1The projected transfers for 2026 are pertinent because this is when the Hospital Insurance (Part A) Trust Fund is expected to be exhausted. Trust Fund exhaustion has been considered the bellwether of the system's financial status, but beginning in 2013, Part A spending will exceed payroll tax revenues, necessitating transfers from the rest of the budget.

2In making these estimates, future federal individual and corporate income taxes are assumed to remain at their 2001 share of Gross Domestic Product (GDP), which was 10.99 percent. This assumption is based on the 50-year average income tax share of GDP, from 1953 to 2002. During that period, income taxes as a percent of GDP ranged from a high of 13.7 percent in 1952 to a low of 9.3 percent in 1984, 1986 and 1992.

3The transfers under the lower growth rate assumptions are 11, 19 and 22 percent of federal income taxes in 2020, 2040, and 2060, respectively.

4The Trustees' high-cost projection assumes that per capita spending grows 2 percent faster than per capita GDP for a number of years beyond the initial 25-year period, and then decreases to the intermediate assumption of GDP plus 1 percent by 2052.

5Based on the Medicare Current Beneficiary Survey Cost and Use File, Medicare's share of prescription drug spending grew to 8.1 percent in 1999.

6The estimates of Medicare's payments as described in House bill H.R. 4954 are derived from the prescription drug expenditure distributions as presented in March 22, 2001, Congressional Budget Office (CBO) testimony.

7The first 10 years of projected prescription drug costs are from the March 2002 CBO testimony. To estimate costs beyond 2012, we used the conservative assumption that costs grow at the same rate as the combined cost of the HI and SMI programs and assumed Medicare will pay for 26.2 percent of prescription drug spending.

8The accrued Social Security benefits depicted in the figure are based on estimates by Goss (1999) and updated by the Social Security Administration. See Stephen C. Goss, "Measuring Solvency in the Social Security System," in Prospects for Social Security Reform, Olivia S. Mitchell, Robert J. Meyers, and Howard Young, eds. (Philadelphia: University of Pennsylvania Press, 1999), for a description of this measure, which is referred to as the "maximum transition cost." More recent unpublished estimates of maximum transition cost are available from the Social Security Administration. The maximum transition cost as reported includes the offsetting effects of the Social Security Trust Fund. Because we do not separately report Trust Fund debts in Figure V, the offset is removed from the maximum transition cost.

9Accrued Medicare benefits are assigned to individuals age 28 and above in 2001. We chose 28 based on the assumption that individuals are fully vested in Medicare by that age. Accrued benefits are equal to 100 percent of the present value of anticipated HI benefits. To reflect the fact that beneficiaries pay premiums equal to 25 percent of SMI, only 75 percent of anticipated SMI spending is included in the calculation. Like the calculation for Social Security, we apply life tables and discount future benefits at the government borrowing rate. Medicare's accrued liability is larger than Social Security's because workers and their spouses are fully vested in Medicare after 10 years of program participation, while workers with fewer than 35 years of participation are only partially vested in Social Security.

10See Liu Liqun, Andrew J. Rettenmaier and Thomas R. Saving, "Meaningful Measures of Fiscal Deficit and Debt: The Case for Incorporating Entitlement Debts," PERC Working Paper No. 0211, May 2002, Private Enterprise Research Center, Texas A&M University.

11The HCC model replaces the Primary Inpatient Diagnostic Cost Group, or PIP-DCG, model currently in use. The PIP-DCG model's major drawback has been that risk adjustments were limited to inpatient diagnoses, which affect only about 20 percent of enrollees. There have also been concerns that its reimbursements for healthy individuals are too high. The HCC promises to produce more accurate reimbursements and to reduce the data collection burden on participating providers.

12The commission also recommended redefining the system's "solvency." Under the definition suggested, insolvency would occur when Medicare's spending from general revenues exceeded 40 percent of the program's total spending. See the Breaux-Thomas proposal at http://thomas.loc.gov/medicare/bbmtt31599.html.

13The inclusion of catastrophic coverage will lead to the demise of medigap insurance, which for most retirees converts Medicare into a first-dollar coverage policy. It is this first-dollar coverage that ensures beneficiaries do not care about cost and have no financial incentive to monitor health care providers. Closing the gaps in Medicare will eliminate the need for medigap.

14Joseph P., Newhouse, Free for All? Lessons from the Rand Health Insurance Experiment (Cambridge, Mass.: Harvard University Press, 1996).

15Andrew J. Rettenmaier and Thomas R. Saving, The Economics of Medicare Reform (Kalamazoo, Mich.: W.E. Upjohn Institute for Employment Research, 2000).

16The data used to produce this graph and the ones that follow are from the Continuous Medicare History Sample. The data are a 1 percent sample of all Medicare beneficiaries and are available from 1974 to 1997. It is thus possible to observe beneficiaries born in 1912 from age 65 to age 85. At the end of the sample period, 42 percent of the beneficiaries born in 1912 were alive (survived to age 86). All reimbursements were converted to 1997 dollars using the Personal Consumption Expenditures price deflator.

17For a presentation of annual Medicare spending by age of death for individuals who died in 1990, where the deflator was based on the rate of change in Medicare spending on elderly beneficiaries, see James Lubitz, Beebe James and Colin Baker, "Longevity and Medicare Expenditures," New England Journal of Medicine, Vol. 332 (15), April 1995, pages 999-1003. The authors also estimate cumulative lifetime spending by age of death using the 1990 cross-sectional experience rather than the present value of spending within birth years.

18See Andrew J. Rettenmaier and Zijun Wang, "Explaining the Growth of Medicare: Part II," NCPA Brief Analysis No. 408, August 6, 2002, National Center for Policy Analysis.

19Each calculation considers the same group of beneficiaries: all those originally in the sample at age 65. Should an individual die between age 65 and the last age in the summation, his or her reimbursements end in the final year of life.

20Predictability of remaining lifetime spending based on past spending patterns is still an open question. Several papers have addressed year-to-year persistence in health care spending, as has the risk-adjusted prospective payment system for Medicare+Choice providers. See Daniel Feenberg and Jonathan Skinner, "The Risk and Duration of Catastrophic Health Care Expenditures," Review of Economics and Statistics 76 (1994), pages 633-47; and Matthew Eichner, Mark B. McClellan and David A. Wise, "Insurance or Self-Insurance? Variation, Persistence, and Individual Health Accounts," in Inquiries in the Economics of Aging, David A. Wise, ed. (Chicago: University of Chicago Press, 1998), pages 19-45.

21We observe this birth year from 65 to 85 years of age. However, total lifetime spending is not known for a significant share of these beneficiaries, given that more than 42 percent are alive at the end of the period for which we have data. To overcome this limitation, we impute remaining lifetime spending for those still alive at age 85. The imputed values are derived from the experience of the beneficiaries born in 1897 who are observed to the age of 100. The imputation procedure is described in the appendix. Combining the imputed values with the actual values, we then consider how spending between ages 65 and 69 is related to remaining lifetime spending. Only individuals who are alive at the age of 69 are included in the comparison.

22John H. Cochrane, "Time-Consistent Health Insurance," Journal of Political Economy, Vol. 103 (3), June 1995.

23Individual beneficiaries may, and probably will, make premium payments to the plans. Such payments do not change the fundamental analysis of this section.

24Spending in the year of death represents, on average, six months of spending.

25For the fraction of beneficiaries who had no reimbursements between 80 and 85 years of age in each cohort, a different imputation procedure is used. First, the location in the remaining lifetime spending distribution is found for members of the 1897 birth year. This location is randomly assigned by sex to members of the 1912 birth year who also had no spending between ages 80 and 85. Based on that location, remaining lifetime spending is assigned.