This paper aims to reveal the asymmetric impact of unexpected news shocks on stock return volatility, and to specify and select the model that most accurately estimates and forecasts volatility. To this end, sign bias tests and size bias tests were conducted on four models—the GARCH model, EGARCH model, AGARCH model, and GJR model—to examine model misspecification. The results show that the GARCH model tends to underestimate the impact of bad news and overestimate the impact of good news compared to the AGARCH and GJR models, which capture asymmetric effects. However, the EGARCH model evaluates conditional variance lower than the GARCH model for both good and bad news, making it unsuitable for explaining the conditional variance of stock returns. Through predictive power comparisons, the GJR model is identified as the most appropriate model for capturing volatility.