DETRENDING AND TURNING-POINTS

This paper examines the sensitivity of turning points classification to different detrending methods and the ability of each method to replicate NBER dating. We use two different turning point rules and a variety of detrending methods to compute the cyclical component of output. We show that while differences are minor with the first rule, the results are extremely sensitive to detrending with the second rule. Many detrending procedures generate false alarms and many miss several commonly classified turning points. The output series detrended with the Hodrick and Prescott filter and with a frequency domain masking of the low frequency components of the series reproduce all NBER turning points with at most two quarters lead or lag, regardless of the dating rule used.