Developing the Methods of the Performance Evaluation of VaR Models Using Fuzzy TOPSIS and Copula-ARIMA-GARCH Model

This study used the Copula-ARIMA-GARCH approach to calculate the VAR of a portfolio of four investment companies and daily data (from March/April 2009 to February/March 2017) to calculate risk more accurately and develop methods of the performance evaluation of value at risk (VaR ) models by a combination of a copula function and ARIMA/GARCH models. In this study, a novel method was proposed to evaluate the performance of VaR models using the fuzzy multicriteria decision-making models. For this purpose, ranks attained by the VaR estimation models were employed considering the unconditional coverage test procedure, Dowd’s loss function procedure, and the rank given the prediction accuracy. The results indicated that risk-taking and risk-indifferent investors assume that the VaR calculated from the Copula-ARIMA-GARCH model is the most accurate model, while risk-averse investors take the generalized extreme value (GEV ) model as the most accurate model considering the high importance of the loss function.