The volatility of financial markets has been the object of numerous developments and applications over the past two decades, both theoretically and empirically. Portfolio managers, option traders and market makers all are interested in the possibility of forecasting volatility, with a reasonable level of accuracy. That is so important, in order to obtain either higher profits or less risky positions. Objective: we have compared different GARCH models with both Gaussian and fat-tailed conditional distribution for residuals in terms of their ability to describe and forecast volatility. Results: The best model based on MSE criteria is GARCH with normal distribution, second model is GARCH with t distribution and third model is EGARCH (1,1) with normal distribution. Conclusion: Results indicate that leverage effect exists in asymmetric models with normal distribution, but this effect does not exist in asymmetric models with t-student and GED distributions.
