<![CDATA[Abstract. Amid persistent global economic uncertainty and elevated inflationary pressure in 2026, accurate inflation forecasting has become essential for effective monetary policy. This study forecasts monthly (Month-over-Month) inflation in Indonesia, South Korea, and Thailand for the period January 1986–June 2026 using a Vector Autoregression (VAR) model, comparing specifications without and with the addition of seasonal dummies. Optimal lag selection based on the Akaike Information Criterion (AIC) produced a VAR(13) model. Diagnostic tests show that the model without seasonal dummies still exhibits residual autocorrelation, while the model with seasonal dummies is free of autocorrelation and shows a reduced ARCH effect, particularly for South Korea. Both models are confirmed stable based on inverse roots of the AR characteristic polynomial. Granger causality tests reveal that only the relationship from Thailand to South Korea remains consistently significant across both models, while four other causal directions change in significance after seasonal dummies are added, indicating that some previously identified relationships were influenced by shared seasonal patterns rather than genuine economic causality. Twelve-month-ahead forecasting results (July 2026–June 2027) show that the VAR(13) model with seasonal dummies consistently produces lower MAE and RMSE for all three countries, making it a more reliable model for monthly inflation forecasting. Abstrak. Di tengah ketidakpastian ekonomi global yang masih berlanjut dan tekanan inflasi yang meningkat pada 2026, kemampuan meramalkan inflasi secara akurat menjadi hal penting bagi efektivitas kebijakan moneter. Penelitian ini bertujuan meramalkan inflasi bulanan (Month-over-Month) Indonesia, Korea Selatan, dan Thailand periode Januari 1986–Juni 2026 menggunakan model Vector Autoregression (VAR), dengan membandingkan model tanpa dan dengan penambahan seasonal dummy. Penentuan lag optimal berdasarkan Akaike Information Criterion (AIC) menghasilkan model VAR(13). Hasil uji diagnosis menunjukkan model tanpa seasonal dummy masih menunjukkan autokorelasi residual, sedangkan model dengan seasonal dummy bebas autokorelasi dan menunjukkan penurunan efek ARCH, khususnya pada Korea Selatan. Kedua model dinyatakan stabil berdasarkan inverse roots pada AR characteristic polynomial. Uji kausalitas Granger menunjukkan hanya hubungan Thailand terhadap Korea Selatan yang tetap signifikan secara konsisten pada kedua model, sementara empat arah kausalitas lainnya mengalami perubahan signifikansi setelah seasonal dummy ditambahkan, mengindikasikan bahwa sebagian hubungan yang teridentifikasi sebelumnya dipengaruhi oleh kesamaan pola musiman, bukan murni hubungan kausal ekonomi. Hasil peramalan 12 bulan ke depan (Juli 2026–Juni 2027) menunjukkan bahwa model VAR(13) dengan seasonal dummy secara konsisten menghasilkan nilai MAE dan RMSE yang lebih rendah pada ketiga negara, sehingga lebih direkomendasikan untuk peramalan inflasi bulanan.]]>
