Currency Unions and Trade: A PPML Re‐assessment with High‐dimensional Fixed Effects
| DOI | http://doi.org/10.1111/obes.12283 |
| Author | Mario Larch,Yoto V. Yotov,Thomas Zylkin,Joschka Wanner |
| Date | 01 June 2019 |
| Published date | 01 June 2019 |
487
©2018 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd.
OXFORD BULLETIN OF ECONOMICSAND STATISTICS, 81, 3 (2019) 0305–9049
doi: 10.1111/obes.12283
Currency Unions and Trade:A PPML Re-assessment
with High-dimensional Fixed Effects*
Mario Larch,†Joschka Wanner,‡Yoto V. Yotov§ and Thomas
Zylkin¶
†University of Bayreuth, CEPII, CESifo, and Ifo Institute, Universit¨atsstrasse 30 95447,
Bayreuth, Germany (e-mail: mario.larch@uni-bayreuth.de)
‡University of Bayreuth, Universit¨atsstrasse 30 95447, Bayreuth, Germany (e-mail:
joschka.wanner@uni-bayreuth.de)
§Drexel University, Ifo Institute, CESifo, ERI-BAS, 1020 G-Hall, 3200 Market Street
Philadelphia, PA 19104, USA (e-mail: yotov@drexel.edu)
¶Robins School of Business, University of Richmond, 1 Gateway Road Richmond, VA
23217, USA (e-mail: tzylkin@richmond.edu)
Abstract
Recent work on the effects of currency unions (CUs) on trade stresses the importance of
using many countries and years in order to obtain reliable estimates. However, for large
samples, computational issues associated with the three-way (exporter-time, importer-
time, and country pair) fixed effects currently recommended in the gravity literature have
heretofore limited the choice of estimator, leaving an important methodological gap. To
address this gap, we introduce an iterative poisson pseudo-maximum likelihood (PPML)
estimation procedure that facilitates the inclusion of these fixed effects for large data sets
and also allows for correlated errors across countries and time. When applied to a compre-
hensive sample with more than 200 countries trading over 65 years, these innovations flip
the conclusions of an otherwise rigorously specified linear model. Most importantly, our
estimates for both the overall CU effect and the Euro effect specifically are economically
small and statistically insignificant. We also document that linear and PPML estimates of
the Euro effect increasingly diverge as the sample size grows.
I. Introduction and motivation
Tous, a plausible methodology to estimate the currency union effect on trade involves panel
estimation with dyadic fixed effects. We [
…]await computational advances to be able to
estimate the Poisson analogues. (Glick and Rose, 2016, p. 86)
JEL Classification numbers: C13; C23; C55; F14; F15; F33.
*Thomas Zylkin is grateful for research support from the NUS Strategic Research Grant (WBS: R-109-000-183-
646) awarded to the Global Production Networks Centre (GPN@NUS) for the project titled ‘Global Production
Networks, Global Value Chains, and East Asian Development’. We thank Andy Rose, Jo˜ao Santos Silva, three
referees, the handling editor James Fenske, as well as participants at the European Trade Study Group 2017 in
Florence and the workshop “International Economics” 2018 in G¨ottingen. Note: The procedure presented in this
paper is implemented in Stata and freely available via ssc (to install, type ‘ssc install ppml panel sg, replace’) or at
https://econpapers.repec.org/software/bocbocode/S458249.htm.
488 Bulletin
Writing at the beginning of a transformative period in the empirical study of interna-
tional trade, Rose (2000) reported the stunning finding that sharing a common currency
more than triples trade between countries. While this estimate was regarded as puzzlingly
high at the time, it succeeded in stimulating a vibrant and ongoing empirical literature
investigating the trade-creating effects of currency unions (CUs), having garnered over
3,100 citations since its original publication in Google Scholar and 356 citations in Web
of Science Core Collection. This literature has notably included frequent re-examinations
of the original evidence by Rose himself – such as Glick and Rose (2002, 2016) – as well
as fervent interest in whether the European Monetary Union (EMU) in particular, as the
largest CU to date, might have had similarly remarkable effects.1
Parallel to this literature, the past two decades have seen the development and wide
adoption of many new econometric best practices for consistently identifying the determi-
nants of international trade. These have most notably included the use of poisson pseudo-
maximum likelihood (PPML) estimation to address issues related to heteroscedasticity
and zeroes (Santos Silva and Tenreyro, 2006), time-varying exporter and importer fixed
effects to account for changes in the ‘multilateral resistance’ constraints implied by the-
ory (Anderson and van Wincoop, 2003; Feenstra, 2004; Baldwin and Taglioni, 2007), and
time-invariant pair fixed effects to absorb unobser vable barriers to trade (such as bilateral
history) and to address the endogeneity of trade policy variables due to time-invariant
unobserved bilateral heterogeneity (Baier and Bergstrand, 2007).2Aiding these devel-
opments, empirical researchers working in trade have also benefited from a new-found
consensus on the theoretical underpinnings of the gravity equation (Arkolakis, Costinot
and Rodr´
iguez-Clare, 2012) as well as recent computational advances that permit swift
estimation of linear models with a large number of fixed effects (Carneiro, Guimar˜aes and
Portugal, 2012; Correia, 2016).
Reassuringly, as these new methods havefiltered into the literature on cur rency unions,
they have led to more reasonable and reliable estimates. In their latest instalment that em-
phasizes the use of time-varying exporter and importer fixed effectsas well as time-invariant
pair fixed effects, Glick and Rose (2016) find – under their most rigorous specification –
that CUs generally increase trade by 40%, that CU entry and exit have symmetric effects
on trade, and that the EMU – which could not be included in earlier studies – has promoted
trade more than other CUs.3
Doing their due diligence, Glick and Rose (2016) also experiment with PPML estima-
tion with two-way (exporter-time and importer-time) fixed effects.4However, as captured
1Along with Glick and Rose (2002, 2016), some of Rose’sother work in this area includes Rose (2001, 2002, 2017).
Contributions by Persson(2001), Nitsch (2002), Levy-Yeyati(2003), Barro and Tenreyro(2007), de Sousa (2012) and
Campbell(2013) are examples of reactions to Rose’sinitial finding. Finally, Micco, Stein and Ordonez (2003), Baldwin
and Taglioni(2007), Bun and Klaassen (2007), Berger and Nitsch (2008), Santos Silva and Tenreyro(2010a), Eicher
and Henn (2011), Olivero and Yotov (2012), Herwartz and Weber (2013) and Mika and Zymek (2018) specifically
investigatethe effect of the EMU. Santos Silva and Tenreyro (2010a) and Rose (2017) sur veyeach of these literatures.
2Of course, endogeneityof common cur rencies mayalso arise from time-varying bilateral effects. Our investigation
does not tackle these sources of selection into currency unions.
3Glick (2017) demonstrates that these results are robust to controlling for EU membership and further shows that
there is heterogeneity in the trade effects between newand old EMU members.
4Glick and Rose (2016) include these results in an earlier working paper availableonline (Glick and Rose, 2015).
They still estimate a generally positive ‘additional effect’for the EMU vs. other CUs, but find the overall CU effect
disappears over time, echoing an earlier finding by de Sousa (2012).
©2018 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd
To continue reading
Request your trial