Two years ago, the public discussion surrounding bank management and regulation gave the impression that all models were fighting for survival. The demand was for simple regulations that would produce comparable results at all banks. This only seemed possible with standardised approaches.

Just as it is not possible to do without models as the basis for weather forecasts, banks and supervisory authorities cannot manage without them either. On the contrary: the latest developments, such as to accounting standard IFRS 9¹, show that the use of models is increasing, even outside their original areas of application.

Rather, what is critical is that they are only used where there is a sufficient basis for forecasts, the recipient of model outputs is aware of their boundaries and the associated risks and limitations are taken into account.

Two major developments in 2016 focused on both aspects: firstly, the reform of the Basel framework on internal models entered its final stages. And secondly, 19 national competent authorities and the European Central Bank (ECB) launched the Targeted Review of Internal Models (TRIM) project under the Single Supervisory Mechanism.

The following reports on the current and intermediate results.

Reform of the Basel framework on internal models

The first approval of internal models for market risk (in 1996 under the market risk amendment to the Basel framework) and for credit and operational risk (in 2004 under the Basel II framework) led the banks to significantly improve the quality of the data used for the models and their risk management processes in order to create an appropriate foundation for the modelling and supervisory approval of risk models. Since then, many credit institutions have prepared for the review of these models in projects that often last for years. Conversely, the internal models have helped improve risk management, in particular through intense risk analysis, high-quality data, risk measurement appropriate to the risk profile and decision-oriented risk reporting. The key requirement for this was and remains the risk sensitivity of the models used.

However, the focus in recent years has not been on the appropriateness of an internal model for a specific institution, but rather comparability of model outputs across different institutions. Internal models were seen as a potential source of undesired variation in capital requirements.

Range determined by outliers

In principle, certain international and European benchmark reviews have shown that the range of risk-weighted assets determined is quite acceptable for the vast majority of banks reviewed. The range actually observed is primarily determined by a small number of outliers, and overall a large proportion of the differences are due to the different risks faced by the banks.²

The BCBS is working on stemming the remaining undesired variation and refining the regulatory framework for banks – Basel III – as appropriate. In doing so, the BCBS is balancing the three goals of "risk sensitivity, simplicity and comparability".

After publishing a progress report in November 2015 on the work since the financial crisis to reform the regulatory framework, in March 2016 the BCBS published a consultative document on reducing variation in credit risk-weighted assets by means of constraints on the use of internal model approaches. The proposals contained in the document are based on an analysis of the ability to use internal models to calculate regulatory capital using various criteria such as data availability, modelling techniques and validation.

The consultative document sets out proposals in the following areas:

• Removing or limiting the option to use the internal ratings-based (IRB) approach to calculate regulatory capital for certain portfolios
• Adopting exposure-level floors for certain model parameters for portfolios where the IRB approaches remain available
• Providing specification and guidelines for parameter estimation in IRB approaches to reduce the variability caused by different practices

The document also announces the likely introduction of an output floor for calculating regulatory capital based on the respective standardised approaches. The design and calibration of this output floor is subject to further consultation.³

Impact study

In parallel to the consultative document, the BCBS carried out a quantitative impact study to assess issues including the extent to which the proposed amendments can be reconciled with the requirement of the Group of Central Bank Governors and Heads of Supervision (GHOS) that regulatory capital not be significantly increased as a result.

The proposals set out in the consultative document were partially modified based on the additional information gained from the impact study and the discussions held in 2016. The work is already at an advanced stage but nevertheless has not yet been completed. German banks use a relatively large number of internal models compared with their international peers. As a result, supervisory authorities in Germany pay very close attention to this issue. From a qualitative standpoint, the requirements of the model experts at BaFin and the Bundesbank with respect to compliant and professional use of internal models by banks are at the upper end of the regulatory framework; however, it goes without saying that the principle of proportionality is observed.

Risk sensitivity as principle

BaFin is heavily involved in the Basel consultations on internal models and, while it believes that limiting the use of internal models to calculate regulatory capital is indeed reasonable in specific cases, it takes the view that risk sensitivity should not be abandoned as a regulatory principle. BaFin and the Bundesbank thus place great value on retaining the key achievements of internal models.

Internal models are not an end in themselves; rather, they aim to strengthen an institution's risk management. Ultimately, introducing internal models significantly increases the quality of the information available to manage a bank. Alternative, non-risk-sensitive approaches can lead to inappropriate incentives and thus entail the risk of mismanagement.

Consequently, risk sensitivity and the requirements for the risk management processes should not be sacrificed for standardisation and simplicity. The framework should be as simple as possible but no simpler, since only then can it match the complexity of products and risk profiles.

SSM project TRIM

The overarching objective of TRIM is to rebuild trust in the use of models to calculate risk and capital requirements, which has collapsed since the financial crisis. Likewise, the aim when comparing the model banks is for the same risks to lead to the same capital requirements and for the supervision of these models within the SSM to be harmonised and improved.

Priority areas

The Basel framework and its implementation through the Capital Requirements Regulation are already extremely comprehensive. The EBA has added numerous explanations and interpretations to the legislation. Despite this, member states interpret the framework differently based on their national idiosyncrasies and have developed different supervisory practices. Likewise, the banks are facing challenges due to the complex regulations, and ask for guidance. Thus, a key harmonisation task within the SSM is to identify these differences, understand the specific underlying causes at national level and – where possible – find standardised approaches.

Accordingly, the TRIM project does not address the broad range of issues surrounding model reviews, but focuses on targeting selected topics. Hence its name: "Targeted Review of Internal Models". A project of this scale on the juncture between supervisory framework and supervisory practice must be closely interlinked with the development of regulatory reforms, ongoing supervision of models and regular model reviews.

Project progress

In 2016, various working groups identified the topic areas to be addressed in TRIM and formulated the respective "supervisory expectations" of banks as well as guidelines and instruments for reviewers. These requirements fit within the prescribed regulatory framework.

The "supervisory expectations" form the basis for the TRIM project and the dialogue with the banks. During the project they will be augmented with the experience gained from supervisory consultations and reviews, and towards the end of the project, following a consultation, they will be adopted by the Supervisory Board of the SSM, on which BaFin is also represented.

The dialogue with the banks began in 2016 with topic-related surveys and supervisory consultations, and will be continued on the specific topics in 2017/18 with the conclusion of the supervisory consultations and model reviews.

Conclusion and outlook

The changes to the Basel framework currently pending a decision will not result in the feared eradication of internal models, but rather are expected to allow internal models in those areas where there is sufficient data and information enabling a forecast and review.

One fundamental concern of the Basel framework was to allow internal models as a risk-sensitive approach to calculating capital, which – precisely because of its risk sensitivity – could also be used for risk management. This concern would still be taken into account, although the output floor referred to above could results in constraints on the benefits of using internal models to calculate capital requirements.

The TRIM project makes a key contribution to achieving uniform interpretation and implementation of these rules. TRIM imposes particularly high demands for the regular and comprehensive validation of models and for suitability tests by staff at the banks who are sufficient in number, qualified and capable of seeing the task through.

In addition, TRIM is a major step towards integrated and harmonised supervision of internal models within the SSM, in which the national supervisory authorities can leverage their proximity to the institutions and their many years of experience with the idiosyncrasies of their banking systems.

The developments presented here will place increased requirements on the use of internal models, which will further strengthen the justified trust in the models. The increased requirements must not remain limited to internal models used in calculating capital requirements, but rather must apply in equal measure to all other models used by a bank.

Footnotes: