Fat Tails and Extreme Events [22]

Go to: Summary | Previous | Next   
Bullet points include: Assume Cornish-Fisher (CF4) is a good estimator of ‘1 in 200’ risk level. Risk then proportional to std . (1 + c . kurt) where c = 0.39 at ‘1 in 200’ level. Principal Components Analysis (PCA) focuses just on standard deviation, Independent Components Analysis (ICA) just on kurtosis, blend on both. Analysis based on monthly MSCI World sector relative returns Jun 1996 to Feb 2009. Sizes of ‘1 in 200’ events potentially underestimated several-fold by PCA, if portfolio built on the basis of ‘meaning’ (e.g. if actively managed)

Contents | Prev | Next | Library

Desktop view | Switch to Mobile