Role of Race and Ethnicity in Private Long-Term Care Insurance Ownership

Abstract

Purpose of the Study: To determine if racial/ethnic disparities exist in the ownership of private long-term care insurance (LTCI) among current Medicare beneficiaries. Design and Methods: This study used the 2011 wave of the National Health and Aging Trends Study. Bivariate analysis and multivariate logistic regression were employed to isolate the independent effects of race/ethnicity on LTCI uptake. Stratified multivariate analyses were used to further examine the effect of race/ethnicity on LTCI ownership. Results: 12.3% of Blacks and 5.8% of Hispanics, compared with 20.2% of Whites (p < .001), reported having LTCI coverage. We found that Hispanics were 48% less likely to have LTCI (p = .005) compared with Whites, whereas no difference was found between Blacks and Whites. Compared with White women, Hispanic women were 81% less likely to be insured (p < .001). Ethnic disparities persisted among individuals who, based on income and assets, are considered appropriate for purchasing private LTCI coverage. Implications: This study demonstrates that ethnic differences exist in the ownership of LTCI among elderly Americans. Additional research is needed to determine what factors are responsible for the apparent underrepresentation of Hispanics in the LTCI market.

When the Obama administration suspended the component of the Patient Protection and Affordable Care Act that would have created a federally sponsored form of voluntary long-term care insurance (LTCI) in October of 2011(Aizenman, 2011), it was a signal that the ways in which long-term care (LTC) is financed in this country would remain unchanged for the foreseeable future. Despite the recent creation of a bipartisan Long-Term Care Commission tasked with examining potential solutions to the challenge of financing LTC for an aging population, it appears that no large-scale policy fixes are imminent (Weiner, 2013). Absent the Community Living Assistance Services and Supports Act or other major reforms, Medicaid and private expenditures will continue to be the primary funders of both community-based and institutional LTC (Kaiser Family Foundation, 2012). Yet, as federal and state governments attempt to control ever increasing Medicaid expenditures, public coverage of custodial LTC services (which account for more than a third of national Medicaid spending [Kaiser Family Foundation, 2009]) appears particularly vulnerable to future budgetary cuts (Bernstein, 2012). It is not clear, however, how individuals who need LTC services will make up for any reductions in Medicaid coverage should these changes be enacted. Given that the median cost of a year in a nursing home is about $84,000 (Genworth, 2013), it seems unlikely that all individuals will be able to absorb such costs out of pocket. Private LTCI has been developed to help address this gap in coverage (Cohen, 2003; Crown, Capitman, & Leutz, 1992; Johnson & Uccello, 2005; Meiners, 1983, 1993).

Should cuts to Medicaid LTC coverage be enacted, seniors of racial and ethnic minorities may have particular need for LTCI policies as they are disproportionately represented among elderly Medicaid enrollees. Specifically, African Americans and Hispanics comprise 30% of those who are dually eligible for Medicare and Medicaid, but just 18% of the Medicare-only population (Kaiser Family Foundation, 2009). However, there is reason to suspect that these products may not adequately meet the needs of elderly people of color. The population of African Americans and Hispanics aged 65 and older is estimated to increase at rates substantially greater than those of non-Hispanic Whites over the next 20 years (Greenberg, 2011), suggesting future increases in the demand for LTC among these minority groups that could significantly outpace the current growth trajectory of the LTCI market. Additionally, it is well documented that nonelderly minorities have traditionally been underrepresented among private and nongroup health insurance enrollees (Kirby, Taliaferro, & Zuvekas, 2006; Lillie-Blanton & Hoffman, 2005; Saver, Doescher, Symons, Wright, & Andrilla, 2003). Given that LTCI plans are purchased largely on an individual basis with limited federal and state tax subsidies (Brown & Finkelstein, 2009; Wiener, Tilly, & Goldenson, 2000), there is potential for similar disparities to exist Within this marketplace. Finally, bivariate analysis by Johnson and Park (2011) has demonstrated that fewer African Americans and Hispanics aged 55 and older report owning private LTCI compared to Whites, providing preliminary evidence of racial/ethnic differences in rates of coverage.

As policy makers hope to encourage the purchase of private LTCI to reduce Medicaid expenditures while maintaining protection for elders from the substantial financial risk of LTC, it is important to understand the role that race and ethnicity may play in the uptake of this product. To date, very little research has been done to examine how LTCI ownership varies across racial and ethnic groups, making it unclear if currently offered products can serve as viable financing alternatives for traditionally disadvantaged groups like Blacks and Hispanics. Our study fills this knowledge gap by examining the independent effect of race and ethnicity on ownership of a private LTCI plan, after controlling for a variety of individual characteristics related to anticipated demand for LTC and socioeconomic status, among a nationally representative sample of Medicare beneficiaries, as well as pertinent subgroups, including those who are likely appropriate for LTCI policies based on their financial resources.

Conceptual Model

This analysis integrates expected utility theory and its implications on the demand for insurance with the Andersen behavioral model of health care access to gain a better understanding of the role a social factor like race/ethnicity plays in the purchasing decision. Traditional expected utility theory suggests that a rational risk-averse consumer will look to minimize uncertainty by choosing a small certain financial loss over the potential for a large unexpected loss (Arrow, 1963; Friedman, 1974). Given the uncertainty regarding one’s future need for LTC and the large financial costs associated with these services, one would predict, using the expected utility framework, that rational consumers may be willing to pay a premium to shed this risk. LTCI is one manner in which an individual can achieve a degree of financial certainty with regard to LTC expenses (a recent study of LTCI claimants found that about 98% of beneficiaries had their claims approved and between 60% and 75% beneficiaries reported that most or all of their LTC costs had been covered; Doty, Cohen, Miller, & Shi, 2010).

One’s demand for a LTCI policy is influenced by several factors including the probability of needing LTC, the cost of LTCI in comparison to competing alternatives, risk tolerance, and preferences regarding formal versus informal LTC and asset protection. For example, individuals currently in poor health may have a greater likelihood of needing LTC at some point in their lifetime compared with those in good health. As such, one should expect unhealthy individuals to have greater demand for LTCI than healthy individuals, all else being equal. Similarly, greater demand for LTCI should also be anticipated if it can be obtained at a lower cost than suitable substitutes like relying on informal care giving by family members or “spending-down” income and assets to meet state Medicaid eligibility requirements. Additional factors, like an individual’s desire to bequest wealth to his or her children or prevent the impoverishment of a spouse, have also been identified as important determinants of LTCI demand within an expected utility conceptualization (Pauly, 1990).

Although expected utility theory suggests a variety of economic factors that are likely to be predictive of LTCI ownership, it does not directly account for social factors that could conceivably influence an individual’s awareness and perception of these products. According to the Andersen behavior model, these social constructs are considered predisposing factors: individual characteristics that determine one’s propensity to have access to the health care system, regardless of the need or ability to pay for services (Andersen & Newman, 1973). Since this model’s creation, race and ethnicity have been identified as two key predisposing elements that help to capture the social barriers to medical care a person faces, as well as his or her cultural attitudes and beliefs about disease and the health care system. With this in mind, we hypothesize that race and ethnicity will be important predisposing factors for LTCI ownership, after controlling for demand-side predictors that are consistent with economic theory, with Blacks and Hispanics being less likely to have coverage.

Design and Methods

We employed a public-use database from the first wave (2011) of the National Health and Aging Trends Study (NHATS). NHATS is sponsored by the National Institute on Aging (grant number NIA U01AG032947) through a cooperative agreement with the Johns Hopkins Bloomberg School of Public Health (NHATS, 2011). It is a nationally representative survey of Medicare beneficiaries age 65 or older, or a proxy if necessary, containing individual and family characteristics in a variety of domains including demographics, finances, social structures, and health. The survey uses a three-stage sampling design, pulling subjects from the Medicare enrollment database and oversampling Blacks and the very old to ensure adequate sample size for these groups. In-person interviews were conducted in the respondents’ homes by a professionally trained interviewer. In Round 1, 12,411 beneficiaries were contacted and 8,245 responded, resulting in an unweighted response rate of 70.9%.

Participants and Variables

The analytical sample included 7,777 community-dwelling Medicare enrollees. The outcome variable of interest was whether the respondent reported owning any form of private LTCI that would pay for a year or more of nursing home care, assisted living facility, or home care. The key independent variable of interest was the respondent’s self-reported race/ethnicity, which was classified into the following mutually exclusive groups: (a) White, non-Hispanic; (b) Black, non-Hispanic; and (c) Hispanic. Two-hundred and twenty individuals who reported their race as “other” were excluded. Subjects with missing data for any of the variables used in the full model were excluded from analysis, restricting the final sample to 6,067 respondents (74% of total).

Control Variables

Several additional factors, which expected utility theory and prior studies have identified as key components in the decision to insure against the financial risk of needing LTC, were also included. An individual’s age directly affects the probability of needing LTC and therefore the price of a policy (Department of Health and Human Services, 2012). Age was collected as a categorical variable in the NHATS questionnaire with the following groupings: 65–69; 70–74; 75–79; 80–85; 86–89; 90+ years old. Gender was also controlled for as women are significantly more likely to need LTC at some point in their lifetime, largely due to their greater life expectancy, and therefore may have a greater demand for LTCI (Brown & Finkelstein, 2007).

Marital status was also included as a spouse may be able to serve as a substitute for formal LTC, thereby reducing one’s expected demand for future LTC services (Brown & Finkelstein, 2009). Conversely, a spouse could also increase demand for LTCI if one wishes to protect the family’s wealth to avoid impoverishing a spouse in the event of prolonged LTC use (Pauly, 1990). Marital status was dichotomized as “single” (separated; divorced; widowed; never married) or “married” (currently married; living with a partner). Similarly, the presence of children may influence insuring behavior. Children can provide another source of informal care that could substitute for formal LTC services, again reducing demand for LTCI (Mellor, 2001). Children could also provide greater motivation to bequest a portion of one’s wealth, in which case an individual would be more likely to insure in hopes of preventing an exhaustion of assets (Brown, Goda, & McGarry, 2012). A categorical variable was included counting the number of children as none; 1; 2–4; and 5+. Health status is another key predictor of LTCI demand. Those in poorer health may have greater demand for LTCI assuming their probability of needing LTC is greater (Norton, 2000). Poor health status that is not observable by insurers, nor required to be disclosed, may therefore increase the likelihood of purchasing LTCI (Finkelstein & McGarry, 2006). Otherwise, poor health status may preclude LTCI purchase due to medical underwriting in this market (Sloan & Norton, 1997; Temkin-Greener, Mukamel, & Meiners, 2000). Self-reported overall health status was selected as the most appropriate measure of an individual’s health in this model because one’s perception of current health is likely to be the most reflective of his or her perception of the future risk of needing LTC. This variable was categorized as poor–fair, good, or very good–excellent.

Education has previously been shown to be an important predictor of LTCI demand (McCall, Mangle, Bauer, & Knickman, 1998; Mellor, 2001) potentially due to greater levels of risk aversion (Sloan & Norton, 1997) or increased willingness to delay gratification and plan for the future (Cramer & Jensen, 2006) among individuals with higher levels of education. Information on this factor was collected on both the sampled person and his or her spouse (if applicable), so a household education variable (highest level of education attained by either the respondent or spouse) was included in the analysis. Education was categorized as less than high school, high school/equivalent, some college/vocational school/associate’s degree, bachelor’s degree, and graduate/professional degree.

We also controlled for income and assets for two important reasons. First of all, these measures account for one’s ability to afford a LTCI plan (Sloan & Norton, 1997). Second, they capture the price a family would face if they were to consciously choose to forgo private LTCI and rely on Medicaid to finance future LTC needs. Although eligibility requirements and spousal allowances vary by state, the price of Medicaid coverage roughly amounts to a premium equal to one’s assets with an annual deductible equal to annual income (Brown, Coe, & Finkelstein, 2007). These variables were created by using self-reported values of annual income, asset worth, and outstanding loan amounts when available. In some instances, however, respondents did not report a point estimate but rather selected an applicable category from a list (i.e., bank account worth between $10,000 and $20,000). When this occurred a point estimate was generated for the respondent by using the median value within the reported range. This technique allowed for the calculation of net assets (total asset worth less outstanding loan amounts), which is a better representation of the wealth one has available to pay for LTC. The continuous estimates of income and net household assets were subsequently converted into categorical variables in order to capture the nonlinear relationship between these factors and the dependent variable. Income categories included $10,000 or less; $10,001–$20,000; $20,001–$35,000; $35,001–$65,000; $65,001–$100,000; $100,001–$200,000; and greater than $200,000. Asset categories included $10,000 or less; $10,001–$75,000; $75,001–$150,000; $150,001–$300,000; $300,001–$800,000; and greater than $800,000.

Analyses

We performed bivariate and multivariate analyses, accounting for the sampling and weighting approaches of the NHATS. In bivariate analyses, we used χ² tests to compare differences in individual characteristics and the rate of private LTCI purchase across racial/ethnic groups. We also used similar tests to compare differences in LTCI ownership rates across groups defined by each covariate factor.

In multivariate analyses, we first estimated a logistic regression model where the dependent variable was an indicator for owning a LTCI policy (1 = yes; 0 = no), and the independent variables were two indicators for Black race and Hispanic ethnicity, respectively (non-Hispanic, White race used as a reference group). The model controlled for other demand-side characteristics previously identified. Odds ratios (OR) and associated p values of the model were reported.

To further isolate the impacts of race and ethnicity on owning private LTCI, we performed a series of stratified multivariate analyses within pertinent population subgroups. Namely, the sample was partitioned by whether or not the respondent’s family was recommended to consider LTCI (based on family wealth), by age (64–74; 75+), by gender, by health status (poor–fair; good; very good–excellent), and by household education level (less than high school; high school graduate and above). The recommendation status of a respondent was determined by comparing income and assets to a set of basic guidelines that are published online by insurance companies and consumer advocacy groups. Specifically, for individuals or families with annual incomes between $25,000 (Guide to Retirement Living, 2012; Long-Term Care Insurance Axis, 2012) and $200,000 or liquid assets (household assets less the value of one’s home and/or business, if applicable) worth between $70,000 (Complete Long Term Care, 2012) and $1,500,000 (Guide to Retirement Living, 2012), LTCI purchase is defined as “Recommended.” For respondents with income and assets below these levels, the relative cost of “spending-down” onto Medicaid when LTC is needed is probably lower than the lifetime cost of LTCI premiums. Meanwhile, those with incomes and assets above this threshold are likely wealthy enough to self-insure. It should be noted that published guidelines on income and asset thresholds vary widely. Furthermore, no upper income threshold could be found among online resources. The published levels used here were chosen to maximize the number of individuals who are recommended to consider LTCI, whereas $200,000 was selected as an upper bound because families with this degree of retirement income should be able to comfortably afford a full year in a private nursing home room for both spouses should the need arise. Determining the actual accuracy of these guidelines is beyond the scope of this article; however, sensitivity analysis was conducted by increasing/decreasing the upper/lower thresholds by 15%, 30%, and 50% to assess how our results changed with varying assumptions. All analyses were performed with the survey estimation routines in Stata version 11.0, Special Edition (STATA Corp., College Station, TX).

Results

Black respondents represented 8% (n = 2,295, 219) of the estimated total population, whereas Hispanic respondents comprised about 7% (n = 2, 112,704; Table 1). Compared with Whites, Blacks and Hispanics were more likely to be under the age of 75 (60% and 56% vs. 53%), less likely to be married (37% and 52% vs. 59%), less likely to report very good or excellent overall health status (26% and 20% vs. 49%), and more likely to have 5 or more children (26% and 27% vs. 15%). Fifty-nine percent of Whites had household education levels beyond high school compared with 36% of Blacks and 25% of Hispanics. Whites were also more likely to have higher incomes and net worth compared with both Blacks and Hispanics. In accordance with this difference, racial and ethnic variation was also found in the proportion of families who were “recommended” to consider LTCI purchase. Specifically, about 64% of Whites fell into the “recommended” category, compared with just 36% of Blacks and 31% of Hispanics (p < .001).

Table 1.

Population Characteristics: National Health and Aging Trends Study, 2011

	Total (N = 6,067)	White (N = 4,339)	Black (N = 1,328)	Hispanic (N = 400)	p a
Population size	28,396,798	23,980,331	2,306,444	2,110,023
Age (%)					.045*
65–69	28.6	28.2	31.4	29.3
70–74	25.1	24.6	28.1	26.8
75–79	19.3	19.4	18.1	19.5
80–85	14.5	14.8	12.4	13.6
86–89	8.6	8.8	6.4	8.1
90+	4.0	4.2	3.7	2.8
Gender (%)					.295
Male	45.1	45.5	41.1	44.8
Female	55.0	54.6	58.9	55.2
Marital status (%)					<.001***
Single, divorced, separated, or widowed	43.3	40.9	62.8	48.4
Married or Living with Partner	56.7	59.1	37.2	51.6
Household education (%)					<.001***
Less than high school	17.8	12.7	36.9	55.7
High school graduate (or equivalent)	27.8	28.5	27.5	19.5
Vocational school, some college, or associate degree	24.3	25.6	19.8	14.0
Bachelors degree	14.2	15.5	8.0	6.5
Graduate or professional degree	15.9	17.7	7.9	4.3
Number of children (%)					<.001***
0	9.0	8.9	9.7	9.2
1	11.4	11.1	14.3	12.4
2–4	63.1	65.5	49.8	51.2
5 or More	16.5	14.6	26.3	27.2
Health status (%)					<.001***
Poor–fair	25.1	21.7	39.1	48.0
Good	30.0	29.3	35.1	31.8
Very good–excellent	44.9	48.9	25.8	20.2
Annual household income (%)					<.001***
$10,000 or less	13.3	10.0	27.2	34.7
$10,001–$20,000	20.7	19.0	29.8	29.3
$20,001–$35,000	21.1	21.8	19.0	15.7
$35,001–$65,000	22.3	23.7	15.7	14.2
$65,001–$100,000	12.4	13.9	4.9	4.2
$100,001–$200,000	7.8	8.9	2.5	1.6
>$200,00	2.4	2.7	0.9	0.3
Net household assets (%)					<.001***
$10,000 or less	23.6	18.7	47.8	52.6
$10,001–$75,000	13.9	13.0	21.3	17.0
$75,001–$150,000	13.6	14.0	13.3	9.4
$150,001–$300,000	16.8	18.1	10.1	9.3
$300,001–$800,000	18.5	20.6	5.5	8.2
>$800,000	13.6	15.6	2.0	3.6
“Recommended” for LTCI					<.001***
Yes	59.2	63.8	36.3	31.2
No	40.8	36.2	63.7	68.8

^aχ² used to test for differences between racial/ethnic groups.

*Indicates significance at the p ≤ .05 level.

***Indicates significance at the p ≤ .001 level.

Bivariate Predictors

Overall, the rate of private LTCI uptake was 18.4% (Table 2). 12.3% of Blacks and 5.8% of Hispanics reported having private LTCI coverage compared with 20.1% of Whites (p < .001). Other factors associated with having private LTCI in bivariate analysis were aged 85 or younger, being married, higher levels of household education, having fewer than five children, being in good to excellent health, and higher levels of income and assets.

Table 2.

Bivariate and Multivariate Predictors of Private Long-Term Care Insurance Status

	Private long-term care insurance status
	Bivariate analysis		Multivariate analysis
	Has LTCI (%)	p a	Odds ratio	p
Total	18.4
Race/ethnicity		<.001***
White (ref)	20.1
Black	12.3		1.04	.652
Hispanic	5.8		0.52	.005**
Age		.010*
65–69 (ref)	19.1
70–74	17.9		1.08	.589
75–79	20.0		1.41	.001***
80–85	19.8		1.45	.009**
86–89	15.6		1.17	.208
90+	10.0		0.79	.181
Gender		.152
Male (ref)	19.4
Female	17.7		1.11	.287
Marital status		<.001***
Single, divorced, separated, or widowed (ref)	14.5
Married or living with partner	21.4		0.91	.395
Household education		<.001***
Less than high school (ref)	6.3
High school graduate (or equivalent)	12.9		1.49	.015*
Vocational school, some college, or associated degree	19.1		2.12	<.001***
Bachelors degree	25.4		2.41	<.001***
Graduate or professional degree	34.2		3.25	<.001***
Number of children		.03*
0 (ref)	19.3
1	16.6		0.85	.41
2–4	19.5		0.95	.74
5 or more	14.9		0.90	.502
Health status		<.001***
Poor–fair (ref)	9.6
Good	17.2		1.42	.003**
Very good–excellent	24.1		1.66	<.001***
Annual household income		<.001***
$10,000 or less (ref)	9.1
$10,001–$20,000	10.4		0.98	.896
$20,001–$35,000	13.7		0.96	.803
$35,001–$65,000	22.8		1.43	.020*
$65001–$100,000	29.8		1.73	.002**
$100,001–$200,000	32.7		1.72	.003**
>$200,00	33.4		1.87	.018*
Net household assets		<.001***
$10,000 or less (ref)	8.0
$10,001–$75,000	12.8		1.48	.019*
$75,001–$150,000	14.5		1.44	.01**
$150,001–$300,000	22.1		1.95	<.001***
$300,001–$800,000	26.3		2.00	<.001***
>$800,000	30.8		1.90	<.001***

^aχ² used to test for differences in private LTCI ownership among independent variable categories.

*Indicates significance at the p ≤ .05 level.

** Indicates significance at the p ≤ .01 level.

***Indicates significance at the p ≤ 0.001 level.

Multivariate Predictors

After controlling for individual and family level characteristics, Black race no longer appeared to be an important predictor of insurance status (OR = 1.04, p = 0.652), While ethnicity remained significantly associated with a lower rate of LTCI purchase (OR = 0.52, p = .005). Several additional factors were predictive of LTCI purchase, with higher education, age between 75 and 85, better self-reported health, and higher household income/assets all associated with a higher likelihood of being insured (Table 2).

Stratified Analysis

Stratifying the sample by whether or not the respondent is recommend for LTCI purchase based on the income and asset thresholds discussed previously revealed that Hispanics were less likely to have LTCI compared with their White counter parts (OR = 0.52, p = .037 [“recommended” subgroup] and OR = 0.49, p = .024 [“not recommended” subgroup]), regardless of financial circumstances (Table 3). Sensitivity analysis showed that narrowing these income and asset thresholds did not substantially alter this finding as Hispanics in both categories consistently had lower odds of having coverage with significance levels that reached or approached significance at the p = .05 level (Table 4). Within the “Recommended” subgroup Blacks were more likely than Whites to hold LTCI (OR = 1.31, p = .02); however, this finding was not robust to alternative recommendation thresholds. As such, it is not clear that this finding represents a true difference in LTCI ownership between Blacks and Whites who meet the recommendation guidelines used here. We did not find significant racial or ethnic disparities in other stratified analyses by age, gender, or health status. However, ethnic differences persisted across education groupings as Hispanics were less likely to have LTCI regardless of whether they had below a high school education (OR = 0.40, p = .034) or a high school diploma and beyond (OR = 0.55, p = .026). Finally, in the stratified analysis for women, Hispanic women were more than 80% less likely to have private LTCI compared with White females (OR = 0.19, p < .001).

Table 3.

Effect of Race/Ethnicity on Private Long-Term Care Insurance Ownership Status: Stratified by Select Respondent Characteristics^a

	Race/ethnicity
	Black		Hispanic
	Odds ratiob	p	Odds ratiob	p
Recommended	1.31	.020*	0.52	.037*
n = 3,205
Not recommended	0.75	.099	0.49	.024*
n = 2,862
Age group: 65–74	0.99	.969	0.2	.122
n = 1,162
Age group: 75+	1.06	.593	0.72	.182
n = 4,905
Gender: male	0.99	.918	0.95	.84
n = 2,617
Gender: female	1.09	.531	0.19	<.001***
n = 3,450
Health status: fair–poor	1.14	.472	0.66	.38
n = 1,769
Health status: good	0.89	.462	0.63	.163
n = 1,904
Health status: very good–excellent	1.18	.301	0.4	.066
n = 2,394
Education: less than high school	1.04	.876	0.41	.034*
n = 1,319
Education: high school graduate+	1.09	.39	0.55	.026*
n = 4,648

^aControl variables for each stratified analysis are the same as in the full model with the exception that the variable on which the sample is stratified is excluded. Income and asset measures remain in the recommended/not recommended analysis as both high income/asset and low income/assets respondents are in the “not recommended” stratum.

^bOdds ratios are in reference to Whites within the same stratum.

*Indicates significance at the p ≤ .05 level.

***Indicates significance at the p ≤ .001 level.

Table 4.

Results of Sensitivity Analysis of Recommendation Guidelines

				Race/ethnicity
				Black			Hispanic
				Odds ratioa	p	n	Odds ratioa	p	n
Baseline	Income	$ 25,000	Recommended	1.31	.020*	452	0.52	.037*	111
		$ 200,000	n = 3,205
	Assets	$ 70,000	Not recommended	0.75	.099	876	0.49	.024*	289
		$ 1,500,000	n = 2,862
15%	Income	$ 29,000	Recommended	1.23	.115	377	0.55	.065	96
		$ 170,000	n = 2,813
	Assets	$ 81,000	Not recommended	0.87	.41	951	0.47	.016*	304
		$ 1,280,000	n = 3,254
30%	Income	$ 33,000	Recommended	1.18	.321	317	0.50	.041*	79
		$ 140,000	n = 2,384
	Assets	$ 91,000	Not recommended	0.95	.709	1011	0.71	.044*	321
		$ 1,050,000	n = 3,683
50%	Income	$ 38,000	Recommended	1.05	.23	224	0.35	.029*	58
		$ 100,000	n = 4,285
	Assets	$ 105,000	Not recommended	1.03	.792	1,104	0.62	.095	342
		$ 750,000	n = 1,782

^aOdds ratios are in reference to Whites within the same stratum.

*Indicates significance at the p ≤ .05 level.

Discussion

After controlling for a variety of individual and family level characteristics, we found that Hispanic seniors were statistically significantly less likely than Whites to hold a private LTCI policy indicating that ethnicity is an important predisposing factor. The observed discrepancy persisted when the population was limited to those who, based on family wealth, were recommended to consider LTCI purchase. Ethnic disparities also appeared particularly pronounced among Hispanic women. However, being Black did not have a significant impact on the likelihood of having private coverage, contrary to our hypothesis.

The overall rate of LTCI uptake in our study, 18.4%, was higher than previous estimates of 10%–15%.(Brown & Finkelstein, 2007; Brown et al., 2012) The higher rate may be attributed to the higher levels of education (30% with a bachelors degree or higher vs. 22%) and income (mean income of $59,000 vs. $46,000) in our sample compared with the Census Bureau’s most recent estimates for the general population aged 65 and older (American Community Survey 1-Year Estimates, 2011). Nevertheless, our findings largely coincide with those presented in existing literature. Namely, family structure variables, such as marital status and number of children were not statistically significant predictors (McCall et al., 1998; Mellor, 2001), and we found no differences between men and women in LTCI purchase (Brown & Finkelstein, 2007). Individuals with higher levels of education, better self-reported health, higher incomes, or higher assets were significantly more likely to hold LTCI (Cramer & Jensen, 2006; McCall et al., 1998; Mellor, 2001; Schaber & Stum, 2007; Sloan & Norton, 1997). This concordance provides some evidence of external validity for the multivariate model used here.

The principal finding that Hispanics are much less likely to have private LTCI is consistent with previous research showing that individuals of this ethnic group are traditionally the least likely to use formal LTC services (Angel, Angel, McClellan, & Markides, 1996; Angel & Angel, 1999). There is evidence to suggest that the low rates of LTC use are a result of cultural preferences, an expectation that family members provide needed care, and a distrust of institutions such as nursing homes (Angel et al., 1996; Herrera, Lee, Palos, & Torres-Vigil, 2008; John, Resendiz, & de Vargas, 1997; Min & Barrio, 2009). In this light, at least part of the low prevalence of Hispanics in the LTCI market may be due to seniors’ preference for and the availability of informal long-term caregivers and therefore should not be interpreted as a true “disparity.” However, more recent findings indicate that the demand for nursing home care among Hispanics has been increasing, a trend attributed to a decreasing supply of informal caregivers as more Hispanic women enter the workforce (Fennell, Feng, Clark, & Mor, 2010). What’s more, a recent examination of LTCI claimants found having such policies gave individuals more flexibility to receive LTC outside of a nursing home (Doty et al., 2010), indicating that private LTCI is not incompatible with an aversion to nursing homes. In conjunction with a rapidly growing elderly population and disproportionately high rates of disability (Markides, Eschbach, Ray, & Peek, 2007), these developments suggest that the lower likelihood of private LTCI in this population may go beyond a simple expression of preferences. A recent survey of California voters further suggests that the demand for formal LTC and the financial strain of paying for these services are growing within the Hispanic community (The SCAN Foundation, 2012). Specifically, Latinos reported the highest rate (70%) of anticipated need for paid LTC in the next 5 years for a family member, as well as the lowest rate of self-reported ability to pay for nursing home or home health care (91% and 86% reported not being able to afford more than 3 months of care in the respective settings).

In light of these findings, it seems unlikely that cultural norms can fully explain the relative absence of Hispanics in the LTCI market. Instead, there may be additional factors that are limiting consumer awareness of these plans, restricting access to them, or both. One possibility is that private insurance companies that offer these plans are not actively marketing them toward Hispanic consumers. A lack of targeted advertising may be of particular importance to ethnic minorities as more than a third of Hispanic individuals aged 65 and older are estimated to speak English at a level less than “very well” (US Census Bureau, 2011). Another factor that may inhibit awareness of the need for LTCI is the erroneous belief that LTC is covered by Medicare. The false belief that these services are covered as part of one’s Medicare entitlement has previously been identified as a barrier to LTCI purchase for all individuals (Brown et al., 2012); however, Hispanics may be particularly susceptible to misconceptions if Medicare advertising and information materials are not adequately distributed in Spanish. This may explain why only about one-fifth of Latinos, compared with 40% of Whites, correctly identified that custodial nursing home care is not covered by Medicare in the previously mentioned survey of California voters. It should be noted, however, that similar rates of confusion about Medicare were found between African Americans and Latinos, indicating that additional factors may help mitigate the effect of these misconceptions for racial but not ethnic minorities. Finally, it is also plausible that the historical absence of Hispanics from the formal LTC system has contributed to a deficit in experiential knowledge about what care options are available, the ways in which they can be paid for, and the need to plan for such life events.

The results of our stratified analysis indicated that Hispanics were consistently less likely to have LTCI across education strata, suggesting that the interaction of ethnicity and education is likely not a primary driver of the differences observed in the overall sample. Stratifying by gender, meanwhile, revealed that the low rates of insurance purchase are most apparent among Hispanic females as this group was 81% less likely to have coverage than White females. It is not clear if this discrepancy reflects greater availability of informal caregivers for elderly Hispanic females or if it is a result of deficiencies in awareness and understanding of the need for LTC and the system through which it is provided. Evidence suggests that women may display lower levels of financial literacy and interest in financial decision making (Chen & Volpe, 2002; Lusardi & Mitchell, 2008). In conjunction with traditional Hispanic beliefs that males should serve as the primary family decision makers (Galanti, 2003), this may create an environment in which females are not adequately prepared to plan for their future financial and health-care-related needs.

It is important to note that not all of the differences observed in LTCI purchase rates for Hispanics should be viewed as suboptimal. When the sample population was stratified by income and assets into those who are and are not recommended to consider LTCI, we see some evidence of greater “rational” behavior among Hispanics in the “not recommended” group (i.e., they were less likely to purchase LTCI than Whites). Given the current structure of the Medicaid system, it is generally accepted that those with low levels of assets and income would be ill-advised to purchase LTCI as the cost of this plan over time would essentially be greater than the cost of “spending-down” onto Medicaid. Additionally, those with large retirement incomes or substantial assets should also be capable of paying for LTC out of pocket without financial ruin. Our analysis suggests that ethnic minorities are more likely to adhere to these guidelines, possibly indicating that White seniors may be more risk averse, have a stronger preference to avoid relying on Medicaid coverage, or have a lack of understanding of Medicaid eligibility requirements. The difference in LTCI coverage between Hispanics and Whites did persist, however, when analysis focused only on individuals recommended for LTCI purchase, suggesting that unwarranted differences may exist in this market.

Several study limitations should be mentioned here. First, information about individuals’ preferences was not available, so the independent effect of taste on LTCI purchase could not be directly identified. Furthermore, we analyzed only whether respondents currently hold LTCI. We do not know whether individuals previously had LTCI and dropped their policy or if they have ever considered buying it in the past. Consequently, we were unable to quantify awareness about this product and were able to only draw conclusions based on observed behavior. However, current LTCI ownership among those aged 65 and older is likely to be predictive of having coverage at the time it is needed, thus it provides a good estimate of how different racial and ethnic groups will pay for LTC in the future. Finally, due to lack of information our analysis did not control for supply side factors such as the price of LTCI policies, the availability of insurance providers, and state Medicaid spend down requirements. Although these factors are expected to play a role in the purchasing decision, they are unlikely to entirely account for the sizeable difference in LTCI uptake among Hispanics.

In conclusion, this analysis suggests that Hispanics are disproportionately absent among current owners of private LTCI and that ethnicity is an important predisposing factor in the purchase of such products. Given that this ethnic group is expected to have growing demand for LTC in the future, they would appear to be particularly vulnerable to future reductions in Medicaid coverage. This suggests that policy makers should consider strategies to improve LTCI uptake by Hispanics in conjunction with any proposed plans to reduce Medicaid coverage of LTC services. A targeted information and marketing campaign aimed specifically at Hispanics who have appropriate income and asset levels is one such possibility. Furthermore, greater outreach may be needed to raise awareness among female Hispanics about the benefits of LTCI. Further research is needed to discern why Hispanics are less likely to purchase LTCI policies as this information will be essential in devising strategies to reduce the observed differences.

Funding

This work was supported by the National Institute on Aging of the National Institute of Health [R01 AG032264]; the Agency for Health Care Quality and Research [T32 HS000044].