Two Authorities, One Price

On 28 August, Chairman Warsh described the method that replaced forward guidance: with the Fed no longer telling markets what to expect, it would look instead to market signals — explicitly including Treasury prices and trading volumes — as inputs into its assessment of economic activity, inflation, financial conditions, and risk.

Nine days earlier, Treasury had intervened in long-end price formation by announcing that its liquidity-support buybacks would at least double, from a $2 billion maximum to at least $4 billion per operation, effective 9 September. Under the existing tentative schedule, the first long-end operation after that date is scheduled for 10 September.

Both institutions are acting coherently within their own mandates. Neither, however, has addressed what the other's behaviour does to its own framework. The result is that long-end Treasury prices now contain policy-relevant information of potentially ambiguous origin — and the Fed has not described how it intends to distinguish market information from policy-induced price effects.

Warsh named a related reflexivity risk himself. He devoted several paragraphs to the hall-of-mirrors problem: markets reading the Fed while the Fed reads markets, both parties blinded to new information and more likely to commit errors. Removing forward guidance is his answer to it. The hall of mirrors he described has two participants. The August record contains three.

Each of the three articles published this month independently identified a source of ambiguity in a market-price channel that the new framework depends on. Together, they establish a dependency in Warsh's method that deserves to be examined rather than assumed away.

Article 1 (6 Aug) — the equity channel does not respond to news in the expected direction.

WARSH vs. WALL STREET: THE NFP SHOWDOWN

Four of six 2026 NFP prints moved the S&P opposite to naive beat/miss logic. The piece asked whether the market was discounting the Fed's path or in denial about it. Warsh now lists the level and change of asset prices across sectors among his inputs. Article 1 therefore showed that the equity-price response was not a reliable one-directional transmission mechanism in that sample. Six observations do not establish a systematic inversion, and none is claimed.

Article 2 (12 Aug) — the curve carries two signals, not one.

The Fed at Crossroads: The Wedge, the Steepener, and the Silence

The front end primarily prices the expected Fed path; the long end incorporates credibility, fiscal supply, institutional risk, inflation, and term premium. Warsh's method requires reading market prices, including Treasury prices. Article 2 proposed that the long end is a composite signal. Reading that composite as a policy signal therefore requires some method of decomposition. Warsh has not stated one — and under a framework that removed forward guidance precisely in order to obtain cleaner signals, the absence of a stated filter is not a technical omission but a gap in the method itself.

Article 3 (22 Aug) — the long-end channel has an identified interference source.

The Treasury Signal That Lasted One Session

An off-cycle announcement of increased buyback capacity produced a nine-basis-point move in the 30-year that fully retraced within one session — before any security had been purchased under the revised schedule, and before the change had taken effect. Article 3 concluded that the intervention may have increased the information content of the long end rather than resolving the underlying supply problem. Jackson Hole is why that matters: the Fed has now explicitly committed to reading that price.

That last point carries further than it first appears. If an announcement alone moved the long end, the transmission channel is expectational rather than mechanical. The contamination does not begin when Treasury buys a bond; it begins when Treasury says it might. The enlarged operations had not occurred on 28 August and have not occurred yet — and the channel was already affected.

The arc: the Fed removed its own signal → the market had to infer the Fed's reaction function from observable prices → Article 2 proposed that those prices contain distinct information channels → and one of those channels is subject to intervention by another arm of government.

Four things in the speech bear directly on the August work.

The target measure is settled. Article 2 argued that the inflation debate was a measurement debate, that core CPI at 2.5% and core PCE at 3.3% were telling opposite stories, and that commentary anchored on CPI was measuring the wrong series. Warsh, third principle: the 2 percent objective, as measured by the PCE price index, is a firm, fixed target. On matched July data the gap is now unambiguous — core CPI 2.5%, core PCE 3.3%, 80 basis points, same month. Article 2 had to present this as a mixed-month comparison. It no longer does.

The reaction function was refused as doctrine, not withheld as tactics. Warsh did not decline to specify a reaction function; he argued against having one, on the grounds that economic understanding does not support a mechanical rule and that quasi-commitments on rates inhibit the freedom to decide correctly later. The closing line — committed to a discipline, not to a decision — is the position stated plainly.

The financial-conditions gap, filled by the Chairman. Article 2 conceded that the hawks' strongest counter was easier financial conditions against an unchanged policy rate, and that it did not model that channel. Warsh made the argument directly: credit spreads near the low end of historical ranges, C&I lending standards on the easier end, strong issuance, and his own conclusion that he would be hard pressed to call broad financial conditions restrictive.

He also closed the breadth gap Article 2 flagged. That note said trimmed-mean measures were the only ones able to establish whether price pressure was narrow or broad, and did not have them. Warsh disaggregated the 199 components of the PCE basket: 54% showed twelve-month price increases above 3%, and 49% did so over six months, against a 32% pre-pandemic norm and a 77% post-pandemic peak.

What he did not settle is how the Fed reads a price series that another arm of government has demonstrated it will move.

The thesis makes a claim that can be tested: if Treasury's action functions like a policy event in the Treasury market, it should look like one in the data.

Method. 164 FRED daily observations, 2 January to 27 August 2026, for DGS2/5/10/30 and DFII5/10/30. The 28 August observation is taken from Treasury's par constant-maturity curves, because FRED had not posted that date at the time of writing; the two sources matched at every tenor on 27 August, but they are distinct series and the comparison is flagged throughout. The measured quantity is the curve gap — the day's change in the 2-year yield minus the change in the 30-year, in basis points. It isolates the front-end/long-end separation Article 2 proposed, and it is what a policy event should move if that separation is real.

Events are of two kinds. Information events are the twenty-one NFP, CPI and PCE releases of 2026 — the complete Jan–Jul calendar for all three series. Policy events are the three Fed communication events since the June transition and the 19 August Treasury announcement.

Result 1 — 28 August was the most curve-dislocating day of 2026

Table 1. CALCULATION from FRED daily series; 28 August from Treasury par CMT. 2026 daily standard deviations: 2Y 4.7bp, 30Y 3.8bp, curve gap 3.9bp.

Two things follow, and the second is the more important.

The front-end move was the largest single-day repricing of the year, and the curve gap was equally extreme. But the breakeven move was unremarkable — the 19.5th percentile, with 28 of 164 days in 2026 showing a 10-year breakeven fall of two basis points or more. A note built on the inflation-compensation decomposition would be building on noise. The distinctive feature of 28 August is what happened to the shape of the curve, not to its inflation content.

Result 2 — policy events and information events do not overlap

Table 2. CALCULATION. Policy events: 17 June FOMC, 29 July FOMC, 19 August Treasury announcement, 28 August Jackson Hole. Information events: seven NFP, seven CPI, seven PCE releases, reference months January to July 2026.

Table 3. FACT (yield levels, FRED and Treasury par CMT). CALCULATION (changes and gaps).

The smallest policy event exceeds the largest information event. Nine basis points against eight. There is no overlap between the two sets, and all four policy events sit in the top six of 165 days ranked by absolute curve gap. Under random assignment, four specified days landing in the top six of 165 has probability of roughly five in ten million.

Information events, by contrast, were not materially different from ordinary days in this sample: mean absolute curve gap of 2.7bp against 2.6bp for the remaining 140 sessions. A Welch two-sample test returns p = 0.88 and a Mann–Whitney test p = 0.49; neither rejects equality, so the comparison is reported descriptively though the inference points the same way. Under this Fed, macro data releases do not move the shape of the curve. Policy communications do.

The boundary is not fragile to sample construction. This study was first run on a twelve-event subset and then re-run on the complete twenty-one-event calendar. The information-event maximum stayed at 8.0bp and the mean moved by a tenth of a basis point. Nine additional releases produced no observation that reached the policy range.

Result 3 — the Treasury announcement sits inside the policy cluster

This is the finding the thesis needed, and it is not one this desk expected to be so clean.

The 19 August Treasury announcement produced a curve gap of 9bp — larger than all twenty-one CPI, PCE and NFP releases of 2026, and sixth-largest of 165 trading days. On the measure that separates policy events from everything else, an announcement by the fiscal authority is indistinguishable from a meeting of the FOMC.

The three Fed events produced different curve configurations: 17 June and 28 August were strongly front-end dominated, while 29 July was strongly long-end dominated (2Y −4bp, 30Y +11bp). The Treasury event was also long-end dominated, but in the opposite direction — the 2-year unchanged to the basis point, the 30-year down nine.

That is the two-channel structure Article 2 proposed, now with event evidence behind it: policy communication can transmit through either segment — the Fed moved the front end in June and August and the long end in July — while the Treasury announcement in this sample transmitted specifically through the long end. And the long end is the segment Warsh has now committed to reading.

What the study does not establish

Four policy events is a descriptive result, not an inferential one. The separation is complete and the ranking probability is small, but n=4 cannot support a claim about the population of policy events in general. A fifth event that overlapped the information range would weaken it materially.

The 28 August observation is cross-source. The two series matched exactly at all seven tenors on 27 August, and the 2-year gap would have to be mis-stated by more than 3bp to move 28 August out of the top rank — but it is not a like-for-like comparison and is not presented as one.

Daily closes cannot isolate the speech from the rest of the session. On 28 August that includes the 10:00 ET preliminary benchmark revision, month-end flows, and pre-weekend positioning. Every observation in the study carries the same limitation, which is the argument for a baseline rather than a single-day claim — but it is a limitation, not a solved problem.

If a Treasury reaction function is operating in the long end, one observable implication is a soft floor under long-end prices: reduced upside yield volatility when yields are already elevated, because market participants anticipate intervention at bad levels.

That prediction was tested and it failed.

Table 4. CALCULATION. Daily 30-year changes bucketed by the prior day’s 30-year yield level; top quartile at or above the 75th percentile (5.055%), bottom at or below the 25th (4.860%). Buckets are 41 and 46 rather than equal because six days share the exact 25th-percentile value (4.86%) and are all included; cutoffs are fixed by the level distribution, not chosen after inspecting the response.

There is no asymmetry. Upside and downside moves in the 30-year are the same size at high yield levels as at low ones, and the ratios are effectively identical across quartiles. The long end does not behave as though a floor exists.

That is a negative result and it is reported as one. It constrains the thesis in a specific way: the interference is observable in event response — the 19 August announcement moved the curve like a policy event — but not in ongoing price behaviour. The market treats a Treasury announcement as news when it arrives. It does not appear to price a standing expectation of intervention.

Stated in advance, because a hostile reader should be able to see exactly what would disprove the argument. This section is not a claim that Treasury must behave in a particular way for the thesis to hold. It specifies the observable future behaviour that would weaken or disprove the interference hypothesis.

Treasury conducts the 10 and 24 September operations at the pre-announcement size and frequency, and the 4 November refunding proceeds on the normal cycle. If the 19 August announcement proves to be an isolated adjustment and subsequent operations restore predictability, the proposed interference channel becomes substantially weaker and the dependency remains theoretical.

The Fed articulates an explicit filter for distinguishing intervention-driven from information-driven movements in long-end Treasury prices. If the framework specifies how such price effects are identified or controlled for, the information-ambiguity problem is materially reduced.

A subsequent policy event produces a curve gap inside the information-event range. The separation in Table 2 is complete across four observations. A single Fed or Treasury communication that moves the curve gap by less than 8bp would show the boundary is an artefact of a small sample rather than a feature of the regime. The 15–16 September FOMC is the next test.

The long end shows no statistically or economically meaningful asymmetry at elevated yield levels. Already run, in Section 04, and already failed. Reported rather than omitted.

Article 2 published four dated falsification conditions and one risk condition. Three have resolved.

Table 5. FACT. Sources: BEA Personal Income and Outlays, July 2026; BLS USDL-26-1425.

A call that was wrong. Article 2 described the 28 August benchmark revision as “an underpriced event” and flagged downside employment risk. It came in at −79,000, or −0.1% of total nonfarm, against a ten-year absolute average of 0.2%. That is half the typical size. The risk was overpriced by this desk, not underpriced by the market.

A claim that is downgraded. Article 2's second claim held that the 29 July bear steepener was primarily a credibility reaction rather than a rate-path reaction. The event study bears on this directly. 29 July registers a curve gap of −15bp, the joint-largest of 2026 — so the event was extraordinary, and Article 2 was right that something structural happened. But the mechanism attributed to it is not supported: if a hawkish Fed policy event moves 10-year breakevens down two basis points, inflation compensation in this market is not responding to Fed credibility messaging in the direction the 29 July reading assumed.

The September decision is genuinely open, and this desk holds no directional view stronger than the pricing. Article 2's base case of a hold survived every testable condition, but the conditions that produced it have moved. Core PCE has made no net progress in four months — 3.3%, 3.4%, 3.3%, 3.3% from April through July. The Chairman has defined the 2% objective in PCE terms, called financial conditions non-restrictive, and produced a breadth measure showing roughly half the basket running above 3%.

What the futures actually repriced

Table 6. FACT. CME FedWatch via QuikStrike. Probability of a target range above 3.50–3.75%. September panel as of 30 August 05:20 CT with one-week reference 21 August; October and December panels as of 31 August 09:06 CT with one-week reference 24 August. Probabilities are risk-neutral, not physical.

Three things follow.

Jackson Hole restored pricing rather than creating it. September hike probability ran 63.4% a month before the speech, fell to 39.9% by 21 August, and closed 28 August at 57.0% — still 6.4 percentage points below where it stood on 30 July. The speech reversed an August drift. It did not take the market anywhere it had not already been. A reader who saw only the one-week move would conclude Warsh moved the market seventeen points; the month gives the opposite impression.

The market prices roughly twice the June SEP median. Weighting each meeting’s distribution by the number of 25bp steps gives 1.47 hikes, or 36.8 basis points, priced by the 9 December meeting. The June SEP median dot of 3.8% against a current midpoint of 3.625% implies 17.5 basis points. Article 2 argued that the 3.8% median was an artefact of a nine-nine split rather than a centre of gravity, and that treating it as the Committee’s view overstated its coherence. The futures market is now positioned at roughly twice that path.

No cut is priced anywhere. Ease probability is 0.0% at all three meetings. Through year-end 2026 the market assigns no weight to the Committee moving the target range lower.

What is tradeable here is not the September decision. It is the information content of the long end.

If policy events move the curve by nearly five times more than data events — 12.5bp against 2.6bp on the measure in Table 2 — and if one of the four policy events in the sample originated at Treasury rather than the Fed, then long-end moves carry policy content whose source is not identifiable from the price. A participant marking a 30-year move must now distinguish three possibilities where the framework assumes two: information, supply, or an authority acting on the level.

The Fed has said it reads that price. It has not said how it separates the three.

Scheduled catalysts

Rates, 27 → 28 August. Nominal: 2Y 4.20 → 4.34; 5Y 4.38 → 4.48; 10Y 4.67 → 4.73; 30Y 5.19 → 5.22. Real: 5Y 2.07 → 2.18; 10Y 2.34 → 2.42; 30Y 2.92 → 2.96. Implied breakevens: 5Y 2.31 → 2.30; 10Y 2.33 → 2.31; 30Y 2.27 → 2.26. Spreads: 2s10s +47 → +39; 2s30s +99 → +88; 10s30s +52 → +49. The identity holds at every tenor: change in nominal equals change in real plus change in breakeven.

Event-study panel. 164 FRED daily observations, 2 January to 27 August 2026, seven series. 2026 daily standard deviations: 2Y 4.7bp, 30Y 3.8bp, curve gap 3.9bp; mean absolute curve gap 2.8bp. Pre-transition (114 days, to 16 June) curve-gap standard deviation 3.6bp; post-transition (50 days) 4.3bp.

Inflation (July, released 26 August). PCE +0.2% m/m, 3.7% y/y; core PCE +0.2% m/m, 3.3% y/y. Core PCE by month: April 3.3%, May 3.4%, June 3.3%, July 3.3%. BEA’s 30 September annual update will revise portions of the historical PCE price index back to 2021, so this four-month path is provisional. Core CPI (July) 2.5% y/y; matched-month gap to core PCE 80bp.

Labor. Preliminary benchmark revision to March 2026: total nonfarm −79,000 (−0.1%); total private −178,000 (−0.1%); ten-year absolute average 0.2%. Final revision February 2027.

Treasury. Press release SB0607, 19 August 2026, 08:32 ET: maximum size of liquidity-support buyback operations for longer-dated nominal coupon securities (10-to-20-year and 20-to-30-year sectors) increased from $2 billion to at least $4 billion per operation, effective 9 September 2026 through 4 November 2026.

Primary. Federal Reserve Bank of St. Louis (FRED), series DGS2, DGS5, DGS10, DGS30, DFII5, DFII10, DFII30, daily, 2 January to 27 August 2026 — the source for the event study. U.S. Department of the Treasury, Daily Treasury Par Yield Curve Rates and Daily Par Real Yield Curve Rates, 28 August 2026 — the source for the 28 August observation only. Board of Governors of the Federal Reserve System, In Our Time, keynote remarks by Chairman Kevin Warsh, 2026 Jackson Hole Economic Policy Symposium, 28 August 2026 — the source for every characterisation of the speech, including the PCE target statement, the reaction-function argument, the financial-conditions assessment, the 199-component breadth disaggregation, the market-signal inputs and the hall-of-mirrors passage. U.S. Department of the Treasury press release SB0607, 19 August 2026. CME FedWatch target-rate probabilities via QuikStrike, panels captured 30 August 2026 05:20 CT (September meeting) and 31 August 2026 09:06 CT (October and December meetings). Bureau of Economic Analysis, Personal Income and Outlays, July 2026, released 26 August 2026. Bureau of Labor Statistics, Current Employment Statistics Preliminary Benchmark (National) — March 2026, USDL-26-1425, released 28 August 2026.

On the 28 August observation. FRED had not published DGS or DFII values for 28 August at the time of writing. Treasury's par constant-maturity series and the Federal Reserve's H.15 constant-maturity series are distinct constructions with distinct fitting methods; they coincided at all seven tenors on 27 August but are not guaranteed to coincide on any other date. The 28 August row is therefore cross-source and is flagged as such wherever it appears. It is not blended into the FRED panel.

On event dates. The information-event sample is the complete set of NFP, CPI and PCE releases for reference months January to July 2026 — twenty-one events, fourteen before the June transition and seven after. NFP dates are those cited in the prior Regime Intelligence articles; CPI and PCE dates are from the published BLS and BEA release schedules. All releases are 08:30 ET. An earlier version of this study used a twelve-event subset; expanding to the full calendar raised the mean absolute curve gap from 2.6bp to 2.7bp and left the maximum unchanged at 8.0bp, so the separation reported in Table 2 is not an artefact of which information events were selected.

Corrections to prior notes. Article 2 described the 28 August benchmark revision as an underpriced event; it came in at half the ten-year average size, and that characterisation was wrong. Article 2's second claim — that 29 July was primarily a credibility repricing — is downgraded to an open question in Section 06.

Analysis of publicly available data and published central bank communications. Not investment advice.