$40 Trillion and Counting

1The 10-Year Treasury

The U.S. 10-year Treasury (US10Y) is the most important bond and maybe the most important financial instrument in the entire financial market. A quick rundown:

It is the interest rate on a Treasury note maturing in ten years, and it is the benchmark for what it costs to borrow dollars for a long time. If you buy a house, your 30-year fixed mortgage is likely priced off it. Companies' borrowing costs are quoted as an amount that is above it. State and local (municipal) borrowing costs move with it. Federal student loan rates are set each May from the 10-year auction (let’s bring these rates down!). Stock analysts use it as the starting point when they value a company's future profits in today's dollars, which means it sits underneath the price of every stock you own whether an analyst mentions it or not (and for the most part, they don't).

On October 16, 2026 Treasury rates were 4.74% at two years, 4.89% at five, 5.09% at ten, and 5.43% at thirty. The 10-year closed at 5.01% on September 16, the first close above 5% since July 19, 2007, and its 2026 high was 5.31% on October 5, the highest close since 2002. The 30-year hit 5.67% on October 7, the highest 30-year close in more than two decades. For scale, the highest the 30-year got in all of 2020 was 2.38%. GULP!

For context, the 10-year closed at 3.97% on February 27, 2026, the last trading day before the United States went to war with Iran.

Rates since 2019

The 10-year's low was 0.52% on August 4, 2020. Its high, until this September, was 4.98% on October 19, 2023. The 2-year's low has come down every year since 2023, as the Fed cut. The 30-year's range has moved up every year since 2024, and its 2026 low of 4.64% is higher than its 2023 low, its 2024 low and its 2025 low. Short-term rates followed the Fed down, but long-term rates did not.

The fed funds rate was 3.88% on October 16, 2026: cuts from 5.33% between September 2024 and December 2025, and then one increase.

The Fed cut by roughly 1.7 percentage points, and the 30-year went up. That is why the Fed is not going to bring the interest bill down. On September 16 the Fed raised, and short-term rates went up while long-term rates stayed up: two days after the increase Treasury rates were 4.76% at two years, 5.01% at ten and 5.34% at thirty, and on October 16 they were 4.74%, 5.09% and 5.43%.

Not a credit problem

It is not pricing credit stress. The extra interest that BBB-rated companies (the lowest tier of investment-grade borrowers) pay over the Treasury, which widens when the market worries about defaults, was 1.60% on October 19, 2023 and 1.06% on October 16, 2026. Corporate borrowers are paying less over the government than they were, not more. The repricing is in the government's own rates, and it is about the cost of lending for a long time, not about fear of default.

What it means for your mortgage

The average 30-year fixed mortgage was 3.48% at the end of June 2016, touched 2.65% on January 7, 2021, reached 7.79% on October 26, 2023, and was 7.40% on October 15, 2026. Against a 10-year at 5.10% that day, that is a gap of about 2.3 percentage points between the mortgage rate and the 10-year, wider than before the pandemic. It reflects both the chance homeowners refinance and pay their loans off early and the fact that the largest buyer of mortgage-backed securities (bonds built from bundles of home loans) for a decade was the same central bank that is now letting its mortgage bonds run down.

The gap itself has a history. At the end of 2016 the 30-year mortgage was 4.32% against a 10-year of 2.45%, a gap of 1.87 points. At the end of 2019 it was 1.82. At the end of 2021, with the Fed still buying mortgage bonds, it was 1.59. At the October 2023 peak it was 2.93. It came in to about 1.9 by mid-September and was back to 2.3 by mid-October, nowhere near where it was when the Fed was buying.

Federal student loans first disbursed between July 1, 2026 and June 30, 2027 carry 6.52% for undergraduates, and an eye-popping 8.07% for graduate students and 9.07% for PLUS loans, and those are set by statute off the 10-year auction. Cities and states price their borrowing off the same rates, and so do insurers and corporate treasurers when they work out what money costs.

Every investment that pays a fixed amount over a long time has been repriced, in a year when the government borrowed $11.6 trillion, counting refinancing, and its central bank was no longer buying long-term debt.

Is this just an American problem? The 10-year Japanese government bond yielded −0.025% in December 2019 and 3.05% in September 2026, and Japan's overnight interest rate went from below zero to 1.04%. The other large government bond market that spent a decade with its central bank as the dominant buyer has repriced by roughly the same amount over the same years, from a lower base. This is talked about more in section nine.

Back on topic, the 10-year sets what almost everyone pays to borrow for the long term: mortgages, companies, cities and student loans. The Fed cut short-term rates by 1.7 percentage points, and the 10-year went up, closing above 5% in September for the first time since 2007. Nearly all of that rise is a higher return lenders want after inflation, not expected inflation, and about 40% of it is the extra pay for lending long: the New York Fed's estimate of it went from −0.68% at the end of 2021 to +0.76% on October 16. The Fed does not control that part, which is why lower Fed rates alone will not bring the government's borrowing costs down.

2People Don’t Believe in The Economy

Let’s start with what is going right.

On September 16 the Federal Reserve said, "Economic activity is expanding at a solid pace. While uncertainty remains elevated owing, in part, to geopolitical developments, domestic spending has been resilient. Productivity growth is strong, and capital investment is robust. Job gains have kept pace with the workforce, and the unemployment rate has changed little." Unemployment was 4.2% in September. Real GDP grew at annual rates of 4.0% and 3.9% in the second and third quarters of 2025, 0.2% in the fourth, and 2.5% and 2.2% in the first two quarters of 2026. The median projection of the Fed's own policymakers has the economy growing 2.3% this year and 2.4% next, with unemployment at 4.1% through 2029. Average hourly earnings have outpaced prices over the last decade (section ten).

Now the mood. I referenced a survey from the University of Michigan above. They have asked households how they feel about their finances and the economy since 1952 (my father’s birth year). The index at the points that probably matter most for this thought exercise:

The index averaged 86.3 from 1978 to 2019. In April 2020, with 14.8% of the labor force out of work, it was still 71.8. In June 2022, with inflation at 9.0%, it was 50.0. In May 2026, with unemployment at 4.3%, it was 44.8, the lowest reading since the survey began. It was 48.1 in September and a preliminary 46.3 in early October.

Trust in the government runs the same way. Pew has asked Americans since 1958 whether they trust the government in Washington to do what is right just about always or most of the time. In its September 2025 survey, 17% said yes: 2% just about always and 15% most of the time. It was 22% a year earlier and 77% in 1964, and it has not reached 30% in any survey since 2008. Among Democrats it was 9%, the lowest Pew has recorded; among Republicans, 26%.

Spending limits have also lapsed. The Fiscal Responsibility Act of 2023 set statutory caps on discretionary spending for 2024 and 2025, enforced by automatic cuts. For 2026 through 2029 it set limits that, in the words of the Congressional Budget Office (CBO), "would be enforced using the Congress's procedures for considering budgetary legislation." There is no automatic enforcement after 2025. The debt limit was raised by $5 trillion in July 2025, in the reconciliation act (a budget bill that can pass the Senate with a simple majority) introduced as the “One Big Beautiful Bill Act.”

Why doesn't a solid economy feel solid? Most of that answer must be chalked up to prices: they are up a truly staggering 30.0% since December 2019 (section six). In May, 57% of the University of Michigan's respondents said, unprompted, that high prices were eroding their finances. A smaller part is the cost of borrowing. The monthly payment on a median house has doubled since 2016, and the average interest rate on a credit card balance is 22% (section ten).

The economy is actually quite good right now. The stock market is booming (the S&P 500 closed at 7,659.44 on October 16, up 11.9% this year, with 28 record closes in 2026), the job market is less than stellar, but it is also steady and unemployment is low, wages have outpaced prices since before the pandemic, and some states' economies have grown by about half in ten years, twice the national pace. Trust in the government, however, is at a near all-time low. There is a belief of “is this it?” The “American Dream” and upward mobility both seem out of reach for so many, so the perception becomes the reality and consumer confidence has evaporated.

3The Interest Bill

There are two numbers for what the government pays to carry its debt. Many people don’t know that. After this, you will.

The fiscal year ended at the end of September and the gross interest on public debt was $1.36 trillion. Net interest (the number that shows up in the budget) was $1.05 trillion. This paper uses net interest, the budget's own number, unless it says otherwise. The gap is interest the government pays to its own “trust funds”, mostly Social Security and Medicare, on the Treasury securities those funds hold. Of the gross bill, $1.07 trillion went to debt held by the public and $291.9 billion went to the Government Account Series (the special securities the trust funds hold). Roughly 21% of gross interest is the government actually paying itself.

Net is what the debt costs taxpayers. Gross is what the debt costs in total. When someone tells you interest now exceeds defense spending, ask which one they used, because on the gross measure that happened earlier and by more.

Full fiscal years, gross interest expense: 2021 $562.4 billion, 2024 $1.13 trillion, 2025 $1.22 trillion, 2026 $1.36 trillion. It has more than doubled in five years.

What the bill is made of

Gross interest is not one payment. It's the sum of what Treasury pays on each kind of security it has outstanding, and the mix tells you where the rate cycle landed first. Three full fiscal years: the last one before the pandemic, the last year of near-zero rates, and the one that just ended.

A full year of bill interest cost $4.2 billion in 2021. In 2026 it cost $254.1 billion. That is about sixty times as much, because a bill is repriced every time it matures and bills mature in weeks. Notes went from $189.2 billion to $534.4 billion. Bonds, the 20- and 30-year bonds, went from $98.1 billion to $179.8 billion, a smaller move than bills or notes, because most of the bonds outstanding were issued years ago at whatever the rate was then and have not come due. TIPS (Treasury Inflation-Protected Securities, whose principal rises with prices) interest is two things, the coupon (the fixed rate the security pays) and the inflation adjustment to principal. The adjustment was $69.7 billion of this year's $92.3 billion, and the 2021 line had already jumped because inflation was moving by the summer of 2021.

The Government Account Series line is the trust funds. It rose from $173.4 billion to $291.9 billion for the same reason the public lines did: the average rate Treasury pays the trust funds went from 2.123% in September 2021 to 3.590% in September 2026. That interest is real income to Social Security and Medicare and a real charge to the general fund. It's money we owe ourselves, and we still have to pay it, which is why the gross and net numbers have to be kept apart.

History of Interest

Have we been here before? As a share of what the government spends, yes, and then it slowed. Net interest against total spending and against GDP, from the first year of the modern series in 1962:

Net interest was 6.4% of federal spending in 1962. It climbed through the 1980s as both the debt and the rates on it rose, reached 14.7% of spending in 1991 and 15.3% in 1995, and then fell for two decades, to 6.2% in 2012 and again in 2016, while the debt itself quadrupled. The low point as a share of GDP was 1.29% in 2016. Nine years later it was 3.20%, and in 2026 interest took 14.2% of everything the government spent.

The share was this high before and came down because rates fell for twenty years and the debt was refinanced at each lower rate as it matured; the debt itself shrank only in the four surplus years. Since 2022 that has run in reverse, slowly (section five).

Who gets paid

So who actually gets the money?

Of the $1.36 trillion gross bill in 2026, $291.9 billion went to federal trust funds and never left the Treasury. Of the $1.07 trillion paid on debt held by the public, the largest single holder is the Federal Reserve, with $4.57 trillion of Treasuries on its balance sheet for the week that ended October 14. What the Fed earns on them it is supposed to remit to the Treasury (by law, the Fed sends the Treasury whatever it earns beyond its operating costs, the dividends it pays its member banks and the surplus it is required to keep). For example, it remitted $109.0 billion in 2021 but remits nothing now, because it is carrying a $233.2 billion deferred asset from the years its interest costs exceeded its portfolio income. So, the interest paid to the Fed is, for now at least, interest the Treasury doesn't get back.

Foreign holders own 28.8% of the debt held by the public. If interest were paid in proportion to holdings, something on the order of $310 billion a year of the public-issue bill would be going abroad. The rest, the majority, is paid to domestic money market funds, banks, pension funds, insurers, mutual funds, and the households and hedge funds that sit in the Federal Reserve's catch-all household line. (The Fed's quarterly Financial Accounts of the United States record who owns what across the economy; the household line is whatever is left after every identifiable holder is subtracted.) That is income to Americans, which is why a large interest bill is not the same thing as a large tax. It is a transfer from taxpayers in general to bondholders in particular. If you keep cash in a money market fund, some of it is coming to you. Who the bondholders are is the subject of the next section.

Where it ranks

2026, by budget function:

Total spending $7.40 trillion, revenue $5.40 trillion, deficit $1.99 trillion. The 2025 column is from Treasury's statement; OMB's final tables (OMB is the White House Office of Management and Budget), used elsewhere in this paper, put 2025 net interest at $970.1 billion.

The seven lines above are the large ones. The whole statement, every function the Treasury reports:

Total spending rose 5.5%. Two lines fell by a quarter or more: natural resources and environment, down 35.3%, and community and regional development, down 26.1%. Three lines rose by more than the interest line: education and training, up 51.5%, mostly because the 2025 figure was pulled down by a $131 billion reduction in the estimated cost of student loans booked in September 2025; administration of justice, up 29.0%, which is where the 2025 act's immigration enforcement money lands; and veterans, up 10.4%. Net interest, at 8.4%, grew faster than both of the lines above it in size. On the revenue side, individual income taxes rose 7.1% and corporate income taxes fell 15.6%, from $452.1 billion to $381.4 billion, which is the 2025 act's expensing provisions taking effect. Customs duties fell 11.1%, because $144.6 billion of tariffs were refunded during the year after the Supreme Court's ruling. Total revenue rose 3.2%.

Net interest is the third largest line in the federal budget, behind Social Security and Medicare and ahead of defense, and it passed a trillion dollars in August, with a month of the fiscal year still left. For part of the year the monthly statements showed it in second place. When the first of a month falls on a weekend, Medicare's payment goes out on the last business day of the month before, so the monthly totals move around: the first of August fell on a Saturday this year, August's Medicare payment went out on July 31, and through August Medicare was up only 1.5% on the year while interest was up 8.9%. For the full year the timing washes out. Medicare finished at $1.07 trillion and net interest at $1.05 trillion, close enough that a year from now the order could flip.

Interest is also more than half of the deficit. Net interest of $1.05 trillion against a deficit of $1.99 trillion is 53%. In 2025 it was $970.1 billion against $1.77 trillion, or 55%. Take interest out and the government's deficit on everything else it does was $940.7 billion, under a trillion dollars, which CBO calls the primary deficit.

Interest against revenue

Net interest in 2025 was $970.1 billion against revenue of $5.24 trillion. That is 18.5 cents of every dollar the federal government collected, gone to interest before a single program got funded. In 2026 it was $1.05 trillion against revenue of $5.40 trillion: 19.5 cents.

Against GDP, you may have seen this one, only it’s typically stated wrong. Net interest was 3.199% of GDP in 2025. In 1991 it was 3.191%. 2025 essentially matched the 1991 peak, and it is the highest reading in a series that starts in 1962. It exceeded the 1991 figure by 0.8 basis points. OMB's own table shows both at 3.2% at one decimal place, so the two are the same number.

A country that spent 3.2% of everything it produced on interest in 1991 had $3.6 trillion of debt. We have $40 trillion now.

The last time the bill sat at this level it produced a third-party presidential candidate that attained the most success in almost a century: Ross Perot. I expand on him more in section eight.

What the Government Owes Itself

Not all the $40 trillion is owed to somebody else.

On August 18, 2026, the day the debt first closed above $40 trillion, $32.27 trillion of it was held by the public and $7.78 trillion was intragovernmental holdings, which is the “trust funds” mentioned above, mostly Social Security and Medicare, holding Treasury securities. That second number is the government's debt to itself. It is the debt behind the $291.9 billion of gross interest that stays inside the government.

People hear that and say we're borrowing from ourselves. The record actually runs the other way.

These are OMB's year-end figures. Treasury's daily record, which section seven uses, puts the September 30, 2025 total at $37.64 trillion; the two keep slightly different books.

We're actually borrowing from ourselves less every year, not more. The share peaked at 43.7% in 2007 and was 19.3% at the end of 2025, the lowest since 1986. Of the $28.4 trillion the gross debt grew between 2007 and 2025, $25.1 trillion of it, 88%, went to the public. The trust funds took 12%.

The reason is the one that shows up again in section seven, but I will spoil it quickly here too. Social Security stopped accumulating. While it ran an annual cash surplus it bought Treasuries with it, which is how the intragovernmental share climbed to nearly 44%. It is drawing down now, so it is not buying, and every marginal dollar the government borrows has to come from somebody who expects to be paid in cash.

That is a large part of why the interest bill grows faster than the debt does. The part of the debt where interest is a bookkeeping entry is shrinking, and the part where it is a cash payment to an outside creditor is taking all of the growth.

The funds are not exempt from the repricing either. The special-issue securities they hold are priced off market yields at issue and have been rolling over into the higher levels like everything else, which is why their average rate is now above what the marketable debt carries. That $291.9 billion is owed to Social Security and Medicare and paid by the general fund, and if you're on Social Security, it's the one part of the interest bill you should want to be large.

4Who Owns The Rest?

Who do we owe it to? The answer you've probably heard is China, and it isn't close to true and hasn't been for a pretty, pretty, pretty long time.

On August 31, 2026, foreign holders of Treasury securities held $9.33 trillion against $32.41 trillion of debt held by the public. That is 28.8%. Foreign governments (their central banks and sovereign funds, as opposed to private investors) held $3.76 trillion, or 11.6%.

In December 2014, foreign holders held $6.16 trillion against $13.02 trillion of debt held by the public. That was 47.3%, with foreign governments at 31.7%.

Foreign governments held $4.12 trillion in December 2014 and hold $3.76 trillion now. That is a decline of 8.9% in nominal dollars, before adjusting for eleven years of inflation. Over the same period, debt held by the public rose 149%.

On top of that, the foreign buyers of our debt since 2014 have been private investors, not foreign governments. The world's central banks, in aggregate, own fewer Treasury securities today than they did in 2014.

China specifically held $590.7 billion in August 2026, down from $699.7 billion a year earlier, a decline of 15.6% in twelve months. Japan held $1.07 trillion, down 9.8%.

Twenty-five years of foreign holdings

The China story was true once, and Treasury's historical table of major foreign holders shows when.

China's holdings rose from $60.3 billion at the end of 2000 to $1.27 trillion at the end of 2013, and every year since 2013 they have been lower or flat. At $590.7 billion they are less than half the peak. Japan passed China in 2019 and peaked at $1.30 trillion in 2021, and is $233 billion below that. Foreign holdings as a whole kept rising, from $1.0 trillion to $9.3 trillion, but after 2008 the debt they are set against rose faster, which is how the foreign share went from a peak of about half in 2008 to under 30% now. The largest foreign creditor the United States has ever had was cutting its position while the debt doubled, and the debt was absorbed anyway.

What the country table measures

Treasury prints a caution on that table. TIC data (Treasury International Capital, the Treasury's monthly count of foreign holdings) is collected from U.S.-based custodians and broker-dealers. It records where a security is held, not who owns it. Treasury's own note says that securities held in overseas custody accounts "may not be attributed to the actual owners" and that the data "may not provide a precise accounting of individual country ownership."

The reason Belgium shows $490.5 billion is that Euroclear, which is a securities depository (a firm that holds bonds for investors all over the world), is domiciled there. Similarly, the Cayman Islands show $463.0 billion, which is investment funds registered there. Luxembourg shows $448.2 billion for the same reasons. The United Kingdom shows $1.01 trillion, a large share of which is third-country money managed out of London.

Between August 2025 and August 2026, Japan and China, the two largest holders that are countries rather than custodial centers, shrank, and the financial centers, where holdings are custodial or fund-based, grew.

Foreign governments sold and private buyers bought: over the same twelve months, foreign governments' holdings fell $164.2 billion and foreign private holdings rose $246.8 billion. The composition of foreign ownership is changing faster than the total.

Japan's holdings reached $1.24 trillion in February 2026 and fell $171.0 billion in the six months to August. China's were $699.7 billion in August 2025, $699.2 billion in September, $684.4 billion in December, $618.0 billion in July and $590.7 billion in August 2026, a decline in almost every month. The United Kingdom went from $901.3 billion to $1.01 trillion over the same twelve months.

Who bought it

The Federal Reserve tracks who holds Treasuries, sector by sector, every quarter. Between the fourth quarter of 2021 and the second quarter of 2026, marketable Treasury debt outstanding rose from $22.56 trillion to $30.88 trillion, an increase of $8.32 trillion.

Treasury holdings by sector:

The Federal Reserve shed $1.97 trillion of Treasuries. The single largest replacement buyer, by a wide margin, was households and nonprofits, which went from $303 billion to $2.83 trillion. That line is not actually households.

That line item is a residual: everything the Fed cannot identify. The Fed measures what every other sector holds, subtracts it from the total, and assigns what is left to households. Most of what is left is likely domestic hedge funds, and the Fed also counts those separately, from the confidential reports large hedge funds file with the SEC:

Hedge fund Treasury holdings rose $1.72 trillion over those four years. The households line rose $2.53 trillion over a slightly longer window. Hedge funds account for roughly two-thirds of it.

In the second row you will see repo liabilities, money the funds borrowed with Treasuries as collateral, nearly tripled, from $1.23 trillion to $3.52 trillion. These are positions that are financed overnight. The lenders split almost evenly between domestic, at $1.80 trillion, and foreign, at $1.72 trillion.

The funds are doing something called the basis trade. A fund buys a Treasury and at the same time sells a futures contract on it (an agreement to deliver that Treasury later at a price fixed today). The two prices are almost the same, and the fund earns the small gap between them, a few hundredths of a percentage point. To make that worth doing, it trades in very large size, mostly with borrowed money, and it borrows in the repo market, overnight, putting up the Treasuries themselves as collateral. Because the fund has sold the future, it is not much affected by whether rates go up or down. What matters to it is whether it can keep borrowing every night at a reasonable rate. The trade keeps futures and Treasury prices tied together and makes the market easier to trade in. It also gets unwound quickly when overnight borrowing costs jump or lenders demand more collateral, and it is now one of the largest sources of demand for newly issued Treasury debt.

When did they buy? Each year-end:

Hedge fund Treasury holdings were no higher in 2022 than in 2019, and fell in 2020, the year the Fed was buying. They rose $676 billion in 2023, the first full year the Fed was letting its holdings shrink, $445 billion in 2024 and $454 billion in 2025, and the borrowing rose with them each year. Their Treasury holdings have more than doubled in three years, and the overnight borrowing that finances them has nearly tripled.

Over those four and a half years the market had to absorb $8.32 trillion of new Treasury debt, plus the $1.97 trillion by which the Fed's holdings shrank, about $10.3 trillion in all. Hedge funds, on the Fed's count through the end of 2025, took $1.72 trillion of it and added $2.29 trillion of overnight borrowing to do it. Along with the money market funds below, that is who replaced the Fed.

Money market funds are the other big buyer, adding $1.47 trillion. They can only own short-term debt: of their $3.28 trillion in Treasuries, $2.32 trillion is bills. A money fund cannot buy a 30-year bond, which means one of the biggest new buyers can only support the part of the curve (the range of Treasury rates, from 1-month bills to 30-year bonds) that reprices fastest.

The Job of the Federal Reserve

For those interested in financial markets, the Fed chairman is basically the second most important person in the world after the president. Jerome Powell, who was chairman from 2018 until May of this year and is still on the Board, has turned into a personal hero of mine. I have started to study the history of the position, and of the whole Board, but that is a whole other paper. The question here is: where do they come into play for bonds?

In two places. First, the Fed is still the largest single domestic holder. As of the week ended October 14, 2026, it held $4.57 trillion of Treasury securities at amortized cost (roughly, what it paid for them), up $374.3 billion from a year earlier. Runoff (letting bonds mature without replacing them) has ended. It is buying again, and specifically buying bills: its bill holdings are $566.0 billion, up $370.6 billion year over year. (The sector table above shows $4.08 trillion for the Fed at the end of June because it counts market value, while the Fed's weekly balance sheet counts amortized cost. They are two valuations of the same portfolio.)

The path of that portfolio: Treasury holdings on the Fed's weekly statement were $2.46 trillion at the end of 2016 and $2.33 trillion at the end of 2019, when the Fed was letting the post-2008 portfolio shrink. They were $2.52 trillion on March 11, 2020, $4.20 trillion on June 24, 2020, $5.65 trillion at the end of 2021, and $5.77 trillion at the peak on June 8, 2022. Then the runoff: $4.40 trillion in August 2024, $4.20 trillion in August 2025. And then the turn: $4.57 trillion in the week ended October 14, 2026, with nearly all of the year's increase in bills. In a decade the central bank's Treasury holdings went from $2.5 trillion to $5.8 trillion, back to $4.2 trillion, and are now growing again at the short end (the shortest-term debt).

OMB's own debt table separates the Federal Reserve's holdings from the rest of the public going back to 1940, and it puts the last six years in proportion. The Fed held $23.8 billion of the $241.9 billion held by the public at the end of fiscal 1946, 9.8%. It held $534.1 billion of $3.32 trillion at the end of fiscal 2001, 16.1%, and $779.6 billion of $5.04 trillion in 2007, 15.5%. After the pandemic purchases it held $5.43 trillion of $22.28 trillion at the end of fiscal 2021, 24.4%, the highest share in the table. At the end of fiscal 2025 it held $4.20 trillion of $30.17 trillion, 13.9%. The share has come down by ten and a half points in four years, and every one of those points was absorbed by a holder who is paid in cash.

Second, the Fed used to be a revenue source for the Treasury and currently is not. The Reserve Banks remitted $54.9 billion to the Treasury in 2019, $86.9 billion in 2020, and $109.0 billion in 2021. Then short rates went above what the portfolio earns, the Banks began operating at a loss, and remittances stopped. Under the Fed's accounting the losses are recorded as a deferred asset, which is the amount the Reserve Banks have to earn back before payments to the Treasury resume. That balance was $133.3 billion at the end of 2023, $216.0 billion at the end of 2024, and stands at $233.2 billion on the Fed's weekly balance sheet for the week ended October 14.

So the institution holding roughly a seventh of the publicly held debt is returning nothing to the Treasury, and will return nothing until it earns back $233 billion.

Some Addition

Most of the money is staying home; we own 71% of our debt. That removes the risk people worry about, that a foreign creditor stops lending and the funding disappears. No lender abroad is in a position to do that, and the China story is a decade out of date.

Three kinds of buyer took its place. The largest identifiable new buyer is a set of leveraged funds financing Treasuries overnight in repo. The second largest can only own the front end (short-term debt) by mandate. And the third, the central bank, is buying bills again while remitting nothing.

I wouldn't consider that a crisis. There is, of course, risk, but it is a different risk than the one you're hearing about, and it's concentrated at the short end of the curve, which is the part of the debt that reprices every year.

And the buyers who replaced the Fed are able to hedge. A basis-trade fund owns the bond and has sold the future; its exposure to the level of rates is small by design, and what it carries is financing risk it can exit overnight. A money fund holds ninety-day bills and is never far from repricing. The Treasury itself chooses the maturity of what it sells. If you're on the other side of the same rate cycle (section ten), you hold a 30-year mortgage or a credit card balance and are not able to do any of those things.

So is this a debt crisis? Nominally it does not look great. $40,000,000,000,000 is actually an unfathomable number. Just look at that. But when you actually break down how the debt has repositioned, we are in a much better place to deal with it now than in other time periods.

5How the Debt Reprices, and What People Get Wrong

A lot of times the way this works is understood backwards, and I had to relearn it myself, so I will try to explain:

Treasury reports the average interest rate on every category of its debt each month. The average rate on all of the debt next to bills (T-bills, the shortest kind of debt that the government issues):

Between September 2024 and September 2026, the rate on T-bills fell 1.28 percentage points. Over exactly the same two years the average rate on the entire amount of debt went up 0.207 points.

The Federal Reserve spent two years cutting rates. The front end of the curve came down substantially, but still the government's average cost of borrowing rose. Why?

Because most of the debt is not repricing at today's rates. It is repricing at today's rates as it matures, and what it is rolling off is the debt issued in 2020 and 2021 at rates that no longer exist anywhere on the curve. For example, the 10-year note that the Treasury sold in August 2020 carries an interest rate (in finance terms, a coupon) of 0.625%. With two reopenings later in 2020 there is $132.4 billion of it, it does not mature until August 15, 2030, and when it does it will be refinanced at whatever the 10-year costs that week. As of October 16, the date this paper is anchored to, that is 5.09%. That single rollover, four years from now, raises the average whether the Fed cuts or not, and there is $3.63 trillion of notes and bonds from those two years still outstanding at an average rate of only 1.243%.

Every kind of debt, 2019 to 2026

The same Treasury table gives the average rate by kind of security. At the end of the last fiscal year before the pandemic, the two fiscal years at the bottom of the rate drops, and the one that just ended:

Across the rows, the shorter the debt, the faster it reprices. Bills went from 0.053% to 3.870% because all of the bills turn over inside a year. Floating rate notes reset weekly and did the same. Notes went from 1.450% to 3.383%, because the average note is a few years long and most of the 2020 and 2021 ones have now matured. Bonds went from 3.124% to 3.466%, a smaller move, because the average bond outstanding was issued long before 2020 at a rate not far from today's, and the low-rate bonds of 2020 will be there until 2050.

The last row is the average of all of that. The debt as a whole has gone from 1.605% to 3.531%. That is still below what bills and floating rate notes pay, because notes and bonds, most of the debt, are still paying the rates they were sold at, when rates were lower. The bonds will keep dragging the average down for a long time, but the notes will not.

Two of those rows reprice without maturing at all. Treasury inflation-protected securities carry a low interest rate, 1.149% on average, but their principal is adjusted to the consumer price index (CPI) every month, and that adjustment is booked as interest expense as it accrues: $69.7 billion of it in 2026, against $22.7 billion of regular interest. When inflation ran at 9%, in 2022, that inflation adjustment rose by almost three quarters, from $82.8 billion in 2021 to $143.4 billion in 2022, the largest increase of any kind of Treasury debt. Floating rate notes, the $708 billion line, reset to the 13-week bill rate every week, which is why their average rate tracks bills closely, 4.141% against 3.870%, and why they are the first line to move in either direction; they moved within a week of the Fed's September increase. Between TIPS and floating rate notes, there is $2.9 trillion of debt that does not wait for a maturity date to reprice.

The government borrows for different lengths of time. T-bills are paid back within a year, so when rates change, their cost changes almost right away. Floating rate notes change their rate every week, and TIPS grow with inflation, so those move quickly too. Notes, which last two to ten years, and bonds, which last twenty or thirty, keep the rate they were sold at until they are paid back. That is why the average cost of the whole debt moves slowly: most of it is notes and bonds still paying older, lower rates, and it only catches up as they come due.

Now, what debt is outstanding?

The Treasury's marketable debt by type, at the end of 2019, the end of 2021, and September 2026, the latest month available on October 16, 2026, the date I decided to anchor this paper to:

Marketable debt has nearly doubled in under seven years and bills have tripled. Their share of the total marketable debt went from 14.5% to 22.4%, and the path was, dare I say, not smooth sailing: bills made up $2.66 trillion at the end of March 2020 and were up to $5.08 trillion at the end of June, an increase of $2.42 trillion in one quarter. This is not meant to be contemporary history, but what might come as a refresher for some is that we used T-bills to finance the initial pandemic laws. The share fell back to 16.7% by the end of 2021 as Treasury termed the borrowing out (moved it into longer-term notes and bonds), and has climbed again since 2023.

How much of it reprices

How much of it? The Government Accountability Office (GAO), Congress's nonpartisan audit agency, put numbers on it this past April. As of September 2025, 33% of federal debt outstanding was scheduled to mature within the next twelve months. Eleven years earlier, in September 2014, that figure was almost a quarter (24%).

In 2026, Treasury had to refinance $9.7 trillion of maturing securities. Add new borrowing to cover the deficit and total gross borrowing needs came to roughly $11.6 trillion.

That number is the whole shebang. The federal government is not slowly layering new expensive debt on top of a stable base of cheap debt. Roughly a third of the entire debt passes through the market and gets repriced at what are now prevailing rates every single year. So the interest bill reflects rate changes from the past year or two, whether or not anyone in our nation’s capital does anything.

The next twelve months

The Treasury's security-by-security table can give us a preview of what actually comes due next year, using October 1, 2026, through September 30, 2027, fiscal 2027, as our dates. Every bill outstanding matures inside that window. This is not a model I made; this is just adding up the securities the Treasury already lists.

That is $10.6 trillion, 33.4% of the marketable debt, that must be refinanced or sold again within a year, before a dollar of the 2027 deficit is ever financed. Two thirds of it are bills, which reprice to the Fed's rate within weeks of a move in either direction. The other third is longer-term debt issued at an average rate of about 3%, and every piece of it refinances above that unless the 2-year falls by more than a point and a half from the 4.74% it closed at on October 16.

What still has to be repriced

The Monthly Statement of the Public Debt lists every note and bond that is outstanding, with its interest rate and its maturity date, so the schedule can be read off directly. By fiscal year of maturity, for notes and bonds only (T-bills are excluded because all of them reprice within the year regardless).

Against a 2-year note at 4.74% and a 5-year at 4.89%, every one of those rows will refinance at a higher rate. The 2027 debt carries an average rate of 2.983%; refinance $2.95 trillion of it about 1.75 points higher and the annual bill rises by roughly $52 billion from that year's maturities alone, before any new borrowing. 2028 is larger and carries a higher average, and so on. What was borrowed in 2020 and 2021 is concentrated in 2027, 2028 and 2031, and the debt paying under 1% runs out in November 2030. That last $950 billion is the cheapest debt the Treasury has ever sold.

The better news is that most of the repricing has already happened. The average rate on the debt rose 0.900 points in the year to September 2023 and another 0.354 points in the year to September 2024. In the two years since, it has risen 0.207 points in total. It is 3.531% on the September statement, against 4.085% on the 13-week bill sold on October 13 and 4.74% on the 2-year note. The gap between what the existing debt costs and what the market charges had largely closed by the summer. The September increase reopened it at the front.

The interest bill from here grows mainly because the debt grows, not because the old debt will keep repricing upward. The exception is the schedule above. The steep part, the 1.25 points the entire debt repriced between September 2022 and September 2024, has already happened.

The weighted average maturity of marketable debt (how long, on average, the debt has left until it comes due) is 71 months, which the GAO notes is near a 26-year high. Did Treasury make a mistake and borrow too short-term? That is the intuitive explanation. I would say no, it did not. Average maturity is actually unusually long by historical standards. The average rate keeps climbing anyway, because even with that longer average, about a third of the debt, roughly $10 trillion, still comes due every year and has to be borrowed again at today's rates.

T-bills are a real part of the story, though. On the September Monthly Statement of the Public Debt, total marketable debt was $31.83 trillion, of which T-bills were $7.12 trillion, or 22.4%. At the end of fiscal 2014 bills were $1.41 trillion of $12.29 trillion, or 11.5%. TBAC (the Treasury's own borrowing advisory committee) has said a bill share averaging around 20% is the right trade between interest cost and rollover risk. It is now a bit above that.

A bill cost 3.870% on average in September and a 30-year bond cost 5.43% on October 16, so every dollar Treasury funds with bills instead of bonds saves about a cent and a half a year in interest, for as long as short rates stay below long rates. What it buys with that saving is exposure: the $7.1 trillion of bills reprices completely every few months, so if short rates go back up, the bill does too, immediately, with no lag. Treasury has chosen to borrow cheaper and shorter.

My read is that the Treasury has been leaning on short-term borrowing because the alternative was selling even more long-term debt (debt that matures in ten years or more), and buyers have not had much appetite for more of it.

6The COVID Spending and the Rate Cycle

Debt doesn't cause inflation; inflation causes debt.

This section is probably the most important one in this paper for where we are today. The debt more than tripled between 1998 and 2016 without producing high inflation or a larger interest bill. Our ally in Asia, Japan, has run more than 150% of GDP for two decades and spent most of that period fighting deflation. A large amount of government debt is not, believe it or not, by itself, inflationary. What can be inflationary is deficit-financed spending arriving at the wrong moment, and the spending of 2020 and 2021 did contribute to the inflation that followed. Now to be clear, people who defend the government spending (myself included, for the most part) will say that the inflation that followed, while not the best thing on earth, was much better than the alternative COVID represented, which was a potential depression-era pullback. I do believe two things can be true: 1) We should do whatever is necessary to make sure we never live through depression-era economic conditions again in this country; and 2) we probably went a little overboard with what we considered necessary. That being said, and this is a policy opinion more than anything else, isn’t the preferred approach to do too much rather than not enough?

Back on topic: Inflation actually can shrink debt by inflating the economy underneath it: total federal debt was 132.7% of GDP in the second quarter of 2020 and 119.4% at the end of 2021, and nobody was paying debt down. It fell because GDP, measured in today's dollars, was rising along with prices. So inflation doesn't cause debt the way a deficit does. It causes debt through what it forces after the fact. It forces the central bank to raise rates; as outlined above, the rates reprice a third of the debt every year; the repricing turns into interest; and the interest is borrowed. Net interest was 1.5% of GDP in 2021 and 3.2% in 2025, and it was 55% of the 2025 deficit. Inflation also raises the spending that is indexed to it directly: the Social Security cost-of-living adjustment was 5.9% for 2022 and 8.7% for 2023, the two largest adjustments since 1982, and the inflation adjustment on the Treasury's own inflation-protected securities was $69.7 billion in 2026. The ratio fell once, but the higher interest recurs every year.

The rate cycle repriced the entire existing debt, not just the new borrowing. That repricing is how COVID still shows up in a 2026 interest bill, and it is, obviously, a much bigger effect than the borrowing itself.

Federal spending in 2019 was $4.45 trillion. In 2020 it was $6.52 trillion, an increase of 46.5% in a single year. 2021 came in higher still at $6.82 trillion. The resulting deficits, $3.10 trillion and $2.77 trillion, are the two largest in the history of the country. Six laws did it, and they are laid out one by one in section eight. Their combined ten-year cost on CBO's tally is $5.13 trillion.

How it was financed

The money had to be borrowed in months, and it was borrowed short-term. Bills outstanding rose $2.42 trillion between the end of March and the end of June 2020 (section five), against $2.74 trillion of new marketable debt over the same quarter, so nearly nine tenths of the new borrowing in the first pandemic quarter was in T-bills, debt that matures inside a year.

The first buyer was the Federal Reserve. Its holdings of Treasury securities were $2.52 trillion on March 11, 2020 and $4.20 trillion on June 24, up $1.67 trillion in fifteen weeks. Set against the $2.74 trillion of marketable debt Treasury added over the quarter, the central bank absorbed the equivalent of 61% of it. The Fed kept buying at a slower pace for two more years. Its Treasury holdings reached $5.65 trillion at the end of 2021 and peaked at $5.77 trillion on June 8, 2022, and its total balance sheet went from $4.31 trillion in March 2020 to $8.97 trillion in April 2022.

M2, the broad measure of money in the economy (cash, checking and savings deposits, and money market funds), was $15.35 trillion in December 2019 and $21.79 trillion in March 2022, up 42% in twenty-seven months. It was $23.34 trillion in August 2026.

Inflation followed. CPI was running 1.32% year over year in December 2020. By June 2021 it was 5.30%, by December 2021 it was 7.17%, and it finally peaked at almost 9% (8.98%) in June 2022. Prices are up 30.0% cumulatively since December 2019. The rate of increase was 3.4% in August and 3.7% in September 2026, well off the peak but still above where it ran in any year of the 2010s.

Economists argue about the split between fiscal support, supply chains, energy, and pandemic labor-force effects, and that argument is not settled. What is not disputed is that the fiscal expansion was a contributor. You can hold that view without holding the view that it was the only cause.

Then the Fed

The federal funds rate, month by month at the turning points:

0% to 5.33% in seventeen months, in what was the fastest tightening cycle since the early 1980s. The rate held there for a year, came down in slow, small cuts through December 2025, and on September 16, 2026 went a quarter point (25 basis points) back up.

A point that gets talked about on CNBC once in a blue moon, if that, is that the Fed's rate-setting committee isn't setting rates around the national debt. It is setting short-term rates, and the entire Treasury market reprices around them. When the Fed went from 0% to 5%, every T-bill and note the Treasury sold, including the ones replacing debt that came due, started selling at the new, higher rates.

So, the average rate on all federal debt was 1.565% in December 2021. It is 3.531% now. Apply that to $40 trillion, and that difference of less than 2 percentage points is about $790 billion a year of interest, and it is already in this year's spending.

So what did the COVID spending actually cost? The direct answer is about $5 trillion in additional borrowing that wouldn’t have existed without the virus. The much larger answer is that it helped end the conditions under which $19 trillion could be carried for the same annual cost as $5.5 trillion (section seven). The borrowing was a one-time charge (or a few charges, across several bills), but the repricing is permanent until rates fall, and rates falling is not the same thing as the average rate falling, as section five showed. Are you staying with me?

The difference in numbers: The six laws cost $5.13 trillion on CBO's tally. At the September 2021 average rate on the debt, 1.605%, carrying that would cost $82 billion a year. At the September 2026 average, 3.531%, it costs $181 billion a year. Only the rate changed between the two dates, and it applies to the other $35 trillion as well.

7Every Budget Since the Last Surplus

The last federal surplus was 2001. Every budget since has been a deficit, where we spend more than we collect in taxes. The year that ended September 30, 2026, made it twenty-five straight years. The first twenty-four add up to $24.4 trillion on OMB's historical tables, and Treasury's year-end statement puts 2026 at $1.99 trillion more, $26.3 trillion in all. The historical tables in this section run through 2025, the last year in OMB's tables; the projections at the end are CBO's.

Year-to-Year

In the early 2000s we were running deficits in the “low” hundreds of billions. The financial crisis put them above a trillion for four straight years and then they came back down, though never to where they started. 2020 and 2021 produced the two largest deficits in the history of the country, $3.10 trillion and $2.77 trillion. And then they didn't go back down to the pre-2020 level. The four years since have averaged $1.66 trillion, which is two and a half times the 2013–2019 average and more than five times the 2002–2008 average.

All twenty-four:

Tax revenue fell in six of those years: 2002 and 2003, after the 2001 tax cut and the dot-com recession; 2008 and 2009, due to the global financial crisis; 2020 due to COVID; and 2023, when they fell $456 billion from a 2022 peak that had been inflated by capital gains people cashed in during the year prior, in 2021. Spending fell in five: 2010, 2012, 2013, 2022 and 2023, the last two being the pandemic programs expiring. The smallest deficit of the period was a now modest $157.8 billion in 2002 and the largest was “you gotta read this a second time” $3.10 trillion in 2020. The median year of the twenty-four ran a deficit of about $730 billion.

Did revenue keep up? Between 2001 and 2025, federal revenue rose 163%, from $1.99 trillion to $5.24 trillion. Spending rose 276%, from $1.86 trillion to $7.01 trillion. Revenue grew a lot, but not like that.

Growth in the broader economy

Government dollars spent over twenty-four years do mislead more than a little, because the economy nearly tripled too. Spending and revenue as a share of GDP, at the points that matter:

Revenue has run between 16% and a little over 19% of GDP in every year in that table, and fell to 14.6% in 2009 and 2010, the recession years that are not in it. The fifty-year averages, 1976 through 2025, are 17.3% of GDP for revenue and 21.1% for spending. Revenue in 2025 was on the average. Spending was two points above it. Net interest alone was 3.2% of GDP in 2025, against 1.3% in 2016.

So, if you’re like me, you may be asking: where did the structural deficit (the part that persists in good years as well as bad) come from? Revenue as a share of the economy is where it has been for sixty years. Spending on programs, as a share of the economy, is roughly where it was in the 2010s. What's new, though, is the interest.

Where the revenue comes from

The mix of what the government collects has barely moved.

Individual income taxes were 49.9% of revenue in 2001 and 50.7% in 2025. Payroll taxes were 34.9% and are 33.4%. Corporate taxes were 7.6% and are 8.6%. Four tax laws in between, in 2001, 2003, 2017 and 2025, changed rates and brackets and the treatment of business income, and the composition of what the government collects came out almost exactly where it started. What changed is the level against the economy in the two years that stand out: 2022, when revenue hit 19.2% of GDP on capital gains cashed in during the 2021 asset boom, and 2023, when they fell $456 billion. The swings in the revenue line come from the stock market and the trend from the economy; neither one is a policy choice.

How the Clinton budgets are still relevant

The 1998 to 2001 surpluses were unified figures, meaning they counted Social Security. On the on-budget measure, which excludes it, only 1999 and 2000 were actually in surplus. The difference was Social Security running a large annual cash surplus and lending it to the rest of the government: $160.7 billion in 2001 alone.

Social Security's balance went negative in 2021 at −$51.5 billion and reached −$147.9 billion in 2025. A line that added $160.7 billion to the federal bottom line in 2001 took $147.9 billion out of it in 2025. That is a $309 billion swing in a single year's budget. The aging of the population will keep widening it.

The split for the last ten years: Off-budget is Social Security and the Postal Service; on-budget is everything else.

Until 2020 Social Security was still adding to the federal bottom line, by $35.5 billion in 2016 and $9.8 billion in 2020. It has subtracted from it in every year since, and the subtraction has grown from $15.2 billion in 2022 to $147.9 billion in 2025. The on-budget deficit, everything except Social Security and the Postal Service, was $1.63 trillion in 2025, lower than in 2024. The total was not lower by as much, because the off-budget line went from −$60.4 billion to −$147.9 billion in one year. That is the aging of the population showing up in the budget, and it will not reverse.

Can we go back to what we did in the nineties if we wanted to? The answer is NOPE. The arrangement that produced those numbers has reversed.

Interest as a line item

Net interest in 1998 was $241.1 billion. In 2016, eighteen years later, it was an almost unbelievable $240.0 billion.

Over those same eighteen years the gross federal debt went from $5.53 trillion to, are you ready for it… $19.57 trillion. The government more than tripled what it owed but the annual cost of carrying it did not move. It actually did more than just stay flat; it fell! Net interest fell from $241.1 billion in 1998 to $153.1 billion in 2003, a decline of 37%, its lowest level since the 1980s, and it did that while the debt grew from $5.53 trillion to $6.80 trillion.

Then the nine years after. 2016 to 2025, gross (which has two meanings here) debt went from $19.57 trillion to $37.64 trillion, which is almost, but not quite, a doubling. Net interest went from $240.0 billion to $970.1 billion, which is a quadrupling. It passed national defense in 2024, $879.9 billion to $873.5 billion, and the gap widened in 2025.

In the first stretch the debt grew 3.5 times and the interest bill was flat. In the second the debt grew 1.9 times and the interest bill grew 4 times. So what's driving the cost of the debt? It very clearly is not simply the size of it.

If you’ve read up to this point you should know, it is the rate. That is why the way the debt reprices (that we talked about in section five) matters more than any single year's borrowing or budget: a rate change reaches the interest bill slowly and automatically, and is harder to reverse than most of the political talking-head commentary assumes, mine included.

What has the borrowing bought?

Twenty-four years of deficits definitely have bought something. What? Let’s examine. Federal spending was $1.86 trillion in 2001 and $7.01 trillion in 2025. Where the $5.15 trillion a year of growth went, by the broad categories OMB uses:

Programs that pay or care for people directly (Social Security, health care, Medicare, income support and veterans' benefits) took 68%. Interest took 15%. Defense took 12%. Physical resources, which is every dollar the federal government spends on transportation, energy, natural resources and the environment, community development, and commerce and housing credit, took 4%. Education, training and employment took an—I have said this probably too many times—almost unbelievable 0.3%.

The 4% or four cents per dollar are the part of the budget people picture when they picture the government building things or working well. Transportation spending has gone from $54.4 billion to $145.3 billion. Community and regional development went from $11.8 billion to $82.4 billion, most of the increase after 2019. Natural resources and the environment went from $25.5 billion to $89.8 billion. Energy went from nothing to $20.8 billion. Housing and business credit programs went from $5.7 billion to negative $28.1 billion; that line now takes in more than it spends. Education, training and employment is the other line that stands out, in the other direction: $57.1 billion in 2001, $109.7 billion in 2016, $136.7 billion in 2019 and $298.4 billion in 2021, with the pandemic education money. It then swung to $677.3 billion in 2022 and to negative $2.2 billion in 2023, as student loan cancellation was booked and then reversed. In 2025, after the pandemic money ran out and new student loan changes were booked, it was $72.0 billion. On the government's own accounts it spent less on education, training and employment in 2025 than in 2016.

Now, what happened to the people the money was spent on. These are from the Census Bureau, the Bureau of Labor Statistics and the Fed's data service: at the start of the period, the middle, the year before the pandemic, and the latest year each series has.

Census's figure for 2025, published in September, is $87,460, a new high (in 2024 it had still been just below 2019, at $85,210): $15,110 above 2001, a gain of 20.9% over twenty-four years, and $2,140 above 2019, a gain of 2.5% in six. Life expectancy is up two years over the period and one tenth of a year since 2019. The homeownership rate is lower than it was in 2001. The poverty rate is about even with where it was. Real median weekly earnings are up 10% over twenty-three years. One more data point, one I wrote about in undergrad for one of the few papers I enjoyed writing. It comes from economist Raj Chetty (I think I used the same source back in the day) and colleagues' research using tax records: about 90% of Americans born in 1940 went on to earn more than their parents had, and about 50% of those born in the 1980s did.

$24.4 trillion of borrowing through 2025, $5.1 trillion a year of added spending, and at the median the measurable result is a household that is somewhat better off than in 2001 and only slightly better off than in 2019. Social Security, Medicare and Medicaid did what they're designed to do, which is keep people maintained. Only four cents of each new dollar built anything, and the data that would show a country’s next generation becoming more prosperous (income, ownership, mobility) do not show it.

And the years since 2020? The $12.5 trillion of deficits run in 2020 through 2025 sits next to a real median income that rose 2.5% between 2019 and 2025. The one thing that moved sharply in that window was child poverty, which fell by half in 2021 and went back the next year (section ten).

The last four years

The interest arrived in the last four fiscal years.

The deficit has been about $1.7 trillion for three years, and net interest doubled between 2022 and 2025. In 2022, a third of the deficit was interest. In 2025 more than half of it was. Strip interest out and the primary deficit, the gap between what the government collects and what it spends on everything except its creditors, was $898 billion in 2022 and $805 billion in 2025. The primary deficit shrank and the total did not, because of the net interest line.

Fiscal 2026, on Treasury's year-end statement, kept the pattern, with one difference. The deficit was $1.99 trillion and net interest $1.05 trillion, 53% of it. The primary deficit widened, to $940.7 billion, an increase of $135.7 billion. Tariff refunds after the Supreme Court's ruling rose by $137.1 billion over the year before, so without them the primary deficit would have been about where it was in 2025.

A deficit before interest that is steady to slightly improving, and an interest bill compounding on top of it: CBO projects the same pattern for the next ten years.

CBO's ten-year projection

CBO published its current-law baseline on February 11, 2026. "Current-law baseline" assumes Congress stops passing new things, which never happens, so treat it as where current law leads rather than a forecast.

The primary deficit (everything the government does other than pay interest) improves. It goes from 2.6% of GDP to 2.1%. The total deficit gets worse though (by almost a percent), from 5.8% to 6.7%, because net interest rises from 3.3% to 4.6%.

The same baseline, year by year, from CBO’s Table 1-1. The "deficit" is defined as the primary deficit plus net interest; the last column is debt held by the public at the end of the year.

Net interest over the next ten years, 2027 to 2036, is projected to total $16.2 trillion. The primary deficit over the same ten years sums to $8.3 trillion. We're projected to borrow twice as much to pay interest on past borrowing as we borrow to run the government. In 2035 interest is projected to be 73% of the deficit. Debt held by the public goes from $32.1 trillion at the end of 2026 to $56.2 trillion in 2036, and $16.2 trillion of that $24.1 trillion increase is due to interest. These numbers are very likely to change, but this is the read if they don’t.

The primary deficit line is $805 billion in 2025 and $971 billion in 2036, and in between it goes as low as $695 billion. In an economy that grows from $30 trillion to something over $40 trillion, that is a primary deficit that shrinks as a share of the economy. The government's position before interest is projected to improve over the decade. It does not matter, because the interest line grows in every single year of the table, by more than the rest of the budget improves.

Debt held by the public crosses the 1946 record of 106.1% of GDP in 2030, at 107.7%. It reaches 120.2% in 2036 and 175.1% by 2056. CBO assumes the 10-year Treasury averages about 4.3% over the window, running 4.2% in 2026 and 4.4% by 2036. The 10-year closed October 16 at 5.09% and the 30-year at 5.43%.

The baseline's interest line depends mostly on one assumption, the rate, and a small miss costs a lot.

Debt held by the public is $32 trillion and heading to $56 trillion. On today's debt, every tenth of a percentage point on the average rate is $32 billion a year of interest once the whole debt has repriced; on the 2036 debt it is $56 billion. CBO's baseline has the 10-year at 4.2% in 2026. It closed September at 5.29% and closed October 16 at 5.09%, about nine tenths of a point above the assumption. What if it stays there? If that difference persists and passes through to all of the debt, the interest line would run about $285 billion a year above the baseline on today's debt and about $500 billion a year above it on the 2036 debt, before counting the interest on the extra borrowing itself. That compounding is why CBO's own long-term projection curves upward rather than running straight.

The other direction is the same size. If the 10-year settled at 4.09% instead of 5.09% and all of the debt repriced to it, the saving would be on the order of $320 billion a year on today's debt. Section five explains why that takes years to arrive: the debt maturing in the next five fiscal years carries an average rate of about 3.2%, so even a 10-year at 4.09% refinances most of it higher, not lower. It works in reverse, but slowly, and only from a level the 10-year has not seen since March.

8Differences in Handling

So, questions may arise: which, if any party is to blame for this? The answer is both. Every version of this analysis depends on framing, and most framing is at best self-admittedly biased and at worst dishonestly biased. I say that, and then immediately say this: I will try to give an honest, unbiased view in this section. The federal fiscal year runs from October 1 to September 30. I believe it is most sensible to assign each fiscal year to whoever controlled the White House, House, and Senate on the day that fiscal year began. For example, 2002 (which began on October 1, 2001) is assigned to President Bush and the 107th Congress. This is defensible, but it is also imperfect, because major fiscal legislation routinely passes in the middle of a fiscal year and sometimes reshapes one that is already underway. Deficits are stated as a share of GDP, because a $400 billion deficit against a $10 trillion economy and a $400 billion deficit against a $30 trillion economy are not remotely in the same ballpark. So while it is not perfect, it is also likely the least biased way to lay this out.

By Presidential Administration

Two of these are distorted by a single event each. Bush's average includes 2009, the year of the global financial crisis, when the deficit was 9.8% of GDP, and Trump's first term includes the 2020 (−14.5%) and 2021 (−12.0%) deficits of the pandemic. Strip out the crisis years and Bush averages −2.39% and Trump −4.20%, which changes the framing a little. Even so, you would notice a trend: Clinton +1.40%, Bush −2.39%, Obama −4.95%, Trump −4.20%, Biden −5.95%. Every administration since the surplus has run wider deficits than the pre-crisis baseline it inherited, and the two parties have alternated the White House the whole time: eight years, eight, four, four, and now four more. 2026, the first full year of Trump's second term, ran a deficit of $1.99 trillion. It joins these tables once its share of GDP can be stated: the third-quarter GDP figure comes out October 29 and OMB's tables in February.

Congressional Control

It is easy to look at that table and draw a conclusion about what the most fiscally responsible mix of president and Congress is—a president of one party facing an opposition majority in Congress, but this too can be a bit dubious. For one, that row has all four of the Clinton surplus years in it, which were produced by a specific and unrepeatable combination: a 1990s revenue boom, rare lack of major international conflict, and a Social Security cushion that no longer exists and never will again, short of sweeping reform of the program.

The worst-looking row is unified Democratic control, but that is also distorted by events. All four of its years came right after a crisis, 2010 and 2011 after the financial crisis and 2022 and 2023 after the pandemic, when spending is typically elevated to get the economy growing again while revenue is still depressed. To be fair, however, in 2022 and 2023 the Biden administration and a Democratic Congress also passed substantial spending bills, some of which had little to do with recovery.

The split Congress row contains 2020 and 2021, the two largest deficits ever recorded in dollars, which occurred under a Republican president (Trump), a Democratic House, and a Republican Senate. Take out those two years and the row's average deficit falls from 5.95% of GDP to 4.12%.

No combination of parties has produced a surplus in twenty-five years, and the crisis years dominate every group they fall in.

Tools Congress Uses

Congress has three tools it uses when it wants to move a large amount of money: reconciliation, emergency designation, and debt limit raises.

The first, and the most common, is budget reconciliation. A reconciliation bill cannot be filibustered, so it clears the Senate with a simple majority: 51 votes, or 50 with the Vice President breaking the tie. Nine of the fourteen laws in the table below were reconciliation bills. Seven of the nine were the 1993 deficit reduction act, the 2001 and 2003 tax cuts, the 2017 tax act, the 2021 relief act, the 2022 climate and health act and the 2025 act. In five of those seven the other party supplied exactly zero Senate votes. In 2001 it supplied twelve, and in 2003 two. The other two, 1990 and 1997, were negotiated between the parties.

The second is the emergency designation, which lets a bill spend outside whatever caps and pay-as-you-go rules are in force at the time. The five pandemic laws of 2020 went through it (the sixth, the 2021 relief act, was reconciliation), and the largest of the five, the CARES Act, passed the Senate 96 to 0.

The third is the debt limit, which authorizes no spending but instead, in a sense, is the negotiation starter. By the Treasury's own count, Congress has raised, extended or redefined the limit 78 times since 1960, 49 of them under Republican presidents and 29 under Democratic ones. Since 2013 it has mostly stopped setting a number and started suspending the limit for a period of time instead: February 2013, October 2013, February 2014, November 2015, September 2017, February 2018, August 2019 and June 2023. Eight suspensions in ten years. It set a number twice in 2021 and again in 2025, when the 2025 act raised the limit by $5 trillion, to $41.1 trillion.

How the three tools were used, in order, from the last time the rules worked to now:

1985 to 2002: targets and caps

The first attempt was a target. The 1985 Gramm-Rudman-Hollings act wrote a maximum deficit into law for each year, falling to zero by 1991, with an automatic across-the-board cut (a sequester) if the projection came in over the line. The Supreme Court struck down the enforcement mechanism in 1986 in Bowsher v. Synar, because it put the trigger in the hands of the Comptroller General, an officer of Congress. The 1987 reaffirmation act moved the trigger to OMB and pushed the zero out to 1993. The targets against the deficits actually recorded from 1986 to 1993:

The targets were unironically missed in every year they were "enforced," and in the year the first set called for zero the deficit was $269 billion (have you ever had a six-year loan and ended up owing $269 billion at the end?). In the fall of 1990, with the targets still being missed, President Bush and congressional leaders agreed at Andrews Air Force Base to a five-year, $500 billion deficit reduction plan, and the act that carried it replaced the targets with something that could be enforced: statutory caps on discretionary appropriations (the spending Congress votes on each year), and a pay-as-you-go rule that required any new entitlement spending (benefits paid by formula, such as Medicare) or tax cut to be offset. The 1990 act passed the House 228 to 200 with 47 Republicans and 181 Democrats, and the Senate 54 to 45 with 19 Republicans and 35 Democrats. The plan was highly unpopular with Bush's own party.

The 1993 act extended the caps and the pay-as-you-go rule (PAYGO) through 1998. It passed the House 218 to 216 and the Senate 50 to 50 with Vice President Al Gore, and not one Republican in either chamber voted for it. The 1997 Balanced Budget Act extended the caps through 2002 and, on CBO's estimate at the time, produced $127 billion of net deficit reduction over 1998 to 2002: $160 billion of gross savings, of which $112 billion came from slowing the growth of Medicare, $21 billion from auctioning spectrum, $7 billion from Medicaid, $5 billion from tobacco excise taxes and $15 billion from other items, offset by $20 billion for children's health insurance and $13 billion to soften the 1996 welfare law. It passed the House 346 to 85 and the Senate 85 to 15.

The four surplus years, 1998 through 2001, all fell inside the window when the caps and PAYGO were in force. The surpluses were $69.3 billion, $125.6 billion, $236.2 billion and $128.2 billion. The caps expired after 2002, and the PAYGO rule was formally terminated in December 2002, and there, of course, has not been a surplus since.

2001 to 2010: the rules have lapsed

The first big fiscal law of the new millennium was a tax cut, passed by reconciliation (what better way to ring in still being here after Y2K). The 2001 act reduced revenue by $1.35 trillion over 2001 to 2011 on the estimate of the Joint Committee on Taxation (JCT), Congress's tax scorekeeper: $874.9 billion from cutting income tax rates, $138.0 billion from phasing out the estate tax, $63.3 billion from marriage-penalty relief, $29.4 billion from education provisions, $25.1 billion from changes to individual retirement accounts, and the rest from the child credit and pension rules. Nearly every provision carried a sunset of December 31, 2010, which put the same cuts back in front of Congress in 2010 and again at the start of 2013. It passed the House 240 to 154 and the Senate 58 to 33, with 28 Democrats in the House and 12 in the Senate voting with Republicans.

Another tax cut in 2003 followed. This—also passed by reconciliation—accelerated the 2001 rate cuts ($171.4 billion), cut taxes on dividends and capital gains ($148.1 billion), added business incentives ($10.1 billion) and sent $20 billion to the states as fiscal relief, for $349.7 billion over 2003 to 2013. The Senate was 50 to 50 and Vice President Dick Cheney broke the tie. Democratic support was much thinner this time around: seven House Democrats and two Senate Democrats.

The same Congress added a prescription drug benefit to Medicare through regular order (the normal process, where 60 senators are needed to end debate). CBO's estimate of the 2003 act was a $394 billion increase in deficits over 2004 to 2013, net of offsets. The largest pieces were $507 billion of payments to drug plans and Medicare Advantage plans for the basic benefit over 2006 to 2013 and $191 billion of subsidies for low-income enrollees, with $88 billion recaptured from the states' Medicaid drug savings among the offsets. The House vote was 220 to 215.

Then came, of course, the global financial crisis. The 2009 recovery act was scored (its budget effect estimated) at $787 billion when it passed and $831 billion by CBO's February 2012 revision, spread over 2009 to 2019 with close to half of it in 2010. CBO grouped what it bought into purchases of goods and services by the federal government, transfers to state and local governments for infrastructure and for other purposes, transfers to individuals, and tax cuts. By CBO's estimate the act raised employment in the last quarter of 2011 by between 0.3 million and 2.0 million people. It passed the House 246 to 183 with no Republican votes in support and the Senate 60 to 38 with three brave souls crossing party lines.

So, between 2001 and 2010 we had two enormous tax cuts, a massive drug benefit and a $787 billion stimulus. That isn’t even mentioning the wars in the Middle East. Iraq, Syria and Afghanistan ended up costing the Pentagon $1.68 trillion through 2023, and the Iran war that began in February has cost about $38 billion thus far (through July), with between $2 billion and $3 billion a month in continued costs while it continues. This has also contributed to rising fuel costs, which in turn, has had a trickle-down effect on other industries, and in an attempt to slow inflation has been impactful in the Fed’s decision to raise interest rates, thus causing more of our money to go to interest than before. And none of these have been offset.

2011 to 2019: caps, with exceptions

The caps came back as the price of a debt limit. On May 16, 2011 Treasury Secretary Timothy Geithner announced that the debt had reached its limit and began the extraordinary measures (accounting steps that let Treasury keep paying bills after it hits the limit) that would carry the government to about August 2. A bill to raise the limit to $16.7 trillion with nothing attached, introduced by Ways and Means Chairman Dave Camp, a Republican, failed in the House 97 to 318 on May 31: every Republican voted no, and Democrats split 97 to 82. Two months later, on August 2, the Budget Control Act was signed.

It raised the limit by $2.1 trillion in three steps ($400 billion that day, $500 billion on September 22 and $1.2 trillion on January 28, 2012), set caps on discretionary appropriations for 2012 through 2021 that CBO scored at $917 billion of deficit reduction, and created a joint committee charged with proposing at least $1.5 trillion of further deficit reduction, with an automatic sequester of $1.2 trillion if legislation achieving at least that much was not enacted. It was not, and the sequester took effect on March 1, 2013. CBO's total for the act, $2.1 trillion over 2012 to 2021, is the caps plus that sequester. The act passed the House 269 to 161, with Democrats split 95 to 95, and the Senate 74 to 26.

Congress then raised the caps four times before they expired. The 2013 agreement raised the 2014 defense and nondefense caps by about $22 billion each. The 2015 act raised the 2016 caps by $25 billion each. The 2018 act raised the 2018 defense cap by $80 billion and the nondefense cap by $63 billion, and the 2019 caps by $85 billion and $68 billion. The 2019 act set 2020 at $667 billion for defense and $622 billion for nondefense, and raised 2021 by $81 billion and $72 billion, to $672 billion and $627 billion. Each of the four also extended the automatic cut to some mandatory programs (ones that run without an annual vote, such as Medicare) by two more years, to 2029 by the last of them, and three of them (2015, 2018 and 2019) carried a debt limit suspension. That is the origin of the suspension as a habit: from 2013 on, Congress mostly chose to suspend the limit for a set period rather than raise it by a set amount. When a suspension ends the limit is reset at whatever the debt has reached, so Treasury is left with no headroom and the extraordinary measures start again. The resets were $19.81 trillion on March 16, 2017 and $21.988 trillion on March 2, 2019.

The other law of the decade was the 2017 tax act, by reconciliation, under a budget resolution that allowed up to $1.5 trillion of new deficits over ten years. The Joint Committee on Taxation estimated it would lose $1.46 trillion of revenue over 2018 to 2027: $1.13 trillion from the individual provisions, $653.8 billion from the business provisions, offset by $324.4 billion raised from the international provisions. The corporate rate cut was made permanent. It passed the House 224 to 201 and the Senate 51 to 48, with no Democrat voting for it in either chamber. Most of the individual provisions were written to expire at the end of 2025, which is what the 2025 act, famously introduced as the One Big Beautiful Bill Act, was built around.

2020 to 2025: COVID and following years

Six spending bills in a little over a year, with CBO's estimate of what each added to deficits over ten years:

The CARES Act was $988 billion of mandatory spending, $408 billion of revenue reduction and $326 billion of emergency appropriations on CBO's April 2020 estimate. It sent $1,200 per adult and $500 per child to households (under an income limit), extended unemployment benefits through the end of 2020 to people who would not otherwise have qualified, created the Paycheck Protection Program (PPP), and appropriated $100 billion to reimburse hospitals and other providers for pandemic costs and lost revenue. The Senate passed it 96 to 0 and the House by voice vote. Four weeks later the fourth law added $321 billion to the Paycheck Protection Program.

The December 2020 act sent an additional $600 per adult and dependent, put $82 billion into an education stabilization fund and $25 billion into rental assistance, and increased SNAP benefits through June 2021. The 2021 act, by reconciliation, sent an additional $1,400 per adult and dependent, which JCT scored at $393.7 billion in 2021 and $16.9 billion in 2022. It appropriated $350 billion of state and local fiscal recovery funds (CBO's spending estimate was $362 billion), $170 billion for education including $123 billion for elementary and secondary schools, $29 billion for restaurants, $15 billion for vaccine activities, and it raised the child tax credit for one year, to $3,600 per child under six and $3,000 per older child, made it fully refundable (paid even to families that owe no income tax), and paid half of it out monthly from July 2021. JCT put the refundable payments of that one-year credit at $88.8 billion, $84.4 billion of it in 2021 and 2022. The act passed the Senate 50 to 49 and the House 220 to 211 without a single Republican vote in either chamber.

In October 2021 Congress raised the debt limit by $480 billion, the first time in almost ten years that it set a number rather than suspending the limit, and in December it raised it again, by $2.5 trillion, to just under $31.4 trillion.

The 2022 act, by reconciliation, was actually the first party-line bill of the period that CBO scored as reducing deficits: $58.1 billion over 2022 to 2031, from $108.7 billion of new revenue against $50.6 billion of new spending. The revenue was a 15% corporate minimum tax ($222.2 billion) and an excise tax on stock buybacks ($73.7 billion); the largest spending items were energy and manufacturing credits and three years of larger premium subsidies under the Affordable Care Act ($32.8 billion). Medicare drug price negotiation reduced spending by $98.5 billion. The act also gave the IRS $79.6 billion, which CBO expected to raise $180 billion of revenue that CBO does not count in its estimate. The Senate voted 50 to 50 with Vice President Kamala Harris breaking the tie, and the House passed it 220 to 207.

With Trump back in office in 2025, the One Big Beautiful Bill, passed by reconciliation, made the expiring 2017 individual provisions permanent and also added to them. In CBO's February 2026 baseline (its projection of the budget if the law does not change) it reduces revenue from individual income taxes by $4.4 trillion over 2026 to 2035 and total revenue by $4.9 trillion; JCT's estimate of the individual, estate and gift tax changes alone is $4.1 trillion of primary deficits over 2025 to 2034, about half of it from the rate reductions. It made the larger estate exemption permanent ($227 billion), set the child tax credit at $2,200 per child permanently ($192 billion of added spending with the earned income credit), appropriated $154 billion of mandatory defense money for shipbuilding, missile defense and munitions, and $165 billion of mandatory funding for the administration of justice, the budget function that includes immigration enforcement. On the other side it cut health programs by $1.0 trillion, most of which is coming from Medicaid, which CBO expects to reduce Medicaid enrollment by 12.9 million and raise the number of uninsured by 7.5 million in 2034; cut SNAP by $187 billion through work requirements and benefit formula changes; and changed the terms of federal student loans. The CBO February 2026 baseline attributes $4.7 trillion of added deficits over 2026 to 2035 to the act, and the act raised the debt limit by $5 trillion, to $41.1 trillion. The Senate voted 50 to 50 with Vice President JD Vance breaking the tie; three Republicans voted against the bill. In the slim Republican-controlled House the vote was 218 to 214.

The bills

Below are the bills, how they were passed, the final votes in the House and Senate, and what each added to or subtracted from the federal deficit on the official estimate (a plus sign adds to the deficit; the sources page gives the basis for each row). Independents, such as Angus King and Bernie Sanders, are counted with the party they caucus with.

One note: in CBO's February 2026 baseline the 2025 act adds $4.7 trillion over 2026 to 2035, more than the $3.4 trillion over 2025 to 2034 in the table, which was CBO's estimate when it passed.

The debt limit, action by action

The votes

Which ones cut the deficit? Six of the fourteen: 1990, 1993, 1997, 2011, 2022 and 2023. Four of the six were bipartisan in both chambers. The other two passed with one party's votes: 1993, by two votes in the House and the Vice President in the Senate, and 2022, by the Vice President breaking the tie in the Senate.

Three of the increases were crisis responses, 2009, 2020 and 2021. The 2009 act got three Republican votes in the Senate and none in the House. The 2020 act passed the Senate 96 to 0. The 2021 act was passed by reconciliation eleven months after that vote, without a Republican vote in either chamber.

The remaining five were deliberate. Three tax cuts, the prescription drug benefit, and the 2025 act. Four of the five were reconciliation bills. The one that went through regular order, the drug benefit, passed the House 220 to 215.

Then there are the rules. Statutory caps and pay-as-you-go rules were in force from 1990 through 2002, the window that holds all four surplus years; came back for 2012 through 2021, raised four times before they expired; and were set again for 2024 and 2025. In the twenty-five years since the last surplus, statutory caps were in force for thirteen.

What happened last time

Each side thinks the other side wastes money, but neither side seems very serious about curtailing the cost, and the public is not pressing them. The last time interest took this much of the economy, in 1991, it produced a third-party presidential candidate who ran on essentially one thing.

Ross Perot took 19,742,267 votes in 1992, or 18.91% of the total 104,426,659 cast, against 43.01% for Clinton and 37.45% for Bush. Zero electoral votes, but it is the best showing by a candidate outside the two parties since 1912, and the platform was the deficit and trade, in that order, and not much else.

The 1992 deficit was $290.3 billion, the largest the country had run in dollar terms to that point. It fell every year after: $255.1 billion, $203.2 billion, $164.0 billion, $107.4 billion, $21.9 billion, and then four surpluses.

Whether that was Perot, the 1993 budget act, a Republican Congress arriving in 1995 across from a Democratic president, or a decade of growth nobody had correctly forecast, is an open debate people still have. It is the only stretch in the modern record where the line went the other way for eight straight years, 1993 through 2000, and for five of those eight years the arrangement, a Democratic president across from a Republican Congress, wasn’t one that either side wanted.

In January 1996 Gallup found 28% of Americans naming the deficit the country’s most important problem, first on the list and the highest the deficit had ever scored on that question. In Gallup’s monthly readings from December 2024 through June 2025 it got between 2% and 7%. In competitor Pew’s July 2026 survey, asked which party has better policies for dealing with the deficit, 33% said Democrats, 31% said Republicans and the remaining third said neither.

9Five Countries

Is this a problem unique to America? Five other large economies carry government debt under different rules but each have their own similar, if not worse problems: Japan, where the central bank owns the debt; Germany, which wrote a limit on borrowing into its constitution and then changed it to borrow for infrastructure and defense; the United Kingdom, where the debt is long, a quarter of it rises with inflation, and the bond market turned negative on the government in 2022; Canada, which (somewhat surprisingly) cut its spending and its debt in the 1990s; and France, where disagreement over the budget has quickly brought down two governments since December 2024, with a presidential election due in the spring of 2027, Marine Le Pen the favorite to win it (44% on Polymarket as of October 25, against 16% for the next candidate), and street protests over the cost of living that began in September and had put 2,500 people in custody and partial curfews on Paris, Lyon, Marseille, Bordeaux, Lille and Nice by October 20, on Reuters' count.

Side by side

Each country measures its own debt differently, so the first column is not a like-for-like ranking. The yields are the OECD's monthly averages for each country's 10-year government bond, for September 2026 (the U.S. figure on this basis is 4.99%; the Treasury's end-of-month close was 5.29%). Average maturity is as of the end of 2025 from Japan's Ministry of Finance, which compiles it from each country's debt office (Japan's is the end of its fiscal 2025), except Canada, which is as of March 2025 from Canada's Department of Finance. n/a means no official figure on a comparable basis.

Every one of the six 10-year yields rose between December 2021 and September 2026, by between 2.4 points (Canada) and 4.4 points (France and the United Kingdom). The United States has the shortest average maturity of the six, which means a new level of rates reaches its interest bill faster than it reaches any of the others.

Japan: when the central bank owns the debt

Japan is the country everyone brings up on this subject. What it shows is what near-total domestic ownership of the debt looks like, and what it took to get there.

As of March 2026, holdings of Japanese government bonds and treasury bills broke down like this, against a total of ¥1,150.1 trillion, according to Japan's Ministry of Finance:

The central bank owns 42.2% of it. Foreigners own 13.7%, which is less than half the American figure of 28.8%. Japanese households directly own 1.7%.

When your own central bank holds the largest block of your government debt, and holds it at yields set by that same central bank, the interest cost of the debt stops being a market outcome. In Japan the central bank carried much of the cost, so it never appeared as a budget line.

Japan: what it costs now

The cost is showing up in the budget again. The Ministry of Finance's own budget document for fiscal 2026:

Interest payments were ¥8.5 trillion in the fiscal 2023 budget, ¥9.7 trillion in fiscal 2024, ¥10.5 trillion in fiscal 2025 and ¥13.0 trillion in fiscal 2026. The Ministry gives two reasons for the ¥2.5 trillion jump, and they are the two reasons in section five of this paper: the rate applied to the new bonds it plans to sell was raised from 2.0% to 3.0%, and "the replacement with higher-interest-rate issues amid a rising government bond balance." The budget rate of 3.0% was set against a recent one-month average long-term rate of 1.9%, with a margin for past episodes of sharp increases. Debt service is 25.6% of the general account. Interest alone is 15.5% of tax revenue, against 19.5% for the United States in fiscal 2026.

The Ministry's own debt ratio, which counts only the central government's bonds, is 168.6% of GDP at the end of the fiscal year that closed in March and projected at 165.5% a year on; broader measures that add local government and social security debt run higher. Japan has limited new borrowing: new bonds under ¥30 trillion for two consecutive years, the first time in seventeen years, and a bond dependency ratio under 30% for the first time in twenty-seven years. Its borrowing rates rose anyway.

The 10-year Japanese government bond yielded −0.025% in December 2019, 0.055% in January 2021, 1.655% in October 2025 and 3.05% in September 2026. The overnight rate the Bank of Japan controls went from −0.038% at the end of 2019 to 1.043% in September 2026. Long-term rates went first and went further, for the same reason as in section one: the buyer that had been setting the price stepped back. In June 2025 the Bank of Japan published a plan to cut its monthly purchases of government bonds from about ¥4.1 trillion to about ¥2 trillion by the first quarter of 2027, confirmed the schedule in June 2026, and kept a clause allowing it to buy more "in the case of a rapid rise in long-term interest rates."

Japan's central bank went to 42.2% of the debt. The Federal Reserve went to roughly a quarter of the debt held by the public in 2021 and holds about a seventh now, and it remits nothing to the Treasury while it works off a $233 billion deferred asset. Japan's interest bill rose 53% in three budget years; the American net interest bill nearly tripled in five fiscal years. One more number: foreigners hold 13.7% of Japanese government debt overall but 55.6% of Japanese treasury bills, so Japan's short-term debt is majority foreign-owned. America's is held by money market funds and leveraged hedge funds, which can also leave quickly.

Germany: the debt brake, and borrowing to build

Germany's constitution limits federal borrowing. Articles 109 and 115 of the Basic Law, the debt brake, hold the federal government's structural deficit to 0.35% of GDP. General government debt (all levels of government combined) was 63.5% of GDP at the end of 2025 on the European Union's Maastricht measure (the EU's standard definition of government debt), up from 62.2% a year earlier.

On March 18, 2025, the outgoing Bundestag changed the rule. By 512 votes to 206, with no abstentions, it amended the Basic Law so that defense spending above 1% of GDP no longer counts against the limit; created a special fund of €500 billion for infrastructure and for reaching climate neutrality by 2045, to be committed over twelve years; and gave the states, which had been allowed no structural borrowing, room of 0.35% of GDP.

The interest on that borrowing is already in the budget. Federal debt service is €33.65 billion in 2026 and €43.61 billion in the government's draft for 2027. The federal finance plan has interest rising to €55.2 billion in 2028, €68.1 billion in 2029 and €80.7 billion in 2030, including the interest on the infrastructure fund, which is €673 million in 2026 and €2.31 billion in 2027. The 10-year Bund (the German government's 10-year bond) yielded −0.38% in December 2021 and 3.47% in September 2026. Germany stopped issuing inflation-indexed bonds in 2024, which Japan's Ministry of Finance attributes to "concerns about rising interest payment costs resulting from rising inflation rates." The average German federal security matures in 7.6 years.

Germany shows that even the most fiscally conservative constitution of the six chose to borrow to build: a €500 billion fund for infrastructure and climate, plus defense spending outside the limit. Its interest bill still roughly doubles by 2029 on the government's own finance plan. Borrowing to build costs money to carry too, but Germany will have something to show for it. In the United States, four cents of each new federal dollar went to physical resources (section seven).

The United Kingdom: long debt, indexed debt, and a confidence shock

The United Kingdom's public sector net debt was 93.8% of GDP at the end of August 2026, 1.3 points lower than a year earlier and, in the Office for National Statistics' words, "at levels last seen in the early 1960s." Three things make it behave differently from ours.

The first is maturity. The average gilt (a UK government bond) matures in 13.8 years on the Debt Management Office's measure for 2025-26 (13.5 on the comparison basis in the table above). That is more than twice the U.S. figure. Debt that long reprices slowly: it takes more than a decade, not a few years, for a new level of rates to work through it.

The second is indexation. A quarter of the UK's debt, 25.3%, is linked to the Retail Prices Index, an older UK measure of inflation. When prices rise, the principal rises with them, and the increase is booked as interest in the same month. Central government debt interest was £8.8 billion in August 2026, the highest August figure since monthly records began in 1997, and £2.1 billion of it was that uplift on index-linked gilts.

The third is 2022. In September of that year, in the Bank of England's own account, vulnerabilities in pension funds' liability-driven investment strategies (leveraged strategies for matching what they owe retirees) "amplified the impact of an abrupt change in fiscal stance into a self-reinforcing spiral in government bond prices." On September 28 the Bank began temporary purchases of long-term gilts "to restore orderly market conditions," on "whatever scale is necessary to effect this outcome." The 10-year gilt yielded 0.84% in December 2021 and 5.24% in September 2026, the highest of the six countries here.

Canada: a change of belief on spending

Canada's Department of Finance publishes its fiscal history as a share of GDP, back to 1966. The mid-1990s look like this:

In 1993-94 federal program spending was 16.4% of GDP and the deficit 5.2%. Four years later program spending was 12.7% of GDP and the budget was in surplus. Public debt charges, Canada's term for interest on the debt, had been 6.5% of GDP in 1990-91, about 38 cents of every dollar of revenue; by 2024-25 they were 1.7% of GDP, about 10 cents. Federal debt peaked at 66.6% of GDP in 1995-96 and was 29.0% twelve years later. The pandemic took it back to 47.2%; it is 41.2% now. Canada's market debt was $1.48 trillion (Canadian) at the end of 2024-25, at a weighted average interest rate of 2.76%, with an average term to maturity of 6.5 years; non-residents held 35% of Government of Canada securities.

France: two governments down

French general government debt was 119.0% of GDP at the end of June 2026. On December 4, 2024, the National Assembly brought down Michel Barnier's government with a motion of censure carried by 331 votes, after the government forced through its social security financing bill; it was the first French government censured since 1962. On September 8, 2025, his successor, François Bayrou, asked the Assembly for a vote of confidence on a statement of general policy and lost it, 194 to 364. Under Article 50 of the constitution the government resigned.

The 10-year French government bond yielded 0.05% in December 2021 and 4.48% in September 2026, against 3.47% for Germany. In August, the latest month with Italy's figure, France paid 4.03% and Italy 3.99%: France paid more to borrow for ten years than Italy did. France's tradable debt matures in 8 years and 142 days on average, as of August 31, 2026 (8.5 on the comparison basis in the table).

Who reprices fastest

Maturity sets how fast a new level of rates reaches the interest bill. The United Kingdom’s debt reprices over thirteen and a half years, Japan’s over nine and a half, France’s over eight and a half, Germany’s over seven and a half, Canada’s over six and a half, and ours the quickest, just under six.

10What It Costs a Household

Household debt does not move Treasury rates. But the same rate increases that repriced the government's debt repriced every mortgage, card and car loan written since.

Wages have beaten inflation over the last decade. From December 2016 to September 2026 the consumer price index rose 38.6%. Average hourly earnings for all private employees rose 45.8%. After inflation that is a gain of 5.2% in ten years, most of it before 2021, with a stretch in 2022 when real wages went backwards. It's a real gain and a small one, but the claim that wages haven't kept up with prices isn't what the data says.

The median household has less to show. Real median household income was $85,320 in 2019 and $87,460 in 2025, in 2025 dollars, up 2.5% in six years. Over roughly the same span, real disposable income per person, which is an average and includes investment income, rose 13.4%.

The things you finance got far more expensive than the things you buy outright.

The house got 34% more expensive and the mortgage rate more than doubled, so the monthly payment on a median house with 20% down went from $1,097 to $2,275. At the end of 2021, with the median house at $414,000 and the rate at 3.11%, the payment on the same terms was $1,416. At the end of 2022, with the median at $442,600 and the rate at 6.42%, it was $2,219. The house is cheaper now than it was at the end of 2022, by $32,000, and the payment is higher. Roughly two thirds of the increase since 2016 is the rate, not the price.

Every rate a household pays

The same thing happened everywhere you borrow. The rates at four points: 2016, the last autumn before the pandemic, late 2021, when borrowing was cheapest (the lows are labelled), and now.

If you borrowed for school in 2020–21, you pay 2.75% for the life of the loan. A student borrowing this autumn pays 6.52%, because the rate is set each May off the 10-year Treasury auction and the 10-year is where it is. The rate on a card balance is 8.8 points higher than it was in 2016, and the rate on a car loan three and a half points higher. On September 17, the morning after the Fed's increase, the prime rate went from 6.75% to 7.00%, and prime is the rate most card and home equity balances are priced from.

Who owes what

Total household debt was $18.8 trillion at the end of June, down $13 billion in the quarter. The largest pieces: $13.1 trillion of mortgages, $1.71 trillion of auto loans, $1.65 trillion of student loans, $1.26 trillion of credit cards and $459 billion of home equity lines. The Fed's household debt service ratio, which is required debt payments as a share of disposable income, was 11.11% in the second quarter of 2026, below the 11.87% of 2016 and far below the 15.85% of 2007.

A large part of the $13.1 trillion of mortgages was written or refinanced in 2020 and 2021 at rates near 3%, and a 30-year fixed mortgage does not reprice. If you have one of those loans, you're carrying it at 2021 rates and will for decades. The 7.40% rate falls entirely on whoever has to borrow now: the first-time buyer, the family that has to move, the person whose card balance rolled over at 22%. Home equity lines of credit grew $13 billion in the quarter, to $459 billion, in an expansion the New York Fed dates to 2022. Households with a 3% first mortgage that need cash are borrowing against the house on a second, floating-rate line rather than refinancing the first at 7.40%.

The one-year credit

The American Rescue Plan Act contained the clearest measured improvement in any quality-of-life series over the pandemic years, and it lasted one year. It was the expanded child tax credit described in section eight: for 2021, up to $3,600 per young child, paid in full even to families with no earned income, half of it in monthly checks. The child poverty rate on the Census Bureau's Supplemental Poverty Measure was 9.7% in 2020, 5.2% in 2021, 12.4% in 2022, 13.7% in 2023 and 13.4% in 2024. The Census Bureau attributes the 2022 rise to the expiration of the expanded child tax credit and earned income credit and the end of stimulus payments. Congress funded one year of it and did not renew it. The 2025 act set the credit at $2,200 permanently.

States and cities

The same rate reaches states and cities. Every municipal bond in the country is priced off the Treasury curve, and the 30-year Treasury hit 5.67% on October 7, its highest close in more than two decades. A higher municipal rate shows up in a property tax bill or in a service that doesn't get funded, and 49 states have a balanced budget requirement in their constitution or statutes on the National Conference of State Legislatures' count, so there is no borrowing through it.

What a city pays was mostly set years ago. Chicago's pension bill went from $978.3 million in 2016 to $2.76 billion in 2026, nearly a tripling, and it is now 16.7% of everything the city spends. The schedule is set by statute, on a ramp to 90% funded by 2055 for the police and fire funds and 2058 for the rest, so the city has to pay whatever the schedule says each year.

The states got the pandemic money and kept much of it in balances. Their combined year-end balances were $121.6 billion at the end of fiscal 2019, 14.0% of general fund spending, and $437.3 billion in 2023, 38.8%.

They are being spent down now, to a proposed $274.5 billion in 2027, and spending is not coming down with them: general fund spending in 2027 is still 50% above 2019. The $350 billion the American Rescue Plan Act sent directly to states and localities mostly replaced revenue lost in the pandemic, on GAO's count, so the programs it paid for are now state obligations, paid out of state revenue, at the same moment states are paying more to borrow for their roads and buildings.

Who can hedge and who can't

The Treasury's average rate on everything it owes is 3.531%. It chooses the maturity of what it sells, and when it needed $2.7 trillion in a single quarter in 2020, its own central bank bought the equivalent of 61% of it. The funds that replaced the Fed as buyers can hedge the level of rates or exit overnight (section four). You can't short a Treasury future against your car loan. Your one hedge was a fixed-rate mortgage taken out before 2022, and that was only available if you already owned a house by then.

The cost of a mortgage taken out today traces back through the mortgage-backed securities market to the 10-year Treasury, which is set in the same auctions that fund the federal deficit. This is the one main connection between the national debt and a household budget. The debt does not make your groceries expensive. The rates set to stop your groceries from getting more expensive are what doubled your mortgage payment.

Most people’s income didn’t double between 2016 and 2026, but the payment on a median house did.

11Conclusion

So is there a catalyst, and if so, what is it?

The economy seems to be doing well, all things considered. The Fed continues to claim it is expanding at a solid pace, unemployment is at a healthy 4.2%, and the stock market continues to provide positive returns. One big problem though: people just don’t believe it. Consumer sentiment hit its lowest reading since the survey began in 1952 this past May, and only 17% of Americans currently trust the government. Perception is reality, and the perception is that something is off.

Spending levels are elevated, but the reason is not any singular program. The problem is the interest the spending now carries. Interest took 19.5 cents of every dollar the government collected in the fiscal year that just ended and over half (53%) of the deficit, and the CBO sees it as 73%(!!!) of the deficit by 2035. One third of the debt comes due every year and is borrowed again at present-day rates, so the bill will rise whether or not anyone in DC does anything. The $26 trillion borrowed since 2001 mostly kept things where they were. Only four pennies of each new dollar of spending built anything, and the median household is only 2.5% better off than it was pre-pandemic, pre-trillions of new dollars injected into the economy.

Each side sees wasteful spending differently, but neither has actually done anything about it. Both parties ran deficits every year they held power since 2002. The debate now is about who should pay more and who should pay less. Democrats have spent the better part of a decade with their economic groupthink dominated by the Bernie Sanders and Elizabeth Warren wings of the party, running on slogans like “tax the rich” and using potential wealth taxes to accomplish some of their spending goals (these proposed taxes typically fall well short of the funding needed for “Medicare For All” or a similar program, but that’s a story for another paper). Democratic-governed states like Massachusetts and Washington have passed legislation that taxes the highest earners (I wrote about Washington state’s here). Republicans have begun their own type of populism, typically aimed at seniors and (freshly) property owners: in Florida this year, a Republican proposal in the state House would have amended the state constitution to end every property tax except the school tax on the homes of people 65 and older (it died in committee in March). Even Jeff Bezos, one of the five richest men of all time, proposed that the bottom 50% shouldn’t be paying income taxes (I also wrote about that here). We are in an age of ideas, some good, some bad, some outlandish, and in the marketplace of public opinion, whoever can push those ideas out to the masses usually has the most success in making their will heard. The institutions that financed the debt can hedge what it costs; households cannot, and the payment on a median house has doubled. There is no kumbaya moment coming to the country; there is simply too much current disagreement over what color the sky is.

The catalyst, if it comes, will come from one of three places.

The first is the debt limit. Congress set it at $41.1 trillion in July 2025. On October 15 the debt stood at $40.39 trillion, which leaves roughly $700 billion of room. Over the last year the debt grew by $2.53 trillion, about $210 billion a month, so at that pace the room is gone in the first months of 2027 and the Treasury goes back to extraordinary measures. The two deficit deals of the last fifteen years, in 2011 and 2023, both came attached to the debt limit, a $2.1 trillion increase the first time and a suspension the second. The next one will be the first negotiated with the interest bill above a trillion dollars, which is sure to make headlines.

The second is Social Security. It currently is on track to be depleted in the fourth quarter of 2032. After that the program would only be able pay 78% of scheduled benefits. This has been a classic kick the can down the road to the next guy routine in Washington, but nobody genuinely expects social security to run out. It would be political suicide as millions of seniors depend on that money as their only source of income. But with that will a new spending bill or funding source. Medicare’s funding is also set to be depleted in second quarter of 2033 to 89% of current levels. There have been some ideas on how to move forward that don’t require tax raises or increased funding from current tax areas in place, but no action as of yet.

The third is the market. The 10-year closed above 5% in September for the first time since 2007, after two years in which the Fed cut and long rates rose anyway, and the bills share of the debt sits above the 20% guideline of the Treasury’s own borrowing advisory committee. CBO’s ten-year path assumes a 10-year near 4.3%; it closed October 16 at 5.09%. If that gap holds and the debt reprices to it, as a third of it does every year, the interest line runs about $250 billion a year above the baseline on today’s debt alone. Markets do not vote, but when they turn, it sure feels like they do.

In January 1996, four years after Perot’s run, 28% of Americans called the deficit the country’s most important problem. In Gallup’s readings from December 2024 through June 2025 it drew between 2% and 7%, and in July Pew found that a third of the country thinks both parties are in the wilderness on it. The midterms are on November 3, days after this is published, and the deficit is not in the top 20 things voters are imploring the candidates they support to talk about at rallies. The last time the bill was this heavy, a third-party candidate built a campaign on it and the budget improved for eight straight years afterward.

So my answer is this: there is no catalyst visible today, and the next two are already on the calendar. Spending gets forced into the conversation by the debt limit in early 2027 and by Social Security in 2032. What happens at those two dates depends on which rate the debt carries when they arrive, which is why the rest of this paper spent so long on the rate.

So what would or could change the path? It’s cliche but education would be a start. Most people never learn how important the Federal Reserve is, how rates are raised or cut, and how their mortgage, or car loans or student loans are set. A country that understood that on a deeper more personal level would notice when a third of its debt reprices every year.

Realistically though, we also probably need rates to fall at some point, because the last time the interest bill came down for good it was rates falling for twenty years and the debt refinancing at each lower step, not the debt shrinking, and because if there is one thing we like doing in America, it is kicking the can and handling the problem later.