Ranking of Parametric Value at Risk Models with Consideration of Trader Position (Application of Asymmetric Distribution Functions in GARCH Models)

Abstract

In this paper, we estimate the value at risk of Tehran stock exchange (TSE) index by using GARCH family models in short and long trading positions. Because of asymmetric behavior of returns for long and short positions in TSE, for enhanced accuracy of model, we apply asymmetric normal and t-student distribution functions. By developing Sener et. al (2012) measurement for considering trading positions in performance assessment of parametric models, we show that EGARCH and GJRGARCH models with asymmetric normal and t-student distribution functions are more accurate than other models. Also complementary forecast ability test explain that, with a benchmark model such as GJRGARCH, other models do not have equal mean error, so the asymmetric distribution functions in EGARCH and GJRGARCH models improve their ranks.

Keywords

• value at risk
• trading position
• parametric models
• performance ranking of models