Testing Both Debt Theses Against the Record: What the Data Supports, What It Cannot Settle, and What It Contradicts (Part 2 of 3)

• The ledger, and the three verdicts it can hand down
• The dataset under the whole case
• The claim that cannot survive without its caveat
• The firmest ground: how bank money is actually made
• The identity that is true, and the inflation claim tested in both directions
• His figures, one at a time
• Mallers: the headline, the mix-up and the interest bill
• The doom loop, the missing figure and the roll
• The long bond, the carry trade and the debasement price
• The ledger closed, and what the third part inherits

Part one set two accounts of the same object against each other and named what would settle the argument between them; this part is that ledger, grading each claim supported, unresolved, or contradicted, with the person left out of it.

DeepDives · 2026 · An AI–human collaboration that weighs every source, mainstream or alternative, under strict rules of truth and logic.

§The ledger, and the three verdicts it can hand down

Part one put Richard Vague, in conversation with Farzin Irani on Connecting the Dots, against Jack Mallers on the Jack Mallers Show. Vague argued that the crisis risk sits in private debt, and that government debt in a country borrowing in its own currency is largely a distraction from it. Mallers argued that the government debt is itself the trap, that the interest bill compounds on a fixed schedule, that the long end of the curve will not clear at the old prices, and that every route out of the problem runs through a weaker dollar.

This part is the ledger part one called for, and the method is deliberately boring. Stock claims are checked against the Treasury’s own debt and interest data, private-borrowing claims against the Bank for International Settlements and its long credit series, ownership claims against the Treasury International Capital series, belief claims against current market prices, and claims about how money is created against the central bank’s own explanation of its system.

Each item is then graded in one of three states. Supported, meaning the number or the mechanism checks out against an outside source. Unresolved, meaning the mechanism is real but the evidence does not settle it, or the figure cannot be reproduced from any source within reach. And contradicted, meaning the outside evidence runs the other way.

§The dataset under the whole case

Start with the dataset, because it is the foundation under Vague’s entire case and it is more checkable than most of what gets said in this argument. On the data site maintained for his book, the crisis audits number forty-three: ten for the United States, nine for the United Kingdom, eight for France, six for Germany, four for China, three for Japan, and three for other countries.

The description on air was looser than the dataset. The phrase used was the six largest economies, a loose way of describing the six countries in the book: the United States, Great Britain, Germany, France, Japan and China. For most of the two hundred years in question, China and Japan were not among the six largest economies.

The abstract says that private-debt over-accumulation does a better job than any other single variable of explaining and predicting crises, and that real-estate lending dominates the run-ups. That is a comparative claim, so it can only be tested against the alternatives, and this is where the sample gets thin: to fault public debt as the explanation, you need episodes in developed economies where public debt ran up hard without private credit doing the same, and the dataset holds very few of those, because the two usually rise together.

The assembly is supported, because a list of forty-three crises, counted and written up, is real work and it is checkable by anyone. The causal claim is unresolved, for a structural reason: a dataset that mostly contains the pattern you are testing for will mostly confirm the pattern you are testing for, and that is a limit on what the work can carry.

§The claim that cannot survive without its caveat

Rising government debt has not once caused a financial crisis in a developed economy. It is not quite the sentence he said, because he attached a condition in the same passage, and the condition is doing all of the lifting: the sovereign has to issue its debt in a currency it controls.

The euro-area crisis from 2010 to 2012 is that condition being tested in public. Greece, Ireland, Portugal, Spain and Italy each faced some form of sovereign funding crisis, and Greece restructured its debt in 2012. That is a sovereign debt crisis in developed economies, inside the post-war period, and it is why the caveat exists: those countries had borrowed in a currency they did not issue and could not create at will, which put them in the position Vague reserves for developing countries.

Write the claim down without the caveat and the euro-area crisis falsifies it; write it down with the caveat and it survives, but it has become a narrower statement about monetary sovereignty rather than a statement about debt levels. The grade follows: supported, but only with the caveat doing the work, and the version without it is contradicted by a well-documented episode in countries usually called developed.

§The firmest ground: how bank money is actually made

His account of how bank money is made is not a heterodox flourish; it is the mechanism as described by a central bank. In the first quarter of 2014, the Bank of England’s Quarterly Bulletin carried an article called Money creation in the modern economy, by McLeay, Radia and Thomas. Its finding is unambiguous: the majority of money in the modern economy is created by commercial banks making loans, and banks do not act as intermediaries lending out deposits placed with them.

That is exactly the mechanism Vague walks through on air with a loan of ten thousand dollars, booked as an asset to the bank and a matching deposit as a liability to the borrower, so the deposit is new money in the system.

He attributes roughly ninety per cent of the growth in deposits to private banks lending to households and businesses, and roughly ten per cent to the Federal Reserve. The mechanism is one thing, but the attribution of shares across a twenty-five year period is his own measurement rather than a central-bank finding, and in a system where reserves are abundant the central-bank half runs through reserves, the interest paid on them, and the portfolio choices of the institutions that hold them.

§The identity that is true, and the inflation claim tested in both directions

Next his accounting argument, which is the one about buyers. His version is that the sale of government debt does not deplete the money supply, because the proceeds are spent in the same instant, so there will be no shortage of buyers. Money supply of twenty-one trillion dollars, a trillion raised, a trillion spent, and the total is back where it started. Behind that sits a real identity: when the government runs a deficit, some other part of the economy is running a surplus of the same size.

But an identity is a statement about quantities, and it is silent about two things that matter more. It says nothing about the price at which the debt changes hands, because a market can clear with no shortage of buyers at a yield of three per cent or at six per cent. It also says nothing about the term structure, about whether the buyers want thirty-year paper or only the bill at the front of the curve.

So the grade is precise: supported as accounting, and silent on the two things under dispute. Vague is right that there is no mechanical shortage of buyers in the aggregate sense, but the claim Mallers is making is not about whether buyers exist in the aggregate; it is about what rate they will demand and for how long they will lend, and the identity cannot answer that question.

Then the inflation claim, which needs testing in both directions because it is stated in both. His claim is that rising government debt does not produce inflation, and if anything the relationship runs the other way. The decade after 2008 is the case he reaches for, and it is a genuinely good case: government debt rose sharply, the central bank expanded its balance sheet enormously, and measured consumer inflation stayed low for years.

The pandemic contradicts the generalisation in the other direction. Federal debt as a share of GDP went from one hundred and six point eight per cent in the first quarter of 2020 to one hundred and twenty-four point one per cent in the third quarter. Consumer inflation then followed it up to about nine per cent in 2022. Debt exploded and inflation exploded, in the same window, in the same economy.

So the generalisation is contradicted, but the narrower claim underneath it survives both cases: there is no stable mechanical link in either direction. Two episodes with large increases in public debt, one followed by low inflation and one followed by high inflation, is exactly what you would expect if the link depended on something else. That puts him closer to Mallers than the volume suggests, because both guests end up saying there is no reliable mechanical relationship between the debt stock and the price level, and they differ only on what fills the gap: Mallers fills it with the currency, and Vague fills it with private credit.

§His figures, one at a time

He gives one hundred and forty per cent of GDP for the United States. The Bank for International Settlements series for total credit to the private non-financial sector puts the United States at one hundred and forty point three per cent in the fourth quarter of 2025, which is as close to exact as a spoken number gets.

The peak he gives does not survive the same test, and this is the first item that fails. He says United States private debt peaked at one hundred and fifty-three per cent of GDP. That same series peaks at one hundred and seventy-two per cent, in the third quarter of 2008, a gap of nineteen percentage points that is not a rounding matter. He may be using a different definition of private credit or a different vintage of the data, but the peak he states could not be reproduced.

Then household wealth, where he is out in a direction that does not hurt his argument. He says United States household wealth is about one hundred and eighty trillion dollars. Household and nonprofit net worth stood at one hundred and ninety-five point eight seven trillion dollars in the second quarter of 2026, and the equivalent concept on the older financial accounts basis gives one hundred and eighty-five point six five trillion. So one hundred and eighty trillion understates the sector by between five and fifteen trillion dollars, depending on which series you prefer, but his point that household wealth dwarfs the government’s forty trillion holds on either number.

His foreign holdings figures are both out, in neat opposite directions. He says Japan holds about nine hundred billion dollars of United States debt and China about eight hundred billion. The Treasury International Capital data for December 2025 put Japan at one thousand one hundred and eighty-five point five billion dollars and mainland China at six hundred and eighty-three point five billion, so the Japanese figure is low by roughly a quarter and the Chinese figure high by roughly a sixth. The conclusion he draws survives the correction, since both are small against a stock of forty trillion, but the numbers themselves are not right.

One more item from his toolkit is a different kind of thing entirely. His warning rule is that growth in the debt-to-GDP ratio of any category above fifteen per cent a year should compel regulators to intervene, but that is not a finding; it is a proposal he would impose on the system, and it has not been run against the crisis list the way his other claims have.

§Mallers: the headline, the mix-up and the interest bill

Then Mallers, on exactly the same terms, starting with the number in his title. The forty trillion is right, and the Treasury’s own count on the twenty-eighth of September 2026 puts total public debt outstanding at forty trillion, one hundred and two billion dollars. Gross debt is about one hundred and twenty-three per cent of GDP, so the headline figure is sound, and it is a gross figure rather than a net one, which matters for the next item.

Because his one hundred and twenty-five per cent mixes two different things inside one sentence. He says the debt is forty trillion against an economy of thirty-one to thirty-two trillion, and then in the same passage speaks about the debt owned by outside lenders as though it were the same quantity. Debt held by the public is thirty-two point four trillion dollars, against an economy of thirty-two point five trillion, which is about one hundred per cent, so he runs two true figures together and inflates the ratio he is defending.

The trillion-dollar interest bill, by contrast, is right on the Treasury’s own arithmetic. Interest expense on public issues reached nine hundred and eighty-one point eight billion dollars in fiscal 2026 through the end of August, eleven months of the fiscal year, and the full year runs past one trillion; fiscal 2025 came in at nine hundred and seventy billion. When he says the interest bill is about a trillion dollars a year, he is not exaggerating and he is not early, because the bill is already on the books.

§The doom loop, the missing figure and the roll

The doom loop needs a different kind of test, because it is a mechanism rather than a figure, and mechanisms are checked against the record. As stated, the loop runs like this: the central bank hikes, the interest bill rises, the deficit widens, more bonds are issued, yields rise, the central bank hikes again. Each step follows from the one before it, and the arithmetic is coherent, so the question is not whether the loop could turn but whether it has been turning, and for how long, and at what pace.

This is where the policy record makes the loop a forecast rather than a description. The federal funds effective rate was three point eight eight per cent on the twenty-eighth of September 2026, up from three point six three per cent in the middle of the month, which is a single hike of twenty-five basis points and matches his own account of a hike the week before. But a year earlier the rate was four point three three per cent, which means the policy rate was cut through 2025 and has been raised once since.

A policy rate of three point eight eight per cent sits far below the long end of the curve, with a ten-year at five point two four per cent and a thirty-year at five point five six, so the short rate is not dragging the long rate up; the long rate is higher on its own account. A loop driven by policy hikes needs the policy rate to be the leading force, and right now it is the lagging one.

Then the item that could not be confirmed at all. He says the United States must refinance eight point three trillion dollars of debt this year, and no Treasury source within reach states that figure. The bill discount in fiscal 2026 implies a bill stock of roughly six trillion dollars, and something like a fifth to a quarter of marketable debt matures within a year, so the order of magnitude is plausible. But plausible is not verified, so it is marked as unconfirmed rather than as wrong.

Now the strongest single claim either guest makes. It concerns the difference between the average cost of the debt already issued and the cost of the debt being issued now. The weighted average interest rate on all marketable Treasury debt was three point four seven five per cent at the end of August 2026. The marginal ten-year was five point two four per cent, and the marginal thirty-year was five point five six per cent, so every refinancing happens at a rate far above the rate it replaces.

Set out what that does mechanically. The stock of debt pays an average of about three and a half per cent, while the money coming in to replace maturing debt costs about five and a half per cent at the long end. Each roll lifts the average a little, and the lift is one-directional for as long as the marginal rate sits above the average rate. No vote is required, no crisis is required, and no forecast about anyone’s behaviour is required; it is a slow transfer from the borrower to the lender, executed by the passage of maturities.

The average rate has been rising because the marginal rate has been above it, and the size of that gap is the source of the movement in the interest bill, which is why the interest bill can keep climbing even if the deficit itself stops growing. It is subtraction.

So the grade on that one is supported, and it is the best quantitative argument in either episode.

§The long bond, the carry trade and the debasement price

His second-strongest claim sits right next to it, and it is a claim about the market. He says the government has struggled to sell long-duration debt, so issuance has floated to the front of the curve and the debt is therefore refinanced constantly at market rates. Two things are being claimed: that long bonds cannot be sold at acceptable prices, and that the mix of what is issued has shifted towards bills and short paper.

The auctions test the first claim directly, and they are on the public record. The thirty-year bond auctioned on the tenth of September 2026: twenty-two billion dollars offered, a bid-to-cover of two point six one, and a high yield of five point three one per cent. The twenty-year auctioned on the fifteenth of September cleared at a bid-to-cover of two point five seven, and the ten-year on the ninth of September cleared at two point seven one, which are healthy coverage ratios. Those bonds sold in size and at a price.

The thirty-year yield is five point five six per cent, the highest since 2004, and the ten-year at five point two four is the highest since 2007. Watch a single maturity: the seven-year auctioned at four point five one two per cent on the twenty-seventh of August and at five point zero eight five per cent on the twenty-fourth of September, fifty-seven basis points of repricing in a month on one maturity, which is a market doing its work.

Which gives the grade: directionally supported, literally too strong. Long-dated debt is being repriced hard, the curve has steepened, and the cost of term is rising for the government every month, but the claim that the government cannot sell long-dated debt is too strong, because a thirty-year bond cleared at more than two and a half times coverage in September. And of the two claims inside the sentence, the price claim is the one that matters, because the price is the variable that feeds the roll.

Then the carry trade, his version of the argument that the dollar system is fragile from the other side. The figures he uses are that the rest of the world owes fourteen trillion dollars in dollar-denominated debt and owns twenty-two trillion dollars of United States assets, a little under half of that being government debt. On that picture, a stronger dollar makes foreign obligations heavier in local currency, and holders who are squeezed have to sell assets to meet those obligations.

What that does imply is worth stating carefully, because it is usually left out of the comfortable version of the reserve-currency argument: a sharp strengthening of the dollar tightens foreign balance sheets in local terms, and a balance sheet under pressure is forced to sell. A system with that structure is fragile to dollar strength, not only to dollar weakness, and anyone who has only considered the second direction has considered half of the exposure.

What it does not imply is the conclusion that gets attached to it, which is that a stronger dollar leaves the United States unable to fund itself or means default. A forced seller is a seller at a price, and the price is what attracts the marginal buyer in, so the mechanism describes volatility and dislocation in the holdings of the debt rather than an end state for the debt itself.

So the grade is unresolved, in two parts: the mechanism is real and worth carrying forward into the third part, and the conclusion is overstated. Notice also the shape of the claim, because it says the system can be broken by a stronger dollar while the guest’s own forecast is a weaker dollar, and both cannot be the operative risk at once. What he has actually described is a system fragile in both directions, which is a larger and more interesting claim than the one he is making.

Then the price test, which is the cleanest in the whole ledger because it involves almost no interpretation. The claim under test is that the market is pricing debasement, that investors expect the purchasing power of the dollar to erode and are positioning themselves for it. If that were true, two things would be visible in prices at once: breakeven inflation, the market’s own estimate of future inflation embedded in bond prices, would be rising, and real yields, the return after expected inflation, would be falling.

On the twenty-eighth of September 2026, the ten-year breakeven inflation rate was two point three four per cent, and the five-year five-year forward breakeven, the market’s estimate of inflation over the second half of the next decade, was two point three six per cent. Neither number is a debasement number; both sit close to what a well-anchored expectation looks like, and both are consistent with the central bank’s target zone rather than well above it.

And the real yield did the opposite of what the debasement story requires. The ten-year real yield was two point nine per cent, up from one point eight four per cent in October 2025, so real yields rose by more than a percentage point across the very period when the debasement trade was most widely discussed. If the market were pricing the erosion of the dollar’s purchasing power, real yields would be falling and breakevens rising. Two honest caveats belong with the grade: prices can be wrong, and these figures measure expectations rather than outcomes, but the claim under test was about what the market is pricing, and on that claim the answer is no.

§The ledger closed, and what the third part inherits

Now close the ledger, item by item. Supported: the money-creation mechanism, documented in the central bank’s own Quarterly Bulletin; the private-debt level of one hundred and forty per cent, exact against the Bank for International Settlements; the forty-trillion headline figure, which is the gross measure; the trillion-dollar interest bill, already running; and the roll, the spread between the average rate on the stock and the marginal rate on new money, the strongest single item on either side. Also supported, and only as accounting, the identity that a government deficit is somebody else’s surplus.

Unresolved, and this is the longest list. The causal claim at the heart of Vague’s dataset, because the sample holds too few run-ups in public debt without run-ups in private credit. The claim that rising government debt has not caused a crisis in a developed economy, unresolved as it was stated on air, because it is defensible only with the monetary-sovereignty caveat attached and fails without it. The carry-trade mechanism, real in its mechanics and overstated in its conclusion. Any claim that public borrowing has a fixed relationship to inflation, in either direction. And Vague’s fifteen per cent warning threshold, which is a proposal rather than a finding.

Contradicted: the one hundred and fifty-three per cent private-debt peak, which could not be reproduced on the series that instead gives one hundred and seventy-two per cent. The claim that government debt and inflation move together, or apart, in a fixed way, once the pandemic and the post-crisis decade are both held in view. And the debasement price, which the market is simply not paying, with breakevens near two and a third per cent and real yields rising rather than falling.

Then the numbers that could not be confirmed at all, which deserve their own line because being unverified is not the same as being contradicted. The eight point three trillion dollars of refinancing this year, plausible in order of magnitude and unsupported by any Treasury source within reach. The claim that only about one hundred billion of Asia’s one point eight trillion dollar surplus is going into long-term United States bonds, where the stock data point the same way but the flow figure could not be reproduced. The projection of interest at four and a quarter per cent of GDP in 2027, which sits above recent official net-interest projections. And the share of household wealth held by people over seventy, which was not checked here and should not be repeated as a fact.

What the ledger shows is not a winner. Vague has the mechanism of money creation, documented by a central bank, and a dataset with a sampling problem he does not address; Mallers has the arithmetic of the roll, which is close to unarguable, and a set of forecasts about the policy rate, the term structure and the dollar that the record has not yet caught up with. Where they overlap, they agree more than either would like to admit.

That is where the checking stops, and the third part is where the synthesis belongs. Part one established the collision between the two accounts, and this part has itemised the claims and put a grade against each; what remains is the harder work of asking what the two accounts look like when they are held together rather than set against each other.

This has been a deep dive into whether the evidence supports either thesis.

Sources: this examination names its sources inline — Richard Vague in conversation with Farzin Irani on Connecting the Dots, and Jack Mallers on the Jack Mallers Show, both weighed as claims rather than as persons; the crisis dataset maintained for Vague’s book; the Bank of England’s Quarterly Bulletin article Money creation in the modern economy, by McLeay, Radia and Thomas; the Bank for International Settlements series on total credit to the private non-financial sector; the Treasury International Capital series; the Treasury’s published debt, interest and auction data; and the market prices for the federal funds effective rate, breakeven inflation and real yields. Every source is weighed with the same skepticism regardless of politics or business model. The audio version of this article, read by DeepDives, is available on Wavlake.