| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||
RESEARCH ARTICLE |
Institute of Gerontology, Wayne State University, Detroit, Michigan.
Address correspondence to Anne Theisen Cramer, Institute of Gerontology, Wayne State University, 87 East Ferry Street, Detroit, MI 48202. E-Mail: atcramer{at}wayne.edu
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Methods. Using data from the 2002 Health and Retirement Survey, we estimated logistic regressions to model consumer decisions to purchase LTC insurance. We explored several alternative measures of the price of a policy.
Results. Price was a significant determinant in decisions to purchase coverage. The demand for coverage, however, was price inelastic, with elasticities ranging from –0.23 to –0.87, depending on the specification of the model. The education level and income of the purchaser were also important.
Discussion. This analysis provides the first estimates of price elasticity of demand for LTC insurance. The finding that demand is very price inelastic suggests that state initiatives that effectively subsidize premiums as a way of stimulating purchases are likely to meet with very limited success in the present environment.
PROVIDING for long-term care (LTC) is one of the major challenges of the 21st century. Care is expensive, and the likelihood of needing LTC at least once in a lifetime is substantial. Today's 65-year-old faces a 24% chance of having a year-long or longer nursing home stay sometime during his or her remaining life (Spillman & Lubitz, 2002
). The cost of this stay, in today's dollars, will be $74,095 on average, and additional care provided at home can cost more than $19 per hour (MetLife Mature Market Institute, 2005).
Who pays for this care? In 2002, public funds accounted for more than $115 billion of LTC spending (more than 64%), with Medicaid's share accounting for close to $85 billion (Georgetown University, 2004). Public policy analysts warn that Medicaid will not be able to provide the amount of LTC funding that will be needed as baby boomers age.
One possible alternative to public funding is private LTC insurance. Private insurance policies for LTC have been marketed since the early 1980s, but even today, few adults purchase them. As of year-end 2002 there were only about 7 million LTC policies in force in the United States, with the overwhelming majority having been purchased in the individual market (America's Health Insurance Plans, 2004).
Researchers have offered several theories as possible explanations for the low LTC insurance purchase rate. For low- and middle-income adults, the availability of Medicaid crowds out the demand for private insurance. Other adults decide to self-insure, meaning they will rely on private savings should the need for LTC arise. Many adults have the misperception that plans they already have, such as Medicare and Medigap coverage, will pay for LTC. Some believe the likelihood of needing LTC services is small. For many, the coverage is unaffordable, or other expenses are more immediate. And some parents prefer to rely on children for caregiving, should the need arise.
Why do so few persons purchase LTC insurance policies? This article addresses this question. Using data from the ongoing Health and Retirement Survey (HRS), we report on an empirical analysis of factors influencing decisions to purchase new LTC insurance policies in 2002. To our knowledge, this analysis is the first to quantify the role of price in purchasing decisions in this market. Knowledge of the price elasticity of demand is useful for policy purposes. For example, it can be used to gauge the effectiveness of using premium subsidies or tax subsidies to increase purchases of private LTC insurance.
What Is LTC Insurance?
Most LTC policies are individually purchased indemnity insurance plans. They begin paying benefits when the insured can no longer perform routine activities of daily living or becomes cognitively impaired. Typically, if the individual develops difficulty with two or more standard activities of daily living—such as bathing, dressing, eating, toileting, and transferring (out of bed or out of a chair)—benefits will be triggered. Covered services usually include skilled care (such as nursing care and professional therapy), as well as personal care services, household services, and custodial services. These services can be provided in a nursing home, assisted living facility, or at home, depending on the preferences of the individual.
LTC insurance is distinctly different from other types of insurance. Unlike health insurance (including Medicare), which covers medical costs associated with recovery from sickness, a medical condition, or an accident, LTC insurance covers services needed when the prospect of regaining health and functioning is unlikely. In addition, unlike most individually purchased health insurance policies, LTC insurance protects against changes in classification risk as an individual ages. In this respect, LTC insurance is similar to term life insurance in that policies are guaranteed renewable with no change in annual premium as a result of aging. LTC insurance differs from disability insurance in that it does not provide income replacement, and it differs from life insurance in that there is no cash value in the event of death.
Why Purchase LTC Insurance?
There are many reasons why a person might want to purchase LTC insurance. First, LTC insurance enables an individual to preserve assets and savings, rather than depleting them in order to pay for care or in order to qualify for Medicaid. As a result, an individual can use assets and savings to provide for his or her own comfort in retirement, to provide for a surviving spouse, or to leave a bequest for children. Second, LTC insurance provides flexibility in choosing care options in the event LTC becomes necessary. Finally, LTC insurance allows for budgeted payments while ensuring that funds will be available when needed.
Over the years, several initiatives aimed at increasing LTC insurance sales have focused on ways to make private LTC insurance policies more attractive. The 1996 Health Insurance Portability and Accountability Act gave taxpayers who itemize deductions the right to consider LTC insurance premiums when figuring eligible medical expense deductions. The Partnership for Long-Term Care, a program funded by the Robert Wood Johnson Foundation and implemented in four states in the early 1990s, combines private insurance and Medicaid in an effort to make private LTC policies more affordable and more attractive (McCall, Mangle, Bauer, & Knickman, 1998). For persons who qualify, these financial incentives are equivalent to a discount on the price of LTC insurance. In a number of states, financial incentives have been supplemented with educational programs intended to educate seniors about the need for LTC planning (National Governors Association, 2004).
Why Don't People Purchase LTC Insurance?
Researchers have offered a number of explanations for the low prevalence of private LTC insurance policy purchases. One is the availability of potential substitutes. For persons with low income, Medicaid is a safety net in the event LTC is needed (Hubbard, Skinner, & Zeldes, 1995). Because Medicaid is a means-tested program whereby individuals are covered only if they do not have enough income, assets, and insurance to pay for the care, individuals at or close to the Medicaid threshold may prefer to forgo LTC insurance. For these people, care is provided free of charge only in the absence of private insurance. Studies show that 14% of all seniors have Medicaid, and 19% of seniors with LTC needs rely on Medicaid for this care (Banthin & Cohen, 1999; Kaiser Family Foundation, 1999).
For persons with high levels of assets and income, self-insurance is a substitute for LTC coverage. Recognizing the uncertainty of needing LTC, some individuals rely on their own accumulated assets to pay for care if it is needed. Out-of-pocket spending accounted for more than 20% of LTC expenditures in 2002 (Georgetown University, 2004), reflecting both planned and unplanned self-funding.
In addition to using income-related substitutes, persons often perceive informal care from a spouse or one's children as a substitute for formal LTC. Pauly (1990) suggested that the potential for informal care might serve as a disincentive for the purchase of LTC insurance. That is, individuals may not desire LTC insurance because its existence might decrease the amount of informal care provided by children, neighbors, or the community.
Another explanation for the low prevalence of private LTC insurance purchases is that many people erroneously believe they already have LTC coverage. A 2001 survey of adults aged 45 and older revealed that 55% believed that Medicare covered extended nursing home stays for "age-related or other chronic conditions," when in fact it does not (AARP, 2001). In addition, 41% of respondents erroneously believed that Medigap (Medicare Supplemental) insurance covered such care.
Several past studies have looked at the determinants of LTC insurance purchases using a multivariate framework. Sloan and Norton (1997) focused on the role of Medicaid and the influence of self-reported risk factors, expectations, and bequest motives. They found evidence of crowding out by Medicaid and of adverse selection based on the perceived probability of needing LTC. Mellor (2001) tested theories about the role of children as informal caregivers and found no evidence that children serve as a substitute for formal LTC. McCall and colleagues (1998) focused on partnership programs and found that education and knowledge were important factors in encouraging LTC insurance purchase. Although these studies contributed significant new findings about the purchase of private LTC insurance, none of them included price as a determinant of demand for LTC insurance. In the present study, we include price along with other potential determinants of the demand for LTC insurance.
|
METHODS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
). In 2002, HRS survey respondents included 18,167 individuals who were born in 1947 or earlier (aged 55 or older in 2002) along with their spouses. We chose to analyze new purchases separately, because there are major differences between purchases of new policies and renewals of existing policies. First, the annual premium (i.e., the price of carrying LTC insurance) is generally much higher for new purchases than for renewals because, much like individually purchased term life insurance, it is based on the individual's age when the policy begins. Second, the rate of purchase of new policies is lower than the rate of renewal of existing policies. Purchasing a new policy is a rare event: Only 4% of individuals purchase LTC insurance, whereas the majority of existing policy-holders choose to renew. Finally, in our preliminary review of the data for both new purchasers and renewers, we pooled both groups and discovered strong statistical evidence that the demand function varies significantly in its parameter values between the two groups.
Because this study centered on modeling the purchase of new LTC insurance (as opposed to the renewal of coverage), we limited our focus to individuals who reported that they did not have LTC insurance in the period prior to the 2002 HRS survey. We further eliminated any individual who was residing in a nursing home at the time of the study, or who was already unable to perform two activities of daily living, indicating that he or she was already eligible for LTC and would therefore not be a candidate for purchase of a new LTC policy. Finally, we eliminated from the sample persons for whom essential data was missing. This yielded a final analytic sample of 9,863 individuals who were eligible for LTC insurance but did not currently have a policy.
Empirical Model
McKenna's (1986) basic model of health insurance purchase describes a consumer who is maximizing utility across several periods, recognizing that there is some unknown probability that he or she will require medical care in the future. McKenna defines the consumer's expected utility with and without insurance, recognizing that insurance will be purchased as long as expected utility in the insured state is at least as great as expected utility in the uninsured state. Using this framework, we considered factors that were recognized in prior studies regarding LTC insurance (Pauly, 1990; Sloan & Norton, 1997; Zweifel & Strüwe, 1996, 1998) and those that have been recently identified in the marketplace.
Our interest was in modeling purchases of new LTC insurance policies. We pursued this by estimating variations of the following equation:
|
|
i was a random disturbance. We assumed i followed a logistic distribution and estimated the parameters by maximum likelihood. The 2002 HRS survey asked participants about basic health insurance (both private and government). Participants were then asked, "Not including government programs, do you now have any long-term-care insurance which specifically covers nursing home care for a year or more or any part of personal or medical care in your home?" Individuals who answered "Yes," were then asked, "Is that one of the plans you have already described, or a different plan?" For the present study, we recorded the dependent variable, Di, as 1 if an individual replied "Yes" to the first question and then replied "A different plan" to the second question. Approximately 4.4% of the sample fell into this category.
Explanatory Variables
Explanatory variables fell into three general categories: price and potential price variables, individual characteristics, and financial variables. A list of explanatory variables along with expected signs is presented in Table 1.
|
) and term life insurance (Brown & Goolsbee, 2002; Pauly et al., 2003) by assigning prices based on industry practices.
We explored three different price variables in this study, each one based on a different industry source. We based the first, GEC Price, on the rate schedule filed by GE Capital in 1996 with the State of Michigan Office of Financial and Insurance Services (State of Michigan, 1996–1998). We based the second, Hancock Price, on a chart of typical rates for LTC insurance published by the Kiplinger Washington Editors, Inc., in February 1999, which was in turn based on rates filed by the John Hancock Life Insurance Company with the State of New Jersey (Kiplinger Washington Editors, 1999). We based the third, HIAA Price, on national LTC insurance rate averages (Health Insurance Association of America [HIAA], 2002). For all three variables, price depended heavily on the respondent's age and health status, as illustrated in Figure 1. We expected that price would be negatively correlated with the demand for LTC insurance.
|
Demographic variables included gender, marital status, number of children, race, ethnicity, and geographic region of residence. We expected that women would be more likely to purchase LTC insurance because they have a longer life expectancy, implying a higher probability of needing care and of needing it for a longer period of time. We also expected that married persons would purchase LTC insurance, because "impoverishing one's spouse [...] seems to be the major fear of many married elderly" (Pauly, 1990, p. 161). Because informal child-provided care is a possible substitute for formal LTC, we expected that the probability of purchasing LTC insurance that would pay for formal care would be lower among those with children than among those without children (Pauly, 1990; Zweifel & Strüwe, 1998). Race and ethnicity variables were intended to capture cultural differences that might affect the decision to purchase LTC insurance. Age was not included as a variable because, policies being age rated, its main effect was already reflected in the price variable. We measured the respondent's geographic region of residence by using a set of regional dummy variables reflective of the nine census regions in the United States. We tested indicators of urban or rural status in an effort to capture likely proximity to a nursing home or intermediate care facility, but these were not statistically significant and ultimately were not included. It is possible that the comprehensive nature of current policies, which often provide for in-home care, makes proximity to a traditional care facility less crucial.
Information and knowledge variables were also important to the model. There is considerable misperception about LTC coverage, and information variables were intended to reflect the individual's working knowledge of LTC insurance. First, previous experience with LTC insurance was measured with an indicator for whether the person reported ever purchasing LTC insurance (even though they did not carry it in the prior wave). In all, about 9% of the sample reported having purchased LTC insurance at some time in the past. We included these individuals in the sample as newly rated purchasers because they were in a position to be rated according to their current age and health status if they decided to purchase a policy; their decision was no longer a decision to renew existing coverage. It is worth noting that when respondents who had dropped LTC coverage were asked directly why they had allowed their policies to lapse, the most common reason given was that the policy had cost too much.
We also considered the circumstances of the respondent's own parents, because they provide a frame of reference or base of experience for the individual. In the present sample, 6% of respondents reported that at least one of their parents or parents-in-law had an extended illness before they died. In addition, 7% reported having a living parent who currently required regular help, and 4% reported a parent who was now deceased and had lived in a nursing home at some time during their life. In the present model, dummy variables capture these parental circumstances.
Prior studies suggest several reasons for including education in the model (Kenkel, 1991). First, education can potentially improve efficiency of health production. Second, education might coexist with LTC insurance because both are forward-looking behaviors that reflect a low discount rate, a concept also described by Ehrenberg and Smith (2002). Finally, education can improve individuals' knowledge about health. Each of these impacts is consistent with a higher probability of purchasing LTC insurance. In addition, education might serve as a proxy for lifetime earnings, positively impacting the purchase of LTC insurance.
Probability of needing LTC is clearly important in deciding whether or not to insure against that possibility. In the present model, we measured the perceived probability of needing care in three ways. First, we included the individual's self-reported health status. This sign of health status is ambiguous. In one respect, good health can reflect a lower probability of needing care and could therefore enter negatively into the purchase decision. Alternatively, one can perceive good health as increasing life expectancy, reflecting a higher probability of living long enough to require care; in this case, good health could enter positively into the purchase decision. The second variable is the self-reported probability of living at least 10 more years. Again, the expected sign is uncertain, as this variable captures two opposing effects. If the probability of living that much longer is high, it could reflect high life expectancy, and therefore a higher probability of needing LTC. Alternatively, it could mean that the need for LTC is less urgent, and therefore could enter into the purchase decision negatively. The third variable was exercise. This was a dummy variable that equaled 1 if the individual regularly performed vigorous exercise. We expected this variable to negatively impact the probability of purchase, because exercise is perceived as a way to lower the probability of needing LTC.
Wealth was another potentially important determinant of purchase decisions, because the high cost of LTC can quickly erode whatever savings and other assets an uninsured person has. In addition, assets are important in determining Medicaid eligibility. We tested two measures of wealth. The first was total assets (measured in thousands of dollars). The second was a dummy variable that equaled 1 if the individual's assets fell within a "middle range" of $200,000 to $1.5 million (this is the asset range for which Consumer Reports recommends purchase of an LTC insurance policy; Consumer Reports, 2003). We expected that one of these asset measures would increase the probability of purchase.
We also included, in addition to accumulated wealth measures, a measure of current income. We calculated income as household income divided by two if the individual was married or by one if the individual was single. Because there is no benefit to having a lapsed policy, experts usually recommend the purchase of LTC insurance only if premiums are expected to be affordable for the rest of the purchaser's life. We expected that individuals at higher income levels would be better able to afford the premiums and would therefore be more likely to purchase LTC insurance.
Our models also included two potential motivation indicators. The first was a dummy variable that indicated whether the individual itemized deductions on their federal tax returns. Approximately 45% of the sample itemized deductions, indicating that they were in a position to deduct LTC premiums. Although individuals must still meet criteria in order to deduct these expenses, we expected itemizers to have a stronger demand for LTC insurance, because they could potentially benefit from this tax break.
Finally, the models included a dummy variable that reflected the likelihood of leaving a bequest. In the HRS survey, individuals were asked to report the probability that they would leave a bequest in varying amounts. In the present sample, we recorded a value of 1 if the individual reported more than a 50% chance of leaving a bequest of any size. If preservation of the bequest is a motive for purchasing LTC insurance, we would expect this variable to be positively associated with the purchase of LTC insurance.
|
RESULTS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
|
The potential price increase if purchase is delayed was also significantly related to plan purchase. This variable is the difference between what the individual would pay next year as a newly rated subscriber versus what he or she would pay next year as a renewing subscriber. Individuals are more likely to purchase LTC insurance when substantial savings accrue from purchasing it now as opposed to a year from now. A price savings of $100 on next year's annual premium, for example, raised the probability of purchase by 0.01 in the GE Capital model (Table 3). The implied elasticity of purchase with respect to the price savings ranged from 0.29 to 0.69, depending on the model.
Prior purchase of LTC insurance was associated with a 0.028 increase in the probability of purchasing it now, even though there had been a break in coverage (Table 4). This may reflect improved awareness of products and options, or a strong desire for coverage, perhaps purchasing it when financial circumstances make it affordable.
|
; Sloan & Norton, 1997). There are a number of possible explanations for the positive relationship between schooling and the purchase of LTC insurance. These include efficient health production, a low discount rate, and more knowledge about health. In the present model, we had already controlled for knowledge of the market by including variables for prior purchase and parental experience. Therefore, the positive relationship between LTC insurance purchase and education is likely due to efficient production or to a high valuation of future consumption (low discount rate). If it is due to a low discount rate, policies aimed at education are unlikely to have much effect. If, however, it is due to more efficient production, then policies directed at education might be effective in encouraging LTC insurance purchase. We initially introduced marital status as a separate explanatory variable, but we found that the effect of marital status on LTC insurance purchase depended on the level of household assets. Therefore, we ultimately introduced four categories: married with assets in the middle range ($200,000 to $1.5 million), married with assets outside the middle range, single with assets in the middle range, and single with assets outside the middle range. Individuals who were married and who had assets that fell in the middle range were more likely to purchase LTC insurance. We theorized that marital status would have two different impacts. First, married individuals might rely on their spouses as substitutes for formal care, thereby negating the need for LTC insurance. Second, the possibility of "impoverishing one's spouse" may lead some married couples to purchase more LTC insurance. The findings here suggest that the latter incentive dominated when the couple had assets in the vulnerable middle range. Marital status was not a significant factor for individuals with assets outside this range. The presence of children was negative and significant. This supports the claim that older adults expect children to serve as substitutes for formal LTC.
Current income was positive and significant; however, demand was inelastic with respect to income. The positive relationship between income and LTC insurance purchase supports the claim that low-income individuals are likely to decline LTC insurance in order to qualify for Medicaid.
Two additional variables that reflect possible motives for LTC insurance purchase were positive and significant. First, we found evidence that individuals who itemized deductions on their tax returns were significantly more likely to purchase LTC insurance. Individuals who itemize are in a position to deduct a portion of LTC insurance premiums. Second, we found evidence that individuals who believed they would leave a bequest were more likely to purchase LTC insurance.
Of the three variables that reflected probability of needing LTC, only self-reported health status was significant. The positive coefficient on health status suggests that individuals who believe they are healthy may expect to live longer and eventually require care. This effect may be understated, as self-reported health status could be correlated with the objective health measures used to assign price.
Surprisingly, the experiences of the individual's own parents were unrelated to the demand for LTC insurance. Specifically, we found that neither having a parent who currently needed care nor having a parent who had had a lengthy illness before dying had an effect on the demand for coverage. We also included gender, race, and ethnicity in the model, but these were not statistically significant.
|
DISCUSSION |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
We found that considerations other than price were key to whether an individual purchased LTC insurance. The importance of income was borne out, with high income individuals more likely to purchase LTC insurance. The tax deductibility of premiums also influenced decisions to purchase. Individuals who could itemize deductions were more likely to purchase LTC insurance than individuals who could not itemize deductions.
Research has shown that many older adults rely on informal LTC provided by their spouses and children. Some observers have suggested that people may forgo the purchase of LTC insurance because they prefer to receive care from family. In this study, we found that the presence of a spouse does not discourage the purchase of LTC insurance. In fact, married persons were more likely to purchase coverage when their assets fell within the middle range (meaning the couple was unlikely to qualify for Medicaid yet had too few assets to self-insure). This suggests that for married individuals, the desire to protect assets for a spouse outweighs the desire to rely on informal care from a spouse if care is needed. The present findings did, however, support the notion that nonpurchasers may be relying on informal care from children in the future. The presence of children did significantly decrease the likelihood of purchase. Having had one's own parent or in-law experience a serious illness had no bearing on whether an individual purchased LTC insurance.
Education was influential. It is possible that individuals with a high valuation of the future tend to invest in the future through both education and the provision of future LTC. It is also possible that education promotes efficient investment in health. The positive influence of prior LTC insurance purchase suggests that knowledge about the industry encourages purchase.
The present findings suggest several policy approaches for LTC. First, the inelastic nature of both price and potential price increases suggests that, in the present environment, initiatives that rely on lowering policy price are likely to meet with limited success. For example, our findings suggest that even with a 25% price discount, demand would only increase by 11.2%, or from a 4.4% to a 4.9% rate of new purchases per year (based on the Hancock price model). Second, evidence that prior knowledge is influential in purchasing decisions suggests that policies directed at increasing knowledge of the product might be effective. Third, the positive influence of education suggests that provision of tools to efficiently process information about LTC insurance might be effective.
The findings from this research also help explain why initiatives to date have met with limited success. Most of these initiatives were focused on price, yet demand does not respond appreciably to price. The presence of an inelastic demand also suggests that insurers are unlikely to engage in vigorous price competition and may instead focus on benefit enhancements as a way to encourage purchase. Finally, the present findings with regard to education and knowledge echo those of McCall and colleagues (1998). Initiatives intended to increase knowledge of the LTC insurance market may be the most sensible.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
Received for publication February 25, 2005. Accepted for publication July 3, 2005.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
| ||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
