<![CDATA[This study evaluates the performance of Auto-Regressive Integrated Moving Average (ARIMA), Error-Trend-Seasonality (ETS), and Holt-Winters frameworks in forecasting state tax revenue collections in Indonesia. Accurate tax revenue prediction is essential for prudent fiscal governance, yet comparative assessments of these methods remain limited in the existing literature. Using monthly tax collection data from January 2016 to March 2024, we employed the period from January 2016 to March 2023 for model identification and the remaining twelve months for out-of-sample validation. Three key findings emerge from our evaluation using Mean Absolute Error (MAE), Mean Squared Error (MSE), and Root Mean Squared Error (RMSE) metrics. First, all three frameworks demonstrate satisfactory predictive accuracy for Indonesian tax revenues. Secondly, each method shows distinct strengths, such as ARIMA performs optimally for withholding tax on employment (PPh 21), ETS achieves superior results for domestic value-added tax (PPN DN), and Holt-Winters provides the most accurate forecasts for personal income tax (PPh OP). Finally, the ETS framework generally outperforms both ARIMA and Holt-Winters across most tax categories. While these findings are specific to Indonesia, the methodological framework offers transferable insights for tax authorities seeking to enhance revenue forecasting and inform evidence-based policymaking.]]>
Copyrights © 2026
