In applied macroeconomics, especially business cycle studies, the most comm
only used filters for removing time trend and noise are Linear filters. In
this paper we introduce a class of nonlinear filters that has been recently
proven very powerful for removing time trend and noise in signal processin
g. Interesting applications to economics are carried out in an example of f
orecasting post-war US recessions. (C) 1999 Elsevier Science S.A. All right
s reserved.