Modelling model risk
Paul Wilmott has words of wisdom for anybody in the financial-services industry who’s putting a model together:
At every stage of valuation and model development you must be asking questions about risk and robustness. It is dangerous to come up with some fancy model and only afterwards start asking questions about model error. Anyone who has ever calibrated a model knows that the methods used to mitigate model risk almost come as an afterthought, and are totally inconsistent with the original model. This need not be the case.
The problem is that developing a model is the sort of thing which (a) quants are trained to do, and (b) can, eventually, make money. While mitigating model risk is a very recondite field which very few people have any expertise with; what’s more, it doesn’t really make money in and of itself. Where will all the model-risk modellers come from?