CheckRisk works with organisations to review their in house risk and financial models and then optimises them against their benchmarks. Effective model validation is about finding the best model for your objectives, without over-fitting it, and reducing confidence in its validity.
The model valuation process involves taking a structural model, for example, then producing a desired forecast. This “forecast” is then measured for its accuracy both in and out of sample. CheckRisk can then make the process dynamic and allow re-calibration and ongoing testing, as though it were ‘live’. Varying the sensitivity of parameters allows us to get a sense of the strengths and weaknesses of a given approach, ultimately providing us with enough evidence to make the most appropriate validation or recommendation for your needs.
The above example, is a simple but good illustration of how traditional value at risk (VaR) models underestimate risk at the time you need them to be most correct. VaR models are commonly used by the financial industry, central banks and other investors and often lead investors to make wrong decisions.
The chart shows how the volatility of FTSE 250 returns exceeds the VaR levels (with re-scaled volatility) of the model at various confidence intervals (95% yellow, and 99% red). Some breaches may be acceptable, but persistent breaches in a given time period mean that your model is not working as it should. Ideally, we should aim for the risk to be a slave to the model, not vice-versa. In this way, the returns will take care of themselves.
CheckRisk can explore alternative methods on behalf of the client to find the best solution. The model above, is from the filtered historical simulation family, and as you can see, is a better fit. By gradually adding more sensitivity and calibrating parameters, we can ensure we tailor make a model that will suit your specific portfolio. To maintain an optimal performance, the model validation process needs to be ongoing and part of the Investment Risk Governance process of an organisation.