How Can a Risk Model Fail?

Christian Meyer and Peter Quell

In the first two chapters, we have presented the basics and typical set-up of QRMs. It has been shown that there are numerous ways in which a QRM can fail, and this will now be examined in more detail by proceeding systematically through the key phases of a model’s life, which will be expanded upon as the chapter progresses.

• In the beginning there is model design. Quantitative risk modelling means making simplifying assumptions about the real world, including human behaviour, and not being aware that the resulting limitations might be an issue.

• Implementation then represents the QRM as a piece of software on a computer system. There might be inappropriate project management or plain errors in computer code (bugs), but also subtle numerical and statistical issues.

• Data (eg, portfolio data, market data) then connect the implementation of the model to observations from reality. Data might be missing or incomplete, inaccurate, outdated, temporarily unavailable, asynchronous, misinterpreted or subject to complex (statistical) modelling.

• Processes constitute sequences of actions that describe how to proceed from the observation of data to the use of model results