View as PDF 17 Nov 2026

How Much AI Growth Will It Take to Grow Us Out of Debt?

Productivity, Longevity, and the U.S. Fiscal Gap

AI-generated research note · Edited by Kurtis Hingl · Version of 17 November 2026

About this document

The research note below was generated by Claude Opus 5.5 under my (Kurtis Hingl’s) direction. If there is an author in the usual sense, it is Claude. However, I am happy to carry liability for all errors, so I will list myself as an editor.

Cite: Claude. 2026. “How Much AI Growth Will It Take to Grow Us Out of Debt? Productivity, Longevity, and the U.S. Fiscal Gap.” Edited by Kurtis Hingl. https://kurtishingl.com/explainers/ai-fiscal/

Abstract

Optimists argue that artificial intelligence will raise growth enough to fix the federal budget. This document asks how much faster growth would need to be and compares the answer with published forecasts. Scaling the sensitivity analyses of the Congressional Budget Office and the Social Security Trustees, it finds that holding federal debt near today’s 100 percent of GDP for thirty years would require AI to add about 1.2 percentage points to annual productivity growth. Closing Social Security’s 75-year shortfall would require about 2.3 points. Most forecasts lie between 0.1 and 1.0 points. Longevity pushes the other way. If AI-driven medicine made death rates fall 2 percent a year, twice the pace of the past century, Social Security would lose the full benefit of an extra point of growth. A simple model built on an expert survey puts the odds that AI growth alone stabilizes the debt at about one in five, and the odds that it rescues Social Security at about one in twenty.

1Introduction

In a talk in early November 2026, Tyler Cowen predicted that AI would add about half a percentage point to annual growth in U.S. GDP per capita.The talk was given in Barcelona. This summary follows Henderson (2026), who also proposes giving each retiree a fixed 20-year window of full benefits, to be drawn whenever they choose. That, he argued, would be enough for the government to pay Social Security without cutting benefits. Later in the same talk he suggested that AI-driven medicine could let people now 40 or younger live to 120. Henderson (2026) spotted the tension. Growth helps the budget, but longevity hurts it. A retiree who claims at 70 and lives to 110 or 120 collects benefits for 40 to 50 years instead of about 18.

The hope that growth will rescue the budget is not confined to economists’ talks. “With 3% growth, we grow our way out of this,” Treasury Secretary Scott Bessent said in September 2026. A month later, CBO Director Phillip Swagel offered the official rejoinder: “growth will help, but it’s probably not plausible that growth alone will stabilize our fiscal trajectory.”Both quotations as reported by Ma (2026). Swagel added that with interest rates of 4–5 percent, growth alone would have to run at 5–6 percent a year to stabilize the debt.

This document asks how likely it is that AI-driven growth gets the United States out of its fiscal hole, once we allow for the possibility that the same technology lengthens lives. It builds no new macroeconomic model. Instead, it takes the official sensitivity analyses at face value, scales them up, and compares the growth they require with the growth that forecasters expect. The findings are threefold:

  1. The bar is high. To keep federal debt from rising above today’s level through the mid-2050s, AI would need to add about 1.2 points to annual productivity growth, sustained for three decades. To close Social Security’s 75-year gap through growth alone would take about 2.3 points, which means real wages growing roughly 3.4 percent a year for 75 years. The past 20 years averaged about 0.9 percent.
  2. Forecasts fall short. Nearly every central forecast, from Acemoglu (2025)’s 0.1 points to Aghion and Bunel (2024)’s 0.8 to 1.3, lies below the federal threshold, and every one lies below the Social Security threshold. Only the boom scenarios clear them.
  3. Longevity is a large offset. Each extra 0.1 point of annual decline in death rates costs Social Security about as much as 0.08 points of extra growth adds. A decline of 2 percent a year erases a full point of growth.

Putting these together in a simple model whose inputs come from an expert survey, the odds that AI growth alone stabilizes the debt are about one in five. The odds that it closes Social Security’s gap are about one in twenty. Section 6 lets you put in your own assumptions.

Several 2026 studies reach related conclusions from different angles. Harris, Mehrotra, and Overcash (2026) find that a large, conventional productivity shock would cut the 2036 deficit from about 6 to about 2 percent of GDP. Here, they write, “the techno-optimists are validated.” But their AI-specific scenarios keep only about half of that gain, or less. Some assume a smaller boom, and all add offsets: longer lives, a shift of income from labor to capital, or higher interest rates. The Yale Budget Lab (2026) finds that a large, persistent productivity boom would make the fiscal path “more sustainable” even after interest rates rise, while warning that falling labor supply would offset part of the gain. Riedl (2026) puts the savings at perhaps $600 billion a year within a decade, a little over one-eighth of the deficit she projects under current policies. CRFB (2026) calculates that stabilizing the debt would take 2.5 percent TFP growth for a decade, a pace never sustained in the postwar era. This document differs in three ways. It puts growth and longevity on a single map. It looks beyond the 10-year window, to the horizon over which Social Security’s wage-indexed benefits catch up with wages. And it turns the comparison into explicit odds.

2The size of the problem

The fiscal problem has two parts: a deadline and a gap. The deadline is close. The Social Security retirement trust fund runs out in late 2032. From then on, the fund can pay only 78 percent of scheduled retirement benefits, a share that falls to 62 percent by 2100. Medicare’s hospital insurance fund runs out in 2033, and the combined Social Security funds in 2034 (SSA Trustees 2026; Medicare Trustees 2026).

The gap is large. Under CBO’s long-term projections, debt held by the public rises from 101 percent of GDP in 2026 to 175 percent in 2056, well past the 1946 record of 106 percent (Figure 1a).CBO released its 2026 long-term projections as data only. The figures here are from that data file. Over the same period, net interest climbs from 3.3 to 6.9 percent of GDP. Keeping the debt at its 2026 level would require primary deficits that average about 2 percentage points of GDP smaller over thirty years (Swagel 2026). Social Security alone faces a 75-year shortfall of 4.42 percent of taxable payroll, about 1.5 percent of GDP, up from 3.82 percent of payroll a year earlier. Spending on Social Security and Medicare together rises from 9.4 to 14.2 percent of GDP by 2100 (Figure 1b), while Social Security’s own income stays between 4.4 and 4.8 percent.

Figure 1: Federal debt and the two big entitlement programs under current law. Notes: Panel (a): debt held by the public, CBO extended baseline, February 2026. Panel (b): Social Security (OASDI) cost and total Medicare spending, intermediate assumptions of the 2026 Trustees Reports. The dashed line is Social Security’s dedicated income (payroll taxes and taxation of benefits). Hover for values.

These projections are not knife-edge. The Social Security Trustees run 5,000 random simulations of the economy and population. In 97.5 percent of them the 75-year deficit is at least 1.86 percent of payroll, and in 97.5 percent the trust funds run dry by 2039. Even the Trustees’ low-cost scenario, in which every assumption breaks favorably at once, still runs out of money in 2048.

3How much growth would it take?

Faster growth helps the budget. But three features of U.S. fiscal law mean it helps less than intuition suggests.

Benefits follow wages. A worker’s starting Social Security benefit is computed from lifetime earnings indexed to economy-wide wage growth. So when AI raises wages, it raises payroll-tax revenue at once, and it also raises the benefits that today’s workers will collect later. The Trustees explain why growth still helps: “Higher wages increase taxable payroll immediately, but they increase benefit levels only gradually,” and benefits after retirement rise only with prices. Still, the offset is large. The 2026 report raised projected real earnings growth for 2025–35 from 1.44 to 1.72 percent a year. Together with the report’s other economic updates, that improved the 75-year balance by only 0.10 percent of payroll, and the combined funds’ depletion date did not move. As Kent Smetters of the Penn Wharton Budget Model put it, a faster productivity boost “barely moves the balance because the initial benefits go up.”As reported in Pringle (2026).

Interest rates rise with growth. In standard models, faster expected growth raises real interest rates (Chow, Halperin, and Mazlish 2026). Furman (2026) reports a simulation in which markets expect the AI buildout to continue through 2031 and productivity growth to rise by half a point. The 10-year Treasury yield rises by about 75 basis points. With debt above 100 percent of GDP, this matters: CBO calculates that interest rates one point higher than projected would push 2056 debt from 175 to 222 percent of GDP (Swagel 2026).

The tax system leans on labor income. If AI’s gains flow mainly to profits, capital, or salaries above the payroll-tax cap, much of the growth never reaches Social Security. Growth in capital income raises only about half as much federal revenue as the same growth in labor income (CRFB 2026). In the most dramatic scenario of Korinek et al. (2026), labor’s share of income falls from 60 to 45 percent by 2030.

With these leaks in mind, the official sensitivity analyses give the thresholds.

Social Security. The Trustees report that each 0.1-point increase in long-run real wage growth improves the 75-year balance by about 0.19 percent of payroll. Extending that slope, Cowen’s half point closes 22 percent of the gap, and Henderson’s full point closes 43 percent. Closing the whole gap takes about 2.3 extra points: real wages would have to grow about 3.4 percent a year for 75 years, compared with the Trustees’ 1.14 percent and roughly 0.9 percent over the past two decades (SSA Trustees 2026, Table V.B1). If only half of AI’s gains reach taxable wages, the requirement doubles to 4.6 points.

Federal debt. CBO’s most recent productivity sensitivity, from May 2025, found that total factor productivity (TFP) growth half a point faster than projected would leave debt at 113 percent of GDP in 2055, against a baseline of 156 percent (CBO 2025). That scenario already lets interest rates rise somewhat. The 2026 baseline is worse, so scaling the same effect implies that holding debt near 101 percent through 2055 would take roughly 0.8 points of extra TFP growth, about 1.2 points of labor-productivity growth.In growth accounting, labor productivity grows by roughly TFP growth divided by labor’s share in production, about 1.5 times TFP once capital accumulates to match. The scaling assumes the debt effect per point of TFP growth is the same in the 2026 baseline as in the 2025 one; because the 2026 baseline starts with more debt, the true effect is probably a little larger.

The federal bar is lower than Social Security’s for two reasons. It sets a modest goal: it stops debt from rising further from a record level, rather than paying every promised benefit for 75 years. And income-tax brackets are indexed to prices, not wages, so real growth pushes revenue up faster than GDP.

4What forecasters expect

Figure 2 sets these thresholds against published estimates of AI’s effect on growth. The forecasts fall into three groups. Skeptics: Acemoglu (2025) estimates a TFP gain of under 0.07 points a year over a decade; CBO builds 0.1 points into its baseline for 2026–36 (CBO 2026a); and the Penn Wharton Budget Model (2025) projects a TFP boost that peaks at 0.2 points in 2032 and then fades. In the middle, estimates from the OECD (Filippucci, Gal, and Schief 2024) and McKinsey (McKinsey Global Institute 2023) run from 0.1 to 0.9 points, and a Dallas Fed illustration uses 0.3 (Wynne and Derr 2025); Cowen’s half point sits here (Cowen 2025). Optimists: Aghion and Bunel (2024) put the boost at 0.8 to 1.3 points over a decade, and Goldman Sachs at just under 1.5 points in the decade after wide adoption (Briggs and Kodnani 2023).

Figure 2: Growth needed versus growth forecast. Notes: Dots are point estimates and bars are published ranges, in percentage points of extra annual labor-productivity growth. TFP estimates are multiplied by 1.5; GDP-growth forecasts are measured against a 1.7 percent baseline. Horizons differ: most estimates cover about a decade, the expert survey covers 2045–49, and Korinek et al. cover 2026–30. Vertical lines are this document’s thresholds (Section 3). Hover a row for details.

The best evidence on what experts actually believe comes from Karger et al. (2026), who surveyed economists, AI industry and policy professionals, superforecasters, and the general public. The median economist expects 2.5 percent annual GDP growth in the late 2040s, about 0.8 points above the 1.7 percent long-run baseline used by the Trustees and CBO. Only the boom scenarios reach the thresholds. Economists give 14 percent odds to “rapid” AI progress, in which AI surpasses humans at many cognitive and physical tasks by 2030. In that world, the median economist forecasts 3.5 percent growth in the late 2040s, or about 1.8 points of extra growth. That clears the federal line but not the Social Security line. Further out are Dario Amodei’s suggestion that 10–20 percent growth “may be possible” (Amodei 2026) and the “extreme” scenario of Korinek et al. (2026), in which growth reaches 15 percent by 2030.

Figure 2 also flatters the optimists in one respect. Most forecasts describe a decade-long boost, but the thresholds require the boost to last thirty to seventy-five years. A temporary acceleration raises the level of wages for good. Social Security, however, eventually pays a higher wage level back out as higher benefits, so a one-time level gain does little for its 75-year balance. Early data are encouraging but not decisive. Nonfarm productivity grew 3.2 percent in 2024 and 2.3 percent in 2025, against 1.5 percent over 2007–19 (BLS 2026). Productivity data are noisy, however, and it is too early to tell how much of the recent strength comes from AI.

5The longevity offset

Henderson’s point is that AI may lengthen lives as well as raise incomes. The Trustees assume death rates will fall 0.73 percent a year. Over the past century they fell about 1.0 percent a year. Each extra 0.1 point of annual decline worsens the 75-year balance by about 0.15 percent of payroll. Set against 0.19 percent of payroll per 0.1 point of growth, this gives a simple exchange rate: a decline in death rates of about 2 percent a year cancels a full point of extra growth. In more intuitive units, each extra year that a 65-year-old can expect to collect benefits costs about 0.27 percent of payroll, so one point of extra growth pays for roughly seven more years of retirement, and no more.Across the Trustees’ three mortality scenarios, period life expectancy at 65 in 2100 is 21.0, 23.6, and 26.5 years. One point of growth is worth 1.92 percent of payroll; one year of life expectancy at 65 costs about 0.27 percent; 1.92 / 0.27 ≈ 7. Under a 2 percent decline, a 65-year-old in 2100 would expect to live to about 96 rather than the Trustees’ 89, collecting benefits for about 31 years rather than 24.

Figure 3: The race between growth and longevity: Social Security’s 75-year balance under current law. Notes: Shading shows the 75-year OASDI actuarial balance implied by adding the Trustees’ one-at-a-time sensitivities to real wage growth and to mortality decline (Tables VI.D4 and VI.D2). The vertical axis is the expected age at death of a 65-year-old in 2100 (65 plus period life expectancy at 65), which maps one-for-one to the rate of mortality decline across the Trustees’ scenarios. Under current law benefits are paid for life, so a longer life means more years of benefits. The Trustees tested growth up to 0.6 points above baseline and ages up to about 91½; everything beyond is linear extrapolation, increasingly rough toward the top. The two blue lines mark the Trustees’ longevity projection and the past century’s pace of mortality decline; the small marker on the age axis is today. Each ring combines one person’s growth estimate with the same person’s longevity speculation. That combination is this document’s construction, not theirs: it shows what current law would cost if both came true, and it is not a policy proposal or an endorsement by anyone named. Black dots are Brookings scenarios, each of which specifies both growth and mortality. Hover anywhere for the implied balance.

Figure 3 puts both forces on one map. Moving right is faster wage growth; moving up is a longer expected life for a 65-year-old, which under current law means more years of benefits. Above and to the left of the black line, Social Security runs short. The map shows that growth-only optimism and health optimism pull in opposite directions:

How plausible are large longevity gains? The leaders of the AI labs are bullish. Dario Amodei expects AI to “compress” 50–100 years of biology into 5–10 and suggests lifespans could double to 150 (Amodei 2024). Demis Hassabis calls the end of disease “within reach” in a decade or so (CBS News 2025). Demographers are skeptical. Olshansky et al. (2024) find that life-expectancy gains in the longest-lived countries have been slowing since 1990. Curing a single disease does surprisingly little, because people spared one cause of death die of another: eliminating all cancer would raise life expectancy at 65 by about 2.2 years, and eliminating heart disease by about 3.2 (Arias, Heron, and Tejada-Vera 2013). Radical life extension would require slowing aging itself, not curing diseases one at a time. The fiscal risk is correlated, though. The worlds in which AI is capable enough to deliver 2 points of growth are the worlds in which it is most likely to transform biology.

Medicare is exposed differently. When longer lives are healthy lives, as they have been recently, they raise Social Security costs more than Medicare costs. Between 1992 and 2019, changes in elderly health and mortality raised expected lifetime Social Security spending by 14 percent but Medicare spending by only 6 percent (Einav and Finkelstein 2026). Medicare’s main risk is health-cost growth, and here AI could help directly. Under current law, payment updates for most non-physician providers are reduced by a ten-year average of economy-wide productivity growth, so an AI boom trims Medicare prices automatically, though with a lag. The catch, which the Medicare Trustees flag, is whether providers can sustain those cuts.

6How likely is it?

To turn the comparison into odds, this document uses a deliberately simple model whose every input you can change below. It starts from the scenario probabilities in Karger et al. (2026): economists put 39 percent on slow AI progress, 47 percent on moderate progress, and 14 percent on rapid progress. Within each scenario s, sustained extra productivity growth g is normally distributed with mean μs and standard deviation σs, and death rates fall at rate ms. The odds of clearing a bar are

Pr(clear)  =  ∑s ps · Pr( λ g ≥ T(ms)  |  s ),

where T(m) is the threshold from Sections 3 and 5, which rises with faster mortality decline, and λ ≤ 1 is the share of growth that does fiscal work. The defaults:

Figure 4 and Table 1: put in your own prior

AI growth alone stops the federal debt from rising
–
 
AI growth alone closes Social Security’s 75-year gap
–
 
About 0.3 points of extra growth
About 1.1 points
About 1.8 points, with wide uncertainty

Weights are rescaled to sum to 100 percent.

Figure 4: Your prior. The solid curve mixes the three scenarios (dashed). Vertical lines are the federal and Social Security bars at the Trustees’ longevity, adjusted for leaks; the odds above also raise each bar with the mortality decline you assign to each scenario.

Table 1: Growth required and the odds that AI delivers it

GoalExtra growth needed
(pts/yr, sustained)
Odds AI delivers it

Notes: Thresholds at the Trustees’ longevity unless stated. The two “if” rows recompute your odds with that one assumption changed.

Under the default inputs, the odds that AI growth alone stops the debt from rising are about 22 percent, or roughly one in five. The odds that it closes Social Security’s gap are about 4.7 percent, roughly one in twenty. Both fall further if faster medical progress comes along with the growth, or if the gains bypass wages. To reach even odds on the federal bar, you would need to put most of your weight on rapid AI progress; to reach even odds on Social Security, you would need rapid progress that also leaves lifespans close to the Trustees’ path.

The last row of Table 1 deserves emphasis. The 2032–34 deadlines arrive before any plausible AI dividend could accumulate. In 97.5 percent of the Trustees’ stochastic simulations, the combined funds run dry by 2039. Whatever AI does to growth, Congress faces a Social Security decision within six years.

7Conclusion

AI is the best fiscal news the United States is likely to get. Even the mainstream forecasts would shrink the gap: half a point of extra growth closes about a fifth of Social Security’s shortfall and about 40 percent of the federal one. But growing out of the problem requires a boom at the top end of expert expectations. It also requires that boom to last for decades, to pay off in wages rather than profits, and to leave interest rates and lifespans roughly where they were. Those conditions are unlikely to hold together. The boom that would deliver the growth is the same boom most likely to deliver the longevity.

The policy lesson follows from the asymmetry. Growth is uncertain and arrives late, while the deadlines are fixed and close. Three implications stand out:

Hope for the boom, but legislate for the median.

References

Acemoglu, Daron. 2025. “The Simple Macroeconomics of AI.” Economic Policy 40 (121): 13–58. Link ↗

Aghion, Philippe, and Simon Bunel. 2024. “AI and Growth: Where Do We Stand?” Working paper, June. Link ↗

Amodei, Dario. 2024. “Machines of Loving Grace: How AI Could Transform the World for the Better.” Essay, October. Link ↗

Amodei, Dario. 2026. “The Adolescence of Technology.” Essay, January. Link ↗

Arias, Elizabeth, Melonie Heron, and Betzaida Tejada-Vera. 2013. “United States Life Tables Eliminating Certain Causes of Death, 1999–2001.” National Vital Statistics Reports 61 (9). Link ↗

BLS (Bureau of Labor Statistics). 2026. “Productivity and Costs, Third Quarter 2026, Preliminary.” News release, November 5. Link ↗

Briggs, Joseph, and Devesh Kodnani. 2023. “The Potentially Large Effects of Artificial Intelligence on Economic Growth.” Goldman Sachs Global Economics Analyst, March 26. Summary: Link ↗

CBO (Congressional Budget Office). 2025. The Long-Term Budget Outlook Under Alternative Scenarios for the Economy and the Budget. May. Link ↗

CBO (Congressional Budget Office). 2026a. The Budget and Economic Outlook: 2026 to 2036. February 11. Link ↗

CBO (Congressional Budget Office). 2026b. The Long-Term Budget Outlook Data: 2026 to 2056. February 25. Link ↗

CBS News. 2025. “Artificial Intelligence Could End Disease, Lead to ‘Radical Abundance,’ Google DeepMind CEO Demis Hassabis Says.” 60 Minutes, April 20. Link ↗

Chow, Trevor, Basil Halperin, and J. Zachary Mazlish. 2026. “Transformative AI, Existential Risk, and Real Interest Rates.” Working paper, revised June. Link ↗

Cowen, Tyler. 2025. “Why I Think AI Take-Off Is Relatively Slow.” Marginal Revolution, February 23. Link ↗

CRFB (Committee for a Responsible Federal Budget). 2026. “Don’t Count on Growth to Fix the Debt.” Blog post, September 24. Link ↗

Einav, Liran, and Amy Finkelstein. 2026. “Elderly Health and Longevity in the US: Evidence and Implications.” NBER Working Paper 35346. Link ↗

Filippucci, Francesco, Peter Gal, and Matthias Schief. 2024. “Miracle or Myth? Assessing the Macroeconomic Productivity Gains from Artificial Intelligence.” OECD Artificial Intelligence Papers 29. Link ↗

Furman, Jason. 2026. “The AI Boom’s Two Engines.” PIIE Realtime Economics, November 9. Link ↗

Harris, Benjamin H., Neil Mehrotra, and William Overcash. 2026. “Can AI Restore Fiscal Sustainability in the US? Accounting for an AI-Driven Productivity Shock.” Brookings Institution Working Paper, June. Link ↗

Henderson, David R. 2026. “Tyler Cowen’s Insights and One Apparent Contradiction.” I Blog to Differ (Substack), November 16. Link ↗

Karger, Ezra, Otto Kuusela, Jason Abaluck, Kevin Bryan, Basil Halperin, et al. 2026. “Forecasting the Economic Effects of AI.” Federal Reserve Bank of Chicago Working Paper 2026-07, March. Link ↗

Korinek, Anton, Charles I. Jones, Szymon Sacher, Tess Cotter, and Peter McCrory. 2026. “Economic Scenarios for Transformative AI.” The Anthropic Institute Working Paper 2026-02, September. Link ↗

Ma, Jason. 2026. “CBO Chief Warns It’s ‘Probably Not Plausible’ That a Strong Economy Alone Can Steady U.S. Debt as 5%–6% Growth Is Needed—More than Bessent’s 3% View.” Fortune, October 10. Link ↗

McKinsey Global Institute. 2023. “The Economic Potential of Generative AI: The Next Productivity Frontier.” June. Link ↗

Medicare Trustees (Boards of Trustees, Federal Hospital Insurance and Federal Supplementary Medical Insurance Trust Funds). 2026. 2026 Annual Report. Link ↗

Olshansky, S. Jay, Bradley J. Willcox, Lloyd Demetrius, and Hiram Beltrán-Sánchez. 2024. “Implausibility of Radical Life Extension in Humans in the Twenty-First Century.” Nature Aging 4 (11): 1635–1642. Link ↗

Penn Wharton Budget Model. 2025. “The Projected Impact of Generative AI on Future Productivity Growth.” September 8. Link ↗

Pringle, Eleanor. 2026. “Trump Plans to Grow His Way Out of $40 Trillion Debt Crisis—It’s a ‘Fantastic Story’ but Virtually Impossible, Says Top Budget Economist.” Fortune, August 25. Link ↗

Riedl, Jessica. 2026. “Even an AI-Sparked Economic Miracle Will Not Save the Federal Budget.” Reason, June 16. Link ↗

SSA Trustees (Board of Trustees, Federal Old-Age and Survivors Insurance and Federal Disability Insurance Trust Funds). 2026. The 2026 Annual Report. House Document 119-163. Link ↗

Swagel, Phillip L. 2026. “Projections of Deficits and Debt Under Alternative Scenarios for Interest Rates and the Budget.” Letter to Senator Jeff Merkley, Congressional Budget Office, September 24. Link ↗

Wynne, Mark A., and Lillian Derr. 2025. “Advances in AI Will Boost Productivity, Living Standards over Time.” Federal Reserve Bank of Dallas, June 24. Link ↗

Yale Budget Lab. 2026. “What Might AI Adoption Mean for the Fiscal and Economic Outlook?” May 6. Link ↗

Colophon. Generated by Claude Opus 5.5 (Anthropic). Research agents gathered sources; every number in the text, figures, and calculator is computed from a single parameter file by code that also produces the typeset PDF. Set in Latin Modern.

Versions. 17 November 2026: first public version.