Evaluating the Robustness and Implementing Double One-Sided Hodrick-Prescott for Cycle Extraction

Abstract

Cycle extraction is a crucial part of a business cycle analysis. This research aims to identify the robustness of cycle extraction methods, which indicates a business cycle’s reliability. Interestingly, this research proposes the Double One-sided Hodrick-Prescott (DoneHP) for cycle extraction. This method was compared to four well-known methods, including the Hodrick-Prescott (HP), Double Hodrick-Prescott (DHP), Christiano-Fitzgerald (CF), and One-sided Hodrick-Prescott (OneHP) analyses. In addition to evaluating the revision of size and signal, the study also assessed the revision of the turning point.

Using four choices of Thailand’s reference series, the results of the expanding window experiment showed the consensus that DoneHP was outstanding and had less revied cycle values, especially at the end of the cycle. Based on the overall ranking, the order of performance for the methods was DoneHP, OneHP, DHP, HP, and CF. The findings suggest DoneHP for cycle extraction in cases where the analysis is made in real-time and minimizing the revision of past estimation is preferred. In detail, DoneHP is recommended, followed by OneHP, HP, DHP, and CF when the priority is minimizing size revisions. However, the emphasis focuses on a steady signal and turning point, preferring as follows DoneHP, OneHP, DHP, HP, and CF.