What the US Treasury and the Federal Reserve should announce:

1) The fiscal house needs to be put in order right now:

a) Introduction of a federal VAT at a rate of 10%.
b) Social Security: the retirement age will immediately be increased by two years. In the future, the retirement age will be linked to US national life expectancy. The same indexation will apply to the Medicare eligibility age.
c) The revenues will be used to turn the Social Security system into a fully funded system. The VAT is what pays for the transition – one generation would otherwise have to pay twice.

2) A nominal freeze on all federal spending until the structural budget balance has been restored. With NGDP growing at 4%, this cuts federal spending as a share of the economy by around 4% a year without a single nominal cut.

3) All tariffs will be scrapped immediately.

4) A framework for nominal stability:

a) A 4% NGDP level target, with today’s NGDP level as the starting point. The target path will be adjusted upwards once, by the estimated one-off price level effect of the VAT net of tariff removal, so that a tax change is not mistaken for a monetary one.
b) Interest rates will be fully market determined.
c) Every quarter, the Fed will announce the growth rate of the money base it will maintain until further notice to keep NGDP on the target path – and it will be judged, and correct itself, against that path rather than against any forecast of its own.
d) The Fed and the US Treasury will refrain from all intervention in fixed income and FX markets, other than the Fed buying a predetermined basket of currencies and commodities to ensure the announced growth rate in NGDP for the coming period.

This would ensure that the budget situation improved dramatically and very fast. NGDP would settle at 4% growth, which would deliver inflation of around 2% over the medium term, and US Treasury yields would stabilise, likely just below 4% at the long end of the curve.

The problem: the US population has not been told the scale of the fiscal challenge facing the nation, and both the Democrats and the Republicans have only unrealistic quick fixes that won’t work. Tariffs and “DOGE” failed, and “taxing billionaires” will fail in the same way.

The federal government is running a structural deficit of around 6% of GDP at full employment, and that is not sustainable for much longer. If Americans don’t realise this very soon, the US is heading for a major fiscal and financial crisis.

If you want to avoid the fate of Argentina over the past 60 years, it is time you stopped behaving like Argentinians – the road there is not hyperinflation overnight, it is a fiscal position that eventually decides monetary policy for you.

The US Treasury said this afternoon (European time), in a press release, that it will at least double the size of its liquidity support buybacks of long-dated government bonds. In the two long sectors, 10-20 years and 20-30 years, the maximum per operation goes from 2 to at least 4 billion dollars.

The change takes effect on 9 September and runs to the end of the refunding quarter. It follows the 30 July refunding, where the frequency of operations in those same two sectors was already doubled from two to four per quarter, and the quarterly ceiling was lifted from 30 to 38 billion dollars.

Officially this is about market liquidity. But look at the week it arrives in. On Monday the 30-year yield closed at 5.31%, the highest since 2007. The week before, the Treasury had to sell 25 billion dollars of 30-year paper at 5.216%, the highest auction yield since 2001. The 30-year real yield has gone from 2.63% in January to 3.06%. And foreign holdings of Treasuries fell in June, with the UK, China and Japan all selling.

It worked today. The 30-year fell almost 9 basis points to 5.196%, the 10-year 6 basis points to 4.647%, and equities rallied. Scott Bessent got the headline he was after, and I expect he is pleased with himself this evening.

I do not think it lasts, and my reason is a simple one. There is no money to buy the bonds with.

Nothing in the till

There is no cash pile sitting in the US Treasury that can be deployed to support the bond market. Quite the opposite. The federal deficit was 432 billion dollars in July alone, the largest month since March 2021, and interest on the debt has cost around 1.2 trillion dollars so far this fiscal year.

Every dollar spent on a buyback has to be borrowed first. The Treasury sells paper at one end of the curve to buy paper at the other, so, as far as I can see, net demand for US government bonds does not rise by a single dollar. And if the operations are funded with bills, duration comes out of the market, which is an Operation Twist run by the wrong institution. The price of that is a shorter government debt, and a budget that becomes more sensitive to the short rate, at a moment when inflation has run above target for five years.

I would also hold on to the scale here. Two billion extra per operation, against 739 billion dollars of net borrowing in this quarter alone.

Only one institution can create dollars out of thin air

That institution is the Federal Reserve. And I note that the Fed is already at it.

Quantitative tightening ended on 1 December last year, and from 12 December the Fed has been buying Treasury bills. Its holdings sat at 195 billion dollars from April 2024 through to the middle of December 2025. On 12 August this year they stood at 534 billion. That is 338 billion dollars in eight months, and the highest level in the history of the series.

Officially these are reserve management purchases and not monetary policy. The Fed buys short paper for two reasons: to keep bank reserves ample, and to move the portfolio towards the short end of the curve, where issuance already sits. The New York Fed has been explicit about it: these purchases are not a change in the stance of policy, and should not be confused with large-scale asset purchases.

Now follow the dynamic, because this is what worries me.

The Treasury is shifting issuance towards the short end because the long end has become expensive.

The Fed is buying at the short end because reserves must be ample. And in February the Treasury’s own advisory committee discussed something worth pausing on: whether it would be reasonable, in the current environment, to meet part of the central bank’s demand for bills through heavier issuance in that sector.

Every time the Fed buys a bill, reserves are created, which is to say money is created. Ample, in the Fed’s own framework, is a range set by a judgement about rate sensitivity rather than a number, so there is no external limit that says stop. The distinction between reserve management and monetary financing of the deficit becomes a question of intent. You cannot read intent off a balance sheet.

If that is the road we are on, I would expect the dollar to take a beating and inflation expectations to go sharply higher.

Warsh’s test

This is where the Fed has to demonstrate that it is an independent central bank. Kevin Warsh left the Board in 2011, in protest at QE2. He has argued ever since that an oversized balance sheet, whatever its operational justification, puts the central bank in a relationship with the Treasury where politicians can pile up debt without feeling it. He now sits at the head of the table, with a balance sheet growing again and a Treasury Secretary busy propping up the long end.

In my view the test will not be whether he raises or cuts. It will be whether he says no when the Treasury’s issuance choices start to determine what the central bank buys.

The man who thinks he is smarter than the market

And then there is Bessent himself. He has an unfortunate habit of believing he can outsmart the market rather than deal with the thing the market is actually reacting to.

I have watched him do it before. When Moody’s downgraded the United States last year, he called it a lagging indicator. In January he dismissed talk of currency intervention on the grounds that sound fundamentals attract capital on their own.

Six months later he sold euros to buy yen, and told the European Central Bank afterwards. The yen moved 3.2%, half of it came back in August, and US long yields went on to a 19-year high. The intervention held for about a week.

I see the same habit today. The man who thought the market had the dollar wrong now thinks it has the long bond wrong. What he has not done, in either case, is put forward a plan to do something about the fundamental problem – the massive US government budget deficit.

You can push markets around in the short run with liquidity operations, and I have no quarrel with the tactic as such.

But Japan and China have no appetite for US Treasuries. Nor, after the way this administration has behaved for the past eighteen months, is there much left in Europe. The market is not mispricing the long end. It is pricing the fiscal policy it has been shown.

Which brings me to an American president worth listening to on the subject:

“You can fool some of the people all of the time, and all of the people some of the time, but you cannot fool all of the people all of the time.”

Abraham Lincoln

US vice president JD Vance has said in public that he does not think the dollar’s reserve currency status is good for the United States.

Back in 2023 JD Vance said this:

“I am not sure that I think the reserve currency is actually good for the United States of America. I think there is a good argument that reserve currency status is akin to coal in Appalachia: it’s a resource curse, right? It allows your consumers to consume very cheaply.

That’s been the story of the American economy over the last 15 or so years, even before that. We can just borrow basically an unlimited way because we have the reserve currency, so that’s a massive subsidy to incurring debt. The debt is cheaper even in a raised interest rate environment…

So we have this massive subsidy to cheap debt for the American consumer. Well, that’s good, right? Consuming is important, especially food, medicine, things like that. But I think it’s a massive tax on American producers.”

What is he saying here?

That it is a problem that the rest of the world will go on financing the large US budget deficit and thereby holds the dollar stronger than it would otherwise be. And that this hits US manufacturing.

And the analysis is right enough. Reserve currency status does make it possible for both the state and the private sector to finance themselves more cheaply. And yes, it is a challenge for manufacturing, if you disregard what happens to wages and prices when the dollar is stronger.

Vance has bought completely into Trump’s obsession with the manufacturing sector, and I think the conclusion is clear. This administration believes that the US economy can be lifted by a cheap dollar, and that reserve currency status stands in the way.

And it is not only talk

On Friday 31 July the US Treasury went into the currency market and bought Japanese yen. The United States had not done that since 1998, which is twenty-eight years ago.

It did not pay in dollars. It sold euros out of its own reserves, and it did not tell the European Central Bank until after the trade had gone through.

On the surface this is two finance ministries, the Japanese and the American, intervening in the currency market, and twenty or thirty years ago we saw that fairly often. A Treasury Secretary did not think a currency was where it ought to be, so he bought and he sold.

Note that these are the finance ministries and not the central banks, which matters more than it sounds. The central bank may execute the trade, and in the US case the New York Fed does it on the Treasury’s instruction, but the decision was US Treasury Secretary Scott Bessent’s.

The two things are one thing. The vice president says he does not want the dollar to be the reserve currency, and the Treasury Secretary goes into the market to strengthen the yen. He does it in euros rather than in dollars, and I would not want anyone to be reassured by that, because buying yen against the dollar is pushing the dollar down whatever currency you pay with.

So the US government is now actively trying to weaken its own currency without saying so directly. The market will work that out soon enough.

The real story is why now, and why after more than twenty years of not doing it. I have watched central banks and finance ministries in currency markets for a long time, and this one does not look like the others.

The official explanation does not survive contact

US President Trump said the Japanese had a weakened currency and would like some help, and that the United States is always there for Japan.

You can of course take the Americans at their word: the Japanese are in trouble because their poor currency has become so weak, the Japanese are our good friends, it is a shame for them, and so we are helping.

But the explanation is strange. Is the US Treasury Secretary to go out and help every friend whose currency weakens, and are we to buy every currency whose president is on good terms with Donald Trump?

The Treasury has already been very willing to support the Argentine peso for President Milei, who is seen as an ally, and the Japanese have recently acquired a conservative and relatively nationalist new prime minister whom Trump apparently likes. Meanwhile the Europeans are the ones we do not like, so we sell their currency and buy our friends’. I would call that a maxim with very little to do with economics.

“A massive subsidy to incurring debt”

The rule of thumb is that if the nominal interest rate on government debt exceeds nominal GDP growth, public debt rises as a share of GDP automatically, without any fiscal easing at all. And then you risk losing control not only of the debt, but of inflation expectations, and therefore of interest rates.

The 30-year Treasury yield, nominal GDP growth, and federal debt as a share of GDP, quarterly, 1980 to 2026. Source: BEA and US Treasury via FRED.

Take a look at how that has worked. The yield beat nominal growth in 89% of quarters from 1980 to 1999, and the debt ratio went from 31% of GDP to 58% over those two decades. It beat growth in 39% of quarters from 2000 to 2019, and in 19% since 2020, which is the only reason the arithmetic looks tolerable at the moment. Federal debt stood at 122.6% of GDP in the first quarter of this year.

Consider what that would mean. If the debt becomes unsustainable in earnest, it stops being only a fiscal problem and becomes a monetary one. The market begins to expect that the Federal Reserve will have to run the printing press into overtime, and then inflation explodes, the dollar collapses, and interest rates go through the roof.

We are not there yet. But underlying real growth in the United States is in my view around 2%, and if the Fed is to deliver 2% inflation, that gives 4% nominal GDP growth, which is well below the thirty-year yield. To secure 2% inflation and stabilise debt as a share of GDP at the same time would require massive fiscal tightening.

“The debt is cheaper even in a raised interest rate environment”

The thirty-year US Treasury yield keeps ticking up day after day. It closed at 5.21% on 13 August, and the last time it was at that level was 12 July 2007. We are at pre-crisis levels, and there is no doubt that this causes real concern in the US Treasury and certainly also at the Federal Reserve.

So I would expect the Treasury Secretary to be occupied with telling investors, who after all are the people financing the enormous US budget deficit, that something is being done about it.

Instead Bessent has turned his attention to the Japanese yen, which he thinks has become far too weak. The yen has weakened since around 2014 and at an accelerating pace since 2020, and it is precisely since 2020 that the long US rate has risen in earnest.

The 30-year US Treasury yield and the yen against the dollar, daily, 2000 to August 2026. Source: US Treasury and Board of Governors via FRED.

Bessent’s logic is therefore this. If he can get the yen strengthened, the close correlation with the US rate must mean that he thereby fixes his own problem. A stronger yen, a lower US rate.

And I think that is exactly as thin as it sounds.

The yen is around 158 today. It strengthened 3.2% in July and lost half a point again in August, so the intervention worked for about a week. And note where the thirty-year closed on the day of the intervention itself: 5.27%, the highest reading of the year.

So what does he do when the yen is back where it started within a month? His mission will have had one result, which is to burn part of the Treasury’s currency reserves, and one has to ask what the purpose was. The US voter might reasonably ask why money is being sent to strengthen the Japanese currency. If the Japanese want a stronger currency, they can raise interest rates like the rest of us.

And there is a reason the division of labour settled where it did. Central banks secure low inflation, finance ministries keep the public finances in order, and currency intervention disappears from the arrangement.

Imagine what happens when it does not. If the US government sold dollars, the dollar would weaken, which amounts to pumping money into the system and creating inflationary pressure. That runs directly against the Federal Reserve’s mandate of 2% inflation, and the Fed would have to soak it up again. A finance ministry cannot run around operating in the currency market entirely independently of what is being done with monetary policy.

You cannot write long books about successful intervention in the currency market, and people have tried.

The norm that was broken

Intervention is normally coordinated, and I would stress how deliberate that is. Economists call the reason beggar-thy-neighbour: if you weaken your own currency to buy competitiveness, everyone else does the same and nobody gets anything out of it.

The great central banks are in close contact for that reason. Under the pandemic they came out together and said that they were easing monetary policy, providing liquidity in size, printing money, doing it together and announcing it together, on the simple maxim that we should not work against each other and should get as much effect out of it as possible.

Everyone in the market could see what happened on 31 July anyway. It was not 800 European pension funds deciding simultaneously to buy yen and sell euros, and anyone who sits in the financial markets knows when a central bank is in the market. In Denmark, where we run a fixed exchange rate, Danmarks Nationalbank intervenes as a fully announced policy and confirms it afterwards.

So the Europeans broke no collegial confidence whatsoever, and I would put it more sharply than that. It is the Americans who stopped operating within the normal institutional frame, and it is the Trump administration’s conduct over the past eighteen months that has changed, not anybody else’s.

We are holding a party in the stairwell without telling the neighbours, and when somebody is annoyed about it, we decide the problem is theirs.

Nobody has asked why the yen is weak

Here is the thing I have not seen mentioned in this discussion at all. Donald Trump started a war with Iran, which has sent global energy prices sharply up, and Japan is extremely dependent on imported energy.

Had that war not been started, and had the oil price stayed nearer $65 a barrel than close to $100, the Japanese yen would in my view have been considerably stronger than it is today. So if that is really what is going on, Trump might look at his own foreign policy rather than at the currency market. And note the paradox: both the Chinese and the Japanese are hit hard by rising oil prices, which forces them to sell US government bonds.

Then there is the Bank of Japan, which raised its policy rate to 1% in June, the highest in thirty-one years, and held there in July. Core inflation was 1.6% in June, below the bank’s 2% target for the fifth month running, and the bank has cut its own forecast for the current fiscal year from 2.8% to 2.5%.

So inflation is not rising as much as expected, there is a repricing of how far the Bank of Japan has to go, and all else equal you get a weaker currency out of that.

And there are no warning lights flashing on Japanese inflation. We are not talking about 4 or 5%, we are talking about inflation just above 2%.

“It allows your consumers to consume very cheaply”

Some countries save and others dis-save, which is where I would start. For the past thirty or forty years the Asian countries have had large savings, while the Americans, and partly the Europeans, have run deficits both on the public finances and on the current account. The Americans simply consume more than they have money for.

Japanese households in particular have saved a great deal, so that Japan as a nation has plenty of savings despite its large public debt, and that money went to the most liquid bond market in the world, belonging to the nation with the greatest need to borrow it. The Americans have been able to live beyond their means for thirty years because there were willing buyers of US government bonds.

The ten-year Japanese government bond yield hit 2.88% on 9 July, the highest since 1996.

Take a look at what that does to a Japanese pension fund. If you can get almost 3% in your own currency with no exchange rate risk, that may well be preferable to placing the money abroad. So buyers disappear from the European and the US bond markets, and yields there are pushed up.

Japanese investors sold nearly $30 billion of US debt in the first quarter of this year alone, and Chinese holdings are down to $683 billion, the lowest since 2008, against a peak of $1,320 billion in 2013.

And they are not being replaced. Chinese ageing means Chinese growth comes down and Chinese savings get spent rather than placed abroad, Japan has been through the same for years, and the problem is no smaller in Europe. The Americans already spend more servicing the debt than they spend on defence.

So that, I think, is Bessent’s real worry. What if the Japanese stop buying our government bonds, and what happens then to the US interest rate?

The man who thinks he is cleverer than the market

Bessent was, in and of himself, a relatively unwritten page. He was known in the international financial world from his time with George Soros, but he had no remarkable career beyond that, and then he went solo and started his own macro fund, meaning a fund that invested on macroeconomic signals.

The truth is that the fund did terribly, really terribly, and it performed miserably against comparable funds right up until he decided to become Donald Trump’s Treasury Secretary.

Then it became interesting. He suddenly began arguing that tariffs were a good idea, and in the run-up to the presidential election he was the one trying to put Trump’s rather unorthodox economic proposals into a rational frame and say that it is not as insane as you think, and not really in conflict with economic theory.

As a theory nerd I get thoroughly irritated every time I hear him, because he uses economic and theoretical concepts wrongly, again and again. He seems like the one trying to be the clever boy in the class and is not, and I think more and more people have arrived at the same view. There is no larger plan, and his boss is completely uncontrollable.

Anyone who has been near the investment world has met the investor who is very, very certain that this is what is going to happen. When someone keeps telling you that the market is wrong, it is usually because he is the one who got it wrong, and once he starts calling the market irrational you can be sure of it. I have been there myself.

Bessent is that type. He thinks he can just fix it, that he is a little cleverer than the market, that he knows where the yen should be, or the US equity market, or the US bond market. That thinking is what made his own macro fund perform so extraordinarily poorly. Now he is doing it with the world economy.

They share a model, and it is taught nowhere

Vance has the same bad habit as his Treasury Secretary, which is to use technical terms they clearly do not understand. But the vocabulary is not the point, because they share an economic model.

It is certainly not taught at the world’s economics faculties, and it is, as Vance says, heterodox.

The model resembles to an uncanny degree Modern Monetary Theory (MMT), the internet’s darling of eight or ten years ago.

Central to MMT is that monetary policy should finance the public sector without inflation. And like MMT, the Vance-Bessent thinking offers a quick fix: you do not have to touch the public finances, because it can be fixed with tariffs and a weaker dollar.

Underneath it all sits the same conviction, which is that the market cannot be left to sort it out. This administration thinks it can set the right exchange rate, the right share price, and the right import prices, and fine-tune the economy that way.

That is extremely interventionist thinking, and it dominated economic policy in the Western world in the 1970s, which from the beginning of the 1980s we said we would get nothing out of. It is worth saying plainly when we are placing this administration on a left-right scale.

Vance is asking for a permanently weaker real dollar, which monetary policy cannot deliver, and he is asking for it in a period when inflation is already too high. The Federal Reserve has not managed to deliver 2% inflation in a single month in six years.

“A massive tax on American producers”

If you want an example of the consequences of this kind of thinking, look at Erdoğan’s Turkey.

Lira per dollar and the Turkish consumer price index, monthly, indexed to their 2016-2025 average = 100. Source: OECD via FRED.

Note first what this was not, because the lira floats and nobody devalued anything. This was monetary policy, and the currency did what a floating currency does when the money is that easy.

And then look at what the two lines do. The lira falling and Turkish prices rising is one event rather than two. The cheaper currency never sat there as a competitive advantage waiting to be used, because it went into the price level.

The bill

We have of course tried the coordinated version too. In September 1985 the G5 met at the Plaza Hotel in New York to get the dollar down, saying officially only that a further orderly appreciation of the non-dollar currencies was desirable, while behind closed doors they had set a target of a 10 to 12% depreciation and $18 billion of intervention over six weeks.

Reagan had cut taxes and expanded defence spending while Paul Volcker ran tight money against it, which is a basic lesson in macroeconomic theory: tight monetary policy and easy fiscal policy push rates up and make the currency stronger. The hole in the US public finances was nothing like the size of the one we have now. And the Americans decided the strong dollar was not only their problem but the rest of the world’s too.

At the Louvre in February 1987 the United States promised to bring the deficit down to 2.3% of GDP in fiscal 1988, Japan promised to ease monetary policy, and Italy turned up and refused to sign. The Japanese said yes and kept the promise, with the deficit coming down to about 2.5% and held there for years.

The great villain then was Japan, and there were stories about the Japanese owning all of Manhattan.

In 1987 Donald Trump, that Donald Trump, took out an advertisement in a number of US newspapers thundering against Japan for exploiting the Americans. When I heard him say a few days ago that Japan were good friends, I thought that the way he thundered against the Japanese nearly forty years ago is the way he now thunders against everybody else.

Nothing has changed. The Americans are still dependent on the Japanese and the Chinese and everybody else financing their large deficits. What has changed is that the problem has become markedly larger, and that the Chinese and the Japanese may no longer have the ability to do what they did before.

In my view neither Plaza nor Louvre had any positive effect on anything at all, other than to let the Americans think their public finance problems were for somebody else to fix. Japanese interest rates came down because of the financial crisis of the 1990s and an ageing population that arrived in Japan earlier than in the rest of the world. That is precisely the thinking behind Bessent and Trump, and the debt problem does not go away by selling euros and buying yen.

So let me put it together. Structurally, the Chinese and the Japanese as the two main buyers no longer have the desire or the ability or the need to buy so many international bonds. The Americans have not got a grip on their public finances. And the Federal Reserve has not been able to get a grip on monetary policy.

Add those three together and you get one thing. Interest rates keep creeping up, and if they do it in the United States they do it in the rest of the world. That is the threat in this for everyone outside America, and it is why I think what we have seen from both Bessent and Vance matters.

It ought to cause real unease on Wall Street and at the Federal Reserve, and not least among US consumers, who can hardly expect lower inflation.

Take them at their word

So my advice to investors and to policy makers outside the United States is simple. Listen to what Vance said in 2023. They – the Trump administration – mean it when they say it, and the mistake being made almost everywhere at the moment is to treat it as rhetoric that will be quietly walked back once somebody explains the arithmetic to them.

It is not rhetoric, because one of them supplies the doctrine and the other executes it in the market with public money. That is what 31 July was.

They will come to regret it, and I think the regret will arrive sooner than they expect. But regret is not a policy, and it does not unsay anything that has already been said to the market.

What we are looking at is a weak dollar policy in all but name. And underneath it sits a refusal: this administration does not want the responsibility that comes with producing the world’s reserve currency, and it would rather hand that job back than pay for it.

Whatever else this is, it can in no way be good for the dollar, and it can in no way be good for the outlook for US inflation. A permanently weaker dollar is not a competitiveness programme. It is synonymous with permanently higher interest rates and permanently higher inflation.

A couple of days ago, on 1 August, Brad Setser posted that The Economist had picked a good week for a currency cover, and that the Big Mac index agreed with his own work in putting the yuan around 30% undervalued. I replied.

Setser is the Whitney Shepardson senior fellow at the Council on Foreign Relations, and before that he was deputy assistant secretary for international economic analysis at the US Treasury and a senior adviser to the US Trade Representative.

He has spent two decades tracking global capital flows, reserve accumulation, and currency intervention, and his blog Follow the Money is one of the few places where the actual balance of payments data get read properly. I have a great deal of respect for Brad’s work, which is why the disagreement is worth having in public rather than in a footnote.

His number is exactly what the index says. In the July 2026 edition a Chinese Big Mac costs 31% less in dollars than an American one, and in January it was 33% less. The question is what that gap is a fact about.

The proposition is not a throwaway. It says that a persistent gap between a currency’s market rate and its purchasing power parity rate is a fact about the country’s price level rather than a fact about its currency, and if that is right then a good deal of what is written about undervalued currencies is a category error.

Let me be clear about where I start, because the reply reads as though I were reaching for Balassa-Samuelson first. I am not. I start with purchasing power parity, and I start there because it is the decisive factor. Prices and exchange rates move together, and everything else in currency analysis operates on the remainder.

I have forecast exchange rates professionally for a quarter of a century, and PPP has been the starting point of every model I have built. Balassa-Samuelson comes second, as the correction that says the parity a country converges to depends on how productive it is. I have tested that correction many times, in many formats, on many datasets. Never on the Big Mac index.

Which is a strange omission, because the index turned forty this summer.

It first appeared in the print edition of 6 September 1986, in a table of thirteen countries, under the heading of a light-hearted guide to whether currencies were at their correct level. Pam Woodall, then the paper’s economics editor, devised it as a semi-humorous illustration of purchasing power parity, and it gave the language the word burgernomics.

She cannot have expected it to still be running four decades later, in more than fifty countries, twice a year. The anniversary was as good an occasion as I am going to get to run the test again on data I had never used. And with a bit of help from Kaggle, Github and Claude it can now be done relatively fast.

What follows is the result: 2,036 observations, 54 currencies, and thirty years of the index, together with income, productivity, institutions, and governance data for every one of them.

What the index actually measures

Start with the theory, because almost every argument about the Big Mac index is really an argument about purchasing power parity, and most of the people having it have not thought carefully about which version they mean.

Purchasing power parity is one of the oldest propositions in international economics.

Gustav Cassel, who is one of my absolute favourite economists, gave it its name and its modern form during the First World War, when the gold standard was suspended and the practical question was which exchange rates Europe should return to afterwards.

And his answer was that the equilibrium rate is simply the ratio of the two countries’ price levels. So if a basket of goods costs 100 kroner in Denmark and 20 dollars in America, the equilibrium rate is five kroner to the dollar.

The Big Mac index applies exactly this logic to a single good: take the local price of a Big Mac, divide it by the American price, and you have an implied exchange rate that can be compared with the actual one. If the actual rate is weaker than the implied rate, the currency is called undervalued.

Note what this is not. It is not a measure of whether a currency will rise. It is a measure of whether the price level and the exchange rate are consistent with each other. Any adjust might happen through changes in the price level or the exchange rate – or both.

Figure 1 puts every observation on a single chart, with the Big Mac parity rate on the horizontal axis and the actual rate on the vertical, both measured in local currency per dollar and both on a logarithmic scale.

That covers seventy-four currencies over forty years and a span of seven orders of magnitude, from the Swiss franc at one end to the Vietnamese dong at the other.

Figure 1. Parity holds across seven orders of magnitude

The regression slope is 1.0162, which means a forty-year test of Cassel’s proportionality postulate cannot reject a coefficient of one, and the explained variance is 0.982.

Cassel was right!

The bit the chart hides

But now look at the same picture from the other side, because the median absolute deviation from the 45-degree line is 23%.

On a chart spanning seven orders of magnitude that deviation looks like nothing at all, while to anybody actually running money a 23% misalignment is enormous. And that gap is what the rest of this piece is about.

The standard explanation is Balassa-Samuelson, after Béla Balassa and Paul Samuelson, who published it independently in 1964.

The mechanism is straightforward once you separate goods into two kinds. Traded goods such as steel, wheat, and semiconductors have their prices arbitraged internationally, while non-traded goods and services such as haircuts, rent, and waiting a table do not.

Rich countries are rich because they are productive in the traded sector, and that productivity bids up wages across the whole economy, including in the barber shop where no productivity gain has occurred. So the price of non-traded goods rises with income, the overall price level rises with it, and rich countries end up expensive.

A Big Mac is itself mostly non-traded, being beef and bread assembled by local labour in rented premises. Not to mention electricity and taxes.

But Balassa-Samuelson is a statement about productivity, and productivity can be measured in a dozen ways that do not agree with each other.

So I ran the exchange rate equation with each of them in turn, alongside the burger price, a governance term, and a dummy for the four city states (See more on that further below). A negative coefficient means a stronger currency.

Eight of the ten carry the sign the theory predicts with a t above two – meaning it statistical significant at a 5% level.

Whether productivity is measured at purchasing power parity, at constant prices, per worker, per hour worked, as total factor productivity, or as how long people live, a more productive country has a stronger currency than the burger parity rate implies.

Human capital and years of schooling are the two that fail, and both are stocks of education rather than measures of output, which is not the same thing.

Output per hour is the theoretically correct variable, and in the latest Penn World Table it covers 52 of the 55 currencies, which it did not in earlier versions. Total factor productivity gives the largest coefficient of all, at −27.52, though it is also the least precisely measured.

Life expectancy is the one I keep coming back to. It contains no prices, no exchange rate, and no monetary quantity of any kind, and it still gets a t of −4.71.

Note what is not in the table. GDP per head measured at market exchange rates is the variable most people reach for, and it has no business being here. Convert income to dollars and you have multiplied by the exchange rate, which is the thing on the left-hand side. The relationship that produces looks strong and means nothing.

Figure 2 shows what the surviving relation looks like, with productivity measured as GDP per head at constant purchasing power prices, which is the measure that ends up in the model.

Figure 2. Richer countries have dearer burgers

The slope is 0.23. A country twice as productive as another has a burger price around 16% higher, which is the Balassa-Samuelson effect in its plainest form.

Note the spread around the line, though. At any given level of productivity the burger price varies by 50 log points or more.

Institutions are hard to separate from income

Every empirical exchange rate economist reaches for institutions at this point, and I am no exception.

The trouble is visible in one chart. Figure 3 plots the Fraser Institute’s measure of legal system quality and property rights against productivity. The correlation is 0.75, which leaves 43% of the variation in one that the other does not account for.

Figure 3. Institutions and income move together

I want to be careful about what that does and does not mean.

It is not that institutions are income, and nothing here says that good institutions are produced by being rich. Rich countries have good institutions and well governed countries are rich, and forty years of burger prices will not tell you which way that runs. What the correlation does mean is that the two are very hard to hold apart in estimation.

That shows up as instability. Put several institutional measures in together and they take turns having the wrong sign, because they are all proxies for the same underlying thing and each one absorbs whatever the others leave behind. Which measure survives a general-to-specific search depends on the order in which the others are removed.

So instead of picking one, here is every institutional measure I have, entered one at a time into the same equation.

All ten point the same way. Three are individually significant. The first principal component of all ten, which takes 84% of their common variation, enters at −3.61 with a t of −1.98.

That is the honest version of the institutional result, and it is a better one than picking the survivor of a search. The sign consistency across ten measures from two independent sources is the finding. The magnitude is not: the coefficients span a factor of fifty because the measures are on different scales, and which one is individually significant depends on nothing more interesting than measurement noise.

The equation below uses voice and accountability because it survived a general-to-specific search. I would not defend that choice. Rule of law or control of corruption has at least as good a theoretical claim, and both give a similar answer.

The model

Put it together and the equation is written the way a currency forecaster writes one, with the exchange rate versus the US dollar on the left.

100 × ln(local currency per dollar) = 2.58 + 0.989 × relative burger price − 15.68 × productivity − 0.437 × governance + 27.89 × city state

n = 2,036, 54 currencies, 1996 to 2025. R-squared 0.9868. Standard errors clustered by currency.

Note what is not on the right-hand side: no exchange rate, in any form, anywhere. The relative burger price is measured in local currency, the productivity term comes from the IMF in constant purchasing power prices, and the governance term is a percentile rank.

A negative coefficient means fewer local units per dollar, which is a stronger currency. So the reading is that productivity and governance both pull a currency above its parity rate, and that the four city states sit around 32% below either.

Every candidate had a sign attached to it before estimation, and anything that came out with the wrong one was removed. Twenty-two candidates went in and four came out. I want the model to be statistically correct and economic theoretically correct.

That procedure is not innocent, and the reported t values are too high because of it.

Selecting on the estimated sign and then quoting the significance of what survives overstates the evidence, and it does so most for the governance term. Read the burger price and productivity coefficients as estimates and the governance coefficient as an indication.

The burger price coefficient is 0.989 with a t against one of −0.99. Again highly significant.

Cassel’s proportionality postulate cannot be rejected, in an equation with no exchange rate on the right, on 2,036 observations across 54 currencies and thirty years.

One more test before the story is told, and it is the one that matters most.

The equation above is estimated in levels across currencies, which means it can be read as a statement about why some countries are dearer than others.

Add currency fixed effects and the question changes to whether the relation holds within a currency over time. It is a harder test, and two of the four terms do not survive it.

The burger price coefficient barely moves, at 0.97 either way, and productivity survives once vintage effects absorb the common movement in the world price level.

Governance does not survive. It changes sign and becomes significant in the wrong direction. Whatever the governance term is picking up, it is a difference between countries and not something that moves a currency when a country’s own institutions change.

I said above that I would treat it as an indication rather than an estimate. This is why.

That is close to being the whole story. Purchasing power parity, plus The Balassa-Samuelson effect, plus a governance term that is real but hard to size, plus a dummy for four places that are cities rather than countries.

Burgernomics is no evidence of Chinese currency manipulation

Which brings me to the argument that started all of this. Brad Setser as mentioned above that the Big Mac index agrees with his own work in showing the yuan substantially undervalued, and every American administration for twenty-five years has said something similar in considerably stronger terms.

The raw index does appear to show it, since China’s Big Mac has averaged 42% below the American price over the whole period.

However, conditional on the model, I find nothing of the sort.

The yuan has averaged within a few percent of the rate the equation implies, and the sign of the gap flips depending on which productivity measure goes into it: slightly undervalued on GDP per head, slightly overvalued on output per hour, and on the final specification 11% overvalued across the sample with the latest reading essentially on the line. Account for Chinese productivity and there is no unexplained undervaluation left to explain.

Figure 4 puts every currency in the January 2026 edition on one scale, with the United States given a residual of its own rather than assumed to be correctly priced.

Figure 4. Price level and exchange rate misalignment

Meanwhile the two largest unexplained undervaluations in the entire dataset are Taiwan at 21% and Hong Kong at 19%.

The Taiwanese burger is now 61% cheaper than the American one against a model prediction of 21%, and the gap has widened by around 55 percentage points since 1994 while Taiwanese income rose sharply.

The country everybody argues about is the one the model fits best. Just have a look below.

Figure 5 puts the fourteen most traded currencies side by side and shown as they are quoted: units of local currency per dollar.

The dashed line is the rate the model implies, which is the Big Mac parity rate adjusted for productivity, governance, and the city state term. The shaded area on the right-hand axis is the gap between the two, in percent.

Figure 5. The fourteen most traded currencies against the dollar

China is among the tightest fits in the panel. The yuan has averaged 7% away from the rate the equation implies, with the Canadian dollar at 0.5%, the New Zealand dollar at 3.7%, the won at 3.3%, and the yen at 5.5% alongside it.

Taiwan is the failure. The Taiwan dollar goes from roughly fair in the mid-1990s to 47% undervalued today, while the rate the equation implies barely moves. The Hong Kong dollar is the same story without the trend: 55 observations, undervalued in all but one of them, and the city state term takes only part of the gap. It is tempting to say that geopolitics might play a role here – both for Taiwan and Hong Kong.

The Swedish krona has averaged 31% overvalued and the krone 37%, which sounds like a standing feature of both economies. It is not. The krona peaked at 94% in 2011 and reads 9% undervalued in the latest edition.

The Norwegian krone peaked at 116% in 2006 and is now 2% below the line.

The euro has gone from 65% in 2008 to half a percent. All three converged on the model from above, at between one and five percentage points a year, and all three arrived.

Switzerland is the exception that did not arrive. The franc is the standing example of an overvalued currency, and the Swiss Big Mac is the dearest in the world at 9.08 dollars.

It has been converging too, from 106% in 1995 to 22% today, but at 1.7 points a year it is the slowest of the four and it still has two decades to go at that rate. Whatever makes Switzerland expensive is more durable than whatever made Scandinavia expensive. That being said – for all these currencies there are clearly elements of safe haven effects that the model does not fully account for.

The commodity dollars went the other way. Both the Canadian and the Australian dollar were 30% or more above the line during the commodity boom of 2011, and both are now below it, at 9% and 20% respectively. For these two the model reads the cycle rather than a level, and the cycle is the price of iron ore and oil.

How the gap closes

An equilibrium relation is only interesting if deviations from it come back. So the last question is whether they do, and how fast.

The object that has to revert is the real exchange rate: the dollar price of a Big Mac relative to the American one.

I estimated an error correction (EC) equation for it on the annual panel, with the lagged deviation from the model, one lag of the dependent variable, and changes in productivity and governance. 1,154 annual changes across 55 currencies, with years of collapse and hyperinflation removed.

It reverts, and the estimate is stable. Roughly nine percent of a deviation closes each year, which puts the half-life between seven and eight years, and adding short-run dynamics moves it hardly at all.

The country-level picture is consistent with the pooled one.

The adjustment coefficient has the mean-reverting sign in 40 of the 41 currencies with at least twelve annual changes, though only 16 of those are individually significant, and the median country half-life is three years rather than eight. That gap between the pooled and the median country estimate is the usual one in this literature: the pooled coefficient is dragged towards zero by the currencies that barely move.

Push the deviation further back and it fades in the way it should. Measured from two years earlier the coefficient is 0.078, from three years earlier 0.052, both still significant on the same sample.

Seven to eight years is slow, but it is also squarely in the range the purchasing power parity literature has settled on since Rogoff’s survey put the consensus at three to five years for the real exchange rate, and rather slower than that, which is what you would expect from a single traded good with a large non-traded component.

What I cannot do with these data is say which side does the adjusting.

The index is relative prices minus the nominal exchange rate, and estimating a separate equation for each side requires putting one of them on the right-hand side of the other, where it is endogenous by construction. A regime split would help, since a pegged currency cannot move by definition, but the available classifications describe each currency against its own anchor rather than against the dollar, and half the currencies my model calls pegged are pegged to the euro. That is a question for a different dataset – or a later blog post.

What forty years of burgernomics say

Four findings, in the order of how much weight they will carry.

Prices and exchange rates move together, and that is most of the story. The coefficient on relative prices in an equation for the nominal rate is 0.99, and it is 0.97 when currency fixed effects are added, so it is not an artefact of the pooling.

Deviations from the implied rate revert at 8 to 9% a year. Cassel’s proposition, formulated in 1918 to work out what exchange rates Europe should return to after the war, is the single most robust thing in this data.

Productivity moves the parity, as Balassa and Samuelson said it would. The elasticity is around 0.2. Eight of ten measures give it, from output per hour to total factor productivity to how long people live. The coefficient survives currency and vintage effects.

This is the correction to Cassel, and it is a real one, but it operates on the residual and not on the main relation.

Institutions probably matter and I cannot prove it. Ten institutional measures from two independent sources all carry the predicted sign, and three are individually significant. That consistency is worth something. But add currency fixed effects and the coefficient changes sign, as the model section showed. Institutional quality also correlates 0.75 with productivity, and no amount of econometrics on 55 currencies will separate the two.

And the index is not a currency forecast. It measures the consistency between a country’s price level and its exchange rate, and when the two are out of line the historical record says the gap takes seven or eight years to close. The model cannot say which side gives. A currency 40% from its implied rate is not a prediction that the currency will move 40%.

That last point is the one I would press hardest, because it is the one the index is routinely used against.

The Economist has presented it since 1986 as a guide to whether currencies sit at their correct level, and that framing invites the reader to treat a negative number as a trade. It is not. It is a measurement of how expensive one country is relative to another, and it is a rather good one.

So the index deserves a promotion and a demotion at once. As a currency signal it is weak. As an indicator of relative price levels between countries it is one of the longest and most consistent series anyone has.

Which brings me back to where I started – my reply to Brad Setser was right about the mechanism – poorer countries, i.e. countries with relatively lower productivity, have “undervalued” currencies.

A currency that has been cheap for decades is usually a poor country, and China is the clearest case in the data. Condition on Chinese productivity and the yuan sits within a few percent of where this framework puts it, with the sign of the gap flipping depending on which productivity measure goes in. The raw 30% is income, not manipulation.

On its fortieth birthday, then, the index says something rather unfashionable. Cassel was broadly right, Balassa and Samuelson were right about the correction, the yuan is roughly where Chinese productivity puts it.

References

Source data for the index are published by The Economist in its big-mac-data repository and are the basis for everything here.

•  Balassa, B. (1964), “The Purchasing-Power Parity Doctrine: A Reappraisal”, Journal of Political Economy 72(6)

•  Cassel, G. (1918), “Abnormal Deviations in International Exchanges”, Economic Journal 28(112)

•  Feenstra, R. C., Inklaar, R., and Timmer, M. P. (2015), “The Next Generation of the Penn World Table”, American Economic Review 105(10). Version 11.0 is used here, covering 185 countries to 2023, published in October 2025

•  Rogoff, K. (1996), “The Purchasing Power Parity Puzzle”, Journal of Economic Literature 34(2)

•  Samuelson, P. A. (1964), “Theoretical Notes on Trade Problems”, Review of Economics and Statistics 46(2)

Data

The Fraser index and the Penn World Table both end in 2023. For those series the last observation is carried forward. The main model uses only the Big Mac index, IMF income at constant PPP prices, and the governance indicators.

Second-quarter GDP figures arrived this week for both the US and the euro area, and the coverage has been almost entirely about a single number.

US real GDP grew at an annualised rate of 1.5% in the second quarter, down from 2.1% in the first, and it came in weaker than the euro area, which managed 1.8%.

The US economy is slowing, it is now slowing faster than Europe, and the natural conclusion might be that the Federal Reserve has kept policy too tight for too long.

However, that conclusion is totally wrong. The Fed does not control real GDP growth – the Fed controls nominal GDP growth and that actually accelerated from 5.8% annualised growth in Q1 to 7.9% annualised growth in Q2.

So the Fed has actually been EASING rather than tightening monetary policy.

I have seen almost nobody make that point this week, and I find that genuinely puzzling, because the nominal number is not a technical footnote to the real number. It is the only one of the two that the Federal Open Market Committee decides.

What a central bank can and cannot do

Start with the identity that every monetarist starts with. MV = NGDP = PY. The money stock times the velocity of money equals nominal spending in the economy, which in turn equals the price level times real output. It is true by definition, so on its own it explains nothing at all, and in my experience that is precisely why it gets waved away by economists who ought to know better.

But it stops being empty the moment we add a single proposition, and the proposition is that real output over the medium term is determined by the supply side rather than by the central bank. By technology, the labour force, the capital stock, regulation, and the quality of institutions. I would concede that the Fed can push real output around for a few quarters, and that it can wreck an economy outright if the error is large enough, but it cannot raise the growth rate of productive capacity by printing money in the long run. Said in another way in the long run the Phillips curve is vertical.

Call that supply-determined growth rate Y*. If the Fed cannot change Y*, and if it has committed itself to 2% inflation, then the Fed’s job reduces to a piece of arithmetic: it has to deliver nominal spending growth of Y* plus 2%. Deliver more than that and you get more than 2% inflation, deliver less and you get less, and that is very nearly the whole of the theory of inflation a central bank actually requires.

Hence we do not need the Phillips curve here, and we do not need an estimate of r*, of u*, of the output gap, or of any of the other unobservable variables the FOMC spends its meetings arguing about. Take nominal GDP growth, subtract trend real growth, and look at what is left over.

That residual is the inflation rate the current stance of monetary policy is delivering, and I will call it implied inflation.

Choosing the starting point

The whole exercise turns on Y*, so let me be open about how I have arrived at it rather than pull a number out of a hat.

I have measured Y* as the average annualised quarterly growth rate of US real GDP from the first quarter of 2022 onwards. Not a Congressional Budget Office projection, not a Fed staff assumption, and not a number chosen because it produced the answer I wanted, but simply what the US economy has in fact delivered over the past four and a half years.

The reason I have chosen this period is that it is post-Covid distruptions. The choice that actually matters is the starting date, and my criterion is that the real economy should be roughly in balance at that point, because a base period taken during a collapse or a rebound measures the cycle rather than the trend.

Take a look at where that criterion leads. Going back to 2019 would mean averaging in the collapse of 2020 and the rebound of 2021, which tells us about lockdowns rather than about productive capacity. Starting in 2023 would mean measuring the supply side over a window short enough that a single strong quarter moves the answer. By the end of 2021 the labour market had normalised, unemployment had returned to something close to the structural rate, and the output gap had closed.

So the first quarter of 2022 is where the real economy stops telling us about the pandemic and starts telling us about the supply side. That gives Y* of 2.4% and therefore an implied NGDP target of 4.4% (2.4% + 2%) a year.

Note that the choice is conservative rather than convenient, because a lower estimate of trend growth would make the Fed look worse than it does here.

US nominal GDP indexed to Q1 of each year against a 4.4% implied target

Each coloured line is one calendar year of US nominal GDP, indexed to 1.00 in the first quarter and carried through to the first quarter of the following year, while the black dashed line is the 4.4% path. The figure in brackets beside each year label is the implied inflation that year’s nominal growth delivers.

Implied inflation came in at 8.9% in 2021, then 5.4% in 2022, then 3.1% in 2023, then 2.2% in 2024, then 3.7% in 2025, and the first half of 2026 is running at 5.5% annualised.

Not one of the twenty-one quarters sits below the line. Not one. So consider what this measure would have been telling the Fed in real time, because the sequence carries more information than any single reading does.

Through 2020 the Fed did the right thing and did it fast, because nominal spending collapsed, the Fed responded aggressively, and the collapse had been reversed inside a year.

From early 2021 the same measure turns from reassuring to alarming, and implied inflation of 8.9% is neither a rounding error nor a data artefact. Furthermore, the Fed could have seen it coming, because markets were pricing it well before the CPI showed it, and it spent that year explaining instead that the pressures were temporary and supply-related.

Then 2022 to 2024 is the stretch the Fed deserves real credit for. Implied inflation falls from 5.4% to 3.1% to 2.2%, which is a genuine disinflation achieved without a recession, and by 2024 nominal spending was within two tenths of a percentage point of the rate consistent with the target.

And then it reversed, with 2025 coming in at 3.7% and the first half of 2026 running at 5.5% annualised. Consider what the Fed was doing while that reversal was under way. It cut the funds rate three times in the autumn of 2025, in September, in October, and again in December, seventy-five basis points in total, into a year in which nominal spending was already accelerating away from the target, and it has held at every meeting since.

Said in another way, the Fed eased into an overshoot it had spent three years correcting.

The explanation everybody reaches for

There is a war in the Middle East, crude oil rose more than 20% in July alone, and the FOMC statement blamed energy prices explicitly. That argument is nonsense, and it is nonsense in a way that ought to be embarrassing for anybody who lived through 2021.

Hence we heard exactly the same argument five years ago, when inflation was supply chains, semiconductors, shipping containers, and pent-up demand for used cars. Every one of those explanations was a claim about relative prices dressed up as a claim about the general price level, and the general price level rose 20% anyway. Milton Friedman’s line survives contact with the evidence rather better than the alternatives do: inflation is always and everywhere a monetary phenomenon.

That being said, supply shocks obviously do move the inflation data from one month to the next, and I would never argue otherwise. What they cannot do is move nominal spending. Oil changes how a given quantity of money chasing goods divides itself between prices and output, and it does not change the quantity.

So don’t tell me that 7.9% nominal GDP growth is an oil story.

The control group is called the euro area

If the US overshoot were the product of an energy shock, then the shock ought to be visible in the other economies exposed to it, because the war in the Middle East is not a US event.

In fact the euro area imports a far larger share of its energy than the US does, so if the oil price were doing the work here, Europe should look worse rather than better.

Euro area nominal GDP indexed to Q1 of each year against a 3.1% implied target

Trend real growth in the euro area since the start of 2022 comes out at 1.1% on exactly the same measure, which gives an implied NGDP target of 3.1%. Implied inflation there runs 9.0, then 6.9, then 3.3, then 2.7, and then 1.7% in 2025, so the 2025 line finishes below the target path with three of its four quarters below it as well.

Take a look at the two starting points before anything else, because in 2021 implied inflation was 8.9% in the US and 9.0% in the euro area. Hence the initial error was not a US error and it was not a European error. It was the same error, made at the same time, on both sides of the Atlantic, and it was enormous.

Anybody who wants to argue that one of these two central banks was prudent in 2021 has not looked at the data.

Implied inflation, US and euro area

Then look at what came afterwards. Through 2024 the two track each other closely, with the euro area consistently a shade higher, and in 2025 the sign flips to 3.7% against 1.7%, while in the first half of 2026 the gap widens to two and a half percentage points.

That is about as close to a controlled experiment as macroeconomics ever gets, with the same shock, the same starting error, and the same global supply conditions producing opposite outcomes. Whatever explains the divergence differs between Washington and Frankfurt, and there is only one serious candidate.

It is not a supply shock. It is US monetary policy that is far too easy.

Credit where it is due

I have spent a lot of the past two decades criticising the European Central Bank and I have not been polite about it, so let me be equally direct in the other direction: the ECB has handled this well.

In June the Governing Council raised rates by 25 basis points, taking the deposit rate to 2.25%, and in doing so it became the first major central bank to tighten in the face of the energy shock. In July it held, unanimously, while making clear that several members had already asked whether a further increase was warranted and that September remains open.

Note that the question everybody in Frankfurt was arguing about is beside the point. Transitory or persistent, second-round effects or not, look through the oil price or do not: in my view none of it matters. Trend real growth in the euro area is somewhere between 1% and 1.5%, which puts the implied NGDP target between 3% and 3.5%, and the job is to keep nominal spending there whatever crude does.

So the transitory debate is not a hard question that the ECB happened to answer correctly. It is a question a central bank aiming at nominal stability never has to answer at all.

Measured against that target the ECB has delivered, because euro area nominal GDP grew 2.8% in 2025, which is at or below the bottom of the range whichever end of it you prefer.

Furthermore, Frankfurt had the harder task, since a target of 3% to 3.5% leaves very little room and the euro area therefore had to squeeze nominal demand considerably harder than the US did to reach the same inflation outcome. It did, and it worked. That European trend growth of barely 1% is dismal remains a damning verdict on European supply-side policy, but it is not a verdict on European monetary policy.

There is an irony here I cannot resist. In 2011 the ECB hiked twice in response to an energy price rise caused by a supply shock and very nearly killed the euro doing it. In 2026 it has tightened citing energy once more and happens this time to be right, not because the reasoning about oil has improved but because nominal spending was where it should be rather than collapsing.

Unfortunately the first half of 2026 is running at something like 4% nominal growth in the euro area on my estimate, so nobody in Frankfurt should be relaxing either. But the difference is one of degree and the degree is large, because an overshoot of half a percentage point is a monetary policy worth watching, while an overshoot of three and a half percentage points is a monetary policy that has lost its anchor altogether.

What happened on Wednesday

The FOMC left the funds rate unchanged at 3.5-3.75% on 29 July, for the fifth consecutive meeting, and three members dissented in favour of a quarter-point increase, which is the most dissents in one direction in a decade.

What followed was a press conference about forward guidance and about whether the chairman should publish his own projection.

Warsh was asked about a monthly consumer price print that had fallen 0.4%, and he answered that it was not much of a consideration for him while calling inflation elevated in the same breath. It is hard not to find Warsh’s performance internally contradictory.

My optimism has cooled sharply

I was optimistic after Warsh’s first press conference, and I said so at the time. I had been sceptical of him for months, because he spent last year campaigning for the job by telling the White House what it wanted to hear about lower rates and an AI-driven productivity boom.

And then his first meeting produced a much shorter statement, no forward guidance at all, and a single blunt sentence committing the Committee to deliver price stability, which is a large part of what I have argued for since I started writing about monetary policy.

I have not totally changed my mind about Warsh. But that optimism has cooled very sharply, and the reason is not the one people assume. The problem is not that he held rates, because holding is a decision like any other and there are meetings where it is the right one.

The problem is that he held rates while nominal spending grows at 7.9%, and then declined to say which number would make him act. A chairman who places credibility above every other virtue has now had two opportunities to tell us what he is targeting, and he has used both of them to explain why he will not.

Of course he is entitled to dislike forward guidance, and in fact I dislike it too. But refusing to pre-commit to a path for the policy rate is one thing and refusing to name the variable you are trying to control is quite another. The first is discretion about instruments. The second is discretion about objectives, and giving that up is what makes a central banker rule-based rather than merely well-spoken.

The pressure is real, and so is the arithmetic

US President Donald Trump has been explicit and public about what he wants from the Fed, including on the day of the meeting itself, when he backed the chairman and called the institution political in the same breath. The debt stock is around 100% of GDP, and the previous chairman faced a criminal referral for refusing to cut.

Unfortunately that configuration has a name. When the fiscal authority moves first and the monetary authority is left to finance whatever gap remains, controlling inflation stops being a question of will and becomes a question of arithmetic, and the arithmetic is unpleasant. So some part of what we are watching is a Fed that would like to tighten and finds that it cannot, and that part will grow rather than shrink as debt service rises.

But this is where I part company with the fatalists. Unpleasant arithmetic is a constraint rather than an excuse, and it binds through expectations, which are exactly what a central bank chairman is paid to manage. A chairman who names his target, explains his reaction function, and then demonstrates that he will act on it makes the fiscal authority’s problem harder and his own easier, while a chairman who names nothing at all makes precisely the opposite trade.

Hence the refusal to give guidance is not a technical preference about communication. It is the surrender of the one instrument that still works.

What I am asking for, and what it costs

I am not asking the Fed to bring nominal GDP back to the pre-2020 path. That path now sits some 15% below the actual level of nominal GDP, and closing the gap would require a contraction severe enough to produce a deep recession and, in all likelihood, a financial crisis alongside it. The mistake of 2021 cannot be undone, because it has been baked into the price level, and pretending otherwise is not toughness. It is vandalism.

What I am asking for is much smaller. Stabilise the growth rate, bring nominal spending back to roughly 4-4.5% a year and keep it there, say so publicly, and then let the market do most of the work of getting you there.

But I should be honest about the cost. Nominal growth has to fall by three and a half percentage points, and a slowdown of that size cannot be promised painless, since some of it comes out of prices and some out of output depending almost entirely on whether the announcement is believed.

Hence the argument for saying it clearly and saying it soon. The longer nominal spending runs at 8%, the more of the eventual adjustment lands on real activity rather than on inflation. And 2024 showed that the target is achievable, because the Fed came within two tenths of a percentage point of it without a recession.

Volcker, Greenspan, Burns and Miller

Let me be precise about what impressed me in June. The emphasis on rules rather than discretion, the scepticism about how far the Fed’s remit had expanded, the refusal to be sorted into either the hawk box or the dove box, and the insistence that money has something to do with monetary policy. That is the language of the Volcker and Greenspan era, and it had been absent from the Eccles Building for a very long time.

But six weeks and one meeting later, the language is still (partly) there and the policy is not.

Volcker’s distinguishing feature was not that he talked about credibility, because every central banker does that. It was that he named a nominal target, accepted an enormous real cost in order to hit it, and did not flinch when the US president’s party lost seats over it. Burns and Miller talked about credibility too, and what they actually did was let nominal spending run while explaining, meeting after meeting, that the inflation was caused by oil, by unions, by food prices, by anything at all except the money they were printing.

So Warsh has now had two meetings. At the first he asked for a good fight and got one, and at the second he got three votes for tightening from his own committee, which he treated as institutional health rather than as a signal that his colleagues think he is late.

Nominal GDP is growing at 7.9%, trend real growth is 2.4%, and the implied inflation embedded in the current stance of US monetary policy is above 5% and rising in a year when real output is growing below trend.

If nominal GDP continues to growth at close to 8% annually then US inflation will soon be heading forward 5% or more.

Warsh, you are way behind the curve. And the longer you refuse to say what you are targeting, the more the comparison stops being Volcker and Greenspan and starts being Burns and Miller.