A COMPARATIVE ECONOMETRIC STUDY OF ARIMA-GARCH MODELS IN MODELING SAUDI ARAMCO STOCK VOLATILITY

Abstract

This article examines the volatility dynamics of the daily adjusted closing prices of Saudi Aramco (SAOC) shares using a hybrid econometric framework that combines ARIMA and CS-GARCH models. Financial time series are typically characterized by volatility clustering, conditional heteroskedasticity, and asymmetric responses to shocks, which limit the effectiveness of purely linear models. To address these challenges, this research applies an ARIMA(2,1,2) model to capture linear dependencies in the conditional mean and a CS-GARCH (0,1) specification to model time-varying conditional variance and decompose volatility into short-run and long-run components. Using data spanning January 2, 2022, to May 8, 2025 (834 observations), the results show that the series is integrated of order one and exhibits significant heteroskedasticity. After model comparison involving more than 750 GARCH-type specifications, the ARIMA(2,1,2)-CS-GARCH(0,1) model with a skewed normal distribution was identified as the optimal specification based on information criteria and diagnostic tests. The findings indicate strong persistence in the long-run volatility component, while short-run shocks display temporary effects that gradually dissipate. Out-of-sample forecasting results confirm that the hybrid model outperforms the conventional ARIMA model across all accuracy measures (MAPE, RMSE, MAE, and MSE). The study concludes that hybrid ARIMA-CS-GARCH models provide a robust and reliable framework for modeling and forecasting stock volatility in highly dynamic markets, offering valuable implications for investors, portfolio managers, and risk analysts in the Saudi financial market.