Detrending time-aggregated data

This paper examines the combined influences of detrending and time aggregation on the measurement of business cycles. The results indicate that time aggregation and detrending do indeed interact in a non-trivial manner. In particular, detrending filters that pass through substantial high-frequency variation tend to distort the basic shape of disaggregate spectra and cospectra when applied to time-aggregated data and can even give the illusion of business-cycle variation in aggregate data when none is present in the basic data.