CSRBB: Why Credit-Spread Risk Matters Beyond Traditional Market Risk

What CSRBB is, and what it is not

When a bank holds a bond or a loan, the yield it earns can be decomposed into a risk-free component, a compensation for expected credit loss on that specific obligor, and a market-priced spread reflecting the general price of credit risk and of liquidity at that moment. The first is interest-rate risk in the banking book. The second is credit risk, managed through provisioning and capital. The third is credit-spread risk in the banking book.

The distinction is not merely taxonomic. During a spread widening event, a portfolio of bonds can lose substantial value without any deterioration in the credit quality of the issuers and without any movement in the risk-free curve. If the institution measures only rate risk and only obligor credit risk, that loss is invisible in the risk framework until it appears in the accounts.

Scope questions

• Asset-side instruments valued at fair value through other comprehensive income are the clearest case, since spread movements flow to own funds.
• Amortised-cost assets still carry CSRBB in economic terms, even where accounting does not recognise the movement immediately.
• Liability-side spread risk — the bank's own funding spread — is treated differently across jurisdictions and requires an explicit policy decision.
• Positions already captured in the trading book or in specific credit-risk frameworks should not be double counted.

Measurement challenges

The core difficulty is that the decomposition of an observed yield into components is a modelling choice rather than an observation. The market quotes a price; it does not quote how much of the spread is compensation for expected loss and how much is a liquidity premium. Different reference-curve choices — swap curves, government curves, overnight index swap curves — produce different measured spreads for identical positions.

1. Select a reference risk-free curve and document why it is appropriate for the currency and instrument type.
2. Determine the expected-loss component using the institution's own credit parameters, so that what remains is genuinely market-priced spread.
3. Group positions into buckets with comparable spread behaviour — by rating, sector, seniority and currency — rather than applying a single shock.
4. Apply shocks calibrated to observed historical spread movements for each bucket, including at least one episode of severe generalised widening.
5. Report both the value impact and, where relevant, the earnings and own-funds impact through the applicable accounting treatment.

Illiquid and unquoted positions present a further problem: there is no observable spread to shock. Proxying from comparable quoted instruments is standard practice, but the proxy relationship itself should be tested and its uncertainty disclosed rather than assumed away.

Interaction with valuation and capital

The transmission channel from spread movement to capital depends on accounting classification. Instruments measured at fair value through other comprehensive income move own funds directly through the revaluation reserve, subject to any prudential filters that apply. Instruments at amortised cost do not, unless they are sold or become credit-impaired — which is precisely the circumstance in which a liquidity-driven sale would crystallise the loss.

This creates a coupling between liquidity risk and credit-spread risk that is easy to miss. A liquidity buffer of high-quality securities held at amortised cost provides no capital volatility in normal conditions, but monetising it during a widening event realises the loss at the worst moment. Sizing and classifying the buffer is therefore a joint liquidity, spread and capital decision.

Governance and reporting

• Document the definition of CSRBB adopted, the scope of instruments included and excluded, and the rationale for each exclusion.
• Set limits on measured spread sensitivity, with escalation thresholds and named owners.
• Report CSRBB alongside IRRBB so that the committee sees the combined banking-book market exposure rather than two partial views.
• Subject the spread-decomposition methodology to independent model validation; it is a model, even where implemented in a spreadsheet.
• Review bucket definitions after any episode in which spreads within a bucket behaved heterogeneously.