Eighteen years ago, Henry Aaron, Barry Bosworth, and

Transcription

1 Abstract - Long term federal outlays for Medicare and Medicaid are projected to increase in the future because of the interaction between demographics and program eligibility. However, the magnitude of the projected increase depends almost entirely on the assumption about excess cost growth. This growth factor is an unexplained residual the gap between growth in per beneficiary program outlays and growth in GDP per capita not explained by demographics alone. In the last several decades, excess cost has been positive and significant on average, though the annual pattern is volatile. This paper considers the implications of alternative assumptions about excess cost growth for federal budget and tax policy. Jonathan A. Schwabish, Julie Topoleski, & Ignez Tristao Congressional Budget Office, Washington, D.C National Tax Journal Vol. LX, No. 3 September 2007 INTRODUCTION Eighteen years ago, Henry Aaron, Barry Bosworth, and Gary Burtless posed the question, Can America afford to grow old? That study was seminal because the long term budget projections were explicitly tied to a macroeconomic/ demographic model of long run growth. 1 The question raised in the Brookings study focused primarily on the issue of Social Security financing, with little attention devoted to health care outlays. The question of whether America can afford to grow old has now come of age, and although there are still outstanding financing issues in the area of Social Security, those problems pale in comparison to the question of whether we can afford to get sick. Health care spending is expected to consume a growing portion of the nation s GDP over the next few decades, which could have significant implications for federal tax and budget policy. Not only are Medicare and Medicaid outlays expected to increase rapidly, but increased total health care costs and the provision and cost of private health insurance also have the potential to affect tax revenues. For example, employment based health insurance premiums and some out of pocket costs are paid with pre tax income, driving down total income tax receipts. As health insurance costs increase, workers may receive even less of their compensation as cash wages. These factors have the potential to reduce 1 The model used for the Aaron et al. (1989) analysis is the intellectual ancestor of the Congressional Budget Office Long Term model (CBOLT) used for this analysis. The macro/budget equations from the Brookings model were adopted for a number of CBO studies in the late 1980s and 1990s, and eventually became the core of the macro sector in CBOLT. 407

2 NATIONAL TAX JOURNAL tax revenues and force America to decide how these reduced tax revenues should be spent. Outlays for Medicare and Medicaid increase over time because of the interaction between demographics and program eligibility. However, the magnitude of the increase in both history and projections is dominated by a factor known as excess cost growth. Excess cost growth is basically an unexplained residual the gap between growth in per beneficiary program outlays and growth in GDP per capita that is not explained by demographics alone. In the last several decades, excess cost has been positive and significant on average, though the year by year pattern is volatile. Given uncertainty about the future of the residual, this paper considers the implications of alternative assumptions about excess cost growth for federal budget and tax policy. Thirty years ago, federal spending on the two main government financed health care programs Medicaid and Medicare accounted for about one percent of GDP; by 2005, this figure had risen to 4.2 percent. The projections below show that health care costs will continue to consume a growing share of the nation s GDP, but these projections are very sensitive to assumptions regarding excess cost growth: the rate at which growth in costs per beneficiary exceeds growth in GDP per capita. PUBLIC HEALTH INSURANCE PROGRAMS IN THE UNITED STATES Medicare and Medicaid are the two primary government financed health care programs that provide health insurance coverage for the elderly, the disabled and the poor. Together, they cover 25 percent of the U.S. population and finance over 45 percent of national health spending, rising from 37 billion dollars in 1980 to 408 billion in Medicare primarily provides health benefits to the nation s elderly population, although about 15 percent of the current Medicare population is disabled and under age 65. In 2005, there were more than 42 million Medicare beneficiaries, and the share of disabled beneficiaries has been growing slowly over time. As the baby boom generation ages, Medicare enrollment is projected to increase to 78.6 million by 2030 (Centers for Medicare and Medicaid Services and Office of the Assistant Secretary for Planning and Evaluation, 2007). Also affecting Medicare outlays is the changing composition of costs. In 1980, inpatient hospital services accounted for 68 percent of total costs (see Figure 1) and physician costs accounted for 24 percent of costs. By 2006, inpatient hospital services accounted for only 36 percent of total costs; physician costs had fallen to 14 percent of the total; and prescription drug costs account for 12 percent of costs. This changing composition of Medicare costs could have implications for future costs. Medicaid is jointly funded by the federal and state governments and pays for health care services for a variety of low income individuals. 2 In terms of enrollment, it is the government s largest health care program, covering more than 60 million people or about 20 percent of the U.S. population. In particular, it pays for 40 percent of all births and provides health insurance to one third of all children. Medicaid is also the main financer of 2 The share of benefit spending paid by the federal government varies across states and services but averages 57 percent. States design and administer their Medicaid program under federal guidelines that specify a minimum set of services that must be provided. There is, therefore, wide variability in eligibility and coverage of health services. In many states, the State Children s Health Insurance Program (SCHIP) is part of Medicaid program. In this paper, we exclude the SCHIP program costs from Medicaid spending. 408

3 Figure 1. Distribution of Medicare Spending, 1980 and 2006 Sources: 1980: CMS, An Overview of the U.S. Health Care System Chart Book, Table 4.4, TheChartSeries/downloads/Chartbook_2007_pdf.pdf. 2006: 2007 Medicare Trustees Report, Table II.B

4 NATIONAL TAX JOURNAL nursing home care, paying for two thirds of all nursing home stays by the time of patient discharge. 3 While children and non elderly, non disabled adults make up almost three quarters of Medicaid recipients, the elderly and the disabled account for 75 percent of total spending (see Table 1). One third of Medicaid s spending goes towards long term care, with elderly and disabled populations accounting for virtually all of that cost. In fact, the disabled account for under 20 percent of beneficiaries and almost 50 percent of costs. Between 1999 and 2004, total Medicaid spending increased by 70.2 percent, rising from $140 billion to $238 billion. Increases in both the number beneficiaries and the cost per beneficiary helped generate this growth. During this period, the number of beneficiaries grew by 39.2 percent, while cost per beneficiary rose by 22.3 percent. Of the four main Medicaid eligible groups children, adults, elderly and disabled it was the adult population that grew the fastest during this period, increasing by 69 percent. Cost per beneficiary, on the other hand, grew the fastest for the disabled population (a 52 percent increase), a group that only makes up about 19 percent of the total Medicaid population. Four main factors contribute to the growth in total spending in the Medicaid and Medicare programs: increases in the number of beneficiaries, real increases in the cost per beneficiary, changes in the composition of beneficiaries, and general inflation in prices. Federal Medicaid benefits are projected to double in nominal terms over the next ten years, increasing from $172 billion in 2007 to $353 billion by 2016 (Congressional Budget Office, 2007). This aggregate increase is expected to be partly being driven by growth in enrollment (especially in the elderly and disabled populations) and growth in costs per beneficiary (especially for prescription drugs and non institutional long term care services) (Congressional Budget Office, 2006). DEFINING EXCESS COST GROWTH We refer to the unexplained residual growth in health spending that exceeds demographic and economic growth as excess cost growth. The excess cost growth concept allows historical cost growth trends to be summarized and is often used as the basis for health spending projections. For example, the Medicare Trustees use an excess cost growth assumption of one percent in Elderly Disabled Children Adults Total TABLE 1 DISTRIBUTION OF MEDICAID RECIPIENTS AND BENEFIT PAYMENTS BY POPULATION CATEGORY, 2004 Benefit Payments Number of Recipients Acute Care Long Term Care 10.0% 18.8% 48.2% 23.0% 100.0% 9.3% 34.4% 13.3% 10.9% 67.9% 18.4% 13.2% 0.2% 0.3% 32.1% Total Spending 27.6% 47.7% 13.5% 11.2% 100.0% Note: Long term care includes payments for care in nursing homes and intermediate care facilities for the mentally retarded, home health services, and other community based services. Managed Care costs are not included in these totals. 3 For a more detailed description of the Medicaid program and the sources of growth in its spending, see Congressional Budget Office (2005, 2006a). 410

5 their intermediate long run projections of Medicare spending (Medicare Trustee Report, 2007). Similarly, the Congressional Budget Office used an excess cost growth assumption of one percent in their intermediate spending scenario for long run projections of spending on Medicare and Medicaid (Congressional Budget Office, 2005). 4 While there is uncertainty in both economic and demographic projections, the dominant source of uncertainty in projecting Medicare and Medicaid is residual growth in spending per enrollee relative to the growth of per capita GDP. Assumptions about the level of excess cost growth can have large effects on long run projections of federal health spending. For example, suppose that Medicare excess cost growth is zero percent. Here, Medicare spending per beneficiary (within any given age group) would grow at the same rate as per capita GDP. Thus, although per beneficiary spending and per capita GDP would continue to grow, the increase in Medicare s share of GDP would only reflect the aging of the population. In the case of one percent excess cost growth, spending per beneficiary would grow at the rate of GDP growth plus one percentage point, which in the long run could generate large differences in Medicare s share of GDP. ESTIMATING AND PROJECTING HEALTH CARE COSTS The projection methodology has four steps. First, micro data are used to create point in time spending estimates by individual characteristics such as age, sex, time until death, and beneficiary type. Second, these spending estimates are combined with aggregate beneficiary counts to infer excess health cost growth in history and in the ten year CBO budget window. Third, the ratio of aggregate beneficiary counts to the overall population are calculated and fixed, allowing us to predict beneficiary populations beyond the ten year projection window. Finally, long run excess cost growth rates are assumed, which ultimately generates the overall projections of federal health outlays. Step 1: Create Spending Indices For both Medicare and Medicaid, the first step is to develop point in time spending indices that vary by relevant characteristics. For Medicaid, the indices vary by age, sex, and population group. For Medicare, those characteristics are age, sex, and time until death. The Medicaid cost indices are constructed using the Medicaid Statistical Information Statistics (MSIS) micro data. For each year, 1999 to 2004, the MSIS contains monthly records for each person in the Medicaid system, their age, sex, state of residence and spending by service type. We calculate national population and expenditure totals for each age, sex, seven service types (prescription drugs, hospitals, institutional services (such as nursing homes), managed care, non institutional services (such as home health and personal care), physicians, and dentists, other practitioners, and other forms of acute care), 5 and four population 4 For a detailed exposition of excess cost growth, see White (2006). 5 Other acute services include outpatient, clinic, laboratory and x ray, hospice, rehabilitation, physical and occupational therapy, target case management, nurse midwife and practitioner, sterilizations, and private duty religious non medical services. We also forecast costs for four smaller categories including vaccines, disproportionate share of hospitals, Medicare premiums, and administrative costs. Forecasting these costs mimics those used to project the ten year forecast and typically grow at about the rate of consumer prices. 411

6 NATIONAL TAX JOURNAL groups (children, adults, disabled and elderly). 6 Because health spending can change during a person s lifetime, especially in the final years of life, the methodology to estimate per beneficiary Medicare costs is slightly more complex. Instead of simply estimating Medicare costs by age and sex, we incorporate measures of time until death (TUD) to explicitly capture those high costs associated with end of life care. This methodology is described in detail in Sabelhaus, Simpson and Topoleski (2004). The micro data used for the Medicare cost estimates are produced by Acumen, LLC and contain monthly records of a five percent sample of all enrolled Medicare beneficiaries between 1989 and Each observation includes monthly information on enrollment and entitlement status (aged fee for service, disabled fee for service, aged managed care, disabled managed care, not enrolled, or dead), six spending categories (inpatient hospital, skilled nursing facilities, hospice, home health, outpatient hospital, and physician and other provider service utilization), dates of birth, and death, race and sex. These data are then used to generate average spending indices that vary by age, sex and time until death for each category of cost. Step 2: Inferring Excess Cost Growth These average cost indices are then used in the calculation of excess cost growth, both historically and over the next ten years. As shown in the following equation, the average cost indices are combined with aggregate beneficiary counts, growth in per capita GDP and excess cost growth to generate future spending levels. For Medicare, this equation is as follows: [1] X t = X t N χ asdt asdt a= 1 s= 1 d= Nasd( t 1) χ asd ( t 1) a= 1 s= 1 d= 1 GDPt ( 1+ η t ), GDP t 1 where η t = excess cost growth rate; X t = Medicare costs in time t; N asdt = number of beneficiaries for each age (a)/sex (s)/tud (d) category in time t; 8 and χ asdt = the cost index. Excess cost growth for Medicaid is similarly inferred, but varies by age, sex, service type and population group. In the ten year projection period, excess cost growth is inferred from the various components by solving equation [1] for η t : [2] Xt ηt = X t 1 GDP GDP t 1 t N a= 1 s= 1 d= a= 1 s= 1 d= 1 1. χ asd( t 1) asd( t 1) N asdt χ asdt 6 Because we aggregate MSIS monthly data into annual totals and individuals may be eligible for Medicaid under different population groups during the year (for instance, a child for some months and disabled for others), we weight the totals by the share of months under these different eligibility statuses. In addition, we make some restrictions on the MSIS population to correct for what we see are inconsistencies in the data. In particular, we only include costs for persons age 18 or under in the children population. Similarly, we only include costs for persons age 65 or over in the elderly population and persons age 19 to 64 in the adult population. Overall, our estimates are very close to publicly available estimates published by the Centers for Medicare and Medicaid Services (CMS) (see CMS Medicare and Medicaid Statistical Supplement Tables, www. cms.hhs.gov/medicaremedicaidstatsupp/lt/list.asp#topofpage). 7 Major outliers beneficiaries with more than $1 million in annual spending were removed from the data set. Sabelhaus et al. (2004) used the same data set through For a more thorough review of the estimation methodology, the reader should see Sabelhaus et al. (2004). 8 TUD categories include dying within on year; dying within two years; and dying outside the two year window. 412

7 Thus, the inferred value of excess cost growth is the residual growth remaining after accounting for economic growth and demographic changes. Total Medicare costs have been growing faster than the economy for years. In 1970, Medicare spending accounted for 0.7 percent of GDP, and by 2005, it had increased to 2.7 percent. Program costs have increased in part because of increased enrollment (see Table 2), but costs per enrollee rose 3.0 percentage points faster than per capita GDP over the period. If the 1970s are excluded, the excess cost growth rate is a smaller 2.4 percentage points, and if the 1980s are excluded, it is smaller yet at 2.1 percentage points. Medicare spending decelerated in the late 1990s but has since rebounded (see also Glied (2003)). This excess cost growth captures both the effect of policy changes and residual growth. The Medicaid program has also grown faster than the economy, increasing from 0.3 percent of GDP in 1970 to 1.5 percent in From 1975 through 2004, Medicaid grew on average 2.4 percentage points faster than GDP per capita (Table 2). 9 The excess cost growth rate between 1980 and 2004 averaged 1.6 percent, and from 1990 to 2004, the average was 1.4 percent. This decrease reflects, in part, the increased enrollment of children into the Medicaid program because children have lower per capita costs than other Medicaid beneficiaries. Presenting only average inferred excess cost growth figures does not show the volatility in the series. This volatility underscores how difficult it is to predict future excess cost growth. Inferred Medicare excess cost growth rates have varied historically, from a low of 5.8 percent to more than ten percent (see Figure 2). Recall that excess cost growth is the residual growth in program costs remaining after accounting for economic growth (growth in GDP per capita) and demographic changes (growth in the cost index multiplied by beneficiary population). As White (2006) notes, this variation may be attributable to policy changes affecting payment policies and regulations. 10 This volatility continues into the projection Years a TABLE 2 GROWTH IN THE MEDICARE AND MEDICAID PROGRAMS Annual Percentage Growth in Federal Outlays Annual Percentage Growth in Enrollment b Medicare Medicaid Annual Percentage Growth in Federal Outlays per Enrollee Annual Percentage Growth in per Capita Gross Domestic Product Annual Excess Cost Growth c a Medicare data are for calendar years; Medicaid data are for fiscal years. b The measure of enrollment used for Medicare reflects the effects of costs on the changing composition of Medicare beneficiaries; the measure of enrollment used for Medicaid does not. The latter measure is based on administrative data from the Centers for Medicare and Medicaid Services. c Excess cost growth is one plus the growth rate of outlays divided by one plus the growth rate of per capita GDP minus one Note that 2004 is the most recent year for which Medicaid enrollment figures are available. 10 See White (2006) for a detailed discussion of how changes in payment policy affected excess cost growth. 413

8 NATIONAL TAX JOURNAL Figure 2. Inferred Excess Cost Growth for Medicare and Medicaid, period. From 2006 through 2016, residual excess cost growth rates are inferred such that we exactly match Congressional Budget Office projections for that period. Inferred excess cost growth for Medicaid is similarly volatile. For example, consider the implications of the addition of a prescription drug benefit to the Medicare program. In the first year of the benefit, there is a level shift in total Medicare costs. Putting this in terms of the definition above (see equation [2]), growth in Medicare outlays (X t /X t 1 ) will be large because of this level shift, but the remainder of the equation would be largely unchanged. The result would be a residual excess cost growth term that is large, at least until the new benefit is fully phased in. At that point, the residual excess cost growth would be expected to return to a lower level. Knowing that historical excess cost growth estimates include the effects of policy changes makes it difficult to make assumptions about the level of excess cost growth going forward. We do not expect 414 the policy changes that have already been made to be repeated, and it is impossible to predict future policy changes. As a result, we consider three alternative assumptions regarding excess cost growth in the future. Step 3: Calculating the Beneficiary Population Ratio The third step requires us to calculate the ratio of beneficiaries in each age and sex cell to the overall population during the historical period. These ratios are fixed and applied to population projections over time to solve for the number of Medicare and Medicaid beneficiaries. Population projections by age and sex are generated using an integrated macro/micro model. Step 4: Assuming Long Run Excess Cost Growth Rates As we stated above, long term projections for Medicare and Medicaid are subject to considerable uncertainty, and

9 the dominant source of that uncertainty is the growth is spending per enrollee relative to the growth of per capita GDP. Here we consider three different assumptions about excess cost growth: Spending per enrollee grows 2.5 percentage points faster than GDP per capita; Spending per enrollee grows one percentage point faster than GDP per capita; and Spending per enrollee grows at the same rate as GDP per capita. The three paths represent very different assumptions about the future of Medicare and Medicaid outlays and are each included separately in the projections of future health costs using equation [1] specified above. These long run assumptions are generally lower than historical averages, but they are consistent with what others have assumed. The Medicare Trustees assume that health care costs will continue to grow faster than per capita GDP, but that pressure will be brought to bear on the health care system to reduce the excess cost growth rate to one percentage point. This reflects the view that higher levels of excess cost growth are not sustainable in the long run because they lead to an implausibly large share of GDP being devoted to health care (Technical Review Panel on the Medicare Trustees Reports, 2000). PROJECTING MEDICARE AND MEDICAID COSTS Total Medicare spending was about $408 billion in 2006, up from about $37 billion only 26 years earlier. In the same year, the federal share of Medicaid accounting for approximately 57 percent of the total was just under half the Medicare total with $202 billion in costs. Together, federal spending on these two programs grew from 1.0 percent of GDP to percent between 1970 and Our forecasts suggest under the intermediate spending path of one percent excess cost growth that total public health care costs could grow to 8.4 percent of GDP by 2025 and 13.0 percent by Spending Projections Under Alternate Excess Cost Assumptions The three paths of spending projections for Medicaid and Medicare are presented in Figures 3 and 4. For Medicaid, the intermediate spending path shows that by 2025 total spending is projected to account for 2.5 percent of GDP, a full percentage point above the current ratio of 1.5 percent (see Figure 3). By 2050, that percentage is projected to climb to 3.7 percent. Under the low cost scenario where health care costs grow at the same rate as GDP per capita Medicaid costs are projected to account for about 1.5 to 1.8 percent of the nation s total output. Under the higher spending scenario, Medicaid spending is projected to be about 2.6 percent of GDP in 2025 and exceed five percent of GDP by the end of the projection period. The analogous spending paths for Medicare are presented in Figure 4, where a similar story is demonstrated: The intermediate spending path shows total projected Medicare spending as a percent of GDP at about nine percent in Under the low cost scenario, this spending is projected to grow slightly more slowly, but is not as flat as in the case of Medicaid. Finally, under the high cost scenario, Medicare spending is projected to make up 7.8 percent of GDP in 2025, rising to over 17 percent of the nation s total output by By 2019, when the Medicare actuaries estimate the Medicare trust fund will be exhausted (Medicare Trustees Report, 2007), we project that Medicare will account for between four and eight percent of the nation s output, or between $980 billion and $1.4 trillion. Fifty years earlier in 1970, total Medicare

10 NATIONAL TAX JOURNAL Figure 3. Total Federal Spending for Medicaid Under Different Assumptions About Excess Cost Growth (Percent of GDP) Figure 4. Total Federal Spending for Medicare Under Different Assumptions About Excess Cost Growth (Percent of GDP) 416

11 Figure 5. Total Federal Spending for Medicare and Medicaid Under Different Assumptions About Excess Cost Growth (Percent of GDP) spending accounted for less than one percent of GDP and equaled $7.2 billion. Finally, in Figure 5, we put the two systems together to show the cumulative share of GDP held by the public health system during the next 50 years. The results, though not unexpected, are striking. Under the one percent excess cost growth scenario, total public health system costs are projected to account for 8.4 percent of the nation s total output in 2025 and 13.0 percent by At the higher spending assumption of 2.5 percent excess cost growth, these percentages are projected to jump to 10.4 percent and 22.6 percent. CONCLUSION This analysis has shown that the magnitude of the increase in Medicare and Medicaid costs in both history and projection is dominated by a factor known as excess cost growth. Differences in assumptions about excess cost growth can shift the share of total public health spending by several percentage points. As the baby 417 boom ages and health care costs climb, the nation will have to decide which tax and spending policies to incorporate and where to place government spending on health on the list of national priorities. Acknowledgments The analysis and conclusions expressed in this paper are those of the authors and should not be interpreted as those of the Congressional Budget Office. The authors wish to thank Carol Frost and Susan Labovich for assistance with the Medicare and Medicaid data. We would especially like to thank Jeanne de Sa and Eric Rollins for assistance with program rules and the Medicaid data. REFERENCES Aaron, Henry J., Barry P. Bosworth, and Gary Burtless. Can America Afford to Grow Old? Paying for Social Security. Washington, D.C.: The Brookings Institution, 1989.

12 NATIONAL TAX JOURNAL Centers for Medicare and Medicaid Services and Office of the Assistant Secretary for Planning and Evaluation. An Overview of the U.S. Health Care System Chart Book. Accessed April, chartbook/report.pdf Congressional Budget Office. The Long Term Budget Outlook. A Congressional Budget Office Study. Washington, D.C., December, Congressional Budget Office. Medicaid Spending Growth and Options for Controlling Costs. Statement of Donald B. Marron, Acting Director, before the Special Committee on Aging, United States Senate. Washington, D.C., July 13, Congressional Budget Office. Fact Sheet for CBO s March 2007 Baseline. Washington, D.C., March, gov/ftpdocs/78xx/doc7861/m_m_schip. pdf. Glied, Sherry. Health Care Costs: On the Rise Again. Journal of Economic Perspectives 17 No. 2 (Spring, 2003): Medicare Trustees Report. The 2007 Annual Report of the Boards of Trustees of the Federal Hospital Insurance and Federal Supplementary Medical Insurance Trust Funds. Washington, D.C.: Center for Medicare and Medicaid Services, April 23, Sabelhaus, John, Michael Simpson, and Julie Topoleski. Incorporating Longevity Effects into Long Term Medicare Projections. CBO Technical Paper No Washington, D.C.: Congressional Budget Office, January, Technical Review Panel on the Medicare Trustees Reports. Review of Assumptions and Methods of the Medicare Trustee s Financial Projections. Washington, D.C., December, White, Chapin. The Slowdown in Medicare Spending Growth. CBO Working Paper No Washington, D.C.: Congressional Budget Office, July,