Key Points:

• We project that the Social Security Old-Age and Survivors Insurance (OASI) Trust Fund will deplete in 2032. If OASI and DI (Disability Insurance) are treated as a combined, pooled trust fund, depletion occurs two years later, in 2034. Annual revenue can then cover only 83 percent of scheduled benefits, decreasing to 64 percent over the next 75 years. We consider five different reform options that vary in the amount of tax increases and benefit cuts.

• Augmenting traditional policy metrics with more comprehensive academic-level modeling highlights two shortcomings of traditional policy metrics. First, reforms that reduce the long-run debt-GDP ratio the most often produce the smallest macroeconomic gains because federal debt does not comprehensively measure total future obligations. Second, conventional distributional summaries of fully-phased-in benefit reductions can suggest benefit losses even for future households who gain from the policy change. A comprehensive analysis, therefore, starts with microeconomic foundations to explicitly account for taxes and transfers between and within generations, and explicitly models how changes in incentives and risk pooling impact growth and human welfare in both the short and long run.

• Despite receiving much less attention than tax reform, changes to Social Security (and entitlement reform more broadly) can produce much larger changes to economic growth and individual welfare.

Background

Social Security is the primary source of retirement income for over half of retired Americans. Its old-age benefits are protected against inflation through annual cost-of-living adjustments and are paid for life, providing a valuable inflation-indexed lifetime annuity. But demographic change and rising benefit obligations per retiree are straining Social Security’s finances.

Under current law, we project that the Social Security OASI (Old-Age and Survivors Insurance) Trust Fund will be depleted in 2032. If OASI and DI (Disability Insurance) are treated as a combined, pooled trust fund, depletion occurs two years later, in 2034. At that point, payable benefits would fall to 83 percent of scheduled benefits, continuing to decline to 64 percent by 2100, the end of the 75-year projection window. In general, restoring balance requires some combination of higher revenues, slower benefit growth, and reduced initial benefit generosity for future cohorts. But how much does each lever contribute, and what are the trade-offs?

Current Policy and Stylized Reforms

Table 1 summarizes the provisions of five policy bundles considered in this brief. The bundles span a spectrum: options on the left rely primarily on raising taxes, while options on the right rely primarily on reducing benefits. All provisions not listed for a given option follow current policy. PWBM can now model numerous Social Security reform options, which will be considered in future analyses. The focus of the current brief is laying some of the groundwork for understanding dynamic analysis.

Current Policy

We begin by describing current policy.

OASDI Tax Rate. The combined Old-Age, Survivors, and Disability Insurance (OASDI) payroll tax rate is 12.4 percent, split evenly between employer and employee at 6.2 percent each.

Taxable Maximum. Only earnings up to the taxable maximum — $184,500 in 2026 — are subject to the payroll tax. This threshold adjusts automatically with economy-wide average wage growth over time.

COLA. After benefits are determined at retirement, they grow each year with consumer price inflation as measured by the CPI-W (Consumer Price Index for Urban Wage Earners and Clerical Workers).

PIA Factors. The initial benefit at retirement is determined in several steps. First, a worker’s previous covered wages (up to the taxable maximum in each year) are indexed to economy-wide average wages using the Average Wage Index (AWI), which generally grows faster than inflation. Second, the worker’s Average Indexed Monthly Earnings (AIME) are calculated using the highest 35 years of indexed earnings. Third, the monthly retirement benefit — the Primary Insurance Amount (PIA) — is computed using a progressive formula applied to the AIME: 90 percent of AIME up to the first bend point ($1,286 in 2026), plus 32 percent of AIME between the first and second bend points ($1,286 to $7,749 in 2026), plus 15 percent of AIME above the second bend point. For example, a worker with an AIME of $5,000 per month would receive a PIA of approximately $2,346 per month: (90% × $1,286) + (32% × $3,714) = $1,157 + $1,188. A higher-earning worker with an AIME of $10,000 per month would receive a PIA of approximately $3,563 per month: (90% × $1,286) + (32% × $6,463) + (15% × $2,251) = $1,157 + $2,068 + $338.

Full Retirement Age. The full retirement age (FRA) is currently 67 for workers born in 1960 or later. Workers may claim reduced benefits as early as age 62 or earn delayed retirement credits by waiting until age 70.

Five Options: A - E

Several of the benefit changes described below are phased in gradually over decades for newly eligible beneficiaries, whereas most tax increases take effect immediately. In particular, the tax rate, taxable maximum, donut hole, and COLA provisions begin in 2027. The Appendix provides additional detail on each provision.

Option A mostly relies on revenue increases. It combines a 1.0 percentage point increase in the OASI payroll tax rate (from 12.4 to 13.4 percent) with a higher taxable maximum ($250,000, up from $184,500 in 2026), a half-rate donut hole, and a switch to the Chained CPI for cost-of-living adjustments. By mixing revenue increases with a modest benefit adjustment, Option A spreads the adjustment across taxpayers and beneficiaries.

Option B adds benefit formula changes to the revenue side. It pairs the higher tax rate and taxable maximum with a Chained CPI COLA and a reduction in PIA factors from 90/32/15 to 90/25/8, phased in over 20 years for newly eligible beneficiaries.

Option C shifts the balance further toward benefits. It retains the higher tax rate but replaces the revenue provisions with a PIA factor reduction (90/25/8, phased in over 20 years starting in 2027) and an increase in the full retirement age from 67 to 69 for individuals age 44 or less in 2027 (phased in at two months per year). The two-year increase in the full retirement age still grows by less than the increase in life expectancy over the next 75 years.

Option D combines the higher taxable maximum with a Chained CPI COLA, the same PIA factor reduction (phased in over 20 years), and the retirement age increase, mixing revenue and benefit changes across multiple levers.

Option E is the most benefit-focused bundle. It makes no tax rate change and raises no additional revenue beyond the Chained CPI’s indirect effects. Instead, it applies the largest PIA factor reduction (80/22/5, phased in over 20 years), raises the retirement age to 69, and switches to the Chained CPI, concentrating the adjustment entirely on the benefit side.

Standard Benefit Metrics

Figure 1 shows changes in Social Security benefits for beneficiaries age 65 and older by 2100 (the end of the standard 75-year accounting window), after all benefit changes have been fully phased in. The results are grouped by baseline benefit quintile. The reader can click Options in the legend of Figure 1 to isolate and compare different option combinations. For any given option, notice that the benefit reductions are mostly progressive, with the largest percent reductions occurring among beneficiaries who would otherwise receive larger benefits in retirement.

Option A generates modest, fairly uniform reductions of about 3 to 4 percent across quintiles by 2100. The Chained CPI grows more slowly than the traditional CPI, so annual COLAs shrink slightly, and the effect compounds over long claiming durations. The remaining options produce larger and increasingly progressive benefit cuts. Option B reduces benefits by 6 percent for the lowest quintile and 19 percent for the highest by 2100. Options C and D reach reductions of 20 to 30 percent across quintiles. Option E produces the deepest cuts: 31 percent for the lowest quintile and 40 percent for the highest by 2100.

Figure 6 in the Appendix extends Figure 1 to report changes in lifetime net tax rates — changes in the difference between the present value of payroll taxes paid and benefits received, expressed as a percentage of the present value of lifetime earnings — for the same policies and income groups as Figure 1. By including taxes paid, Figure 6 is more comprehensive than Figure 1. But both figures rank the five options in roughly the same order of severity. Options A and B impose the smallest costs (for example, Option A produces benefit cuts of only 3–4 percent and net tax rates of 1.2–2.5 percent), while Options C through E impose increasingly larger costs (up to 40% benefit cuts; up to 4.2% net tax rates). In other words, the options that cut benefits the most also produce the highest net tax rates.

Figures 1 and 6, however, are both “conventional” metrics that measure direct program flows without behavioral responses, changes in risk pooling and macroeconomic dynamics. The more comprehensive dynamic analysis presented below indicates that future generations facing fully-phased-in benefit reductions gain the most from larger benefit reductions (Figure 1) or net tax increases (Figure 6).

Standard Accounting Metrics

Table 2 reports several standard Social Security metrics. Under the baseline (current policy), the long-range deficit equals 4.75 percent of covered payroll. In words, increasing the current combined employee-employer payroll tax rate from 12.4 percent to 17.15 percent of payroll would provide just enough resources to cover benefits over the next 75 years on a conventional basis, not including any labor market, capital market or health care market responses. We project that the OASI trust fund depletes in 2032 (OASDI in 2034). Payable OASDI benefits (benefits that can be covered by contemporaneous revenue without a trust fund) equal 83 percent, falling to 64 percent of scheduled benefits by 2100.

All five options, A through E, improve the long-range balance but none fully restores solvency, whether measured using the trust fund depletion dates or the 75-year imbalance. The revenue-focused bundles delay depletion the most because revenue increases are generally implemented faster than benefit decreases. Option A delays depletion to 2058. Option B, which mixes revenue and benefit changes, reaches 2038. Option C pushes depletion to 2035, Option D to 2034. Option E barely delays depletion, by just a few months.

However, notice that options that are most effective at delaying the depletion date are the least effective at reducing the 75-year imbalance. Compare Options A and E. Table 2 shows that Option A delays the OASI trust fund depletion to the year 2058 and the OASDI trust fund depletion to the year 2070. That leaves a shortfall (including DI) equal to −1.56 percent of covered payroll. Option E barely moves the trust fund depletion dates but also reduces the 75-year balance to -0.94 percent of covered payroll, more than Option A.

Figures 2 and 3 illuminate the differences between Options A and E in more detail. Figure 2 traces the cumulative OASDI trust fund reserves over time. Figure 3 reports the annual non-interest balance over time. The non-interest balance is a standard metric that presents annual cash flows equal to projected revenue received less benefits paid in a given year. To focus, click Options B, C, and D in the legends of Figures 2 and 3 to temporarily hide them. Click again to unhide.

Notice that the near-term revenue raisers in Option A produce a jump in annual cash flows (Figure 3) that delays the trust fund depletion date more than Option E. However, notice that the negative slope of the trust fund in Figure 2 for Option A near its depletion year is steeper than for Option E. The reason is that, relative to Option E, Option A does little to contain costs over time. So, by 2055, the annual balance (cash flows) in Option A falls below Option E, leading to a faster rate of decumulation in Option A.

Under current policy (baseline), Social Security program costs rise over time for several reasons. One important factor is increasing longevity without a commensurate increase in the full retirement age (FRA). The 1983 reforms gradually raised the FRA from 65 to 67, and that transition is now nearly complete; for someone born in 1960, the FRA is 67. Looking ahead, both the Social Security Office of the Chief Actuary and PWBM project further gains in life expectancy between 2026 and 2100, though the magnitudes differ. Using the Chief Actuary’s projections, life expectancy at birth increases by about 7 years for males and 6 years for females, while PWBM projects smaller gains of about 4 years for males and 3 years for females. Conditional on reaching age 65, the Chief Actuary projects roughly a 4-year increase in life expectancy at 65, compared with about 2 years under PWBM. As a result, reforms that do not continue to increase the FRA have a harder time preventing costs per beneficiary from rising over time as beneficiaries collect benefits for longer periods.

Macroeconomic Effects

All five options increase long-run GDP relative to the current-policy baseline, which assumes scheduled benefits are financed with additional federal debt. The effects are small at first and grow over time as two channels accumulate: reduced crowding out of private capital and greater household saving. We first report the outcomes and then explain them in more detail.

Results

Private capital. Lower federal debt frees private savings for productive investment. By 2060, private capital increases by 4.4 percent under Option A and 13.5 percent under Option E. The benefit-focused options generate additional capital growth because reduced expected benefits incentivize households to save more for retirement on their own.

Wages. Higher private capital raises labor productivity and wages. By 2060, wages increase by 1.6 percent under Option A and 5.7 percent under Option E.

Hours worked. Labor hours vary slightly across options. Higher wages create an incentive to work more, but higher payroll taxes offset this by lowering the after-tax wage. By 2060, hours decline by as much as 0.2 percent (Option C) or increase by as much as 0.3 percent (Option E). Option E, which does not raise the payroll tax rate, sees the most favorable labor supply response.

GDP. The combined effects on labor and capital increase GDP for all options in the long run, though some options produce short-run declines. By 2040, most options increase GDP modestly, up to 1.7 percent. By 2060, all options raise GDP, ranging from 2.0 percent (Option B) to 6.1 percent (Option E).

Federal debt. All options reduce the debt-to-GDP ratio relative to baseline. By 2060, the reduction ranges from 12.6 percentage points under Option C to 22.4 percentage points under Option A.

Discussion

The use of a dynamic overlapping generations (OLG) model with idiosyncratic risk, like that used by PWBM, is crucial for understanding macroeconomic effects of entitlement programs, including Social Security, Medicare, Medicaid and, more generally, health care policy. The OLG model explicitly links microeconomic behavior (heterogeneous households, firms, international investors, and government policy) with the macroeconomy (GDP, capital, labor supply, and prices). Virtually all academic analyses of entitlement reform published in leading economics journals use the OLG framework. PWBM scales the OLG model to the size of the U.S. and world economies, including rich modeling of non-stationary demographics, corporate and non-corporate sectors, government tax and spending programs, and a fully endogenous health care sector (public programs and private) that produces large interactions with the rest of the economy. A full treatment of PWBM’s OLG model is beyond the scope of this brief, but several key points deserve attention.

First, notice that the increase of private capital formation and GDP is not proportional to reductions in the amount of federal Treasury debt relative to GDP. Options A and B significantly reduce the debt-to-GDP ratio but have the least impact on private capital formation and GDP. It is generally understood that one could improve Social Security’s finances using general revenue, which increases the amount of federal debt. The point here, however, goes one step further: even the amount of federal debt is not a holistic measure of obligations being passed to future generations. It is well understood by public finance economists that pay-as-you-go financing transfers resources from future generations with the same effects (both microeconomic and macroeconomic) as federal debt. Since tax payments equal benefits at each point in time, pay-as-you-go financing does not require any new explicit federal debt. These pay-as-you-go transfers are instead known as implicit debt (or the “labeling problem”). The microeconomically founded OLG model captures all inter-generational transfers, whether explicit or implicit.

While the literature on pay-as-you-go implicit debt receives very little attention inside Washington DC, its importance is more than just theoretical. Under baseline law, PWBM estimates that pay-as-you-go transfers are twice as large as explicit Treasury debt. Put differently, if these transfers were formally booked, similar to FASB rules for private companies, the US debt-GDP ratio would currently be reported in excess of 300 percent.¹ In the context of Social Security reform, our previous analysis of the Social Security 2100 Act showed that reductions in explicit federal debt stemming from Social Security reform can produce less capital formation and GDP if the reduction in federal debt is more-than offset by increases in pay-as-you-go transfers from future generations.

Options C and D produce better macroeconomic outcomes — despite reducing explicit debt by less — than Options A and B because Options C and D reduce the sum of implicit (pay-as-you-go) and explicit (Treasury) liabilities by more. Despite receiving less attention, entitlement reform generally has a much bigger impact on the macroeconomy and individual lifetime outcomes than most tax changes due to changes in implicit debt.

Second, PWBM’s OLG model explicitly recognizes that payroll taxes are generally less distorting to labor supply than other forms of taxes, provided that Social Security benefits increase in lifetime wages and taxes paid. In fact, this marginal tax-benefit linkage implies that if, purely hypothetically, “r = g”³ and Social Security did not contain intra-generational redistribution, Social Security taxes would behave more like “contributions” than distorting taxes; these contributions, therefore, would largely substitute for private household savings and produce little economic effect.⁴ As a result, policy options that reduce the marginal tax-benefit linkage increase labor supply distortions. These policy options include traditionally more “liberal” policies (Options A and B) that increase taxes for a given level of benefits as well as more “conservative” options (Options D and E) that reduce benefits for the same tax contributions; flat benefit proposals (not shown here) fully delink taxes paid from benefits received. The net effect on labor supply requires computation.⁵

Third, being grounded in microeconomics, the OLG model also incorporates a standard insight from the field of public finance: the incidence and economic impact of a tax does not depend on whether the statutory tax is levied on employees or employers (Rosen and Gayer, 2013; Stiglitz and Rosengard, 2015; Gruber, 2022). While Social Security reforms that mostly target employers instead of employees might have a different public appeal, employers only care about the total cost of labor while employees only care about the total benefit from working. The statutory division of taxes does not matter. Of course, the incidence might fall more on employers or employees, depending on elasticity of labor demand versus supply, with the evidence pointing to it being borne more by labor (Gruber, 1997). But common reduced-form models used in policy analysis often lack a foundation in microeconomics and assume that a statutory tax on employers falls more on capital than labor relative to a statutory tax on employees.

Fourth, PWBM’s model accommodates “good debt” to complement investments (e.g., public infrastructure and early childhood education). Social Security’s implicit debt reduces capital formation because these transfers finance more immediate consumption rather than investments.

Fifth, PWBM’s OLG model recognizes that the private market often fails to provide insurance against major risks, including insurance against variation in wages as well as inflation-indexed lifetime annuities. Explaining the exact reasons for this private market failure is beyond the scope of this brief. But the value of the government provision of risk reduction is captured within a comprehensive lifetime measure of a policy’s impact. The value of risk pooling is often strong enough to make some level of progressive taxation and social insurance optimal, even with a negative fiscal externality from reducing precautionary labor supply and capital savings in the presence of other second-best taxes (Nishiyama and Smetters, 2005, 2007).

These five factors impact how we measure the lifetime impact of policy, which we consider next.

Dynamic Distributional Effects

A policy’s equivalent variation (EV) is the paradigm metric used in the economics literature for capturing the benefit from a policy change on a holistic lifetime basis. These factors include:

• Direct conventionally-measured wins and losses from a policy change at the individual level
• Microeconomic changes in incentives (household labor supply, savings, purchase of health insurance and paying for OOP expenses)
• Changes in the value of social insurance (risk pooling) against idiosyncratic risks
• Changes in the fiscal externality as changes to household labor supply and savings interact with the existing tax and transfer system.
• Macroeconomic changes in prices (in the PWBM model: wages, interest rates and competitive health care premiums)

The equivalent variation measure is much richer in detail than the change in benefits alone shown earlier in Figure 1 (or net taxes shown in Figure 6 in the Appendix). Figure 4 adds the dollarized (2026) equivalent-variation metric (positive y-axis) to the conventional benefit change metric provided earlier in Figure 1 (negative y-axis). Figure 4 focuses on the long-run impact of the policy by considering a person born 25 years after the policy is enacted, who will be age 50 in 2100, at the end of the 75-year window. (The reader can click Options in the legend of Figure 4 to isolate and compare different option combinations.)

The reader can use mouseover to show specific values for each bar. As an example, Figure 5 shows two separate mouseovers for Option E in the “2nd Quintile”. The first mouseover shows a reduction of fully-phased-in benefits of 30 percent while the second mouseover shows a positive EV value (positive y-axis) of $62,300. In words, in the long run, a person in the 2nd-quintile sees a 30 percent reduction in benefits along with a $62,300 one-time increase in value from the policy, reflecting a combination of higher economic growth, an increase in wages, and other factors noted above.

Notice that Figure 4 shows that reforms that produce the smallest reductions in standard fully-phased-in benefits (negative y-axis, same as Figure 1) without dynamics also produce the smallest gains on a lifetime basis with dynamics; conversely, the reforms that produce the greatest reductions in standard fully-phased-in benefits also produce the largest gains on a dynamic basis.

As another example, consider an individual in the “Bottom 20%”. Under Option A, their standard fully-phased-in benefits (negative y-axis) fall by 3 percent (hover your mouse on the blue bar to see the exact values). On a dynamic basis (positive y-axis) lower-income individuals gain $22,800 in lifetime value from Option A. For Option E, a 31 percent loss in standard benefits (negative y-axis) is matched with a $43,300 lifetime gain on a dynamic basis (positive y-axis).

Intuitively, incentives to save, work and purchase health insurance are improved along with macroeconomic outcomes. These factors are not included in standard metrics (Figure 1 or Figure 6) but are captured with dynamics (Figure 4). In sum, conventional metrics (Figure 1 and Figure 6) and more comprehensive dynamic metrics (Figure 4) can produce the opposite sign and magnitudes, even across a fairly wide range of policy options.

Of course, long-run gains often come at the cost of short-run losses. Table 4 reports the EV for each option across age cohorts and income quintiles. The reader can select an option from the pulldown in Table 4. Under Option A, someone age 60 in 2026 in the middle quintile loses $30,745, while someone born in 2051 in that same quintile gains $42,025. Under Option E, the same comparison is a loss of $60,970 versus a gain of $81,932. Younger and future cohorts gain more because they benefit from improved macroeconomic conditions over their entire working lives, whereas older cohorts bear transition costs with fewer years to recoup them.

Table 4 avoids a direct comparison of equivalent-variation welfare metrics with traditional Social Security benefit-change metrics, as such a comparison would be challenging, especially in the short run. For example, individuals in the bottom quintile for Social Security benefits are often not the same individuals in the bottom quintile of earnings at age 60. Individuals in the bottom quintile of Social Security benefits typically had lower realized income throughout most of their working years. In contrast, individuals who happen to be in the bottom quintile of income at age 60 typically had income above the bottom income quintile at earlier ages; their sorting into the bottom quintile at age 60 reflects a combination of exogenous shocks and endogenous labor choices driven by past asset accumulation and even inheritances within a household setting. As a result, there is considerable heterogeneity in household attributes, including wealth, that is naturally “averaged out” in any summary table, including Table 4. In the future, PWBM will consider additional ways to present information that allow for a more direct comparison between holistic equivalent-variation analysis and more traditional metrics. PWBM invites feedback on this issue, since these types of comparisons are intrinsically hard, especially for individuals who are older at the time of reform.

Still, an important lesson emerges from Table 4: achieving the largest long-run gains for future generations does not always require the largest short-run sacrifices. To be sure, it is often the case that achieving the largest long-run gains for future generations requires larger short-run costs to current generations. That inter-generational trade-off generally holds in the hypothetical presence of non-distorting (e.g., lump-sum) taxes and no changes in risk sharing or precautionary savings (or with fully contingent claims). But that trade-off is not singularly determinant in a richer model setting.

For example, select Option A in Table 4. Notice many people alive today at age 60 would be harmed by this policy change, including a change in price indexing of Social Security benefits and macroeconomic prices. Both Table 4 and Figure 4, though, show that Option A delivers the smallest gains to future generations relative to the other options. Now select Option C in Table 4. Most people alive today at age 60 gain from this policy change, mostly driven by macroeconomic gains enjoyed throughout the remainder of their lifetime. The long-run gains from Option C are also larger than those from Option A, despite the smaller short-run sacrifice. While inter-generational transfers are important, other factors noted above — including microeconomic incentives, macroeconomics and changes in risk sharing — also play a role.

Appendix

Policy Provision Details

Tax rate: Increase in the OASI payroll tax rate

This provision increases the Old-Age and Survivors Insurance (OASI) payroll tax rate by 1.0 percentage point, from 12.4 percent to 13.4 percent, beginning in 2027. The higher rate applies to all covered earnings up to the taxable maximum in each year and is shared equally between employers and employees. By raising the contribution rate broadly across the covered workforce, this provision increases program revenues immediately and persistently.

Tax max: Increase in the taxable maximum to $250,000

The taxable maximum ($184,500 as of 2026) for OASI contributions is increased to $250,000 starting in 2027. Earnings above the current-law taxable maximum and up to $250,000 become subject to payroll taxes, but these additional contributions do not generate higher benefits. The provision raises revenues primarily from higher earners while weakening the link between contributions and benefits at the top of the earnings distribution.

Donut hole: Half donut hole above $400,000

This provision applies 50 percent of the OASDI payroll tax to earnings above $400,000, creating a second taxable tier above the current-law taxable maximum. The $400,000 threshold is fixed in nominal terms and is not indexed, so over time a growing share of earnings becomes subject to the payroll tax. No additional benefits are credited for taxes paid on earnings above this threshold.

COLA: Switch to the Chained CPI

This provision reduces the growth rate of Social Security benefits by basing the cost-of-living adjustment on the Chained Consumer Price Index starting in 2027. Because the Chained CPI grows more slowly than the traditional CPI, annual benefit increases are smaller, leading to gradual reductions in real benefits relative to current law. The effect compounds over time and primarily affects beneficiaries who receive benefits for longer durations.

PIA: Large or small PIA factor reduction

Both PIA reforms reduce scheduled benefits for future retirees by lowering the PIA formula factors gradually over a 20-year phase-in beginning in 2027, applying only to newly eligible beneficiaries. Under the large reduction, the 90/32/15 factors shift to 80/22/5, substantially lowering replacement rates across the earnings distribution. Under the small reduction, only the second and third factors decrease — from 32/15 to 25/8 — while the 90 percent factor remains unchanged, preserving benefit adequacy for low earners while reducing benefits for middle- and higher-earning workers.

Full retirement age: Increase from 67 to 69

This provision raises the full retirement age from 67 to 69, phased in at a rate of two months per year beginning with individuals who are age 44 in 2027 (born in 1983). The maximum age for delayed retirement credits increases from 70 to 72, maintaining the same accrual rate. The earliest eligibility age remains unchanged. By encouraging later claiming and reducing lifetime benefits, this provision lowers program costs while increasing labor supply at older ages.

Net Tax Rates (Cohorts Born 2000–2009)

The net tax rate measures the difference between the present value of payroll taxes paid and benefits received, expressed as a percentage of the present value of lifetime earnings. A higher net tax rate indicates that a cohort pays more into Social Security relative to what it receives in benefits. Figure 6 reports changes in net tax rates by baseline benefit quintile for cohorts born between 2000 and 2009, corresponding to each policy change option.

Options with larger benefit reductions (C, D, and E) produce higher net tax rates for lower quintiles, ranging from about 2.7 to 4.2 percent. In contrast, Options A and B produce more modest net tax rates of 1.0 to 2.7 percent across all quintiles. The top quintile shows relatively uniform net tax rates across all five options (2.5 to 2.9 percent), reflecting the fact that higher earners already receive benefits that replace a smaller share of their earnings under current law.

More specifically, the effects of tax provisions vary substantially across earnings groups. High-income workers have a larger share of earnings above the taxable maximum and therefore face lower effective payroll tax rates on total earnings under current law. As a result, their lifetime net tax rates are especially sensitive to reforms that expand taxation above the taxable maximum, such as a higher taxable maximum or a donut-hole provision. By contrast, a uniform increase in the payroll tax rate has a smaller effect on high-income workers’ effective tax rate on total earnings because a portion of their earnings remains above the taxable maximum and is not affected.

Benefit provisions also differ in how they affect groups. Broad benefit reductions — such as slower COLA growth or an increase in the normal retirement age — tend to be more consequential for lower-income workers because their replacement rates are higher, so a given percentage reduction in benefits translates into a larger change relative to lifetime earnings. In contrast, benefit changes designed to protect low earners — such as a PIA adjustment from 90/32/15 to 90/25/8, which preserves the lowest (90%) replacement factor — can mitigate the impact on lower-income workers while still reducing benefits for higher earners. Notably, the effects of slower COLA are generally present across cohorts because COLA applies to current beneficiaries as well, and the impact grows over time as smaller annual adjustments compound as beneficiaries age.

Consistent with these mechanisms, packages that rely on contributions from the currently untaxed portion of earnings above the taxable maximum (such as Option A) raise lifetime net tax rates primarily for the highest quintile. In contrast, packages that rely on broad benefit reductions (such as Option E) tend to have larger effects for lower-income workers, while higher-income workers see smaller changes when effects are expressed relative to lifetime earnings. Option D appears to distribute effects more evenly across income groups, whereas Option B appears more progressive, with smaller changes for lower-income groups and larger changes for higher-income groups.

References and Recommended Readings

Abel, A. B., Mankiw, N. G., Summers, L. H., & Zeckhauser, R. J. (1989). “Assessing dynamic efficiency: Theory and evidence.” The Review of Economic Studies, 56(1), 1-19.

Auerbach, Alan J., and Laurence J. Kotlikoff. “National Savings, Economic Welfare, and the Structure of Taxation.” In Behavioral Simulation Methods in Tax Policy Analysis, edited by Martin Feldstein, University of Chicago Press, pp. 459–498. 1983.

Auerbach, A. J., Gokhale, J., & Kotlikoff, L. J.. “Generational Accounts: A Meaningful Alternative to Deficit Accounting.” In D. Bradford (Ed.), Tax Policy and the Economy, Volume 5 (pp. 55–110). MIT Press, 1991.

Auerbach, Alan J., and William G. Gale. “Then and Now: A Look Back and Ahead at the Federal Budget.” National Bureau of Economic Research, Working Paper No. 34455, November 2025.

Biggs, Andrew. “What Might a Flat Benefit Social Security Reform Look Like?” 2025. (https://www.aei.org/profile/andrew-g-biggs/).

Diamond, Peter A. and Peter R. Orszag. Saving social security: A balanced approach. Rowman & Littlefield, 2006.

Gokhale, Jagadeesh, and Kent Smetters. “United States Federal Indebtedness and Fiscal Policy Trade-Offs.” Public Budgeting & Finance, vol. 45(1), pages 21–56, March 2025.

Gruber, J. “The incidence of payroll taxation: evidence from Chile.” Journal of labor economics, 1997, 15(S3), S72-S101.

Gruber, Jonathan. Public Finance and Public Policy. 7th ed., Worth Publishers, 2022.

Krueger, Dirk, and Chunzan Wu. “Taxes on Lifetime Income: A Good Idea?” National Bureau of Economic Research, # w33664. 2025.

Nishiyama, S., and Kent Smetters. “Consumption taxes and economic efficiency with idiosyncratic wage shocks.” Journal of political Economy, 2005, 113(5), 1088-1115.

Nishiyama, S., and Kent Smetters. “Does social security privatization produce efficiency gains?” The Quarterly Journal of Economics, 2007, 122(4), 1677-1719.

Primus, Wendell, Ben Graham, Chloe Zilkha, and Sonia Jain. “Insufficient financing should not provoke dramatic changes to Social Security”, 2026 (https://www.brookings.edu/articles/insufficient-financing-should-not-provoke-dramatic-changes-to-social-security/)

Rosen, Harvey S., and Ted Gayer. Public Finance. 10th ed., McGraw-Hill Education, 2013.

Seiter, Grant, and Sita Nataraj Slavov. “Do Older Adults Accurately Forecast their Social Security Benefits?”, Applied Economics Letters 32 (2), 2025: pp. 183 - 187.

Stiglitz, Joseph E., and Jay K. Rosengard. Economics of the Public Sector. 4th ed. New York: W. W. Norton & Company, 2015.

Vickrey, William. “Averaging of Income for Income-Tax Purposes”. Journal of political economy 1939, 47.3, pp. 379–397.

---

This analysis was produced by Seul Ki (Sophie) Shin under the direction of Kent Smetters who helped write the brief. John Huntley provided computations of dynamics from the PWBM overlapping generations model.

Footnotes

1. This value assumes that the federal government is an infinitely-lived entity, whereas private companies must report a shutdown liability. A shutdown liability measure for the federal government would be even larger. ↩

2. See, for example, Auerbach and Kotlikoff (1983) and Auerbach, Gokhale and Kotlikoff (1991). To be sure, there are technical legal differences between explicit and implicit debt. Explicit debt is backed by the “full faith and credit” of the U.S. Treasury, but its value can be eroded by future inflation since most Treasury debt is nominal. Implicit debt, by contrast, is backed by the policy of the U.S. Congress, but its value is generally harder to erode with inflation alone due to automatic adjustments. For given policy and inflation paths, the two debts are identical. ↩

3. This condition implies that the capital supply is at the golden-rule level that maximizes consumption. Specifically, $(1+r) = (1+n)(1+x)$ where r is the risk-adjusted return to capital, n is the population growth rate and x is the rate of labor-augmenting (Harrod-neutral) technological change. It is common to define g as $(1+g) = (1+n)(1+x)$ Pay-as-you-go Social Security produces a windfall to early generations that is compensated with a permanent (r − g) tax on subsequent generations, essentially a lifetime tax in the spirit of Vickrey (1939). ↩

4. One exception would be households that are borrowing constrained. Social Security benefits can endogenously create these borrowing constraints when payroll taxes exceed optimal savings rates. The Social Security program has also been credited with solving personal planning problems associated with beta-delta preferences as well as observationally-equivalent social gaming problems known as the Samaritan’s Dilemma. However, these last two problems could also simply be solved with mandatory savings. ↩

5. In the case of a smaller flat benefit, multiple trade-offs emerge. A flat benefit increases labor supply distortions, which can increase or decrease labor supply depending on income and substitution effects. If the flat benefit is small enough, risk sharing falls, thereby inducing more precautionary labor supply. The smaller average benefit (and reduced risk sharing) also encourages more household savings and capital accumulation. At the microeconomics level, there is no algebraic solution to this problem, at least not with plausible household utility preferences (e.g., CRRA) that induce the standard Inada condition (where marginal utility goes to infinity as consumption falls to zero). At the macroeconomics level, simple aggregation representation is not possible, in part, due to a lack of homotheticity. Instead, computation is required. ↩