In brief - 93.0% of 1,234 decisions matched the economists' consensus.
- Holds are easy to call: 2.0% of them surprised. Moves are not: 23.3% did.
- The European Central Bank surprised least (2.3%), the Reserve Bank of Australia most (13.2%).
- The hardest years to call: 2008 (76.3% matched), when most surprises were lower rates, and 2022 and 2023 (82.4% in each), when most were higher.
- A count of the rate only. It does not measure surprises in a statement, a vote or a forecast.
What was counted
Every decision on FXStreet's economic calendar from Jan 10, 2008 to Nov 3, 2026 by the Federal Reserve, the ECB, the Bank of England, the Bank of Japan, the Bank of Canada, the
Reserve Bank of Australia, the Reserve Bank of New Zealand, the Swiss National Bank, Norges Bank and the Riksbank, where the calendar has both a result and a consensus. A decision
"matched" when the two are the same to the basis point. For the Fed the figure is the top of the target range; for the ECB, the deposit rate.
Decisions without a listed consensus are left out, most of them older ECB decisions, listed on the refinancing rate. The calendar's rows for Norges Bank and the Riksbank start later than 2008. Three calendar rows were corrected first because they were plainly wrong: a consensus
removed for the SNB in December 2018 (it listed -1.25% with the rate at -0.75%) and for the Bank of Japan in April 2024 (it listed 0% on a hold at 0.1%), and the Riksbank's July 2017 result, which the calendar
gave as -0.6%; the bank held at -0.5%.
Bank by bank
Banks that meet more often have more chances to surprise and more easy holds, so the shares are not a ranking of how clear each bank is.
The holds flatter the figure
Most meetings change nothing, and a hold is usually easy to call. Of 946 holds, 19 surprised (2.0%), usually because economists expected a move that did not come. Of 288 hikes
and cuts, 67 did (23.3%), in size or in having happened at all. The rate was below the consensus 50 times and above it 36 times.
Year by year
| Year | Decisions | Away from consensus | Share matched | |
|---|
| 2026 | 62 | 2 | 96.8% | |
| 2025 | 75 | 4 | 94.7% | |
| 2024 | 70 | 6 | 91.4% | |
| 2023 | 74 | 13 | 82.4% | |
| 2022 | 74 | 13 | 82.4% | |
| 2021 | 72 | 2 | 97.2% | |
| 2020 | 70 | 4 | 94.3% | |
| 2019 | 76 | 2 | 97.4% | |
| 2018 | 68 | 0 | 100.0% | |
| 2017 | 66 | 1 | 98.5% | |
| 2016 | 63 | 4 | 93.7% | |
| 2015 | 51 | 5 | 90.2% | |
| 2014 | 46 | 0 | 100.0% | |
| 2013 | 58 | 1 | 98.3% | |
| 2012 | 63 | 4 | 93.7% | |
| 2011 | 63 | 2 | 96.8% | |
| 2010 | 61 | 2 | 96.7% | |
| 2009 | 63 | 7 | 88.9% | |
| 2008 | 59 | 14 | 76.3% | |
The bar is the share that did not match. The 2026 row covers decisions to Nov 3, 2026.
The furthest from the consensus
| Central bank | Date | Consensus | Result | Gap |
|---|
| Bank of England | Nov 6, 2008 | 4.00% | 3.00% | -100 bp |
| Swiss National Bank | Jun 16, 2022 | -0.75% | -0.25% | +50 bp |
| Norges Bank | Mar 13, 2020 | 1.50% | 1.00% | -50 bp |
| Reserve Bank of New Zealand | Jan 29, 2009 | 4.00% | 3.50% | -50 bp |
| Reserve Bank of Australia | Oct 7, 2008 | 6.50% | 6.00% | -50 bp |
| Reserve Bank of New Zealand | Oct 28, 2026 | 3.00% | 2.75% | -25 bp |
| Norges Bank | May 7, 2026 | 4.00% | 4.25% | +25 bp |
| Reserve Bank of New Zealand | Oct 8, 2025 | 2.75% | 2.50% | -25 bp |
| Sveriges Riksbank | Sep 23, 2025 | 2.00% | 1.75% | -25 bp |
| Reserve Bank of Australia | Jul 8, 2025 | 3.60% | 3.85% | +25 bp |
Does a surprise move the currency more?
This site has one-minute candles only since January 2025, so this part is thin. Of the decisions measured since then, 6 surprised and 130 matched. The first 15 minutes after a surprise covered a median of
5.3× the same pair's hour before, against 1.2× after a match. 6 decisions is too few to rest anything on. The pairs are each bank's own
(for the RBA, AUD/USD).
A move that was expected can also go the other way: after 16 hikes in the same period, the bank's currency was weaker an hour later in 9. That is also a small count.
What this does not show
- The consensus is one number. When economists are split, the calendar still lists one figure, so a close call can count as a surprise or a match. Market pricing can differ from it, and is not measured here.
- It is the calendar's figure. We cannot audit how FXStreet gathered each consensus, or when, over 18 years. Three plainly wrong rows are corrected above; others may remain, so check any single row before quoting it.
- The rate is not the whole decision. A hold with a changed statement can move markets more than a hike everyone expected.
- It describes the past. It is not a forecast of any meeting, and not advice to trade one.
Check it yourself
Each bank's rows, with the consensus, are in the CSV on its rate history page, such as the Fed's. The data page lists every file. To report an error, write to
dylan@forexiapro.com; what we fix goes in the corrections log.
To cite: Dylan Forexia, "How often do central banks surprise?", InterestRateDecision.com, data to Nov 3, 2026. https://interestratedecision.com/blog/how-often-do-central-banks-surprise/
Questions
Central bank surprises: questions answered
How often do central banks surprise the markets?
Measured against the economists' consensus in FXStreet's calendar, 86 of 1,234 decisions by ten central banks since 2008 came out away from it: 7.0%. A surprise here means the rate itself differed from the consensus. A statement, a vote or a forecast can surprise on a day the rate does not, and is not counted.
Which central bank surprises the most?
Of the ten, the Reserve Bank of Australia: 26 of its 197 decisions came out away from the consensus (13.2%). The European Central Bank surprised least, 2 of 87 (2.3%).
When were central banks hardest to predict?
2008, when 76.3% of decisions matched the consensus and most surprises were lower rates, then 2022 and 2023 (82.4% in each), when most were higher rates. Over the whole period, 93.0% matched.
Can I quote or reuse these figures?
Yes, with a link to this page. Say they are counts of past decisions against FXStreet's listed consensus. Questions about the method go to Dylan Forexia, dylan@forexiapro.com.