Government spending and heterogeneous consumption dynamics

Abstract

What is the effect of government spending on consumption? Neoclassical and new Keynesian models deliver opposing predictions and empirical studies differ in their conclusions. This paper takes a data-driven approach and exploits information from over 200 variables via a Bayesian factor-augmented VAR model. I use sign restrictions for identifying effects that both classes of models agree upon. This approach imposes a minimal set of restrictions on the empirical model, leaving the sign and magnitude of the effect of government spending on consumption free to be determined by the data. I find that: (i) government spending increases aggregate consumption; (ii) the estimated spending multiplier is close to 2; (iii) there exists heterogeneity even within durable, nondurable, and service consumption variables which is undocumented in the literature. Finally, I show that my identified structural shocks are not predictable by economic agents and are uncorrelated with traditional monetary policy shocks, suggesting that these effects are not confounded by monetary policy.