NBER Working Paper 23329; How the Growing Gap in Life Expectancy May Affect Retirement Benefits and Reforms

By Alan J. Auerbach, Kerwin K. Charles, Courtney C. Coile, William Gale, Dana Goldman, Ronald Lee, Charles M. Lucas, Peter R. Orszag, Louise M. Sheiner, Bryan Tysinger, David N. Weil, Justin Wolfers, and Rebeca Wong
National Bureau of Economic Research, April 2017

(Note: These excerpts from this paper are quite wonkish so you may want to skim down to the Comment, and then return to read this only if you are still interested in the specifics.)


Older Americans have experienced dramatic gains in life expectancy in recent decades, but an emerging literature reveals that these gains are accumulating mostly to those at the top of the income distribution. We explore how growing inequality in life expectancy affects lifetime benefits from Social Security, Medicare, and other programs and how this phenomenon interacts with possible program reforms. We first project that life expectancy at age 50 for males in the two highest income quintiles will rise by 7 to 8 years between the 1930 and 1960 birth cohorts, but that the two lowest income quintiles will experience little to no increase over that time period. This divergence in life expectancy will cause the gap between average lifetime program benefits received by men in the highest and lowest quintiles to widen by $130,000 (in $2009) over this period. Finally we simulate the effect of Social Security reforms such as raising the normal retirement age and changing the benefit formula to see whether they mitigate or enhance the reduced progressivity resulting from the widening gap in life expectancy.

From the Introduction

We have several major findings. First, consistent with other recent studies, we confirm that life expectancy at older ages has been rising fastest for the highest socioeconomic groups. For those born in 1930, the gap in life expectancy at age 50 between males in the bottom 20 percent and top 20 percent of lifetime income is 5 years, according to our estimates. For males born 30 years later (in 1960), the projected gap at age 50 between the highest and lowest quintiles widens to almost 13 years, an increase of nearly 8 years. Second, we find that there is a growing gap by lifetime income in projected lifetime benefits from programs such as Social Security and Medicare. For the 1930 cohort, the present value of lifetime benefits at age 50 is roughly equal for those in the highest and lowest quintile of lifetime income, as those at the top receive more from Social Security while those at the bottom receive more from Disability Insurance, Supplemental Security Income, and Medicaid. For the 1960 cohort, by contrast, there is a $130,000 gap in benefits between the highest and lowest quintiles, as those in the top quintile are increasingly likely to receive benefits over longer periods of time, relative to those at the bottom. Finally, we show that a number of commonly-discussed Social Security reforms would make the program more progressive, although their impact on progressivity tends to be small compared to the changes arising due to differential changes in life expectancy.

Policy Reform Simulations

Our final goal is to analyze policy reforms to determine how they would affect the progressivity of government programs and interact with projected changes in life expectancy. The reforms we simulate – five that affect Social Security and one that affects Medicare – were chosen because they are either frequently mentioned in policy discussions or meet objectives with which many stakeholders would agree. Unfortunately, the structure of the FEM (Future Elderly Model) made it impossible to simulate certain reforms, such as raising the Social Security maximum taxable earnings amount. The policy simulations we study include: raising the Social Security EEA (Early Eligibility Age) by 2 years (to age 64), raising the Social Security NRA (Normal Retirement Age) by 3 years (to age 70); reducing the cost-of- living adjustment applied to benefits by 0.2 percent per year; reducing the top PIA factor by one- third (from a 15% to 10% rate); reducing the top PIA factor (Primary Insurance Amount) to 0 above median AIME (Average Indexed Monthly Earnings); and raising the Medicare eligibility age by 2 years (to age 67).

There are two mechanisms by which a policy change may translate into change in
benefits. The first, which can be characterized as the “mechanical effect,” results directly from the policy change, holding behavior constant. For example, if the NRA were raised by 3 years, a worker claiming benefits at age 67 would see the monthly benefit amount fall from 100 percent of the PIA to 80 percent, experiencing a 20 percent reduction in benefits. The second channel, which can be characterized as the “behavioral effect,” results from changes in individual behavior in response to the policy. For example, the individual may claim Social Security later, work longer, or be more likely to claim DI. These responses can be captured by the FEM.

We show results for all reforms on Table 4, reporting only the change in net benefits as a fraction of inclusive wealth for brevity. We begin with the increase in the EEA. At first glance, it might seem that this policy would have little effect on the present value of benefits given the common belief that the actuarial adjustment is roughly actuarially fair. We find that this reform raises net benefits as a share of wealth by 0.1 for males in the lowest quintile under the 1960 mortality regime and by 0.4 for males in the highest quintile. Under our assumptions, the actuarial adjustment for delayed claiming is slightly more than fair, so when individuals are forced to claim later by this policy change, lifetime benefits increase, particularly for high- income individuals who have longer life expectancies. The policy change is thus mildly regressive, although its effects are fairly small.

Raising the NRA has a much bigger effect – we estimate that lifetime Social Security benefits fall by $30,000 (or 25% of the pre-reform value) for the lowest quintile of males in the 1960 mortality regime and by $59,000 (20%) for the highest quintile. The percentage drop need not be the same in the two quintiles because the behavioral response of the two groups to the policy change (captured by the FEM) could differ; also, the same response – say, postponing retirement and claiming by one year – could have a different effect on lifetime benefits because of differences in life expectancy. As low income males experience the larger percentage drop in benefits, this policy might be considered regressive. Yet the policy change reduces benefits as a share of wealth by 4.8 percent for males in the lowest quintile and by 5.2 percent for males in the highest quintile, as the larger dollar loss for high income males ends up being a slightly larger share of their lifetime wealth (as captured by our inclusive wealth measure). Viewed by this metric, the policy change is progressive. Thus, the progressivity of this policy change is somewhat sensitive to the particular measure used.

Reducing the cost-of-living adjustment (COLA) has a modest effect on benefits, reducing them by 0.4 percent of wealth for males in the lowest quintile under the 1960 mortality regime and by 0.6 percent for males in the highest quintile. The larger effect for high income men is due to their longer life expectancy, since the effect of a lower COLA is cumulative over time. Reducing the top PIA factor by one-third has a fairly modest impact of 0.1 percent of wealth for low-income males and 0.3 percent for high-income males; the larger effect on high-income males is expected, since the top PIA factor applies only to earnings past the second bend point of AIME (e.g., at higher earning levels). A related policy with a much bigger impact is reducing the top factor to zero and moving the second bendpoint to the median of AIME. This policy, which would reduce benefits for the top half of earners, is chosen an as example of a substantial benefit cut designed to have a smaller impact on low earners. We find that this policy would reduce benefits as a share of wealth by 1.1 percent for males in the lowest quintile and by 3.4 percent for men in the highest quintile. Finally, we simulate raising the Medicare eligibility age. This policy has a fairly similar effect across quintiles in dollar terms, reducing lifetime Medicare benefits by $8,000 for low-income males in the 1960 mortality regime and by $7,000 for high- income males. Measured as a share of inclusive wealth, there is a loss of 1.4 percent for low- income males and 0.5 percent for high-income males, indicating a regressive policy.

Overall, most of these policy changes would make overall net benefits more progressive. The exceptions are raising the EEA or the Medicare eligibility age, which make benefits less progressive. When compared to the changes in progressivity occurring due to mortality trends, however, the effect of these policies on progressivity is generally fairly small. For example, consider the policy reducing the top PIA factor to zero above median AIME, which is the most progressive of the policies we simulated. Absent any policy change, the gap in lifetime Social Security benefits between males in the highest and lowest income quintiles grows from $103,000 in the 1930 mortality regime to $173,000 in the 1960 mortality regime. Implementing this policy would eliminate 60 percent of the increase, so that the gap under the 1960 regime would be $131,000. This illustrates the scale of the policy reform that would be needed to counteract the changes in progressivity of government benefit programs that we project are occurring due to the widening gap in life expectancy.

From the Discussion

Life expectancy at older ages has been growing steadily in the U.S. over the past several decades. Yet there is growing awareness that these gains are not being shared equally. Our study confirms a substantial increase in the life expectancy gap between those with higher and lower income. For men, we project that the gap in life expectancy at age 50 between males in the highest and lowest quintiles of lifetime income will grow from 5 years for the 1930 cohort to nearly 13 years for the 1960 cohort. Estimates for women (from Committee, 2015) are somewhat less reliable, but show a similar if not larger change over time.

We also assess the effects of the growing gap in life expectancies among older adults on the major entitlement programs. The larger life expectancy gap means that higher-income people will increasingly collect Social Security, Medicare, and other benefits over more years than will lower-income people. We estimate the value of net lifetime benefits for different income groups from Social Security, Disability Insurance, Supplemental Security Income, Medicare, and Medicaid. Our estimates suggest that these net lifetime benefits are becoming significantly less progressive over time because of the disproportionate life expectancy gains among higher-income adults. The changes in life expectancy between the 1930 and 1960 mortality regimes generate an increase in benefits equivalent to an increase of 6.9 percent of wealth (measured at age 50) for men in the highest income quintile, while benefits for men in the lowest income quintile are essentially unchanged. While we caution that one cannot rely too heavily on the specific numbers we estimate, since these are projections that necessarily rely on assumptions and our analysis does not allow us to construct confidence intervals around these projections, our results clearly indicate that lifetime benefits are increasingly accruing to those in the top of the income distribution due to the widening gap in life expectancy.

We then consider how the differential changes in mortality would affect analyses of some possible reforms to major entitlement programs in the face of population aging. For example, many proposals to increase the normal retirement age under Social Security are motivated by the rise in mean life expectancy. The mean, however, masks substantial differences in mortality changes across income groups. We show the impact of that proposal and other possible Social Security and Medicare reforms on lifetime benefits across income groups and in a manner that reflects their different life expectancy trajectories. We find that while there are policy reforms that tend to raise the progressivity of government programs, the effect of these reforms are fairly small when viewed next to the reduction in progressivity that is occurring due to the growing gap in life expectancy. This suggests that policy changes that (alone or in combination) are more progressive than those we simulate here would be needed to undo the effect of the widening longevity gap on the progressivity of government programs.

Social Security, Medicare, and the other programs included in our study face rising expenditures over time, straining the ability of existing revenue sources to fully fund benefit promises at current tax rate. The US is far from unique in this regard – rising longevity, falling birth rates, and slowing economic growth threaten the long-term solvency of entitlement programs in many countries, particularly where financed via a pay-as-you-go mechanism or out of general revenues. Many countries have already implemented reforms to public pension, disability insurance, and other social insurance programs, for example by raising retirement ages or altering benefit formulas in a way that reduces program generosity, and many countries continue to contemplate implementing (further) reforms. As policy makers continue to debate the future of social programs in the United States, they would do well to consider the welfare implications not only of improved longevity, but also the increasing gap in life expectancy by socioeconomic status.


This paper was authored by a team of prominent economists (some names you may recognize) who are members of the National Academies of Sciences’ Committee on the Long-Run Macroeconomic Effects of the Aging U.S. Population. It is an important paper because it demonstrates the disproportionate distribution of Social Security, Medicare, and other public benefits to wealthier individuals by virtue of the fact that they live longer and thus collect benefits for a longer period of time. Since the life expectancy differential is increasing, the inequity of distribution of benefits is also increasing.

This team begins with the basic assumption that entitlement programs, especially Social Security and Medicare, are growing too rapidly and will create an excess burden for the federal budget. In their concluding remarks they note that other nations have raised retirement ages and reduced generosity of benefits. So this analysis seems to aim for that same goal of controlling spending. Thus they direct their consideration to various proposals designed to reduce federal spending for these programs while not giving consideration to increasing revenues to cover the anticipated growth in spending. We’ll get back to this in a moment.

It is widely acknowledged that, over the past several decades, there has been an unacceptable transfer of income and wealth upwards such that the wealthier are becoming much wealthier while the incomes and wealth accumulation of America’s working families have largely stagnated. Public policies, especially tax policies, are now required to correct these inequities.

The authors of this report do understand the inequities and thus describe how the various proposals to reduce spending on entitlements are either regressive or progressive. They conclude, “most of these policy changes would make overall net benefits more progressive. The exceptions are raising the EEA (Early Eligibility Age for Social Security) or the Medicare eligibility age, which make benefits less progressive.”

An interesting observation is that they did not evaluate one of the more important proposals – raising the Social Security maximum taxable earnings amount – simply because the economic model they selected (Future Elderly Model or FEM) made it impossible to do such a simulation. Since this is one of the more important proposals that would improve progressivity, you would think that they might look for or create a model that would take this into consideration, but then this would be counter to their goal of decreasing spending rather than increasing revenues.

In decreasing spending they are concerned about the progressivity of the various proposals since currently the wealthier are favored because of their increased longevity resulting in greater accumulated benefits, thus reductions in spending should apply more to the wealthier individuals. But their concern is about relative progressivity; they still anticipate that reductions would also occur amongst those with lower incomes and wealth, except at a lesser percentage.

The difficulty is that they are addressing the dubious problem as to whether we can afford the increases in entitlement spending, which would require greater revenues, when the problem that they should be addressing is whether or not the benefits are adequate for America’s workforce. Clearly, reducing retirement and health benefits for the workforce is the wrong direction, especially considering the background of wealth moving to the top.

So instead of policies designed to reduce spending, we should be looking at policies designed to ensure adequate retirement and health security. That means protecting and expanding benefits for those with lower incomes (e.g., improving Medicare). Wealthier individuals do not need expanded benefits, but in an egalitarian system it is far easier to provide the same benefits to everyone. To make the system more equitable, revenues from the wealthier individuals with higher incomes should be increased. For those who say that the “economy cannot afford it,” that’s nonsense. Social Security checks are spent, moving that money into the economy, and health care is one of the largest, most robust sectors of the economy, which is supported by the Medicare program. The economy thrives on Social Security and Medicare.

Unfortunately, right now the Trump administration is recommending the opposite. The tax policy recommendations released this week would greatly reduce revenues from those with higher incomes and totally eliminate estate taxes which tax wealth.

The groupthink of this committee is on target to be concerned about the progressivity of entitlement reform, but they have totally missed the boat by suggesting that harmful reductions in benefits for the majority are acceptable merely because the wealthy have a greater percentage reduction. Though that percentage reduction may be greater, the accumulated benefits for the wealthy still remain much greater, and they certainly are not harmed by these reductions, as those with little wealth are. But instead of increasing administrative complexity by selectively reducing benefits for the wealthy, we can offset that inequity by increasing revenues from the wealthy.

We can afford to ensure that absolutely everyone has a decent retirement, including freedom from fear of medical debt though a well financed Medicare program (make that a Medicare program for everyone). Normative economics needs to have a prominent place in groupthink deliberations.