Real-time US GDP gap properties using Hamilton’s regression-based filter

Abstract

It has been previously documented that the Hodrick–Prescott (HP) filter exhibits real-time instability such that the estimates of trend and cycle components from an aggregate time series are revised both as more data become available and previously available data are revised. An alternative to the HP filter has recently been suggested by Hamilton (Rev Econ Stat 100(5):831–843, 2018). The current article investigates and compares the HP and the Hamilton filters with respect to real-time stability in US GDP gap estimation. The results reveal that the Hamilton filter outperforms the HP filter when it comes to real-time revisions. The source of the inferior performance of the HP filter is found to be the fact that component estimates close to the end of the sample are revised to a large extent for the HP filter even when only a few more data points are added to the sample.