Are Fiscal Multipliers Estimated with Proxy‐SVARs Robust?*

Published date01 February 2023
AuthorGiovanni Angelini,Giovanni Caggiano,Efrem Castelnuovo,Luca Fanelli
Date01 February 2023
DOIhttp://doi.org/10.1111/obes.12521
OXFORD BULLETIN OF ECONOMICS AND STATISTICS, 85, 1 (2023) 0305-9049
doi: 10.1111/obes.12521
Are Fiscal Multipliers Estimated with
Proxy-SVARs Robust?*
GIOVANNI ANGELINI,† GIOVANNI CAGGIANO,‡,§
EFREM CASTELNUOVO§ and LUCA FANELLI†
†University of Bologna, Bologna, Italy (e-mail: g.angelini@unibo.it)
‡Monash University, Melbourne, Victoria, Australia (e-mail: giovanni.caggiano@monash.edu)
§University of Padova, Padova, Italy (e-mail: efrem.castelnuovo@unipd.it)
Abstract
We estimate government spending and tax multipliers in the United States via a f‌lexible
proxy-SVAR model, where identif‌ication is achieved by combining f‌iscal and non-f‌iscal
instruments with additional parametric restrictions. We f‌ind that, while the spending
multiplier is robustly estimated to be larger than one across different models, the estimate
of the tax multiplier is sensitive to the combination of instruments that we use. We unveil
that the key factor behind these heterogeneous estimates is the assumption of orthogonality
between total factor productivity (our main non-f‌iscal proxy) and the tax shock. If this
assumption is imposed, the tax multiplier is around one. If it is not imposed, the tax
multiplier is three times as large. Our results point to the need of accounting for the large
uncertainty surrounding the tax multiplier for the design of optimal f‌iscal policies.
I. Introduction
The COVID-19 pandemic has triggered the most severe recession of the post-war era
for all major economies (International Monetary Fund (2020)). With policy rates at
their effective lower bound, f‌iscal policy stimulus packages of unprecedented scale have
been implemented in most countries with the aim of sustaining aggregate demand. How
effective these interventions are depends on the size of the f‌iscal multipliers, whose
JEL Classif‌ication numbers: C52, E62.
*We thank Anindya Banerjee (Editor) and two anonymous referees for their helpful comments and suggestions.
We also thank Ralf Br¨
uggeman, Dario Caldara, Fabio Canova, Pedro Gomis-Porqueras, Giovanni Pellegrino,
Bruce Preston, Søren Hove Ravn, Antti Ripatti, Xuan Zhou and participants at seminars held at the Universities
of Bologna, Helsinki, Adelaide, Deakin University. We also thank participants of the 30th EC 2Conference on
”Identif‌ication in Macroeconomics” (Oxford University, December 2019) and the IAAE 2002 Annual Conference
(King’s Business School, June 2022) for useful feedback. Special thanks are due to Valerie Ramey and Sarah
Zubairy for an email exchange on a f‌iscal proxy available on Ramey’s website and to Giorgio Primiceri for sharing
the marginal eff‌iciency of investment shock series estimated in Justiniano, Primiceri, and Tambalotti (2011).
Giovanni Angelini thanks the University of Melbourne (where part of this research was developed) for its kind
hospitality. Financial support by the Australian Research Council via the Discovery Grants DP160102281 and
DP190102802 and the Italian MUR via the PRIN 2017, Grant 2017TA7TYC are gratefully acknowledged. This
work is dedicated to Edoardo, Giovanni Angelini’s son, who arrived the very same day this paper got accepted
for publication.
95
©2022 Oxford University and John Wiley & Sons Ltd.
96 Bulletin
quantif‌ication is a diff‌icult task. The reason is that spending and tax revenues are in large
part endogenous, and tackling this endogeneity issue to identify the output effects of
exogenous variations in f‌iscal variables – i.e. f‌iscal shocks – is challenging.
One way to identify causal effects that have recently gained a lot of traction is the
‘proxy-SVAR’ (or ‘SVAR-IV’) approach, which relies on the use of instruments for the
identif‌ication of the shocks of interest (see Stock and Watson (2012) and Mertens and
Ravn (2013) for early contributions, and Stock and Watson (2018)forareview).
1There
is, however, lack of consensus on the size of the estimated multipliers. Using a narrative
measure of unanticipated exogenous variations in tax revenues, Romer and Romer (2010)
and Mertens and Ravn (2011b,2012,2013,2014) f‌ind tax multipliers to peak between 2
and 3. Differently, Caldara and Kamps (2017), who employ Fernald’s (2014) measure of
total factor productivity (TFP) to identify exogenous changes of output from f‌iscal policy
rules from which f‌iscal policy shocks are recovered, estimate the tax multiplier to peak
in the range between 0.5 and 0.7. Turning to the spending multiplier, Caldara and Kamps
(2017) estimate it to peak between 1 and 1.3, larger than the estimates documented
in the papers surveyed by Ramey (2019), which point to a 0.6– 1.0 range. In light of
policymakers’ need to get reliable and robust indications on the size and relative strength
of the spending and tax multipliers, the heterogeneity of the estimates provided by the
extant literature is problematic.
Contributions of this paper
This paper employs f‌iscal and non-f‌iscal policy instruments jointly in a proxy-SVAR
model and provides a framework to reconcile the previous heterogenous estimates. We
make two main contributions to the literature.
First, we show that the size of the tax multiplier is extremely sensitive to the
orthogonality (exogeneity) condition between the non-f‌iscal proxy used to identify
output shocks (TFP, as in Caldara and Kamps (2017)) and the tax shock. When this
assumption is imposed, the peak tax multiplier is estimated to be below 1, as in Caldara
and Kamps (2017). When this assumption is relaxed as suggested by the procyclicality
of tax revenues or by theories of endogenous labour productivity, the tax multiplier is
three times as large and is found to be in line with Mertens and Ravn (2014). We then
show that the hypothesis that the TFP proxy is uncorrelated with the tax shock is not
strongly supported by US data. Considering for example the same time series, time span
and predictive regression approach as in Caldara and Kamps (2017), the hypothesis of
no correlation between the TFP proxy and the tax shock can be rejected if inference is
robustif‌ied using heteroskedasticity and autocorrelation consistent (HAC) standard errors.
We further show that empirical evidence against the orthogonality between the TFP
proxy and tax shock holds across a number of proxy-SVAR specif‌ications, starting from
a standard three-variate f‌iscal VAR and then progressively enlarging the information set.
Overall, we interpret our empirical f‌indings as evidence that the orthogonality condition
between the TFP proxy and the tax shock commonly imposed in f‌iscal proxy-SVARs
is at least questionable. To reiterate, our paper shows that imposing or not this condition
1We will use the terms ‘instruments’, ‘proxies’ and ‘external variables’ interchangeably throughout the paper.
©2022 Oxford University and John Wiley & Sons Ltd.

Get this document and AI-powered insights with a free trial of vLex and Vincent AI

Get Started for Free

Start Your Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant

  • Access comprehensive legal content with no limitations across vLex's unparalleled global legal database

  • Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength

  • Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities

  • Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting

vLex

Start Your Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant

  • Access comprehensive legal content with no limitations across vLex's unparalleled global legal database

  • Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength

  • Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities

  • Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting

vLex

Start Your Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant

  • Access comprehensive legal content with no limitations across vLex's unparalleled global legal database

  • Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength

  • Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities

  • Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting

vLex

Start Your Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant

  • Access comprehensive legal content with no limitations across vLex's unparalleled global legal database

  • Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength

  • Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities

  • Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting

vLex

Start Your Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant

  • Access comprehensive legal content with no limitations across vLex's unparalleled global legal database

  • Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength

  • Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities

  • Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting

vLex

Start Your Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant

  • Access comprehensive legal content with no limitations across vLex's unparalleled global legal database

  • Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength

  • Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities

  • Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting

vLex