The Hodrick-Prescott Filter with a Time-Varying Penalization Parameter. An Application for the Trend Estimation of Global Temperature.

In this paper we use the Hodrick-Prescott filter for analysing global temperature data. We are especially concerned with a reliable estimation of the trend component at the end of the data sample. To this end we employ time-varying values for the penalization parameter. The optimal values are derived by a comparison with an ideal filter. The method is applied to temperature data for the northern hemisphere from 1850 to 2012. The main result is that for the optimal specification of the flexible penalization the trend component of temperature is still increasing, possibly with a somewhat lower pace.