Tests of Policy Interventions in DSGE Models
| Author | M. Hashem Pesaran,Ron P. Smith |
| DOI | http://doi.org/10.1111/obes.12224 |
| Date | 01 June 2018 |
| Published date | 01 June 2018 |
457
©2017 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd.
OXFORD BULLETIN OF ECONOMICSAND STATISTICS, 80, 3 (2018) 0305–9049
doi: 10.1111/obes.12224
Tests of Policy Interventions in DSGE Models*
M. Hashem Pesaran†,‡ and Ron P. Smith§
†Department of Economics & USC Dornsife INET, University of Southern California, Los
Angeles, CA, USA (e-mail: pesaran@dornsife.usc.edu)
‡Trinity College, Cambridge, UK
§Birkbeck, University of London, London, UK (e-mail: r.smith@bbk.ac.uk)
Abstract
This paper considers tests of the effectiveness of a policy intervention, defined as a change
in the parameters of a policy rule, in the context of a macroeconometric dynamic stochastic
general equilibrium (DSGE) model.We consider twotypes of inter vention, first the standard
case of a parameter change that does not alter the steady state, and second one that does
alter the steady state, e.g. the target rate of inflation. We consider two types of test, one a
multi-horizon test, where the postintervention policy horizon, H, is small and fixed, and a
mean policy effect test where His allowed to increase without bounds. The multi-horizon
test requires Gaussian errors, but the mean policy effect test does not. It is shown that
neither of these two tests are consistent, in the sense that the power of the tests does not
tend to unity as H→∞, unless the intervention alters the steady state. This follows directly
from the fact that DSGE variables are measured as deviations from the steady state, and the
effects of policy change on target variablesdecay exponentially fast. We investigate the size
and power of the proposed mean effect test by simulating a standard three equation New
Keynesian DSGE model. The simulation results are in line with our theoretical findings
and show that in all applications the tests have the correct size; but unless the intervention
alters the steady state, their power does not go to unity with H.
I. Introduction
This paper considers testing the effectivenessof a policy intervention given time-series data
on outcome variables, both before and after the policychange. The policy effect is measured
as the difference between the policy outcome, the postintervention realized values, and a
counterfactual. The counterfactual is constructed assuming no policy intervention, using
JEL Classification numbers: C18, C54, E65.
*Weare grateful to the editor, Anindya Banerjee, and an anonymous referee for constructive comments; to Karrar
Hussain and Alex Chudik for their help with the calibration and simulation exercisesrepor ted in the paper. We have
also benefited from discussions with Oscar Jorda, Adrian Pagan, Ivan Petrella and Glenn Rudebusch. An earlier
version of this paper was called ‘Testsof policy ineffectiveness in macroeconometrics’.
458 Bulletin
parameters estimated on the preintervention sample.1While there are many ways that one
could construct such a counterfactual, this paper considers the case where it is obtained from
a dynamic stochastic general equilibrium (DSGE) model whose variables are measured
as deviations from the steady state.2The realized policy outcomes will reflect both a
deterministic component, the effect of the intervention, and a stochastic component, the
postintervention disturbances or shocks.
In the DSGE literature, a typical policy intervention is a monetary policy shock, calcu-
lated as a one standard error displacement of the structural disturbance of a policy equation,
such as a Taylor rule. The impulse response function (IRF) is the time profile of the de-
terministic component of the effect of such a displacement, and as discussed in section
II, yields ex ante information about the way the model responds to such a displacement,
not an ex post evaluation of the effectiveness of an actual policy intervention. As we shall
see, IRFs ignore the cumulative uncertainty associated with the stochastic component, the
postintervention disturbances. While one can construct tests for such displacements, we
focus on interventions that change policy parameters. The first type of intervention, such
as changing parameters of the Taylor Rule, does not alter the steady state. The second type,
such as changing the target rate of inflation, does alter the steady state. We show that if the
intervention does not alter the steady state, the power of the tests will not go to unity as
the postintervention horizon, H, gets large. This is an inherent consequence of the fact that
DSGE models use variables measured as deviations from the steady state and the effects
of policy changes decay exponentially fast. Thus, unless the intervention alters the steady
state, we cannot be sure that it has had an effect.
Tests based on the differences between realizations and counterfactuals are standard
in the statistical literature and have been used to examine a range of macroeconometric
questions. Abadie and Gardeazabal (2003) examine the effect of terrorism on the Basque
country using a ‘synthetic control region’ as a counterfactual. Hsiao, Ching and Wan
(2012) examine the effect on growth in Hong Kong of political and economic integra-
tion with mainland China, using a panel data approach to construct a counterfactual, us-
ing predictions from similar economies. Synthetic control and panel data counterfactuals
are compared by Gardeazabal and Vega-Bayo (2016). Pesaran, Smith and Smith (2007)
examine what would have happened to the economies of the UK and the eurozone had
the UK joined the euro in 1999, using ‘euro’ restrictions on a GVAR model to construct
a counterfactual. Fagan, Lothian and McNelis (2013) examine whether the Gold Standard
was in fact destabilizing, constructing the counterfactual by replacing the pre-1914 US
money supply process with a Taylor rule in a DSGE model.
Rather than comparing actuals with counterfactuals, the mainstream macroeconomic
literature has tended to emphasize estimation issues in the context of structural vector
autoregressions (VARs) and DSGEs. For instance, in the case of the Volcker disinflation
which marked the transition from an era of macroeconomic turbulence and high inflation
to an era of ‘Great Moderation’ and low inflation, Primiceri (2006) provides an explana-
1The Lucas Critique does not apply since counterfactuals are estimated using the preintervention sample, and the
policy-induced parameter change gets reflected in the realized postintervention outcomes, which embody the effect
of the change in the policy parameters and any consequent changes to expectations.
2Policy ineffectivenesstests where the counterfactuals are obtained from reduced or final form specifications are
considered in Pesaran and Smith (2016).
©2017 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd
To continue reading
Request your trial