The LIBOR settlements, 2012Medium

A number that set payments on contracts all over the world was produced by asking banks a question rather than by observing a market. The reform that followed was a change of method, not of enforcement.

4 min read · 703 words

What happened

• The benchmark — a set of daily reference rates in several currencies and maturities, produced by asking a panel of banks the rate at which they believed they could borrow unsecured, trimming the extremes and averaging the rest.
• What it referenced — floating-rate loans, mortgages, floating rate notes and an enormous volume of interest rate derivatives, all of which paid according to whatever the published figure said.
• 2012 onwards — regulators in several jurisdictions reached settlements with a number of banks, and criminal proceedings followed in some cases. As a matter of public record, findings included submissions influenced by traders' derivative positions, and submissions influenced by how the bank's own funding cost would be perceived.
• The reform — benchmark administration was brought under regulation, oversight of submissions was formalised, and jurisdictions moved to replace the old rates with benchmarks computed from actual overnight transactions. The transition ran for years and required amending existing contracts.

The mechanism

• The input was an opinion, and by then it had to be. Unsecured interbank lending at three, six and twelve months had become rare, so panel banks were being asked to estimate a rate at which they were not actually borrowing. An estimate is a judgement, a judgement has an author, and an author with a position has an interest.
• The conflict was structural rather than incidental. The person answering the question worked at a firm holding contracts that paid according to the answer. That is a governance defect independent of anybody's intentions, and the reform's first move was to separate the two functions.
• A rate that embeds bank credit moves for reasons the borrower did not cause. An unsecured interbank borrowing rate widens when banks look less creditworthy, which shifted the coupon on every referencing contract during a banking crisis. That is a feature in one context and a serious problem in another.
• The trimmed average made a small number of submissions decisive. Discarding the extremes is meant to blunt an outlier; it also means the remaining submissions carry the whole weight, so a modest shift in one of them is not diluted the way a large sample would dilute it.
• An input to a published figure is worth more than the figure. This is the general case, and it is why benchmark manipulation is treated as its own category in market abuse: influencing a reference that thousands of contracts point at is leverage of an unusual kind.

What it teaches

• Ask what a published number is computed from. A transaction-based rate and a survey-based rate answer the same question with different epistemology, and only one of them can be checked against something that happened. Reference rates is the mechanism in full.
• A market that thins out does not announce it. The benchmark carried on being published exactly as before while the activity underneath it disappeared. The publication schedule is not evidence that the input still exists.
• Separate the person who submits from the person who benefits. The general rule this produced applies well beyond benchmarks — it is the same argument that separates a fund's manager from its administrator.
• Infrastructure changes slowly and that is the cost of being infrastructure. Replacing a rate written into millions of contracts took years, required a fixed spread adjustment computed from history, and needed legislation where consent was impossible.
• A conflict that has always been there is still a conflict. The arrangement was long-standing and widely known, which made it feel like a fact about the world rather than a design choice somebody could revisit.

The mechanisms behind this

• Reference rates — submission-based against transaction-based, and the fallback problem.
• Market abuse — why benchmark manipulation is its own category.
• Interest rate swaps — the contracts that referenced it in the largest volume.
• Curve construction — what these rates feed once they are published.