The Efficiency of GARCH Models in Realizing Value at Risk Estimates

Abstract

Market risk is an important type of financial risk that is usually caused by price fluctuations in financial markets. One determinant of market risk comprises Value at Risk (VaR), which is defined as the maximum loss that can be achieved within a certain time horizon and at a given reliability level. The aim of the article is to determine the importance of selecting conditional volatility model within the parametric and semi-parametric approach for VaR estimation. The results ascertained show that the application of these models tends to provide more accurate predictions of actual losses as compared to traditional approaches to VaR estimates. Overall, the application of conditional volatility models ensures that VaR estimates are more flexible to adapt to changing market conditions – especially in the periods associated with higher return volatility. Furthermore, the results show that the differences between individual models of contingent volatility are primarily determined by selecting the specific distribution of the standardized residue series.