Notes on the Neglected Premisses of the Hodrick-Prescott Detrending and the Hamilton Regression Filter

Abstract

The Hodrick-Prescott filter is a convenient and therefore widely and routinely applied detrending method in macroeconomics working with empirical data. However, James Hamilton has recently gained attention with his vigorous advice against it and a proposal of a better alternative. Before abandoning Hodrick-Prescott and uncritically switching to the Hamilton regression filter, or before by force of habit ignoring Hamilton's contribution altogether, this paper, in a nontechnical and elementary manner, provides a little methodological reflection about the premisses behind the two approaches. In addition, it sets up a stylized oscillatory scenario in which the Hamilton filter dramatically misjudges the trend. On the other hand, it sketches a modification of the Hodrick-Prescott approach and also a search strategy that, at least under similar conditions, can help find a more appropriate degree of trend smoothing than the conventional choice of lambda = 1600 for quarterly data.