Abstract

We estimate nonlinear effects of government spending shocks on US macroeconomic activity using quantile regression methods. This amounts to allowing regression parameters to depend on how far output or the unemployment rate are from their means, conditional on past explanatory variables. Applying quantile methods to vector autoregressions and local projections as an alternative way to estimate impulse response functions, we find the output effects of fiscal policy to be notably larger for lower quantiles of the conditional output distribution. Conversely, higher government spending appears to lower the rate of unemployment considerably only at its highest deciles.

You do not currently have access to this article.