Social Security’s trust funds may have a little more runway than many people expected, but the program’s long-term finances still point to painful choices ahead. A new forecast from the Penn Wharton Budget Model, shared exclusively with CNBC, puts the Old-Age and Survivors Insurance fund’s depletion in February 2033.
That is slightly later than the Social Security trustees’ June 9 estimate, which says the fund could run out in the fourth quarter of 2032. When the disability insurance fund is included, PWBM sees depletion in February 2035, compared with the trustees’ third-quarter 2034 projection.
What happens if Congress does nothing
The depletion dates assume lawmakers take no action to shore up the system. Social Security would not go bankrupt when the trust funds are exhausted, because payroll taxes would still keep coming in.
Still, benefits would have to be cut to match incoming revenue. PWBM expects 86% of scheduled benefits would be payable once the combined funds are depleted, slipping to 60% by 2100. The trustees project 83% would be payable at depletion, falling to 65% by 2100.
Why the gap between forecasts has narrowed
PWBM’s latest outlook is notable because its depletion estimate has now moved closer to, and slightly past, the trustees’ number. Kent Smetters, Wharton professor and faculty director of PWBM, said the change does not lessen the need for action.
“We’re still talking about a pretty sizable increase that would be necessary in terms of taxes or benefit cuts going forward, and if we don’t take action soon, that number just simply goes up,” Smetters said.
PWBM projects an actuarial deficit of 4.65% of taxable payroll, slightly wider than the trustees’ 4.42% estimate. Closing that gap would require lifting the current 12.4% payroll tax rate for workers and employers to 17.1%, a 4.7 percentage point increase, or making an equivalent benefit cut or mix of both.
Why the models disagree
PWBM uses a microsimulation model built from individual-level data on earnings and family structure. The trustees begin with aggregate assumptions, such as fertility and wage growth, and build long-range projections from those inputs.
That difference matters because the two models do not treat the same forces in the same way. In PWBM’s new report, fertility, life expectancy and wage growth are outputs of the model rather than assumptions.
What changed in the latest trustees report
Karen Glenn, the Social Security Administration’s chief actuary, said during a June 10 virtual briefing hosted by the Committee for a Responsible Federal Budget that four major changes affected this year’s trustees report.
| Change | Effect |
|---|---|
| Total fertility rate | Lowered to 1.75 children per woman from 1.90 |
| Immigration outlook | Updated to reflect recent historical data and future expectations |
| Productivity and earnings | Projected to grow faster in the near term |
| Income tax changes from Trump’s “big beautiful bill” | Reduced revenue to the trust funds from taxes on benefits |
PWBM differs on two of those points. It projects a long-term fertility rate of about 1.6 births per woman and does not separate out the effect of the “big beautiful bill” in the same way.
Smetters said the law does not directly end taxes on Social Security benefits, which would have had a larger financial impact. He added that the revenue loss is partly offset by short-run economic gains, leaving the net effect within the usual margin of forecast error.
Other factors could still change the outlook
Future projections could move again if health or technology trends shift. Smetters said GLP-1 drugs could worsen the long-run shortfall if they extend life expectancy, since people would collect benefits for longer.
He also said PWBM’s research suggests artificial intelligence could boost productivity and GDP over time, but a burst AI bubble could create negative macroeconomic effects. For now, the broad message from both sets of projections is the same: Social Security still faces a large funding gap, even if the deadline is slightly less urgent than before.