Jackson Hole 2026: Warsh's "Work to Do" Took September Odds to 57.5% — the 2-Year Closed +14bp and the Dollar Had Its Best Week in Five

Kevin Warsh spent most of his first Jackson Hole keynote as Fed chairman arguing that central banks should stop telling markets what they are going to do next. Then, in the last section, he said the Fed's "predominant focus right now should be on prices" and that if underlying inflation is not moving to target "clearly and at sufficient speed", then "we have work to do". Fed funds futures moved from a 35.4% implied probability of a September increase on Thursday to 57.5% on Friday. By the close the 2-year Treasury had risen 14 basis points to 4.34%, the 30-year 3 basis points to 5.22%, and the dollar index was up 0.35% at 99.46 — its biggest weekly gain in five weeks.

This page was published on 23 August as a preview, updated on 26 August with the July core PCE print, and argued that the fork Warsh himself described on 29 July — a "big questions" address or a conventional autumn set-up — had not been closed by the data. The answer, delivered at 10:00 a.m. Eastern on 28 August, was both at once, in that order, and the ordering turned out to matter more than the content.

This update, on 29 August, replaces the intraday levels this page carried on Friday with the official closing figures — and corrects one of them. The midday tape showed the long end of the Treasury curve slightly lower while the front end sold off, which read as a term-premium compression. The Treasury's official par yield curve for the day says otherwise: every maturity closed higher. The flattening was real, but it was a bear flattening, and the distinction changes what the day tells you.

What actually happened at 10:00 a.m. Eastern

The keynote, titled "In Our Time", ran in four parts, and Warsh flagged the structure with a joke that was also a policy statement: "You can call it an outline . . . you can call it a trail map . . . just don't call it forward guidance."

Part one was artificial intelligence — not as a market story but as a measurement problem for a central bank. He cited reports putting annualised token sales for the two leading labs alone above $100 billion, an increase of more than 500 percent from a year ago, and listed the open questions: whether the technology raises productivity in a sustained way and when, whether token usage complements or competes with labour, whether the next model generation demands greater capital intensity, and where the returns eventually land. He was explicit that a new internal task force on productivity and jobs will report later and has "no bearing on decisions we make in the current policy conjuncture".

Part two was the argument against forward guidance, which has been the through-line of his chairmanship so far. "Transparency in communications about future policy decisions is not a virtue unto itself," he said; the practice, adopted in the 2008 crisis, "has overstayed its welcome". He described the danger in the language of the academic literature — a hall-of-mirrors problem — and named the cost directly: "If markets rely materially on the Fed's guidance and the Fed relies on market prices, we are all more likely to be blinded." He then declined the obvious fallback of publishing a reaction function, on the grounds that no simple rule such as a Taylor rule can be "rigorously relied upon" when geopolitics, supply chains and technology are all moving at once.

Part three was a set of principles, of which the third and the fifth are the operative ones. Third: "The Fed's price-stability objective of 2 percent, as measured by the personal consumption expenditures (PCE) price index, is a firm, fixed target," followed by the sentence that did most of the work in the bond market — "Price stability is not self-executing, nor is inflation necessarily mean-reverting." Fifth: short-term interest rates are "the predominant tool" for the dual mandate, and unconventional policies "should otherwise be used sparingly, if at all".

Part four was the economy, and it is the part that repriced September.

Yields are the official daily par yield curve published by the US Treasury. Equity and dollar closes as reported by CNBC. Spread changes are calculated from the Treasury figures.

The measurement question got answered — in the hawkish direction

The 26 August update to this page flagged one thing worth listening for above the rate question: whether Warsh would address which inflation gauge he weighs, given that July's core PCE at 3.3% had been flattered by portfolio management and investment advice prices, and that BEA's market-based core measure rose only about 0.15% on the month. A chair who leaned on the softer sub-measures would have been making a dovish argument without issuing any guidance.

He did address it, and he went the other way.

He framed the problem on the headline index: "The Fed's preferred measure of inflation, the 12-month change in the PCE price index, stands at 3.7 percent, while the six-month change is 4.1 percent." Both are above the 3.3% annual core figure the market had been anchored to, and the six-month rate is the one that shows momentum rather than base effects. He added that the CPI equivalents "are also elevated, as are the core measures of both PCE and CPI inflation", and that "none of these measures are perfect, but they all tell a similar story".

Then he supplied his own breadth test, which is the most concrete thing in the speech. Disaggregating the 199 individual components of the PCE price index, 54 percent showed 12-month price increases above 3 percent — down from a post-pandemic peak near 77 percent, but well above the 32 percent that prevailed in the two decades before the pandemic. Over the past six months, 49 percent showed annualised increases above 3 percent.

The rest of the assessment removed the growth objection to tightening. Capital expenditure on equipment and intangibles is growing around 9 percent on a four-quarter basis, its fastest since 2021, with more than half of this year's growth ascribable to the AI buildout. Profits for firms in the S&P 500 have grown more than 20 percent over the past year, with margins elevated relative to history. Credit spreads on corporate bonds and leveraged loans sit near the low end of their historical ranges, issuance has been strong, and banks report commercial and industrial lending standards "on the easier end of their historical range". Real consumer spending has risen more than 2 percent over four quarters and private domestic final purchases at nearly 3 percent this calendar year. Unemployment is 4.1 percent and four-week average claims are near their lowest in decades, which he read as "consistent with full employment".

Put together: a labour market that meets the mandate, financial conditions he "would be hard pressed to describe" as restrictive, and inflation that is broad rather than narrow. That is the full architecture of a case for tightening, assembled without a single sentence of guidance.

Why the front end rose nearly five times as fast as the long end

This is the mechanism worth taking away from the day, and it survives a correction that is worth stating plainly. In the hours after the speech, wire reports had short-dated yields higher and the long end little changed to lower, and this page said so. The Treasury's official par yield curve, published after the session, shows every maturity closed higher. Nothing rallied. What happened was a bear flattening — the entire curve rose, and the front end rose nearly five times as much as the back.

A 2-year Treasury note is, to a first approximation, the market's forecast of the average federal funds rate over the next two years. Raise the odds of a September increase and it must move up. It did, hard: 14 basis points, from 4.20% on Thursday to 4.34% on Friday, and 17 basis points above the 4.17% close of 25 August.

A 30-year bond is a different instrument. Only a small part of its yield is a forecast of policy rates; the larger and more volatile part is term premium — the extra compensation investors demand for holding duration through decades of inflation and fiscal uncertainty. Through August, that premium had been the dominant story, with the long end selling off on issuance and fiscal concerns rather than on the policy path. A chairman who says the 2 percent target is fixed, that inflation is not self-correcting, and that "the responsibility for 65 months of sustained, elevated inflation sits squarely with the central bank" is speaking directly to the risk that premium exists to compensate for.

So the long end had two forces acting on it in opposite directions: a higher expected policy path pushing yields up, and a more credible inflation-fighting message pulling term premium down. The close says the first one won, but only just — the 30-year rose 3 basis points to 5.22% and the 20-year 3 basis points to 5.21%, against 14 at the 2-year and 11 at the 3-year. The 10-year, which sits between the two forces, rose 6 basis points to 4.73%.

The cleanest way to read that is as a spread rather than a level. The 2s30s curve narrowed from 99 to 88 basis points and 2s10s from 47 to 39. A market that genuinely doubted the Fed's willingness to control inflation would have sold the long end harder than the front on a speech acknowledging inflation is broad, because the news would be that the problem is bigger than assumed. It did the reverse. That is the credibility signal, and it does not require the long bond to rally to be visible — it only requires the long bond to lag.

BMO's US rates strategist Vail Hartman described it as "a deliberately hawkish speech that will put to rest any concerns about the Fed's willingness to raise rates to restore price stability". That framing also explains an equity response that was mild in the wrong direction: the S&P 500 gave up 0.25% to close at 7,711.76, the Nasdaq Composite 0.52% to 26,402.42 — weighed by semiconductors, with Marvell down more than 10% on a soft gross-margin guide — and the Dow 0.02% to 53,559.99. All three still finished the week higher, by 0.5%, 0.9% and 0.5%, the Dow's first winning week in three. A speech that raises near-term rate risk while lowering long-run inflation risk is not a straightforward negative for risk assets, which is why the term-premium story that drove August is the necessary context for reading Friday's tape.

The paradox: no guidance, twenty-two points of repricing

There is an obvious tension in a speech that argues against telling markets what the Fed will do, and then moves the implied probability of the next meeting by more than twenty points. It is worth being precise about why it is not actually a contradiction.

Forward guidance, in the sense Warsh is retiring, is a statement about future policy — a promise, a projected path, a conditional commitment. What he delivered instead was a statement about current conditions: here is what the data says, here is the standard I apply to it. Markets drew the inference themselves. That is precisely the regime he described wanting, in which "market participants themselves should be tracking real information across the economy" and drawing "their own conclusions".

The practical consequence for anyone reading this Fed is that the information has moved from the guidance paragraph to the assessment paragraph. Statements have been shortened to what he calls "just the facts". The projection apparatus has been pulled back. What remains is the chairman's description of the economy, and Friday demonstrated that a sufficiently pointed description is a fully adequate substitute for a signal. He was candid about the standard he applies: "We must be confident that underlying inflation is moving to our objective, clearly and at sufficient speed. Otherwise, we have work to do. That's our job . . . our mandate . . . and our charge to keep."

The July minutes, published 19 August, had already recorded that "many participants assessed that policy tightening would likely be necessary if inflation did not decline". On the named gauge, it did not decline. Friday's assessment turned that conditional into something markets could price — and it is worth noting that the minutes themselves produced a far smaller reaction when they landed.

What it means for the dollar

The rate factor is where a speech lands, and it landed cleanly. An implied probability of a September increase moving from roughly a third to better than even raises expected US short rates relative to the rest of the majors, and the 2-year Treasury at 4.34% now sits roughly 71 basis points above the 3.625% midpoint of the target range — a front end pricing tightening risk rather than easing.

This time the currency followed. The dollar index rose 0.35% to 99.46, touching 99.592 intraday, its highest since 19 August, and closed the week up around 0.7% — the biggest weekly gain in five weeks. The move was broad rather than concentrated in one cross.

Moves as reported by Reuters via CNBC. Note that two of these crosses had their own domestic news the same day: Tokyo core inflation accelerated in August for a third consecutive month, which firms the case for a Bank of Japan move and makes the yen's loss the more striking, while the Canadian dollar was carrying a second week of tariff escalation alongside a sharp Q2 GDP rebound.

Two caveats keep this from being a simple story for the US dollar.

First, the pricing has round-tripped rather than broken out. September hike odds stood at 55.8% a month ago before falling to roughly a third; 57.5% restores a level the market has already visited this summer. Rate differentials returning to where they were are a weaker impulse than differentials making new highs.

Second, one clean day does not re-establish the rate channel as the driver. For most of August the dollar's direction had nothing to do with the policy path: the currency rose against all six other majors on 25 August — EUR/USD to 1.1662, USD/JPY to 159.24, GBP/USD to 1.3632, USD/CAD to 1.3860 — in a week when front-end yields were falling. When a currency and its own rate expectations point in opposite directions for weeks, the marginal buyer is trading something else: in this case confidence in the fiscal and issuance outlook, which sits on the risk-sentiment and positioning side of the ledger rather than the rate side. Friday put the two back in alignment, which is informative precisely because it had not been true for a month. The test is whether they stay aligned through the September data.

So the honest read is narrow. Friday improved the interest-rate factor for the dollar, and by holding the long end nearly still while the front end jumped, it kept the term-premium problem from getting worse — which is a smaller claim than compressing it. It did nothing at all to the supply of Treasuries, which is what had been setting the direction. Those are separate channels, and the five-factor framework exists so that a read does not have to collapse them into one number.

What would change the picture

Three things, in order of how much they would move.

The August employment report, early September. This is the last labour print before the vote, and the only remaining input that could break the "consistent with full employment" half of Warsh's assessment. July delivered payrolls falling 23,000 against a consensus near +83,000 while the unemployment rate fell to 4.1% from 4.2% — a combination he explained on Friday as low turnover plus barely-growing labour supply, a reading that makes weak job gains benign rather than alarming. A print where employment and the jobless rate deteriorate together is the one that does not fit that explanation. Source is the BLS employment situation.

August CPI on 11 September. The only inflation reading between here and the decision, and it carries more weight than usual because BEA's next personal income release lands on 30 September, two weeks after the Committee announces. July core PCE at 3.3% is the last reading of the named gauge the meeting will ever see. Warsh's breadth statistic offers a way to read the CPI that does not depend on the headline: if the share of components running above 3 percent keeps falling, the argument he made on Friday weakens on its own terms.

The 16 September decision itself. Better-than-even is not a settled matter, and a chairman who has spent four months refusing to pre-commit has preserved exactly the room he says he wants. The three July dissenters — Hammack, Kashkari and Logan — already preferred a quarter-point increase, so the arithmetic of the Committee has not changed. What changed on Friday is that the chair's public description of the economy now reads closer to theirs than it did a week ago, and that is a fact about communication rather than a forecast about votes.

Educational macro context only — not investment advice.