Does Structural Breaks Improve Forecast Accuracy for Malaysian Macroeconomic Indicators? An ARIMA Model Analysis

Abstract

This study employs the Autoregressive Integrated Moving Average (ARIMA) model and the Chow test to identify structural breaks in macroeconomic indicators, which are crucial for understanding the dynamics of economic systems. By analyzing a dataset of macroeconomic indicators, this research aims to detect and model the changes in these indicators over time, providing valuable insights for policymakers and researchers. Structural breaks, often caused by economic events or policy changes, can sometimes significantly impact the accuracy of time series models. The presence of structural breaks is tested using the Chow test, and the results are compared to those without breaks. The analysis focuses on three ARIMA models with different parameters and evaluates their performance using root mean squared error (RMSE) and mean absolute percentage error (MAPE). The results indicate that the models with structural breaks exhibit higher RMSE and MAPE values compared to those without breaks. Specifically, the ARIMA (11,0,2) model shows a significant increase in RMSE and MAPE when a structural break is introduced, while the ARIMA (12,0,4) model exhibits a smaller but still noticeable increase. In contrast, the ARIMA (9,1,11) model demonstrates relatively better performance with and without structural breaks. The results show that traditional ARIMA models provide more accurate forecasts than ARIMA models adjusted for data breaks. Incorporating structural breaks results in less accurate forecasts. The presence of a structural break negatively impacts the forecasting performance of this model, leading to larger errors. Our findings suggest that structural breaks of minor magnitude in time series data should be disregarded by policymakers and economists to improve the reliability of their forecasts.