Macroeconomic forecasting based on high-dimensional datasets: a smooth transition three pass regression filter model

This study delves into the ramifications of incorporating various structural state alterations observed within macroeconomic datasets on the forecasting accuracy of macroeconomic variables. To explore the existence of mechanism transitions, we developed a Smooth Transition Three-Pass Regression Filter (STR-3PRF) model based on the foundation of the three-pass regression filter. The extended model can capture dynamic changes and mechanism shifts within the economic environment. Empirical results indicate that high-dimensional datasets provide effective predictive information for macroeconomic variables. Furthermore, the STR-3PRF model captures the dynamic non-linear relationships between the macroeconomic dataset and macroeconomic variables. Noteworthily, comparative assessments reveal the STR-3PRF model’s significantly superior predictive efficacy relative to alternative models. Additionally, the STR-3PRF model has passed various robustness tests. Overall, our findings affirm the profound influence of regime transitions in macroeconomic datasets on the predictability of macroeconomic variables.