FRANKFURT SCHOOL

BLOG

Validation of the entire bank: new MaRisk module for the application of models
Executive Education / 23 January 2024
  • Share

  • 2359

  • 0

  • Print
Henning Heuter is a managing partner at a specialised consultancy firm. In his work, which includes risk management issues and their regulatory treatment, he advises banks from all groups of institutions in Germany and abroad. His activities also include the design and further development of the limit system for the entire bank, the credit risk strategy and counterparty risk management at the overall bank level. He has been a highly regarded lecturer in risk management at Frankfurt School for many years.

To Author's Page

More Blog Posts
Erforschung der Auswirkungen von KI auf die Zukunft der Arbeit
Key strategies for business development in the dynamic world of business
Master the integration process and achieve your merger goals!

With the publication of the 7th MaRisk Amendment, Module AT 4.3.5 sets out new requirements for the use of models that institutions must comply with from 1 January 2024. The new module is the result of the implementation of the EBA Guidelines on loan origination and monitoring (EBA/GL/2020/06). However, the requirements of MaRisk are broader, as they do not only apply to models in credit processes. The new module contains a definition of the term “model” and extends the scope to all decision-relevant models in banking operations that are covered by the MaRisk processes. The requirements do not refer to specific technologies and include automated models, technology-enabled innovation and artificial intelligence (AI). The requirements also focus on model governance, with particular emphasis on model validation.

Regulatory background

MaRisk AT 4.3.5 point 5 requires regular validation of the models used in the institution and a critical review of their limits, restrictions, assumptions and input data. During validation, the appropriate handling of model results and the accuracy of the model must be checked. It is also necessary to analyse the model results on a regular basis. This includes their accuracy as well as the stability and consistency of the process.

The notes also point out that recalibrations can play an important role in some models. These can strongly influence assumptions and weightings. It is therefore important to consider the impact of recalibrations when analysing model results.

Point 6 requires that models are also sufficiently explainable. This refers to the traceability of causal relationships between input and output variables and is particularly relevant for technology-based innovation and artificial intelligence. We interpret point 6 to mean that this should be done in the course of validation, as explainability must be checked “in addition to the desired accuracy”, which is mentioned in point 5 as part of the requirements for validation. In particular, the term “model” should be expanded to include a wide range of other model categories, such as:

  • Models used in the ESG environment and in the credit process (including AI models)
  • Models to estimate the input parameters for risk models or to estimate cash flow structures
  • Pricing models and models to estimate value adjustments

Practical implications for the validation

Previously, the minimum regulatory requirement was that the risk models used to calculate risk-bearing capacity had to be validated (MaRisk, AT 4.1 (in particular points 8, 9 and 10)). It was not uncommon for such models to be validated in order to determine liquidity capacity. However, further validation of other model categories was the exception rather than the rule.

The RBC validation focuses on overarching issues, as the risk models for the main risk types are checked in advance. These include, for example, data quality and procedural issues, the completeness of input variables and the scope of defined scenarios to analyse the economic and normative perspective.

The expansion of the scope of models to be validated by AT 4.3.5 means that these and other types of models (see examples above) are now also the focus of validation. Ultimately, it is expected that the existing and ongoing requirements for risk models under MaRisk, AT 4.1, will have to be applied to the entire modelling context.

The requirements for method verification and validation continue to grow. The risk model validations established in recent years need to be adapted to changing conditions, while the validation of the entire bank often requires further development in order to integrate overarching requirements and, in particular, the normative perspective.

It is essential for professionals and executive managers in banks to acquire specialist knowledge in order to develop their risk management skills. Advanced Training, especially in risk management, knowledge transfer from experienced lecturers and an established network allow you to acquire practical know-how and exchange expertise.

 

Co-authors

 

Dr. Walter Gruber
Lecturer at Frankfurt School

 

 

Dr. Walter Gruber is a managing partner of 1 PLUS i GmbH. He has authored numerous publications, particularly in the areas of banking supervision, market and credit risk modelling and derivative financial products. He is a lecturer in risk management at Frankfurt School.

 

Dr. Christian Stepanek
Partner at 1 PLUS i GmbH

 

 

Dr. Christian Stepanek is a partner at 1 PLUS i GmbH. His consulting activities focus on rating procedures and the validation of risk measurement procedures (credit and market price risk or macroeconomic stress tests). He gained extensive knowledge of the requirements for IRB procedures by supporting the ECB TRIM exercise in 2016 and the TRIMI audit in 2017. He has extensive experience in supervisory data queries such as the EBA stress test (e.g. as a project manager and in the NII cluster). He was in charge of the technical implementation and group coordination of the SRB data queries for MREL at Landesbank. In addition to his consultancy work, he is also a speaker and author on the topics mentioned above. Dr. Stepanek holds a degree in physics and a doctorate in empirical capital market research and applied econometrics.

0 COMMENTS

Send