What was LIBOR, and why did the world have to replace it?
First: what a reference rate is
A reference rate is a published number that thousands of contracts point at instead of negotiating a rate each time. A floating-rate loan says "the benchmark plus 150 basis points." A swap exchanges a fixed rate for a published floating one. A floating-rate note resets its coupon to a fixing.
This is why benchmark integrity is not an academic concern. A reference rate is not a price you choose to trade at; it is a fixing that determines cash flows for people who never traded at it.
The arithmetic of one basis point. A $50 million loan pays the benchmark plus 150bp. If the benchmark averages 4.20% over a 90-day period, the coupon is 5.70%:
50,000,000 × 0.0570 × 90 / 360 = $712,500
Move the fixing by a single basis point and the interest changes by:
50,000,000 × 0.0001 × 90 / 360 = $1,250
Trivial on one contract. Now scale it across hundreds of trillions of notional. That single line explains both why manipulation was worth attempting and why benchmark administration is now a regulated activity.
LIBOR: an estimate, not a transaction
The London Interbank Offered Rate was administered by the British Bankers' Association and, from 2014, by ICE Benchmark Administration. Each business day at 11am London, panel banks answered a question: at what rate could you borrow funds, were you to do so by asking for and then accepting interbank offers in a reasonable market size? The top and bottom quartiles of submissions were trimmed and the remainder averaged. At its widest, LIBOR was published in five currencies across seven tenors, from overnight to twelve months.
Read the question again. "Could you borrow." It asked for an expert judgement, not a record of a trade. That was tolerable while the underlying market was busy. It stopped being tolerable.
Two failures, not one
Failure one — misconduct. Because submissions were internal judgements, nudging one by a basis point cost the submitter nothing and could be worth a great deal to a large derivatives book. From 2012, investigations documented traders at panel banks asking their own submitters to shade the number. Separately, during 2007 to 2009, some banks submitted artificially low rates to avoid signalling distress — a bank that admits it is being charged more than its peers is a bank people stop lending to. Industry fines ran to roughly $9 billion: Barclays about $450 million in June 2012, UBS about $1.5 billion in December 2012, RBS about $610 million in February 2013, Deutsche Bank about $2.5 billion in April 2015. Several individual traders were convicted in the UK and US; some of those convictions were later overturned on appeal.
Failure two — and this is what actually killed it — the market underneath had disappeared. After 2008, banks largely stopped funding themselves through large unsecured three-month interbank loans. They used repo, deposits and central bank facilities instead. So a benchmark referencing an estimated $400 trillion of contracts at its peak was resting, on many days, on almost no transactions at all. Perfect ethics would not have fixed that. The pyramid was enormous and its base had eroded.
The wind-down
The Wheatley Review of 2012 reformed governance and criminalised manipulation. In July 2017 the UK's Financial Conduct Authority announced it would not compel panel banks to submit beyond the end of 2021. Most settings — all sterling, euro, Swiss franc and yen tenors, plus one-week and two-month US dollar — ceased after 31 December 2021. The remaining US dollar settings ceased after 30 June 2023, with a temporary "synthetic" US dollar rate published under FCA compulsion until 30 September 2024. Legacy contracts that could not be renegotiated were handled by statute; the US LIBOR Act of 2022 substituted a SOFR-based rate by law.
The lesson is not "surveys are evil"
EURIBOR did not die. It was reformed onto a hybrid, transaction-anchored waterfall methodology and is still published today. The principle that survived the whole episode is narrower and more useful: a benchmark must be anchored in a market that actually exists, and its governance must be robust enough that no participant's judgement can move it.
In the data
The replacement's defining property is what gets published beside the rate. Here is one SOFR fixing with it:
On 24 September 2026 SOFR fixed at 3.88%, the median of $2,990 billion of actual repo trades; half the volume traded between 3.86% and 3.92%, and almost all of it between 3.81% and 3.96%. A submitted rate could publish neither line, because there was no volume and no spread of trades behind it, only a panel's answer to a question. And LIBOR itself is no longer published at all.
Try it now
- Below is every dollar reference rate for one day, 24 September 2026. List every one that is a single overnight fixing rather than an average or an index, then note what is absent: LIBOR is not one of them.
- Open the terms of a floating-rate note issued after June 2023. Supranational issuers publish theirs in full: the World Bank (IBRD) and the European Investment Bank both post the final terms of their dollar floating-rate notes on their investor-relations pages. Identify the benchmark and the spread the note names.
- In one sentence each, state LIBOR's two independent failures — and which one made replacement unavoidable.