Loading reports … Fetching the current index of long-form research.
Reports
Loading reports … Fetching the current index of long-form research.
Depth, speed and duration are three different damages that need three different toolkits.
Depth is how far output falls, duration is how long the damage lasts, and speed is how fast the fall and the recovery happen. The Depression, the 1970s and COVID each did their worst on a different one of the three, and each needed a different response. This report holds the axes apart and asks what misreading them would cost a government whose rescues now cost more to finance than in 2019.
The economic shocks that define modern American memory each damaged the economy through a different channel, on a different time horizon, and made a different demand of the state. Reading them on one severity scale hides the question that matters when the next shock arrives: which mechanism is at work.
This report uses three measurements that are usually conflated. The first is depth: how far real output falls from its prior peak, best read in annual data. The Depression has no postwar rival here, with a peak-to-trough loss of 26.3 percent in annual real GDP across 1929 to 1933.1 The second is duration: how long output stays below that peak, and how long balance sheets, labor markets and price expectations stay damaged. Duration turns a recession into a regime, and it is where the 1970s stand out. The third is speed: how fast the collapse and the recovery happen. Speed is invisible in annual data, decisive in quarterly data, and the axis on which COVID is unique.
The "worst since the Depression" framing collapses the three into one number. A shock can be deep without being abrupt, abrupt without lasting damage, and inflationary for a decade without a deep output collapse. Each axis implies a different stabilization problem and a different set of tools, so the comparison becomes useful only once the axes are read separately.
In annual data the Depression produced catastrophic depth. Real GDP fell 8.5 percent in 1930, 6.4 percent in 1931 and 12.9 percent in 1932, and contracted again in 1933. Output fell from about $1,191 billion in 1929 to $877 billion in 1933, in chained 2017 dollars, a loss of more than a quarter of national production. No postwar recession approaches those figures.2
COVID produced remarkable depth in quarterly data: a fall of 7.9 percent in 2020 Q2 alone, which the national accounts report as a −28.0 percent annualized rate. In annual data the same episode barely registers, with real GDP down 2.1 percent in 2020 and up 6.2 percent in 2021. The gap between the two readings is the signal: the damage was concentrated into one quarter and then largely reversed.
Duration is where the Depression and the 1970s diverge most. The Depression kept real output below its 1929 peak until 1936, the seventh year, when the annual index first cleared 100 (Table A2). The 1970s output declines were mild, yet the damage to inflation expectations, wage setting and productivity persisted across the decade and ended only with the Volcker disinflation of 1982 and 1983. COVID output cleared its 2019 Q4 peak in 2021 Q1, five quarters on.3
Speed is the axis that makes COVID unique. The NBER dates the 2020 recession from February to April, two months, the shortest in its chronology.4 Unemployment jumped from 3.5 percent in February to 14.8 percent in April, and output rebounded at a 34.9 percent annualized rate in 2020 Q3. The interventions had to be authorized, designed and disbursed within weeks, inside the window in which they could still matter.
Grading every shock against the Depression works tolerably well when recessions share a mechanism, and it fails when they do not. The 1970s were a supply repricing that forced policymakers to choose between stabilizing output and stabilizing prices, and a political economy that repeatedly chose the softer option until inflation became self-reinforcing. The tools for a 1930s-style demand collapse, fiscal stimulus, easier money and credit support, were the wrong tools for that shock, and using them extended the damage.
COVID was a sudden stop: an abrupt interruption of activity caused by a public-health intervention, with no financial or demand failure behind it. Its toolkit was different again. It needed speed above all, with bridge financing to carry firm and household balance sheets through a temporary cessation of activity.
The core error is misreading the mechanism. An analyst who concludes that the next shock is most like the Depression when it is most like the 1970s, or the reverse, will reach for the wrong tools, and the wrong tools applied at the wrong time compound the shock they were meant to absorb.
The 2020 case shows the cost in real time. On 14 April 2020 the IMF headlined the global outlook "Worst Economic Downturn Since the Great Depression" and projected US real GDP down 5.9 percent for the year; the outturn on the current vintage is down 2.1 percent.5 The comparison drew on unemployment of 14.8 percent in April 2020, the highest since the monthly series began in 1948. The mechanism was a two-month administrative stop with intact household and corporate balance sheets behind it, and the policy that worked was the one the mechanism called for: bridge financing, disbursed in weeks. A planner who took the Depression frame at face value would have built for a multi-year reconstruction and arrived late to a shock that was over in five quarters.
The reason this matters now is fiscal. Part V prices a 2020-scale transfer of $3.1 trillion at about $65 billion a year of interest at 2019 yields and $112 to $134 billion at April 2026 yields. The difference, $47 to $69 billion a year, is about 0.15 to 0.22 percent of 2026 GDP. That is small against output. It recurs for as long as the debt is outstanding, and it lands on a budget where net interest already runs above 3 percent of GDP and above defense. A supply repricing misread as a demand collapse, and met with a transfer of that size, would pay roughly double the 2019 carry for a response that feeds the price problem, and a central bank facing rising prices could not cut to bring yields down. That is the sense in which misgrading has become more expensive.
The Depression was severe on all three axes at once and then compounded year after year. What sets it apart is the combination of catastrophic depth with a state structurally unequipped to absorb it.
Unemployment reached 25.6 percent in May 1933 on the NBER macrohistory monthly reconstruction, a series that predates the modern household survey; Lebergott's annual figure for 1933 is 24.9 percent.7 The Depression was a multi-year compression in which bank failures destroyed deposits, lost deposits cut spending, lower spending closed firms, and closed firms cut employment, which fed back into more bank failures. The spiral ran for four years before the state deployed enough capacity to interrupt it.
The 1929 crash and the commodity-price collapse were a large demand shock, but no larger than several shocks that produced ordinary recessions. What converted it into the Depression was the banking system's amplification of that shock across four years without effective countervailing policy. The shock started the contraction, and the absence of institutions to stop it made the contraction the Depression.
Between 1930 and 1933 more than 9,000 US banks failed.8 The mechanism was the bank run: once depositors believed a bank might be insolvent, the rational response was to withdraw at once, and that response made insolvency self-fulfilling. Runs destroyed solvent banks alongside insolvent ones because depositors could not tell them apart, and depositors in failed banks bore the losses, with no insurance to absorb them.
The destruction of deposits was the destruction of money. Broad money fell by about a third between 1929 and 1933, and a smaller money supply meant less spending, revenue, employment and income.9 Falling prices made the debt burden heavier in real terms, so a farmer who owed $1,000 in 1929 carried a larger real burden in 1932 on the same nominal debt. Irving Fisher called this spiral debt deflation, and it remains the most precise description of the Depression's self-reinforcing character.10
There was no deposit insurance until the FDIC was created in 1933, and no federal unemployment insurance worth the name. The Federal Reserve, instead of expanding money to offset the bank failures, raised its discount rate in October 1931 to defend the gold standard, and the Hoover administration tried to balance a budget whose revenues were collapsing. There was no fiscal stabilizer of meaningful scale and no political consensus to build one quickly.
The strongest rival account widens the frame without overturning it. Barry Eichengreen and Ben Bernanke trace the depth and spread of the slump to the gold standard itself: countries that left gold earlier recovered earlier, so the Fed's 1931 defense of gold was one instance of a constraint every gold country shared.11 On either reading the United States lacked the institutions that would have broken the spiral, and the New Deal mattered most for building them: deposit insurance, the SEC, Social Security as a permanent automatic stabilizer, and the separation of commercial from investment banking. That architecture is what the country has used for nine decades to keep a Depression from recurring.
CPI fell 2.7 percent in 1930, 8.9 percent in 1931, 10.3 percent in 1932 and a further 5.2 percent in 1933, on the annual average, and turned positive only in 1934.12 Deflation raised the real value of nominal debt, which forced debtors to sell assets to meet their obligations, depressed asset prices further and raised real debt burdens again.
Annual output did not recover its 1929 level until 1936, and the labor market stayed damaged for the rest of the decade. On the Lebergott basis unemployment was still 14.3 percent in 1937 and rose to 19.0 percent in 1938, after the Federal Reserve doubled reserve requirements in 1936 and 1937 and the federal deficit was cut, before wartime mobilization cleared the backlog. The institutions built afterward, deposit protection, a central bank charged with stabilizing prices, and stabilizers that expand without a vote of Congress, have made a Depression-grade duration very hard to repeat in a normal recession, and as Part V shows, they have consumed fiscal capacity in doing so.
The 1970s output loss was modest. The damage was a decade of entrenched inflation and the unemployment that Volcker’s disinflation required to break it, a different kind of damage that demanded a different toolkit.
The 1970s sit awkwardly in the standard recession comparison because the output numbers are unimpressive. Annual real GDP fell 0.5 percent in 1974 and 0.2 percent in 1975, and again 0.3 percent in 1980 and 1.8 percent in 1982. Quarterly contractions of −3.7 percent in 1974 Q3 and −4.8 percent in 1975 Q1, annualized, were painful but unremarkable in postwar context.
The damage was in prices. Annual CPI inflation reached 11.1 percent in 1974 and 13.5 percent in 1980, and energy prices ran 33.7 percent above their prior-year level by September 1974.13 Unemployment hit 9.0 percent in May 1975 and 10.8 percent in November 1982. Nonfarm business labor productivity, which had grown about 2.8 percent a year from 1948 to 1973, slowed to about 1.2 percent from 1973 to 1979.14 The Federal Reserve's trade-weighted dollar index against major currencies fell about 12 percent between January 1973, where the series begins, and the 1980 annual average.15 Stagflation, rising unemployment alongside rising inflation, had been treated as theoretically impossible within the dominant postwar framework, and the experience reshaped central-bank practice.
The 1974 to 1975 recession produced a peak-to-trough real GDP decline of 3.1 percent, from the 1973 Q4 peak to the 1975 Q1 trough, smaller than any single annual Depression decline. The right metric for the decade is how persistently prices rose and what it cost to bring them down. Annual inflation exceeded 6 percent every year from 1973 through 1981 except one, and expectations stayed dislodged from the Federal Reserve's implicit target for a decade. Once inflation is expected it is written into wage contracts, supply contracts and asset prices, and removing it takes a recession engineered to be painful enough to break those expectations.
The 1973 OPEC embargo was the trigger, and the ground was prepared. The United States had ended the Bretton Woods gold link in 1971, productivity growth had slowed, and the embargo then repriced energy across the economy.16 Energy enters nearly every production process, so a sustained rise in its price pushed up costs in sector after sector at once. That is a supply shock: inflation that starts from a cost increase. Adding demand on top of it pushes prices higher still, so it leaves policymakers a choice between accepting the inflation and accepting the output loss needed to bring prices down.
The monetarist reading puts more of the decade on the Federal Reserve than on OPEC, and the report's own data give it weight: inflation was already 6.2 percent in 1973, on the annual average, before the embargo of October that year could move it much. Both readings converge on the same trap. Policymakers repeatedly accepted the inflation, because unemployment is visible and immediate while inflation is diffuse at first, and each accommodation made expectations more entrenched and the eventual disinflation costlier. The Nixon wage and price controls of 1971 suppressed the visible price level and left the pressure intact, and it returned when the controls were lifted.
Paul Volcker's appointment as Federal Reserve Chairman in 1979 broke the pattern. The federal funds rate averaged 19.1 percent in June 1981, its monthly peak, and the daily effective rate touched 22.4 percent on 22 July 1981.17 Unemployment peaked at 10.8 percent in November 1982. That was the price of breaking expectations a decade in the making. The Depression needed demand stimulus and the 1970s needed demand suppression: fiscal stimulus against the 1970s inflation would have worsened it, and monetary tightening against the Depression deepened it, which is what the 1931 rate rise did.
The dangerous moment in a 1970s-type episode is when inflation becomes embedded in wages and expectations, because embedded expectations are self-fulfilling and the Fed's ability to break them without a severe recession falls as they deepen. The difference between 1979 and 2026 that matters most is the speed of the anchoring mechanism. In 1979 contracts were typically annual and expectations updated with a lag; in 2026 markets price expected Fed policy within minutes, and long-term expectations are anchored to the Fed's target with more precision. That anchoring rests on the Fed's commitment and its record, both tested in 2021 and 2022.
COVID compressed a collapse and a near-full recovery into a window no prior modern shock approaches. Reading it as the worst since the Depression captures one quarterly print and misses the mechanism that produced the recovery.
COVID is the episode most often compared to the Depression, and the comparison misleads most consistently. Annual data make the recession look modest: real GDP fell 2.1 percent in 2020 and rebounded 6.2 percent in 2021. Quarterly data show what households and firms lived through. Real GDP contracted at a −5.2 percent annualized rate in 2020 Q1 and −28.0 percent in 2020 Q2, then rebounded 34.9 percent in Q3.18 Quarterly output cleared its 2019 Q4 level in 2021 Q1, five quarters after the peak.
Annualizing projects one quarter's experience across four, so the −28.0 percent rate needs careful reading. The quarter-over-quarter decline was 7.9 percent, extraordinary and far smaller than the annualized figure. The two readings are the same number stated two ways: real GDP fell from $20,709.2 billion in 2020 Q1 to $19,078.0 billion in 2020 Q2, and compounding that 7.9 percent fall across four quarters gives the 28.0 percent annualized rate.19
The quarterly data make the mechanism visible. Public-health restrictions on movement, gathering and the operation of some businesses stopped activity almost at once, and households cut spending before the restrictions required it. Hotels, restaurants, airlines, live entertainment and personal services lost revenue within days. The recovery followed from the same mechanism: the stop restricted activity that would otherwise have continued, and it left capacity, balance sheets and expectations largely intact. Firms reopened, consumers resumed deferred spending, and employment in the hardest-hit sectors recovered faster than in most prior recessions, partly because firms had kept their workforces attached through enhanced unemployment benefits and the forgivable-loan program.
The state intervened on a scale without precedent. The CARES Act authorized about $2.2 trillion of headline fiscal support, which CBO and the Joint Committee on Taxation scored at about $1.8 trillion of added deficits over 2020 to 2030. Of that, $349 billion went to the Paycheck Protection Program, and successive rounds carried its cumulative approvals to about $800 billion across roughly 11.8 million loans by the time it closed on 31 May 2021. Expanded unemployment benefits added $600 a week to state payments through 31 July 2020, and direct payments of $1,200 per adult reached most Americans.20 The Federal Reserve cut the federal funds rate to the zero lower bound in March, established facilities to support corporate and municipal credit, and expanded its balance sheet by about $3 trillion, from $4.16 trillion on 26 February to $7.17 trillion on 10 June 2020.21
The CARES Act was signed on 27 March 2020, fourteen days after the national emergency declaration of 13 March and eight days after the first statewide stay-at-home order. The mechanism made a fast recovery possible, and the rescue was built to protect it: a reversible stop stays reversible only if firms survive it and households keep spending power through it. The evidence on what the rescue bought is mixed. David Autor and coauthors estimate that PPP preserved 2 to 3 million job-years over fourteen months at $169,000 to $258,000 per job-year, and that only 23 to 34 percent of its dollars reached workers who would otherwise have lost jobs.22 The rescue was costly per job and fast, and speed was the property the mechanism required. The Depression comparison treats the 2020 outcome as given, when it depended on a rescue the state could finance at near-zero rates.
The rescue left an inflation tail. CPI inflation ran at 1.2 percent in 2020, then 4.7 percent in 2021, 8.0 percent in 2022, 4.1 percent in 2023 and 2.9 percent in 2024; the 2022 reading was the highest since 1981. Energy CPI rose 41.5 percent in the twelve months to June 2022 after the invasion of Ukraine.23 Ben Bernanke and Olivier Blanchard attribute most of the initial surge to price shocks in energy, food and shortage-hit sectors, and a growing share of the later, stickier part to a tight labor market fed by strong demand.24 The demand share is the part policy chose. The Fed's first reading of the surge as transitory delayed the tightening, and the tightening that followed, from zero to above 5 percent between 2022 and 2023, brought inflation down without a Volcker-scale recession at the cost of some credibility.
A shock that sheds 14 percent of payroll employment in two months, from 152.3 million to 130.4 million, cannot be met by a fiscal response that takes a year to design.25 It needs pre-positioned automatic stabilizers, standing facilities with established legal authority, and institutions that can disburse at the pace of the shock. The 2020 response was fast partly because the 2008 crisis had already built the legal and administrative rails: the Federal Reserve re-established its 2008 commercial-paper and primary-dealer facilities on 17 March 2020, four days after the emergency declaration.26
The sudden-stop structure also meant that much of the lost activity was deferred and could resume. Transfers held household incomes up through the stop, so when the health constraint lifted, pent-up demand and the means to act on it were both real. A demand collapse of the 1930s type has no such built-in rebound, and a financial-system failure, a trade-network collapse or a sovereign debt crisis needs years to rebuild trust and credit. Calibrating future risk to the COVID recovery timeline is optimistic by construction whenever the mechanism differs.
The federal balance sheet that absorbed the COVID shock could not absorb a comparable shock today on the same terms. The capacity to act is intact. The price of a transfer at the 2020 scale has risen.
The variable that has changed most since the last downturn is the state's room to maneuver. The United States entered 2020 with fiscal credibility broadly intact, debt service that did not yet crowd out other outlays, and a political coalition that could authorize a multi-trillion-dollar rescue inside a month. The first two have since changed in ways this part measures; whether the third still holds is untested, and Part VI carries it as an assumption.
Federal net interest passed national defense in fiscal 2024, at $880 billion against $874 billion on the OMB basis, and has stayed above it.27 CBO's February 2026 outlook puts net interest at 3.3 percent of GDP in 2026 rising to 4.6 percent by 2036, and debt held by the public at 101 percent of GDP in 2026 rising to 120 percent, past the 1946 record of 106 percent.28 Interest is a non-discretionary claim that grows as debt rolls over at today's rates.
The price of a rescue can be measured where it is paid, at the yield on new borrowing, against 2019 as the last normal year. The three-month bill averaged 2.06 percent in 2019 and 3.61 percent in April 2026; the ten-year note averaged 2.14 and 4.32 percent. A rescue the size of the fiscal 2020 deficit, $3.1 trillion, costs about $64 to $66 billion a year in interest at 2019 yields and $112 to $134 billion at April 2026 yields, financed in bills or in ten-year notes. In April 2020 itself, with the bill at 0.14 percent and the note at 0.66, the same sum cost $4 to $20 billion.29 Those yields are the measured content of the claim that a rescue now costs more. The obvious reply is that yields fell in 2020 because of the crisis itself, and would fall again. In a demand collapse they would, as the Federal Reserve cut and investors fled to Treasuries. In a supply shock they need not fall, because a central bank facing rising prices cannot cut, and that asymmetry is why the 1970s template is the hard case for the fiscal constraint as well as for the monetary one.
The share of outlays is the most quotable version of the claim and the weakest. Fiscal 2021 is its most flattering base, because roughly $2 trillion of emergency spending inflated that year's denominator: from fiscal 2021 the share rises from 5.2 to 13.8 percent, and from fiscal 2019, a normal year, it rises from 8.4 to 13.8. The harder objection sits in Table A4: net interest took 12.5 percent of outlays in fiscal 2000, within a point and a half of today, in a year of budget surpluses. A high outlay share is compatible with abundant capacity, so the share is supporting evidence only, read against fiscal 2019.
What differs between 2000 and 2026 is the stock behind the share and its direction. In fiscal 2000 debt held by the public was 33.3 percent of GDP and falling, and interest was a large share of a small budget in surplus. In 2026 the stock is 101 percent of GDP and rising. The load-bearing measure is net interest as a share of GDP, which removes the denominator problem. CBO's 2026 outlay share dips to 13.5 percent only because its projected outlays of $7.4 trillion grow faster than the interest bill; as a share of GDP the climb continues.
The 1946 peak was financed in a different rate environment. Treasury bills were pegged at 0.375 percent and the long bond held under a 2.5 percent ceiling until the March 1951 Fed-Treasury Accord ended the peg, which kept the carrying cost of the war debt low while growth and inflation reduced it.30 Today's debt was largely issued in the decade of near-zero rates, and the notes and bonds sold in 2020 and 2021 carried coupons well below today's market yields, so each maturing issue refinanced at current rates adds to net interest without any rise in the stock.
The strongest objection is that the debt-service share is a weak proxy for fiscal capacity while the dollar holds reserve status. In a crisis, demand for Treasuries is a flight toward the issuer, which is how the United States financed 2020 at a negative real cost with a deficit of about 15 percent of GDP. On that reading a rising interest share describes the budget's composition and leaves what the state can borrow in an emergency unchanged, and 1946 shows a larger stock carried and then retired.
The evidence favors the objection on its own terms. The ten-year term premium is the price the market charges for the risk this report describes, and it has risen little: the Federal Reserve Board's Kim-Wright estimate averaged −0.18 percent in 2019 and about +0.65 percent in April 2026. Applied to a debt stock of about 100 percent of GDP, that rise would eventually add about eight-tenths of a point of GDP a year to interest, arriving only as the debt rolls over across several years. That is a modest charge, and it is the single strongest datum against the thesis. A longer frame weakens the thesis further: the premium ran well above zero through the 1990s and early 2000s, so +0.65 percent may simply be a return toward the pre-2010 range after a decade of central-bank balance-sheet policy. The level alone cannot separate the two readings.
The assembled opposing case uses this report's own evidence. The amplifier that turned 1929 into the Depression has been dismantled. The financial channel was interrupted in 2008 and in March 2020 by facilities that now stand as permanent capability, and response speed is cumulative: 2020 was fast because 2008 had built the rails. The 2022 disinflation carried CPI from 8.0 percent to 2.9 percent in two years without a Volcker-scale recession, with long-run breakevens near 2.2 percent.31 On that reading the state has rarely been better equipped, and its one degrading margin is a slow structural budget problem.
This report's claim is narrower than its title, and it concedes the capability built since 2008 without qualification. What has not been built is a way to finance the use of that capability at the 2020 price. The facilities that interrupted the financial channel are lending facilities: they expand a balance sheet against collateral and are repaid. The 2020 rescue was mostly transfers, made in a year when the deficit reached about 15 percent of GDP, and transfers are not repaid. Nothing in the post-2008 apparatus makes a transfer of that size cheaper to fund, and market yields on new borrowing are several points higher. The higher bill also raises the political price: each further dollar is voted in a Congress where interest already exceeds defense. The claim concerns one axis, the price of a fiscal transfer. Its full test is behavior under stress: if a growth scare arrives and term premia widen instead of compressing, the safe-haven bid has stopped absorbing the supply. Two readings would show the constraint tightening before any crisis: net interest above 3.5 percent of GDP in an outturn, and a term premium above 1 percent held for a quarter outside a growth scare.
Each mechanism would meet the constrained state at a different joint. A demand collapse of the 1930s type would test a fiscal authority that can no longer respond at the 2020 scale on 2020 terms. A supply repricing of the 1970s type would test a Federal Reserve without a second Volcker disinflation in its credibility budget, pressed to stimulate exactly when stimulus would worsen prices. A sudden stop of the COVID type would test legislative speed, which Part VI carries as an assumption of six to nine months for an emergency without pandemic-level clarity.
The financial template deserves a hearing first. Leveraged spending on AI compute and the growth of private credit concentrate maturity and valuation risk outside the insured banking system, and both have grown faster than the supervision that would see them. The financial channel is also the one the post-1930s architecture was built to interrupt, and it has been interrupted twice by facilities that now exist as standing capability. A financial shock would be the more dangerous if it arrived; it is also the one the state is best equipped to meet.
The supply channel has no equivalent backstop. The 2022 release of 180 million barrels from the Strategic Petroleum Reserve was the largest in its history and drew the reserve to 347 million barrels by July 2023, its lowest since 1983; it held 398 million in late April 2026,32 and tariff relief, export-control adjustments and strategic stockpiles work on scales and timetables that no energy or semiconductor shock respects. There is no discount window for the price of a physical input, and in this channel the constrained state and the constrained central bank bind at the same moment. That is why this report ranks the 1970s first as the template to plan for: it is the hardest to answer with the tools now available.
The ranking is conditional on a variable the report can read today. The 1970s injury was dislodged inflation expectations, and expectations are anchored: the five-year, five-year forward breakeven sits near 2.2 percent, and Part III argues that anchoring is more precise now than in 1979. So the 1970s template is a case to plan for, and the evidence does not yet support forecasting it. What would change that is a breakeven above 2.5 percent, held through a cost-push episode the Federal Reserve is seen to accommodate. Tariffs are the live test. A tariff is a one-time shock to the price level, and it drops out of the twelve-month comparison unless monetary policy accommodates it into wages; that distinction is most of the difference between 1973 and 2026, and the breakeven says accommodation is not expected.
The Federal Reserve's credibility was partly spent in the transitory episode of 2021 and 2022 and partly restored by the tightening that followed. The net position is weaker than in February 2020, so the next request to look through a temporary spike will meet more scrutiny. The problem bites hardest if the next shock carries weak demand and cost pressure at once: the Fed would be pressed to cut for demand and to hold for prices, and the outcome would turn on which it chose.
The CARES Act passed the Senate 96 to 0 on 25 March 2020.33 That vote reflected the clarity of a pandemic, a visible, shared and immediate danger that removed the normal resistance to emergency spending. A demand collapse, a supply-shock inflation or a financial stress would lack that clarity, which is why Part VI assumes six to nine months for such a case. The 2020 package took fourteen days. The bottleneck is the legislative capacity to authorize funds before the economy has stabilized on its own or deteriorated further.
Three variables would confirm or weaken the claim that the price of a 2020-scale rescue has risen. Debt service is rising and confirms it. The term premium has not moved the way the argument needs. Legislative speed is an assumption no source measures, and it is carried as one.
Three variables carry the fiscal half of the thesis. The first is the cost of servicing the debt, read as a share of GDP and cross-checked against the share of outlays. The second is the term premium on long-dated Treasuries during a growth scare, which shows how the bond market prices the credibility of a response. The third, legislative speed, is an assumption built from four observed enactments.
Debt service is rising, and it is the one measured variable that confirms the thesis. CBO puts net interest at 3.3 percent of GDP in 2026 rising to 4.6 percent by 2036, and the outlay share reached 13.8 percent in fiscal 2025 against 8.4 percent in fiscal 2019.34 Each point of outlays absorbed by debt service is a point unavailable for a discretionary response, and the trajectory points to further deterioration absent a change in the structural deficit or in interest rates.
The term premium has risen from −0.18 percent on the 2019 average to about +0.65 percent in April 2026.35 A premium that widens during a growth scare, when a flight to Treasuries would normally compress it, would be the clearest market signal that fiscal credibility is under stress. That has not happened. This variable does not confirm the thesis, and on the pre-2010 frame Part V sets out, its level carries no fiscal-risk charge at all. One measured variable confirms, one does not, and the third is an assumption: enough to make the case worth monitoring, too little to say the buffer is degrading on every margin at once.
Legislative speed matters most for sudden stops, and it is the hardest to measure. The observed cases bracket the range, each timed from its own trigger. TARP moved from Treasury's three-page draft of 20 September 2008 to enactment on 3 October, thirteen days, under an acute and visible market failure. CARES ran fourteen days, from the emergency declaration of 13 March 2020 to enactment on 27 March. The Recovery Act was signed on 17 February 2009, 105 days from the 4 November 2008 election that produced its majority. The December 2020 relief package was enacted on 27 December, five months after the CARES unemployment supplement expired on 31 July, in a Congress that had voted 96 to 0 on the same subject. The two fast cases both carried a shock the whole chamber could see, and a contested diagnosis removes that shock, so the report treats six to nine months as its assumption for a contested emergency, an extrapolation past all four cases that no source measures.36 For a COVID-type sudden stop, a six-month fiscal response arrives after the moment when bridge financing mattered.
Exhibit 8. One measured variable moves the way the thesis predicts, one does not, and the third is an assumption
The three watch variables, each on its own scale
| Then | Now and next | |
|---|---|---|
| Debt-service share of outlays | 8.4 percent of federal outlays in fiscal 2019 | 13.8 percent in fiscal 2025; CBO puts net interest at 3.3 percent of GDP in 2026 and 4.6 by 2036 |
| Treasury term premium (not confirming) | Minus 0.18 percent on the 2019 average, a balance-sheet-era reading | About plus 0.65 percent in April 2026, which is also consistent with normalization toward the pre-2010 range |
| Emergency authorization time (assumption) | Thirteen days for TARP from Treasury’s draft; fourteen for CARES from the emergency declaration; 105 days for ARRA from the 2008 election; five months for the December 2020 package from the supplement’s expiry | Six to nine months assumed for an emergency without pandemic-level clarity, an extrapolation beyond all four observed cases, each of which is timed from its own trigger |
Source: CBO, Budget and Economic Outlook: 2026 to 2036, February 2026, OMB Historical Tables 3.1 for fiscal 2019 and Treasury outturns for fiscal 2025 (debt service); Federal Reserve Board Kim-Wright model via FRED THREEFYTP10 (term premium); enactment dates for P.L. 110-343, P.L. 111-5, P.L. 116-136 and the Consolidated Appropriations Act, 2021, with the six-to-nine-month figure a stated Power Curve assumption extrapolating beyond all four. A monitoring scorecard.
The report sets its confirming readings as mirrors of the falsifiers in the watchlist that follows: a term premium that widens through the next growth scare, net interest above 3.5 percent of GDP in an outturn, or an emergency without pandemic-level clarity left unauthorized beyond sixty days. Any one of them would move the thesis from a planning case toward a constraint already binding.
Allocators and risk managers should stop using annual output as their main measure of recession severity. Liquidity planning, covenant exposure and labor-cost flexibility deserve separate stress tests against each axis. In 2020 the quarterly data ran ahead of the annual data by most of a year, so a framework that waits for annual GDP to cross a threshold misses a sudden stop entirely. The 1970s showed that an inflation regime can do more damage to a balance sheet than a deeper but cleaner contraction, because duration mismatch becomes a balance-sheet loss when rates stay high.
Policymakers and boards should treat the 1970s as the template to plan for first, on the condition Part V states: the expectations channel that defined the decade is anchored today. Anyone modeling the state's response should assume a smaller and slower fiscal reaction than the 2020 baseline, because the report assumes six to nine months to authorize relief in an emergency without pandemic-level clarity, against fourteen days in March 2020, and because its financing would cost more, although the term premium is not yet pricing that difference. The analog also points to what would help most: pre-positioning that reduces the need for discretionary action under stress, through automatic stabilizers, standing credit facilities with pre-authorized capacity, and fiscal consolidation that rebuilds headroom before the shock arrives. In a constrained environment the wrong tool does active harm: demand stimulus against a supply shock spends fiscal and monetary headroom while making the price problem worse. The watchlist gives the decision rule: a breakeven above 2.5 percent held for a quarter, or net interest above 3.5 percent of GDP in an outturn, is the signal to move contingency plans from the 2020 template to the 1970s one.
Five variables will show whether the thesis holds, each observable in public data within a year: the three scorecard variables, and the policy rate and the breakeven that condition them. The triggers are the readings that would falsify it, set on the debt-service burden, the behavior of term premia under stress, the political clarity of the next trigger, and whether inflation expectations are firmer than the current read suggests.
| Variable | What a move would mean | Reading | Falsifying trigger | Horizon |
|---|---|---|---|---|
| Net interest, share of GDP | The interest cost of a rescue stops rising, and the load-bearing claim weakens | 3.2% (FY2025) | Below 3.0% in two consecutive fiscal years | FY2027 outturn |
| Ten-year term premium, Kim-Wright | The market stops charging for fiscal risk, and the argument loses its market signal | +0.65% (April 2026) | Compresses through the next growth scare, or holds below +0.25% for two quarters | Next growth scare |
| Federal funds target range | A lower policy rate cuts the rollover cost driving the debt-service climb | 3.50–3.75% (April 2026) | Below 3.00% with the interest share falling | FY2027 |
| Five-year, five-year forward breakeven | Expectations stay anchored, so no 1970s-type duration dynamic can start | About 2.2% (April 2026) | Below 2.0% held for two quarters | Rolling |
| Days from trigger to enacted relief, each case on its own trigger | A fast authorization shows the legislative-speed constraint is not binding | Five months, December 2020 package, the last test; 13 days to five months across four cases | An emergency without pandemic-level clarity enacted inside 60 days | Next downturn |
Each falsifying trigger is set in the direction of the move that would change the view. Every row is a falsifier: the reading that would weaken this report. Confirming readings sit in Part VI. Readings are as of the April 2026 cutoff, except the legislative-speed row, which reports the last observed test, the December 2020 package, because no emergency has been put to Congress since.
Source: OMB via FRED (FYOIGDA188S); Federal Reserve Board Kim-Wright model (FRED THREEFYTP10); FOMC statements of 18 March and 29 April 2026; FRED T5YIFR; enactment dates from the public laws named in Part VI. Data through April 2026.
Each falsifier is achievable, and none has happened. The three axes serve at once as a diagnostic framework, a monitoring framework and a planning tool, and their value is greatest under ambiguity, which is how most serious downturns arrive: the 2007 to 2008 crisis looked like a housing recession in early 2007, COVID looked like a normal demand disruption in January 2020, and the 1979 to 1980 episode looked like an ordinary post-oil-shock slowdown until Volcker's tightening showed that something different was required. The next shock will not announce its type, and a response matched to the wrong one now costs nearly twice as much to finance as it would have in 2019.
This report was published on 21 October 2026 on data through April 2026. The readings below are as of that cutoff and have not been refreshed for publication.
CBO's February 2026 outlook keeps net interest above defense in fiscal 2026, as it has been since fiscal 2024, rising to 4.6 percent of GDP by 2036, with debt held by the public at 101 percent of GDP this year and 120 percent by 2036. The Federal Open Market Committee held its target range at 3.50 to 3.75 percent on 18 March and again on 29 April 2026, after cuts in late 2025, so the notes issued at near-zero coupons in 2020 and 2021 still refinance well above those coupons as they mature. No agreement on medium-term fiscal consolidation has been announced. Tariff increases remain the live source of input-cost pressure, the mechanism closest to the 1970s analog. Of the watch variables, debt service has moved the way the thesis predicts; the term premium has not, and long-run breakevens remain anchored.
Every figure traces to a named, dated source. Real output is chained 2017 dollars; quarterly rates are annualized, with the quarter-over-quarter reading stated where the difference matters. Inflation is headline CPI, annual average, on one basis for all three episodes. Outlay shares are fiscal years: OMB published shares to 2023, Treasury outturns for 2024 and 2025, and CBO's projection for 2026, labeled wherever it appears. Vintage: data run to the April 2026 cutoff; national-accounts figures are the BEA vintage read on 24 September 2026, so 2020 Q2 reads −28.0 percent annualized and −7.9 percent quarter over quarter. Limitations: the 1929 to 1933 accounts are a BEA backcast, so the 26.3 percent is precise to that backcast and not to the decimal; Depression-era unemployment and prices rest on NBER and Lebergott reconstructions; the term premium and breakeven are as read from FRED on 24 September 2026; and the six-to-nine-month legislative interval is a stated assumption.
| Dataset | Source | Frequency | Coverage | Retrieved | Used in |
|---|---|---|---|---|---|
| Annual real GDP, level and growth | BEA, via FRED (GDPCA, A191RL1A225NBEA) | Annual | 1929 to 2021 | Sep 24, 2026 | Exhibits 1–2, Parts I–IV |
| Quarterly real GDP, level and growth | BEA, via FRED (GDPC1, A191RL1Q225SBEA) | Quarterly | 1973 Q4 to 2021 Q2 | Sep 24, 2026 | Exhibits 2, 5, Parts I, III, IV |
| Unemployment rate | BLS (UNRATE); NBER; Lebergott | Monthly | 1929 to 2020 | Sep 24, 2026 | Exhibit 3, Parts I–IV |
| Consumer price index, all items | BLS, via FRED (CPIAUCNS) | Annual average | 1929 to 2024 | Sep 24, 2026 | Exhibit 4, Parts II–IV |
| Policy rate and Treasury yields | H.15 via FRED (FEDFUNDS, TB3MS, GS10); FOMC | Monthly | 1979 to April 2026 | Apr 29, 2026 | Parts III, V, watchlist |
| Net interest and national defense outlays | OMB Historical Tables 3.1; Treasury Combined Statement; CBO Feb 2026 | Fiscal year | 2000 to 2026 | Feb 2026 | Exhibit 6, Parts V–VI |
| Federal debt held by the public | OMB via FRED (FYPUGDA188S); CBO | Fiscal year | 1939 to 2026 | Apr 13, 2026 | Exhibit 7, Part V |
| Ten-year term premium | Federal Reserve Board Kim-Wright model, via FRED (THREEFYTP10) | Daily | 2019 to April 2026 | Sep 24, 2026 | Parts V and VI, watchlist |
| Bank failures, 1930 to 1933 | FDIC, The First Fifty Years | Episode | 1930 to 1933 | Historical | Part II |
| Fiscal and monetary response, 2020 | P.L. 116-136; SBA; Federal Reserve H.4.1 (WALCL) | Event | Feb to Jul 2020 | Sep 24, 2026 | Part IV |
Every national-accounts figure is the current BEA vintage, read September 24, 2026. Two move on that choice: 2020 Q2, whose superseded advance estimate carried a −9.5 percent quarterly fall against the −7.9 used here, and the 2021 rebound, raised from 5.9 to 6.2 percent.
The 2021 reading is 6.2 percent on BEA's current vintage; 5.9 percent was the 2022-vintage print and is retired.
| Year | Episode | Change |
|---|---|---|
| 1930 | Depression | −8.5 |
| 1931 | Depression | −6.4 |
| 1932 | Depression | −12.9 |
| 1933 | Depression | −1.2 |
| 1974 | 1970s | −0.5 |
| 1975 | 1970s | −0.2 |
| 1980 | 1970s | −0.3 |
| 1982 | 1970s | −1.8 |
| 2020 | COVID | −2.1 |
| 2021 | COVID | +6.2 |
Source: BEA via FRED (A191RL1A225NBEA), chained 2017 dollars. Data through 2021, read September 24, 2026. Behind Exhibit 1.
The Depression column is annual and indexed to 1929; the other two are quarterly and indexed to 1973 Q4 and 2019 Q4. COVID crosses 100 at 1.25 years, which is 2021 Q1, five quarters from the peak. The 1970s column clears 100 at 2.00 years, which is 1975 Q4.
| Years from peak | Depression, annual | 1970s, quarterly | COVID, quarterly |
|---|---|---|---|
| 0.00 | 100.0 | 100.0 | 100.0 |
| 0.25 | n/a | n/a | 98.7 |
| 0.50 | n/a | n/a | 90.9 |
| 0.75 | n/a | n/a | 98.0 |
| 1.00 | 91.5 | n/a | 99.1 |
| 1.25 | n/a | 96.9 | 100.5 |
| 1.50 | n/a | n/a | 102.2 |
| 2.00 | 85.6 | 100.6 | n/a |
| 3.00 | 74.6 | n/a | n/a |
| 4.00 | 73.7 | n/a | n/a |
| 5.00 | 81.6 | n/a | n/a |
| 6.00 | 88.9 | n/a | n/a |
| 7.00 | 100.3 | n/a | n/a |
Source: BEA via FRED (GDPCA, annual; GDPC1, quarterly). Data through 2021 Q2, read September 24, 2026. Behind Exhibit 2.
The quarter-over-quarter column is the annualized rate taken to the one-quarter power, which is the identity the report uses to show that −28.0 annualized and −7.9 quarter over quarter are the same number stated two ways.
| Quarter | Annualized | Quarter over quarter |
|---|---|---|
| 1974 Q3 | −3.7 | −0.9 |
| 1975 Q1 | −4.8 | −1.2 |
| 2020 Q1 | −5.2 | −1.3 |
| 2020 Q2 | −28.0 | −7.9 |
| 2020 Q3 | +34.9 | +7.8 |
Source: BEA via FRED (A191RL1Q225SBEA, GDPC1). Data through 2020 Q3, read September 24, 2026. Behind Exhibit 5.
Three bases sit on one line and the column says which is which. The 2026 row is CBO's February 2026 projection of $1.0 trillion of net interest and $885 billion of defense against $7.4 trillion of outlays. Net interest passed defense in fiscal 2024, $880 billion against $874 billion.
| Fiscal year | Net interest | National defense | Basis |
|---|---|---|---|
| 2000 | 12.5 | 16.5 | OMB published share |
| 2009 | 5.3 | 18.8 | OMB published share |
| 2019 | 8.4 | 15.4 | OMB published share |
| 2021 | 5.2 | 11.1 | OMB published share |
| 2023 | 10.7 | 13.4 | OMB published share |
| 2024 | 13.1 | 13.0 | Treasury outturn |
| 2025 | 13.8 | 13.1 | Treasury outturn |
| 2026 | 13.5 | 12.0 | CBO projection |
Source: OMB, Historical Tables 3.1, for 2000 to 2023; Treasury Combined Statement and OMB totals for 2024 and 2025 (FRED FYOINT, FYONET); CBO, Budget and Economic Outlook: 2026 to 2036, February 2026. Data through February 2026. Behind Exhibit 6.
Hart, Brendan. 2026. "Depth, Speed, Duration." The Power Curve, October. https://brendanhart.com/reports/depth-speed-duration.
