Multiple Dimensions of Private Information: Evidence from the Long-Term Care Insurance Market

A. Theoretical Effect of Asymmetric Information on Market Equilibrium

Standard models of asymmetric information consider individuals who differ only in terms of their (privately known) risk type. A robust prediction of these models is that in equilibrium there will be a positive correlation between the amount of insurance and the occurrence of the risky event (Chiappori and Salanie, 2000; Dionne et al., 2001; Chiappori et al., forthcoming). This observation has motivated the standard test for asymmetric information used in the literature, which is to test for a positive correlation between the amount of insurance coverage and the ex post occurrence of the (potentially) insured risk.²

This “positive correlation” can arise from either adverse selection or moral hazard, both of which result in a market that is inefficient relative to the first best. In the case of adverse selection, the insured is assumed to have ex ante private information about his risk type relative to what the insurance company knows; those with private information that they are high risk select contracts with more insurance than those with private information that they are low risk (see, e.g., Michael Rothschild and Joseph E. Stiglitz, 1976). In the case of moral hazard, the causality is reversed and the informational asymmetry occurs ex post: insurance coverage lowers the cost of an adverse outcome and thus increases the probability or magnitude of the risk occurrence (see, e.g., Richard J. Arnott and Stiglitz, 1988).

Recent theoretical work has enriched these standard asymmetric information models to allow for private information about preferences as well as about risk type. With these multiple forms of private information, a positive correlation between insurance coverage and risk occurrence may be neither a necessary nor sufficient condition for the presence of asymmetric information about risk type (Michael Smart, 2000; Achim Wambach, 2000; David de Meza and David C. Webb, 2001; Bruno Jullien et al., 2002; Chiappori et al., forthcoming). For example, if unobserved preferences are positively correlated with insurance demand and negatively correlated with risk occurrence, then the correlation between insurance coverage and risk occurrence in equilibrium can be negative, despite the presence of asymmetric information about risk type.³

Just as with a single dimension of private information, an equilibrium with multiple forms of private information is unlikely to be efficient relative to the first best. This result applies even absent a positive correlation between insurance coverage and risk occurrence. Intuitively, consider the case in which private information about risk type and about risk preferences offset each other to produce an equilibrium with no correlation between insurance coverage and risk occurrence. In this case, one could well imagine two different groups of individuals purchasing a given insurance policy: low-risk, high-risk-aversion individuals, and high-risk, low-risk-aversion individuals. If these groups pay the same price for the insurance policy, but have different expected costs, both cannot pay an actuarially fair price, and the quantity of insurance purchased by at least one group will therefore not be first best. However, whether the equilibrium is inefficient relative to the second-best, or whether there is scope for Pareto improvement for government intervention, varies across models, just as in the case of unidimensional models of private information (Keith J. Crocker and Arthur Snow, 1985).⁴

B. Long-Term Care Expenditure Risk and Private Insurance

We focus on the long-term care insurance market for two main reasons. First, long-term care expenditure risk is among the largest financial risks faced by today’s elderly, making an investigation of this market of particular interest. Annual expenditures on long-term care in the United States totaled $135 billion in 2004, comprising over 8.5 percent of total health expenditures, or roughly 1.2 percent of GDP (U.S. Congressional Budget Office, 2004). These expenses are distributed unevenly. For example, Christopher M. Murtaugh et al. (1997) estimate that while 60 percent of individuals who reach age 65 will never enter a nursing home, one-fifth of those who do will spend at least five years in the institution.

These figures suggest that there exist potentially large welfare gains from insurance to reduce this expenditure risk. Yet most of the risk is uninsured. Only about 10 percent of individuals age 65 or older have private long-term care insurance, and most of the private policies provide only limited coverage (Health Insurance Association of America (HIAA), 2000b; Jeffrey R. Brown and Finkelstein, 2004a; Marc Cohen, 2003). As a result, only 4 percent of expenditures are reimbursed by private insurers, while one-third are paid for out of pocket (Congressional Budget Office, 2004; National Center for Health Statistics, 2002).

Second, the long-term care insurance market is unusual among insurance markets in the United States in that, during the period of our analysis, there was essentially no direct regulation of the prices charged or policies offered (aNational Association of Insurance Commissioners, 2002a, 2002b; Stephanie Lewis et al., 2003). In a regulated market, it is more difficult to infer which aspects of the equilibrium are inherent to the market itself and which stem from regulatory constraints.⁵

The average age of purchase for long-term care insurance is 67 (HIAA, 2000b). Once purchased, policies are guaranteed renewable for the lifetime of the individual at a prespecified constant nominal premium.

A. Data

Our individual-level survey data are from the Asset and Health Dynamics (AHEAD) cohort of the Health and Retirement Study (HRS). At baseline this cohort is representative of the non-institutionalized population born in 1923 or earlier and their spouses. Appendix A provides more detail on the data and our sample. The average age of our respondents when we observe them in 1995 is 78, and 11 percent have long-term care insurance. These respondents are followed over time, allowing us to observe actual nursing home use from 1995 to 2000. Sixteen percent of our initial community-based sample enters a nursing home at some point during this five-year period.

Central to our analysis is a measure of individual beliefs about nursing home use. We take advantage of a question in AHEAD that directly asks respondents about their perceived likelihood of nursing home use. The specific question we use from the 1995 survey is: “Of course nobody wants to go to a nursing home, but sometimes it becomes necessary. What do you think are the chances that you will move to a nursing home in the next five years?” Individuals are asked to give a response on a scale of zero to 100, which we rescale to be between zero and one. Their responses may reflect information about their health status and thus their chance of needing nursing care. They may also reflect their alternatives to nursing care should they need it, such as the willingness of a spouse or child to take care of them, or their insurance coverage.

On several dimensions the subjective responses to this question appear reasonable. Individual predictions appear to be correct on average; the average self-reported probability of nursing home use over the five-year period is 18 percent, while 16 percent of the responders actually enter a nursing home over the five-year period. We also find that self-reported nursing home entry probabilities covary in sensible ways with known risk factors; they are higher for women than for men, and increase monotonically with age and with deteriorating health status. These results are consistent with other work that has found sensible covariance patterns for self-reported mortality probabilities and characteristics such as the individual’s age or health status (Daniel S. Hamermesh, 1985; V. Kerry Smith et al., 2001; Michael D. Hurd and McGarry, 2002).

One well-known drawback with self-reported probabilities, however, is the propensity of respondents to report round figures such as 0, 50, or 100 percent (Hurd and Mc-Garry, 1995;Li Gan et al., 2005). The preponderance of focal responses suggests that individuals may not be comfortable reporting probabilistic answers, and may not in fact even think in these terms. If individuals use probabilistic information in making insurance purchase decisions, but are unable to translate these latent probabilities into numbers when faced with a survey question, our results are likely to produce underestimates of the extent of the individual’s information.⁶

We supplement the AHEAD data with information obtained directly from insurance companies about the information they collect from applicants and their risk classification practices. We draw on the application forms used by several large long-term care insurance companies which reveal the set of individual characteristics the companies observe, and on the industry’s actuarial model of nursing home utilization, which itself is a function of these observed characteristics.

B. Econometric Approach

The first step of our analysis is to examine the relationship between an individual’s beliefs about his subsequent nursing home utilization, on the one hand, and his actual subsequent nursing home utilization or his current long-term care insurance holdings, on the other. We therefore estimate the following two probits:

CARE is a binary variable for whether the individual went into a nursing home in the five years between 1995 and 2000. LTCINS is a binary variable for whether the individual has long-term care insurance in 1995. The coefficient of interest in each equation is that on B, the individual’s self-reported beliefs (measured in 1995) of his probability of entering a nursing home between 1995 and 2000. X is a vector of covariates to control for the risk classification that would be assigned to the individual by insurance companies in 1995.⁷

The second step of our analysis is to implement the standard positive correlation test for asymmetric information. To do so we employ two approaches used in the literature (the results are quite similar across the two). In one approach, we follow Chiappori and Salanie (2000) and estimate a bivariate probit of insurance coverage and risk occurrence, conditional on the risk classification variables (X). This is equivalent to estimating the probits in (1) and (2) simultaneously, with the beliefs variable (B) omitted from the regression. The key variable of interest is the correlation between the error terms (ρ). A unidimensional model of asymmetric information predicts that residuals will be positively correlated (ρ > 0) and an inability to reject the null hypothesis that ρ = 0 constitutes a failure to reject the null of symmetric information.⁸

Our second approach follows the work of Finkelstein and Poterba (2004) and estimates a probit model of nursing home use as a function of insurance coverage, controlling for risk classification:

The positive correlation prediction is that β₂ > 0. Note that β₂ does not have a causal interpretation. In a pure moral hazard model, the coefficient would represent the causal effect of insurance coverage on care utilization. In an adverse selection model, however, the causality is reversed and private information about expected care utilization affects insurance demand.

In the final step in our empirical work, we provide several examples of additional individual characteristics that are not used in the insurer’s risk classification, but that when added on the right-hand sides of equations (1) and (2), have opposite signed relationships with CARE and with LTCINS.

C. Controlling for Risk Classification

Any analysis of private information, whether the approach we undertake in the first part of our analysis or the standard positive correlation test in our second step, requires that we condition on the risk classification of the individual by insurance companies (X). This conditioning allows us to test for the existence of residual private information, which is the economically meaningful question.

Using insurance applications from five leading long-term care insurance companies, we determined what individual characteristics insurance companies observe. All of them collect a limited set of demographic information—age, gender, marital status, and age of spouse—as well as similar and extremely detailed information on current health and on health history. This same information is observable in AHEAD, which collects extremely rich and detailed information on current health and medical history, as well as standard demographic data. We can therefore accurately replicate the insurer’s information set.

Our preferred approach is to control for the insurance companies’ actuarial prediction of the individual’s risk type, because it is this measure that is used to generate the price the individual is charged. We generated the insurance companies’ prediction of the probability the individual will go into a nursing home over a five-year period using the same actuarial model that is employed by many of the firms in the industry; the model and its pedigree are described in detail in James Robinson (1996), Robinson (2002), and Brown and Finkelstein (2004a).⁹ The predictions depend nonparametrically on the individual’s age, sex, and membership in one of seven different health states (defined by the number of limitations to instrumental activities of daily living (IADLs), the number of limitations to activities of daily living (ADLs), and the presence or absence of cognitive impairments). As noted, all of this information is available in the AHEAD. This measure provides a parsimonious way of controlling for nonlinear (and nonparametric) interactions between the observed characteristics of the individual used by the actuaries in pricing insurance and care utilization.¹⁰

One potential issue with this approach is that we are implicitly assuming that none of the information that is collected by the insurance company, but omitted from the actuarial model, is used in pricing the policies. This assumption appears broadly consistent with insurance company practice. Companies offer age-specific prices with only two or three broad health-based rate classifications within each age (Murtaugh et al., 1995; American Council of Life Insurers, 2001; Weiss, 2002), and do not further adjust premiums over time if the characteristics of the individual change.¹¹ However, we cannot rule-out the possibility that this information is somehow used by the insurance company in determining which of the two or three broad health-based rate classifications to assign the individual to, or whether the company agrees to insure the individual in the first place. Therefore, we also present results from an alternative approach, which we term the “application information” specification. In this specification, we attempt to control for everything insurance companies observe about the individual. We include a full set of single year of age dummies, all of the demographic information that insurance companies collect in their applications (sex, marital status, and age of spouse), and over 35 indicator variables for each of the detailed current health and health history characteristics collected by any insurance company.¹² To be conservative, we also include indicator variables for the household’s income quartile and asset quartile, even though we found only one company that collected any financial information.¹³ Finally, to allow for possible nonlinearities among these various characteristics, we also include a complete set of two-way and three-way interactions between age, sex, and the health variables that are included in the actuarial model (ADLs, IADLs, and cognitive impairment). The “application information” specification thus invokes a more finely defined categorization of risk than insurance companies likely use in pricing.¹⁴

A. Private Information about Risk Type and Its Relation to Insurance Coverage

Table 1 reports the marginal effects from probit estimation of equation (1) of the relationship between beliefs and subsequent nursing home use. The results in column 1 indicate that individual beliefs about the likelihood of entering a nursing home are a statistically significant, positive predictor of subsequent nursing home experience. Our finding of a relationship between risk perception and actual risk of nursing home use complements similar findings for mortality risk (e.g., Smith et al., 2001; Hurd and McGarry, 2002).

The results in column 2 indicate that, perhaps not surprisingly, the insurance companies’ prediction is more highly correlated with subsequent nursing home use than is the individual’s. Comparing columns 1 and 2, we see that a 10-percentage-point increase in the self-reported probability of nursing home use is associated with a 0.9-percentage-point increase in eventual use, whereas a 10-percentage-point increase in the insurer’s prediction is associated with a 4-percentage-point increase in eventual use. This is not, however, the relevant metric for testing for asymmetric information; as long as the individual has residual private information—conditional on the menu of choices offered by the insurance company—asymmetric information can operate as in the theoretical models.¹⁵

Indeed, most importantly, columns 3 and 4 indicate that, controlling for the insurer’s risk classification, individual beliefs remain a positive and statistically significant predictor of subsequent nursing home use. Individuals thus appear to have residual private information about their risk type.¹⁶

Table 2 reports the results from estimating the relationship between beliefs and insurance coverage in equation (2). The results indicate that individuals who believe that they are of higher risk are also more likely to have insurance. By contrast, the insurance company’s prediction of the individual’s risk is negatively related to insurance coverage; this is consistent with the results below that, in fact, on average risk type and insurance coverage are negatively correlated. It also supports our use of the insurance company prediction as a proxy for insurance pricing; conditional on the individual’s risk assessment, a higher insurance company prediction implies a higher price relative to the individual’s perception of an actuarially fair price, and therefore reduces the probability of purchase.

Taken together, the results in Tables 1 and 2 indicate that individuals have residual private information that predicts their risk type and is positively correlated with insurance ownership. This provides direct evidence of asymmetric information. It does not, however, allow us to distinguish between ex ante private information (adverse selection) and ex post private information (moral hazard). Other empirical evidence suggests that demand for nursing home use is relatively price inelastic (David Grabowski and Jonathan Gruber, 2005); we therefore suspect—but cannot directly corroborate—that at least some of what we are detecting reflects ex ante private information.¹⁷

We now turn to an examination of what the results from the standard positive correlation test would suggest about asymmetric information in this market. This test also does not distinguish between adverse selection and moral hazard.

B. Long-Term Care Insurance and Long-Term Care Use

The standard test for residual asymmetric information, based on a positive correlation between insurance coverage and risk occurrence conditional on insurance company risk classification, has been applied across a variety of insurance markets with differing results. In the case of health insurance, David Cutler and Richard Zeckhauser (2000) review an extensive literature that tends to find evidence of this positive correlation. The positive correlation also appears in annuity markets (Finkelstein and Poterba, 2002, 2004; McCarthy and Mitchell, 2003). Several papers, however, find no evidence of a positive correlation in life insurance markets (Cawley and Philipson, 1999; McCarthy and Mitchell, 2003) or in automobile insurance markets (Chiappori and Salanie, 2000; Georges Dionne et al., 2001; and Chiappori et al., forthcoming).

Table 3 shows the results of this standard test in the long-term care insurance market. The top row shows the correlation of the residuals from a bivariate probit of long-term care insurance and nursing home use, as in Chiappori and Salanie (2000). The bottom row shows the marginal effect from probit estimation of nursing home use on long-term care insurance (equation (3)), as in Finkelstein and Poterba (2004). Both approaches yield the same findings. With no controls for the insurers’ information set, the relationship between coverage and risk occurrence is negative and statistically significant. This finding is consistent with other aggregate data on relative rates of nursing home use for the insured and general population (Society of Actuaries, 2002). When we control for the insurance companies’ risk classification in columns 2 and 3, as required by the positive correlation test, we are unable to reject the null hypothesis of zero correlation.

Proper implementation of the positive correlation test requires that we examine insurance demand among individuals who face the same set of possible insurance contracts. The controls for risk classification—and hence the premium the insurer would offer the individual—represent an important component of this choice set. They may not, however, fully capture all determinants of the individual’s choice set. We therefore use two approaches to verify that our failure to reject the null hypothesis of symmetric information with the positive correlation test is not an artifact of our inability to control completely for the options offered to the individuals.

First, we implement the positive correlation test in a more homogenous subsample of individuals who are likely to face the same option set. We identified two additional characteristics of the individual that are particularly likely to affect his effective choice set: his wealth, and the presence of certain adverse health conditions.¹⁸ The public long-term care insurance provided by Medicaid offers a substitute for private insurance; indeed, Brown and Finkelstein (2004b) estimate that Medicaid may have substantial crowd-out effects on demand for private long-term care insurance. However, because Medicaid requires that individuals exhaust virtually all of their financial assets before it will cover long-term care expenses, Medicaid is a substantially better option for lower-wealth individuals. In addition, individuals with some observably very poor health conditions are often denied insurance coverage, at least by the large long-term care insurance companies (Murtaugh et al., 1995; Weiss, 2002).¹⁹ For such individuals, simply controlling for the price they would face, if they were to be offered coverage, may produce misleading results. We therefore re-implement the positive correlation test limiting our analysis to the 20 percent of individuals who are in the top quartile of the wealth distribution and who have none of the health conditions that could trigger a denial.²⁰ Consistent with the better option set available to the healthier and wealthier subsample, 17 percent of our selected sample have long-term care insurance, compared to only 11 percent of the total sample. Table 4 shows the results from implementing the positive correlation test on this much more homogenous subsample. There is again no evidence of a positive correlation between coverage and nursing home use. Indeed, there is now statistically significant evidence of a negative correlation.²¹

Second, to identify more precisely the options that individuals face and the policies they choose, much of the existing literature testing for asymmetric information has relied on proprietary data provided by insurance companies themselves. These data typically include the individual’s risk classification as assigned by the company, details about the policy he has selected from the menu of available policies, and his ex post risk experience (see, e.g., Chiappori and Salanie, 2000; or Finkelstein and Poterba, 2004). To implement this alternative approach, we obtained proprietary data from a large long-term care insurance company which contain all of the information used in similar analyses, including the individual’s choice from the menu of contracts, his exact risk classification, and his ex post risk experience.²² This analysis complements the analysis with the AHEAD data because it allows us to examine the relationship between the quantity of insurance coverage (conditional on having insurance) and risk occurrence. Using these data, we once again fail to reject the null hypothesis of no positive correlation. The data and results from this exercise are described more fully in Appendix B.

We also undertook numerous additional tests of robustness in the AHEAD data, many of which we present in detail in the working paper version (Finkelstein and McGarry, 2003). For example, we verified that the positive correlation does not manifest itself in other measures of care utilization such as the intensity of care use (i.e., number of nights in a nursing home), or home health care use. We also verified that the positive correlation does not emerge if the relationship between insurance coverage and risk occurrence is analyzed over a longer time horizon than the five-year period studied here. Finally, although policies once purchased are guaranteed renewable for life, some individuals stop paying their premiums and thereby forfeit some or all of their potential future nursing home benefits; we therefore verified that the results are unaffected by excluding from the sample the 10 percent of insured individuals in 1995 who subsequently report having dropped their insurance coverage.²³

The results presented in Tables 1 and 2 point to the presence of asymmetric information, even though the results in Tables 3 and 4 indicate that the standard positive correlation test is unable to reject the null of symmetric information. This suggests that our beliefs-based test may be a more discerning test for asymmetric information than the standard positive correlation test. Moreover, as noted previously, because our beliefs measure is a highly imperfect proxy for an individual’s private information, our findings in Tables 1 and 2 likely understate the amount of private information in this market.

Nonetheless, a natural question is whether the private information we detected in Tables 1 and 2 is sufficiently large that we should have expected to obtain a positive correlation between insurance coverage and long-term care use in the absence of any offsetting selection. A simple decomposition suggests that this is indeed the case. Specifically, we used the results from estimating equation (2) to decompose long-term care insurance coverage into the portion explained by individual beliefs and a residual that is not explained by these beliefs. We then reestimated the nursing home utilization equation (3) substituting these two components of long-term care insurance for the LTCINS variable. In doing so we found a positive and statistically significant relationship between nursing home utilization and the portion of insurance coverage predicted by individual beliefs, with a coefficient of around 0.5 when controls for risk classification are included. This suggests that the relationship between nursing home utilization and insurance coverage would indeed be positive in the absence of any offsetting selection effects. The next section presents more direct evidence of the existence of these offsetting selection effects.

C. Private Information about Preferences for Insurance

Our finding that insured individuals are no more likely to enter a nursing home than individuals without insurance might ostensibly seem at odds with our finding that individuals have private information about their risk type that is positively correlated with insurance coverage. A natural reconciliation of these two sets of results, however, is the existence of other (unobserved) characteristics of the individual that are positively correlated with insurance coverage but negatively related to insurance use. These factors must be items that are omitted from the pricing formulas used by insurance companies and that have the opposite correlation with insurance coverage and care utilization. We now turn to an empirical examination of what some of these offsetting preference-based factors might be.

It is worth noting at the outset that if individuals’ self-reported beliefs are a sufficient statistic for their private information and if individuals efficiently incorporate all available information when forming their beliefs, then conditional on these beliefs other characteristics of the individual should have no predictive power in explaining subsequent utilization. In practice, however, our belief measures are likely to be an imperfect proxy for individuals’ beliefs. Moreover, evidence from Smith et al.’s (2001) study of belief formation about mortality prospects suggests that in fact individuals do not efficiently incorporate all available information in forming and updating their beliefs.

We focus on two potential dimensions of offsetting preference-based selection—wealth and cautiousness—which have previously attracted theoretical attention (de Meza and Webb, 2001; Bruno Jullien et al., 2002). Neither of these factors is used by insurance companies in pricing long-term care insurance.

As we noted earlier, Medicaid offers a substantially better substitute for private insurance for lower-wealth individuals. This makes it likely that an individual’s wealth—a factor not used in pricing private policies—may be an important source of preference heterogeneity in private insurance demand. The top panel of Table 5 reports the results of adding indicator variables for the individual’s financial net wealth quartile to equations (1) and (2). Individuals in higher wealth quartiles are less likely to go into a nursing home (odd columns) but more likely to have long-term care insurance (even columns).

We measure an individual’s cautiousness by his investment in risk-reducing activities. We observe, in 1995, whether the individual undertook various gender-appropriate preventive health care measures in the previous two years. These activities are: whether the individual had a flu shot, had a blood test for cholesterol, checked her breasts for lumps monthly, had a mammogram or breast x-ray, had a Pap smear, and had a prostate screen. The median individual undertakes two-thirds of gender-relevant activities; 7 percent report doing nothing and 30 percent report engaging in all relevant activities. The insurance company applications we reviewed did not solicit any of this information.

Individuals who invest more in such preventive activities may also be more risk averse and thus place a higher value on insurance (de Meza and Webb, 2001). However, more cautious individuals are not necessarily more risk averse; preventive activity affects the mean as well as the variance of the risk distribution (see, e.g., Dionne and Eeckhoudt, 1985, and Jullien et al., 1999). Thus the sign of the correlation between cautious behavior and insurance coverage is an empirical question. The results in Table 5, panel B, indicate that individuals who undertake a greater fraction of potential preventive health activities (i.e., more cautious individuals) are in fact more likely to own insurance; they are also less likely to enter a nursing home.²⁴

One interpretation of the negative relationship between preventive health behaviors and nursing home use is that the preventive behaviors endogenously lower the individual’s risk type by forestalling or preventing nursing home admissions. For example, flu shots reduce the risk of pneumonia which is a nontrivial contributor to nursing home use among the elderly. We also find, however, that individuals who invest more in our measured preventive health activities are substantially less likely to have a hip fracture (another important contributor to nursing home use), yet none of the measured activities themselves would be expected to affect bone density or agility. We therefore suspect that these preventive health activities are correlated with other health investments that themselves cause lower rates of institutionalization. It is also possible that these preventive health activities proxy for other unmeasured preference-related characteristics that themselves have a causal effect on nursing home utilization. We find, for example, that individuals who engage in more preventive health activity tend systematically to overestimate their risk probability relative to actual experience. This suggests the existence of an unobserved “pessimism” factor; Kostas Koufopoulos (2003) conjectures that pessimism simultaneously increases demand for insurance and demand for risk-reducing activities, which lower risk type.²⁵

One concern with our findings, however, is that preventive health measures may themselves be determined by insurance coverage or health status. We suspect this is unlikely, as everyone in our sample is covered by Medicare (which covers the expenses from these preventive health measures at least to some extent), virtually our entire sample of elderly individuals has seen a doctor in the last two years, and our analysis includes detailed controls for health status. As a further check, however, we turn to a different type of precautionary behavior that is less likely to be affected by health status or health insurance coverage: seat belt use. The results in panel C of Table 5 indicate that individuals who report that they always wear their seat belt in a car (which 77 percent do) are also less likely to go into a nursing home and more likely to own insurance.²⁶

Interestingly, other than preventive activity and wealth, the characteristics of the individual that we can measure and that insurance companies do not use in pricing appear to have the same correlation with insurance coverage and risk occurrence. For example, individuals with more children are both less likely to have insurance and less likely to use nursing home care. The same is true for non-whites relative to whites, Hispanics relative to non-Hispanics, and less educated individuals relative to more educated.

Construction of Some of the Health Measures Collected by Insurance Companies

Appendix Table B.

Hazard of Receiving Nursing Home Care for Hundredth Consecutive Day

1. American Council of Life Insurers. Life insurers fact book, 2001. Washington, DC: American Council of Life Insurers; 2001. [Google Scholar]
2. Arnott Richard J, Stiglitz Joseph E. The Basic Analytics of Moral Hazard. Scandinavian Journal of Economics. 1988;90(3):383–413. [Google Scholar]
3. Brown Jeffrey R, Finkelstein Amy. NBER Working Papers: No. 10989. National Bureau of Economic Research, Inc; 2004a. The Interaction of Public and Private Insurance: Medicaid and the Long-Term Care Insurance Market. [Google Scholar]
4. Brown Jeffrey R, Finkelstein Amy. NBER Working Papers: No. 10782. National Bureau of Economic Research, Inc; 2004b. Supply or Demand: Why Is the Market for Long-Term Care Insurance So Small? [Google Scholar]
5. Cardon James H, Hendel Igal. Asymmetric Information in Health Insurance: Evidence from the National Medical Expenditure Survey. RAND Journal of Economics. 2001;32(3):408–27. [PubMed] [Google Scholar]
6. Cawley John, Philipson Tomas. An Empirical Examination of Information Barriers to Trade in Insurance. American Economic Review. 1999;89(4):827–46. [Google Scholar]
7. Chassagnon Arnold, Villeneuve Bertrand. Optimal Risk-Sharing under Adverse Selection and Subjective Risk Perception. Canadian Journal of Economics. 2005;38(3):955–78. [Google Scholar]
8. Chiappori Pierre-André. Econometric Models of Insurance under Asymmetric Information. In: Dionne Georges., editor. Handbook of insurance. London: Kluwer Academic; 2000. pp. 365–93. [Google Scholar]
9. Chiappori Pierre-André, Salanie Bernard. Testing for Asymmetric Information in Insurance Markets. Journal of Political Economy. 2000;108(1):56–78. [Google Scholar]
10. Chiappori Pierre-André, Jullien Bruno, Salanie Bernard, Salanie François. Asymmetric Information in Insurance: General Testable Implications. RAND Journal of Economics. (forthcoming) [Google Scholar]
11. Cohen Alma, Einav Liran. NBER Working Papers: No. 11461. National Bureau of Economic Research, Inc; 2005. Estimating Risk Preferences from Deductible Choice. [Google Scholar]
12. Cohen Marc. Private Long-Term Care Insurance: A Look Ahead. Journal of Aging and Health. 2003;15(1):74–98. doi: 10.1177/0898264302239015. [DOI] [PubMed] [Google Scholar]
13. Cox David R. Regression Models and Life-Tables (with Discussion) Journal of the Royal Statistical Society B (Methodological) 1972;34(2):187–220. [Google Scholar]
14. Cox David R. Partial Likelihood. Biometrika. 1975;62(2):269–76. [Google Scholar]
15. Crocker Keith J, Snow Arthur. The Efficiency of Competitive Equilibria in Insurance Markets with Asymmetric Information. Journal of Public Economics. 1985;26(2):207–19. [Google Scholar]
16. Cutler David M, Zeckhauser Richard J. The Anatomy of Health Insurance. In: Culyer Anthony J, Newhouse Joseph P., editors. Handbook of health economics. 1A. Amsterdam: Elsevier Science North-Holland; 2000. pp. 563–643. [Google Scholar]
17. de Meza David, Webb David C. Advantageous Selection in Insurance Markets. RAND Journal of Economics. 2001;32(2):249–62. [Google Scholar]
18. Dionne Georges, Eeckhoudt Louis. Self-Insurance, Self-Protection, and Increased Risk Aversion. Economic Letters. 1985;17(1–2):39–42. [Google Scholar]
19. Dionne Georges, Gourieroux Christian, Vanasse Charles. Testing for Evidence of Adverse Selection in the Automobile Insurance Market: A Comment. Journal of Political Economy. 2001;109(2):444–53. [Google Scholar]
20. Finkelstein Amy, McGarry Kathleen. NBER Working Papers: No. 9957. National Bureau of Economic Research, Inc; 2003. Private Information and Its Effect on Market Equilibrium: New Evidence from Long-Term Care Insurance. [Google Scholar]
21. Finkelstein Amy, McGarry Kathleen, Sufi Amir. Dynamic Inefficiencies in Insurance Markets: Evidence from Long-Term Care Insurance. American Economic Review; Papers and Proceedings; 2005. pp. 224–28. [DOI] [PubMed] [Google Scholar]
22. Finkelstein Amy, Poterba James M. Selection Effects in the United Kingdom Individual Annuities Market. Economic Journal. 2002;112(476):28–50. [Google Scholar]
23. Finkelstein Amy, Poterba James. Adverse Selection in Insurance Markets: Policyholder Evidence from the U.K. Annuity Market. Journal of Political Economy. 2004;112(1):183–208. [Google Scholar]
24. Gan Li, Hurd Michael, McFadden Daniel. Individual Subjective Survival Curves. In: Wise David., editor. Analysis in economics of aging. Chicago: University of Chicago Press; 2005. pp. 377–411. [Google Scholar]
25. Grabowski David C, Gruber Jonathan. NBER Working Papers: No. 11723. National Bureau of Economic Research, Inc; 2005. Moral Hazard in Nursing Home Use. [Google Scholar]
26. Hamermesh Daniel S. Expectations, Life Expectancy, and Economic Behavior. Quarterly Journal of Economics. 1985;100(2):389–408. [Google Scholar]
27. Health Insurance Association of America. LTC Insurance in 1997–1998. Washington, DC: Health Insurance Association of America; 2000a. [Google Scholar]
28. Health Insurance Association of America. Decade study of buyers and non buyers. Washington, DC: Health Insurance Association of America; 2000b. Who buys LTC insurance in 2000? [Google Scholar]
29. Health Insurance Association of America. Who buys long-term care insurance in the work-place? Washington, DC: Health Insurance Association of America; 2001. [Google Scholar]
30. Hurd Michael D, McGarry Kathleen. Evaluations of the Subjective Probability of Survival in the Health and Retirement Survey. Journal of Human Resources. 1995;30(5):S268–92. [Google Scholar]
31. Hurd Michael D, McGarry Kathleen. The Predictive Validity of Subjective Probabilities of Survival. Economic Journal. 2002;112(482):966–85. [Google Scholar]
32. Jeleva Meglena, Villeneuve Bertrand. Insurance Contracts with Imprecise Probabilities and Adverse Selection. Economic Theory. 2004;23(4):777–94. [Google Scholar]
33. Jullien Bruno, Salanie Bernard, Salanie François. Should More Risk-Averse Agents Exert More Effort? Geneva Papers on Risk and Insurance Theory. 1999;24(1):19–28. [Google Scholar]
34. Jullien Bruno, Salanie François, Salanie Bernard. Screening Risk Averse Agents under Moral Hazard. Unpublished Paper. 2002 [Google Scholar]
35. Koufopoulos Kostas. Asymmetric Information, Heterogeneity in Risk Perceptions, and Insurance: An Explanation to a Puzzle. Unpublished Paper. 2003 [Google Scholar]
36. Lewis Stephanie, Wilkin John, Merlis Mark. Publication No. 6073. Menlo Park, CA: Kaiser Family Foundation; 2003. Regulation of private long-term care insurance: Implementation experience and key issues. [Google Scholar]
37. Life Insurance and Market Research Association. Individual medicare supplement and long-term care insurance. Windsor, CT: LIMRA International, Inc; 2001. [Google Scholar]
38. McCarthy David, Mitchell Olivia S. NBER Working Papers: No. 9975. National Bureau of Economic Research, Inc; 2003. International Adverse Selection in Life Insurance and Annuities. [Google Scholar]
39. Mehta Kala M, Yaffe Kristine, Langa Kenneth M, Sands Laura, Whooley Mary A, Covinsky Kenneth E. Additive Effects of Cognitive Function and Depressive Symptoms of Mortality in Older Community Living Adults. Unpublished Paper. 2002 doi: 10.1093/gerona/58.5.m461. [DOI] [PMC free article] [PubMed] [Google Scholar]
40. Mellor Jennifer M. Long-Term Care and Nursing Home Coverage: Are Adult Children Substitutes for Insurance Policies? Journal of Health Economics. 2001;20(4):527–47. doi: 10.1016/s0167-6296(01)00078-9. [DOI] [PubMed] [Google Scholar]
41. Murtaugh Christopher M, Kemper Peter, Spillman Brenda C. Risky Business: Long-Term Care Insurance Underwriting. Inquiry. 1995;32(3):271–84. [PubMed] [Google Scholar]
42. Murtaugh Christopher M, Kemper Peter, Spillman Brenda C, Carlson Barbara L. Amount, Distribution, and Timing of Lifetime Nursing Home Use. Medical Care. 1997;35(3):204–18. doi: 10.1097/00005650-199703000-00003. [DOI] [PubMed] [Google Scholar]
43. National Association of Insurance Commissioners. Executive summary of amendments to the long-term care insurance model regulation. Kansas City, MO: National Association of Insurance Commissioners; 2002a. [Google Scholar]
44. National Association of Insurance Commissioners. State survey of long-term care insurance rating practices and consumer disclosure amendments. Kansas City, MO: National Association of Insurance Commissioners; 2002b. [Google Scholar]
45. National Center for Health Statistics. Health, United States, 2002 with chartbook on trends in the health of Americans. Washington, DC: U.S. Government Printing Office; 2002. [Google Scholar]
46. Robinson James. A Long-Term Care Status Transition Model. Presented at Bowles Symposium on the Old-Age Crisis: Actuarial Opportunities; September 26–27, 1996.Georgia State University; [Google Scholar]
47. Robinson James. A Long-Term Care Status Transition Model. Unpublished Paper. 2002 [Google Scholar]
48. Rothschild Michael, Stiglitz Joseph E. Equilibrium in Competitive Insurance Markets: An Essay on the Economics of Imperfect Information. Quarterly Journal of Economics. 1976;90(4):630–49. [Google Scholar]
49. Sloan Frank A, Norton Edward C. Adverse Selection, Bequests, Crowding out, and Private Demand for Insurance: Evidence from the Long-Term Care Insurance Market. Journal of Risk and Uncertainty. 1997;15(3):201–19. [Google Scholar]
50. Smart Michael. Competitive Insurance Markets with Two Unobservables. International Economic Review. 2000;41(1):153–69. [Google Scholar]
51. Smith V Kerry, Taylor Donald H, Jr, Sloan Frank A. Longevity Expectations and Death: Can People Predict Their Own Demise? American Economic Review. 2001;91(4):1126–34. [Google Scholar]
52. Society of Actuaries. Society of Actuaries, Transactions: 1988 –89 –90 Reports of mortality, morbidity and other experience. Schaumburg, IL: Society of Actuaries; 1992. Report of the Long-Term-Care Experience Committee: 1985 National Nursing Home Survey Utilization Data; pp. 101–64. [Google Scholar]
53. Society of Actuaries. Long-term care experience committee intercompany study 1984–1999. Schaumburg, IL: Society of Actuaries; 2002. [Google Scholar]
54. U.S. Congressional Budget Office. Financing long-term care for the elderly. Washington, DC: U.S. Government Printing Office; 2004. [Google Scholar]
55. Wambach Achim. Introducing Heterogeneity in the Rothschild-Stiglitz Model. Journal of Risk and Insurance. 2000;67(4):579–91. [Google Scholar]
56. Weiss Ratings, Inc. Weiss ratings’ consumer guide to LTC insurance. Jupiter, FL: Weiss Ratings, Inc; 2002. [Google Scholar]