In Defence of VaR

Risk measurement and Backtesting in times of Crisis

Since the Financial Crisis, Value at Risk (VaR) has been on the receiving end of a lot of criticisms. It has been blamed for, amongst other things, not measuring risk accurately, allowing banks to get away with holding insufficient capital, creating over reliance on a single model, and creating pro-cyclical positive feedback in financial markets leading to ‘VaR shocks’.

In this article we do not seek to defend VaR against all these claims. But we do seek to defend the idea of modelling risk, and of attempting to model risk accurately. We point out that no single model can meet mutually contradictory criteria, and we demonstrate that certain approaches to VaR modelling would at least provide accurate (as measured by Backtesting) near-term risk estimates. We note that new Market Risk regulation, the Fundamental Review of the Trading Book (FRTB) could provide an opportunity for banks and regulators to treat capital VaR and risk management VaR sufficiently differently as to end up with measures that work for both rather than for neither.

We show that an Exponentially Weighted Moving Average Vol Scaled VaR would have performed far better than 'vanilla' VaR both in the financial crisis and in the recent COVID-19 related market turbulence.

Did you find this useful?