3.2 Non-Default Spread

3.2 Non-Default Spread risk

Non-default Spread Risk (NDSR) aims to capture unexpected changes in the level or volatility of spreads over the risk-free interest rate term structure, excluding the default component which is captured in Credit risk.

NDSR charge is calculated as a relative bi-directional stress applied to both assets and liabilities. The NDSR charge is calculated as the maximum of an upward and downward stress, subject to a floor of zero.
All liabilities sensitive to changes in spreads are taken into account in the calculation of the NDSR charge, with the exception of financial instruments issued by the IAIG that qualify as capital resources.
All assets that contribute to the calculation of the spread adjustments for valuation purposes, are taken into account in the calculation of the NDSR charge, with the exception of sovereign assets.
The upward and downward stresses used for the calculation of the NDSR charge are a relative stress of -75% and +75% of spreads at each maturity up to the Last Observable Term (LOT).

Initially, the NDSR charge was set as an absolute bi-directional shock (up/down) to the balance sheet affecting both assets and liabilities, capped by a relative limit. The assumption made in the calibration was considering half of the observed spreads were not implied by default losses and should therefore be taken into account in the NDSR charge.
The results of the 2018 Field testing exercise showed unjustified discrepancies between NDSR charges for different currencies, which were driven by the underlying differences in the spread levels for the different currencies. Based on expert judgement the NDSR charge was subsequently lowered, assuming that only a quarter of the spread should be used for the NDSR charge calibration.
The calibration work was performed using historical data. The Spreads of Markit Corporate bond indices (iBoxx), with the following date ranges of the input data were considered:

a.EUR: 01/06/2004 -31/12/2016
b.GBP: 02/01/2003 -29/12/2017
c.USD: 04/01/1999 -29/12/2017
– Based on that data using rolling averages, a historical 1-year 99.5% VaR was calculated. The calibration resulted in a single stress in basis points for each credit rating, but did not depend on for example currency or maturity. This calibration of the absolute values was the starting point for the re-calibration when the switch to the relative approach for NDSR was made due to the changes in the discounting approach (using maturity dependent spread adjustments instead of a flat spread adjustment).

In 2022, the NDSR charge was adjusted to a relative bi-directional stress to reflect the dynamics of the term-structure approach introduced for Market Adjusted Valuation. The relative stress was calibrated to 75% based on the calibration work done earlier as described in 102. The re-calibration to a relative stress was done in a way such that the NDSR charges would be comparable to the NDSR charges of the previous absolute methodology. The decision to calibrate the relative stress to a single value was based on the granularity of available data (eg with respect to currencies, maturities and credit ratings) and accompanied with expert judgement to strike the balance between accuracy and complexity.

The relative stress approach can lead to very low stresses in low spread environments and very high stresses in high spread environment. To avoid a misestimation of the risk by a pure relative approach, a floor of 40 basis points (bps) and a cap of 150bps for spread movements for the up-shock were introduced.
Calibration was conducted based on time series (01/01/2010-30/06/2023) of spread data available for EUR, GBP and USD determining the 99.5% and the 0.5% quantiles of spread movements. To keep the balance between accuracy, the limited availability of granular spread data and complexity of the model, a single value for all maturities, currencies and credit ratings was implemented. This approach was also supported by analysis indicating that results would not substantially differ with a more granular approach for most currencies.