Acta Univ. Agric. Silvic. Mendelianae Brun. 2015, 63, 1287-1295
Published online 2015-09-02

Assessing Efficiency of D-Vine Copula ARMA-GARCH Method in Value at Risk Forecasting: Evidence from PSE Listed Companies

Václav Klepáč, David Hampel

Department of Statistics and Operation Analysis, Mendel University in Brno, Zemědělská 1, 613 00 Brno, Czech republic

The article points out the possibilities of using static D-Vine copula ARMA-GARCH model for estimation of 1 day ahead market Value at Risk. For the illustration we use data of the four companies listed on Prague Stock Exchange in range from 2010 to 2014. Vine copula approach allows us to construct high-dimensional copula from both elliptical and Archimedean bivariate copulas, i.e. multivariate probability distribution, created from process innovations. Due to a deeper shortage of existing domestic results or comparison studies with advanced volatility governed VaR forecasts we backtested D-Vine copula ARMA-GARCH model against the VaR rolling out of sample forecast from October 2012 to April 2014 of chosen benchmark models, e.g. multivariate VAR-GO-GARCH, VAR-DCC-GARCH and univariate ARMA-GARCH type models. Common backtesting via Kupiec and Christoffersen procedures offer generalization that technological superiority of model supports accuracy only in case of an univariate modeling – working with non-basic GARCH models and innovations with leptokurtic distributions. Multivariate VAR governed type models and static Copula Vines performed in stated backtesting comparison worse than selected univariate ARMA-GARCH, i.e. it have overestimated the level of actual market risk, probably due to hardly tractable time-varying dependence structure.


The paper was supported by the Internal Grant Agency FBE MENDELU under project: Vývoj stochastických metod pro predikci korporátních bankrotů (No. 36/2014).


24 live references