Is cyclical real GNP really more persistent than the trend?
Using an observed components decomposition to estimate permanent and temporary components of US real GNP, Campbell and Mankiw (1987) found that the estimated temporary component displays greater persistence than the estimated permanent component. This suggests that attempts to estimate separate permanent and temporary components in real GNP are unlikely to succeed. This finding is re-examined here using a longer real GNP series and a different decomposition, which guarantees that the estimated temporary component will be stationary. The resulting estimates of permanent and temporary output suggests that the temporary component displays no persistence and that the permanent component exhibits almost precisely the persistence of a random walk with drift. Cyclical real GNP is not more persistent than the trend.