An essay with data · The United States, Britain, the euro area and Bulgaria, 1946–2026

Do Governments Benefit from Inflation?

The inflation of 2021–23 moved about a trillion dollars from the people who had lent to the US government to the Treasury. Then the voters removed the people running it. Both things happened in most rich countries, and the distance between them is the answer to the question.

Teodor Dermendzhiev ·

There is a belief, held about equally by people who distrust governments and people who manage bond funds, that governments like inflation. A state owes trillions in a currency it can print. Every rise in prices makes that debt lighter. So when prices rise, the suspicion goes, somebody in the finance ministry is quietly pleased. Keynes gave the idea its best sentence in 1923, in a chapter called “Inflation as a Method of Taxation”: “It is the form of taxation which the public find hardest to evade and even the weakest Government can enforce, when it can enforce nothing else.”

Two facts from the last five years sit awkwardly together. In fiscal 2022 the US Treasury paid an average of 2.1% on its debt while prices rose 7.1%. In real terms its lenders paid it 4.6% for the privilege of lending. Over three years inflation took 13 points off the federal debt ratio. And then, in 2024, the governing party lost vote share in every one of the twelve developed western countries that held a national election, something the Financial Times's John Burn-Murdoch found had not happened before in almost 120 years of records.

If inflation is such a good deal for governments, it is odd that it destroys them. I live in Bulgaria, which ran this experiment at full strength in 1997, and where the question is not academic. So I did what I did for the last two essays: collected the data and the papers, and checked. Every US Treasury security outstanding at the end of 2020; eighty years of federal budgets; the IMF's figures for 45 countries; the British income-tax schedule; and about thirty studies, from Keynes to the European Central Bank. The short version:

$1.07tn
lost by the holders of US Treasuries outstanding at the end of 2020, to inflation they had not been paid for. That is 5% of GDP, and it is the Treasury's gain
29 pts
taken off the US debt ratio by inflation in 1946–55, twice what real growth did. The war debt was not grown away
$391bn
of extra interest the Treasury paid in fiscal 2025 alone, compared with its 2020 borrowing rate. The gain came once; this comes every year
12 of 12
governing parties in developed western countries lost vote share in 2024. The treasury gained; the government did not

The windfall

Start with the cleanest case. On 31 December 2020 there were $14.6 trillion of ordinary fixed-rate US Treasury bills, notes and bonds in the hands of someone other than the Federal Reserve, equal to 68.5% of GDP. Each is a promise to pay a fixed number of dollars on fixed dates. That day the bond market expected inflation of about 2% a year for the next decade: the five- and ten-year “breakeven” rates were 1.95% and 1.99%.

Prices then rose 29.6% between December 2020 and October 2026. At 2% a year they would have risen 12.2%.

I took the Treasury's list of every security outstanding that day, removed the Fed's holdings security by security, and valued each remaining coupon and repayment twice: once at the prices lenders expected, once at the prices that arrived. The difference is what the lenders lost. Since the Treasury is the one paying, it is also what the Treasury gained.

The longer the loan, the bigger the loss
Share of expected real value lost by holders of US Treasuries outstanding on 31 December 2020, by time left to maturity on that day
Treasury Monthly Statement of the Public Debt (security level) less Federal Reserve holdings from the New York Fed; CPI-U from the BLS. Payments deflated by actual CPI against a 2% path; payments after October 2026 assume the price gap stays where it is. Undiscounted. Inflation-indexed and floating-rate securities excluded.

The total is $1.07 trillion: 5.0% of 2020 GDP, or 7.3% of what the holders expected to get back. The answer moves with the assumption about what they expected. If it was 2.5% inflation, the loss is $0.88 trillion; if it was the Cleveland Fed's 1.4% estimate, $1.32 trillion. Most of it, $926 billion, was locked in by the end of 2023.

Maturity decides who lost. Someone holding a bill due within a year lost 1.7%, because they got their money back in months and could lend it again at a higher rate. Someone holding paper with five or more years to run lost 13%. Roughly 43% of the marketable Treasuries outside the Fed were held abroad, so a large part of this was paid by foreign central banks and foreign savers.

Other people, using other methods, find the same order of magnitude. Robert Barro and Francesco Bianchi estimate that across 37 OECD countries roughly 80% of the extra pandemic spending was financed by inflation eroding existing debt. The IMF finds that in countries with debt above 50% of GDP, each point of surprise inflation cuts the debt ratio by 0.6 of a point. The ECB puts the euro-area gain from the 2022 surprise at close to 5 points of GDP by 2024.

What makes the episode remarkable is that the market had put a price on it beforehand. Jens Hilscher, Alon Raviv and Ricardo Reis used inflation options to ask how likely it was that inflation would erode US debt. On prices from 2017, the market gave a 0.3% probability to an erosion of more than 4% of GDP (on a somewhat narrower measure of the debt than mine). Their reasons were sensible: much of the debt is short, and nobody expected inflation to stay high for years. The paper was published in 2022, by which time the one-in-three-hundred event was under way.

Verdict on “inflation lightens the debt”: true. About 5% of GDP for the United States, taken from lenders in five years, with no vote.

It has worked before

The case everyone cites is the debt from the Second World War. The United States ended fiscal 1946 owing 106% of GDP and by 1974 owed 23%, without ever defaulting and without a decade of austerity. The comfortable account is that the country grew out of it.

The change in a debt ratio can be split exactly into four parts: interest adds to it, inflation and real growth shrink it by enlarging the economy it is measured against, and the budget excluding interest adds or subtracts depending on whether it is in deficit or surplus. Here is that split for the first nine years after the war, next to the three years after the pandemic.

What moved the US debt ratio
Contributions to the change in federal debt held by the public, in percentage points of GDP. Blue reduced the ratio; red raised it.
Own calculation from OMB Historical Tables (net interest, debt held by the public) and fiscal-year GDP and GDP deflator from FRED. “Other” is the residual: asset purchases, loans and timing differences.

From 1946 to 1955 the ratio fell by 50 points. Inflation accounts for 29 of them and real growth for 14. Budget surpluses, which people forget the Truman and Eisenhower governments ran, account for 17. Over the whole period to 1974, inflation took off 48 points and growth 41.

Inflation worked so well because the interest rate did not respond to it. Until 1951 the Federal Reserve held long-term rates at 2.5% by agreement with the Treasury, and for decades afterwards regulation kept rates low and kept banks and pension funds buying. In fiscal 1947 the Treasury paid 1.7% on its debt while prices rose 11.9%.

For a generation, lending to the Treasury lost money
The real interest rate the US Treasury paid on its debt each fiscal year: net interest divided by debt, less inflation. Blue: below zero.
Own calculation: net interest (OMB Table 3.1; Monthly Treasury Statement for 2024–25) over the previous year's debt held by the public, less the change in the fiscal-year GDP deflator.

The real rate was negative in half of the 34 years from 1947 to 1980. Carmen Reinhart and Belen Sbrancia call this financial repression and estimate that it “liquidated” debt worth 3.2% of GDP a year in the United States and 3.6% in Britain between 1945 and 1980, about a fifth to a quarter of tax revenue. Julien Acalin and Laurence Ball rerun the history without the surpluses, the surprise inflation and the pegged rates, and find the debt would have fallen only to 74% of GDP by 1974.

Seen from the lender's side: someone who held long US government bonds from January 1946 to September 1981, reinvesting every coupon, ended with 51% less purchasing power than they started with. That is thirty-five years of saving to lose half.

Lenders noticed. The average maturity of US government debt was more than nine years in 1947. By 1976 it was 31 months. Nobody would lend long at a fixed rate to a borrower who had just done this, and a short debt reprices before inflation can do much to it. George Hall and Thomas Sargent, who reconstructed the wartime bond portfolio and found that by 1951 it was worth 70% of its pre-war value in real terms, conclude that the government “will have more difficulty using inflation to impose real losses on its creditors today.”

The broader record is less kind to the idea than the American case. Barry Eichengreen and Rui Esteves assembled debt reductions across 183 countries and 220 years. Moderate inflation helped with the large ones, they find, but successful reductions were less frequent in periods of relatively high inflation.

Verdict on “we grew out of the war debt”: mostly false. Inflation with pinned-down interest rates did twice the work of growth in the first decade. It needed controls that no longer exist, and it is the reason lenders stopped lending long.

The two smaller ways

Debt is one of three channels. The other two are older and get more attention than they deserve.

Printing the money

This is the one Keynes meant. Banknotes are a loan to the state at zero interest, and inflation is the rate at which the loan is being written off. Economists call it the inflation tax: the tax rate is the inflation rate, and the tax base is the cash people are willing to hold. The trouble for the government is that the base runs away. The faster prices rise, the less cash anyone keeps.

For the United States I computed it for every year since 1960. The inflation tax on currency averaged 0.13% of GDP a year in the 1960s, 0.33% in the high-inflation years 1970–81, and peaked at 0.57% in 2021. Federal revenue is about 17% of GDP. Over 2021–23 holders of dollar cash lost $345 billion of purchasing power, against the $129 billion that 2% inflation would have taken. It is real money and a rounding error in the budget.

It is a serious source of revenue only for states that have lost the ability to tax in the ordinary way. Stanley Fischer, Ratna Sahay and Carlos Végh went through 94 economies: the highest take ever recorded was a little over 6% of GDP, in places such as Israel, Chile and Argentina in their worst years, and across countries revenue peaks at an inflation rate of about 174% and falls beyond it. Alex Cukierman, Sebastian Edwards and Guido Tabellini found that the countries leaning on it are the ones with unstable, polarised politics. That is Keynes's sentence with data behind it: the inflation tax is what a government collects when it can collect nothing else.

Taxing numbers that inflation pushes up

If tax thresholds are fixed in money and wages rise with prices, people drift into higher bands without being any richer. The government collects more with no change in the law. This channel depends entirely on a decision not to adjust the thresholds, and Britain has made that decision. Its personal allowance and higher-rate threshold have been frozen since April 2021.

Same real pay, higher tax rate
UK income tax as a share of pay, for workers whose pay rose exactly in line with prices from April 2021 to September 2026 (+31%)
2021–222026–27, thresholds frozen
Own calculation: income tax only (no National Insurance), rest-of-UK rates, personal allowance £12,570 and higher-rate threshold £50,270 in both years; CPI from the ONS. Pay shown in 2021 pounds.

A worker on £30,000 in 2021 whose pay has kept exact pace with prices now pays 13.6% of it in income tax instead of 11.6%. At £45,000 the rate has gone from 14.4% to 18.7%, because part of the pay has crossed into the 40% band. At £100,000 it has gone from 27.4% to 34.5%. None of these people had a real pay rise. The Office for Budget Responsibility expects the freezes to raise £44.6 billion a year by 2028–29, 1.4% of GDP, with nearly four million more people paying income tax and three million more paying the higher rate.

Where thresholds are indexed, the channel is nearly shut. The United States has indexed its brackets since 1985; before that, the Congressional Budget Office reckoned income-tax revenue rose about 1.5% for each 1% rise in incomes. In 2021–23 the extra revenue from this effect was under 0.2% of GDP in Germany and France, which adjust their schedules, and 0.6–1% in Italy and Spain. Bulgaria has a flat 10% income tax with no allowance, so there is no band to drift into.

Verdict on “governments print their way out”: false for any state that can still tax. Verdict on “inflation is a stealth tax rise”: true where thresholds are frozen, and only there.

It pays in proportion to what you owe

The debt channel is simple arithmetic, and the arithmetic has a consequence people forget. Inflation shrinks a debt ratio by the inflation rate times the debt. A government with little debt gains little, however fast prices rise.

Inflation helped the heavily indebted, not the heavily inflated
Points of GDP that inflation took off each country's public debt ratio in 2021–23, against the size of the debt at the end of 2020
advanced economiesothers
Own calculation from the IMF World Economic Outlook, April 2025, via DBnomics: general government gross debt, GDP deflator. 43 countries shown; Turkey and Argentina are off the scale and discussed in the text. US debt here is the IMF's general-government measure, larger than the federal figure used above.

Debt ratios fell between 2020 and 2023 in 26 of the 33 advanced economies in the IMF's data, while governments were running the pandemic deficits. The typical country got about 10 points of relief from inflation. Greece, which owed 210% of GDP, got 24 points from a 14% rise in prices. Japan got 10 points from a 4% rise, because it owes 258%. Bulgaria had one of the highest inflation rates in the European Union, 34% in three years on this measure, and got 6 points, because it owed 23% of GDP. Its debt ratio ended almost exactly where it began.

The currency of the debt matters as much as its size. Argentina's prices rose more than sixfold over the same three years, and its debt ratio went up, from 104% to 155% of GDP, because much of the debt is in dollars and the peso's collapse made every dollar owed more expensive. Inflation erases what you owe in your own money and nothing else.

One result did not come out the way I expected. If inflation surprises were what drove the improvement, the countries with the biggest price surprises relative to the IMF's October 2020 forecasts should show the biggest debt surprises. Across the advanced economies they do not; the correlation is weak and has the wrong sign. Everyone had roughly the same shock, and what each government then did with its budget mattered more.

Verdict on “high inflation is a gift to the finance ministry”: depends on the debt. Large for Greece, Italy and the United States. Close to nothing for a low-debt country, and negative for one that owes dollars.

The bill

A government gets the windfall once, on the debt it had already issued at the old interest rates. Everything it borrows afterwards, including the refinancing of that old debt as it matures, is priced by lenders who have just been burned.

US federal net interest was $345 billion in fiscal 2020, 1.6% of GDP. In fiscal 2025 it was $970 billion, 3.2% of GDP. Some of that is simply more debt. To isolate the rate, I compared each year's interest bill with what it would have been at the 2020 average rate of 2.06% on the same debt.

The windfall came first. The interest bill is still arriving.
Cumulative, in billions of dollars: what inflation took from holders of the end-2020 debt, and the extra interest paid because the Treasury's borrowing rate rose above its 2020 level
erosion of the old debtextra interest from higher rates
Own calculation. Erosion: as in the first chart, by calendar year. Extra interest: net interest less 2.06% of the previous year's debt, by fiscal year (ending 30 September). The two lines are not a profit-and-loss account; see the text.

In 2021 the Treasury was still paying less than in 2020. By fiscal 2025 it had paid $830 billion more over five years, and the gap that year was $391 billion. At that pace the running total will have overtaken the $1.07 trillion windfall during fiscal 2026, which ended in September and which I have not yet computed.

That comparison is harder on the government than it should be, and the chart needs a caveat. Part of a higher interest rate is compensation for higher inflation, and that part costs the borrower nothing in real terms. After inflation, the Treasury paid 0.7% and 0.9% on its debt in 2024 and 2025, against an average of 0.5% in 2015–20. So the accurate description is this: the Treasury had three years of deeply negative real rates, and is now back to paying roughly what it paid before, on a larger debt. The windfall was real. It was also all there was.

Other costs do not have that excuse.

The central bank. The Federal Reserve bought trillions of long-dated bonds and paid for them with reserves on which it pays the overnight rate. When it raised that rate to fight inflation, its costs passed its income. It sent the Treasury $107 billion in 2021, stopped in September 2022, lost $114 billion in 2023, and as of this month still has $232 billion to earn back before it sends anything again. This is why I left the Fed's holdings out of the windfall: on that part of the debt the government had, in effect, lent to itself at a floating rate.

Indexed spending. Almost a third of euro-area government spending rises automatically with prices, mostly pensions, mostly with a year's delay. The ECB's own assessment is that the 2022 surprise helped debt ratios and hurt budget balances.

Indexed debt. A large part of British government debt is linked to the retail price index. UK debt interest reached £111.6 billion in 2022–23, 4.3% of GDP and a post-war record; the OBR's rule is that each extra point on the index adds £5.9 billion in the same year.

Inflation everyone expects. The CBO publishes rules of thumb for what happens to the US budget when the economy changes. If inflation and interest rates run a tenth of a point higher every year, revenues rise by $370 billion over a decade and spending by $634 billion. Expected inflation makes the deficit bigger. The gain exists only in the gap between what lenders were promised and what they got.

Verdict on “inflation is free money for the state”: false. It is a one-off transfer on old debt. New debt is priced accordingly, and the rest of the budget is mostly indexed.

The full-strength version: Bulgaria, 1997

What happens when a government takes the whole dose is on record in the country I live in. In 1996 Bulgaria had loss-making state enterprises, banks lending to them, a central bank financing the banks and the budget, and a domestic government debt of 65% of GDP that could only be rolled over at shorter and shorter maturities. In the winter of 1996–97 the currency collapsed, and in February 1997 prices more than tripled in a single month.

By the debt arithmetic it was a triumph. According to the OECD's 1999 survey, domestic government debt fell from 65.3% of GDP to 16% in a single year. The part denominated in leva fell from 31.5% to 5.5%. Interest payments, which were running at 20% of GDP around the turn of the year, were 1.2% by the fourth quarter of 1997. (Some of the fall was IMF money replacing domestic borrowing; most was inflation.)

Everything else went the other way. Taxes are assessed on one day and paid weeks later, and in a hyperinflation the delay eats the revenue; add collapsing output and evasion, and the IMF's Anne-Marie Gulde records tax receipts falling from almost 40% of GDP to 14.7% in February 1997. Household savings in leva were erased by the same inflation that erased the debt. The government resigned after weeks of street protests, and its party lost the election that April. In July the state gave up monetary policy altogether and tied the lev to the German mark through a currency board, an arrangement it kept until it joined the euro this year.

The debt was wiped out together with the savings, the tax system and the government. The thing that survived was a rule that the central bank may never lend to the state again.

Verdict on “a government can inflate its debt away”: true, once, if it is willing to lose everything else.

Who paid

A tax has payers, and this one is not levied in proportion to income. It falls on whoever holds claims fixed in money: bonds, bank deposits, cash, pensions that are not indexed. It benefits whoever owes fixed sums: the government first, and after it anyone with a fixed-rate mortgage.

Matthias Doepke and Martin Schneider worked this out for the United States before anyone thought it would be tested. A surprise inflation moves wealth from old, wealthy bondholders and from foreigners to the government and to young households with mortgages. In their example of five extra points of inflation for ten years, the government gains between 5% and 13% of GDP and foreigners lose between 5% and 8%.

For 2021–23 a group of economists, most of them at the ECB, did the count for the euro area. The loss to households was about 3% of three years' income in France and Spain, 7% in Germany and 9% in Italy. Retirees lost up to 14%. About half of 25- to 44-year-olds came out ahead, because inflation shrank their mortgages faster than it shrank anything they owned. And “with the possible exception of Germany, national governments were net winners, even when accounting for the additional spending” on energy subsidies and pensions.

In my Treasury calculation the losers are whoever held paper with three or more years to run at the end of 2020: foreign central banks, pension funds, insurers, banks. The winner is the Treasury, which in the end means the future taxpayers who will not have to service that trillion dollars.

Verdict on “inflation hurts everyone”: false. It moves money from lenders to borrowers, from old to young, and from foreigners to the state that issued the currency.

Why the people in charge lost anyway

So the finance ministries of the rich world had a good three years, and the people who ran them were voted out. The Financial Times count is twelve of twelve developed western incumbents losing vote share in 2024, with the governments of India and Japan losing ground as well. Inflation was not the only cause. Burn-Murdoch's own summary lists stagnation, immigration and a general disillusionment alongside it. But it is the one thing all those countries had in common at the same moment.

The explanation, I think, is that the gain and the cost land in different places.

The gain is invisible. No budget has a line for “erosion of outstanding debt”. It shows up as a debt ratio that is lower than it would have been, which nobody experiences. Even the winners do not know they won: Stefanie Stantcheva surveyed Americans on why they dislike inflation and found that people believe their wages do not keep up, blame the government and business, and see no offsetting benefit at all. Half of the young households in the ECB study were net gainers. I doubt many of them would have said so.

The cost is visible every day, in every shop, and it has an obvious address. A treasury is an institution with a balance sheet and no term limit. A government is a group of people with four years. The first can be better off after an inflation that ends the careers of the second.

The same gap explains why governments do not do it deliberately and repeatedly, which is what the cynical version of the belief implies.

It only works as a surprise. Inflation that lenders expect is already in the interest rate. The IMF and the CBO find the same thing from opposite directions: surprise inflation cuts debt, expected inflation does nothing or makes the deficit worse.

It works less well each time. Lenders who have been surprised shorten the loans they will make and ask for a premium. The United States went from nine-year average maturities to 31 months in thirty years. In 1946 the long bonds were there to be eroded. In 2020, 41% of the debt outside the Fed was due within a year.

Governments have tied their own hands. Independent central banks, indexed tax brackets and indexed bonds all exist to remove the temptation, and they were mostly built after the 1970s by governments that had seen what it cost. Thomas Sargent and Neil Wallace described in 1981 what happens when the restraints fail: if the deficit is fixed first and lenders will not finance it, the central bank “is forced to create money and tolerate additional inflation.” That is a description of Bulgaria in 1996.

Verdict on “governments want inflation”: false for governments, which are made of politicians. Half true of treasuries, which are not.

What the evidence doesn't say

  1. It doesn't say anyone did this on purpose. Nothing here is evidence of intent. The market gave the outcome a 0.3% chance, the central banks that could have prolonged it raised rates instead, and they did so at a loss to themselves and to their treasuries.
  2. It doesn't say $1.07 trillion to the dollar. The figure is undiscounted, runs from $0.9 to $1.3 trillion depending on what lenders expected, counts state governments and foreign central banks as ordinary holders, and assumes the gap in the price level stays where it was in October 2026. One outside estimate I saw quoted but could not trace to its source is somewhat higher.
  3. It doesn't say the interest bill is the price of the windfall. Rates rose everywhere, part of the rise is compensation for inflation, and the Treasury's real borrowing cost is back near where it was. What the evidence says is narrower: the gain does not repeat, and the debt it applied to has since been refinanced at higher rates.
  4. It doesn't say inflation decided the 2024 elections. It says every incumbent in the sample lost ground in the same year, that inflation is the common factor, and that surveys show voters blame governments for it. Immigration and weak growth are in the same data.
  5. The data has edges. The cross-country figures use the IMF's gross general-government debt and GDP deflators, not consumer prices. The British calculation is income tax only. The Bulgarian figures for 1997 come from OECD and IMF accounts written two years later, in a period when the statistics themselves were unstable. Several of the published estimates I cite come from models, Barro and Bianchi's most of all.

So do governments benefit?

A government that owes a lot of long-dated debt in its own currency gains from inflation it did not announce. It gains once, at the expense of the people who lent to it, and in proportion to how much it owed. That was worth about 5% of GDP to the United States after 2020 and about 29 points of the debt ratio after 1946. It was worth almost nothing to Bulgaria in 2021–23, and less than nothing to Argentina.

The other two channels are small or optional. Printing money raises real sums only for states that are already failing. Fiscal drag raises real sums only where a government chooses to freeze its thresholds, which is an ordinary tax rise that happens not to be announced as one.

And the gain is paid for. It is paid for in interest on everything borrowed afterwards, in a central bank that stops paying dividends, in indexed pensions, and in shorter and more suspicious lending for a generation. Mostly, though, it is paid for by the people in office when it happens. That is the part I had not understood before doing the numbers. The folk belief treats “the government” as one thing with one interest. The balance sheet and the politicians are two things, and inflation is good for the first and usually fatal to the second. Which is probably the best protection savers have: the people who would have to choose inflation are the ones who lose their jobs for it.