Model validation: pain or gain?
These days you no longer have to explain to people that mathematical models are used in the financial world. They are more than often pointed to when people try to find what started the crisis! As an econometrist and consultant on risk management, I find it rather difficult to agree with that. But I cannot deny either that misuse of the models has played a role in the financial crisis.
Now that the dust whipped up by the crisis has settled to some extent, we can take stock. Risk management is high on everyone’s agenda, new and amended rules have been introduced for banks and insurance companies, and people are much more critical of what risk management has to tell us. In my view this more critical stance may well be the best improvement. The entire organization has to know about risk management, has to be able to trust it and understand it; the entire organization has to recognize its importance and act accordingly.
To this end it is essential that the models used by risk management are fit for purpose. To assess this, the regulators have developed the concept of ‘model validation’. At regular intervals all risk management models now have to be validated in terms of their data, assumptions, techniques, etc. This goes beyond checking whether the Hull-White scenarios are correctly generated.
Model validation assesses whether the models are deployed in the right way. It looks at the processes and evaluates the governance. This means that, in addition to the audit function, another control function has been added to the organization which takes a critical view of the models and how they are used. Banks are already familiar with this, but it’s something new for insurance companies. It will attract even more interest in future because of the planned introduction of Solvency II.
As risk managers, do we really want to find fault with everything? To do so would be hugely time-consuming – in interviews, explanations and discussions, additional documents and compliance with more or less relevant recommendations. But if you want to abide by the rules, you have to bite the bullet, I suppose.
The prospect of model validation is not all that dramatic anyway in my view. I think that it will bring a huge improvement and development in the discipline. After all, model validation is about guaranteeing a consistent, uniform and correct use of models. It must offer the assurance that the models are appropriate for the business and its objectives.
Model validation is thus far more than ticking the boxes to confirm that the regulations are being followed. Model validation constitutes an additional layer for risk management. A layer that ensures risk is used in the right way. I predict that this will only broaden the base for riskoriented thinking within the organization. It will help to ensure that risk is treated appropriately within the financial organization. Eventually, the ‘risk’ aspect will influence everyone’s thinking and actions.