THE EFFECTS OF FISCAL POLICY SHOCKS IN SVAR MODELS: A GRAPHICAL MODELLING APPROACH
| Published date | 01 September 2011 |
| Author | Matteo Fragetta,Giovanni Melina |
| DOI | http://doi.org/10.1111/j.1467-9485.2011.00558.x |
| Date | 01 September 2011 |
THE EFFECTS OF FISCAL POLICY
SHOCKS IN SVAR MODELS:
A GRAPHICAL MODELLING APPROACH
Matteo Fragetta
n
and Giovanni Melina
nn,nnn
Abstract
We apply graphical modelling (GM) theory to identify fiscal policy shocks in SVAR
models of the US economy. Unlike other econometric approaches – which achieve
identification by relying on potentially contentious a priori assumptions – GM is a
data based tool. Our results are in line with Keynesian theoretical models, being also
quantitatively similar to those obtained in the recent SVAR literature a
`la Blanchard
and Perotti (2002), and contrast with neoclassical real business cycle predictions.
Stability checks confirm that our findings are not driven by sample selection.
I Intro ductio n
In macroeconomics there is still no widespread consensus on the impact and
transmission channels of fiscal policy on many variables. Both theoretical and
empirical studies generally find a positive response of output and hours worked to a
positive shock to government purchases and the disagreement is usually about the
magnitude and timing of the response. On the contrary, the sign of the responses of
variables such as consumption, wages and investment is still a matter of debate.
In the theoretical literature, on one hand neoclassical Real Business Cycle
(RBC) theory claims that a positive government spending shock triggers a
negative wealth effect that dampens consumption, fosters labour supply and
curbs real wages (e.g. Baxter and King (1993)). On the other hand, Keynesian
theories and recent dynamic stochastic general equilibrium (DSGE) models such
as Linnemann (2006), Ravn et al. (2006) and Galı
´et al. (2007) among others,
assert that an expansionary fiscal policy boosts consumption, hours worked and
real wages. In addition, while RBC theories predict that real output should rise
less than proportionally to the increase in government spending, due to the
crowding-out effect on consumption, Keynesian theories foresee that the
increase in consumption should amplify the expansionary effect on output.
The empirical literature of the late 1990s, such as Ramey and Shapiro (1998)
and Edelberg et al. (1999), mostly relying on vector-autoregressions (VAR)
n
Univesity of Salerno
nn
University of London
nnn
University of Surrey
Scottish Journal of Political Economy, Vol. 58, No. 4, September 2011
r2011 The Authors. Scottish Journal of Political Economy r2011 Scottish Economic Society. Published by Blackwell
Publishing Ltd, 9600 Garsington Road, Oxford, OX4 2DQ, UK and 350 Main St, Malden, MA, 02148, USA
537
employing a narrative approach to identify discretionary fiscal policy shocks,
supports RBC predictions. More recent empirical studies, starting from
Blanchard and Perotti (2002), adopt structural VARs (SVAR) for the purpose
of identification and obtain results more in line with Keynesian claims. In
addition, Perotti (2007) and Monacelli and Perotti (2008), propose a variant to
the narrative approach that yields similar results to the SVAR literature.
We consider VAR analysis as a tool to understand what happens in actual
economies and to evaluate competing theoretical economic models used for
policy evaluation. SVARs, however, generally require the imposition of a
number of restrictions, which is often a complex and contentious task, as these
may be based on possibly arguable assumptions. This is why we believe that
relying on a data-oriented tool, that allows to empirically derive those
restrictions, may help shed light on this issue.
Therefore, in this paper we conduct a SVAR analysis for the US economy
that combines Graphical Modelling (GM) theory. GM is a data-oriented tool, as
it allows us to obtain identifying restrictions directly from statistical properties
of the data. Reale and Wilson (2001) and Wilson and Reale (2008) show how
this theory can be used in a VAR, while Oxley et al. (2009) is a good example of
how the method can be applied to macroeconomic analysis. However, our paper
is the first piece of research that analyses the effects of fiscal policy by making
use of such an empirical methodology. The starting point of the technique is the
computation of partial correlations among the variables in the model and the
subsequent construction of a Conditional Independence Graph (CIG), a
graphical representation of all statistically significant interconnections among
all variables. From the CIG, based on well defined statistical rules, we derive
Directed Acyclic Graphs (DAGs), graphical representations of the many
possible structural VARs, which are later evaluated by means of statistical
information criteria. Granger and Swanson (1997) have applied a similar
strategy to sort out an order among contemporaneous variables based on the
partial correlations of the innovations derived from the reduced form and obtain
residual orthogonalization. GM offers a systematic procedure to obtain residual
orthogonalization and hence identify a SVAR model.
Our results are in line with the recent SVAR literature (Blanchard and Perotti
(2002), Perotti (2005), Caldara and Kamps (2008) among others) and hence give
credit to Keynesian claims. In response to a positive government spending
shock, we detect a partially deficit-financed fiscal policy and obtain a fiscal
multiplier of output greater than one. Adding more recent data increases the
magnitude of the fiscal multiplier compared with earlier studies such as
Blanchard and Perotti (2002). Private consumption shows a positive and
persistent response to a spending shock. While non-residential investment is
significantly crowded out by the fiscal expansion, residential investment rises,
comoving with output. However, the crowding-out effect on non-residential
investment is not stable over the sample considered. Lastly, a positive response
of real wages coexists with an increase in hours worked. As far as the effects of a
positive tax shock are concerned, we find that peak responses of consumption,
non-residential investment and the initial response of hours worked show signs
M. FRAGETTA AND G. MELINA538
Scottish Journal of Political Economy
r2011 The Authors. Scottish Journal of Political Economy r2011 Scottish Economic Society
To continue reading
Request your trial