Report

Extracting business cycle fluctuations: what do time series filters really do?

Abstract: Various methods are available to extract the "business cycle component" of a given time series variable. These methods may be derived as solutions to frequency extraction or signal extraction problems and differ in both their handling of trends and noise and their assumptions about the ideal time-series properties of a business cycle component. The filters are frequently illustrated by application to white noise, but applications to other processes may have very different and possibly unintended effects. This paper examines several frequently used filters as they apply to a range of dynamic process specifications and derives some guidelines for the use of such techniques.

Keywords: Business cycles; time series analysis;

Bibliographic Information

Provider: Federal Reserve Bank of New York

Part of Series: Staff Reports

Publication Date: 2007

Number: 289