Why CBO Changed Its Approach to Projecting Mortality

September 24, 2013

A common measure of the sustainability of the Social Security program is its estimated actuarial balance over a given period—that is, the sum of the present value of projected tax revenues and the current trust fund balance minus the sum of the present value of projected outlays and a target balance at the end of the period. For Social Security, that difference is traditionally presented as a percentage of the present value of taxable payroll (that is, the total amount of wages and self-employment income that is subject to the Social Security payroll tax).

Over the next 75 years, if current law remained in place, the Social Security program’s actuarial shortfall would be 3.4 percent of taxable payroll, CBO estimates—as detailed in The 2013 Long-Term Budget Outlook released last week. In other words, to bring the program into actuarial balance through 2087, given CBO’s projections, payroll taxes could be increased immediately and permanently by 3.4 percent of taxable payroll, scheduled benefits could be reduced by an equivalent amount, or some combination of tax increases and spending reductions of equal present value could be used. That shortfall is substantially greater than the 1.9 percent of taxable payroll that CBO estimated last year.

The single largest factor contributing to the 1.5 percentage-point increase in that actuarial deficit is an increase in CBO’s projection of life expectancy, which accounted for 0.6 percentage points. This year CBO based its long-term budget projections on its own projections of mortality rates, whereas in previous years it used the projections in the annual report of the Social Security trustees. The trustees’ 2013 report incorporates an assumption that mortality rates, adjusted for the age and sex composition of the population, will decline at an average pace of 0.80 percent a year over the next 75 years—a rate of decline smaller than the one seen for the past several decades. CBO, by contrast, projects that mortality rates will decline at an average pace of 1.17 percent a year—as they did between 1950 and 2008. The faster projected decrease in mortality rates, compared with the decrease assumed in last year’s long-term budget outlook, leads to an increase in projected life expectancy and thus an increase in projected spending for Social Security and Medicare.

The Social Security trustees note that, "Many factors are responsible for historical reductions in death rates, including increased medical knowledge, increased availability of health-care services, and improvements in sanitation and nutrition”; their projections, they say, are based on “the expected rate of future progress in these and other areas."1 In the past decade, the trustees have increased their projection of the decline in mortality rates: from 0.70 percent a year for the last 50 years of the 75-year projection period in their 2004 report to 0.80 percent for all 75 years of the projection period in their 2013 report. As a result, the trustees’ projection of life expectancy in 2060 has increased from 79.4 years in the 2004 report to 83.6 years in the 2013 report (see the table below).

Various Analysts' Estimates of What Life Expectancy Will Be in 2060

Many demographers, however, have argued that mortality rates will probably decline more rapidly than the trustees assume. In 1992, Ronald Lee and Lawrence Carter found that age-specific mortality rates fell at a relatively constant pace between the early 1930s and the late 1980s.2 That is, gains in life expectancy have been fairly constant over time despite changes in behavior, medical technology, and public health conditions; in contrast, the trustees assume that gains in life expectancy will be smaller in the future than they were in recent decades. In 2006, John Bongaarts reviewed evidence of increases in life expectancy in 16 high-income countries and found no signs that the pace of improvement in life expectancy was slowing down.3 In 2012, Samir Soneji and Gary King developed a different approach to projecting life expectancy, using U.S. data and adjusting for trends in smoking and obesity; they also concluded that mortality will decline faster than the trustees assume.4 In addition, the Census Bureau’s 2012 projections show a more rapid decrease in mortality rates than the trustees assume.5

Every four years, the Social Security Advisory Board convenes an independent panel of demographers, actuaries, and economists to review the assumptions and methods used by the Social Security trustees. All four of the technical panels that have met since 1999 have recommended that the assumed rate of mortality improvement be based on an extrapolation of past trends.6

The 2011 panel also recommended that such an extrapolation be adjusted, because in their view, the net effect on mortality of changes in smoking and obesity will diminish over time, causing the decline in mortality rates to speed up. The negative impact of smoking on life expectancy peaked in the 1990s and is expected to continue to dwindle because of decreases in the prevalence of smoking that have already occurred. Past improvement in mortality was depressed by increases in smoking, so future improvement will be greater than it otherwise would be because of the decrease in smoking. However, the panel also noted that some of that improvement is likely to be offset by an increase in obesity rates. The panel argued that the Social Security trustees’ mortality projections should be based mainly on an extrapolation of past trends and should not be adjusted for many specific factors—but that the effect of smoking (and, to a lesser extent, of obesity) was large enough and predictable enough to merit explicit adjustment.

CBO chose to take a simpler approach—extrapolating from past trends without adjustment—for a few reasons. First, a number of past factors (such as improvements in medical technology, environmental conditions, and health behaviors) have had as significant an impact on mortality rates as smoking has. Second, there is great uncertainty about how such factors will affect mortality in the future. Projecting a continuation of past improvement requires no subjective judgments about the roles that specific changes will play. That approach implies that although a great deal of uncertainty exists about the impact of the many factors that will influence mortality, the effects on mortality of future changes in those factors will be such that mortality rates will continue to decline at their long-term average pace.

Projecting mortality from past trends requires choosing a historical period on which to base the projection. The pace at which mortality improved was different in the first and second halves of the 20th century, with a noticeable break around 1950. Like the 2011 technical panel, CBO based its projection on the average improvement beginning in 1950.

Joyce Manchester is Chief of the Long-Term Analysis Unit in CBO’s Health, Retirement, and Long-Term Analysis Division.


1 Social Security Administration, The 2013 Annual Report of the Board of Trustees of the Federal Old-Age and Survivors Insurance and Federal Disability Insurance Trust Funds (May 31, 2013), p. 80, www.ssa.gov/OACT/TR/2013/index.html. For additional discussion of those assumptions, see Social Security Administration, Office of the Chief Actuary, The Long-Range Demographic Assumptions for the 2013 Trustees Report (May 31, 2013), www.ssa.gov/oact/TR/2013/2013_Long-Range_Demographic_Assumptions.pdf (620 KB).

2 Ronald D. Lee and Lawrence R. Carter, "Modeling and Forecasting U.S. Mortality," Journal of the American Statistical Association, vol. 87, no. 419 (September 1992), pp. 659–671, http://tinyurl.com/ono94tn.

3 John Bongaarts, "How Long Will We Live?" Population and Development Review, vol. 32, no. 4 (December 2006), pp. 605–628, http://dx.doi.org/10.1111/j.1728-4457.2006.00144.x.

4 Samir Soneji and Gary King, “Statistical Security for Social Security,” Demography, vol. 49, no. 3 (August 2012), pp. 1037–1060, http://dx.doi.org/10.1007/s13524-012-0106-z. (Those projections extend only to 2031, so they are not included in the table on the previous page.)

5 Census Bureau, Methodology and Assumptions for the 2012 National Projections (2012), www.census.gov/population/projections/data/national/2012.html.

6 See Social Security Administration, Technical Panel on Assumptions and Methods, Report to the Social Security Advisory Board (September 2011), pp. 55–64, www.ssab.gov/Reports/2011_TPAM_Final_Report.pdf (6.3 MB). For additional background, see Hilary Waldron, “Literature Review of Long-Term Mortality Projections,” Social Security Bulletin, vol. 66, no. 1 (September 2005), www.socialsecurity.gov/policy/docs/ssb/v66n1/v66n1p16.html; and John R. Wilmoth, Overview and Discussion of the Social Security Mortality Projections, working paper for the 2003 Technical Panel on Assumptions and Methods (Social Security Advisory Board, May 5, 2005), www.ssab.gov/documents/mort.projection.ssab.pdf (480 KB).