Implications for U.S. Health Insurers of Future Growth in the Age 65+ Population

Transcription

1 Implications for U.S. Health Insurers of Future Growth in the Age 65+ Population Dr. Etti G. Baranoff, FLMI Associate Professor of Insurance and Finance Department of Finance, Insurance and Real Estate School of Business Virginia Commonwealth University Snead Hall, 301 West Main Str., Suite B4167 Richmond VA Tel: , Fax: Cell: or & Research Director, Insurance and Finance The Geneva Association Thomas W. Sager Professor of Statistics Department of Information, Risk, and Operations Management The University of Texas at Austin, CBA Austin, Texas (512) Bo Shi Assistant Professor of Finance School of Business and Public Affairs Morehead State University Morehead, Kentucky (606)

2 Implications for U.S. Health Insurers of Future Growth in the Age 65+ Population Abstract This paper explores potential impacts of growth in the 65+ age cohort on the U.S. health insurance industry. We assess impacts on insurers cost of coverage, capital, profitability, and appetite for risk. We use historical data from to identify relevant trends and to understand the relationships of these trends to insurers cost of coverage, capital, profitability, and risk. Medicare Supplement insurance provides the most cleanly accessible data for assessing the impact of these trends on insurers. The methodology is a 2-stage simultaneous equations model to allow for interacting main variables. With that understanding as a base, we forecast the future of health insurers cost of coverage, capital, profitability, and risk appetite by incorporating projected growth in the number, longevity, and medical utilization of the 65+ age cohort. We find mostly manageable impacts. -2-

3 Implications for U.S. Health Insurers of Future Growth in the Age 65+ Population I. Introduction For many years in the U.S., the cost of health care has been rising much more rapidly than the rate of inflation. According to the National Health Expenditure (NHE) projection, annual U.S. healthcare expenditure is expected to grow to $4.4 trillion by 2018, from $2.34 trillion in Moreover, in 2011, the first year of the Baby Boom cohort reached age 65. For the next 20 years, the over-65 population is projected to grow by about 16% every five years. 2 Not only will its numbers grow, but also its longevity. However, as longevity lengthens, the over-65 population may become relatively sicker, especially if medical advances continue to transform formerly fatal illnesses into chronic conditions. A sicker population most likely will use medical services more intensely. The implications of these trends for U.S. society have received more attention than the implications for U.S. health insurers. In the U.S. health care system, private health insurance companies generally occupy intermediary positions between patients and providers. Health insurers collect health insurance premiums from individuals and/or their employers and funnel payments to doctors, hospitals, and other providers. The stress upon this system from an aging population is about to intensify. How will the stress impact U.S. health insurers? In this study, we investigate this question. We find that although there are some potentially troubling implications of over-65 trends, health insurers should be able to adapt. 1 Klees, B. S., Wolfe, C. J. and Curtis, C. A. (2009) Brief Summaries of Medicare & Medicaid, Title XVIII and Title XIX of the Social Security Act, Health Care Financing Review/2009, Center for Medicare & Medicaid Services. 2 U.S. Census Bureau Population Projection at: This URL got cut off. I just started it we never got it Bo?} -3-

4 For the most part, the health care needs of the over-65 U.S. population are covered by a government program, Medicare, which pays for comprehensive health coverage. Medicare is funded largely by a tax that workers and their employers pay. 3 At age 65, most U.S. workers become eligible for Medicare. Thus, there is a sharp demarcation between the over-65 age cohort and the under-65 age cohort in the role that health insurers play in the U.S. health care system. For those over age 65, the role of the government expands and that of health insurers contracts. Medicare covers hospitalization (Part A) and doctor visits (Part B). Parts A and B are comprehensive and include catastrophic coverage, but there are limits. For example, Part A limits the number of days of hospitalization coverage; Part B limits payments for diagnostic procedures. For costs that exceed limits, Medicare participants incur out-of-pocket expenses, copayments and deductibles. 4 However, since Part A and B are paid by the government, health insurers have no at-risk involvement. Therefore, trends in the over-65 age cohort will not affect health insurers financially through Parts A and B. But there are additional Medicare and Medicare-related plans in which health insurers may have some risk and through which insurers may be affected by trends in the over-65 population. However, we shall see that the at-risk involvement of private health insurers with the over-65 population is limited in its nature. The two principal ways that health insurers incur financial exposure to trends in the over-65 population are Medicare Supplement and Medicare Part C. 3 Medicare has two central parts A and B. During their working lives, employees and employers each pay 1.45% of the total employee salary for Medicare Part A (covering hospital services only). Thus, the combined tax is 2.9% of salaries. Eligibility for benefits under Part A begins at age 65, at which time beneficiaries may add Part B (covering services by doctors and other providers) by paying a modest premium that is subsidized by the government. In addition, Medicare A and B participants have a variety of optional Medicare and Medicare-related plans that they may buy at extra cost. Participants retain comprehensive A and B coverage under all optional plans, which may add drug, vision, or dental coverage, among other things. These additional options are discussed more fully in the Appendices. For this paper, two of them Medicare Supplement (also called Medigap) and Medicare Part C (also called Medicare Advantage) are particularly relevant and receive attention in the text. 4 In the early 2000s, Medicare covered 45 percent of the individual health care expenditures of the over-65 U.S. population, according to the Kaiser Family Foundation (2005) -4-

5 Participants in Parts A and B of Medicare may elect optional supplemental health insurance sold by private health insurers - called Medicare Supplement insurance or Medigap (described in Appendix A). Medicare Supplement insurance pays some of the out-of-pocket medical expenses not covered by Parts A and B. For example, Medicare Supplement may cover more days of hospitalization than Part A and may cover certain doctor procedures more generously than Part B, so that out-of-pocket expenses are lower. However, Medicare Supplement insurance is separate from Parts A and B and does not duplicate the A and B coverage. Medicare Supplement policies are highly regulated products only ten are allowed in the market (see Appendix A). 5 Moreover, the extra benefits that Medicare Supplement provides are relatively small in comparison with the comprehensive coverage of Medicare Parts A and B and do not include catastrophic coverage. However, the important point about Medicare Supplement policies for our purposes is that health insurers are financially at risk. Therefore, trends in the over-65 population may affect the financial indicators of health insurers through involvement with Medicare Supplement. Thus, in this study, we analyze the effects of over-65 trends on Medicare Supplement insurers. A second form of involvement of health insurers with the over-65 age cohort is Part C of Medicare also called Medicare Advantage. This form essentially bundles the basic Part A and B of Medicare with additional medical coverages. Although the structure of Part C protects the financial interests of health insurers, Part C insurers are partly at risk. However, when health insurers report to the NAIC, they do not unbundle their data so that the at-risk portion of Part C may be separated from the not-at-risk portion. Therefore, we do not include Part C within the 5 See Choosing a Medigap Policy: A Guide to Health Insurance for People with Medicare developed jointly by the Centers for Medicare & Medicaid Services (CMS) and the National Association of Insurance Commissioners (NAIC). -5-

6 scope of our analysis. However, we shall see that our analysis of Medicare Supplement will allow us to infer bounds for Part C effects on health insurers. The most significant trends in the over-65 population are expected to be an increase in numbers, in longevity, and in utilization of healthcare services. We expect that the primary impact of increasing numbers, increasing longevity, and increasing medical utilization among the over-65 population will be felt through utilization. As numbers increase, health insurers may have more Medicare Supplement customers a development that insurers may welcome rather than dread but if relative utilization patterns do not change, then financial outcomes will remain proportionate, subject to economies or diseconomies of scale. As lifespans lenthen, people will remain in the over-65 cohort longer, thus adding to their number. Longevity gains could be positive, negative, or neutral for utilization. It depends upon health. We think that a likely outcome of longer lives may be that people will remain sick longer. This may shift the distribution of over-65 health toward a sicker state as medical science enables more people to linger in impaired states who previously would have succumbed. Although the literature has not studied directly the potential impact of the growing over- 65 population on health insurers capital and risk behavior, there are studies that looked specifically at Medicare Supplement. Some studies conclude that higher risk people use more Medicare Supplement. Wolfe and Goddeeris (1991) and Ettner (1997) and Hurd and McGarry (1997) empirically show the existence of adverse selection in Medicare Supplement insurance. Cutler, Finkestein, and McGarry (2008) find the opposite in the Medicare Supplement insurance market and call it advantageous selection. Fang, Keane, and Silverman (2008) attribute advantageous selection in the Medicare Supplement market to higher income, education and cognitive ability. Additional literature discusses the interaction between Medicare and Medicare Supplement insurance. Buchmueller, Couffinhal, Grignon, and Perronnin (2002) find that -6-

7 individuals with supplemental coverage have substantially more physician visits under the French health care system, which resembles the U.S. Medicare and Medicare Supplement relationship. Finkelstein (2004) finds that Medicare doesn t affect the coverage in residual private insurance market. Maestas, Schroeder, and Goldman (2009) try to explain wild price variation on the standard Medicare Supplement products after OBRA (1990). 6 Unlike prior work, this paper focuses on the impact on financial characteristics of health insurers as the over-65 population grows and increases its utilization of the health system. More specifically, we create a model that allows us to gauge the possible impact of the trends in utilization upon major financial measures for health insurers their cost of coverage (price), capital, profitability, and investment risk. For reasons discussed above, we reduce the relevant over-65 trends to medical service utilization and, more specifically, to Medicare Supplement utilization. Our models allow us to compare the effects of over-65 utilization to the impact of below-65 utilization that results from a working population that is covered under private comprehensive health insurance. In brief, our methodology will be to produce, first, historical trends in utilization, second, projections of those historical utilization trends into the future, third, a statistical model for the impact of utilization upon pricing, capital, profitability, and investment risk, and finally, estimates of future impacts from applying our statistical model to utilization projections. We use the National Association of Insurance Commissioners (NAIC) health insurers annual filing dataset from 2001 to This is a comprehensive panel dataset, which provides very rich data of health insurers. 6 Budget Reconciliation Act of 1990 (OBRA 1990) is effective in This act standardizes the Medicare Supplement insurance products into A-J ten types of contracts as shown in Appendix A. This is to help Medicare beneficiaries select the right residual insurance coverage. -7-

8 The paper is structured as follows: Section II describes the data and profiles Medicare Supplement insurers as a subset of health care insurers. Section III provides the methodology and results. II. Data and Profile of Medicare Supplement Insurers Endogenous Variables Predetermined or Exogenous Variables Variables Table 1. Variable Definitions Description Medicare Supplement Total Medical Encounters / Cost of Coverage Medicare Supplement Total Premiums / Medicare Supplement Members Capital Capital / Total Assets Health Underwriting Gain Health Underwriting Gain / Health Premium Income Investment Risk Opportunity Asset Risk 65+ Per Premium Medical Encounters 1 Medicare Supplement Premium Income 65+ Per Premium Inpatient Services 2 Medicare Supplement Loss Ratio 65- Per Premium Medical Encounters Per Premium Inpatient Services Population Loss Ratio Size LogRBCratio Return on capital States of licensure Medicare Supplement Inpatient Services / Medicare Supplement Premium Income Medicare Supplement Underwriting Deduct / Medicare Supplement Premium Income (Comprehensive Total Medical Encounters + Federal Employee Health Benefits Plans Total Medical Encounters)/(Comprehensive Premium Income + Federal Employee Health Benefits Plans Premium Income) (Comprehensive Inpatient Services + Federal Employee Health Benefits Plans Inpatient Services)/(Comprehensive Premium Income + Federal Employee Health Benefits Plans Premium Income) (Comprehensive Underwriting Deduct + Federal Employee Health Benefits Plans Underwriting Deduct) / (Comprehensive Premium Income + Federal Employee Health Benefits Plans Premium Income) Geometric Mean of (Total Assets, Total Liabilities, Total Writings) Logarithm of the risk-based capital Return on Capital = Total Income / Total Book Capital Number of States of Licensure In group? In Affiliated Group (1), Not in Affiliated Group (0) Stock insurer? Stock Firm (1), Non-Stock Firm (0) Use derivatives? Indicator of Derivative Activity (1=Yes) 1 Medical encounters include both physician and non-physician encounters. 2 Inpatient services are the sum of inpatient days and inpatient admissions. -8-

9 Table 2. Summary Statistics* Variable N Mean S.D. N Mean S.D. Capital Ratio Health Underwriting Gain Investment Risk Per Premium Medical Encounters Per Premium Inpatient Services Medicare Supplement Loss Ratio Per Premium Medical Encounters Per Premium Inpatient Services Population Loss Ratio Size logrbcratio Return on Capital States of Licensure Stock insurers (1/0) In group (1/0) Use derivatives (1/0) * Our analysis focuses on health insurers active in Medicare Supplement product identified by positive Medicare Supplement premium income. There are 778 firm-years of records in the dataset, 2 firm-years are deleted since they experienced capital ratio below 0 or above 1. Table 3 and Table 4 profile key facts about the health insurance industry, as well as health insurers active in Medicare Supplement insurance for 2004 and For example, we see that Medicare Supplement insurers are large health insurers, as a group, with nearly half of industry premiums, 60% of industry assets, and more than one-third of covered members. Moreover, their mean premiums and assets are 3-4 times as large as mean industry premiums and assets. We also see that Medicare Supplement insurance accounts not only for a small proportion of health insurers business but also accounts for a small proportion of Medicare Supplement insurers business. -9-

10 Table 3. Profile of Medicare Supplement Insurers and Health Insurance Industry 2004 and 2008 Medicare Supplement Active Medicare Supplement Active All Health Insurers Insurers Insurers compared to All Health Insurers Number of % 12% Insurers Total Assets $55,626,211,148 $89,073,589,547 $98,614,207,955 $151,848,704,624 56% 59% Mean Total Assets Total Health Premium Income Mean Health Premium Income Total Health Members $561,880,921 $817,188,895 $128,403,917 $172,948, % 473% $96,483,119,611 $158,131,822,721 $227,909,116,611 $344,814,916,344 42% 46% $974,576,966 $1,450,750,667 $296,756,662 $392,727, % 369% 47,338,756 57,161, ,511, ,824,911 37% 36% Table 4. Medicare Supplement Insurance Compared to All Health Products Medicare Supplement Insurance Medicare Medicare Supplement Supplement Insurance Compared to All Insurance Health Products of Insurers Compared to All Active in Medicare Health Products for Supplement Insurance All Health Insurers Premium Income $5,724,059,271 $7,546,285, % 4.77% 2.51% 2.19% Covered Members 3,382,457 3,840, % 6.72% 2.61% 2.45% III. Methodology and Results In this paper, we study the impact of trends in the over-65 U.S. population on health insurers. The most prominent of these trends are growth in the numbers of over-65 people, increases in their longevity, and growth in their utilization of medical services. We have argued -10-

11 that the impacts of these trends will be largely felt by the U.S. Medicare program, rather than by health insurers, because the U.S. government has pre-empted most of the risk in intermediary roles between patients and providers that private health insurers could occupy in the provision of comprehensive care for the over-65 population. To be sure, private health insurers earn considerable fees in the Medicare Advantage (Part C) program. However, the structure of Part C limits insurer risk and data on the portion of Part C for which insurers may be at risk are not broken out. The one area in which private health insurers are exposed to the risks of trends in the over-65 population and for which itemized data are available is Medicare Supplement insurance. Thus, we concentrate on the effects upon the subset of health insurers that write Medicare Supplement insurance. At the end, we shall return for a comment on how our analysis provides bounds to the effects of Part C. We have also argued that the effects of growth of numbers, longevity, and utilization in the over-65 population combine and operate through utilization. We examine the major impact of growing Medicare Supplement utilization upon Medicare Supplement insurers cost of coverage, capital, profitability, and investment risk. To this end, there are four steps in our methodology: 1. We analyze historical trends in Medicare Supplement utilization; 2. Using historical growth rates, we project trends in Medicare Supplement utilization into the future; 3. We analyze the historical relationship between Medicare Supplement utilization as an explanatory variable and Medicare Supplement insurers cost of coverage, capital, profitability, investment risk; and 4. We project the impact of increasing Medicare Supplement utilization upon Medicare Supplement insurers cost of coverage, capital, profitability, investment risk, and cost of -11-

12 coverage by combining projected utilization trends from #2 with our understanding gleaned from #3. Medicare Supplement utilization measures To represent Medicare Supplement utilization, we create several measures of the intensity of Medicare Supplement utilization. Two utilization measures are based upon the number of healthcare system contacts per covered member. 7 Two other utilization measures are based upon the number of healthcare system contacts per premium dollar collected. Healthcare system contacts may be measured by counting the number of encounters with healthcare providers or by tallying the number of hospital admissions and inpatient days. Per member utilization measures (used in the cost of coverage model) 1. pmemmedtotalencounter is healthcare provider utilization per Medicare Supplement member, computed as the total number of Medicare Supplement encounters with healthcare providers divided by the total number of Medicare Supplement members. That is, pmemmedtotalencounter = # encounters / # members. 2. pmemmedinpatient is hospital utilization per Medicare Supplement member, computed similarly as the total number of Medicare Supplement inpatient services divided by the total 7 There is a distinction that is made between insured and member. The insured is the person in whose name the contract is issued. If family members are also covered, they are counted as members, rather than as insureds. Our measures are based upon members, counting in the utilization tally every person entitled to use the coverage. -12-

13 number of Medicare Supplement members. Inpatient services are the sum of number of inpatient admissions and number of inpatient days. That is, ppremmedinpatient = # inpatient services / # members. Per dollar utilization measures (used in the capital, profitability and investment risk models) 3. ppremmedtotalencounter is healthcare provider utilization per Medicare Supplement premium dollar, computed as the total number of Medicare Supplement encounters with healthcare providers divided by the total number of Medicare Supplement premium dollars collected. That is, ppremmedtotalencounter = # encounters / $ premiums. 4. ppremmedinpatient is hospital utilization per Medicare Supplement premium dollar, computed similarly as the total number of Medicare Supplement inpatient services divided by the total number of Medicare Supplement premium dollars collected. Inpatient services are the sum of number of inpatient admissions and number of inpatient days. That is, ppremmedinpatient = # inpatient services / $ premiums. Measures of effects upon Medicare Supplement insurers We assess the effects of Medicare Supplement utilization upon the following variables. Thus these variables are the endogenous response variables in our statistical model. -13-

14 1. Cost of coverage. We use total Medicare Supplement premium income per Medicare Supplement member. This calculation may also be thought of as the price of the insurance. 2. Capital. For capital we compute the logarithm of the ratio of book capital to total insurer assets. 3. Profitability. For profitability we use Health Underwriting Gain calculated the overall underwriting gain from all health related products, including Medicare Supplement, as a proportion of total premiums from all health lines Investment risk. For investment risk, we use the logarithm of opportunity asset risk, a volatility-of-returns measure that comports with notions of portfolio risk in finance. Historical trends in utilization Table 5 shows the historical trends in per member utilization for the over-65 population, juxtaposed for comparison beside the corresponding trends in per member utilization for the under-65 population. Also shown are future utilization projections based upon the historical growth rate. 8 Health insurers underwrite a variety of lines: Comprehensive (group and individual), Federal Employee Health Benefits Plans, Dental, Vision, Medicare, Medicaid, Medicare Supplement, Long-term care, Disability, and Prescription Drug. -14-

15 Year N Table 5: Per Member Medical Service Utilization * 65+ Per Member Utilization Mean Per Member Utilization Mean 2 Medical Encounters 3 Inpatient Services 4 Medical Encounters 3 Inpatient Services Compound Growth Rate 9.93% 3.79% 4.68% 2.42% Projection at Historical Compound Growth Rate * Our sample includes all health insurers recording positive premium income in Medicare Supplement insurance. There are 778 insurer-years in the sample. All of these insurer-years also experience positive premium income for Comprehensive or Federal Employee Health Benefits Plans age group refers to members with Medicare Supplement coverage age group refers to member with either Comprehensive or Federal Employee Health Benefits Plans. 3 Medical Encounters include both physician and non-physician encounters. For each insurer, total encounters are divided by total covered members in each coverage. And the mean of related insurers across the industry is taken. 4 Inpatient services are the sum of inpatient days and inpatient admissions. For each insurer, inpatient services are divided by total covered members in each coverage. And the mean of related insurers across the industry is taken. -15-

16 Table 6 shows the historical trends in per dollar utilization for the over-65 population, as well as future projections based upon the historical growth rate. It should be noted that the dollar figures that we present in all projections are real 2008 dollars. However, we have not attempted to address possible excess inflation in the cost of medical services, although recently the overall costs of medical services have been rising more rapidly than general inflation. That is, our dollar projections implicitly equate medical services inflation to general inflation rates. -16-

17 Year N Table 6. Medical Service Utilization per Premium Dollar * 65+ Mean Utilization per Premium Dollar Mean Utilization per Premium Dollar 2 Medical Encounters 3 Inpatient Services 4 Medical Encounters 3 Inpatient Services Compound Growth Rate 3.80% -0.17% -3.87% -4.77% Projection at Historical Compound Growth Rate * Our sample includes all health insurers recording positive premium income in Medicare Supplement insurance. There are 778 insurer-years in the sample. All of these insurer-years also experience positive premium income for Comprehensive or Federal Employee Health Benefits Plans age group refers to members with Medicare Supplement coverage age group refers to member with either Comprehensive or Federal Employee Health Benefits Plans. 3 Medical Encounters include both physician and non-physician encounters. For each insurer, total encounters are divided by total covered members in each coverage. And the mean of related insurers across the industry is taken. 4 Inpatient services are the sum of inpatient days and inpatient admissions. For each insurer, inpatient services are divided by total covered members in each coverage. And the mean of related insurers across the industry is taken. -17-

18 The data for Tables 5 and 6 are obtained from annual statements filed with the NAIC for years for all insurers that file as health insurers. Tables 5 and 6 also show the utilization projections for 2015, 2020, and 2025 obtained by applying the compounded annual growth rate for each series. Tables 5 and 6 complete steps 1 and 2 of our methodology. Next we develop the historical relationship between utilization and four key effects: cost of coverage, capital, profitability, and investment risk. This part of the methodology develops a statistical model for each of the four effects as dependent variables, with utilization and other appropriate predictors and controls as independent variables. We used NAIC data for insurers that filed as health insurers for We limited our analysis to insurers that had some involvement with Medicare Supplement insurance for reasons argued above. Thus, we omitted health insurers with zero levels of Medicare Supplement premiums. Each year yielded about 100 or fewer insurers involved with Medicare Supplement, for a total of about insurer-years of observations, depending upon the variable analyzed. This step in the analysis is implemented by a series of four interrelated regression models. Each model has one of our effects variables as the response variable. The models are interrelated because the effect variable in one model may be one of the predictor variables in another model. That is, our overall model is a set of simultaneous structural equations. The appearance of the response variable in one equation as a predictor in another creates an endogeneity issue that makes separate estimation of each equation by ordinary least squares statistically invalid. 9 Therefore, we adopt a two-stage approach to estimating coefficients. In the first stage, statistically valid instruments are created for each of the four response variables: We regress each response upon a set of predictors that lack an endogeneity issue. In the second stage, 9 The OLS coefficient estimators are biased and inconsistent. -18-

19 the instruments (estimates of the response variables created in the first stage) replace each instance of a response variable used as a predictor in the structural equations. After the instruments have replaced the response predictors, ordinary least squares may be run on the modified structural equations to yield coefficient estimates that are unbiased and consistent in large samples. In our selection of predictors and controls to use in each equation, we were guided by theoretical considerations that suggested appropriate choices, by the empirical importance of predictors, and by the need to maintain identifiability of parameters. 10 We implemented this procedure using GEE (generalized estimating equations) in SAS. Table 7. Model for Cost of Coverage 65+ Group (766 Obs) 65- Group (729 Obs) Parameter Estimate P-Value Estimate P-Value Intercept Predicted Capital Ratio Predicted Investment Risk Size logrbcratio In Group (1/0) Return on Capital Per Member Medical Encounters < <.0001 Per Member Inpatient Services < < Loss Ratio Loss Ratio Scale R Identifiability essentially means that there is only one set of coefficients that can satisfy the structural equations. -19-

20 Parameter Table 8. Model for Capital Ratio (650 Obs) Estimate S.E. Z Pr > Z Intercept Predicted Health Underwriting Gain <.0001 Predicted Investment Risk Lag of LogCapital Ratio <.0001 Size <.0001 LogRBCratio <.0001 Return on Capital < Group Per Premium Medical Encounters < Group Per Premium Medical Encounters Group Per Premium Inpatient Services Group Per Premium Medical Expenses Scale R

21 Parameter Table 9. Model for Profitability (645 Obs) Estimate S.E. Z Pr > Z Intercept Predicted Log Capital Ratio <.0001 Predicted Investment Risk Lag of Health Underwriting Gain <.0001 Size <.0001 LogRBCratio Group Per Premium Medical Encounters Group Per Premium Medical Encounters Group Per Premium Inpatient Services Scale R

22 Parameter Table 10. Model for Investment Risk (652 Obs) Estimate S.E. Z Pr > Z Intercept <.0001 Predicted Log Capital Ratio Predicted Health Underwriting Gain Lag Investment Risk <.0001 Size < Group Per Premium Medical Encounters Group Per Premium Medical Encounters Group Per Premium Inpatient Services Scale R The key figures from Tables 7-10 are the coefficients of the utilization variables. They show the impact of utilization upon the four financial measures. Together with the projections from Tables 5 and 6, they provide benchmark projections of the effects of over-65 trends that we seek. In the final step of our analysis, let us consider what our equations can tell us about the effects of future changes in Medicare Supplement utilization. The cost of coverage equation (Table 7) shows that the ceteris paribus effects of increases of one each in provider encounters per member and in hospitalization days per members on premiums per member are $ and $ per annum, respectively. Table 5 shows that if historical trends continue, utilization in the form of encounters per member may be expected to grow from a mean of in 2008 to in This is an absolute increase of per member. According to the cost of -22-

23 coverage model (Table 7), an increase of this magnitude may be expected to increase the per member per annum cost of coverage by $ x = $433.31, ceteris paribus. Of course, if the mean utilization for the entire industry were to nearly double, as in this illustration, the whole model may change as well. It would be difficult to maintain the status quo in other predictors. Nevertheless, this calculation presents an informative benchmark. Similarly, a growth in mean per member in-patient services from 2.75 in 2008 to 3.57 in 2015 could be expected to increase the cost of coverage by $ x ( ) = $94.06 per member per annum. By contrast, projected growth in under-65 encounters among Medicare Supplement insurers could be expected to increase their mean cost of coverage by only $ x ( ) = $ per member per annum in 2015, and in-patient services growth to increase under-65 mean cost of coverage by $ x ( ) = $ Table 11 shows the expected increases in cost of coverage per member calculated by this method for each of the projections in Table 5. Table 11. Projected changes in mean cost of coverage (price) per member per year Over 65 Under 65 Encounters In-patient Encounters In-patient 2015 $ $94.06 $ $ $ $ $ $ $1, $ $ $95.68 Using the same methodology as for the cost of coverage model, we can calculate benchmark projections for the effects of utilization changes on capital, profitability, and investment risk. The main difference is that utilization is measured as encounters or in-patient services per dollar for the capital, profitability, and investment risk models, instead of as encounters or in-patient services per member terms for the cost of coverage model. For example, -23-

24 in the capital equation, the coefficients of 65+ (Medicare Supplement) encounters per dollar and in-patient services per dollar are and , respectively. Table 6 shows that if historical trends continue, utilization in the form of dollars per encounter may be expected to grow from a mean of $ in 2008 to $ in This is an absolute increase of $ per encounter. According to the capital model (Table 8), an increase of this magnitude may be expected to increase the capital ratio by x = , ceteris paribus. In 2008, the mean capital ratio for Medicare Supplement insurers was So the indicated increase is a little more than 1% of total insurer capital. Similarly, a decrease in mean per dollar in-patient services from in 2008 to in 2015 could be expected to decrease the capital ratio by x ( ) = By contrast, in Table 8, the effect on capital of an increase in under-65 utilization among Medicare Supplement insurers is negative. However, under-65 utilization is expected to decline. The capital impact of the expected mean change from 2008 to 2015 is calculated as x ( ) = We conjecture that large expected increases in the cost of coverage (premiums) for both the over-65 and under-65 age cohorts (see Table 11) may account for the accumulation of capital for insurers of both cohorts. Table 12 shows the expected changes in the capital ratio calculated by this method for the projections in Table 6. The expected changes in capital are each no more than 6% of total 2008 capital. -24-

25 Table 12. Projected changes in capital ratio Over 65 Under 65 Encounters In-patient Encounters Tables 13 and 14 show the corresponding expected changes in profitability and investment risk for the projections in Table 6. Table 13. Projected changes in profitability Over 65 Under 65 Encounters In-patient Encounters Mean underwriting gain in 2008 was negative ( ), although in most years it was between Most of the expected profitability impacts shown in Table 13 are beneficial. The negative impacts for the under-65 population are relatively modest and offset by positive impacts for the over-65 population. -25-

26 Table 14. Projected changes in investment risk Over 65 Under 65 Encounters In-patient Encounters To put these figures into context, consider that the mean investment risk in 2008 is On account of the financial crisis of that year, this figure is much larger by a factor of about two than preceding years. Thus, many of the expected changes shown in Table 14 are quite substantial. On the other hand, the asset portfolios of health insurers are quite liquid and relatively small a fraction of their annual health premiums. Health insurer portfolios probably bear closer monitoring in the future. Finally, we return to the question of the exposure of health insurers to over-65 trends through insurers at-risk activities in Medicare Part C (Medicare Advantage). We previously noted that the Part C data reported to NAIC are not unbundled into components that would permit a direct analysis such as we have conducted for Medicare Supplement. However, we can gain some insight through the following table of per member utilization for health insurers that are active in Part C that is, all health insurers that report premiums for Part C. -26-

27 Table 15. Per member utilization for health insurers active in Medicare Part C Encounters Inpatient Part C encounter utilization begins in 2001 at about the same level as Medicare Supplement encounter utilization (see Table 5) but increases by only about 1.8% compounded annually. Part C inpatient utilization also begins at about the same level but actually decreases slightly. Therefore, projected future utilization for Part C would be much lower than projected utilization for Medicare Supplement. To apply our projection methodology fully to Part C, we would also need models for the relationship between our four key financial indicia and utilization and other controls for Part C active health insurers. We do not have such models. However, if the active Medicare Supplement insurers are representative of Part C insurers (some of them are the same), then application of our methodology would yield much smaller impacts of over-65 trends on Part C than we have calculated for Medicare Supplement insurers. In conclusion, our analysis shows that there are some potentially troubling aspects of trends for health insurers in the over-65 population; however, it will not be disastrous. Health insurers should be able to adapt. There are even some positive aspects, such as modest projected increases in profitability. Changes in capital are relatively modest and manageable over time by -27-

28 prudent planning. The troubling aspects that we see in the projections are the multiplication of investment risk and growth in cost of coverage. However, growth in cost of coverage is more directly a problem for consumers than for insurers, whereas increases in investment risk are for regulators and legislators to monitor. Indirectly, continuing significant growth in cost of coverage could be problematic for insurers if it prices too many consumers out of the market and triggers regulatory intervention. -28-