Comparing traditional VaR with copula based VaR in extreme market volatility : Empirical Evidence from Finland between 2004 and 2009

Abstract

This thesis examines the effectiveness of copula-based Value at Risk (VaR) compared to traditional VaR methods under extreme market volatility within the Finnish financial markets from 2004 to 2009. Traditional VaR has been widely used in risk management due to its simplicity; however, it often fails to capture the nonlinear dependencies, tail risks and stylized facts in financial time series. This inadequacy is critical during periods of significant market stress, such as the 2007-2009 financial crisis.

To address these limitations, this study employs copula-based VaR, which integrates copulas to model dependency structures from the marginal distributions, potentially offering enhanced accuracy in risk assessment during volatile periods. Utilizing historical data from Finnish stock and bond markets, the research applies ARMA-GARCH models to estimate the marginal distributions and Monte Carlo simulations to compute VaR. The VaR forecasts are done in rolling fashion manner, imitating a more practical approach. The performance of both copula-based and traditional VaR models is tested with two backtesting methods.

The results indicate that copula-based VaR, particularly the Student’s t copula, provides a more accurate and reliable measure of risk under extreme market conditions compared to traditional VaR. These results not only underscore the potential limitations of traditional VaR in extreme market conditions but also highlight the advantages of copula-based approaches in enhancing risk assessment frameworks.