Previous studies argue that financial variables do not help forecast U.S. output growth. F statistics for excluding financial variables from output growth equations depend on the sample period and the inclusion of 1974:12 in the sample. Also, an autoregressive model of output growth often provides better forecasts than models with lagged financial variables included. I decompose output into permanent and cyclical components and ask whether financial variables help forecast either component in isolation. The paper-bill spread does improve in-sample forecasts of cyclical output, but no financial variable helps forecast either cyclical output or permanent output growth out of sample. Copyright 2002, Oxford University Press.