Uncertainty and Monetary Policy: Assessing their Nonlinear Interactions

This thesis assesses the interactions between uncertainty and monetary policy by means of nonlinear econometric methods. It consists of three separate chapters. The first chapter is concerned with the effects of monetary policy shocks conditional on different levels of uncertainty. On the basis of the theoretical literature, several explanations are thought to be able to reduce the effectiveness of monetary policy during uncertain times (e.g., real option effects, firm-price setting behavior and precautionary savings). In order to empirically assess theoretical predictions I estimate a nonlinear Interacted VAR model, where, in a novel way with respect to the literature, I model the conditioning indicator - uncertainty, in my case, which discriminate "high" from "low" uncertainty states - endogenously in the VAR. This implies the necessity to adopt the Generalized Impulse Response Functions à la Koop, Pesaran and Potter (1996). This strategy enables me to consider both the possible endogenous reaction of uncertainty to the policy shock and its feedbacks on the dynamics of the system. My findings suggest that monetary policy shocks are significantly less effective during uncertain times, with the peak reactions of a battery of real variables being about two-thirds milder than those during tranquil times. I also find that uncertainty decreases after an expansionary monetary policy shock. Further, I show that, consistently with Vavra's (2014) predictions, the reaction of prices appears greater during firm-level uncertain times. The second chapter (coauthored with G. Caggiano and E. Castelnuovo) is concerned with the impact of uncertainty shocks at the zero lower bound (ZLB), which has been hit since December 2008 in the U.S. On the theoretical side, several recent studies suggest that when monetary policy is constrained by the ZLB, uncertainty shocks should generate a much larger and persistent drop in real activity (see Fernandez-Villaverde, Guerron-Quintana, Kuester, and Rubio-Ramirez (2015), Johannsen (2013), Nakata (2013), and Basu and Bundick (2014, 2015)). However, on the empirical side, no analysis explicitly modeling the nonlinearity of the real effects of uncertainty shocks due to the ZLB has been proposed so far. To this aim we employ a parsimonious nonlinear Interacted-VAR model to examine whether the real effects of uncertainty shocks are greater when the economy is at the ZLB. Our results show that the contractionary effects of uncertainty shocks are statistically larger when the ZLB is binding, with differences that are economically important. Such differences are shown not to be driven by the contemporaneous occurrence of the Great Recession. The third chapter returns on the argument of Chapter 1 with the aim of enquiring on the structural reasons behind the lower effectiveness of monetary policy shocks during uncertain times. To do so I adopt the lens of the workhorse New Keynesian model. In particular, I propose a simple state-conditional Minimum Distance estimation strategy of a DSGE model, which I apply to the Altig, Christiano, Eichenbaum and Lindè's (ACEL, 2011) model. The estimator matches as closely as possible the regime-dependent responses coming from an unrestricted Threshold VAR model with the corresponding model-based responses. This approach may capture possibly unmodelled mechanisms through regime-specific estimates of structural parameters. I find the ACEL model to be remarkably able to match the VAR impulse responses, particularly in tranquil times. This performance is driven by very different estimated values for some key-structural parameters in the two states. A higher slope of the new-Keynesian Phillips curve, a higher cost of the variation in capital utilization, and a lower degree of habit formation in consumption are shown to be behind the model ability to predict the lower real effects of monetary policy shocks in periods of high uncertainty.