Seasonal adjustment and the univariate testing of asymmetry

Abstract

Following Holly and Stannett, a standard approach has emerged in the implementation of Sichel's tests of business cycle asymmetry, whereby seasonally adjusted data is detrended via Hodrick-Prescott filtering. While Speight and MacMillan have considered the impact of alternative methods of detrending upon these tests, the role of seasonal adjustment has yet to be examined. In this letter the impact of seasonal adjustment is examined by means of an application to durable, non-durable and aggregate UK consumers' expenditure. Whereas alternative means of detrending most noticeably effect the deepness test and its significance, alternative methods of seasonal adjustment are shown to primarily effect the steepness statistic and its size.