In a bivariate Vector Autoregressive approach Blanchard and Quah [1989] used the unemployment rate as a stationary indicator to disentangle the effects of permanent and transitory innovations to output. This paper examines the stability of their results with respect to a wider class of time series models and the choice of the stationary variable. We take a Vector Autoregressive Moving Average approach to allow for a more parsimonious representation of the process. Furthermore, capacity utilization and the rate of inflation are used as alternative cyclical indicators. It turns out that the decomposition is fairly stable, with the exception of the model where we used the GNP-deflator as stationary variable. Final1y,,we discuss the interpretation of permanent and transitory innovations as supply and demand innovations.