Macroeconomics Seminar Series: Uncertainty-Driven Comovements in Booms and Busts: A Structural Interpretation

We estimate a nonlinear VAR to quantify the responses of output, consumption, investment, and hours to a financial uncertainty shock in booms and busts in the post-WWII U.S. data. We find evidence of comovements both in expansions and in recessions, with stronger responses of all real activity indicators in the latter state. We interpret this state-dependent responses with a version of the Basu and Bundick (2017) model in which an uncertainty shock conceptually comparable to the one used in our VAR analysis generates comovements in real activity. A state-contingent estimation of this model conducted via Bayesian direct inference points to counter-cyclical risk aversion as the crucial ingredient to replicate the evidence produced with our nonlinear IVAR. An exercise focusing on the great recession suggests that the nonlinear DSGE model is able to replicate about 50% of the cumulative output loss in the 2009-2014 period, twice as much what the same model would predict if estimated conditional on a linear VAR.