Governing by Deadline

What this paper establishes, before the argument.

Three times in thirteen months the United States federal government has run out of the legal authority to spend its own money. It happens again on 11 December, thirty-five days from now, and the deadline falls to a Congress that was voted out five weeks before it arrives. Not one of the twelve appropriations bills for the current fiscal year has been enacted, and none has passed the Senate.

What follows is an account of what that costs, what it does not cost, and why the market has been right to ignore almost all of it. Seven findings, with the evidence for each in the sections that follow.

• 43 days. Longest government-wide shutdown on record
• 116 days. Border patrol and ice unfunded, 2026
• 0 of 12. Fy2027 appropriations bills enacted

• Length is not severity, and almost every account of this subject assumes it is. The 2026 lapse at Homeland Security ran 76 days, with Border Patrol and Immigration and Customs Enforcement unfunded for 116, the longest any part of the United States government has ever gone without appropriations. It left no measurable footprint in the national accounts at all. Scope decides the damage.
• The reason it left no footprint has just been taken away. About 85 per cent of that department kept working because it was being paid out of reconciliation money rather than appropriations. The same instrument raised the debt limit by US$5 trillion in 2025. Both were products of one party holding the House, the Senate and the presidency, which ended on 3 November. The next lapse is therefore capable of doing more measurable damage than the last two did, even if it is shorter.
• The direct economic cost is real and small. The Congressional Budget Office put the permanently unrecovered output loss from the 43-day shutdown at US$7bn to US$14bn, about US$11bn on the scenario that occurred. Fourth-quarter 2025 growth fell to 0.5 per cent annualised from 4.4 per cent, and the following quarter recovered to 2.1 per cent.
• Markets did not price it, and were right not to. The S&P 500 rose about 2.8 per cent across the 43 days and the ten-year Treasury yield fell eight basis points. The week of the midterm made the point at far greater volume: on 5 November the S&P 500 fell 4.4 per cent in its worst session since March 2020, five weeks out from a funding cliff, and none of it ran through the appropriations cycle.
• The bond market tells a shutdown from a debt-limit date to the basis point. Bills maturing around the 2023 debt-limit date carried roughly 100 basis points of distortion. The 2025 shutdown produced none anywhere in the curve. A shutdown never touches debt service. An exhaustion of borrowing authority does, and that date is expected during 2027.
• The damage that does not reverse is to the measurement system, and it is permanent. The October 2025 household survey was never fielded and the October 2025 consumer price index was never collected, breaking series that had run continuously since 1948 and 1921. Neither can be reconstructed, and every model priced off those series now has a hole in it.
• The binding constraint is net interest, and the market has just bought gridlock as a substitute for fixing it. Net interest reached a record US$1.05 trillion in fiscal 2026, about 19 per cent of federal revenue. Treasuries rallied on the election because a divided Washington passes less and therefore issues less. Capping new borrowing is not the same thing as correcting the trajectory, and only the second needs a legislature that can agree.

Sections 1 to 4 set out the record, the mechanism and the measured cost. Section 5 is the market evidence. Section 6 is 11 December and the 2027 debt limit. Section 7 is what follows from it.

Shutdowns are not the disease. They are the test result.

Three lapses in thirteen months.

Appropriations lapsed across all twelve bills on 1 October 2025 over the expiry of the enhanced Affordable Care Act premium tax credits. It ran 43 calendar days and ended with Public Law 119-37, after the Senate passed it 60-40 and the House concurred 222-209. No official count of furloughed employees exists; the defensible range is 1.3 to 1.6 million federal employees affected.

The side that forced the shutdown lost. The deal bought a guaranteed Senate vote on the subsidies and no House vote. The Senate votes failed on 11 December 2025 and the credits expired. Average monthly net premiums rose 58 per cent from US$113 to US$178, average deductibles rose 37 per cent to a record US$3,786, and sign-ups fell by more than a million to 23.1 million, the sharpest single-year fall since the exchanges opened. Forty-three days of closure and about US$11bn of permanently lost output bought nothing.

Ten weeks later the same instrument was used again. Customs and Border Protection agents killed Alex Pretti and Renee Good in Minneapolis in January 2026, Senate Democrats withdrew support for the Homeland Security bill, and appropriations lapsed on 31 January. Homeland Security funding then lapsed alone from 14 February for 76 days.

The mechanics of that third lapse matter more than its length. About 85 per cent of the Homeland Security workforce, some 200,000 people, kept working, because Border Patrol and Immigration and Customs Enforcement were funded from the 2025 reconciliation law and about US$10bn of multi-year reconciliation money paid some 70,000 law enforcement personnel. The department was shut and its principal enforcement arms were not. The bite fell on the Transportation Security Administration, whose officers worked unpaid, missed a full pay packet on 13 March, and of whom more than 1,110 out of roughly 50,000 resigned by 27 April.

The scopes are not comparable and should not be summed into a single headline. Days of government-wide lapse: 43. Calendar days on which at least one agency was unfunded: 164 of the 437 days to the next deadline.

All twelve fiscal 2026 bills were finally enacted on 30 April 2026, seven months into the fiscal year. For fiscal 2027, a continuing resolution passed the Senate 90-6 on 8 August and the House 370-48 on 1 September, and was signed on 2 September, funding agencies to 11 December and removing the 1 October cliff ahead of the midterms. Three of the twelve bills have passed the House. None has passed the Senate. None has cleared Senate Appropriations.

Why it recurs, which is a question about machinery rather than personalities.

The law converts a disagreement into a closure, and guarantees it saves nothing

The Antideficiency Act bars federal agencies from obligating funds without an appropriation, with removal and, for a wilful violation, criminal penalties. It was written in the nineteenth century to stop agencies overspending and then presenting Congress with a bill it had to pay.

Before 1980 agencies generally kept working through a lapse and Congress ratified the obligations afterwards. A Comptroller General decision in March 1980 and the two Civiletti opinions of April 1980 and January 1981 ended that, holding that agencies must suspend operations and furlough staff. Civiletti expressly noted that the earlier opinion had not addressed a general congressional failure to appropriate on time. The regime that now governs the recurring case was never designed for it.

The Government Employee Fair Treatment Act of 2019 then guaranteed retroactive pay. The combination is routinely missed and it is decisive. A shutdown defers payroll, pays it in full later, and destroys the output that would have been produced in between. There is no fiscal argument for a shutdown. The only argument is leverage.

Regular order stopped producing appropriations a generation ago

Congress has enacted all twelve regular appropriations bills on time four times in roughly fifty years, most recently for fiscal 1997. It passed continuing resolutions in 44 of the 47 fiscal years from 1977 to 2023, more than two hundred in total. The twelve-bill process is no longer the normal mechanism of federal funding. The stopgap is, and every stopgap creates a new deadline and a new hostage.

The supermajority applies to exactly the spending that can lapse

Appropriations bills and continuing resolutions require sixty votes for cloture. Reconciliation requires fifty-one. Discretionary spending, the only category that lapses, is therefore the one category the majority cannot move alone, while mandatory spending and tax policy can be legislated around the minority entirely.

The 2026 Homeland Security episode is the clean demonstration. The minority's leverage over discretionary appropriations was not overcome. It was routed around: Border Patrol and Immigration and Customs Enforcement were funded through reconciliation in June 2026, on 52-47 in the Senate and 214-212 in the House. Funding operating agencies that way makes them immune to shutdown leverage, which lowers the value of taking the hostage. It also moves spending out of the process that exists to control it, on bare partisan majorities, for multiple years, with weaker direction and oversight. If that becomes the pattern, shutdowns get less effective and the congressional power of the purse gets weaker at the same time.

There is no overlap left between the parties

On the first dimension of the standard congressional voting scale, the distance between party means in the 119th Congress is 0.910 in the House and 0.923 in the Senate, the highest readings in a series running from 1879. Under ten per cent of members in either chamber are moderates. Party overlap is exactly zero and has been since 2001 in the House and 2003 in the Senate: no Republican is less conservative than the most liberal Democrat, or the reverse. Senate polarisation now exceeds the House's, which matters because the Senate is the chamber where forty-one votes is a veto.

The tactic survives its own failure, and the reform that would stop it is stalled

A tactic repeated after it has demonstrably failed is not being used to obtain an outcome. It is being used to establish a position. The Prevent Government Shutdowns Act would impose rolling automatic continuing resolutions and confine members to Washington with no votes on anything but appropriations. It has twenty-four bipartisan cosponsors and has sat on the Senate calendar since 1 June 2026 without floor action. The Senate did find time to vote 99-0 for cloture on a resolution withholding senators' own pay during shutdowns, which binds only the Senate, takes effect only after the November midterms, and does not have the force of law.

No comparable economy has this problem, for identifiable reasons

Germany's Basic Law authorises the federal government to keep making necessary payments if the budget is not adopted. Canada has Governor General's Special Warrants. The United Kingdom has votes on account and a confidence convention under which losing a supply vote can bring down the government. Japan passes a provisional budget. Australia, the closest structural analogue with an elected upper house able to block supply, resolved its one real episode in 1975 by dismissing the Prime Minister.

The United States combines separated powers, a supermajority threshold in one chamber, and a statute that criminalises obligation without appropriation, with no default spending authority of any kind. No comparator has all three. In every other system the penalty for deadlock falls on the government. In the United States it falls on federal employees and the public.

A quarter of growth, substantially recovered.

The Congressional Budget Office estimated that the shutdown would reduce fourth-quarter 2025 annualised growth by 1.0 to 2.0 percentage points depending on length, and that US$7bn to US$14bn of output would never be recovered. The realised drag was about 1.0 percentage point from reduced federal labour services, with federal spending subtracting 1.15 points, the largest such drag since the first quarter of 1994, as federal spending fell at a 16.6 per cent annualised rate.

The 2025 shutdown was 1.23 times the length of the 35-day shutdown of 2018 and 2019 but cost roughly 3.7 times as much permanent output, US$11bn against US$3bn. The difference is scope. The 2026 Homeland Security lapse completes the proof in the opposite direction: far longer than any precedent, and no measurable national-accounts effect, because reconciliation money substituted for appropriations across most of the department.

The documented transmission is into capital formation, not consumption

Three channels are properly evidenced.

• Securities registration stopped. About 90 per cent of the Securities and Exchange Commission's staff were furloughed, more than 900 registration statements were filed into a closed agency, and more than a dozen companies revised listing timelines. Registration statements cannot be declared effective while the Commission is shut.
• Cross-border review stopped. CFIUS filing reviews were tolled for the entire 76 days of the Homeland Security lapse and resumed only on funding. This is the one clean, verifiable 2026 transmission channel, and it is institutional rather than a price.
• Credit stopped. The Small Business Administration reported US$5.3bn of lending to about 10,000 small businesses blocked during the 43 days. Federal contractors, unlike federal employees, have no statutory right to back pay.

Beyond those, no credible quantified evidence of broader private investment deferral attributable to the shutdowns could be found, and none is asserted here. The closest authoritative precedent, the Federal Reserve's Beige Book of March 2019 on the 2018 and 2019 shutdown, found consumption effects across about half the districts and did not report investment deferral.

One attribution error to avoid

Federal employment has fallen from a January 2025 peak of 3.010 million to 2.674 million in August 2026. This is overwhelmingly a policy-driven workforce reduction through deferred resignations, incentivised exits and hiring freezes. The Government Accountability Office, which recorded a net reduction of about 256,000 positions across twenty-two agencies in the thirteen months to January 2026, does not attribute any of it to the shutdowns. The workforce effects properly attributable to the shutdown cycle are the 3,605 reduction in force notices issued during the 2025 lapse and later rescinded in full, and the Transportation Security Administration attrition in 2026.

The damage that does not reverse.

The Bureau of Labor Statistics cancelled nine releases outright. For most of them the release was cancelled and the data published later. Two were different.

The October 2025 household survey was never fielded. The Bureau has confirmed the data were not collected for that reference period and will not be collected retroactively. There is no October 2025 unemployment rate, no participation rate and no household-survey measure of any kind, and there never will be. That series has run since 1948. The Bureau also collected no consumer price data between 1 October and 12 November 2025 and declined retroactive collection, since price surveys are not recall-based. No October 2025 all-items or core index was ever published. The consumer price index has been published every month since 1921.

• 1948. Unemployment rate series, now with a hole in it
• 1921. Consumer price index series, now with a hole in it
• 64%. Current population survey response rate, nov 2025 record low

The Bureau of Economic Analysis cancelled the third-quarter 2025 advance GDP estimate outright and had to build the national accounts across the gap by imputing October price indexes. The Consumer Expenditure Survey lost both October and November 2025, leaving fourth-quarter consumer units at about one third of the usual number, and the adjustments made for that feed through into the February 2027 consumer price index weight update.

The quality cost has been measured by the agency itself

Carry-forward imputation for the October 2025 housing component understated the twelve-month change in rent by 0.25 to 0.55 percentage points and in owners' equivalent rent by 0.28 to 0.61 points, against four research alternatives. The distortion converged by April 2026 when the panel was next collected, so that particular wound healed. The blackout is nonetheless still degrading published output: twelve-month percentage changes for October 2026 will not be published for indexes where the October 2025 comparison is unavailable.

The revision record shows what degraded source data does to a number people trade on. Fourth-quarter 2025 growth was published at 1.4 per cent on 20 February 2026, cut to 0.7 per cent in March and finalised at 0.5 per cent on 9 April. Anyone positioned on the first print was positioned on a figure that turned out to be wrong by nearly a full percentage point.

The structural weakness is worse than the blackout, and it is not being fixed

The acute episode has largely washed out. The structural position has not. All thirteen principal statistical agencies have lost staff since January 2025; six have lost at least a third and two more than two thirds. The Bureau of Labor Statistics is down 20.5 per cent, the Census Bureau 19.3 per cent and the Bureau of Economic Analysis 20.9 per cent, and the fiscal 2027 budget request proposes further cuts to the first two. Only six of the thirteen agencies have permanent heads. The Census Bureau has had no Director since early 2025 and no nominee.

The Bureau's own report to the appropriations committees is the most damaging document in the file. The 2026 sample of 60,000 eligible households is the same as in 1981, against a population base 61 per cent larger. Fewer than fifty net changed responses can move the headline unemployment rate by a tenth of a point. Twenty years ago a monthly change of 0.18 points was detectable within one month; it now takes two. The proposed remedy costs about US$60m a year.

Monetary policy ran on the degraded series, and the subject has since disappeared

In October 2025 the Federal Reserve chair described the problem as driving in fog, and said the response was to slow down. The December 2025 minutes recorded participants relying on private-sector and limited government data. None of the six 2026 policy statements mentions the shutdown, delayed releases or data gaps. The September 2026 minutes do not mention funding risks or data availability. The October 2026 Beige Book contains no reference to either.

The absence is noted because the absence is the finding. A measurement system with permanent holes in it, a fifth of its staff gone and survey response below seventy per cent has been absorbed as normal within a year. On 16 September 2026 the Committee raised the target range to 3.75 to 4.00 per cent on a 12-0 vote, its first increase since 2023, with headline consumer price inflation at 3.7 per cent.

Private substitutes do not close the gap. For September 2025, one commercial payroll series reported 32,000 jobs lost while another estimated 60,000 gained. A divergence in both sign and magnitude cannot be resolved without an official anchor, and many private series are benchmarked to the official data they are supposed to replace.

Not the shutdown cycle.

The S&P 500 opened the 43-day shutdown at 6,664.92 on 1 October 2025 and closed at 6,850.92 on 12 November, a gain of about 2.8 per cent, with the Nasdaq Composite up around 4 per cent. On the first morning the index was down 0.1 per cent and within reach of an all-time high. Output is recovered, employees are paid in arrears by statute, the spending affected is a small share of the economy.

The week of the midterm was a cleaner experiment than any shutdown

On Wednesday 4 November, with both chambers called for the Democrats, the Dow rose more than a thousand points and the ten-year Treasury yield fell to 4.84 per cent, its lowest since 10 September. The stated reasoning was explicit and it was about supply: a divided Washington passes less, and a government that passes less issues less. On Thursday 5 November the same market had its worst session since March 2020. The Nasdaq 100 fell more than 6 per cent, the S&P 500 4.4 per cent, the Philadelphia semiconductor index 12 per cent into a bear market 38 per cent below its June peak, and the VIX rose 11.57 points to 28.18. Treasuries were bid again, with yields down as much as 12 basis points.

Neither session was about the appropriations cycle. The rally priced an expectation about how much legislation would pass. The rout priced artificial intelligence valuations and the prospect of a policy shift on data centres. Thirty-six days from a funding deadline for which there is no agreed plan, with none of the twelve bills enacted, the American equity market moved four and a half per cent in a session for reasons that had nothing to do with it.

The bill curve distinguishes a shutdown from a debt-limit date, precisely

A shutdown is a lapse in appropriations: agencies cannot spend, but the Treasury's borrowing authority and debt service are untouched. An exhaustion date is the end of borrowing authority: debt service itself is in question.

Bills maturing around the August 2023 debt-limit date yielded roughly 100 basis points away from bills maturing in other months, and credit default swap implied risk-adjusted default probabilities peaked near 6 per cent in 2011 and 4 per cent in 2013 and 2023. Across the 43-day shutdown of 2025 the one-month bill moved from 4.20 to 4.17 to 4.20 per cent and the three-month from 4.02 to 4.01 to 4.03, with no kink and no identifiable premium anywhere in the curve.

The long end has repriced hard, for reasons that are not about governance

The ten-year Treasury yield rose above 5.35 per cent on 7 October 2026, its highest since 2002, with the thirty-year at 5.70 per cent. The drivers are identifiable and governance is not among them: inflation, with energy prices lifted by the Iran conflict; a central bank raising rates with more priced beyond; heavy long-dated corporate supply, with artificial intelligence related issuance estimated at US$489bn year to date; and real yields, with the ten-year inflation-protected yield rising from 1.90 to 2.94 per cent across the year. Above all this is a global move. Japanese ten-year yields are at three-decade highs, gilts at their highest since 2008 and French ten-year yields approaching 5 per cent. Models attribute most of the 2026 rise in the US ten-year to expected short rates rather than to the term premium.

That peak did not hold: the November rally took roughly half a percentage point back out of it. Selloff and rally were both global moves about inflation, supply and risk appetite, and neither was a verdict on the appropriations cycle.

Term premium estimates themselves disagree by about 45 basis points across the three standard models. No single one of them should be quoted as a fact.

There is a structural change underneath that does bear on fiscal credibility, and it is about who has to be paid to hold the paper. The foreign share of Treasuries outstanding has fallen from 49 per cent in September 2013 to about 33 per cent in September 2025, and private investors now hold roughly 73 per cent of the market against about half a decade ago, with official bill flows swinging from a net purchase of US$134bn in the twelve months to June 2025 to a net sale of US$41.9bn in the twelve months to June 2026. Private holders are more price-sensitive than official ones. That is the channel through which fiscal credibility is actually transmitted.

The rating agencies price polarisation, and one of them now names shutdowns

All three major agencies rate the United States one notch below triple-A with a stable outlook. S&P, affirming on 27 June 2026, cited the distance between the parties, the elusiveness of bipartisan cooperation on deficits, and the political system's limited ability to reverse the deterioration in the fiscal position. It did not mention shutdowns. It did say it expects Congress to keep resolving the debt ceiling, which is a stated assumption and therefore a stated exposure.

Fitch, affirming the United States at AA+ with a stable outlook on 13 August 2026, warned that gridlock and government shutdowns may become both more likely and more protracted.

What cannot be attributed to the shutdowns

Nothing in the 2026 market record can be attributed to the shutdowns. A United States and Israeli air campaign against Iran began on 28 February 2026, two weeks before the Homeland Security lapse. The Strait of Hormuz closed on 4 March, Brent rose 55 per cent by 27 March, and the S&P 500 bottomed on 30 March. Month-end market reviews for February 2026 attribute the month to technology valuations, sector rotation and the Middle East, and do not mention the shutdowns at all. The 2026 lapses are not a usable natural experiment and are not treated as one here.

Fiscal 2026 closed with outlays of US$7.396 trillion against receipts of US$5.403 trillion, a deficit of US$1.992 trillion or about 6.1 per cent of GDP, worse than the US$1.9 trillion projected in February. Net interest reached a record of about US$1.05 trillion, roughly 19 per cent of federal revenue. Debt held by the public is 101 per cent of GDP, the first time since the Second World War, and is projected at 120 per cent by 2036 with net interest reaching US$2 trillion. That is the exposure, and no shutdown created it.

Monetary policy is not helping that line. The Federal Reserve held its target range at 3.75 to 4 per cent on 28 October with two members dissenting in favour of a rise, against September consumer price inflation of 3.7 per cent and producer prices up 6.3 per cent year on year. Futures priced roughly an 83 per cent chance of an increase in December. A tightening central bank makes that arithmetic harder.

11 December, and the confrontation behind it.

The live disagreements are the defence against non-defence topline and a possible defence reconciliation bill of nearly US$100bn; Homeland Security and border funding again; disaster relief, nutrition programmes, housing and public health; policy riders; whether to maintain the restriction on the Office of Management and Budget finalising its grants rule; and the administration's authority to transfer, withhold, rescind or condition appropriated funds. Enhanced Affordable Care Act premium assistance, the issue that caused the 43-day shutdown, remains unresolved.

The impoundment question is the one to watch. In September 2026 the administration proposed rescissions of about US$810m from eleven accounts expiring at the end of the fiscal year, on the theory that funds proposed for rescission within forty-five days of year end can simply be run out. The Government Accountability Office concluded on 29 September 2026 that this does not permit withholding past the end of the fiscal year. Three suits are pending and one judge has paused grant cancellations. There is no Supreme Court ruling on pocket rescissions.

A dispute about whether the executive must spend what Congress appropriates is a dispute about whether an appropriations agreement is worth anything. That makes agreement on the bills themselves harder, because the minority has less reason to trust a negotiated number.

The arithmetic and the calendar were already unfavourable. The election made them worse

The midterms of 3 November returned both chambers to the Democrats. CBS News projects a House of roughly 256 Democratic seats on more than forty flips, and a Senate of at least 54, with Mississippi and Alaska unresolved at the time of writing and the Mississippi margin inside two thousand votes.

None of that changes who must clear the cliff. The new Congress convenes on 3 January 2027, so the bills belong to the one elected in 2024. Congress returned on 9 November to roughly five legislative weeks. Three of the twelve bills have passed the House, none has passed the Senate, none is enacted, and the live dispute is a record defence request of about US$1.5 trillion that Senate Democrats reject. Surface transportation authorities, the National Flood Insurance Program and the pause on the grants rule all expire on the same date, with a further set of health programmes expiring on 31 December.

An outgoing majority has little left to trade, and an incoming one has no reason to settle in December what it can write in January. The orderly outcome is a short continuing resolution into the new Congress. The disorderly one is a lapse that neither side has much incentive to end quickly, which is the configuration that produced 43 days last time.

What the change of majority does to the budget process

From January the arithmetic is unusual. A Democratic House and Senate have to agree appropriations with a Republican president who has to sign them. The sixty-vote threshold in the Senate does not disappear, it changes sides: the incoming majority sits near 54 and still needs votes from the other party to close debate. Neither side is within reach of the two-thirds that overrides a veto. The configuration that produced three lapses in thirteen months has not been removed. It has been redistributed.

The consequential change is to reconciliation, and it is being largely overlooked. Reconciliation is the one route that bypasses the sixty-vote threshold, and for practical purposes it is now unavailable to anybody. The administration has lost the majorities that carried it. A reconciliation bill written by the new majority still needs a presidential signature. For roughly two years the instrument is dormant.

Reconciliation has been doing quiet work throughout this cycle. It raised the debt limit by US$5 trillion in July 2025, which is the reason there has been no debt-limit confrontation since. And it is the reason the 76-day Homeland Security lapse left no measurable footprint in the national accounts: roughly 85 per cent of the department kept working on reconciliation money. Both of those cushions were products of one party holding the House, the Senate and the presidency at the same time. Neither is available for the next lapse, or for the next limit.

The practical consequence runs against the intuition. The scope available to be protected outside the annual cycle has just narrowed, which means the next lapse is capable of doing more measurable damage than the 2026 lapses did, even if it is shorter than they were. The FY2028 cycle, which begins with a presidential budget request in February 2027, will be the first in four years drafted by an administration without a majority in either chamber.

Prediction-market quotes on the December deadline are not worth leaning on, and two calibration points show why. The contract on the 1 October 2026 deadline peaked at 35 per cent in late July and collapsed to about one cent when the stopgap was signed. During the Homeland Security lapse, prediction markets centred on a 22-day shutdown against a realised 76 days. These markets have not been good at pricing the duration of political brinkmanship.

The debt limit is the confrontation that would genuinely matter

The limit was raised, not suspended, by US$5 trillion to US$41.1 trillion in the reconciliation law signed on 4 July 2025. Federal debt passed US$40 trillion in August 2026. The limit is expected to be reached during 2027, with extraordinary measures running for some months after that. The Committee for a Responsible Federal Budget puts the point at which it must be raised again in mid to late 2027. There is no official exhaustion date and it should be held as a range rather than a date.

With reconciliation gone, what remains is bipartisan agreement or abolition of the filibuster. A shutdown closes offices. A debt limit breach touches debt service, and the bill market prices that difference at around 100 basis points when it becomes live. The two should not be assessed with the same calm, and from January the second is a negotiation between a president and a Congress controlled by the other party, with no procedural shortcut available to either side.

What follows from the evidence.

What the market bought on 4 November, and what it did not

The bond market’s response to the election was to buy Treasuries on the expectation that a divided Washington will pass less, and within a day that trade had been rewarded twice over. It is a coherent view about issuance. It is not a judgement that the fiscal position is improving, and the two are being run together.

Capping new borrowing and correcting an existing trajectory are different operations, and only the second requires a legislature capable of agreement. The interest bill, the debt stock, the tightening central bank and a limit that binds during 2027 all need legislation to improve and only gridlock to persist. The market has priced the gridlock as good news. On a twelve-month view it may well be. On the horizon the long end is actually discounting, it is the problem itself.

Practical implications

• Do not trade the shutdown. Positioning into funding deadlines has been a reliable way to pay spreads for no return.
• Price the governance constraint into long-dated United States dollar duration. Not because shutdowns damage Treasuries, but because the same gridlock caps the probability of the fiscal correction the long end is implicitly asking for, and because the buyer base has shifted decisively toward price-sensitive private holders.
• For transaction timing the transmission is administrative, not financial. Securities registration, CFIUS review, agency approvals, small business lending and the economic data calendar all stop in a lapse. Anyone planning a pricing window, a regulatory filing, a notifiable transaction or a federally dependent approval between 11 December and the seating of the new Congress on 3 January should build in the possibility of a two to four week administrative freeze. The lame-duck configuration raises that probability rather than lowering it.
• For anyone whose models consume United States official statistics, assume further holes and widen the error bars rather than the conclusions.
• Treat the 2027 debt limit as the date that matters. It now sits inside a divided government with no reconciliation route available, and it is the one deadline where the calm recommended elsewhere does not apply.

The claim this paper does not make

The temptation is to argue that the shutdown cycle is quietly destroying American creditworthiness. The evidence does not support that and it is not argued here. The evidence in sections 3 and 5 runs the other way throughout. A market that reprices on artificial intelligence valuations and ignores a funding cliff five weeks out is not expressing doubt about American creditworthiness.

The narrower claim holds, and it is harder to dismiss. The United States has a fiscal trajectory that requires legislative correction, and thirteen months of evidence that its legislature cannot complete routine tasks, let alone hard ones.

One rating agency has now put that in writing. It is a statement about probability, not about imminent distress, and it should be held that way.

What would prove this wrong.

Four developments would change the assessment set out above, in ascending order of significance.

Equally, the case against this paper is straightforward and should be stated. If the December cliff passes with a routine continuing resolution, if the February 2027 weight update lands without incident, and if the debt limit is raised in 2027 without a confrontation, then the shutdown cycle will have been exactly what markets have priced it as: noise. There is a cleaner test than that one. If a lapse on 11 December produces an identifiable premium in bills maturing in December or January, the central distinction in this paper between a lapse in appropriations and an exhaustion of borrowing authority is wrong, and the calm recommended here is wrong with it. The 2025 shutdown produced no such premium. That is the number to watch, and it is observable daily.

Where every claim in this paper comes from.

Who wrote this.

Paul Gardner Brook is a merchant banker with more than thirty years in international investment banking, including senior roles at Lehman Brothers and Nomura, and is Chief Executive Officer of PanEuro Special Situations Limited, London and Singapore. He writes on capital markets, energy, critical minerals, technology and public policy at paulgardnerbrook.com.

This paper may be cited as: Paul Gardner Brook, “Governing by Deadline. The United States Shutdown Cycle and What It Costs”, Paul Gardner Brook, 6 November 2026.