I am deeply honored to be here with you today to deliver the 2026 Martin Feldstein lecture. I first met Marty more than 30 years ago in the Fall of 1995 as a graduate student in his Public Economics course. I learned so much from him during that course about Social Security and many other topics including individual and corporate income taxation, unemployment insurance, and public goods. I also was fortunate during the next three years of my graduate studies to serve as a Teaching Fellow in both his Principles of Economics and American Economic Policy courses and to have him as one of my advisors on my PhD research. He was a phenomenal teacher and advisor who cared deeply about all of his students. This dedication extended to the entire academic economics profession as evidenced by his inspiring leadership for many years of the NBER.

Marty was a prolific scholar who authored nearly 400 publications on a broad set of topics from his time as a graduate student at Oxford beginning in the early 1960s. In many of these publications, Social Security was the primary focus, including his distinguished lecture at the American Economic Association meetings in January 1996 titled “The Missing Piece in Policy Analysis: Social Security Reform.” In this lecture, Marty made a convincing case that Social Security’s steadily deteriorating financial condition would force major reforms in the future: “whether these reforms are good or bad … will depend in part on whether we, as economists, provide the appropriate analytical framework for analyzing reform alternatives.” And Marty did his best to contribute to this cause by writing (with various coauthors) another 29 papers on Social Security during the next 10 years, including many that highlighted the economic case for reform.

This message resonated far outside of the ivory tower and with policymakers on both sides of the political aisle. Consistent with this, both President Bill Clinton and President George W. Bush highlighted the need to strengthen Social Security’s finances in their (1998 and 2005, respectively) State of the Union addresses. As both presidents and Marty pointed out, delaying reform would make the ultimate and inevitable changes to Social Security much more painful than they otherwise might be. And federal finances at this time were much stronger than today; the ratio of federal debt held by the public to GDP declined from 47 percent in 1996 to 36 percent in 2005. But despite efforts by both administrations and by other policymakers in Washington, D.C. in the late 1990s and early 2000s, federal policy changes to improve Social Security’s finances remained elusive and no significant changes were made.

In the two decades since, Social Security has largely been absent from substantive policy discussions in Washington, D.C., despite the fact that the projected date of the program’s insolvency has come steadily closer. This is no doubt in part because any reform that improves the system’s finances would include some combination of tax increases and benefit reductions (as the last major program changes in 1983 did). But the most recent projections from the Social Security Administration indicate that Social Security’s revenues and trust fund assets will be insufficient to finance retirement and survivors benefits for 70 million beneficiaries of OASI (Old-Age and Survivors Insurance) in just six years. Absent any policy change before then, their benefits would be automatically lowered by an estimated 22 percent. This date of insolvency could be extended by about two years until 2034 if the program’s OASI and DI (Disability Insurance) trust funds were merged, when benefits would be lowered by about 20 percent for the projected 81 million OASDI recipients.

Given these projections and the financial hardship these changes would impose on many Social Security recipients, it is natural to expect policymakers to begin work in earnest soon on legislation that could address the program’s structural deficit. But as Marty recognized in his AEA lecture more than 30 years ago, Social Security reform is much less appealing to policymakers now than it would have been then since the magnitude of the required tax increases and/or benefit reductions are substantially greater today and because a much larger fraction of the population is receiving Social Security or will be soon. Further amplifying this challenge, the overall federal budget is in much worse financial shape today than it was then, with debt held by the public at approximately 100 percent of GDP and the recent deficit at about 6 percent of GDP.

Before proceeding to a discussion of reform options, I will briefly summarize the legislative history of the OASDI program. In August of 1935, the Social Security Act was passed. The program paid benefits to retired workers aged 65 and up and was financed with a 2 percent payroll tax on the first $3,000 in annual earnings. Ten years later, there were 1.1 million Social Security recipients and 46 million covered workers. In the subsequent years, a series of reforms expanded the set of individuals who were covered by the program, added disability insurance to the program, increased the tax rate and the taxable wage base, introduced the option of claiming early retirement benefits at age 62, and changed the calculation and indexing of benefits.

By 1980, the number of OASDI recipients had increased thirtyfold (to 35 million), and the ratio of covered workers to recipients had fallen from about 40:1 to just 3.2:1. Social Security was running modest annual deficits at that time and its trust fund was gradually drifting toward zero, though the financing gap was much smaller than it is today. Furthermore, the overall federal financial outlook was much healthier with the debt-to-GDP ratio at just 28 percent of GDP.

Policymakers responded to several consecutive years of annual deficits in Social Security by passing legislation in 1983 that increased the full retirement age, accelerated planned payroll tax increases, and introduced the taxation of benefits for high-income recipients. The most important of these changes for the program’s long-term financial outlook was the gradual increase in the program’s full retirement age from 65 to 67. One key rationale for this change was the several-year increase in life expectancy since the program was created, which reduced the number of workers per recipient. The political opposition to this benefit cut was somewhat limited because benefits for individuals born in 1937 or earlier (aged 46 or older in 1983) were unaffected by the policy change while those between the ages of 29 and 45 were only partially affected (with a retirement age increase of one year or less). The full two-year retirement age increase applied only to individuals aged 23 or younger at the time. Because of this gradual change, individuals who were affected had many years to prepare and adjust their savings and/or retirement plans accordingly.

Since the passage of the 1983 Social Security amendments, there have been no changes to the program that have significantly altered its revenues or expenditures. As a result, the declining ratio of covered workers to Social Security beneficiaries summarized in Figure 1 has resulted in growing financial pressure for the program. As the figure shows, this ratio actually rose somewhat after 1983 due to the labor market entry of the baby boom cohort but has been declining for the past 25 years and is projected to fall still further in the coming years as a result of declining fertility, rising life expectancy, and reduced rates of immigration relative to previous years.

An additional source of pressure for Social Security’s finances in recent decades is that average benefits have become increasingly generous over time. For example, and as shown in Figure 2, the average, inflation-adjusted Social Security benefit for retired workers increased by more than 150 percent from 1960 to 2025. This increase has been financed both by an increase over time in the OASDI payroll tax rate and in the program’s taxable wage base. More specifically, the tax rate stood at 6.0 percent in 1960 versus 12.4 percent today. Similarly, the ratio of OASDI taxable earnings to workers’ average annual earnings in the economy was just 1.2 in 1960 versus about 2.4 today. Both of these policy levers have been essentially unchanged for decades, however, with the 12.4 percent tax rate unchanged since 1990 and the ratio of 2.4 in effect since the mid-1980s.

The 1983 Social Security amendments substantially improved Social Security finances and, as shown in Figure 3, coincided with a significant increase in the ratio of OASDI trust fund assets to program expenditures. This ratio increased from a low of just 0.2 in 1983 to a high of 3.6 by 2008. This measure of financial health has declined steadily in the years since and is projected by SSA to reach 0 by 2032 (for OASI only) or 2034 (for OASI and DI combined). At the time of the 1983 amendments, Social Security was projected to remain solvent for another 75 years (through 2058). For multiple reasons, including rising earnings inequality and slower-than-expected growth in average earnings during the past 40 years, the program is on track to become insolvent 25 years earlier than was expected back in 1983.

The magnitude of the gap between program revenues and program expenditures will be substantially greater in the next several years than it was leading up to the 1983 amendments. Additionally, this gap is projected to increase substantially over the next few decades, as shown in Figure 4, which assumes that benefit payments are unaffected even after trust fund expiration. The Social Security deficit as a share of GDP is projected to double from 2025 (0.74 percent of GDP) to 2055 (1.48 percent of GDP) and to triple by 2075 (2.22 percent of GDP). While there is, of course, considerable uncertainty about the magnitude of this gap multiple decades into the future, the combination of declining fertility, rising life expectancy, and declining immigration is the key driver of these projections.

While Figure 4 assumes that benefit payments are unaffected by trust fund expiration, this is not accurate. More specifically, if policymakers do not act before the trust fund reaches zero, there will be an instantaneous benefit cut required by law of 22 percent for those receiving retirement and survivors’ benefits. The magnitude of this benefit cut grows steadily over time, reaching 37 percent by 2080, which underscores the severity of the financing challenge. But this challenge has been known for many years—indeed Marty warned about this in his AEA lecture more than 30 years ago.

To make this financing challenge even more difficult, the federal debt is now much larger relative to the size of the economy than it was in 1983. There were large federal budget deficits in the early 1980s too, but the debt-to-GDP ratio was less than 30 percent then versus approximately 100 percent today, as shown in Figure 5. This figure, which utilizes data from the Congressional Budget Office (CBO), includes historical data from 1940 to the present and projections through 2055. The current debt-to-GDP ratio is similar to the level near the end of World War II at 100 percent and is projected to increase to 175 percent of GDP over the next 30 years (Figure 5). And according to CBO, more than half of this projected increase is attributable to Social Security, since these projections also assume that benefits will be paid even if no policy changes are made and the trust fund reaches zero. It is also worth noting that these projections do not incorporate the effects of possible future recessions or other unexpected events such as we recently experienced with the pandemic.

Before proceeding to a discussion of Social Security reform options, it is useful to summarize the key policy levers that influence revenues and expenditures for the program. Revenues for the program last year (2025) were $1.45 trillion, with 91 percent of this coming from the 12.4 percent payroll tax imposed on each worker’s taxable earnings, up to a maximum of $184,500. This tax is paid half by the employer and half by the employee, with the entire 12.4 percent paid by self-employed workers. The taxation of Social Security benefits for high-income recipients provides an additional 4 percent of OASDI revenue, with interest on the accumulated trust fund accounting for the remaining 5 percent. Social Security expenditures were 11 percent higher than revenues at $1.61 trillion in 2025, with retired workers and their dependents receiving 84 percent of program expenditures. The remaining 16 percent is divided approximately equally between benefits to disabled workers and their dependents (9 percent) and those paid to surviving children and spouses (7 percent).

A key determinant of the benefits that a retired worker or a disabled worker (or their dependents, including survivors) receives is the program’s 90-32-15 benefit formula, which is used to calculate what is called the “primary insurance amount.” The primary insurance amount (PIA) is the monthly benefit amount that a person receives if they claim retirement benefits at their full retirement age or that a disabled worker receives if awarded benefits. A worker’s average indexed monthly earnings (AIME) over their highest 35 (indexed) earnings years are used to calculate the PIA. As shown in Figure 6, the formula replaces 90 percent of the first $1,286 of a worker’s AIME, 32 percent of the next $6,463 (if applicable), and 15 percent of any remaining AIME above $7,749. The two “bend points” in this benefit formula are indexed to average wage growth each year. While the formula appears to be quite progressive, with low-income workers replacing a much higher share of their earnings than their high-income counterparts, previous research has shown that differences in life expectancy undo much of that progressivity.

The Social Security benefits that a person receives after claiming are then increased each year (to adjust for cost-of-living increases) with the Consumer Price Index that is calculated each month by the US Bureau of Labor Statistics. There is an actuarial adjustment for Social Security’s retired worker benefits as well that rewards those who delay claiming with a higher monthly benefit. Seventy percent of the PIA is paid to those claiming benefits at age 62, and this steadily increases to reach 124 percent of the PIA at age 70. Workers with lower incomes and shorter life expectancy tend to claim their retirement benefits earlier than do those with higher incomes and longer life expectancy.

In June of 2026 there were 71 million Americans receiving Social Security benefits for an average benefit of approximately $1,900. Retired workers are receiving an average of $2,100 per month versus $1,600 per month for the average disabled worker. The lower benefits for disabled workers largely reflect the fact that they had lower average earnings during their work years than did retired workers. Widows and widowers are between those two extremes, receiving $1,900 a month, while spouses and children of retired and disabled workers are paid about $1,000 per month. More than one in five Americans receive Social Security benefits and the fraction of the population receiving OASDI benefits is steadily increasing given the changing demographics in the country.

One little-noticed feature of this benefit formula is that Social Security taxes are a “pure tax” for a worker in any year outside of the 35 best earning years. For example, and as Marty often pointed out when explaining this benefit formula, it is plausible that many workers would have their highest earnings from age 25 to age 59. In this case, the program would impose the highest effective tax rate on workers in their early twenties or in their sixties and seventies. This coincides with the ages at which labor supply elasticities are highest, thus amplifying the distortionary effects of this revenue source.

As I mentioned previously, there has been a substantial increase over time in the taxable wage cap for Social Security, which is now about 2.4 times average earnings versus just 1.2 times average earnings in 1960. This change has led to a larger increase in the Social Security benefits that retired or disabled high-income workers receive relative to low and middle-income workers. Furthermore, the actuarial adjustment of retirement benefits encourages those with longer life expectancy to delay claiming. The 1983 amendments increased the adjustment beyond the full retirement age from 3 percent to 8 percent per year, and high-income workers have differentially benefited. As a result, a high-income worker who claims retirement benefits at age 70 this year will receive more than $5,000 per month from Social Security. The two policy changes—the increase in the taxable wage base and in the actuarial adjustment beyond the full retirement age—mean that average retirement benefits for high-income workers have increased much more rapidly than for other workers.

Figure 7 summarizes the distribution of monthly Social Security benefits in 2025 among the 54 million retired workers. As the figure shows, approximately 8 percent of them have a monthly benefit that is less than $900 a month, whereas about 10 percent of them have a benefit over $3,300 a month. In dollar terms, a 22 percent across-the-board benefit cut (which would result if the trust fund is allowed to reach zero in 2032 and no other policy changes occur) would be larger for the high-income group captured on the right on this figure than for the low-income group summarized on the left. Put another way, a Social Security recipient with a $900 monthly benefit would experience a $198 benefit cut that would almost surely put considerable pressure on their spending for housing, food, and medical care.

The fact that high-income workers in the US receive large benefits from Social Security was nicely summarized recently by the Committee for a Responsible Federal Budget. The data in Figure 8 lists the maximum public pension amount for a couple in 2024 in 10 industrialized countries. More specifically, if a person earned above the taxable maximum throughout their working years, and they claimed at the full retirement age, and their spouse claimed at the full retirement age, what would their combined Social Security benefit be? Here in the US, it’s $94,000 a year. If you take that 24 percent actuarial adjustment for delaying until age 70, it’s about $117,000 per year. Even though the US, relative to other industrialized countries, has a smaller ratio of people 65 and older to the overall population, the fact that our Social Security benefits are so much more generous makes the financing challenge for our Social Security program just as challenging as in France, Germany, and Italy.

Social Security is arguably the most important source of income for elderly Americans, certainly lower-income elderly Americans. More than half of elderly people get more than half of their income from Social Security. A quarter of elderly people get more than 90 percent. And the average shares vary considerably by age. For example, those aged 80 and up rely on Social Security to a much greater extent than those aged 65 to 69. Not surprisingly, these shares also vary with education, income quintile, race, and gender, with the variation by income quintile especially large. Social Security looms very large for America’s elderly and especially for those with lower incomes throughout their working years. And this is even more true when considering disabled workers.

I would now like to shift gears to discuss reform proposals along with a few criteria for evaluating each proposal relative to the “do-nothing” alternative. The first criterion that I use is, what does any proposal do for the long-run solvency of the program? That seems like an important criterion for policymakers, academics, and others to consider when evaluating reforms.

An additional criterion is the distributional effect of any reform and the effect on benefit adequacy for workers with very low earnings. The effect of a reform on the incentive to work and to save must also be considered, along with any effects on the rest of the federal budget. For example, to the extent that an increase in the Social Security taxable wage base reduces earnings and employment, this could reduce income tax revenues as well.

A final criterion that I will mention is resilience to changes in the economy. There are some quite significant changes going on currently because of the diffusion of artificial intelligence and the declining share of GDP going to labor income. A reform that makes the program more resilient to these and other potential changes in the future would be more desirable than one that does the opposite. The declining labor share displayed in Figure 9 is likely already putting a strain on Social Security finances, and this is likely to become increasingly the case over time. As the figure shows, the labor income share of GDP is down to just 53.7 percent, which is its lowest level since this series began about 80 years ago. Given that Social Security is so dependent on taxable earnings to finance its benefits, these changes put the program’s solvency at even greater risk.

In general, Social Security is a complicated program that many workers may not fully understand. It is incumbent on economists and others who research Social Security and on policymakers to explain the program and the effects of potential reforms so that the program is less confusing for the typical worker and their dependents. All else equal, a simpler, more transparent reform is preferable to one that’s complicated and opaque.

There are many decisions facing policymakers in the very near future. The first is whether any changes should be made. In the absence of any policy change, tens of millions of Social Security recipients will see their benefits substantially lowered in just six years. This would surely impose significant hardship on a large fraction of these recipients, including especially those with low incomes, for whom Social Security is the primary and often the only source of income. On the other hand, the debt-to-GDP ratio would be much lower than projected and GDP would be higher as well (since individuals would likely increase their labor supply in response to the reduction in current or future benefits).

One alternative to this “no reform” scenario could involve no change to Social Security itself but instead legislation that would allow Social Security to use general revenues to finance its large and increasing deficits after the trust fund expires. One potential concern with this is that it would amplify pressure on other categories of federal spending. Social Security spending already represents nearly 5 percent and it is projected to increase to more than 6 percent on its current path. This represents a very large and growing share of projected federal spending, which would likely exert pressure on other categories of spending such as Medicare, Medicaid, and national defense. Proponents of this approach would argue that plausible increases in income tax rates would likely place less of an economic burden on lower-income individuals than would an increase in Social Security’s payroll tax rate. Given the increase in earnings and life expectancy inequality since the last major changes to Social Security were made 43 years ago, one could argue that high-income individuals should bear relatively more of the burden from financing Social Security in the future.

Should the reform retain a pay-as-you-go structure rather than move to a system based partially or entirely on individual accounts? If so, it would likely involve some reduction in benefits from what has been promised. The 1983 amendments only gradually reduced benefits by increasing the full retirement age in two-month increments from 65 to 67 and delaying its implementation for almost 20 years. This change applied to both high- and low-income workers, though a future reform could instead limit benefit reductions to those with high incomes, insulating low- and middle-income recipients. Whether through an across-the-board change such as a retirement age increase or a targeted change such as a reduction in indexing for high-income recipients, gradual benefit reductions will not be sufficient in the current situation given the magnitude of the Social Security financing gap.

An alternative or complementary set of policies could increase revenues for the program through an expansion of the OASDI earnings tax base beyond the current $184,500 or through the inclusion of other sources of income, including interest, dividends, and capital gains. The Social Security tax rate of 12.4 percent has not increased in 36 years despite a substantial increase in life expectancy during that time. Increasing either the taxable wage base or the tax rate would retain the current structure.

There are outside-the-box options that have been proposed in recent years too. For example, Senators Bill Cassidy and Tim Kaine recently proposed a plan called the Big Idea Plan, under which Social Security would borrow $300 billion a year for five years and invest those funds in equity markets. Any returns on those investments in excess of the cost of servicing the debt would be used to finance Social Security deficits.

As economists, policymakers, and others deliberate in the months and years ahead about the best path forward, it is worth considering what has changed since the last major reform in 1983. Debt held by the public is now about three times higher. There has also been a significant increase in earnings, income, and wealth inequality, and a lower share of earnings today is subject to OASDI taxes than back then. The share of all earnings subject to OASDI taxes was about 83 percent last year versus 90 percent in 1983. The labor income share has declined quite significantly, reaching a low recently of just 53.7 percent. This further reduces the available revenues for Social Security.

It is also instructive to reflect on how little political pushback there has been on the gradual increase in the full retirement age, which was equivalent basically to a 12.5 percent benefit cut for those born in 1960 or later relative to what they would have received in the absence of a policy change. As for increasing the taxable wage base, there is precedent, as federal legislation substantially increased the Medicare tax base (by 160 percent) in 1991 and removed the cap on earnings subject to Medicare taxation in 1994. While there has been little evidence of a decline in labor supply resulting from this tax increase, a corresponding increase for Social Security would involve a tax increase more than four times as large (12.4 percent versus 2.9 percent).

It is also true that since 1983, there has been a much larger increase in OASDI benefits for high-income workers than for the average worker. The two reasons for this are that the taxable wage base has grown substantially and the reward for delaying claiming beyond the full retirement age has also increased. As a result, the maximum retired worker benefit has grown by more than twice as much (in percentage terms) as the average benefit since 1990. This, along with the much lower increase in life expectancy among low-income workers, further strengthens the case for any policy-induced reduction in benefits to be larger for high-income recipients.

In addition to the tax rate, the taxable wage base, and the full retirement age, there are other plausible levers to improve the financing of Social Security. For example, the program’s 90-32-15 benefit formula has two bend points that are indexed to average wage growth. If policymakers chose to slow the growth in the 32-15 bend point and leave unchanged the 90-32 bend point, this would reduce benefits for high-income recipients while leaving benefits for low- and middle-income recipients unchanged. And the impact of this change would be gradual, allowing affected recipients to adjust their planned date of retirement, savings, and so forth to cushion any adverse effects.

Many subtle changes in Social Security benefits, in taxes, or in the taxation of benefits are also possible, such as increasing the number of years of earnings used in calculating benefits or altering the indexing or actuarial adjustment of benefits. Reforms like the ones discussed above could also be bundled with changes that provide a minimum benefit for retired and disabled workers. The incremental cost of such a change is of course dependent on the size of this minimum benefit, but the change could help cushion low-income workers against the effects of a potentially more turbulent labor market in the future.

An adjustment as gradual as the one policymakers made in 1983 would not be nearly enough to address the magnitude of Social Security’s financial shortfall today. According to SSA actuaries, a nearly 5-percentage-point increase (from 12.4 to 17.3 percent) in the tax rate would be necessary to keep the program solvent through 2100. And that increase would not be sufficient to achieve sustainable solvency, since the trust fund would be projected to reach zero again around that time.

Perhaps the most widely discussed policy change has been eliminating the cap on earnings subject to Social Security taxes, as was done with Medicare in 1994. While this would eliminate a substantial fraction of the financing gap (assuming that it was a “pure tax” that did not increase Social Security benefits), it would push marginal tax rates well beyond 60 percent for many high-income individuals in states with relatively high state tax rates, such as California and New York. This increase would potentially reduce labor supply and thus other revenue such as from individual income taxation.

All of the policy levers that I have discussed above would involve substantial sacrifices for a large segment of the population. The current structure of Social Security is not designed to weather the economic effects of significant changes such as declining fertility, rising life expectancy, declining immigration, rising earnings inequality, and a declining labor income share. All of these forces are conspiring to create a perfect storm for Social Security financing.

One possibility that may be more palatable to policymakers would involve a partial reform that financed a growing fraction of the financing gap with Social Security benefit reductions and/or tax increases and the remainder with funds from general revenues. General revenues could, for example, contribute a fixed percentage of the program’s revenues, as is done for Medicare Part B. These revenues could help to cover Social Security’s deficit at the time of trust fund expiration. At the same time, gradual reductions in benefits and/or increases in taxes could be implemented to finance an increasing share of the current financing gap going forward. Such a gradual change to Social Security benefits and/or taxes would make it easier for current and future beneficiaries to plan for the future.

There is of course considerable uncertainty about how many of the factors that influence Social Security financing will evolve in the future, such as productivity, immigration, life expectancy, and much more. Adding to this is the uncertain impact of artificial intelligence and adjacent technologies on workers’ earnings and the labor market. In light of this, it seems that any changes to Social Security that make the program more resilient to unanticipated changes will have substantial economic benefits now and in the future.