1.25% and a Weaker Yen (18 September 2026): The Bank of Japan Hiked to a 31-Year High, and the Two Dissents Did the Talking

Three of the eight majors set policy inside 72 hours, and the last one inverted the textbook. On Friday 18 September the Bank of Japan raised the uncollateralised overnight call rate to around 1.25 percent — the highest in Japan since 1995 — and the yen weakened, the 10-year government bond yield fell 4.9 basis points to 2.947% and the Nikkei 225 rose 1.5%. All three are the opposite of the standard response to a rate rise. The hike was not the news: every one of the 52 economists in Bloomberg's survey had it. The news was a 7–2 vote in which the two dissenters wanted no hike at all, from a Board whose only dissenter at each of the previous two meetings had wanted a bigger one. Earlier in the week the Federal Reserve raised to 3-3/4 to 4 percent on a 12–0 vote, and the Bank of England held at 3.75% by 6–3 while voting unanimously to unwind £488 billion of gilts to zero. Three decisions, and in every one of them the vote carried more information than the rate.

What actually happened at the Fed

The FOMC statement of 16 September is four sentences longer than a tweet and contains no forward guidance at all. The Committee "decided to raise the target range for the federal funds rate by 1/4 percentage point to 3-3/4 to 4 percent," by a 12–0 vote. Economic activity is "expanding at a solid pace," domestic spending "has been resilient," job gains "have kept pace with the workforce," and "inflation remains elevated." Then the line that carries the weight: "Today's policy action will support a timelier return to the Committee's 2 percent goal."

Timelier is doing a lot of work. It concedes that the previous path was going to get there eventually and says the Committee decided eventually was not good enough — which is the argument the three July dissenters made and lost, and have now won.

What is genuinely unusual is the unanimity. A committee that split 9–3 in July did not become of one mind in seven weeks. A hike is simply harder to vote against than a hold: a member who thought the move premature had to weigh a public dissent against a chair and a large majority at the precise moment the Committee was trying to establish that it would, in the statement's words, "deliver price stability." Unanimity here is evidence about the cost of dissenting, not about the range of opinion. For the range of opinion you have to read the dots.

The dots moved much further than the rate did

Three months ago the median participant already had one hike in their 2026 projection. September did not just confirm it — it added a second one and then refused to take either back.

Source: the Federal Reserve's September Summary of Economic Projections, in which eighteen participants submitted projections.

Read the first two rows as a pair. A 2026 median of 4.1% against a post-meeting midpoint of about 3.88% is one more quarter-point increase this year. But the 2027 median is also 4.1% — the Committee's central view is now that the rate it reaches at the end of this year is the rate it still holds a full year later. In June, the 2027 median sat 20 basis points below the 2026 median, which is a curve with easing in it. That easing has been deleted.

The distribution behind the 2026 median is where the remaining argument lives. The central tendency — which strips the three highest and three lowest projections — runs 4.1% to 4.4%, and the full range runs 3.9% to 4.4%. A range that starts at 3.9% means at least one participant thinks the Committee is already done; a central tendency that starts at 4.1% means at least fifteen of the eighteen have another increase in their own path. CNBC's read of the grid put it at 16 of 18 expecting another rise, four of them seeing two.

The curve flattened, which is not the same as a selloff

The obvious expectation from a hawkish set of projections is that yields rise across the board. They did not. Using the US Treasury's official daily par yield curve, the 16 September close against the 15th:

The 2s10s spread narrowed from 33 to 27 basis points on the day. The front end absorbed the new policy path; the long end did not move, and the very long end fell.

That shape recurs every time a central bank tightens into an inflation problem rather than a demand boom. A long-dated yield is roughly the expected average of future short rates plus a term premium that includes compensation for inflation risk. When a committee credibly commits to more near-term tightening, two things happen at once: the expected path of short rates rises, pushing long yields up, and the compensation demanded for inflation running away falls, pushing them down. When the second effect roughly cancels the first, you get a 7 basis point move in the 2-year and a 1 basis point move in the 10-year. A market that disbelieved the Fed would have sold the long end hardest.

That last step is the whole of the FX transmission. A currency does not respond to an inflation print or to a hike everyone expected; it responds to the change in the expected policy path relative to somewhere else. Interest rates are one of the five factors the meter scores across the eight majors, and the two-year sector is where that factor reads a policy expectation most cleanly. The dollar page tracks it for the US, the pound page for a committee that held while its own forecast rose, and and the yen page for the bank that proved the point on Friday.

Why an energy shock splits a committee

Every one of these disagreements is a version of the same argument, and it is a genuinely hard one rather than a disguised political fight.

A supply shock raises prices and lowers output at the same time. A central bank tightening into it is deliberately making the output half worse in order to stop the price half from becoming self-sustaining. The textbook answer is to look through it — the price level steps up once and inflation returns to target mechanically a year later, so long as expectations and wages do not start chasing the step. The dissenters' answer is that you cannot know which world you are in until it is too late to act cheaply.

Treasury Secretary Scott Bessent put the look-through case to CNBC on 31 August: "It is my belief that we've seen a supply shock, and traditionally you don't raise into a supply shock unless you see second- or third-order effects," adding that core inflation "has remained very, very restrained." Fed Chair Kevin Warsh made the other case at Jackson Hole three days earlier, acknowledging that inflation readings had been soft but saying they did "not tell me that underlying trends have meaningfully improved," and that the Committee "must be confident that underlying inflation is moving to our objective, clearly and at sufficient speed. Otherwise, we have work to do."

The August CPI report is the evidence both sides used. The headline index rose 0.4% on the month to 3.4% annually, with gasoline alone — up 3.9% — accounting for more than a third of the increase, while the annual core rate was 2.4%. One report, two readings: an energy-driven headline that is not the Fed's target, and a core that ticked up when it was supposed to keep falling. Those figures are confirmed in the Bureau of Labor Statistics' own consumer price index series. Wednesday resolved which reading the Committee acts on — for now, and by twelve votes to nothing.

What actually happened at the Bank of England

The Monetary Policy Summary and minutes of 17 September record that the Committee voted by a majority of 6–3 to maintain Bank Rate at 3.75%, with Megan Greene, Catherine L Mann and Huw Pill preferring an increase of 0.25 percentage points to 4%. That is the same margin, on the same proposition, from the same three names as on 29 July.

Between those two meetings a great deal changed except the count. The minutes put Brent crude at $106 a barrel and UK wholesale gas at 207 pence a therm at the close on 14 September — rises of 36% and 78% respectively since the July Monetary Policy Report. Ofgem's headline energy price cap for October to December was set at £1,723, higher than the Committee had assumed in July, and is expected to rise substantially again in 2027 Q1. On those prices, Bank staff now expect CPI inflation to reach around 3¾% in 2026 Q4, against the 3.2% projected in July, and slightly above 4% in 2027 Q1.

A forecast that moves by more than half a percentage point and a vote that does not move at all is the thing worth explaining.

Start with the print. The ONS reported August CPI at 3.1%, up from 2.9% in July and the highest since March, with transport and motor fuels the largest upward contribution and core unchanged at 2.6%. A headline more than a full point above target mechanically triggers the exchange of open letters between the Governor and the Chancellor that was published alongside the minutes. But the Committee's own decomposition does the holders' work for them: of the 1.1 percentage point overshoot, around 0.7 points came from the direct effects of energy prices, mostly motor fuels. Services inflation was 3.4% in August, unchanged from July and down from 4.5% in March. Private sector regular Average Weekly Earnings grew 2.9% in the three months to July, down from 3.3% at the start of the year — though the Committee judges underlying private sector wage growth to be nearer 3½%, slightly above the rate it considers consistent with the target.

So the domestic gauges are still easing while the headline climbs. The minutes state the position plainly: there has been "little evidence so far of material second-round effects in price and wage-setting," but the risk of them "is greater the longer higher energy prices persist or are more volatile."

Why the count froze while the forecast moved

The Bank publishes each member's reasoning, so the disagreement is legible in a way the Fed's is not. Read together, the nine statements are an argument about one question: does restrictive policy already in place do the job the dissenters want a hike to do?

Alan Taylor, holding, put the arithmetic of that case most directly — Bank Rate is "materially above my estimate of neutral at 3%," and the market curve now sits higher at every maturity from two years out than it did at the final hike of the 2022–23 cycle, when inflation stood at 6.8%. Andrew Bailey held while conceding that if the conflict persists and second-round risks build, "it is likely that policy may have to tighten." Clare Lombardelli held while writing that the case for raising "is building the longer the conflict continues."

The three who voted to hike are not disputing that policy is restrictive. They are disputing what restrictiveness buys you. Megan Greene's statement argues that the slack the holders rely on may already have peaked, and that "waiting for definitive evidence of second-round effects before acting would leave policy behind the curve." Catherine Mann frames a hike as risk management — it "avoids a worse outcome whereby inflation becomes embedded, which requires even tighter policy later." Huw Pill's case is about signalling amid what he calls "the fog of geopolitical conflict and data noise."

That is why the split is frozen rather than drifting. A member who holds because financial conditions are already doing the work has to see the work stop working before switching. On the Committee's own account of the labour market — unemployment estimated at 4.9% in the three months to July, labour demand weak, "still judged to be a margin of slack" — that evidence has not arrived. What has arrived is a growth picture that is quietly firmer: GDP rose 0.4% in Q2, a tenth above the July forecast, and Bank staff now track Q3 at 0.4% against 0.1% projected in July. Resilience is the one thing that moves a holder toward a hiker, which is why Greene's "slack may have already peaked" is the sentence to watch in November.

£488 billion, and the seller that just stepped back

The unanimous decision of the day was not about Bank Rate. On 16 September the stock of UK government bonds held for monetary policy purposes stood at £488 billion, and the Committee voted unanimously to take it to zero — a multi-year plan unwinding the remainder at an average pace of £46 billion a year by the end of 2034, through annual sales of £20 billion alongside maturing gilts.

The headline sounds like more supply. The accompanying Market Notice shows it is mostly the opposite, because it splits the portfolio into three very different buckets.

Two things follow. The first is that the very long end of the gilt curve has been permanently retired from the Bank's sale programme — of its holding of the 1.75% 2049 gilt, £7.8bn is retained to back banknotes and £15.0bn is to be sold, and everything maturing after that is kept. The second is larger: the Bank has paused its APF auctions altogether while it works with HM Treasury and the Debt Management Office on a model in which the gilts are sold to the Government rather than into the market, at market prices and in a pre-announced manner. Progress is to be reviewed before April 2027, so that if it proceeds, the purchases can be folded into the DMO's annual financing remit.

Hold the two halves of that apart, because they pull in opposite directions and only one of them is new.

The stock decision is unchanged in substance: the same £20 billion a year, the same terminal date. Quantitative tightening has not been abandoned. What has changed is the channel. A gilt sold by the Bank into an auction must be absorbed by a price-sensitive buyer, and that buyer demands a concession — a bit more yield, most of it at the maturities being sold. A gilt sold to the DMO, which then issues correspondingly less to the public, nets out the same in consolidated government terms but removes a known, calendar-driven seller from the screen. Duration still has to be placed with someone; it simply stops arriving through a mechanism whose timing everybody can see in advance.

This is the distinction that catches people out. A long gilt yield is roughly the expected average of future Bank Rate plus a term premium. When the premium component falls because a seller withdrew, long yields can decline while the policy path is unchanged — and that is not the same signal for sterling as a decline driven by the market pricing cuts, nor the same as a rise driven by the market pricing hikes. Interest rates are one of the five factors the meter scores across the eight majors, and it is the expectation component, read in the short end, that carries the currency channel. The pound page tracks how that factor is scoring Britain now; the balance-sheet backdrop, and why the gilt market became so sensitive to supply in the first place, is covered in the note on the Bank's Financial Stability Report.

What actually happened at the Bank of Japan

The statement of 18 September records that the Policy Board decided, by a 7–2 majority vote, to encourage the uncollateralised overnight call rate to remain at around 1.25 percent, effective 24 September. By the same 7–2 margin the rate on the complementary deposit facility goes to 1.25 percent and the basic loan rate to 1.5 percent. It is the highest policy rate in Japan since 1995, and it came three months after the previous increase where the earlier steps in this cycle were six months apart.

The reason given is a risk, not a reading. The Board judges that underlying CPI inflation "has been approaching 2 percent" and that "there is a risk that it will deviate upward to a level above the price stability target of 2 percent," citing firms shifting toward raising wages and prices and rising medium- to long-term inflation expectations. That is a pre-emptive hike by a bank whose measured inflation is the lowest of the four majors discussed here: Japan's Statistics Bureau put August national CPI at 1.9% year on year that same morning, with the ex-fresh-food measure at 1.7%, down from 1.8% in July.

None of which was the news, because none of it was unpriced. All 52 economists surveyed by Bloomberg between 4 and 10 September expected the quarter point. CNBC's own survey had almost 90% of respondents on the same side — and, it reported afterwards, they had correctly named the dissenters too.

The dissent changed sides

Read the last column downwards. For two meetings the only member out of step wanted rates higher; in September he got precisely the level he had twice voted for, and joined the majority. The disagreement did not resolve. It crossed the table.

The Bank publishes each dissenter's reasoning in a note attached to the statement. Asada Toichiro dissented "considering that, with the rate of increase in the CPI (all items less fresh food) being below 2 percent recently, it could not necessarily be said that the economic situation was strong, and it was therefore desirable for the Bank to maintain the guideline." Sato Ayano dissented "considering that current economic and price developments did not appear to have substantially accelerated compared to before." CNBC reports that both were appointed to the Board by the Takaichi government earlier this year and are regarded as reflationists.

That is the same structural fact the Fed demonstrated on Wednesday, running in reverse. A committee that grants the minority's demand does not become unanimous; it relocates the argument. In Washington the relocation was into the dot plot. In Tokyo it was into the vote count itself.

Why a rate rise weakened the currency

The ordinary response to a central bank raising rates is a firmer currency, higher bond yields and pressure on equities. On Friday Japan produced the reverse of all three, which CNBC catalogued at the time.

This is not a market malfunction. It is what mechanically happens when the level is fully discounted and the only new information is softer than the consensus assumed.

A currency does not trade a policy rate. It trades the expected path of that rate against the expected path somewhere else, and a move that 52 out of 52 forecasters saw coming is already inside that path before the Board sits down. So the statement's marginal information content was everything except the number — and two of those items read dovish. The first was the vote. "The two dissenting votes in favor of keeping rates unchanged came as a surprise," Hirofumi Suzuki, chief FX strategist at Sumitomo Mitsui Banking Corporation, told CNBC. A board with two members who did not want this hike is, other things equal, a board slower to deliver the next one.

The second was a calendar accident. The Bank publishes its Outlook for Economic Activity and Prices quarterly — January, April, July and October — so a September meeting comes with no revised forecasts attached. Masahiko Loo, senior fixed income strategist at State Street Investment Management, told CNBC this "limited the BOJ's ability to reinforce a hawkish message through revised forecasts." A central bank that wants to raise rates and sound hawkish about the next one usually does it by marking up its own inflation path. This meeting had no path to mark up, and the statement's language tracked July's closely enough that Oxford Economics' Shigeto Nagai described the tone to CNBC as "less hawkish than financial markets had hoped for."

Both banks moved a quarter point, so the gap did not move at all

Here is the arithmetic that the day's headlines obscure. On Wednesday the Federal Reserve raised the federal funds target range to 3-3/4 to 4 percent, a midpoint of roughly 3.88%. On Friday the Bank of Japan raised the call rate to around 1.25 percent. Both moved 25 basis points, in the same direction, inside 48 hours. The differential between the two policy rates — the single largest mechanical input into where a dollar-yen rate sits — finished the week exactly where it began it.

That is the honest answer to why a 31-year high in Japanese rates did not, on the day, produce a stronger yen. Carry is a relative quantity. A currency with a 1.25% overnight rate facing one with a 3.88% overnight rate is in the same relative position it was in at 1.00% against 3.63%. Closing a gap requires moving faster than the other side, and this week neither side moved faster.

That arithmetic is also the backdrop to the official response. On 3 August Japan's Ministry of Finance and the US Treasury conducted a coordinated yen-buying operation, the first joint intervention by the two in roughly fifteen years, described by Al Jazeera and confirmed by both treasuries; Treasury Secretary Scott Bessent said the "coordinated foreign exchange actions countered disorderly yen movements," and Finance Minister Satsuki Katayama said Japan would not hesitate to act again. CNBC reports that at this month's G20 finance ministers' meeting Bessent told Governor Ueda to take "decisive market and monetary steps."

Intervention and rate policy address the same variable through different channels, and only one of them is durable. Selling dollars for yen changes the supply of the currency for as long as the operation lasts; changing the policy rate changes the return on holding it for as long as the rate stays. Neither is a prediction of where the pair goes — what is knowable is that the first works against the flow and the second works with it, and that on Friday the second one did not widen in Japan's favour. Interest rates are one of the five factors the meter scores across the eight majors, and the mechanics of that channel for this currency are set out in the note on what actually moves the Japanese yen, with the intervention history in the note on the yen at a 40-year low.

What to watch from here

At the Bank of England, the thing that has to move before the rate does is the evidence, not the rhetoric. Every one of the six holders wrote some version of the same conditional — Bailey's "policy may have to tighten", Lombardelli's case that "is building", Ramsden's "could be a case" — which means the November meeting turns on whether the specific things they are waiting for show up. Those are nameable: second-round effects in the 2027 pay round, a labour market that stops loosening, and the indirect energy pass-through that the minutes say may have been "delayed rather than diminished". The Committee's own Agents now expect food inflation around 4% at the end of 2026, against 6–7% feared in April; that number drifting back up would do more to move a holder than another motor-fuel-driven headline. Britain's inflation and labour data between now and November are set out in the notes on how the UK CPI release is constructed and what actually moves the pound.

At the Fed, the question has changed rather than closed. A 2026 median of 4.1% and a 2027 median of 4.1% means the next meeting is live and the year after it is not, on the Committee's current view — and December is the next projection round, which is the next scheduled chance for that view to move. The distinction between what the policy rate sets and what it reaches is set out in the note on what the Fed's rate does and does not set; the read of the August inflation data that preceded all of this is in the August CPI note.

At the Bank of Japan, the next scheduled chance for the Bank to move its own forecasts rather than only its rate is the October Outlook Report, and the question the dissenters have put on the table is whether a 1.7% ex-fresh-food print can carry a fourth increase. The statement commits the Board to keep raising "in response to developments in economic activity and prices as well as financial conditions" while watching three named risks — the situation in the Middle East, the expansion in AI-related demand, and developments in foreign exchange rates. No terminal rate has ever been published. Economists quoted by CNBC after the decision put the next move around December: EFG International's Sam Jochim sees roughly quarterly steps toward a peak of 1.75% to 2% in 2027, while Moody's Analytics' Stefan Angrick expects one more around the turn of the year and notes that weak demand-driven inflation and real-wage growth would limit what comes after. Those are forecasts made by other people, not commitments made by the Bank. The build-up to this decision is set out in the July meeting note, and the euro leg of the same week is covered in the note on the September ECB decision.

None of this is a forecast. What is knowable is the mechanism, and three decisions in three days supplied an unusually clean demonstration of it. One committee pre-empted an energy shock and buried its disagreement in a projection round. One waited and published its disagreement name by name, alongside a balance-sheet decision that changes who absorbs gilt supply without changing how much of it there is. One delivered the increase everybody had forecast and watched its currency fall anyway, because the only unpriced thing in the room was a vote count. A rate is one number. What reaches a currency is the distribution behind it — and on Friday the distribution was the whole story. More on how these pieces fit together is on the about page.

Educational macro context only — not investment advice.