Ambiguous business cycles: a quantitative assessment

Abstract

In this paper, we examine the cyclical dynamics of a Real Business Cycle model with ambiguity averse consumers and investment irreversibility using the smooth ambiguity model of Klibanoff et al. (2005, 2009). Ambiguity of belief about the productivity process arises as agents do not know the process driving variation in aggregate TFP, and they must make inferences regarding the true process at the same time as they infer the behavior of the unobserved temporary component using a Kalman filtering algorithm. Our findings may be summarized as follows. First, the standard business cycle facts hold in our framework, which are not altered significantly by changes in the degree of ambiguity aversion. Second, we demonstrate a role for information and learning effects, and show that lower initial ambiguity or greater confidence coupled with learning dynamics lowers the volatility and increases the persistence in all of the key macroeconomic variables. Third, comparing the performance of our model to the New Keynesian business cycle model of Ilut and Schneider (2014) with maxmin expected utility, we find that the version of their model without nominal and real frictions turns out to have limited success at matching the moments for the quantity variables. In the maxmin expected utility framework, the worst case scenario instills too much caution on the part of agents who, in the absence of a key set of nominal and real frictions, end up excessively reducing their responses to TFP shocks.