Estimating nonlinear effects of fiscal policy using quantile regression methods

Abstract

We use quantile regression methods to estimate the effects of government spending shocks on output and unemployment rates. This allows to uncover nonlinear effects of fiscal policy by letting the parameters of either vector autoregressive models or local projection regressions vary across the distribution of macroeconomic activity. In quarterly US data, we find that fiscal output multipliers are notably larger if GDP is predicted to be below trend. Conversely, higher government spending appears to significantly lower unemployment only if the unemployment rate is in the largest deciles of its conditional distribution.