Race/Ethnicity and the Relationship Between Homeownership and Health

Abstract

Objectives. We investigated whether race/ethnicity moderates the association between homeownership and health and whether this association is the same for racial/ethnic minorities as for non-Latino Whites.

Methods. With data on US-born Latinos, African Americans, and non-Latino Whites from the 2003, 2005, 2007, and 2009 California Health Interview Survey, we used weighted multivariate regression techniques in fully adjusted models, controlling for socioeconomic and demographic factors, to test the association between homeownership and number of psychological health conditions, number of general health conditions, self-perceived health status, and health trade-offs.

Results. Race/ethnicity significantly moderates the effect of homeownership on self-perceived health status, incidence of general health conditions, and health trade-offs, including delays in accessing medical care and delays in obtaining prescription medication. Although homeownership was a robust, independent predictor for each health outcome in the non-Latino White population, the association disappeared in statistical significance for racial/ethnic minorities.

Conclusions. The mechanisms that create a significant association between homeownership and health seem not to be operative for racial/ethnic minorities or are countervailed by other processes, such as possible housing insecurity, that may create an adverse association. Homeownership provides a baseline for future investigations.

Homeownership is valued as a means to develop personal wealth, increase social opportunities, prevent financial insecurity,¹ and maximize emotional and physical well-being.² The positive externalities of homeownership on health operate at various levels (e.g., individual and neighborhood) through various mechanisms. For example, homeowners are more likely than are renters to maintain their homes according to individual tastes and preferences, resulting in increased life satisfaction.^2–4 Homeowners have improved social utility as a result of increased tenure and greater attachment to their community,^5–7 higher levels of civic participation because of their stakeholder status as homeowners,^8,9 and increased emotional well-being from the accumulation of financial assets.¹⁰

Health may be adversely affected by substandard physical conditions of the home, such as poor ventilation, mold, or pest infestation, which may lead to infectious disease, injuries, and chronic conditions.^11,12 Although poor physical and psychosocial conditions can exist in both rental and owner-occupied homes,¹³ these conditions may be more difficult to resolve in rental homes because unresponsive property owners may not correct hazardous conditions and there may be fewer pecuniary resources to make repairs.^11,12 Health can also be affected by features of the natural, built, economic, and social environment, including levels of air and noise pollution, dilapidated buildings, and lack of access to primary care.^11,14,15

Although this previous research helps concretize the relationship between homeownership and health, whether this relationship exists among racial/ethnic minorities is unknown. Large gains in homeownership throughout the 1990s and early 2000s,¹⁶ especially among Latino and African Americans in California,¹⁷ provide an opportunity to examine this issue. Because of the distinct historical and social circumstances of wealth inequality and housing discrimination that racial/ethnic minorities experience in the United States, it is plausible that the relationship between homeownership and health may not be robust. In part because of the financial hangover of past discrimination, racial/ethnic minorities may be more likely to purchase homes that are financially unaffordable, in poor condition, and located in neighborhoods with fewer social resources. We examined whether race/ethnicity moderates the relationship between homeownership and health and whether there is a significant association between homeownership and health outcomes among racial/ethnic minority populations compared with non-Latino Whites.

CONCEPTUAL FRAMEWORK

The association between racial/ethnic health disparities and socioeconomic status in the United States, particularly wealth, is well documented.^18–20 Wealthier individuals may have better health because of higher social position, whereas lack of wealth may produce feelings of inadequacy and loss of power, resulting in psychosocial stress and poorer health.¹⁹ Wealth estimates in the United States are lowest among Latinos and African Americans, who have 12 cents and 10 cents, respectively, per 1 dollar of White net worth.²¹ Although a core economic strategy to build wealth in the United States is homeownership,²² formal discrimination through written law, realtor discrimination, and discriminatory lending has constrained this opportunity for racial/ethnic minorities.^23,24 These forms of housing discrimination have significantly contributed to residential segregation, which is a primary determinant of racial/ethnic inequalities and is inextricably linked to minority health disparities.^25–27

Beginning as early as the Homestead Act of 1862,²⁸ housing discrimination continued with Federal Housing Authority underwriting guidelines predicting the presence and probability of undesirable groups “invading” White neighborhoods, including African Americans, Mexican Americans, and American Indians.^29,30 The Federal Housing Authority later used residential maps to designate areas as creditworthy on the basis of race/ethnicity, effectively preventing residents in minority neighborhoods from acquiring mortgage loans.³¹ These acts and other legal statutes the judicial system enforced were instrumental in the institutionalization of redlining³² and restrictive covenants to limit housing options for racial/ethnic minorities.²⁵ Although the Civil Rights Act of 1968 illegalized housing discrimination, realtor and lending discrimination against racial/ethnic minorities persisted.

In the 1970s and 1980s, realtors could refuse services to racial/ethnic minorities, discourage racial/ethnic minorities from purchasing in predominately White neighborhoods to protect prejudiced clientele,³³ or quote higher interest rates,³⁴ resulting in fewer home purchasing options for racial/ethnic minorities.³⁵ Although realtor discrimination decreased in the late 1980s,³⁶ it continues to exist in more obfuscated ways, such as denying financing assistance³⁶ or steering racial/ethnic minorities away from homes located in healthier, safer neighborhoods,¹² including those with greater educational and recreational resources, such as schools and parks. Because racial/ethnic minorities may have a limited choice of housing units, they may settle for older, smaller, or less appealing homes, resulting in overcrowding or increased risk of home-related environmental health concerns, such as lead-based paint and asbestos, influencing long-term illness and poor psychological health.³⁷

In the 1990s, racial/ethnic minorities faced greater information asymmetry and stricter credit requirements in obtaining mortgage loans and homeowner insurance.³¹ Still, homeownership rates among racial/ethnic minorities increased because of the proliferation of subprime loans that eliminated traditional lending policies^38,39 and had higher processing and closing fees,⁴⁰ higher and adjustable interest rates, prepayment penalties, and balloon payments.²⁴ Both middle and higher income racial/ethnic minorities received a disproportionate share of subprime loans, regardless of credit scores. Among loan applicants with the highest FICO scores (720 or higher), only 2.6% of Whites received high-cost loans, compared with 13.5% and 12.8% of Latinos and African Americans, respectively.⁴¹

Previous research details how unaffordable mortgage loans can result in housing insecurity or inability to sustain homeownership because of life uncertainties such as job loss and lack of wealth to cover high mortgage debt during these times.⁴² Financial strain resulting from housing insecurity is associated with higher rates of chronic disease, psychological conditions, and overall mortality.^43,44 Stress incurred throughout the home-buying process can also decrease emotional and physical well-being.⁴⁵ Finally, self-reported discrimination as a socially derived stressor can have direct effects on high blood pressure, mental health status, and substance abuse.⁴⁶

STUDY AIMS

No studies to date have assessed the association between homeownership and health outcomes specifically by race/ethnicity.⁴⁷ We have contributed to the literature by offering insight into the health effects of public policy and private endeavors to increase homeownership among racial/ethnic minorities. Using pooled data from the California Health Interview Survey (CHIS), we hypothesized that race/ethnicity significantly moderates the relationship between homeownership and health. Furthermore, we hypothesized that the association between homeownership and better health outcomes would be statistically significant among Whites but not among racial/ethnic minorities.

METHODS

We focused on 2 of the largest racial/ethnic populations living in the United States—Latinos and African Americans—both of which experienced major growth and decline in homeownership in the past 20 years, particularly in California.^17,48 The sample includes US-born Latino (n = 15 450), US-born African American (n = 8008), and US-born White (n = 107 656) adults aged 18 years and older. The sample excludes foreign-born populations to increase comparability between the groups and to reduce the likelihood that any significant associations between homeownership and health are related to an “immigrant paradox,” the epidemiological finding that immigrants across all racial/ethnic groups tend to have health outcomes that are comparable to, or better than, US-born residents.^49–51

Data Source and Methods

Data came from 2003, 2005, 2007, and 2009 CHIS public use files, a population-based random-digit dial landline and cellular telephone survey of California’s population. CHIS is a serial, cross-sectional survey that the University of California, Los Angeles Center for Health Policy Research in collaboration with the California Department of Public Health, the Department of Health Care Services, and the Public Health Institute conducts every 2 years. CHIS uses a multistage sampling design to provide estimates for large- and medium-sized counties in the state and for groups of the smallest counties (on the basis of population size) and statewide estimates for California’s overall population, its major racial and ethnic groups, and several ethnic subgroups. CHIS is the largest state health survey in the United States and collects information on a wide variety of health topics, providing a detailed representation of health outcomes and utilization of California’s population.⁵²

The weighted, overall adult response rates were 33.5%, 26.9%, 18.7%, and 17.7% in 2003, 2005, 2007, and 2009, respectively, and are similar to response rates of other scientific telephone surveys, such as the California Behavioral Risk Factor Surveillance System, which is highly comparable in scope and size to CHIS.⁵³ Survey response rates are lower in California and have been declining, as they have nationally, since 2001.⁵³ Because of the complex design and administration of the survey, household- and person-level weights are computed using a raking method and applied to the sample data to compensate for the probability of selection and a variety of other factors, to ensure that estimates are consistent with population control totals, and to produce correct population estimates.

Measures

Three health outcomes were measured: (1) self-perceived health status—measured by 5 levels (1 = poor, 2 = fair, 3 = good, 4 = very good, 5 = excellent); (2) number of psychological health conditions (derived from the Kessler 6, a standardized and validated measure of non-specific psychological distress)—a nonnegative variable index ranging from 0 to 6 that measures the number of psychological health conditions (including nervousness, hopelessness, restlessness, depression, worthlessness, and feeling that everything is an effort) experienced in the past 30 days; and (3) number of general health conditions—a nonnegative variable index ranging from 0 to 4 that measures the number of physical health conditions (including hypertension, cardiovascular disease, diabetes, and obesity) an individual reported having. Additionally, because housing insecurity may pose health risks as a result of forced trade-offs between housing payments and covering costs associated with health-inducing goods such as medications,^43,54 we measured 2 health trade-off outcomes: whether in the past 12 months an individual (1) delayed needed medical care or (2) delayed obtaining prescription medication (0 = no, 1 = yes). We measured all outcome variables consistently across the survey years.

Primary regressors of interest.

We used 3 primary explanatory regressors to predict health outcomes. The first—homeownership—is a dichotomous variable measuring self-reported home tenure (0 = rent, 1 = own). The second—US-born racial/ethnic minority—is a dichotomous variable (0 = US-born non-Latino White, 1 = US-born racial/ethnic minority [US-born Latinos and US-born African Americans]). The third primary regressor is an interaction term between homeownership and US-born racial/ethnic minority that tests the hypothesis that race/ethnicity significantly moderates the association between homeownership and health.

Length of residence and neighborhood safety variables.

Positive neighborhood attributes are associated with the average length of residence in a neighborhood,^15,55 and perceptions of neighborhood safety are linked to psychological, physical, and self-rated health.⁵⁶ Therefore, we controlled for length at current residence, a continuous variable measured in months ranging from 1 to 1080, and perceptions of neighborhood safety, an ordinal composite variable measuring, “How often do you feel safe in your neighborhood?” (1 = none or some of the time, 2 = most of the time, 3 = all the time) and “Many people in this neighborhood are afraid to go out at night” (1 = strongly agree or agree, 2 = disagree, 3 = strongly disagree).

Socioeconomic and demographic variables.

We controlled for socioeconomic measures found to capture a socioeconomic gradient in health among racial/ethnic minorities, including US federal poverty level derived from household income and size (measured by 4 dummy variables, with the lowest federal poverty level, 0%–99%, as the referent category), the natural log of household income (income of all working adults living in the household), educational attainment (measured by 5 dummy variables, with < high school or no formal education as the referent category), and employment status (measured by 4 dummy variables, with not looking for work as the referent category). We also controlled for demographic factors that may confound the association between homeownership and the outcomes of interest, including age (continuous), gender (0 = male, 1 = female), marital status (measured by 3 dummy variables, with never married as the referent category), rural residence (0 = no, 1 = yes), and health insurance coverage (measured by 7 dummy variables, with uninsured as the referent category). We also controlled for smoking history (measured by 3 dummy variables, with never smoke as the referent category) and time (measured by 4 dummy variables per wave, with 2003 as the referent category).

Multivariate Regression Analyses

We used ordinal logistic regression in fully adjusted models to test the association between homeownership and the 5-level self-perceived health status measure. We assessed the validity of the proportional odds assumption in the ordinal logistic regression models by formally testing the similarity of odds ratios (ORs) for contrasts between each level of the dependent variable (results available upon request). Parameter estimates from these models indicate the odds of having a higher level of self-perceived health status above any threshold compared with those below that threshold.

We used zero-inflated negative binomial and zero-inflated Poisson regression models to predict incidence risk ratios (IRRs) and their 95% confidence intervals (CIs) for 2 outcomes: (1) number of psychological health conditions, and (2) number of general health conditions. Likelihood ratio tests confirmed that incidence was inflated by zero counts for both count outcomes and overdispersion in the number of psychological health conditions only. Thus, we used zero-inflated negative binomial regression to adjust for excess zeros and control for overdispersion to measure the number of psychological health conditions, and we used zero-inflated Poisson to adjust for excess zeros to measure the number of general health conditions. Finally, we used logistic regression to predict ORs and their 95% CIs for the 2 health trade-off outcomes: delay in needed medical care and delay in obtaining prescription medication. We used replicate survey weights and robust standard errors to estimate the variances correctly to produce unbiased estimates. We used weighted bivariate analysis to produce descriptive statistics. We used an α level of 0.05 for all statistical 2-tailed tests. We conducted all analyses using Stata version 11 (Stata Corp, College Station, TX).

The first set of regression models included interaction terms between race/ethnicity and homeownership. It is problematic, however, to make straightforward conclusions about the significance of the interaction effects using estimations that nonlinear models produced.⁵⁷ Using the MARGINS command in Stata,⁵⁸ we estimated the interaction effect for each dependent variable by examining the difference in the cross-partial derivative of homeownership when minority changes from 0 (White) to 1 (racial/ethnic minority). This process involved 3 critical steps. First, we performed multivariate regression models that included an interaction term between homeownership and race/ethnicity for each of the dependent health outcomes. Second, we estimated the predicted conditional probability that the health outcome = 1 (i.e., the dichotomous outcomes; otherwise a 1-unit increase for the count and ordinal logistic outcomes) for a change in homeownership as race/ethnicity changed from 0 to 1, holding each of the covariates at its mean. Because we conducted the analyses using nonlinear regression, the effects of the covariates were also nonlinear, meaning that the interaction effect would differ depending on the different values of the covariates. The third and final step was to estimate the difference in the predicted probabilities and determine whether the difference was significant at the P < .05 level.

Results from this process (data available as a supplement to the online version of this article at http://www.ajph.org) confirmed that race/ethnicity significantly moderated the effect of homeownership at the P < .05 level in 4 of the 5 health outcomes: self-perceived health status, number of general health conditions, and both health trade-offs. Race/ethnicity did not significantly moderate the association between homeownership and the number of psychological health conditions. The results from this final step provide the basis for stratifying the sample into 2 subgroups. We then performed a second set of regression models for these outcomes in stratified models for racial/ethnic minorities and Whites.

RESULTS

Table 1 presents descriptive statistics, including weighted means and SDs, for the full sample by race/ethnicity and home tenure. Whites had the highest percentage of homeowners (71%) and African Americans had the lowest (45%). Homeownership rates by race/ethnicity for the full sample were representative of homeownership rates in California.⁵⁹ Among homeowners, Whites had the highest percentage (78%) living above the 300% federal poverty level, whereas Latinos had the highest percentage (8%) living below the 99% federal poverty level.

TABLE 1—

Demographic Features of the US-Born Latino, African American, and Non-Latino White Population: California Health Interview Survey, 2003, 2005, 2007, 2009

	Latino (n = 15 450)		African American (n = 8008)		Non-Latino White (n = 107 656)
Variable (Weighted)	Rent, % or Mean (SD)	Own, % or Mean (SD)	Rent, % or Mean (SD)	Own, % or Mean (SD)	Rent, % or Mean (SD)	Own, % or Mean (SD)
Rent or own home	42.5	57.5	54.7	45.3	28.7	71.3
Perceived neighborhood safety	2.23 (0.53)	2.46 (0.49)	2.13 (0.61)	2.40 (0.74)	2.34 (0.67)	2.55 (0.64)
No. of mo at current residence	63.10 (61.50)	151.20 (105.40)	59.40 (63.70)	177.80 (136.70)	60.10 (84.30)	171.40 (168.70)
Age, y	33.00 (10.70)	40.00 (13.40)	41.60 (13.30)	49.00 (15.40)	41.40 (17.10)	51.80 (19.10)
Female	52.3	49.2	56.2	53.5	50.2	51.2
Rural	7.6	10.0	3.2	3.9	13.9	17.4
Marital status
Married	25.9	51.1	18.2	51.7	30.3	68.5
Separate	28.7	16.0	36.5	25.4	36.4	19.2
Never married (Ref)	45.4	32.9	45.3	22.9	33.3	12.2
Federal poverty level,a %
> 300	29.7	58.9	28.6	64.9	50.7	78.7
200–299	17.7	16.3	15.6	15.1	17.3	11.4
100–199	28.0	16.5	26.7	12.9	19.3	7.4
≤ 99 (Ref)	24.6	8.2	29.1	7.1	12.7	2.4
Household income, $1000	40.90 (29.70)	72.20 (39.90)	34.70 (28.40)	73.90 (49.20)	54.70 (48.40)	96.50 (75.90)
Education
> bachelor’s degree	4.2	6.8	4.4	12.9	12.0	19.0
Bachelor’s degree	8.9	13.4	11.7	18.8	22.2	24.5
> high school diploma	29.1	31.5	34.2	32.5	31.3	26.8
High school diploma	38.2	36.0	34.9	28.3	26.4	25.1
< high school or no formal education (Ref)	19.6	12.3	14.7	7.1	8.1	4.6
Health insurance type
Private (self)	3.1	4.4	2.3	2.9	7.5	7.7
Employer-based	39.5	59.3	35.6	58.7	47.9	59.9
Medicare	0.7	1.3	1.5	2.4	1.7	2.2
Medicare and Medicaid	2.6	2.1	5.7	6.1	2.5	1.6
Medicare and other	1.7	8.9	2.6	11.9	7.3	19.5
Medicaid	23.2	7.1	28.4	6.9	11.7	1.9
Other public insurance	3.7	2.5	4.2	3.6	3.9	2.0
Uninsured (Ref)	25.4	14.3	19.7	7.5	17.6	5.1
Work status
Active work with a worksite	58.7	61.0	50.1	55.6	61.0	57.4
Active work without a worksite	2.4	2.6	3.3	3.9	2.9	3.3
Looking for work	12.8	7.2	12.3	6.3	7.6	2.8
Not looking for work (Ref)	26.1	29.2	33.4	34.2	28.5	36.5
Health outcomes
Self-perceived health status	3.37 (0.81)	3.53 (0.80)	3.25 (0.93)	3.44 (0.94)	3.56 (1.09)	3.75 (1.14)
No. of psychological conditionsb	0.47 (0.72)	0.24 (0.48)	0.61 (0.91)	0.26 (0.64)	0.36 (0.94)	0.16 (0.64)
No. of general health conditions	0.30 (0.44)	0.38 (0.50)	0.50 (0.59)	0.58 (0.66)	0.35 (0.63)	0.45 (0.77)
Delayed prescription medicinec	12.9	10.8	18.1	13.7	17.3	10.6
Delayed needed medical carec	19.8	15.2	19.9	14.2	25.6	14.1

Note. Columns sum to 100% down rows in variable blocks. Percentages may not sum to 100 because of rounding error.

^aAs determined by the California Health Interview Survey.

^bData unavailable for 2003.

^cData unavailable for 2005.

Education levels also differed, with Latinos (7.0%) and African Americans (13.0%) having lower percentages of homeowners with graduate degrees than did White homeowners (19.0%); the percentage of Latino (13.0%) and African American (19.0%) homeowners with a bachelor’s degree was less than the percentage of White renters (22.0%) with a bachelor’s degree; and Latinos had the highest percentage (12.0%) of homeowners with less than a high school diploma or no formal education. Latinos also had the highest percentage of uninsured homeowners (12.0%), whereas Whites had the lowest (5.1%).

Self-Perceived Health Status

Table 2 presents adjusted ORs of homeownership on self-perceived health status. The results from the stratified models suggest that although the association between homeownership and self-perceived health status was positive for racial/ethnic minorities, the effect was statistically nonsignificant (OR = 1.06; 95% CI = 0.96, 1.16). For Whites the association between homeownership and self-perceived health status was both positive and statistically significant (OR = 1.10; 95% CI = 1.04, 1.16).

TABLE 2—

Ordinal Logistic Regression Estimates of Homeownership on Self-Perceived Health Status Among US-Born Racial/Ethnic Minorities and US-Born Non-Latino Whites: California Health Interview Survey, 2003, 2005, 2007, 2009

Variable	Model 1: Racial/Ethnic Minority (n = 23 458), OR (95% CI)	Model 2: Non-Latino White (n = 107 656), OR (95% CI)
Own home	1.057 (0.96, 1.16)	1.102 (1.04, 1.16)
Perceived neighborhood safety	1.291 (1.21, 1.38)	1.437 (1.38, 1.49)
Length at current residence	0.999 (0.99, 1.00)	1.000 (0.99, 1.00)

Note. CI = confidence interval; OR = odds ratio. ORs represent the association between a covariate and being in the next higher level of self-perceived health. We also controlled models for age, gender, marital status, smoking history, residence in a rural community, federal poverty level, income, level of education, health insurance coverage type, current working status, and survey year.

The association between perceived neighborhood safety and perceived health status was positive and statistically significant for both racial/ethnic minorities (OR = 1.29; 95% CI = 1.21, 1.38) and Whites (OR = 1.44; 95% CI = 1.38, 1.49). Length of residence did not have a significant effect in either racial/ethnic minorities (OR = 0.99; 95% CI = 0.99, 1.00) or Whites (OR = 1.00; 95% CI = 0.99, 1.00).

General Health Conditions

Table 3 presents the results for number of general health conditions. Among Whites, homeownership was significantly associated with a 2% lower incidence rate of general health conditions (95% CI for IRR = 0.97, 0.99), adjusting for all other factors. The association was statistically nonsignificant (IRR = 0.99; 95% CI = 0.97, 1.02) among racial/ethnic minorities.

TABLE 3—

Count Regression Estimates of Homeownership on Number of General Health Conditions Among US-Born Racial/Ethnic Minorities and US-Born Non-Latino Whites: California Health Interview Survey, 2003, 2005, 2007, 2009

Variable	Model 1: Racial/Ethnic Minority (n = 23 458), IRR (95% CI)	Model 2: Non-Latino White (n = 107 656), IRR (95% CI)
Own home	0.997 (0.97, 1.02)	0.982 (0.97, 0.99)
Perceived neighborhood safety	0.989 (0.96, 1.00)	0.963 (0.95, 0.97)
Length at current residence	0.999 (0.99, 1.00)	0.999 (0.99, 0.99)

Note. CI = confidence interval; IRR = incidence risk ratio. We also controlled models for age, gender, marital status, smoking history, residence in a rural community, federal poverty level, income, level of education, health insurance coverage type, current working status, and survey year.

Perceived neighborhood safety and length of residence were significantly associated, with a 4% (95% CI for IRR = 0.95, 0.97) and 1% (95% CI for IRR = 0.99, 0.99) lower incidence of general health conditions among Whites, respectively, adjusting for all other factors. Among racial/ethnic minorities, however, perceived neighborhood safety (IRR = 0.98; 95% CI = 0.96, 1.00) and length of residence (IRR = 0.99; 95% CI = 0.99, 1.00) were statistically nonsignificant.

Health Trade-Offs

Homeownership was a significant and protective factor against delays in needed medical care (Table 4) among Whites (OR = 0.78; 95% CI = 0.72, 0.86), adjusting for all other factors. Homeownership was not significantly associated with delays in needed medical care for racial/ethnic minorities, however (OR = 0.96; 95% CI = 0.79, 1.18). The association between perceived neighborhood safety and delays in needed medical care was significant for both racial/ethnic minorities (OR = 0.80; 95% CI = 0.72, 0.90) and Whites (OR = 0.73; 95% CI = 0.68, 0.78), adjusting for all other factors. Length of residence was also significantly associated with delays in needed medical care for racial/ethnic minorities (OR = 0.99; 95% CI = 0.99, 0.99) and Whites (OR = 0.99; 95% CI = 0.99, 0.99), although the associations were small.

TABLE 4—

Logistic Regression Estimates of Homeownership on Delays in Needed Medical Care and Delays in Obtaining Prescription Medication Among US-Born Racial/Ethnic Minorities and US-Born Non-Latino Whites: California Health Interview Survey, 2003, 2005, 2007, 2009

	Racial/Ethnic Minorities (n = 18 016)		Non-Latino Whites (n = 82 340)
Variable	Delayed Medical Care, OR (95% CI)	Delayed Prescription Medication, OR (95% CI)	Delayed Medical Care, OR (95% CI)	Delayed Prescription Medication, OR (95% CI)
Own home	0.964 (0.79, 1.18)	0.916 (0.77, 1.09)	0.783 (0.72, 0.86)	0.790 (0.73, 0.86)
Perceived neighborhood safety	0.804 (0.72, 0.90)	0.784 (0.69, 0.89)	0.730 (0.68, 0.78)	0.784 (0.73, 0.84)
Length at current residence	0.999 (0.99, 0.99)	0.999 (0.99, 1.00)	0.999 (0.99, 0.99)	0.999 (0.99, 0.99)

Note. CI = confidence interval; OR = odds ratio. We also controlled models for age, gender, marital status, smoking history, residence in a rural community, federal poverty level, income, level of education, health insurance coverage type, current working status, and survey year.

Homeownership was a significant and protective factor against delays in obtaining prescription medication (Table 4) among Whites (OR = 0.79; 95% CI = 0.73, 0.86), adjusting for all other factors. Among racial/ethnic minorities, the association was nonsignificant (OR = 0.91; 95% CI = 0.77, 1.09). Perceived neighborhood safety was a significant and protective factor for Whites (OR = 0.78; 95% CI = 0.73, 0.84) as well as for racial/ethnic minorities (OR = 0.78; 95% CI = 0.69, 0.89), adjusting for all other factors. Length of residence was nonsignificant among racial/ethnic minorities (OR = 0.99; 95% CI = 0.99, 1.00) and significant for Whites (OR = 0.99; 95% CI = 0.99, 0.99), although the association was small.

Sensitivity Analyses

Because 3 of the 4 waves of CHIS were conducted during the housing crisis, it is possible that the nonsignificant results for racial/ethnic minorities reflected unobservable health effects of home foreclosure. Indeed, studies conducted shortly after the housing crisis have found that home foreclosures have a significant effect on poorer health status among African Americans and in minority neighborhoods.^60,61 We therefore reestimated replicate models using only data from the 2001 CHIS. As in the full model, race/ethnicity significantly moderates the association between homeownership and each health outcome (we did not assess the relationship to psychological health conditions because of inconsistent measures in the 2001 CHIS). Results in stratified models suggest that among racial/ethnic minorities, homeownership was not significantly associated with self-perceived health status (OR = 1.03; 95% CI = 0.88, 1.20), number of general health conditions (OR = 1.03; 95% CI = 0.99, 1.08), delays in needed medical care (OR = 0.85; 95% CI = 0.66, 1.09), or delays in obtaining prescription medication (OR = 0.89; 95% CI = 0.71, 1.11).

We considered whether the inclusion of the rural population in the full sample could render biased results, given that the mechanisms by which homeownership functions to influence health might differ in rural versus urban settings. Reestimated models excluding the rural population showed that race/ethnicity significantly moderates the association between homeownership and each of the health outcomes, with the exception of psychological health conditions. In stratified models, among racial/ethnic minorities, homeownership was not significantly associated with self-perceived health status (OR = 1.08; 95% CI = 0.98, 1.19), number of general health conditions (OR = 0.99; 95% CI = 0.97, 1.02), delays in needed medical care (OR = 0.95; 95% CI = 0.77, 1.19), or delays in obtaining prescription medication (OR = 0.92; 95% CI = 0.77, 1.09).

DISCUSSION

These data partially confirm our first hypothesis that race/ethnicity significantly moderates the association between homeownership and self-perceived health status, number of general health conditions, delays in needed medical care, and delays in obtaining prescription medication, even after adjusting for demographic and socioeconomic factors. Race/ethnicity did not significantly moderate the association between homeownership on psychological health. We are uncertain why, although 1 possibility might be that the measure we used may not have captured psychological effects associated with homeownership, including self-esteem, social utility, and life satisfaction. In stratified analyses by race/ethnicity, the data also suggest significant associations between homeownership and health among Whites and nonsignificant relationships among racial/ethnic minorities, thus confirming our second hypothesis.

We found significant associations between neighborhood safety and 2 of the 3 health outcomes among both racial/ethnic minorities and Whites, suggesting that perceptions of neighborhood safety may be more influential on health than is homeownership. If true, racial/ethnic minorities could be better served by policies that seek to improve neighborhood safety and increase resources that promote health, such as access to parks,^15,62 versus policies solely focused on increasing homeownership rates.

Despite the housing crisis and the slow-recovering economy, homeownership continues to be framed as a surefire strategy to stable housing, greater life satisfaction, and better health.⁶³ The findings here cast doubt on this last claim for racial/ethnic minority homeowners. Policies should emphasize housing security and neighborhood quality as the primary goals over policies that solely focus on expanding homeownership.^24,64,65 Because there is a reciprocal relationship between health and housing that changes over time owing to loss of employment, income, and housing insecurity,⁶¹ longitudinal studies are needed that examine trends in health outcomes among racial/ethnic minority homeowners as well as racial/ethnic minority homeowners living in neighborhoods with high foreclosure rates.

Limitations

This study has several limitations. First, it is unclear whether these findings are generalizable to other US native- or foreign-born populations living in California or elsewhere. However, given California’s long-established US-born racial/ethnic population, our findings may represent the most optimistic scenario of what is occurring among Latino and African American homeowners across the United States. Future studies should examine the relationship between homeownership and health in other racial/ethnic groups, particularly the foreign-born.

Second, each of the health outcome measures is self-reported, which could result in recall bias and measurement error. In particular, we noted potential differential perceptions of health status by race/ethnicity, which may also increase reporting bias. This limitation would bias our results only if it somehow interacted with homeownership, which seems unlikely. Third, the measure for number of psychological health conditions is constrained because of the inconsistency of measuring mental health in the data source.

Fourth, because of limitations in the data, we were unable to assess the relationship between homeownership and health in and between various levels of home tenure, such as that in homeowners with and without mortgage debt. This would have helped us determine whether a housing gradient in health among racial/ethnic minorities existed. Fifth, because of the unavailability of wealth indicators and measures of mortgage arrears in the data set, we were unable to control for the confounding effects of housing insecurity. We were also unable to control for mortgage default and foreclosure processes. Including these data would reduce omitted variable bias.

Sixth, analysis of CHIS public use files limited our ability to control for the effects of residential segregation. Future research should use geocoded data to determine how health outcomes might differ among racial/ethnic minority homeowners living in segregated neighborhoods. Finally, because health status can constrain the ability to purchase a home, endogeneity bias between homeownership and health status may exist, which could lead to upward bias in the estimated effects of homeownership on health if those who buy homes are also those who engage in other health-seeking behaviors. However, we believe that any endogeneity bias would exist in each of the racial/ethnic populations and would not be specific to any single subpopulation. Because these data are cross-sectional, we did not attempt to infer causality but rather to elucidate important racial/ethnic differences in the relationship between owning a home and health.

Conclusions

This research, which falls at the crux of numerous public health concerns including racial/ethnic health disparities, socioeconomic inequalities, and the health implications of housing discrimination, offers a first glimpse into whether homeownership confers health benefits to racial/ethnic minorities similarly to non-Latino Whites. The mechanisms that create significant associations between homeownership and the health outcomes we measured for Whites seem not to be operative for racial/ethnic minorities, or they are countervailed by other processes that create nonsignificant associations. Still, these results should not be seen as suggesting that homeownership fails to deliver any health benefits to racial/ethnic minorities but instead should provide a baseline for future investigations.