
Asians, although relative
mortality varies for specific causes of death (Adler and Rehkopf, 2008; Singh and Hiatt,
2006; Williams et al., 2010). A large body of research and reviews also
show that socioeconomic status (SES) remains a fundamental cause
of health disparities. Individuals at higher SES tend to do better on
most measures of health than their lower SES counterparts (Lynch
and Kaplan, 2000; Braverman et. al., 2010; Williams and Collins,
2016). Although SES often account for a large part of racial/ethnic
differences in health, racial/ethnic disparities in health persist
(Kayitsinga and Martinez, 2008; Adler and Rehkopf, 2008; Williams,
1999; Williams, Priest, and Anderson, 2016).
This study is important for policymakers and scholars to
understand and highlight the continued and persistent racial/
ethnic and SES disparities in health in the United States. The
main objective of this study is to determine the main, relative,
and combined influences of race/ethnicity and SES on the health
status of adults in the United States. This study addresses three
main questions: (1) What is the influence of race/ethnicity on
health? (2) What is the influence of SES on health? and (3) To what
extent does SES explain racial/ethnic disparities in health?
BACKGROUND
RACE/ETHNICITY AND HEALTH
A large body of research highlights racial/ethnic differences
in health (Williams and Collins, 1995; Williams and Sternthal,
2010). Williams and Sternthal (2010) showed that the racial gap in
health is large and persistent overtime. Race/ethnicity is a socially
constructed category that has a tremendous effect on health.
Factors such as racism, segregation, discrimination, and lack of
better economic opportunities, create social and spatial contexts
that may expose individuals to poor health conditions and death.
There is mounting evidence that racism adversely affects health
through multiple mechanisms (Williams and Mohammed, 2013;
Williams and Mohammed, 2009). Racism, in both its institutional
and individual forms, remains an important determinant
of racial/ethnic disparities in health (Williams, 2012).
Williams and Collins (2016) argue that racial residential
segregation is a fundamental cause of racial disparities in
health. Using the example of African American segregation,
they argue that the physical separation of the races by enforced
residence in certain areas is an institutional mechanism
of racism that was designed to protect whites from social
integration with African Americans. Despite its legal nature,
residential segregation remains extremely high for most
African Americans in the United States. They also show that
segregation is the primary cause of racial differences in SES by
determining access to education and employment opportunities.
For Latinos, especially those of Mexican background, it remains
a paradox why they exhibit better health than Whites despite
having lower levels of SES and relatively lower levels of access to
health insurance in the United States (Markides and Eschbach,
2005). What has been evident is that newly arrived immigrants
exhibit better health than similar natives do, and immigrants’
health advantage deteriorates with increasing duration in the
U.S. and greater levels of acculturation (Cho et al., 2004; Hummer
et al., 2007). The rationale is that the longer immigrants stay
in the U.S. the greater the likelihood of losing their traditional
lifestyle, which buffers against unhealthful behaviors. There
is some evidence that second generation Latinos have poorer
health than Latino immigrants despite having higher levels
of SES than their first-generation peers (Collins et al., 2001).
SOCIOECONOMIC STATUS (SES) AND HEALTH
Another social construct that captures differential exposure to
conditions of life that have health consequences is SES. A large
body of research evidence shows that SES remains a fundamental
cause of health disparities (Williams and Collins, 2016). Individuals
at higher SES do better on most measures of health than their
lower SES counterparts (Lynch and Kaplan, 2000; House, 2000;
Braverman et. al., 2010). Braverman and colleagues (2010) show
that individuals with the lowest income and who were least
educated were consistently unhealthy, but for most indicators,
even groups with intermediate income and education levels were
less healthy than the wealthiest and most educated. They showed
that gradient patterns were seen often among African Americans
and Whites, but less consistent among Latinos.
The pathways through which SES affects individuals’ health include
exposure to both health-damaging conditions and health-protecting
resources (Adler and Rehkopf, 2008). Some exposures have
direct effects on health, while others influence psychosocial and
behavioral factors such as cognition and emotion (e.g., depression,
hopelessness, hostility, and lack of control) and behaviors (e.g., use
of cigarettes, alcohol, and other substances) (Adler and Rehkopf,
2008). Health-damaging exposures include early life conditions,
inadequate nutrition, poor housing, exposure to lead and other
toxins, inadequate health care, unsafe working conditions,
uncontrollable stressors, social exclusion, and discrimination (Adler
and Rehkopf, 2008; Williams and Collins, 1995).
Living in disadvantaged neighborhoods also expose individuals to
greater uncertainty, conflict, and threats for which there are often
inadequate resources to respond effectively. These experiences
accumulate to create chronic stress among individuals subjected
to prolonged exposure to such conditions (Adler and Rehkopf,
2008). Poor and low-income individuals are disadvantaged with
respect to lifestyles, as they are more likely to engage in unhealthy
behaviors such as smoking, unhealthy eating and drinking
practices, and lower levels of physical activity across adulthood
(Cockerham, 2005). In contrast, the upper and middle classes tend
to adopt healthier lifestyles by engaging in leisure-time sports and
exercise, healthier diets, moderate drinking, less smoking, more
physical checkups by their physicians, and greater opportunities
for rest, relaxation, and coping with stress (Cockerham, 2005;
Robert and House, 2000; Snead and Cockerham, 2002).

INTERSECTION OF RACE/ETHNICITY AND SES AND HEALTH
Race/ethnicity and SES are interlinked and both influence
conditions of life that have health consequences. SES
accounts for a large part of racial/ethnic differences in health. Nonetheless, racial/ethnic disparities in health persist (Kayitsinga
and Martinez, 2008; Adler and Rehkopf, 2008; Williams,
1999; Williams, Priest, and Anderson, 2016). Williams and
colleagues (2010) reviewed studies that show that differences
in SES across racial groups are a major contributor to racial
disparities in health. However, they add that race reflects
multiple dimensions of social inequality and individual and
household indicators of SES capture relevant but limited
aspects of this phenomenon. Therefore, to understand the
widening gaps in health status, one must look at the separate
and combined effects of race/ethnicity and SES on health.
HYPOTHESIS
Because SES remains a fundamental cause of health disparities
(Williams and Collins, 2016), this study hypothesizes that
individuals with higher SES will likely report better health
than their lower SES counterparts. This study further
hypothesizes that SES will account for a large part of racial/
ethnic differences in health, but racial/ethnic differences in
health will remain. More specifically, this study hypothesizes
that gaps in self-rated health are more likely to narrow among
African Americans than Whites and among Asians than Whites
but are likely to increase more among Mexicans and other
Latinos than Whites once SES is taken into consideration. The
intent of this paper is to estimate racial/ethnic differences in
health, SES differences in health, and how much SES might
contribute to racial/ethnic disparities in self-rated health.
DATA AND METHODS
DATA
Data are from the National Health Interview Survey (NHIS) in 2019
to 2021. The three years were merged together to provide enough
sample size to assess racial/ethnic and socioeconomic status
(SES) differences in health. The 2016 – 2025 NHIS sample design
is a multi-stage probability sample of U.S. households with new
households interviewed each year. The survey conducts household
interviews throughout the United States (U.S.) and collects
information on health status, health-related behaviors, and
information on sociodemographic and economic characteristics,
including race/ethnicity, gender, SES, and other household
characteristics from the U.S. civilian non-institutionalized
population. The NHIS interview begins by identifying everyone
who usually lives or stays in the household. One adult aged 18 years
and older and one child aged 17 years and younger are randomly
selected for an interview. Information about the sample child is
collected from a parent or adult who is knowledgeable about and
responsible for the health care of the sample child.
Due to the COVID-19 pandemic, NHIS data collection in 2020
switched to a telephone-only mode beginning March 2020.
Personal visits resumed in September 2020. In addition, from
August through December 2020, a subsample of adult respondents
who completed the NHIS in 2019 were re-interviewed by
telephone and asked to participate again in the survey. The 2020
sample adult file is hence composed of both the re-interview
cases and the 2020-sampled cases (n = 31,568). Adding the
2019 NHIS sample of 31,997 adult respondents, and the 2021
sample adults (n = 29,482), the 2019–2021 combined sample is
composed of 93,047 adults age 18 years and older. This study
uses data on 91,713 Latino, African American, Asian, and
White respondents, excluding other races (n = 2,334 (2.5%).
MEASURES
Health. The dependent variable is self-rated health status.
Self-rated health is measured with the question that captures the
subjective measure of general health status: “Would you say your
health in general is excellent, very good, good, fair, or poor?”
Responses to the item were reverse coded so that higher values
indicate better health: 1 = poor, 2 = fair, 3 = good, 4 = very good,
and 5 = excellent. The reliability and validity of self-rated health is
well established (Idler and Benyamini, 1997).
Race/ethnicity. Race/ethnicity is constructed from selfreported
ethnicity and race categories. First, Latino adults
are distinguished from non-Latino adults. Among Latinos,
Mexicans are distinguished from Other Latinos. For non-
Latinos, race is categorized as White, Black, Asian, or Other
race categories. Other races include Native Americans and
Alaska natives, and other single or multiple races. For this
study, Other race groups are excluded in the analyses.
Socioeconomic status (SES). SES is measured by two variables:
educational attainment and family income-to-poverty ratio (IPR).
Educational attainment was measured in the number of years
completed and was coded into four dummy variables: less than
high school, high school diploma or equivalent, some college, and
bachelor’s degree or higher (reference category). IPR is a categorical
variable based on family income and poverty thresholds. IPR was
coded into six dummy variables indicating percentages of family
income to poverty ratio: less than 100, 100 – 149, 150 – 199, 200 – 299,
300 – 399, and 400 percent or more (reference group).
The following sociodemographic variables are controlled
in all analyses: age (in years), gender (1 = female, 0 = male),
immigrant (1 = foreign born, 0 = U.S. born), marital status
(married (reference group), cohabiting, widowed, divorced/
separated, and never married dummy variables (1 = yes, 0
= no)), employment status (1= employed, 0 = not employed),
housing ownership (1 = yes, 0 = no), length of residence in house/
apartment (less than one year to 3 years, 4 -10 years, and 10 years
or more (reference group) dummy variables (1 = yes, 0 = no)),
and residential location (1 = nonmetropolitan, 0 = metropolitan
(reference group)). The following health-related variables are also
controlled: weight status categories based on body mass index
(BMI): underweight (BMI < 18.5 kg/m2), healthy weight (18.5 kg/
m2 ≤ BMI < 25 kg/m2) (reference group), overweight (25 kg/m2 ≤
BMI < 30 kg/m2), and obese (BMI ≥ 30 kg/m2) dummy variables
(1 = yes, 0 = no), and smoking (three dummy variables indicating
current smoker, former smoker, and never smoker (reference
group). Descriptive statistics (mean and standard errors) for
variables used by race/ethnicity are presented in Table 1.

ANALYTICAL PLAN
The analysis proceeds in four steps. First, descriptive statistics
of health differences by race/ethnicity are presented. Second,
descriptive statistics of health differences by different dimensions
of SES are presented. Third, regression models of health
status are estimated to highlight the relative and combined
influences of race/ethnicity, SES, and control variables. Finally,
interaction terms between racial and ethnicity and SES are
added to the final model to better estimate racial/ethnic
differences in health that may be attributed to racial/ethnic
differences in SES. To account for the NHIS sampling design, all
analyses are weighted using the final annual sampling weights
(WTFA_A). Both weights were divided by 3 to produce correct
population estimates in the pooled analysis. Point estimates
and estimates of their variances were calculated using STATA
15.1 software to account for the complex sampling design of
NHIS, considering stratum and sampling unit identifiers. The
Taylor series linearization was chosen for variance estimation.
RESULTS
DESCRIPTIVE ANALYSIS
Figure 1 displays mean self-rated health by race/ethnicity. The
results show that Mexicans and African Americans have on
average lower health than Whites whereas Asians exhibit higher
average health than Whites. There was no significant difference
between the average health of Other Latinos and that of Whites
(Figure 1). Figure 2 displays self-rated health by educational
attainment. As expected, education is positively associated with
self-rated health. The mean health among college-educated
adults is significantly higher than the mean health among adults
with less than a high school, high school or equivalent, and some
college education, respectively (figure 2). Figure 3 displays selfrated
health by family income-to-poverty ratio (IPR). As expected
also, income-to-poverty ratio is positively associated with selfrated
health. Adults in poverty (< 100% IPR) have significantly
lower mean health than those at the end of the income
spectrum (≥400% IPR) (figure 3). Table 1 presents summary
statistics for all variables in the analysis by race/ethnicity.
MULTIVARIATE ANALYSIS
Table 2 shows odds ratios for self-rated health by race/ethnicity
from ordered logistic regression models. The results in model 1
(table 2) show that there are significant racial/ethnic differences
in self-rated health. The odds of reporting excellent health
decreases 19 percent (i.e., [(1 - .810) x 100] more for Mexicans
and 27 percent more for African Americans than for Whites,
respectively. In contrast, the odds of reporting excellent health
increases 26 percent (i.e., [(1.26 – 1) x 100] more for Asians than
for Whites. The main effects of race/ethnicity in subsequent
models in table 2 reflect the baseline model of the self-rated health
differences between Latinos, African Americans, and Asians
and Whites. Other covariates in table 2 can be interpreted in the
same way that conventional ordered logistic regression models
are interpreted. Exponentiation of the values for intercept 1 – 4 represents the odds of reporting different levels of health status
for the reference group.
Next, model 2 adds educational attainment as one of the
covariates to determine how education contributes to differences
in self-rated health by race/ethnicity. A comparison of models 1
and 2 shows that controlling for respondent’s education results
in substantial change in self-rated health by race ethnicity.
These results suggest that racial/ethnic differences in education
partially explain differences in self-rated health between Latino,
African American, and Asian racial/ethnic groups and Whites.
Specifically, the odds ratios of excellent health for Mexicans
increases more by 20 percent, 17 percent for Other Latinos, and
16 percent more for Asians than for Whites, respectively, net
of the effects of educational attainment on self-rated health. In
contrast, the odds of excellent health for African Americans
decreases 17 percent more than Whites, net of the effects of
educational attainment on self-rated health. The results in model
2 also show that the odds ratios of excellent health for adults with
less than a high school education decreases 78 percent more, 56
percent more for adults with a high school diploma or equivalent,
and 40 percent more for adults with some college education
than for adults with a college degree or higher, respectively.
Model 3 (table 2) adds another SES measure: the income-topoverty
ratio to see if it too explains racial/ethnic differences
in health. The results suggest that adding income-to-poverty
ratio results in substantially change in self-rated health by
race/ethnicity. The odds ratios of excellent health increases 37
percent more for Mexicans, 34 percent more for Other Latinos,
and 24 percent more for Asians than Whites, respectively, net
of the effects of both educational attainment and income-topoverty
ratio on health. The odds of excellent health for African
Americans further decreases and become not significantly
different from those of Whites once the effects of education and
income-to-poverty ratio are taken into consideration in model
3 of Table 2. As expected, the results in model 3 show that the
odds ratios of excellent health incrementally decrease as incometo-
poverty ratio increases, i.e., as one moves from poverty and
lower class to middle and then upper classes. More specifically,
the odds of excellent health decreases 60 percent more for adults
in poverty (IPR < 100%), 54 percent more for adults with IPR
between 100 – 149 percent (i.e., those near poverty), 44 percent
for those with IPR between 150 and 199 percent, 37 percent
more for IPR between 200 – 299 percent, and 25 percent more
for adults with 300 – 399 percent than adults with IPR greater
or equal to 400 percent, respectively, net of the effects of race/
ethnicity and educational attainment. Educational attainment
also remains negatively associated with self-rated health.
Estimated effects of all the other covariates, displayed in model
4 (Table 2), are in the expected direction. Specifically, the odds
of reporting excellent health decline with age, and they are
smaller for cohabiting, divorced/separated, and never married
compared to married adults. They are also lower for adults who
had resided in their homes/apartments for less than 10 years
compared to those who resided in their homes more than ten years. They are also lower for adults who were underweight,
overweight, and obese compared to those with a healthy weight
status. They are also lower for current smokers and former
smokers compared to never smokers, and those residing in
nonmetropolitan areas compared to those in metropolitan areas.
In contrast, the odds of reporting excellent health were higher for
adults currently employed compared to those not employed and
not working, and higher for homeowners compared to renters.

RACE/ETHNICITY AND SES VARIATIONS IN SELF-RATED HEALTH
Model 5 (Table 2) adds interaction terms between race/ethnicity
and education, and between race/ethnicity and income-to-poverty
ratio. The odds of reporting excellent health remain smaller
for Mexicans, African Americans, and Asians than those of
Whites, net of the effects of other covariates in the model.
The odds of reporting excellent health remain significantly
lower for respondents with lower levels of education and lower
income-to-poverty ratios. Only significant interaction terms
(p < .05) were retained in the final model. To illustrate the results
in model 5, figure 4 displays the estimated predicted probability
of reporting excellent health by educational attainment and
race/ethnicity, whereas figure 5 displays the estimated predicted
probability of reporting excellent health by income-to-poverty
ratio and race/ethnicity. The results show that the probability of
reporting excellent health increased rapidly by education levels
among whites, followed by other Latinos, African Americans,
and Asians while it increased at a fairly rapid and almost a constant rate among Mexicans, leading to an increasing gap in
self-rated health between Mexicans and Whites. The probability
of reporting excellent health for African Americans and Asians
significantly increased as education increases. Nonetheless, gaps
in self-rated health between African Americans and Whites, and
between Asians and Whites slightly increase as education levels
increase, leading to larger gaps for those with college education or
more than those with less than high school education (Figure 5). Figure 5 displays the estimated predicted probability of reporting
excellent health by income-to-poverty ratio and race/ethnicity.
Mexicans in poverty (< 100% IPR) are less likely than Whites
to report being in excellent health. However, the predicted
probability of reporting excellent health increases at a slower rate for Mexicans relative to Whites as income-to-poverty ratio
increases, leading to larger gaps in self-rated health at the
higher end of the income spectrum (≥ 400% IPR). The predicted
probability of reporting excellent health for Other Latinos
increases significantly as income-to-poverty ratio increases,
but it increases at the same pace relative to whites, leading to no
significant differences at the higher end of the income spectrum.
The predicted probability of reporting excellent health for
African Americans also increases as income-to-poverty
increases. However, it increases almost at the same pace relative
to whites, except from poverty to near-poverty (100 – 149% IPR)
where it increases at a slightly higher rate than Whites and then
at the end of income spectrum, where it dramatically decreases,
leading to a larger gap in self-rated health relative to Whites.
The predicted probability of excellent health for Asians is higher
than that of Whites at poverty level, but it decreases thereafter
up to a 150 – 199% IPR and then increases as income-to-poverty
increases but remains lower than that of whites at the end of the
income spectrum.
DISCUSSION
The analysis in this paper based on the pooled data from NHIS
2019 – 2021, shows that there are significant differences in selfrated
health by race/ethnicity. The odds of reporting excellent
health are significantly lower for Mexicans and African
Americans than they are for Whites. In contrast, the odds of
reporting excellent health are significantly higher for Asians
than Whites. As expected, the odds of reporting excellent
health increase as SES increases both in terms of education and
family income social classes. One important conclusion is that
SES accounts for a large proportion of racial/ethnic disparities
of self-health, but in the end racial/ethnic disparities remain.
The odds of reporting excellent health significantly increase
more for Mexicans, Other Latinos, and Asians than Whites,
whereas the odds of excellent health for African Americans
decreases more than Whites once education is accounted for
in explaining self-rated health. Similarly, the odds of reporting
excellent health significantly increase more for Mexicans,
Other Latinos, and Asians than Whites, whereas the odds of
excellent health for African Americans decreases to being
non-significant as compared to those of Whites once family
income-to-poverty ratio is factored in explaining self-rated
health. The contribution of SES to racial/ethnic self-rated health
gap remains substantial and statistically significant across
subsequent models that control for other sociodemographic
characteristics such as age, gender, immigrant status,
marital status, employment status, home ownership, length
of residence, and health indicators such as weight status and
smoking, and nonmetropolitan/metropolitan residence. In
the end, the odds of reporting excellent health remain smaller
for Mexicans, African Americans, and Asians than those of
Whites, net of the effects of SES and those other covariates.
The other conclusion for this study is that accounting for SES and
its interaction with race/ethnicity provide further explanation for persistent racial/ethnic disparities in self-rated health. The
probability of reporting excellent health increases rapidly by
education levels among whites, followed by other Latinos, African
Americans, and Asians, but it increases at a fairly rapid pace
and then at almost a constant rate among Mexicans, leading to
a large gap in self-rated health between Mexicans and Whites
for those with college education or higher. Gaps in self-rated
health between African Americans and Whites and between
Asians and Whites slightly increase as education levels increase
but remain larger for those with higher education than those
with lower education. Gaps of reporting excellent health persist
across family income levels and become significantly larger
between Mexicans and Whites and between African Americans
and Whites at the higher end of the income spectrum. Gaps of
reporting excellent health between Asians and Whites vary by
family income levels with Asians reporting higher probability of
excellent health than Whites at poverty and lower income levels,
but lower than that of Whites at the end of the income spectrum.
These findings are consistent with results from studies that show
that race/ethnicity and SES are interlinked and both influence
conditions of life that have health consequences. Although
SES accounts for a large part of racial/ethnic differences in
health, racial/ethnic disparities in health persist (Kayitsinga
and Martinez, 2008; Adler and Rehkopf, 2008; Williams, 1999;
Williams, Priest, and Anderson, 2016). In summary, individuals
with higher SES are likely to report better health than their lower
SES counterparts. This study further shows that SES accounts for
a large part of racial/ethnic differences in health, but racial/ethnic
health gaps remain and become larger at higher SES levels for
Mexicans, African Americans, and Asians relative to Whites.
This study is limited in focusing on subjective health. Future
studies on racial/ethnic and SES health disparities need to
look at other health outcomes such as mortality rates, chronic
health conditions, mental health, and activity limitations. This
study is also limited in focusing only to individual and family
characteristics’ effects on health. Future research should link
NHIS data to other census data to account for neighborhood/
community residential contexts and their potential effects on
racial/ethnic health disparities. ⏹

---