Risk Model Validation

Book Preface

The last decade of the 20th Century and the first decade of the 21st Century had
seen the use of quantitative risk models (QRMs) become a cornerstone of financial regulation. Financial institutions now have to determine capital buffers based on increasingly complex modelling techniques. The computation and allocation of capital buffers to business lines now share more features with the solution of multidimensional optimisation problems than with classical business model analysis.

As a consequence, the Turner review (FSA 2009) speculated that many in top management face difficulties in assessing and exercising judgement over the risks being taken by their institutions. According to the review, these difficulties contributed to one of the worst crises that our financial system has ever experienced.

Of course, the Turner review also mentions many other factors that may have played a prominent role, such as the heightened complexity of the structured credit market, the fact that financial institutions took on too much leverage and the role of the rating agencies. However, this crisis has definitely emphasised the need for rigorous and critical analysis of the use and misuse of risk models and the results they can produce.

Although the existing literature on risk model validation presents many contributions from specialists into the use of risk modelling, ranging from qualitative aspects to sophisticated statistical validation techniques, their highly specialised nature means they are not hugely accessible to managers without a training in quantitative finance. This report, based on the authors’ practical experiences of establishing models for both market risk and credit risk in the banking industry, will focus on providing a holistic perspective aimed at the ‘informed layperson’. It will attempt to answer these key questions:

• How can we establish a practical framework for thinking about risk?
• Were there risk models before the Basel framework?
• What are the common features of today’s QRMs?
• How can risk models fail?
• What are the limits to risk modelling?
• What are the challenges when implementing risk models in software?