Skip to Content

More than a floor, a basis: Basel IV, output floors and the discriminatory power of risk models

Capgemini
July 24, 2020

However, was the quality of the previous models really insufficient? What are the new challenges that face financial institutions, and how will they deal with them?

A regulation full of nooks and crannies

The supervisory authorities see the completed Basel III framework (also known as Basel IV) as a rectification for the numerous perceived shortcomings of banks’ internal risk models, an increase in the validity of the parameters, an improvement on the transparency and comparability between financial institutions, and the assurance of a certain level of regulatory capital.

From 2023 onwards the minimum capital requirements for credit risk, as derived from the internal ratings-based approach (IRBA), will be at least 50% of the minimum capital calculated according to the standardized approach for credit risk (CRSA). This minimum threshold is a so-called “output floor”, by which all calculated regulatory capital must be at this level or higher. Following the initial roll-out, this lower output floor limit is raised by 5 percentage points per year until 2028, when it will reach 72.5% of the regulatory capital calculated according to the CRSA.

A deeper view on the extent to which the minimum capital requirements calculated by the IRBA differ from the output floors can be read in the article Basel IV: How close are IRBA banks to the output floor?

Additionally, the completed Basel III reform explicitly requires the appropriate separation power of risk models. This means that the challenges cannot be met by recalibrating current models alone.

The possibilities for adjustment are manifold

We consider three possible scenarios of how financial institutions can react to this new regulation:

  1. Adjustment of the RWA calculated according to IRBA to the output floor
  2. Recalibration of the IRBA models so that the calculated RWA do not fall below the output floor
  3. A hybrid mixture of IRBA recalibration and CRSA use

Let’s look at the individual scenarios in detail.

  1. An RWA adjustment to the output floors could be the fastest, although not necessarily the most cost-effective option. It does not involve the abandonment of qualified resources or the recalibration of the models. For this reason, the discriminatory power of the models would not be affected by the new regulation. However, institutions that would choose this option would implicitly admit that their models are not designed to calculate the minimum capital requirements correctly, but rather to minimize them. Such an admission could lead to considerable reputational damage.
  2. The recalibration of your own models to meet the required output floors has the advantage of leveraging existing resources and expertise while ensuring regulatory compliance. Two problem areas arise in this scenario. Firstly, this exercise is a time-consuming undertaking that ties up many resources. Secondly, recalibrating the model weightings when adapting to an exogenous result inevitably affects the predictive power of the model – especially if this is done independently from other recalibrations. Thus, the regulator’s requirement for better discriminatory power is not met.
  3. The introduction or maintenance of the IRBA represents a significant investment on the part of the institutions that is justified by the benefits of improved risk management. However, a financial institution may conclude that the costs of the IRBA are no longer justified and it can abolish the IRBA and revert to using the CRSA. This could reduce the flexibility of decisions and risk-bearing capacity, as well as damaging their image (even leading to a rating downgrade). Embedded in the new regulation is also the possibility of applying the IRBA only to selected asset classes and not to the entire portfolio – partial IRBA use. This allows non-IRBA institutions to consider introducing the IRBA and IRBA institutions may be able to apply the CRSA to selected portfolios. The decision to switch must be justified in advance by a comprehensive cost-benefit analysis. Please read our article Basel IV: Is it still worthwhile using IRBA?

Scenarios 2 and 3 require the bank to take a close look at the models used, to investigate recalibration or implementation options, and to explore possibilities to adapt the models in order to meet the exogenously specified minimum capital requirements.

More than brains: artificial intelligence

Against this background, we foresee an increased need for credit risk modelling and management. To meet this need, we offer AI-based partner solutions to automate model development processes while meeting regulatory requirements such as output floors and discriminatory aspects. All relevant macroeconomic, socioeconomic and company-specific influences including their early warning are considered when modeling risk parameters.

Co-authored by Rita Motzigkeit