Analisis Volatilitas dan Forecasting Return Saham Bank Mandiri Menggunakan ARIMA-GARCH: Proyeksi 5 Bulan ke Depan

Abstract

ANALYSIS OF VOLATILITY AND FORECASTING RETURN OF BANK MANDIRI SHARES USING ARIMA-GARCH: PROJECTIONS FOR THE NEXT 5 MONTHS

Indonesia’s capital market continues to demonstrate strong dynamics, particularly among blue-chip stocks such as PT Bank Mandiri (Persero) Tbk (BMRI), which exhibit high volatility influenced by macroeconomic variables, global conditions, and market sentiment. To obtain a more accurate projection of BMRI’s price movement, this study applies the ARIMA-GARCH model to forecast stock prices for the five-month period from November 2025 to April 2026. The dataset consists of daily closing prices from November 2020 to November 2025. Preprocessing involved handling outliers using winsorizing and conducting a stationarity test, in which the Augmented Dickey-Fuller result indicated that the return series was stationary with a p-value of 0.0000. Based on information criteria, ARIMA(0,0,1) was selected as the optimal mean model with an AIC value of –3985.7041. Volatility was subsequently modeled using GARCH(1,1) with a normal distribution, showing a persistence level of 0.8598, suggesting that BMRI volatility remains elevated and requires time to stabilize after a shock. Diagnostic tests confirmed the absence of residual autocorrelation (Ljung-Box p-value > 0.05) and no remaining ARCH effect (p-value 0.1744), validating the adequacy of the combined model. Forecasting through 1,000-iteration Monte Carlo simulation produced an average projected final price of Rp 4,960, up from Rp 4,730, indicating an expected change of +4.82%. Risk metrics showed a 95% Value at Risk (VaR) of Rp 3,550 and an Expected Shortfall of Rp 3,340. The model’s performance on the test dataset yielded an RMSE of 0.014704 and an sMAPE of 75.7663%.