Joel Litman on Debt, the Dollar and Gold, Compared With Our Own Reading

• The guest, the venue, and what this episode sets out to do
• His system, stated as a system
• What our own credit check found, and the tail it does not price
• The valuation argument, and the denominator that cuts both ways
• Where he improves on us: the canary and the long end
• The artificial intelligence buildout, and where he locates the failure
• Forty trillion dollars, the electrician, and the corrections
• The inflation claim, gold, and the three ways of measuring the dollar
• His case at its best, and the sentence that cannot fail
• The verdict, the markers, and the close

A guest whose firm restates four decades of company accounts says the credit market’s silence means no crash and no dying dollar, and this examination holds that claim against its own published reading — granting him the instrument on which he improves us, and refusing the step where a calm average is asked to carry the weight of a conclusion.

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

§The guest, the venue, and what this episode sets out to do

This examination holds a guest’s argument against our own published reading, without pretending the two agree. The argument belongs to Joel Litman, chief investment strategist at Altimetry and president and chief executive of Valens Research. He put it to Michelle Makori, president and editor in chief of Miles Franklin Media, in an interview published on the tenth of September, 2026, under a headline asking whether a crash was coming and whether the dollar was dying. His answer is that the data says no, against a month in which this series has described forty trillion dollars of debt, interest above a trillion a year, and a Treasury buying back the long bond with money raised at the short end.

Litman is an accountant by training who has spent decades on what a company’s earnings actually are. Valens Research produces the uniform accounting dataset behind almost everything he says: thirty-two thousand companies, more than forty years of restated numbers, on a claim that reported figures understate the underlying cash generation by about half. Altimetry sells subscription research built on that dataset, and Miles Franklin Media is the media arm of Miles Franklin Precious Metals, a dealer whose advertisement the host reads at the mid-roll.

The incentives point two ways. The venue’s interest runs toward a gold bid; the guest’s runs toward equity subscriptions, since a bullish call to a large audience is the top of his funnel. This series has the opposite problem: it holds the assets it discusses and has said on air that the trade can be right about the world and lose money anyway.

§His system, stated as a system

The claim is narrower than the headline: he is not forecasting that nothing will go wrong. He is saying the thing which precedes trouble is visible months before the equity market sees it, and that it is not visible now. The instrument is corporate credit, the terms on which companies borrow and the price buyers charge for that risk.

His framework has three steps. Every sustained American bear market was preceded by a corporate credit crisis, without exception. The lead time is long enough to act on. And credit conditions today are the most benign he has measured in decades.

He dates the pattern to 1929, to the 1970s and to 2008. The lead time arrives as six to eighteen months in one passage, about sixteen to eighteen in another, and as a garbled six-and-sixteen in a third, and a six-month floor and a sixteen-month floor imply different things about the next year.

The evidence that the canary is quiet: companies are issuing debt at phenomenal rates. The credit spread, the extra yield a corporate borrower pays over a matching government bond, is very small; issuers pay about fifty to sixty basis points over that rate, a basis point being a hundredth of a percentage point. And the market he trusts most is credit default swaps, where the price of protection against a company failing is, he says, unbelievably low.

§What our own credit check found, and the tail it does not price

The check we ran on investment-grade credit ahead of this episode found single-A corporate spreads at sixty-six basis points against a long-run average of a hundred and twenty-four, so on the safest corporate borrowers he is right: the compensation for risk is half its normal level.

What a thin spread means is where the press belongs. Being paid fifty to sixty basis points over a risk-free bond to take a company’s credit risk is not a description of a safe world; it is a market in which the compensation for bearing risk has been competed away, and the worst point in a cycle at which to begin one.

His fair reply is that a tight spread also reflects actual default experience and the earnings servicing the debt, and defaults have been low. A tight spread is not a lie; it is a price, and prices can be right. But a price speaks in the present tense, and his claim is about the next eighteen months.

Then the single point we would put to him. High-yield credit spreads stand at two hundred and seventy-six basis points, the extra yield the weakest borrowers pay over government debt,, the twentieth percentile of history; triple-C spreads stand at one thousand and eighty-five basis points, the ninety-ninth percentile. The average borrower is priced as though nothing can go wrong, and the weakest is priced as though something already has.

He is right about the average: if his instrument is the broad high-yield index it is near the calmest reading in a generation. But a credit event does not arrive through the average company; it arrives through the marginal borrower, whose paper has to be refinanced into a market that has repriced, and the ninety-ninth percentile is the market saying out loud that it has begun to discriminate. The objection is not that he is misreading credit, but that he is reading the average of an index whose average is doing the work of a conclusion while the tail does the work of a warning.

§The valuation argument, and the denominator that cuts both ways

His firm restates accounts so earnings are comparable across decades, arguing accounting conservatism has cut how representative reported earnings are by about half against twenty or thirty years ago. The charges: mergers, where the 1980s allowed no goodwill amortisation while today carries impairment write-offs, and research and development, expensed here and capitalised almost everywhere else.

If the denominator is understated by half, a multiple of thirty-nine times is nearer nineteen or twenty. So the long-run adjusted measure that carries Shiller’s name, near forty point six times, becomes in his reading a measurement artefact rather than a valuation warning. He makes the same move on the Buffett indicator, which was never Buffett’s own, and adds that the index is a global earnings stream measured against domestic output, with more than forty per cent of revenue earned abroad.

Two of those points we concede. The accounting critique is correct in direction, and we have said the same of reported earnings and of the index itself. His point about 2020 is the sharper version: small businesses failed in numbers while the index finished up about eleven per cent.

Where we part is what follows. A denominator disputed by half cuts both ways, and if it does, the multiple he offers as reassurance cannot be reproduced outside his firm, while the multiple the market prices can be read by anybody.

§Where he improves on us: the canary and the long end

Two places where he improves on us. First, credit. Our published dashboard ranks fifteen stressors out of a hundred and twenty-five, and we have never carried the corporate credit market itself as a live leading indicator the way he does. He is right that the price of insuring against default moves before the equity market does, and right that in 2008 the swaps were widening in the closing months of 2007 while the index was still near its high.

Second, he refines our reading of the long end. We have described the long bond as being in a buyers’ strike since late June, meaning the auctions clear only at yields buyers resent, with the Treasury as buyer of last resort. He offers a supply explanation instead, pointing to a hundred-year corporate issue and a fifty-odd billion dollar borrowing programme, and says the long end is heavy because of the volume of long-dated paper rather than because foreign holders have walked.

§The artificial intelligence buildout, and where he locates the failure

The artificial intelligence question is where his case and ours come closest to being testable. His argument is that the circular financing everyone points at is old news: chip companies took equity in their own customers decades ago, that is how the industry built the equipment makers, and what it produces is a small number of firms that keep competitors out. He points to revenue at the model developers, and to a teddy bear retailer whose shares rose about eighty per cent, as proof the productivity shows up outside the technology sector. He concedes the objection before it is put: he agrees the railroad comparison is a real concern, and says companies overbuild and the crash comes after that.

Where he locates the failure is the tension worth naming. He identifies the weakest link as a software and cloud company that is a heavy borrower for data centres, says he would not own it, that it sits one notch above junk, and that its slice of the supply chain is the least profitable. The rating reference checks out: S&P Global Ratings cut that company to triple B minus on the ninth of July, 2026. Then he caps the exposure at one company out of a hundred, and the largest capital spending cycle in the market becomes a single containable name.

There is an internal problem here. If the buildout is financed by a handful of near-junk borrowers, and he says himself the agencies merely repeat what the swap market already says, then his own instrument is the one that would tell us, and he is asking us to relax on the strength of a market whose compensation for risk he has just called the lowest in a generation. He treats ratings as closet copies of the swap market at one moment and uses a rating as his warning at the next.

§Forty trillion dollars, the electrician, and the corrections

Then the debt, forty trillion dollars of it, and the claim that makes him comfortable. He does not say the arithmetic is fine. He says Congress has been irresponsible and the promises made for pensions and health care are astronomical, and his claim is about capacity.

Federal revenue this year is around five point six trillion dollars; five to seven years ago it was three point six trillion, so the revenue base has grown by about two trillion. At a five per cent interest rate, two trillion of new borrowing costs a little under a hundred billion dollars a year in additional interest, so revenue growth of that size covers the service cost of the new debt.

He then scales the country down by ten million: an electrician whose after-tax pay is five hundred and sixty thousand dollars a year, or forty-six thousand a month, holding ten million dollars of assets. Multiply by ten million and you get four hundred and sixty-six billion dollars of monthly tax revenue, ninety-three billion of monthly debt service, and federal assets of a hundred trillion dollars at his low end and four hundred trillion at his high end, with gold on the books at forty-two dollars an ounce. His conclusion: more than enough money to service forty trillion dollars of debt, and a real problem at least ten and possibly twenty years away.

Now the corrections. First, the corporate tax claim: he says corporations now pay close to triple what they paid in 2018. The direction is right, the magnitude is overstated. Receipts on corporate income in 2018 came off a depressed base, the first year of the rate cut, and the most recent annual figure we can locate is about four hundred and ninety-seven billion dollars against about four hundred and ninety-two billion the year before, on a base near two hundred and five billion dollars: roughly two point four times, not three.

Second, the revenue base. Five point six trillion dollars this year is consistent with the budget office’s February 2026 outlook, but the three point six trillion he places five to seven years back is closer to the 2018 fiscal year, when revenue was about three point three trillion dollars. His two trillion dollar delta survives even though the year attached to it does not.

Third, an inconsistency in his own description of the dataset: he says thirty-two thousand companies in one passage and coverage of well over seventy thousand later in the same interview, and the two are never reconciled. The federal assets of a hundred to four hundred trillion dollars are not usable as evidence either, because they value land, mineral and spectrum rights without a source. And on the interest bill, the budget office’s net figure is about a trillion dollars in 2026 rising to two point one trillion by 2036, against the one point two five trillion a year we have put on air; his ninety-three billion a month sits between them, the gap being largely gross versus net.

Which leaves the fiscal difference between us. Ours is that the binding constraint is not the interest bill but the total claim on receipts: claims against revenue run at about one hundred and five per cent of federal receipts and are growing faster than receipts. His case is about the increment rather than the stock, and about the difference between servicing new debt and honouring every promise already made. On the second he concedes the field, because his answer to the pension and health care liabilities is to discount the future revenue as well and compare the two, which only works if the revenue growth is durable. That is the load-bearing assumption, and he supports it by noting the budget office has repeatedly under-forecast receipts, though he is not entitled to convert a habit of under-forecasting into permanent growth. And he agrees a breaking point comes; he disagrees only about when.

§The inflation claim, gold, and the three ways of measuring the dollar

Inflation first, because it feeds the gold case. He argues the headline measures overstate inflation by one to two hundred basis points, that real inflation is closer to two per cent, and he prefers an independent index tracking a very large number of items, reading near one point nine per cent. We checked that index and it will not carry the weight he puts on it: it printed about one point eight seven per cent in January 2026 and about zero point six eight per cent on the eighth of February, a swing of a hundred and twenty basis points in six weeks, and its own July 2026 report forecasts the official measure rising to three point six per cent over the year. His read of the central bank matches ours: a hawk who holds rather than raises, targeting two per cent, with the energy shock through the strait as the swing factor.

Gold is the part of the interview the venue most wants. His phrase is Goldilocks for gold and stocks, and his core claim is that you do not need the dollar to crash in order to love the metal. His bull case is other people’s weakness: property in China and the inability to sell what is built, Russia running a wartime economy that produces destruction, and German mid-sized manufacturers crushed by the price of electricity. Central banks and sovereigns keep buying on that list, and he floats the possibility that the United States itself might buy more. He frames price levels at four to five thousand dollars and asks what takes gold to seven thousand, and his answer is the supply and demand asymmetry: demand rising about five per cent against supply that can grow only about two.

Three things about that, and the third is a correction of a number rather than of a view. First, he is describing a genuine structural bid, and our own published data agrees with the direction: official-sector purchases of two hundred and eighty-nine tonnes in the second quarter, bought while the price was falling, which is the behaviour of a buyer not buying for a return. Second, our own record does not agree with itself on that tonnage: one of our episodes calls a seventy-four per cent increase against the year before the highest quarterly purchase on record, a later one calls the same two hundred and eighty-nine tonnes a sixty-two per cent increase and the strongest second quarter, and neither is settled.

Third, the supply and demand numbers he uses. Our published figure for total first-half demand is two thousand five hundred and twenty-two tonnes, up two per cent on the year, so his five per cent figure needs a base and a period the interview does not supply. And his price framing needs its own correction: he speaks of four to five thousand dollars as a destination, when our record has gold above five thousand four hundred dollars an ounce in late January and near four thousand four hundred later, a fall of roughly nineteen per cent.

Then de-dollarisation. His formulation is that the world is dollarising rather than de-dollarising and that the idea does not hold water: nobody in China would rather hold the local unit than dollars, and the caution adversaries show toward American government debt is about the risk that the United States refuses to pay them, not a rejection of the dollar. He has called the currency bloc idea ridiculous for three years and considers it more ridiculous now, because the members do not trust each other enough to hand any of them the issuance power, and because any holder of such a unit would trade it for dollars to buy American assets. His positive case is that the world’s asset pools are dollar pools, which is why the American equity market is bigger than the rest of the world put together.

He is right about the mechanism of substitution: a reserve currency is not replaced by a decree or a summit but by a deep market in assets people want to own, and on that test nothing currently competes with the American capital market. His point about sanction risk is ours too. So the difference is what is being measured. The dollar’s share of reserve holdings, its share of trade invoicing and settlement, and the size of dollar-denominated asset pools are three separate quantities: he is arguing, correctly, about the second and third, and both are stable, while our published position is about the first. The reserve share is at a thirty-one-year low, and the erosion is the quiet kind, measurable in decades rather than months, driven not by a rival currency but by the fiscal arithmetic. We have also put on air that the weaponisation of the dollar and the freezing of reserves are recruiting posters for alternatives. He explains adversary gold buying purely as a consequence of weak local currencies. It can be both, and treating the second explanation as vindication of the first is where his argument runs faster than its evidence.

§His case at its best, and the sentence that cannot fail

His case at its best is stronger than a headline reading allows. Valuation has no timing content: anyone who sold equities because the long-run adjusted multiple was high has been wrong for years. What does have timing content is credit, because a company that cannot refinance must do something on a schedule — a maturity falls due, a covenant is tested, a rating action follows — and that makes credit a dated instrument in a way a ratio is not.

Credit widened through the back half of 2007 while the index was near its peak, and the index was down only a few per cent in the opening months of 2008, which is when he stood in front of an audience and said so. In the late 1990s, venture financing dried up before the equity market peaked in April 2000. The mechanism is a sequence with dates in it, and he is honest about the horizon he claims: he is not making a five-year prediction, he is claiming a lead time.

His off-ramps are dated too: he promises to be the first to say so if the numbers change inside six months, and to call a bear market if he measures a real credit breakdown inside nine months. He names seven to ten per cent as the yield range that would be a problem, on the precedent of the 1970s, and says five and a quarter per cent is not it.

Then the falsification problem, because the honest version of his claim is narrower than the interview’s framing. His lead time means he will get six to eighteen months of notice: if credit breaks down and equities fall, he was right; if credit breaks down and equities rise for a while first, he was right; if credit does not break down, the question has not yet been asked. There is no state of the world in which that sentence is false, and that is the definition of a claim that is not a forecast.

His two bearish triggers share that property. One is a genuine multiple of forty times with earnings growth at the economy’s rate, and the public measure already sits near forty, so the escaping move is to say the multiple is not really forty, which only his own model can adjudicate. The other is that the numbers change inside six months, without specifying which numbers or at what threshold. His 2008 template is also less reassuring than he presents it: by his account the index was down only four to six per cent in January and February of that year, which left the signal’s followers holding through the first half of a very bad year.

§The verdict, the markers, and the close

The verdict. He is right on the mechanism and he improves us on one instrument. Credit does lead equities, and this series has built its picture of the coming transition around private credit and the financing of the buildout without ever carrying the corporate swap market as a live dial. He also improves us on valuation: his point that the denominator is unstable across decades is a real limit on how much weight our long-run adjusted multiple can carry, even though his replacement for it is unauditable.

Where he goes wrong is by sequence rather than direction. He treats a quiet credit market as evidence that the fiscal story does not bind, and the tail we have described is why that step does not follow: the fiscal story binds slowly, in the currency and in real yields, and does not require a credit event to operate. So the synthesis is neither his outcome nor ours. Barring a decision to default, the likely path is not a credit collapse and not a dollar triumph but a long relative erosion: of what the dollar buys, of where reserves sit, and of the compensation demanded for lending long to the Treasury.

Then the markers, dated, so this can be settled rather than continued. His position is confirmed if over the next four quarters high-yield spreads stay below three hundred basis points while the equity market makes further highs on earnings growth rather than on multiple expansion, in which case our emphasis on the fiscal constraint has been a distraction for another year. If the long bond holds below five per cent without the Treasury widening its buyback programme again, his supply explanation beats our buyers’ strike reading. If official-sector gold purchases fall below seven hundred and fifty tonnes for the year, his metal case is early rather than wrong.

And the other side. Our reading gains if high-yield spreads widen past four hundred basis points without a recession, and faster if triple-C spreads stay near the ninety-ninth percentile while the composite stays calm. It gains if the ten-year inflation-protected yield falls back below one per cent while the debt ratio keeps rising, if the broad dollar index breaks below one hundred, or if the central bank buys long-dated debt in size; that yield went from one point eight eight to two point six two per cent this year, and he does not discuss it. And where our own record conflicts, we leave it open: one of our episodes describes a balance sheet still being talked about as shrinking, with no long bond bought, and a later one says it has stopped shrinking and is expanding again, up about two per cent to six point seven four trillion dollars.

One closing thought. His paywall boundary is stated on air: he names three large caps, the chip equipment maker, the power equipment maker and the defence contractor, then says his favourite ideas sit in the small and micro cap space and that you will have to subscribe. That is not a hidden interest, it is a disclosed one, and it points toward publishing a bullish call to a large audience, which is exactly what he did. The venue’s interest points toward gold, and the venue got a gold case, but not the one it advertises: he does not give it a dollar collapse, he gives it a relative case built on everybody else’s weaknesses. The conclusion that an investor should hold American equities and gold costs neither party anything, and it serves the guest’s funnel and the host’s dealer at the same time. That does not make the argument wrong. It does mean that when a thesis is good for the person selling research and the person selling metal at the same time, the burden of proof should sit on the falsification conditions rather than on the tone. The average is calm and the tail is not, and the difference between those two facts is the whole argument.

Sources: this examination names its sources inline — the interview between Joel Litman of Altimetry and Valens Research and Michelle Makori of Miles Franklin Media, published on the tenth of September, 2026; the uniform accounting dataset of Valens Research and the subscription research of Altimetry; the rating action of S&P Global Ratings of the ninth of July, 2026; the budget office’s February 2026 outlook and its interest projections; the independent inflation index he cites, checked against its recent prints and its July 2026 report; this series’ own published stressor scores and credit spread readings, its official-sector gold data, and its earlier episodes on the debasement trade and the fiscal condition; and the public market records for high-yield and triple-C spreads, for the gold price and for reserve composition. Every source is weighed with the same skepticism regardless of politics, business model or holdings. The audio version of this article, read by DeepDives, is available on Wavlake.