A Comparison of Conditional and Unconditional VaR Models
Krzysztof Echaust , Małgorzata Just
Abstract: This paper presents the empirical research on comparison of two different approaches for Value at Risk (VaR) measurement. The research objective is to compare the accuracy of out-of-sample VaR forecasts between conditional and unconditional models. We examine four unconditional models: Gaussian, alpha-stable, Normal Inverse Gaussian (NIG) and Generalized Pareto (GP) distributions and four conditional models: Generalized Autoregressive Conditional Heteroskedasticity (GARCH) model with Gaussian and Student’s t innovations, Exponentially Weighted Moving Average (EWMA) and conditional Extreme Value Theory (GARCH-EVT) approach. Calculations are performed on the basis of 5 world indices, 4 exchange rates and 4 commodity futures and the results are presented for left and right distribution tails. Backtesting methods indicate the GARCH-EVT as the model that outperforms all others.
|Publication size in sheets||0.55|
|Book||Jedlička Pavel, Firlej Krzysztof, Marešová Petra, Soukal Ivan (eds.): Hradec Economic Days : Double-blind Peer Reviewed Proceedings of the International Scientific Conference Hradec Economic Days 2020, Hradec Economic Days, vol. 10, 2020, University of Hradec Králové, ISBN 9788074357763, 906 p.|
|Keywords in Polish||VaR; rozkład stabilny; NIG; GPD; EVT; GARCH; EWMA; GARCH-EVT|
|Keywords in English||VaR; stable distribution; NIG; GPD; EVT; GARCH; EWMA; GARCH-EVT|
|Score||= 5.0, 15-05-2020, ChapterFromConference|
* presented citation count is obtained through Internet information analysis and it is close to the number calculated by the Publish or Perish system.