Interest rate risk in the banking book

The economic value of equity discounts all future cash flows from assets, liabilities and off-balance-sheet positions and asks what rates do to the net present value. It is a long-horizon measure and it captures a maturity mismatch in full: a thirty-year fixed-rate mortgage book funded by overnight deposits shows up here immediately.

Net interest income compares the interest earned and paid under a shock against a baseline, over a one-year or two-year horizon, usually with assumptions about new business. It is the measure a management board feels, because it hits the profit and loss account in the current year, and it can move in the opposite direction from EVE. A bank whose assets reprice quickly gains income when rates rise and loses little present value; a bank with long fixed assets does the reverse. Measuring only one of the two hides the half of the risk the bank happens not to look at.

Article 98 of Directive 2013/36/EU requires supervisors to act where a bank's IRRBB exceeds a defined level, and the EBA technical standards set two tests. The EVE outlier test triggers where a shock reduces the economic value of equity by more than 15 percent of Tier 1 capital. The NII outlier test triggers where a shock reduces net interest income by more than 2.5 percent of Tier 1 capital.

The EVE threshold was tightened from the earlier 20 percent of own funds to 15 percent of Tier 1, which is a stricter test on a smaller capital base. Breaching a test is not a breach of law; it puts the bank in a conversation with its supervisor that can end in an additional own funds requirement, a limit on the activity producing the exposure, or a prescribed modeling assumption. That last outcome is the one banks work hardest to avoid, because it removes their own judgment from the measurement.

For the EVE test a bank applies six scenarios to each currency: a parallel shock up, a parallel shock down, a steepener with short rates down and long rates up, a flattener with the reverse, a shock to short rates up, and a shock to short rates down. The shock sizes come per currency from the standards, so a euro book and a dollar book are not shocked by the same number of basis points.

For the NII test two scenarios apply, a parallel shock up and a parallel shock down. The asymmetry is deliberate: the earnings measure is sensitive to the level of rates over the horizon, and the non-parallel scenarios add less there than they do to a present value calculation.

Alongside the internal measurement system a bank may be required, or may choose, to use the standardized approach the EBA technical standards define, with a simplified version available to small and non-complex institutions. The standardized approach is also what a supervisor can impose where it finds the internal system inadequate.

Credit spread risk in the banking book is the risk from changes in the market credit spread and the market liquidity spread, separate from the risk that a specific borrower defaults and separate from a general interest rate move. The EBA guidelines require a bank to identify, measure and monitor it, and they set the scope wide: all banking book assets and liabilities are in scope, and a product may be excluded only where the absence of sensitivity to credit spread risk is proven.

That proof requirement is where the work sits. A bank cannot exclude its loan book from CSRBB because loans have no market price; it has to show why the position carries no market spread sensitivity, or bring it into the measurement. Instruments held at fair value are inside the scope in every case. The distinction from credit risk matters too, and our page on credit risk in German banking covers the borrower-default side.

Two positions have no contractual maturity and therefore have to be modeled. Non-maturity deposits can be withdrawn at any moment and in practice stay for years, so a bank splits them into a stable core and a volatile portion and assigns a repricing profile to the core. Loans with a prepayment option behave differently as rates move, because a borrower refinances when it pays to.

The guidelines cap the behavioral repricing maturity of certain non-maturity deposits at five years, which limits how far a bank may assume its deposits are long-term funding. That cap changed hedging practice directly: a bank that had modeled a ten-year core deposit profile had to shorten it, and the swap book behind the assumption with it.

These assumptions are model risk in the ordinary sense, and they belong in the model inventory with a validation history. The deposit behavior observed in 2022 and 2023 invalidated models calibrated on a decade of near-zero rates, which is the practical argument for backtesting a behavioral model against a period that contains a rate move.

Unrealized losses on securities held at amortized cost. When rates rose, bond portfolios bought at low yields fell in value, and an institution holding them to maturity recognized no loss in its accounts while the economic value was gone. German supervisors tracked this closely: the Bundesbank Financial Stability Review reported on the hidden burdens in the banking system's securities holdings and on the interest rate risk carried by savings banks and cooperative banks in particular.

The link to liquidity is the part that proved dangerous elsewhere. A portfolio with an unrealized loss is still liquid, but selling it crystallizes the loss, so a bank under funding pressure faces a choice it did not model. Our page on liquidity stress testing covers that side, and the two risk types are measured in the same scenario for exactly this reason.

Through the supervisory reporting framework, on the EBA implementing technical standards for IRRBB reporting, which became compulsory from September 2024. A German bank files through the Deutsche Bundesbank, which collects banking supervision reports and passes them to BaFin and, for a significant institution, to the ECB.

The templates carry the EVE and NII results per scenario and currency, the outlier test outcomes, the modeling assumptions and the volume data behind them. Separately, the earlier Basel interest rate shock notification under German law retains its role for institutions outside the full framework, and a bank should check which reporting population it sits in before building a template.