Empirical Analysis of Value at Risk Models: SIAS Equity Portfolio Risk Model Selection and Formulation

This paper is to determine an appropriate Value-at-Risk model that can improve the overall management of the SIAS fund, particular for the two equity portfolios. We consider the four candidate models: Historical Simulation, Dynamic Conditional Correlation Generalized AutoRegressive Conditional Heteroskedastic , Filtered Historical Simulation, and Hybrid Approach. Using historical information from 2003, all the models are implemented, and their specifications and performances are discussed in detail and examined with four backtesting procedures, including Unconditional Coverage, Independence, Conditional Coverage, and Quantile Regression tests. Our findings confirm that the Historical Simulation model performs poorly in capturing the volatility dynamics, and we also have a comprehensive discussion about the factors that are used in the Hybrid Approach model. Those two are highly rejected from all the test procedures. On the other hand, Filtered Historical Simulation is the only model that passes the likelihood ratio tests. However, the likelihood ratio test may be flawed and biased; therefore, we employ Quantile Regression test that is believed to be a more powerful backtesting procedure. The results turn out that DCC GACRH is the best model among others. In addition, its other properties allow the risk management process to be more in depth. Therefore, DCC GACRH is strongly recommended.