The study looked at how stock returns behave in the Macedonian Stock Exchange and how accurately different models can predict volatility. They found that the stock returns in Macedonia show patterns like volatility clustering and high kurtosis. The asymmetric models showed some evidence of a leverage effect. The mean equation only weakly showed a risk premium. The results were consistent across different error distributions. Models with non-Gaussian error distributions were better at forecasting accuracy than those with normal distributions.