CMStatistics 2016: Start Registration
View Submission - CFE
A1545
Title: Portfolio optimization based on GARCH-EVT-copula CVaR and mean-variance models Authors:  Andreas Stephan - Linnaeus University (Sweden) [presenting]
Maziar Sahamkhadam - Jonkoping University (Sweden)
Ralf Ostermark - Abo Akademi (Finland)
Abstract: Regarding GARCH-EVT-Copula models, most studies have been performed on Student T and Gaussian copulas, with a few considering Clayton copula. By adding Frank and Gumbel, we utilize GARCH-EVT-Copula models for optimizing and comparing Min-CVaR and Mean-Variance portfolios. In order to perform the portfolio optimization, ten stock markets are considered including S\& P 500, FTSE 100, DAX 30, EURO STOXX 50, MSCI World, CAC 40, OMXC 20, OMXH, OMXS 30 and TOPIX. The sample period starts at August 26, 1996 which includes 5218 total observations. By dividing the sample into three periods and applying rolling portfolio, different rebalancing strategies are considered in order to investigate the effects of Financial Crisis. To estimate the optimal weights, Linear and Quadratic Programing are implemented in Min-CVaR and Mean-Variance portfolios, respectively. The results show that Frank copula leads to the lowest CVaR. Moreover, before the crisis, Student T and Gaussian copulas have the highest Sharpe Ratios with quarterly and monthly rebalancing strategies. During the crisis, Clayton copula results to a higher Sharpe Ratio. Interestingly, after crisis, Student T and Gaussian copulas outperform other models with daily and semi-annually rebalancing. Finally, in most cases, Mean-Variance portfolio shows better results than Min-CVaR portfolio.