Marko Milojević, Ivica Terzić
2014 CBU International Conference Proceedings  
The need for understanding financial risk management and unique models for measuring risk in transitional capital markets increasingly gains in importance and becomes a very current issue. This article studies predictive ability of various classes of Value-at-Risk (VaR) models focusing on Serbian equity market in both stressed and normal market conditions. The five VaR models adopted in our evaluation procedure include: historical simulation with rolling window of 500 days, Risk Metrics,
more » ... isk Metrics, exponentially-weighted moving average (EWMA) with optimized decay factor, VaR based on models from GARCH family under three distributional assumptions (normal, generalized error, and Student-t), and Filtered historical simulation. In order to verify the forecasting performance of different VaR models, we employ a backtesting procedure, which consists of statistical tests. The results indicate that VaR based on conditional volatility models with asymmetric distribution of innovations behave reasonably well in both tranquil and crisis period. Standard VaR models developed for liquid and efficient markets seriously underestimate risk forecast in Serbian equity market under all circumstances.
doi:10.12955/cbup.v2.455 fatcat:oywd7ygorjdcrawn6q6t7c6gai