Using vector autoregressive models, this paper finds that shocks that have no immediate effect on the price level explain essentially all short-run variance of aggregate output while shocks that immediately affect price explain virtually none of that variance. Similar findings are obtained in the bivariate model using sectoral, industry-level, seasonally adjusted and non-seasonally adjusted data. In models with aggregate data, shocks that immediately raise the price level eventually cause output to fall while shocks that affect price with a lag immediately raise output and eventually cause the price level to rise. These responses combined with the variance decompositions suggest the short-run aggregate supply curve is nearly horizontal and the aggregate demand curve is essentially vertical. A statistical model that identifies shocks that don't affect prices for at least two months is also developed. In general, shocks with the slowest effect on prices explain most, if not all, of the short-run output variance. Our robust findings reject theories in which prices adjust rapidly to clear markets. However, the evidence that output and price are nearly orthogonal for a substantial amount of time is difficult to reconcile with standard models of sticky price adjustment. We also find that the vast majority of the long-horizon variance of aggregate output is explained by shocks that raise output and the price level. That finding provides support for theories of aggregate demand non-neutrality in which a positive movement in aggregate demand permanently increases the level of output.