Warsh hikes, the dots come to the market, and the long end refuses to follow

14:00 ET — the statement

The Committee raised the target range by a quarter point to 3-3/4 to 4 percent, its first increase since 2023, and the statement carries none of the hedging that usually accompanies a first move. Economic activity is "expanding at a solid pace", domestic spending "has been resilient", productivity growth is "strong" and capital investment "robust". On the labour market: "job gains have kept pace with the workforce, and the unemployment rate has changed little" — no softening, no reference to downside risks to employment. On prices: "inflation remains elevated", followed by the sentence that does the work, "today's policy action will support a timelier return to the Committee's 2 percent goal." That is a Committee saying it was not returning to target fast enough on the old setting. Balance-sheet language is unchanged: "continuing its policy of maintaining ample reserves in the banking system."

The vote was 12–0. July's meeting produced three hawkish dissents — Hammack, Kashkari and Logan all wanted 25bp then — and all three vote this year, so the unanimity is what you would expect from a Committee that has given the hawks what they asked for. What is more notable is the absence of a dissent in the other direction: nobody on the Committee was willing to put their name to a hold. We could not extract the named voting roster from the statement page across five attempts spanning two sessions, so Governor Cook's individual vote remains unconfirmed; she is a sitting governor after the Supreme Court rejected the removal attempt on 29 June, and the Fed's own FOMC and Board pages list her, but that is inference from a listing rather than a report and is labelled as such here for the second consecutive edition.

The projections — the actual event

The hike was 92% priced going in. The Summary of Economic Projections was not, and it is where the session's information was. The end-2026 median moved to 4.1% from 3.8% in June, with a distribution of 12 participants at 4.125%, 4 at 4.375% and 2 at 3.875%. The central tendency is 4.1–4.4% and the full range is 3.9–4.4% — the arithmetic of that range is the single most quotable fact in the document, because it means no participant, not one, projects the funds rate finishing 2026 below tonight's level. The cut is not a minority view on this Committee; it is not a view at all.

Further out the revisions are larger. End-2027 went to 4.1% from 3.6%, a 50bp move; end-2028 is 3.9%; 2029 is 3.6%; longer-run nudged to 3.2% from 3.1%. Read across the years, the Committee is projecting policy above its own estimate of neutral for at least three more years. The economic projections explain why: 2026 PCE inflation revised up to 3.7%, core PCE 3.4%, unemployment revised down to 4.1% and GDP up to 2.3%. Hotter inflation with a tighter labour market and better growth is the combination that makes a hiking cycle look like arithmetic rather than a choice.

14:30 ET — Warsh's first press conference as Chair

The framing was narrower than the mandate. "We must be confident that underlying inflation is moving to our objective, clearly and at sufficient speed. Today, the FOMC decided that this standard has not been satisfied." And: "The plain fact is that inflation is too high and has been for too long." Asked about the balance of the mandate, he said "the predominant focus is on the price stability side". That is a reaction function stated as a test — "clearly and at sufficient speed" — rather than as a forecast, which is consistent with the abandonment of forward guidance he set out at Jackson Hole. We could not obtain direct quotations on balance-sheet policy or on Fed independence and do not paraphrase what we did not read.

The market's read is visible in what moved after 14:30 rather than after 14:00. Gold was still around $4,350 at 13:43 ET, above $4,300 and higher on the day; it closed near $4,269. That $80 arrived during and after the press conference, not on the statement. The same is true of the equity drift into the close: TheStreet's mid-afternoon capture had the S&P at −0.45% and the Dow at −1.07%, and the Dow finished around −1.40%.

The US close, and a number that does not reconcile

S&P 500 ≈7,552.8, −0.43%. That figure is derived two ways: a settled SPY close of $754.10 (−0.43%, "At close: Sep 16, 2026, 4:00 PM EDT") scaled by Tuesday's verified SPX/SPY ratio gives 7,552.8, and an independent whole-point market update prints 7,552. A third source, a widely-read live blog, carries 7,527.80 (−0.76%), and we do not publish it: a 33bp gap between SPX and a settled SPY close is not a tracking difference, and the ETF-scaled route is what caught it. The Dow is the mirror case — that same live blog's 51,363.01 (−1.40%) agrees closely with a settled DIA close of −1.44%, while the market update's −1.2% is the outlier, so we publish the lower figure. The Nasdaq Composite is genuinely unresolved between 25,978 (−0.01%) and 25,875.21 (−0.41%) and both are shown; QQQ, the Nasdaq-100 proxy, closed +0.02%. Russell 2000 ≈2,856.96 (−0.46%), single source, against a 15 September base we never verified. VIX 17.77 (+3.3%) — and the two-session VIX gap is closed at the same time, because FRED's official series now confirms 17.10 on 14 September and 17.20 on 15 September, which retrospectively vindicates discarding the "15.69" figure that turned out to be a 2025 row.

US data ahead of the decision leaned the Fed's way. August import prices +0.7% m/m against +0.4% consensus, reversing July's −0.3%, and +7.0% y/y — the largest twelve-month gain since a 7.7% rise ending August 2022; export prices +0.6% m/m, +8.6% y/y. Imported inflation at a four-year high two hours before a hawkish hike is a tidy piece of corroboration. August housing starts and permits could not be sourced: every route returned 2025-vintage data.

Europe — a rally into the decision, and a gilt bid

European cash closed hours before the Fed and closed higher: Stoxx 600 637.10 (+0.46%), which reconstructs exactly to Tuesday's verified 634.18; DAX 25,617.00 (+0.75%); CAC 8,140.59 (+0.62%); FTSE MIB ≈52,099 (+0.8%, two agency reports agreeing); IBEX 19,619 (+0.32%). The FTSE 100 and the SMI could not be closed at all — every source returned Tuesday's figures under Wednesday headlines, a failure mode this library has now logged four times.

The real European story was in gilts. UK August CPI printed 3.1% headline, 2.6% core, 3.4% services — all exactly on consensus, with PPI output 3.7% against 3.3% expected. Headline up on energy while core and services stand still is the arithmetic of a hold, and the market took it that way: 2-year gilts rallied 15–18bp and 10s 10bp to 5.32%. Goldman's published call into tonight's meeting is a hold with a hike flagged for November. A widely-circulated "Goldman sees a 6–3 vote" headline turns out to be a December 2025 article and is discarded. On the continent, Econostream's speaker feed has Santos Pereira cautiously hawkish on 16 September — noting inflation "was not broadening as quickly as in the previous inflationary episode" — and Moulin tying rising inflation expectations to the global bond selloff, while its own tone meter declined after the 10 September hike. October ECB pricing is unchanged around 29%.

Asia and Australia — risk-on, before the fact

Every Asian and Australian market in this note closed on Wednesday afternoon, ten to eighteen hours before the FOMC. They closed higher, and that positioning is the overhang into today's open rather than a judgement on the decision. Nikkei 63,923.00 (+0.69%) on a 713-point range, verified against the exchange's own archive; TOPIX 4,061.72 (+0.6%, single source). JGBs bull-flattened after Tuesday's bear steepener — 10y 3.00% (−4.2bp), 20y 3.86%, 30y 4.12%, 40y 4.17%. KOSPI 6,717.97 (+1.37%) snapped a four-session slide on chip bargain-hunting, Samsung +2.01% and SK hynix +4.08%. CSI 300 4,480.27 (+0.68%) and Shanghai 3,891.60 (+0.71%) both reconstructed exactly to their verified priors; Hang Seng 24,713 (+0.19%), which is a second consecutive session of H-shares underperforming the mainland. Sensex +0.45%, Nifty +0.43% with midcaps and smallcaps negative. TAIEX +0.65%, level not obtainable.

Australia: ASX 200 ≈8,696.5, +24.0 points, +0.28% — which corrects the 8,691.1 / +18.6 / +0.21% this desk published yesterday from a single source. Energy led on the Saudi supply story; Reliance Worldwide rose about 5% on a reported $4bn Brookfield approach; Life360 −6.2% and Telix −5.2% were the notable losers. Sector performance, breadth, volume and the A-VIX are all missing this morning because every evening-wrap URL resolved to cached articles dated March and June. The front end rallied hard — ACGB 2y 5.02% (−5bp), 3y 5.01% (−5bp), 10y 5.35% (−8bp) — taking 3s10s to 34bp.

Fed detail — the pricing caveat matters more than usual today. The only fed-funds read available at filing is stamped 14:05 ET. That is five minutes after the statement and twenty-five minutes before Warsh began speaking, and the press conference demonstrably moved assets — gold gave up $80 during and after it. The page refreshes after 17:00 ET, which is after this edition files. So the October 43.6% and the December distribution are a pre-press-conference snapshot and should be assumed stale in the hawkish direction. This is the specific failure that produced a 3.4-point error on 15 September, and the rule written then is being applied rather than quietly ignored. The settled read is the first verification target tomorrow.

BoE detail — tonight, and the asymmetry is unchanged. Bank Rate 3.75%, no Monetary Policy Report at this meeting, hold priced around 73%. The August CPI that landed yesterday was the most on-consensus print of the month — headline 3.1%, core 2.6%, services 3.4%, every one of them matching — with motor fuels the largest upward contribution and motor fuel inflation running 23.0% y/y. Headline driven by energy while core and services stand still is the textbook case for holding through a supply shock, and the gilt market agreed emphatically. The asymmetry is therefore dovish and the vote split is the trade, not the level. We carry no vote-split forecast: the "Goldman 6–3" figure circulating this week is from a December 2025 article about a different meeting and has been discarded.

RBA detail — the tracker and the bonds disagree again. The hike probability for 29 September rose to 78% from 76% on a session when ACGB 2s and 3s both rallied five basis points. Yesterday the same tracker printed an unchanged 76% while the front end sold off nine. Two consecutive sessions of the OIS reading moving independently of, or opposite to, the physical front end is a reason to treat the number as a soft input rather than a hard one. The physical market's own statement is unambiguous and unchanged: a 2-year at 5.02% against a 4.35% cash rate embeds roughly 67bp of tightening, which is closer to three hikes than one, and that is the figure to size against. Bullock appears before the House Economics Committee tomorrow at 09:30 AEST alongside Hauser, Hunter and Jones — semi-annual parliamentary testimony, not a speech, and a correction to yesterday's framing. A separate CEDA fireside chat follows on 22 September.

BoJ detail. The decision is Friday, the 18th; the meeting runs the 17th and 18th and third-party calendars persistently quote the start. Around 72% is priced for a move to 1.25% with a hawkish dissent already on the record from July. The complication is the calendar behind it: Tokyo is shut Monday, Tuesday and Wednesday next week — 21, 22 and 23 September, three consecutive sessions, now confirmed from the exchange after two editions of uncertainty about the 23rd. A hike on Friday with the market shut for three days afterwards is a gap-risk problem rather than a drift problem, and positions should be sized for the reopen rather than for the decision.

Closed — V014 Gold, long, WRONG. Opened 7 September at $4,355.80; closed ≈$4,269, −1.99%. The invalidation was "a close below $4,300 without a bid", and No. 008 pre-committed the close in writing: "if the settled mark confirms a close below $4,300, this view closes tomorrow with no argument about which vendor was right." Three independent vendors put Wednesday's close between $4,244 and $4,298 and not one prints a close at or above $4,300, so the argument the pre-commitment was written to prevent never had to happen. That is the fifth consecutive close to fire on a condition written down in advance. Worth recording honestly: the view was above its stop at the intraday high of roughly $4,350 two hours before the close, and the $80 that took it out arrived during and after Warsh's press conference. Not re-opened as a short. The observation that replaces it is about structure rather than direction — 56.4% of open interest, gross longs 8.99× shorts, only 29,047 shorts remaining, and last week's net build coming almost entirely from short-covering — which means there is nobody on the other side of an unwind. That is a reason not to own it, which closing achieves, and not an edge in owning the other side.

Closed — V012 DXY, neutral two-way, SCRATCH. Opened 7 September at 99.17; closed 100.24, +1.08%. Its published trigger had two clauses and the second fired cleanly: "a dot plot that extends the path materially." The 2026 median moved 30bp and the 2027 median 50bp, which is as material as a dot plot gets. The view was constructed around one question — whether the Fed-independence channel could make a hawkish Fed dollar-negative — and the session answered it: the dollar broke 100 to a seven-week high on the hike, and the independence channel did not dominate. It scores as a scratch because a neutral view has nothing to be right or wrong about, and that is the point worth taking from it rather than the score. This is the second time a deliberately two-sided view has scratched while the underlying delivered a clean directional move — V001 in US 2s did exactly the same thing on 14 September — and it echoes V015's lesson from a week earlier that "neutral was the wrong shape". Three data points now say the same thing: when the question is binary and dated, a neutral view is a decision not to have one, and it should be logged as an abstention rather than dressed as a position.

Nothing opened, for a fourth consecutive edition, and the reasons are specific. Three candidates were live. A 2s10s flattener is the direct read of the session — the SEP removed the front end's dovish optionality while the long end refused to follow — and is declined because it is a mechanical consequence of a hike that has just been delivered, which makes entering it one session later a level trade rather than a mechanism trade. A Dow-versus-Nasdaq rotation has the better mechanism, but the evidence is one session long, it sits inside a negative-gamma expiry week that swamps fundamental signals, and it would mean buying the trade 53% of the Fund Manager Survey calls the most crowded on the board. A short-gold successor is declined on the principle above. The honest characterisation is that the book is now thirteen views and shrinking by attrition rather than by decision, and that is itself something for the ledger to judge rather than a neutral fact.

Scorecard. Thirteen open, thirteen closed: two right, seven wrong, four scratch — two of nine on decided views, from two of eight. That is a fourth consecutive weekly deterioration and it is recorded as one. What is holding up is the process rather than the hit rate: five consecutive closes have now fired on conditions written down before the fact — V011's "a close outside the range", V019's "a weekday close above $2,500", V026's distillate clause, V018's "cloture fails" and now V014's $4,300 — and none required a judgement call at the moment of closing. Working this morning: US credit, iron ore, the ASX underweight, the S&P hedge, copper, ACGB 3s10s, AUD/NZD, OAT–Bund, Brent, USD/JPY, and 5s30s, which survived its worst event. Against: AUD/USD at 23 pips, and the China/HK H-over-A leg at two of its three-session trigger.

Portfolio-level read. The four-week pattern took a fifth data point and it held again: every view expressing a mechanism worked through the FOMC, and the two views that closed were the two expressing a level or an abstention. 5s30s survived a hawkish hike because the fiscal thesis it encodes does not depend on the policy path. ACGB 3s10s worked in the opposite direction from the day before, which is the signature of a real curve view. Credit widened in the specific place the view says it should. Meanwhile gold was taken out by a number and DXY was taken out by having no opinion. The structural fact to carry into Friday is that the decision is behind us and the mechanics are not: the index is below its gamma flip with the largest expiry of the year tomorrow, the corporate bid is absent, Tokyo shuts for three days after a probable BoJ hike, and the Saudi restart binary resolves inside a week into that same thin liquidity. Carry less gross, own gamma rather than direction through Friday, and note that the two positions the market is most crowded into — semis at 53% of the survey and a Treasury complex carrying a validated 6,863,118-contract leveraged short — are both trades where the exit is narrower than the entrance.

These are analytical framings for a professional reader, expressed in the vernacular of a macro desk; they are not personalised investment advice and carry no position sizing. The "what changes the view" column is the accountability mechanism — every view is logged and scored in the project's views ledger, including the ones that lose.