Basis risk
Regulatory Definition
Risk arising from the impact of relative changes in interest rates on interest rate sensitive instruments that have similar tenors but are priced using different interest rate indices. Basis risk arises from the imperfect correlation in the adjustment of the rates earned and paid on different interest rate sensitive instruments with otherwise similar rate change characteristics.
EBA GL/2022/14
What This Actually Means
You've got assets and liabilities that both float, but they're priced off different benchmarks. Say your lending is priced off the Bank of England base rate while your funding is on SONIA. Both are overnight, so there's no tenor mismatch between them — but the spread between base rate and SONIA can still move, and when it does, you're exposed.
CRITICAL: before measuring basis risk, work out what your gap framework has already captured — above all, product beta. When you reprice deposits at a behavioural beta, the typical parallel and Basel scenarios already produce the deposit-versus-market spread, so we count that as gap risk, and the chance the beta is wrong as beta risk. Basis is then left to pick up only the genuine index spreads not already embedded in your repricing assumptions.
Where It Matters
First, a word on product beta, because it sits in a grey area and is the most common thing people fold into basis risk. Two different effects hide inside it. One is the spread the beta throws off within a scenario: model a deposit repricing at 60% of a base-rate move and a 100bp rise lifts the deposit rate by 60, leaving a 40bp spread. That spread is already in your earnings number, so we treat it as gap risk, not basis. The other is the chance the 60% is wrong, unstable or non-linear — the movement in the beta itself. Some regulators call that basis risk outright — the HKMA's MA(BS)12 return measures it with scenarios that move managed rates 200bps independently of all other rates, capturing the managed-rate divergence directly. We'd treat the same move as gap risk (the spread it throws off in the scenario) and beta risk (the chance the assumption is wrong), a practitioner concept rather than a regulatory category. Same exposure, different label — which is why scope is worth pinning down before you measure.
Once the beta is handled that way, very little additive basis risk is left. The timing in a tenor mismatch (3-month against 6-month, say) is gap risk too — a parallel shock through your repricing buckets has already picked up much of it, up to three months of repricing timing, because the two legs reset on different dates. And materiality runs the same way: gap risk takes a much larger shock than basis — a parallel move of a couple of hundred basis points, against a far smaller basis stress — so the gap risk from a tenor mismatch usually dwarfs whatever basis is left. One-month against three-month LIBOR was a classic basis in the trading book; in the banking book the tenor gap shock largely swamps it. Strip all of that out and the residual basis risk collapses to one thing: the spread between two market indices moving independently. The cleanest cases are overnight against overnight, where there's no tenor for gap risk to have captured and the whole exposure is the spread. Two examples: base rate against SONIA within sterling, and SOFR against SONIA as a cross-currency basis. In both, the legs reset on the same overnight frequency, so what is left is pure index spread.
There's a practical wrinkle, too: in the banking book, basis risk is far less documented than gap risk. Market risk has long had the tooling to measure and hedge basis; IRRBB is still catching up. Gap has well-worn metrics — repricing gaps, PV01, EVE and NII sensitivities — with years of supervisory expectation behind them. The boundary was argued over in the drafting: after the EBA's 2018 guidance muddled basis with the “timing difference neglected by gap analysis,” the EBF pushed to drop the “similar tenors” restriction — but the 2022 definition kept it, settling basis as a same-tenor, different-index risk and leaving the tenor difference to gap. The most concrete measurement standard is the EBA's standardised approach (Article 21 of its 2022 SA RTS): it buckets floating instruments by reference term — overnight out to 12 months, plus a policy-rate and an 'other' category — and shocks each term against overnight to produce a net interest income add-on. Tellingly, that add-on only bites where non-overnight floating tops 5% of assets, it adds to the repricing calculation (Article 6) rather than replacing it, and it takes the lower of the two shock scenarios. But the shocks are bank-estimated and the add-on is SA-specific, so internal-model practice still varies — two banks with the same exposure can measure it quite differently.
Two cautions pull in opposite directions. Where the same exposure is counted twice — the partial pass-through sitting in gap risk and then again in basis — a Pillar 2 number that sums the two without reconciling the overlap may overstate the risk. But where the risks are genuinely distinct, a regulator may fairly take the other view: one balance can carry gap risk from its tenor mismatch and basis risk on top, because the two curves may not move in unison, and may expect both to be capitalised in the ICAAP. The open question is simultaneity. Capitalising for a parallel shock and a basis widening at the same time is a shock on top of a shock, which assumes the two coincide — so whether to hold capital for both in full, or for the most material and take some diversification, is something banks settle in their own internal framework.