Remember me

Register  |   Lost password?


 
Attached please find copy of some new research I have been working on that may be of interest.
In it I show that the returns for long only (ETF's) are non-normal 15 - 20% of the time and for Hedge Funds 30 - 40% of the time. I test the  'normality' of returns via a panel of all of the major normality tests including ( Anderson Darling, Shapiro-Wilk, Cramer von Mises, Kolmogorov-Smirnov (Lilliefors), Jarque Bera etc ). 
I show how the Cornish Fisher modification to the normal Value at Risk ( VaR ) methodology - which is the current state of the art in widespread use in hedge fund land - gives misleading and potentially inaccurate results up to 50% of the time and should generally not be used at high confidence levels ( above 95% ) or without testing whether it is appropriate to use. This problem does not only occur for funds with high levels of excess skewness or kurtosis as most people believe. I show this and also provide a simple test.
Replacing the normal and modified distributional assumptions with a proper distribution fitting methodology where multiple distributions are tested to the data allows for better fitting up to 80% of the time. This leads naturally to a non-linear form of correlation via a bi-variate copula approach. Whilst these are obviously sensitive to change and the influence of individual outliers they provide significant additional information on which to base the selection of fund A over fund B etc. It is my strong belief that funds that start to go bad exhibit increased tail behaviour well before the event ( this was certainly the case for Amaranth ) and these methods can be used to monitor for this. 
One way this can be used is to identify different low correlation pairs. The Best Fit correlation methodology generates different pairs from the classical Pearson (linear) correlation methodology about 34% of the time. If one adds an additional return threshold constraint then this drops to around 10% but is still useful in selecting different potential pairs for diversification and hedging purposes. The main value add here is that you are not using the same hedge pairs as everyone else. Let me know if you have any feedback.
Kind regards,