Reconciling Tail Risks and Non Tail Risks ...Or Not.

Traditional measures of risk, most obviously sensitivities like the greeks and percentile measures like VaR, capture only selected dimensions of risk. As Nassim Taleb has memorably pointed out, they fail to capture the extreme losses (tail risks) that could derail the organization. Ironically protecting against these major losses is the whole point of risk capital. For example, both credit (Internal Ratings Based approaches) and market risk (Internal Models Approaches) and operational risk (Advanced Measurement Approaches) use the VaR concept as the definition of minimum regulatory capital. Even worse than this, is the fact that careless use of such VaR measures actually makes tail risks more likely, as traders go long exposures in the tail that have little effect on VaR, but potentially subject the institution to devastating downside. Tail risk measures such as conditional VaR and expected shortfall, do help in this regard, but the truth is, no single measure ever completely captures a complete distribution. So the only solution here is more management understanding of what distributions are about – and this means a more sophisticated management audience for risk results.