Abstract This paper presents new evidence on the persistence of fluctuations in real GNP. We estimate two measures of persistence nonparametrically using post-war quarterly data from Canada, France, Germany, Italy, Japan, the United Kingdom, and the United States. We compare these estimates with Monte Carlo results from various AR(2) processes. For six out of seven countries, the point estimates indicate that a 1% shock to output should change the long-run unvariate forecast of output by well over 1%. Low-order ARMA models yield similar conclusions. Finally, we examine the persistence in relative outputs of different countries.