A real-time simulation study on the behaviour of alternative trend-cycle decomposition methods during the pandemic : 2024 edition

This paper conducts a comparative analysis of the trend and cycle estimates derived from the Hodrick-Prescott method, the Christiano-Fitzgerald approximation and the Harvey and Trimbur model for selected euro area (EA) indicators. These three methods are applied to data spanning the last 3 decades, including the COVID pandemic. An additive outlier correction is introduced to handle the most volatile observations in the period of the COVID pandemic. A case study is also presented that repeats the estimation over 30 consecutive vintages, starting in January 2020 and ending in May 2022. The results indicate that the estimates for the euro area's GDP and employment are very similar across the three filters. However, the estimates for the euro area's industrial production index (IPI) provided by the Harvey and Trimbur model indicate that the trend is higher in periods of IPI expansion and lower in periods of contraction such as the financial crisis and the COVID pandemic. Moreover, the case study shows that the estimates are very stable across vintages from winter 2020 when outlier corrections are added to all the methods. Before the introduction of outlier corrections, the estimates are often too flat and, in some cases, unreliable.