Elsevier

Economics Letters

Volume 116, Issue 3, September 2012, Pages 422-425
Economics Letters

Inflation-regime dependent effects of monetary policy shocks. Evidence from threshold vector autoregressions

https://doi.org/10.1016/j.econlet.2012.04.027Get rights and content

Abstract

We use a threshold vector autoregression to study the effects of monetary policy shocks on the US. Depending on the level of inflation we note important regime dependence in the inflation response to monetary policy shocks.

Highlights

► We estimate a VAR with threshold effects for the US over the 1965–2007 period. ► We present evidence for significant threshold effects depending on inflation. ► Standard effects of monetary policy shocks are found for the high inflation regime. ► In the low inflation regime inflation declines quicker and there is no price puzzle.

Introduction

Since the mid 1990s a very successful research program has studied the effects of monetary policy on macroeconomic variables. These effects have been identified by estimating the dynamic responses of output, inflation and other variables to “monetary policy shocks” in vector autoregressive (VAR) models of the economy.

This paper investigates the stability of these results by studying threshold effects in the standard “monetary policy” VAR model. Our results show strong evidence for nonlinearities in the effects of monetary policy shocks on output and inflation dependent on the level of the inflation rate.

Our analysis is related to the recent literature concerning the robustness of the conventional VAR evidence about monetary policy shocks. For example, estimating the canonical VAR model on post-1985 observations leads to results that differ from the standard evidence in important respects (Mojon, 2008). In particular, the responses of output and inflation to a monetary policy shock are not significantly different from zero. Mojon (2008) argues that these differences are the result of shifts in the mean of inflation.

Instead of being exogenous these changes might actually be triggered by the state of the economy. In this paper we focus on the level of inflation to cause switches between regimes. This can be motivated theoretically, for example, by the effects of high inflation on central bank credibility and on the formation of the private sector’s inflation expectations which might affect the Phillips curve relationship and inflation dynamics. The level of inflation might also cause changes in the central bank’s monetary policy reaction function. For example, Orphanides and Wilcox (2002) argue that the strength of the central bank’s response to an inflationary shock depends in a nonlinear fashion on the deviation of inflation from its target and becomes much stronger if the shock pushes the inflation rate outside of a target zone.

A straightforward way to model nonlinearities like these empirically is the estimation of a threshold model that allows for different sets of model parameters depending on the state of the economy. Univariate threshold autoregressions originally introduced by Tong (1978) have been extended to a multivariate context by Tsay (1998) and Balke (2000).

Section snippets

Econometric methodology

The threshold vector autoregressive (TVAR) model with two regimes can be written as: Yt=μ1+A1Yt+B1(L)Yt1+(μ2+A2Yt+B2(L)Yt1)I(ctd>γ)+ut.Yt is a vector of endogenous variables. I is an indicator that equals 1 when the threshold variable ctd exceeds γ and 0 otherwise. If I=0 the dynamics of the VAR are given by the vector of constants μ1, the matrix of contemporaneous interaction coefficients A1 and the matrix of lag polynomials B1(L). If I=1 the relevant coefficients are μ1+μ2, A1+A2 and B1(L)

Threshold tests and estimates

Yt includes the standard variables from the VAR literature on monetary policy shocks (e.g. Christiano et al., 1999). We use quarterly observations from 1965Q3 to 2007Q2 on real GDP, the GDP deflator and the monetary aggregate M1. The indicator for monetary policy is the Federal Funds Rate. As in standard VAR studies we also include an indicator of commodity prices (e.g. Christiano et al., 1999).

Including non-stationary data in the VAR might lead to spurious non-linearities (Calza and Sousa, 2006

Conclusions

This paper presents evidence for regime dependent effects of monetary policy shocks in the US The standard results from the literature are obtained for a regime of higher inflation. A second regime with lower inflation shows no sign of a price puzzle and a stronger decline in inflation. The explanatory power of output growth, inflation and monetary policy shocks for unexpected changes to these variables is shown to differ across regimes as well.

Acknowledgments

I am indebted to Peter Tillmann and an anonymous referee for helpful comments.

References (8)

There are more references available in the full text version of this article.

Cited by (0)

View full text