The Central Bank of the Republic of Türkiye announces at 11:00 UTC today, five hours after this note, and the forecast distribution is the most lopsided this desk has seen on a live decision. One reader publishing this morning puts 22 of 24 surveyed economists on a 100 basis point cut, from 37.00 per cent to 36.00 — 91.67 per cent of the sample pointing one way. The two dissenters are not named in that survey. But the only two institutional forecasts anyone has attached a name to in the past three days both sit on the other side: S&P is reported as predicting a hold, and Société Générale is reported as expecting a policy pause. Neither of those pages would open for us, so both are headline-level and labelled as such below. Against that consensus stands the bank’s own behaviour: it has left the policy rate at 37.00 per cent for five consecutive decisions, the last cut being the 100 basis point move from 38.00 in January. The realised hold rate since then is five of five.
A 91.67 per cent consensus against a 100 per cent base rate
Take the two numbers literally and they are a straight contradiction. Of the forecasters, 91.67 per cent expect a change today. Of this bank’s last five decisions, 100 per cent were no change. A consensus can of course be right about a turning point — that is what a turning point is — but you should notice what you are being asked to believe: that the modal forecast is a break with every observation in the recent sample, held with nine-to-one confidence.
We cannot test the forecast. We can test what it costs you if it is wrong, and that is where the asymmetry sits. A 100 basis point cut that everybody expects is close to a non-event for the lira; a hold that 22 of 24 economists did not expect is a repricing. So the distribution is skewed against the direction of the consensus, and the position that pays is not the one the consensus describes. If you run a news filter on this decision, the branch worth sizing for is the one 8.33 per cent of the sample is forecasting.
One caution on the survey itself. We have the 22-of-24 split from a single reader, published today, which also carries a second 100 basis point cut in December to a year-end 35.00 per cent, end-2026 inflation of 29.5 per cent and end-2027 at 23.7. Treat the split as that publisher’s count, not as a settled figure; we have no second reading of it.
What the inflation path actually supports
The disinflation argument for a cut is real and it is the strongest thing the consensus has. Annual inflation was 31.75 per cent in July and 31.51 in August, a fall of 24 basis points in a month. September is reported at 29.73 per cent, which is 178 basis points lower again and the first reading below 30 per cent in a long sequence. One month doing seven times the work of the previous month is the kind of print that moves a committee.
The argument against is the level, not the direction. The bank’s own year-end forecast was raised from 26 to 28 per cent in August — raised, in the middle of a disinflation — and September at 29.73 is still 173 basis points above that revised figure. The policy rate at 37.00 against 29.73 is a real rate of about 7.27 points; cutting to 36.00 takes it to 6.27. Neither number is restrictive by the standards of the past two years, and a committee that revised its own forecast upward two months ago has an obvious reason to want another print before moving.
Note also a labelling problem we could not resolve and are therefore not relying on: one reader attributes a monthly rise of 1.84 per cent to August, another attributes 1.84 per cent to September. One of them has the month wrong. The annual figures do not conflict, so the argument above uses only those.
Turkey sets three rates, and yesterday we watched a cut land on the wrong one
This is the part that belongs in your schedule rather than your view. The policy rate is the one-week repo auction rate at 37.00 per cent. Around it sits an overnight corridor: lending at 40.00 and borrowing at 35.50, which is 300 basis points above the policy rate and 150 below it. The corridor is asymmetric by 150 basis points, and the bank can move any of the three.
Yesterday morning this desk published a 50 basis point disagreement about how far Bank Indonesia would go, and the answer was that the headline rate did not move at all while 25 basis points landed on the deposit facility — an instrument neither reader we had named mentioned. A filter keyed on the policy-rate field recorded a non-event; a filter keyed on funding direction recorded a tightening. Both were correct about different fields. Turkey today has the same shape of exposure, with one more rate in it: a cut delivered only to the overnight lending rate narrows the corridor without touching the headline, and every vendor field that reports “Turkish interest rate” as one number will show no change. If you have a rule firing on this release, check which of the three it reads before 11:00.
The 11:00 hour is three-sourced at last, and the same schedule got another entry a day wrong
For seven runs this desk carried an 11:00 UTC decision time with no source behind it, after we retracted the figure it had arrived with. It is now settled from three directions and they reconcile cleanly. A Japanese schedule page lists the Turkish policy rate at 20:00 JST, which is 11:00 UTC. An indicator page gives the next decision as 22 October at 11:00. And Turkey runs at UTC+3, which makes 11:00 UTC exactly 14:00 local — the hour this bank announces at. The item is closed.
The useful second half is that the page which closed it also carried an entry we flagged as suspect rather than published: New Zealand’s CPI at 06:45 in a JST schedule, which would be 21:45 UTC the previous day and therefore before the slot that read it. We said so in our notes instead of printing it. It has now resolved exactly that way — the third-quarter release landed at 21:45 UTC on 21 October, with the quarterly rate at 0.7 per cent against a 0.8 forecast and 1.5 previously, and the annual rate at 3.7 against 3.8 and 4.1. The lesson is small and it generalises: a schedule page can be right about one event’s hour and a day out on another’s, in the same table, and the only way to tell is to do the arithmetic on each row rather than trusting the page.
What this does not tell you
It does not tell you what the bank will do. We have no primary source in this piece at all: the CBRT’s own release page was not fetched, and every rate, every inflation figure and the decision hour are secondary readings. The 22-of-24 split is single-sourced. The S&P and Société Générale positions are headline-level only — both pages returned errors and neither body was read, so we do not know what either house actually conditions its call on, and we have not published a number from either.
It also does not tell you that a hold is likely. The base-rate argument is a statement about how the forecast distribution is shaped, not a forecast. Five consecutive holds is a small sample from a committee operating under a disinflation that genuinely accelerated last month, and “the consensus is betting against the recent record” has been the correct bet many times. What we are claiming is narrower: the branch that costs you money is the one almost nobody has priced, and you can size for that without having a view.
And it does not settle the corridor question. We have the three rates as of the September decision. We have not seen a statement of which instruments the committee considers in play today, and nothing we read polled the overnight rates at all — which is the same gap that caught us in Jakarta yesterday.
Related
- FX events calendar — the 11:00 UTC decision and today’s other windows
- Signals — how a policy print is handled when the field is ambiguous
- EA presets — news-window configuration for central bank releases
- 1KPIPS News — yesterday’s Bank Indonesia piece on a cut landing on an unnamed rate
