How Much Do Official Exchange Rates Disagree? We Measured 31 Central Banks for a Year
Ask thirty central banks for the official EUR/USD rate on the same day and you will not get one answer. You will get thirty — and they are further apart than most people writing FX policies assume. We ingest the published rate tables of 35 central banks daily, which makes the question directly measurable, so we measured it: every same-day EUR/USD rate implied by every bank's own published table, over 264 trading days from August 2025 to August 2026.
Key findings
- The same-day official EUR/USD differed by a median of 142 basis points (~1.4%) between the lowest and highest of ~30 banks; the widest day hit 615 bp (~6%).
- Even across 12 major banks (ECB, Fed, BoE, BoC, RBA, RBNZ, BoJ, Riksbank, Norges, CNB, NBP, SARB), the daily spread was a median 32 bp — on a €10M position, that is ~€32,000 of pure methodology.
- ECB vs Fed: 16 bp apart on an average day; 42 bp on a bad one; 131 bp at worst.
- The driver is timing, not quality: the Fed and BoE fix ~1 hour apart and differ by only 6.4 bp on average — while the ECB, fixing 4–5 hours earlier, sits ~16 bp from both.
Method, in one paragraph
For each bank and each trading day we computed the EUR/USD rate implied by that bank's own published table: directly where the bank quotes the pair, otherwise as the cross via the bank's home currency (for example, Norges Bank's NOK-per-EUR divided by its NOK-per-USD). Where a bank publishes only buy/sell, we used the midpoint. Monthly and quarterly publishers (SNB, HMRC, US Treasury) were excluded — the study compares same-day snapshots. For each day with at least ten banks reporting, we took the cross-bank median as the reference and measured every bank's deviation from it in basis points (1 bp = 0.01%). 264 qualifying days, 31 banks. The underlying data is queryable by anyone through the central bank rates API.
The headline: "the official rate" is a range, not a number
The median low-to-high spread across all banks was 142 bp per day — the 90th percentile was 251 bp, and the widest single day (8 April 2026) reached 615 bp, driven by banks whose euro fixings are thin local crosses rather than deep-market readings. Peru's central bank, whose published EUR rate reflects a market where euros barely trade against the sol, averaged 104 bp from the global median — perfectly official, and 1% away from what Frankfurt printed the same day.
Restricting to twelve major, deep-market banks does not make the effect disappear: median 32 bp, p90 59 bp, worst day 133 bp. There is no configuration of "respectable" sources in which the same-day official EUR/USD is a single number.
Timing explains most of it
The cleanest evidence in the dataset: the Fed and the Bank of England differ by just 6.4 bp on average — their snapshots (noon New York, 16:00 London) are about an hour apart. The ECB, which fixes at ~14:10 CET, sits 15–16 bp from both on an average day. Same market, same quality of institution — four extra hours of market movement between the photographs. Divergence between official rates is, to first order, a measure of how much EUR/USD moves between fixing times.
| Pair of banks | Fixing gap | Mean same-day |diff| | p95 | Worst day |
|---|---|---|---|---|
| Fed vs Bank of England | ~1 h | 6.4 bp | 16 bp | 30 bp |
| ECB vs Bank of England | ~3 h | 15 bp | 36 bp | 131 bp |
| ECB vs Fed | ~4–5 h | 16 bp | 42 bp | 131 bp |
The tightest and widest banks
Mean absolute deviation from the daily cross-bank median, over the year:
Closest to global consensus
| # | Bank | Mean |dev| | p95 |
|---|---|---|---|
| 1 | Central Bank of Türkiye | 4.8 | 14.5 |
| 2 | South African Reserve Bank | 6.7 | 20.9 |
| 3 | Hong Kong Monetary Authority | 6.9 | 17.9 |
| 4 | Narodowy Bank Polski | 6.9 | 20.4 |
| 5 | National Bank of Romania | 7.1 | 20.7 |
| 6 | Bank of Israel | 7.6 | 21.4 |
| 7 | Reserve Bank of India | 7.8 | 20.3 |
| 8 | Bulgarian National Bank | 7.9 | 28.6 |
| 9 | Central Bank of Azerbaijan | 7.9 | 22.2 |
| 10 | Central Bank of Sri Lanka | 7.9 | 23.4 |
Furthest from global consensus
| Bank | Mean |dev| | p95 |
|---|---|---|
| Central Reserve Bank of Peru | 103.9 | 230.5 |
| Bank of Russia | 35.7 | 107.9 |
| Central Bank of Bahrain | 34.4 | 191.3 |
| Bank Indonesia | 26.1 | 67.7 |
| National Bank of Ukraine | 21.1 | 59.4 |
The wide end is not a list of bad institutions — it is a list of thin local EUR markets (Peru, Bahrain's EUR cross of a USD-pegged dinar), sanctioned or managed markets (Russia), and administratively set fixings. Their dollar rates cluster far more tightly; it is the euro leg, fixed against illiquid local trading, that drifts. Which is itself a finding: how "official" a rate looks depends on which leg of it you audit.
What this means in practice
For contracts: "at the official exchange rate" is ambiguous by ~1.4% a day across banks — real money at any size. Name the bank, the series, and the date convention. (Our guide to choosing between ECB, Fed, and BoE rates covers how.)
For multinationals: two subsidiaries can each use their impeccable local official rate and produce intercompany figures 32+ bp apart on an ordinary day. Eliminations that "never quite net" are often this, not an error.
For anyone comparing a rate to Google: an official fixing from 14:10 CET is supposed to differ from a live quote at 21:00 — the difference is definitional, not a data problem. (More on that split in official vs market rates.)
Cite or reproduce this study
This analysis is reproducible end-to-end from published central bank data: every input rate is queryable via the AllRatesToday central bank API (the latest tables are free), and each bank's product page documents its source and fixing time. If you reference these figures, please link to this page as the source; we plan to refresh the study annually.
The data behind this study
Official rates from 35 central banks, as JSON — latest tables free, no credit card.
Get your free API key