The European Central Bank main refinancing rate, 1999 to 2026: every change

The line above is a step, not a curve. The rate holds one value until a decision moves it, so the flat stretches are not missing data — they are the data. Moving the cursor across the chart reads the rate on any date; the buttons narrow the window to the last year, three, five or ten.

The series holds 10,084 daily observations, and it repeats the same number every day until it changes, because that is how Banca centrale europea, Data Portal. Due serie ufficiali unite perche' descrivono lo stesso tasso in due regimi d'asta diversi: a tasso fisso fino al 27 giugno 2000 e dal 15 ottobre 2008 (MRR_FR), a tasso variabile con tasso minimo di offerta fra le due date (MRR_MBR). Le due serie non hanno una sola data in comune: 7.053 + 3.031 = 10.084 punti su 10.084 giorni, senza salti. publishes it. Keeping the repeats rather than collapsing them to the change dates matters: it is what lets the chart show how much of the period was spent at each level, which is the part that gets forgotten.

The highest the rate reached in this record is 4.75%, on October 6, 2000. The lowest is 0.00%, on March 16, 2016. It began the record at 3.00% and ends it at 2.40% — −60 basis points over 28 years.

For 6 years and 4 months of that period the rate sat at zero or below. That stretch is worth naming rather than smoothing over: it is a state that had no precedent in the record before it happened, and any average taken across the whole period is pulled by it.

The median level across every day in the record is 2.00%, and the rate spent 41% of the period above it. A median is a better middle than a mean here: the rate spends long spells parked at one number, so an average is dominated by how long the flat stretches lasted rather than by where the rate usually was.

Of the 60 changes in the record, 29 were increases and 31 were cuts. The largest single increase was +75 basis points on September 14, 2022, taking the rate from 0.50% to 1.25%; the largest single cut −75 basis points on December 10, 2008, from 3.25% to 2.50%.

Changes are not evenly spaced. Averaged over the record there is one every 6 months, but that average describes nothing that happened: the rate moves in runs — several decisions close together, then years of nothing. The column showing how long the previous level had held makes those runs visible.

The longest the rate held one level is 6 years and 4 months, at 0.00%, from March 16, 2016 to July 26, 2022. Long flat stretches are the normal state of a policy rate and the part a chart of "recent changes" hides completely.

The current level, 2.40%, has held since June 17, 2026 — 2 months at the last point in the record. Whether that is a long time depends on what it is compared with, which is why the table below lists every stretch rather than describing this one on its own.

Year by year: of the 28 years the record covers in full, 7 ended higher than they began, 11 ended lower, and 10 ended exactly where they started. That last count is the one that surprises people — a year in which nothing changed at all is a common outcome, not an anomaly.

A policy rate does not move one step and stop. It moves in runs: several decisions in the same direction, then a pause. The record contains 10 such runs — 5 upward and 5 downward — and reading them as runs rather than as 60 separate events is the difference between seeing a pattern and seeing noise.

The widest of them ran upward from 0.00% to 4.50% — +450 basis points across 10 decisions between July 27, 2022 and September 20, 2023, 1 year and 2 months from the first move to the last. A run of that size is not visible at all in a table of individual changes, and it is the thing most likely to have mattered to anyone borrowing through it.

The table lists every run. Where a run contains a single decision it means the direction reversed immediately afterwards, which is itself informative: reversals of that kind are rare in this record, and the ones that happened are visible at a glance.

Another way to read the same record: not when the rate changed, but where it spent its time. The level it held longest is 0.00%, for 6 years and 4 months in total in a single unbroken spell. In all, the rate has taken 28 distinct values in this record.

15 of those levels were visited more than once — the rate left them and came back. That matters when a number is described as unprecedented: in this record several levels that felt like a floor or a ceiling at the time were revisited later, and the table shows which.

Decisions are not spread evenly across the years either. The busiest decade in this record is the 2000s, with 29 changes; the quietest the 1990s, with 2. A decade is long enough that the difference cannot be explained by chance in the calendar — it reflects how much the conditions the rate responds to were themselves moving.

Reading the table downward also shows something a single number cannot: whether a decade was mostly a tightening decade, mostly an easing one, or genuinely two-sided. The count of increases against cuts in each row answers that directly, and the level the decade closed at says where it left the rate for the one that followed.

The record begins on January 1, 1999 and cannot say anything about what came before it. It also cannot say why any of these decisions were taken: the series carries dates and levels, not reasons, and attaching an explanation to each move would mean adding something that is not in the data.

A single policy rate is also not the rate anyone pays or receives. What a household is charged on a mortgage, or paid on a deposit, is set by a lender and moves with this rate only partly and with a delay. Reading a change here as a change in your own borrowing cost would be a mistake this page would rather you did not make.

And nothing here is a forecast. The table of past changes shows how often and how far this rate has moved; it contains nothing that determines the next decision, and this site has no view on what that decision will be.

The series is Tasso sulle operazioni di rifinanziamento principali della BCE, identifier FM.D.U2.EUR.4F.KR.MRR_FR.LEV + FM.D.U2.EUR.4F.KR.MRR_MBR.LEV, published by Banca centrale europea, Data Portal. Due serie ufficiali unite perche' descrivono lo stesso tasso in due regimi d'asta diversi: a tasso fisso fino al 27 giugno 2000 e dal 15 ottobre 2008 (MRR_FR), a tasso variabile con tasso minimo di offerta fra le due date (MRR_MBR). Le due serie non hanno una sola data in comune: 7.053 + 3.031 = 10.084 punti su 10.084 giorni, senza salti.. It was read on August 13, 2026 and the file sits in this site's repository, which is why every language edition of this page shows the same numbers and why the date of the last observation is stated rather than implied.

The source is linked below. Pull the same series and the tables here should reproduce exactly — there is no smoothing, no interpolation and nothing dropped. If they do not reproduce, that is an error on this page, and it will be published as a correction rather than quietly edited away.