Tesfalidet Asfaha, Anthony F. Desmond, Getu Hailu, Radhey Singh
This paper develops a skewness and leptokurtic modified VaR model with a mixture weight parameter that blends the Cornish-Fisher and EWMA methods. We estimate and evaluate five existing parametric VaR specifications using weekly returns for Canadian feedlot cattle feeding margin data and Maple Leaf Foods stock return data. The estimation of VaR based on EWMA method yields the most satisfactory results particularly for returns with positive skewness or leptokurtic tails. Meanwhile, the VaR forecasts obtained using the Cornish-Fisher method provides a relatively better tracking of the observed returns compared to the other methods, and therefore, has lower forecast error. Our proposed model allows users to determine the value of VaR based on their own risk preferences.
Publication:Journal of Statistical and Econometric Methods
Date: January, 2014