PEMODELAN HARGA BERAS: EVALUASI POLA MUSIMAN, EFEK VOLATILITAS, DAN IDENTIFIKASI MODEL TERBAIK

Abstract

Rice is a strategic staple food commodity in Indonesia whose price dynamics are heavily influenced by seasonal patterns and potential supply shocks. This study aims to comprehensively analyse wholesale rice price modelling by evaluating seasonal behaviour, testing for conditional volatility effects (ARCH/GARCH), and identifying the best forecasting model. The dataset consists of monthly national average wholesale rice prices from January 2010 to December 2025 (192 observations). The analytical framework involves stationarity testing (ADF and Box-Cox), ARIMA/SARIMA model estimation, residual diagnostic checking (Ljung-Box test), and heteroskedasticity testing (ARCH LM test). The empirical findings reveal that rice price movements are strongly governed by annual 12-month seasonal cycles. The model emerged as the optimal mean specification, yielding the lowest information criteria, and remarkable forecasting accuracy. All estimated parameters were statistically significant (), and the residuals satisfied the white noise assumption. Furthermore, tests on squared residuals showed no evidence of ARCH/GARCH effects (), confirming homoskedastic error variance. This indicates that government price stabilization policies have successfully mitigated extreme volatility clustering. Consequently, complex volatility modeling is redundant, rendering the pure as the most efficient best-fit model to serve as an Early Warning System for national food security.

Keywords:  Rice Price, Seasonal Patterns, Volatility, SARIMA, ARCH/GARCH