Does the choice of detrending method matter in demand analysis?

This paper examines the selection of optimal detrending methods in demand analysis with time series data by measuring the distortionary effect of linear filters on the spectrum of detrended time series. The paper identifies substantial distortions of popular detrending methods in economics (including first-differencing and deterministic linear detrending), while the Hodrick-Prescott and Baxter-King filters produce much smaller distortions in detrended time series. Consequences of alternative detrending approaches are further illustrated by estimating the almost ideal demand system in Japan for major consumption categories.