Comparison of ARCH and GARCH Model Performance in Estimating LQ45 Stock Return Volatility

Abstract

Volatility is a crucial indicator in capital market analysis because it reflects the risk level of an investment instrument. Accurate volatility estimation is essential for investors, portfolio managers, and regulators to support informed decision-making. This study aims to compare the performance of the Autoregressive Conditional Heteroskedasticity (ARCH) and Generalized Autoregressive Conditional Heteroskedasticity (GARCH) models in estimating the return volatility of stocks included in the LQ45 index on the Indonesia Stock Exchange. The LQ45 index was chosen because it comprises stocks with high liquidity and market capitalization, making it a key representative of stock market movements in Indonesia. The data used are daily LQ45 returns over a specific research period. The analysis begins with a heteroscedasticity test using the Breusch-Pagan test to ensure the presence of non-constant variance. Next, ARCH and GARCH model estimations are performed, as well as model performance evaluations based on the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC). The results show that the GARCH model tends to provide more stable volatility estimates and better fits the characteristics of the financial data compared to the ARCH model. This finding indicates that GARCH can be used as a more reliable model for measuring LQ45 stock return volatility.