LOESS Asymmetric Filters for Real Time Economic Analysis

The local polynomial regression predictors developed by Cleveland (1979) have been applied to estimate the short-term trend of season-ally adjusted economic indicators and implemented in several statisti-cal packages. The main purpose of this study is to introduce a RKHS representation of the LOESS smoother with particular emphasis on the asymmetric lters applied to most recent observations. The lters obtained by means of the RKHS are shown to have superior proper-ties relative to the classical ones from the view point of signal passing, noise suppression and revisions. We compare the performance of the kernel representations relative to the classical lters using real life se-ries.