MIT LIBRARIES
DEWEfi
3 9080 03316 3343
ho
.
I
O' 01
Massachusetts Institute of Technology
Department of Economics
Working Paper Series
Political Limits to Globalization
Daron Acemoglu
Pierre Yared
Working Paper 10-1
January 15, 2010
RoomE52-251
50 Memorial Drive
Cambridge, MA 02142
This paper can be downloaded without charge from the
Social Science Research
Network Paper Collection
http://ssm.com/abstract "=l 540243
at
Political Limits to Globalization*
Daron Acemoglu
MIT
Pierre Yared
and CIFAR
January
Columbia
15,
2010
Abstract
Despite the major advances in information technology that have shaped the recent
wave
of globahzation, openness to trade
can change with shifts
particular threat
may be
in
militarist sentiments that
size of the military,
a political choice, and trade policy
domestic political equilibria.
and a limiting factor
around the globe today.
is still
We
to globalization
This paper suggests that a
and
appear to be on the
its
rise
future developments
among many
nations
proxy militarism by spending on the military and the
and document that over the past 20
years, countries experiencing
greater increases in militarism according to these measures have had lower growth
in trade.
Focusing on bilateral trade flows, we also show that controlling flexibly
country trends, a pair of countries jointly experiencing greater increases
for
in militarism
has lower growth in bilateral trade.
Keywords:
JEL
globalization, trade, military spending, militarism, nationalism
Classification: FOl, FIO, F52
for Advanced ReColumbia University, pyared3columbia.edu. We would like to thank
Philippe Aghion. Alex Debs, Philippe Martin. Thierry Mayer, Kevin O'Rourke and Mathias Thoenig
for suggestions and Dmitriy Sergeyev for excellent research assistance.
'Acemoglu:
Massachusetts Institute of Technology and Canadian Institute
search, daronQmit.edu; Yared:
Digitized by the Internet Archive
in
2011 with funding from
Boston Library Consortium IVIember Libraries
http://www.archive.org/details/politicallimitstOOacem
Introduction
1
We
live in
an unprecedented age of globalization, where technology,
of production
and goods are increasingly mobile across national boundaries.
current wave of globalization
major
is
Nevertheless, globalization
to international trade, finance
make, and despite the
The
distinguished from previous ones in part because of the
role of information technology.
Openness
ideas, factors
and technology
is
a choice that countries
facilitating role of information technology,
many
leading players in the world
Brazil
and Russia, could decide
economy including the United
to close their borders.
A
not irreversible.
is
many countries,
even
States, China, India,
major cause of the end of
the previous (also historically unprecedented) 19th century wave of globalization was
disillusionment with the international economic order, in large part precipitated by
the Great Depression
(e.g.,
though not necessarily
and international
Glick and Alan
Harold James, 2001). Another, somewhat
less
important cause Avas the
conflict (e.g.,
M.
rise of
less
emphasized
nationalism, militarism
Ronald Findlay and Kevin H. O'Rourke, 2007, Reuven
The previous wave
Taylor, 2006).^
of globalization took place in the
context of the 100 years following the end of Napoleonic wars, which were vmusually
peaceful for European powers;
conflict that
human
In this paper,
society
it
came
had experienced
we emphasize
until then.
World War
I.
that globalization, which depends on political deci-
sions of nation states, has political limits,
tionalism and militarism.
an end following the most widespread
to
and that these
limits are related to na-
Despite the increasing reach of globalization, anecdotal
evidence suggests that nationalism and militarism are strong around the world, in
countries ranging from the United States to China, Russia and India
Kagan, 2008). To go beyond anecdotal evidence,
^Militarism
is
in this
Robert
paper we proxy nationalist
defined as the doctrine or policy of "'aggressive military preparedness," which
typically leads to a country maintaining a strong military capability to defend or
interests.
(e.g.,
promote
its
national
and
by military spending.'^ In addition to being a useful proxy,
militarist sentiments
military spending might itself impact trade, for example, because
tensions or leads to skirmishes between countries.
Figure
1
shows the evolution of
world trade and total military spending between 1988 and 2007.
trade over the past two decades, during which
rise of
ditures
and the
size of military
known, the increase
trends, e.g., Findlay
declining for a
in the
It
depicts the steady
we have data on
military expen-
personnel across a large number of countries (as
volume of trade during
and O'Rourke, 2007).
number
contributes to
it
also
It
this
time period
reflects
is
well
longer-term
shows that military spending,
after
of years, started increasing from the mid-1990s onwards. This
pattern indeed indicates that there might be more than anecdotal evidence pointing
to a strengthening in nationalist sentiments
Our main contribution
in this
paper
is
and miUtarism.
to
show that military spending,
interpretation as a proxy for nationalist sentiment and militarism,
sociated with trade.
We
present two types of evidence. First,
is
we show
in our
negatively asthat between
1985 and 2005, countries that experience a greater increase in military expenditure
or the size of the military
show a
relative decline in the
volume of trade (compared
to
other countries in the sample). Moreover, countries whose trading partners ("neighbors"
show greater
)
show a
increases in military expenditure or the size of the military also
similar decline.
in different
These patterns are robust across
different specifications
and
subsamples. Second, we investigate bilateral trade patterns again between
1985 and 2005. The data suggest that trade between two countries grows
less
when both become more
of findings,
this pattern
In
is
militarized.
While not as robust as the
first set
rapidly
generally present in a variety of different specifications.
summary, the data point
to a negative correlation
between militarization and
^This is true almost by definition of militarism. Tom W. Smith and Lawrence Jarkko (2001)
provide several measures of "national pride" constructed from survey evidence, some of which are
strongly correlated with military spending across countries.
trade.
Although we cannot ascertain a causal relationship, the evidence
is
broadly
consistent with an association between the strength of nationalist sentiments and
militarism, as proxied by military expenditures or the size of the military,
and
Overall, this evidence suggests that there might be political
ternational trade.
in-
and
military limits to, and dangers against, globalization.
Our paper
is
related to three separate literatures. First, several studies in economic
history have investigated the causes of the end of the 19th century globalization
James, 2001, Barry Eichengreen and Douglas A. Irwin, 2009).
large
and growing
based on ideas
war
literature in international relations
first
less likely (see,
articulated by Montesquieu
Second, there
on the so-called
(e.g.,
is
a
"liberal theory,"
and Kant, that greater trade makes
example, John R. Oneal and Bruce Russett, 1999, Simon
for
Polachek, 1980). Several papers in this literature have simultaneously estimated the
effect of trade
on war and of war disruptions on trade
Mayer and Mathias Thoenig, 2008,
find a negative effect of
militarization
this
we
focus on
by showing that the
in military conflict are
to
war on
is
or
who document
is
from
effect survives
worth noting, however, that the
this disruption effect,
when
excluded from the
Anna Maria Mayda and Dani Rodrik
(2001),
Philippe Martin, Thierry
Havard Herge, Oneal and Russett, 2009), and
trade. It
distinct
(e.g.,
all
data."^
effect of
and we demonstrate
countries or country pairs engaged
Finally, our
paper
is
also related
(2002) and O'Rourke and Richard Sinnott
a negative relationship between attitudes towards trade and
nationalist sentiment in survey responses.
may nonetheless capture the fact that greater military spending by a country and
neighbors increases the likelihood of military conflict in the future, the anticipation of which
discourages trade. We view this as part of the impact of militarism on trade in which we are
''Our results
its
interested.
Data
2
Our measure
of trade (opeuness)
is
a country's trade share of
We
from Summers-Heston dataset, 2009.
GDP
in constant prices
GDP
use this same dataset to measure real
per capita and total population.'*
Bilateral trade data are from the International
2009 (DoT)
Statistics,
s,
CD-ROM.
Let A'^^ denote bilateral between
meaning the sum of exports from
calculate Xijs for
j to
i
country pairs
to j
i
and exports from
in year s for
j to
both are
i
to j or
available,
available.
Using
GIF
(cost, insurance,
we take the
this
j as a fraction of
and
freight)
i
and
FOB
(free
i's
total trade
and we multiply
j as a fraction of
z's
imports by j from
this
measure by
FIRST
GDP)
Database.^
is
We
i's
is
i
trade share
bilateral trade
multiply this fraction by total
GDP
per capita and population from the Summers-Heston dataset to create a
size
measures total military personnel and
a mecisure of the militarization of "neighbors".
inverse distance in kilometers between
i's
capital
and
GDP
data are missing
for
Russia
for
size data,
Specifically, let
and
j's capital.^
a weighted average of the log military spending of country
^
When
from the Stockholm International
from FIRST.'' Using the mihtary spending and military
''Trade
i.
GDP.
measure of military spending. Military
also
on board)
measure, we construct a measure of bilateral trade between
Peace Research Institute
GDP
We
and from
to j
i
s.
and otherwise we use whichever measure
average,
Military spending (as a fraction of
using
j in year
in year
i
from the Summers-Heston dataset so as to achieve a measure of the
between
and
i
which both flows from
are available. These flows can be measured using either
exports from
and
all
Monetary Fund Direction of Trade
i's
is
we construct
tOij
represent the
We
then calculate
neighbors using the
1988 so we use the 1989 value.
Facts on International Relations and Security Trends Database, http://first.sipri.org/
'^Mihtary spending in 1988
from 1989
is
missing for China and for Ru.ssia, so we use the mihtary spending
both observations.
'^Distance between capital cities
for
is
from Kristian
S.
Gleditsch and Michael D.
Ward
(2001).
ujij^s
as the (relative) weight of country j (7^
method
in this calculation.
i)
We
use the
same
to calculate the weighted average of the log military size of neighbors. This
measure captures both the militarization of neighboring countries which would pose
the greatest military threat to country
with which country
i
i
and
also the militarization of those countries
should trade most heavily according to the standard gravity
models.
For the regressions using mihtary spending, the beginning date of the sample
1988 and the end date
is
date of the sample
1985 and the end date
is
2005. For the regressions using military size, the beginning
maximize the number of countries
Finally, as a robustness check
in our
is
2003.
These dates are chosen to
sample given data
we exclude
availability.
countries (or country-pairs) in which a
country (at least one of the countries in the pair) experiences a
war between 1985 and 2005. For
Institute
this exercise,
we
civil or
international
use the International Peace Research
and Uppsala Conflict Data Program Armed Conflict Dataset,^ and we code
a country as experiencing a
war
is
intensity,
primary party
civil
war
if it
and we code a country
in
an interstate armed
experiences an intrastate armed conflict of
as experiencing
an international war
if it is
a
conflict of wai^ intensity.
Cross-Country Evidence
3
We
start
tarization
with cross-country evidence and investigate the relationship between mili-
and trade using the following long-difference regression model:
Ay,
(1)
^
Armed
Conflict Dataset.
= aAm, +
Ax'/3-f-it,.
http://www.prio.no/CSCW/Datasets/Armed-Conflict/UCDP-PRIO/
Here Ay,
is
the change in the trade share of
and the end dates of the sample;
spending or log mihtary
Am;
is
size) of either
GDP
of country
OECD
differential trends in
some
between the beginning
the change mihtarization
country
or of country
i
represents the change in a vector of covariates (log
and
i
GDP
specifications);
i's
is
log military
neighbors;
Ax^
per capita, log population,
and
u,
an error term.
is
medium-run trends
focus on long-difference specifications to capture
annual fluctuations. Naturally, this specification
(i.e.,
We
as opposed to
algebraically equivalent to a panel
data regression with two observations per country and a
of country and year
full set
fixed effects.^
Table
reports estimates of equation
1
sured as (log) military spending, while in Panel
Throughout we report standard
In Panel A, militarization
(1).
B
militarization
is
is
mea-
(log) military size.
errors that are robust against arbitrary heteroskedas-
ticity.
Column
6.3).
1 in
Panel
decades, holding
2,
reports an estimate of
This coefficient, which
an economically large
GDP
A
effect:
all else
a
of -17.6 with standard error (s.e.=
statistically significant at less
is
than 1%, also implies
a 10 percent increase in military spending over two
constant,
is
associated with a reduction in trade share to
of approximately 1.8 percent. ^'^ This result
which displays a residual plot of trade share
is
illustrated graphically in Figure
vs.
military spending and shows a
strong negative relationship.
Columns
countries,
'The
2-4 explore the robustness of this pattern. In
many
results are similar
full set
equivalent of a
2,
if
the level equation corresponding to (1)
is
estimated on annual data
of country
common
"time effect" in a panel regression).
The
log military spending of .16 over
is
captured by the constant (the
regression
whether a country's variation around the global trend towards greater trade
militarization.
we exclude Asian
of which have experienced an increase in militarization simultane-
and year fixed effects.
^"The median country in the sample experienced an increase in
the sample period. The increase in trade during the same period
with a
column
is
is
therefore exploiting
due to variation
in
ously with rapid increases in international trade, which
related to the increasing abihty of
Asia.
The estimate
of
«
may have had
other causes,
Western companies to outsource and offshore to
increases to -22.6 (s.e.= 7.1), confirming that the relationship
between militarization and trade
is
stronger without Asian countries.
Column
3 in-
vestigates whether our results capture the impact of trade disruption caused by wars;
it
excludes countries engaged in
civil
wars or in international wars. Interestingly, our
results are stronger once these countries are excluded (coefficient -24.5, s.e.= 9.3).
This bolsters our belief that our results are related to the relationship between trade
and militarism.
column
Finally,
4 includes a
differential trend in trade for this
dummy
in trade, this
which
colunm
almost identical to that
Columns
OECD
countries, allowing
group of countries. Even though
appear to experience faster growth
is
for
in
1.
5-8 are analogous to cohunns 1-4, but use the log military spending
coefficient estimate of
column
neighbors
1,
is
is
-33.0 (s.e.= 14.3).
militarization of
column
5,
the
almost twice that
Columns
6-8
show
robust to the same set of specification checks as in columns 2-4.
is
9 includes both log militarization of a country
neighbors.
its
column
1
and
The
to those in Panel A, with a
larger effects
effect of
own
is
is still
and the weighted
military expenditure
statistically significant at less
In Panel B, militarization
and
is
In
increases in militarization by geographically proximate
tliat
neighbors' military expenditure
rization
This estimate
associated with even a larger relative decline in trade.
column
to that in
a
suggesting
that this pattern
Finally,
countries
has no effect on our estimate of a,
of neighbors, as defined in the previous section, in place of m,.
in
OECD
is
log
very similar
than 1%, while the
effect of
large but no longer statistically significant.
measured
as log military size.
somewhat smaller quantitative
The
effects
results are similar
from own
milita-
from militarization of neighbors.
Overall, the cross-country evidence shows a relatively robust association between
changes in military expenditure or size of military personnel and changes
tional trade
between 1985 and 2005.
in interna-
,
Evidence from Bilateral Trade
4
We
next investigate more disaggregated bilateral trade flows data. This allows us to
control not only for global trends but also for differences between country pairs, so
that
we can
directly look at
trading partner that
whether a country
becoming more
is
will
reduce
More
militarized.
its
trade with a (potential)
specifically,
we estimate
the
following gravity-type equation:
Ay,j
(2)
= aAm,mj +
where Ay,j now represents the change
country
i's
GDP; Am^mj
is
partner country fixed
i
in the trade
+
7]^
+
u,j,
with country j as a fraction of
the change in the interaction of
and the vector Ax,j includes change
population of countries
Ax',jp+7]^
and
effects;
j; ?/f
and
in the interaction of log
and
i(,j
is
7/^ represent
a
and
i
GDP
full set
the error term which
of
is
j's militarization;
per capita and log
home and
trading
allowed to have an
arbitrary pattern of heteroskedasticity with clustering at the country-pair level.
Analogously to equation
moves
all
eral trade
all
focuses on long differences and re-
country-pair specific characteristics that might simultaneously affect bilat-
and militarization of
either country. In addition, the specification removes
country specific trends (which are the focus of Table
of
home country and
is
a which measures
its
(1), this specification
trading partner.
Table
trading partner country dummies.
1)
by including a
The
full set
coefficient of interest
the effect of a country's militarization interacted with that of
The
results
from the estimation of equation
2.
8
(2) are
reported in
Columns
1
1-4 use log military
shows the interaction
and
effect of
spending as our measure of militarization. Column
home and
statistically significant at less
there
is
than
trading partner's militarization
1%
=
(a
and
-0.090
a strong negative association between a joint
rise in
is
s.e.=: 0.029),
negative
so that
the militarization of the
country pair and bilateral trade. ^''-
Column
interaction effect
is
The
Asian countries.
2 excludes
results are similar
both larger and more precisely estimated.
country pairs where at least one country experienced a
the results are also stronger than those in column 1}^
two
dummy
variables,
one country being
OECD). The
Columns
one
in the
for
both countries being
OECD
(the omitted group
results are again similar to those in
civil or
Finally,
in the
is
and stronger; the
Column
international war;
column 4 includes
OECD, and
The estimates
in these
one
for
only
neither country being in the
column
1.^'^
5-8 are analogous to columns 1-4, except that militarization
as log military size.
3 excludes
is
measured
columns are also negative, though generally
imprecisely estimated and insignificant.
5
Concluding Remarks
This short paper emphasized that, despite the major advances in information technology encouraging globalization, openness to trade
is still
a political choice.
This
suggests that changes in domestic political equilibria might introduce limits to the
process of globalization.
militarization,
"The
We
illustrated this general point
which has seen a recent
quantitative magnitudes of
revival,
tliis effect is large.
on country-level and
A
GDP
^^None of these
ratio in 2005
results
change
if
is
effect of
bilateral trade.
(tlie median
by about 0.529c. The median
5.75 point value of AniiVij
increase in the sample period) reduces bilateral trade relative to
bilateral trade to
by focusing on the
GDP
about 0.06%.
we instead only exclude countries which are
directly fighting
against each other.
^'In addition, the estimates
one country in the
OECD
show that country
pairs in the
have experienced slower growth
OECD
and country pairs with only
than the baseline.
in bilateral trade
The evidence we presented
suggests that countries experiencing greater mihtarization
and those witnessing greater mihtarization among
have seen relatively smaller increases
mented that country
their neighbors (trading partners)
in trade over the past 20 years.
We
also docu-
pairs experiencing greater joint increases in militarization have
seen relative declines in bilateral trade.
Our
results
come with
several caveats.
First,
it is
unclear to what extent these
empirical patterns reflect causality since trade and militarization simultaneously affect
each other and
an explanation
may
themselves be affected by a third factor. Second,
for the
apparent
rise in
we do
not have
nationalism and militarization around the
world.
10
References
Eichengreen, Barry and Douglas A. Irwin (2009) "The Slide to Protectionism in
the Great Depression:
Who
Succumbed and Why?"
Findlay, Ronald and Kevin H.
and the World Economy
m
the
NBER Working
Paper.
O'Rourke (2007) Power and Plenty: Trade, War,
Second Millennium, Princeton University Press,
Princeton.
Glick,
Reuven and Alan M. Taylor (2006)
and the Economic Impact
Gleditsch,
Kristian
S.
of
War,"
NBER
"Collateral
Damage: Trade Disruption
Working Paper.
and Michael D. Ward (2001)
"Measuring Space:
A
Minimum-Distance Database and Applications to International Studies," Journal of
Peace Research, 38, 739-758.
Herge, Havard, John R. Oneal, and Bruce Russett (2009) "Trade Does Promote
Peace:
The
Perils of
Simultaneous Estimation of the Reciprocal Effects of Trade on
Conflict," forthcoming in Journal of Peace Research.
James, Harold (2001) The End of Globalization: Lessons from the Great Depression,
Harvard University Press, Cambridge.
Kagan, Robert (2008) The Return of History and
House,
New
the
End
of Dreams,
Random
York.
Oneal, John R. and Bruce Russett (1999) "Assessing the Liberal Peace with Alternative Specifications:
Trade
Still
Reduces Conflict," Journal of Peace Research,
36, 423-442.
Martin, Philippe, Thierry Mayer, and Mathias Thoenig (2008) "Make Trade Not
War?", Review of Economic Studies,
75, 865-900.
Mayda, Anna Maria and Dani Rodrik (2002) "Why Are Some People (and Countries)
More Protectionism than Others?" Working Paper.
O'Rourke, Kevin H. and Richard Sinnott (2001) "The Determinants of Individual
11
Trade Policy Preferences: International Survey Evidence," Brookings Trade Forum,
157-196.
Polachek, Simon (1980) "Conflict and Trade," Journal of Conflict Resolution 24,
57-78
Smith,
Tom W.
and Lars Jarkko (2001) "National Pride
in Cross-National Per"
spective,"
Working Paper.
..
12
Sample
IS
a balanced pane! 19fi8-2007 for
for each year Militar>'
spending
is
the
sum
which trade share of
across countries and
assigned to 1988, 1991 military spending for Russia
is
Figure corresponds to the residual plot of regression
GDP and mil nan,' spending is available for all years Trade share of GDP is the average across comitnes
is m 1*^96 dollars For reasons of data availability, military spending for Russia and China m 1989 is
interpolated, and 1990 openness for Russia
in
column
1
of panel
A of Table
1
See
is
assigned
to
1988 and 1989 See text for data definitions and sources
lexl for data definitions
and sources
5^
S
OS ^
O "
-J OS
—
—
On
OO
Q O
O
Q
OO
::y
O
'^
—
so
o
r-J
^
rsl
rn
rsi
i/^
a^
o
U
OS
W < —
I
OS
-o
OS o^
°^
o
22
on"
c
._aj
1
C/5
^ J
Q O
(v^
IN
y
o
t^
'O
-Si
5
dD
o
C2
m
OO
c^
OO
o o
o o
^ O
D O
—
X-I
^
— o
5
S
^
3
w^
(^1
^ ~ - ^
Q .1*
^
OO
v£;
^
5^
^
o; r^
^ o
Llj
13"
=
S 2
5-
'~>
QD
\0 CO
Q U <
^ o
^^
J O
^
?
s
D C -
g;
-a
e:
Q
u
o
Q
O
UJ
-^
u
c-
3-
G
O
Q
O
o
O
Di
O
"c^
a^
T3
O o
9 o
O
— o
o
^
—
r-i
m
oo
r-
O
o o
-rr
OO OO
U^- OO
OO
.—
^
11
00
op
5
<S fc
5
o
I
CJ
u-
-o
.
^
-^
"
-a tu -g
,5 i^
C
a
5
^
—
CO
r'l
o
^
U < ~
W
;3
s
•a
<
9 O
<^ fc C«
_!
't:
-a
Q
fS
—
O o
9 o
O
S
=0 ON
Q o c cu
c
o
J
1
p o
^ 9 p.
<s
^
£
-Si
-Q
GC
o
Q
•5
2
(3
&
u
t/i
-a tU
c
—
"S:s
E
S-
00
c
i2
c
!=
2^1
^
-g
«
C
T= UJ
<
o
O o
9 o
*"
e:^
H
c
^
C
c
DlJ
"5
rt
c3
X
c
t:
-^ 1^
£
cp
f-
X
E
N
UJ
O
C^
2
Mj
O
o
1
CO
ij:^
Q
U
w
O
c/:
;j=
p3
2
~
DO
O
a
-^
3 DXJ
c o
= -J
u
Q
u
w
o
iT
a