Portfolio Risk Measurement with Asymmetric Tail Dependence in Tehran Stock Exchange

Portfolio risk measurement has always been one crucial aspect of finance. Several approaches have been modeled through time and some traditional approaches are criticized by researchers. Traditional methods use the Gaussian approach that it been used normal distribution for assets return. Empirical studies indicate that the return of assets has a stylized fact. The stylized facts include the fat tail, negative skewness and volatility clustering. Furthermore, some studies show that there is asymmetric tail dependence. This means in market downsides, dependence is more than upside market. So this concept should be considered in portfolio risk measurement, otherwise, the estimates will end up having a bias.

To consider the stylized fact, different methods had been proposed. For modeling fat tail and negative skewness, this article uses extreme value theory and to consider volatility clustering, GARCH, EGARCH and GJR are applied. For modeling nonlinear and asymmetric dependence different methods are proposed. In the present article T-skewed copula is used. Mean-variance model is used to compute the weight of the portfolio. To analyze the result, backtesting is examined and for comparing methods, QPS function is applied.

In this article TSE data have been used. At first, stylized facts have been checked. For testing fat tail and negative skewness, JB test is used. to analyze the volatility clustering, Liung Box test is applied. Also to check asymmetric tail dependence, exceedance correlation and paired T-test are used. Findings indicate the existence of the stylized facts in the 30 stock return of TSE. After running the models, the only approach based GJR and EGARCH is accepted in backtesting. Also based on QPS function, findings represent the strength of the applied approach to common ones.

Results show that the use of the GJR model is better than other models for Portfolio risk measurementthat it shows the importance of leverage effect in TSE for stock return modeling.