Custom Search
|
|
A New Measure Of Risk - Hybrid Var
With each financial failure, the credit crisis generates a new bout of criticism of Value-at-Risk (VaR) and practice of risk management in general. This article does not seek to discuss how most of the criticism is often ill-placed, that VaR is not the only risk measure available in risk management, that VaR can be produced with models other than Gaussian Variance/Covariance, and so forth. We want instead to focus our attention on two serious failures of VaR (and similar measures, including Expected Shortfall, or Conditional VaR) during this ongoing crisis: • The significant failures in backtesting • VaR pro-cyclicality [1] These two criticisms are certainly valid and while, as risk managers, we can find justifications for the first (weather forecast tends to be less reliable when dealing with tornados), the second criticism is more serious and requires greater attention. Exponentially Weighted (EW) risk models that give greater weight to more recent history are very popular and widely used, but they are also the most pro-cyclical of the models. In this article we will demonstrate that their pro-cyclicality is not balanced by better performance in backtesting terms, and will present a new measure of risk “Hybrid VaR”. The first part of the article looks at the methodology for computing Hybrid VaR, and the second part will demonstrate that it performs better in backtesting than pure historical VaR or EW VaR (both Gaussian and Historical-Adjusted). In addition, Hybrid VaR is anti-cyclical in boom periods, discouraging risk-taking in good times, and much more stable than other measures. This significantly mitigates the problem of pro-cyclicality while also providing a measure that is more reliable than other risk measures. [1] i.e. increases in risk measures during periods of market stress, increasing the likelihood that risk limits are violated, causing further liquidation of positions in a falling market. This pro-cyclicality thereby amplifies negative market movements Article Directory: http://www.articledashboard.com The full 18-page whitepaper is available to download for free. www.statpro.com/eventsandeducation/whitepaperdetails.php?id=6&pid=sr Dario Cintioli is Head of Risk at StatPro Group |
|