The Empirical Implications of the Interest-Rate Lower Bound

Abstract

Using Bayesian methods, we estimate a nonlinear DSGE model in which the interest-rate lower bound is occasionally binding. We quantify the size and nature of disturbances that pushed the US economy to the lower bound in late 2008 as well as the contribution of the lower bound constraint to the resulting economic slump. We find that the interest-rate lower bound was a significant constraint on monetary policy that exacerbated the recession and inhibited the recovery, as our mean estimates imply that the zero lower bound (ZLB) accounted for about 30 percent of the sharp contraction in US GDP that occurred in 2009 and an even larger fraction of the slow recovery that followed. (JEL C11, C32, E12, E23, E32, E43, E52, G01)

The zero lower bound: frequency, duration, and numerical convergence

“…For example, when φ π = 1.5, p 11 = 0.95, and p 22 = 0.5, the average ZLB event is only two quarters, but the maximum ZLB event in a quarter simulation is 15 quarters, which is closer to ZLB events observed in the data. Furthermore, Gust, López-Salido, and Smith (2013) argue that this average duration is consistent with the expectations found in financial market and survey data.…”

“…The data prefers highly persistent shocks (i.e., ρ β > 0.95) with a standard deviation that is over four times the maximum value inside the convergence region. While the model is slightly different, the estimates of the constrained nonlinear model in Gust, López-Salido, and Smith (2013) are also well outside of the convergence region. They estimate that ρ β = 0.88 and σ υ = 0.0025, which may be possible in our model with a persistent Taylor rule or more price stickiness.…”

“…A recent segment of the ZLB literature uses global solution methods to solve fully nonlinear models with a ZLB constraint (e.g., Wolman 2005; Basu and Bundick 2012;Fernández-Villaverde et al 2012;Nakata 2012;Aruoba and Schorfheide 2013;Gust, López-Salido, and Smith 2013;Mertens and Ravn 2014;Richter, Throckmorton, and Walker 2013;Gavin et al 2014). However, all of the work on determinacy uses a perfect foresight setup (e.g., Benhabib, Schmitt-Grohe, and Uribe 2001a,b;Alstadheim and Henderson 2006).…”

L’étude des fluctuations macroéconomiques est-elle « scientifique » ?

“…Dans le présent contexte, « paradoxe du labeur » (paradox of toil), selon lequel les chocs d'offre négatifs deviennent expansionnistes en trappe à liquidité (en raison de leur impact inflationniste), fournit l'expérience cruciale recherchée 4 . -la seconde approche consiste à spécifier un modèle d'équi-libre général complet, dans lequel le mécanisme de spirale déflationniste est présent, pour ensuite l'estimer empiriquement (voir, par exemple, Christiano et al, 2015 ;Gust et al, 2017). Cette approche permet de mesurer l'ensemble de la chaîne causale postulée par la théorie, puis de construire des scénarios alternatifs (« contrefactuels ») qui décrivent comment l'économie se serait comportée si cette chaîne causale avait été brisée (par exemple si la banque centrale avait pu mettre en oeuvre des taux d'intérêt négatifs).…”

Forecasting with Shadow-Rate VARs

“…Similarly, Iwata and Wu (2006), Berg (2017), and Chan and Strachan (2014) consider only censoring of the VAR's left-hand side variables, without tracking the underlying, uncensored shadow rate as a potential predictor. The inclusion of lagged shadow rates as VAR predictors could, however, be potentially relevant as a means of tracking make-up policies at the ELB, as discussed by, among others, Reifschneider and Williams (2000), Gust, et al (2017), and Billi (2020). 7 Johannsen and Mertens (2021) provide an out-of-sample forecast evaluation for short-and long-term nominal interest rates in a model smaller than our VARs, and find their unobserved components shadow-rate model to be competitive with the no-arbitrage model of Wu and Xia (2016), but do not consider forecasts of other variables.…”