The Comparison of Volatility Forecasting Models in VaR Calculations and Backtesting according to Basel II: An Application on ISE 100 Index

For determining the Value-at-Risk number with statistical models volatility must be the primary calculation. There are different volatility estimation methods on VaR calculation. The traditional volatility estimation methods are inadequate for modeling “stylized facts” which are often observed on the financial price series. In this study, different volatility models are introduced and the differences are illustrated among each other. In the empirical application 14.5 years of daily closing values of ISE 100 Index are being used for estimating the different volatility models. Estimated volatility numbers are being used for calculating the VaR numbers and the results are tested by backtesting method based on Basel II. Among all calculations “Rolling window” method is used for updating parameters daily; specifically to determine the success of modeling special characteristics of financial price series and four different time periods are being used. According to the findings obtained, volatility clustering on financial price series, changing variance, leverage effect, peakedness is preferable to be modeled by advanced models such as EWMA and GARCH.