Content uploaded by Kerim Can Kavakli
Author content
All content in this area was uploaded by Kerim Can Kavakli on May 18, 2021
Content may be subject to copyright.
The Power to Hurt and the Effectiveness
of International Sanctions
Kerim Can Kavaklı, Bocconi University
J. Tyson Chatagnier, University of Houston
Emre Hatipoğlu, SabancıUniversity
Although costs of trade disruption play a central role in theories of interstate conflict, scholars have had difficulty in
constructing appropriate measures of trade wars, and few have explored how states can mitigate the resulting costs,
reducing vulnerability to economic coercion. We study these questions in the context of economic sanctions, arguing that
each side’s comparative advantage in exports and domestic production capabilities determine its ability to minimize costs
while maximizing its power to hurt the adversary. We find support for our hypotheses, using commodity-level trade data.
Sanctions are more likely to succeed when sanctioners have a comparative advantage in goods exported to the target, but
more likely to fail if the target’s export portfolio is diverse or the target has a comparative advantage in exports. This is
particularly true for imposed sanctions. These findings open the black box of sanction costs, improving our understanding
of when economic coercion succeeds.
q1 In the aftermath of the February 2014 Ukrainian revolution,
a crisis began to unfold in the south of the country. Days
after Russian-speaking Ukrainians held a series of antirev-
olutionary protests, Russian forces entered Crimea, and by
March, the peninsula had joined the Russian Federation,
following a fraudulent referendum. The West responded with
a series of economic sanctions intended to force Russia to
change its Ukrainian policy. Initially, this move was criticized
as ineffective (see, e.g., Chapman 2014) or even counterpro-
ductive (Sher 2014). But during the second half of 2014, crude
oil prices plummeted more than 50%, and Russia, a major oil
exporter, found itself plunged into economic crisis. Suddenly,
commentators were much more optimistic about the effec-
tiveness of economic sanctions against Russia and argued that
nowwasthetimetopressPutinforcompromise(e.g.,Saun-
ders 2014).
The confidence with which many commentators imme-
diately denounced the sanctions as ineffective when imple-
mented, and subsequently praised them following the oil
crash highlights the fact that we know relatively little about
when and why trade sanctions work. This is an important
shortcoming of our discipline, because economic sanctions
present an attractive alternative tothe use of force and are used
with increasing frequency around the world. For instance,
whereas sanctions were rarely mentioned in US National Se-
curity Strategy documents before 2010, they received nine men-
tions in 2015 alone. But sanctions are used widely throughout
the rest of the world as well. Russia imposed a broad range of
sanctions against Turkey in 2015 after the latter shot down a
Russian military jet. In recent years, China has used trade
embargoes to punish countries that host critics of the regime
or have border disputes with China. Finally, in the summer
of 2017, four Arab countries led by Saudi Arabia imposed a
blockade against Qatar, accusing it of supporting terrorism
and Iran in the Middle East.
1
In this global context it is clearly
important for scholars to analyze the conditions under which
economic sanctions succeed.
An important, yet underexplored, topic is how states adapt
to the disruption of trade. When sanctions are imposed, we
often see target states attempt to minimize their losses and
Kerim Can Kavaklı(kerimcan@gmail.com) is an assistant professor of social and political sciences at Bocconi University, 20100 Milan, Italy. J. Tyson Chatagnier
(jtchatagnier@uh.edu) is an assistant professor of politicalscience at the University of Houston, Houston, TX 77204. Emre Hatipoğlu(ehatipoglu@sabanciuniv.edu)
is an associate professor of political science at SabancıUniversity, 34956 Istanbul, Turkey.
Data and supporting materials necessary to reproduce the numerical results in the article are available in the JOP Dataverse (https://dataverse.harvard.edu
/dataverse/jop). An online appendix with supplementary material is available at https://doi.org/10.1086/707398.
1. For Russian sanctions against Turkey, see Girit (2016); for Chinese sanctions, see Chellaney (2017); for sanctions against Qatar, see Fattah (2017).
170878.proof.3d 1 04/29/20 21:13Achorn International
The Journal of Politics, volume 82, number 3. Published online Month XX, 2020. https://doi.org/10.1086/707398
q2020 by the Southern Political Science Association. All rights reserved. 0022-3816/2020/8203-00XX$10.00 000
impose costs on the sanctioner in order to avoid acquiescing.
In an early article, Galtung (1967) suggested that sanction
success depends on how easily a target can replace its sanc-
tioning trade partners. Even a major trade partner will not
have much leverage over a target if the target can easily switch
to other suppliers, which is what the USSR did in 1980, when
its biggest supplier of grain, the United States, initiated an
embargo against it (Paarlberg 1980). Despite this early atten-
tion, these ideas have not yet been integrated into a modern
bargaining framework to produce testable hypotheses. Un-
surprisingly, in the absence of theoretical development, exist-
ing empirical studies have resorted to problematic measures
of the costliness of sanctions and arrived at sometimes-
contradictory findings regarding when sanctions hurt a coun-
try’s economy and how these costs translate into outcomes
(Kaempfer and Lowenberg 2007). Here we provide a relevant
theoretical framework—in which a sanction episode is con-
ceptualized as a part of the bargaining process, analogous to
war in the conflict bargaining literature—from which we de-
rive novel hypotheses.
In our framework, sanctions are costly for both sides, and
both the target and sanctioner will take steps to minimize their
own costs while maximizing the other side’s costs. The sanc-
tioning process allows the two sides to use their respective
abilities to cause economic harm to the opponent and to
prevent such harm from befalling themselves. This shapes the
bargaining landscape, eventually leading the state that can
no longer inflict pain on its opponent to yield. We argue that
states are able to subvert the opponent’s ability to inflict pain
with a combination of external and internal substitution. Ex-
ternal substitution involves switching to new trade partners.
This was the strategy of Apartheid South Africa during the
1980s, which allowed it to recover 86% of its lost export rev-
enue within a year (General Accounting Office 1992, 14).
Internal substitution is the tactic of overcoming import re-
strictions by producing the requisite good at home. This route
was taken by Rhodesia in the 1960s and 1970s, allowing it
to prosper even in the face of severe economic sanctions
(Chikuhwa 2006). These tactics determine a country’s outside
options, but they involve costs and may not be available to all
states. Even for South Africa, which was one of the most
successful cases of external substitution, there was still a sig-
nificant loss (about 14% of export revenue). Therefore, de-
pending on the context, cutting the same level of trade can
have different effects on different states. From this perspec-
tive, it is clear that the level or change in total trade between
countries as a result of sanctions is an ambiguous measure of
the policy’s costs for either side. Our framework enables us to
make predictions about not only how costly sanctions will be
in the aggregate but also how these costs will be shared by
former trade partners. In particular, we highlight the impor-
tance of three factors: market power (a combination of export
volume and comparative advantage), domestic production
capabilities, and export portfolio variety.
Our empirical contribution is to operationalize and test
explicit mechanisms that link trade to sanction success. Our
measures require the type of commodity-level data that have
been used in other areas of international politics (see, e.g.,
Chatagnier and Kavakli 2017; Li and Reuveny 2011) but have
not yet been used to examine sanctions. Therefore, we employ
commodity-level data on international trade between 1962
and 2000 to test our hypotheses. Our findings suggest that the
more market power a trading partner has (whether sender or
target), the more likely that that partner will achieve a fa-
vorable outcome in a sanction episode. Target states that have
the ability to produce a greater variety of products are harder
to defeat through sanctions. Interestingly, targets whose ex-
port portfolios are concentrated on a small number of goods
do not seem to be more vulnerable to sanctions; further anal-
yses indicate that an autocratic target whose income from
exports depends on a limited number of goods will be more
resilient to sanctions. These findings are robust to a variety of
specifications and estimators, including a novel nonparametric
estimator, which accounts for bias from selection or strategic
incentives in states’decisions to threaten and impose sanctions.
We also find that our measures provide a better fit with data
compared to previous measures of the costliness of sanctions.
This theoretical framework and its empirical operational-
ization can be applied to other areas of international relations
as well. For instance, scholars have argued that the potential
costs from trade disruption can pacify relations between states
(Russett and Oneal 2001) and motivate diplomatic and mili-
tary third-party interventions (Crescenzi et al. 2011; Stojek
and Chacha 2015). However, if the costs of this disruption can
be managed, states may be more willing to fight trade partners
and less willing to intervene in others’wars than existing work
leads us to believe.
The remainder of the article is organized as follows. We
begin with a brief review of the previous research on the effect
of trade ties on sanctions success. We then discuss our theo-
retical model of effectiveness in greater detail, present corre-
sponding testable hypotheses, and develop our measures of
economic importance. Next, we analyze the data and inter-
pret the results. Finally, we draw conclusions and make sug-
gestions for future research.
DETERMINANTS OF SANCTION SUCCESS
The logic behind economic sanctions is captured well by the
bargaining model of conflict (Bapat and Kwon 2015; McLean
and Whang 2010). A “sanction crisis”begins with a threat by
170878.proof.3d 2 04/29/20 21:13Achorn International
000 / Effectiveness of International Sanctions Kerim Can Kavaklı, J. Tyson Chatagnier, and Emre Hatipoğlu
a group of one or more sanctioning states (the “senders”)to
limit economic interaction with a target unless it changes a
particular policy. If the target concedes, then the crisis ends
with a policy change, and the two parties’utilities reflect this
change in policy. If the target resists, then the sender im-
plements sanctions, and trade is disrupted until one side gives
in. Once a crisis reaches this level, both sides are hurt eco-
nomically, and neither side knows, ex ante, which will concede
first. In other words, sanctions are a process, rather than a
one-shot interaction.
This implies that, during crises, actors may have to display
their respective abilities to inflict and bear costs in order to
demonstrate their bargaining power. Indeed, Schelling (1966,
v) calls “the power to hurt . . . a kind of bargaining power.”It
aims to compel the other side to make concessions, in order to
stave off further punishment. Therefore, a sender’s ability to
impose painful sanctions on a target provides it with signifi-
cant leverage, allowing it to demand more from its target.
However, as Slantchev (2003) notes, if the power to hurt
is bargaining power, then its denial undermines the oppo-
nent’s bargaining strength. When a state realizes its inability
to gain an advantage through punishment, it is more likely
to yield, especially when punishing the other side is costly for
the sender as well.
2
In short, both the power to hurt and the
power to bear costs constitute independent sources of bar-
gaining power.
What is needed here is a theoretical framework for the
sanctioning process. In particular, we must explain how
states are able to use economic tools to inflict pain on their
adversaries while finding ways to avoid similar costs them-
selves. The theory that we outline below helps us to under-
stand why states might resort to sanctions even when fully
informed about the costs to both sides, and it tells us which
side is likely to give in (i.e., whether the sanctions will suc-
ceed or fail). All of these insights depend critically on an
understanding of the sources of economic costs, and our
conceptualization of how the power to hurt comes into play.
While both the power to inflict and the power to bear
costs are crucial, few sanction studies separate and explicate
these costs in detail. The most commonly used sanction data
sets, Hufbauer et al. (2007) and Morgan, Bapat, and Koba-
yashi (2014), include ordinal measures of the costliness of
sanctions for the target and the sender individually. How-
ever, the ordinal nature of these variables makes it difficult
to understand which particular economic factors make sanc-
tions more costly. These indicators are also based on subjective
assessments of sanction costliness and suffer from severe
missing data problems.
3
As an alternative, some scholars (e.g.,
McLean and Whang 2010) measure a sanction’s costliness by
the difference in trade levels before and after sanctions are
imposed. One important drawback of this measure is that it
is only available for imposed sanctions and not for cases in
which sanctions were threatened but not carried out. More
recently, Bapat and Kwon (2015) and Peksen and Peterson
(2016) have explicitly modeled costs to parties from sanction
episodes using aggregate trade figures. However, such mea-
sures are unable to separate the sanctioners’and targets’costs
from trade disruption—which can have diametrically oppos-
ing effects on sanction success—and they ignore the target
state’s ability to replace lost trade through internal substitu-
tion. To remedy these problems, we propose a model of
sanctions below that defines power in terms of costs inflicted
and borne.
Although we approach sanctions as an interstate interac-
tion, we are aware that domestic politics can play an impor-
tant role as well (Kobayashi 2017). Sanctions are sometimes
“symbolic”in the sense that senders do not expect sanctions
to change target policy, but they impose them nevertheless to
demonstrate their displeasure with the target to domestic and
international audiences (Whang 2011). However, the exis-
tence of such cases will only make our task more challenging.
In cases in which senders intentionally design weak sanction
regimes by not employing their full leverage against the target,
our measures of sender power should overestimate effective-
ness, attenuating our estimates and making it more difficult
for us to find significant results. In other cases, senders may
employ sanctions against a highly resolved target that they
expect not to acquiesce to economic coercion. An example of
this might be the Arab states’sanctions against Israel in the
1980s. Although our empirical models include proxies for
target resolve, the concept is difficult to measure (Kertzer
2016). Therefore, we may be unable to explain such cases
adequately. Ultimately, however, this should once again bias
us against finding results in the aggregate. Therefore, the true
2. If sanctions have symbolic value and minimal costs for the sender, then
they may continue long after they are revealed to be ineffective. The best
example here is the Cuban embargo.
3. Among the 1,102 sanction episodes the Threat and Imposition of
Sanctions (TIES) data set includes for the time period we cover in this study,
our measure provides data for 1,017 episodes. The missing data in our
measurement occur in casesin which the target was the European Union (EU;
42 cases), a microstate, or a state enjoying its first year of independence. By
contrast, using the anticipated cost to the target and anticipated cost to the
sender variables of the TIES data set leaves us with 657 cases. Additionally,
these measures show little variation: for anticipated sender economic costs,
which are missing for 30% of the TIES data set, 95% of nonmissing obser-
vations are coded as “minor.”The other two categories—“major”and “se-
vere”—make up only 5% of the nonmissing cases.
170878.proof.3d 3 04/29/20 21:13Achorn International
Volume 82 Number 3 July 2020 / 000
impact of the economic factors that we identify should be
slightly stronger than the estimates that we provide.
COSTLINESS OF TRADE DISRUPTION
Understanding the costs associated with trade disruption re-
quires the discussion and operationalization of the two criti-
cally important types of bargaining power: the power to in-
flict costs and the power to bear or mitigate costs inflicted by
others. When applied to trade sanctions, however, these con-
cepts can be disaggregated further, allowing us to generate
novel hypotheses about economic coercion.
Sanctions impose costs on both the target and the sender,
as trade disruption reduces the economic benefits enjoyed by
both parties. Moreover, these costs are not completely pre-
dictable. Even if the sender attempts to impose sanctions that
will cause little harm to itself, it cannot ensure that disrup-
tion will be limited to those commodities included in its
sanction regime. The target can retaliate by disrupting trade in
commodities that are important to the sender. An example of
this phenomenon is Russia’s ban of agricultural imports from
the EU, in response to European sanctions over the Ukraine
crisis. A particularly powerful target can make it difficult for
senders to maintain the sanctions, reducing their likelihood
of success.
4
A state’s ability to bear costs, either as a target or a sender,
depends on its ability to find alternative buyers for its prod-
ucts and alternative suppliers of its imports. If a state benefits
from exporting goods, then sanctions hurt by reducing prof-
its. While a state cannot completely eliminate sanction costs,
the harm will be significantly smaller if its exports are highly
sought after in the global marketplace (due either to low prices
or high quality). Finding new buyers is easier when the state
has a large comparative advantage in the production of that
commodity. Likewise, import substitution is easier if there are
other suppliers in the world who sell low-cost or high-quality
versions of the relevant goods.
5
In short, the exporter’s comparative advantage affects
sanction effectiveness in two ways. First, greater comparative
advantage in a given good allows the exporter to find alter-
native buyers more easily. Second, it makes it more difficult
for the importer to find alternative sellers. If the sanctioner is
the exporter, comparative advantage makes sanctions costlier
for the target. If the target is the exporter, it reduces the costs
inflicted by the sanctions and gives the target more power to
hurt the sanctioner back. The effect of trade on sanction
success, then, depends on who is selling which commodity
to whom and on the seller’s comparative advantage in that
commodity.
Iran’s countermeasures against Western sanctions illus-
trate the link between sender market power, target ability to
resist, and the power to hurt. Despite its vast crude oil reserves,
Iran’s limited refining capacity forced the country to import
more than 40% of its gasoline when sanctions began in 2010
(Dombey, Morris, and Blitz 2009). The sanctions aimed
to exploit this weakness by banning the export of refining
equipment and technology to Iran. However, the United
States soon discovered that Western companies had limited
market power in the global production of refining equipment,
which allowed Iran to obtain such equipment from non-
Western (mostly Chinese, Indian, and Malaysian) companies,
reducing the burden of sanctions (Van de Graaf 2013).
From our bargaining model of sanctions effectiveness, we
can derive two hypotheses. First, we expect sanctions coming
from sender states with greater market power to be more ef-
fective in accomplishing their goals.
H1. The more easily the target can externally substitute
for sender’s exports, the lower the likelihood of sanc-
tions success.
By the same logic, senders will be less sensitive to sanction
costs if their dependence on the target or target’s comparative
advantage is low. Under these conditions, the sender will find
it easier to substitute for the target’s exports externally, al-
lowing it to bear the costs of sanctions.
H2. The more easily the sender can externally sub-
stitute for the target’s exports, the higher the likeli-
hood of sanctions success.
At this point, we reemphasize that tests of hypotheses 1 and 2
require the disaggregation of total trade into imports and
exports. We expect these two (weighted) components of trade
to have opposite effects on sanctions success. This is in stark
contrast to previous works that measure target vulnerability
using the amount or decline of total trade (Bapat and Kwon
2015; McLean and Whang 2010; Whang 2010).
The second major factor that determines a state’s vulner-
ability to trade disruption is its ability to use internal substi-
tution. A country that has the domestic capabilities to produce
a large variety of goods at home should be more resilient to
4. We do, however, see some instances of sanctions being implemented
against high-power targets. Just as private information over resolve can ex-
plain war between rational states, it may also account for sanctions against
powerful targets. Senders are likely to be uncertain about the target’s ability to
withstand costs. If the target is highly resolved, then even high-cost sanctions
may fail; conversely, because substitution is costly, even states that are able to
switch may be unwilling to do so if their resolve is not sufficiently high.
5. In addition to comparative advantage, dyad-specific factors (e.g.,
distance) may also affect trade volumes.
170878.proof.3d 4 04/29/20 21:13Achorn International
000 / Effectiveness of International Sanctions Kerim Can Kavaklı, J. Tyson Chatagnier, and Emre Hatipoğlu
import restriction. A diverse industrial base will allow the
targeted state to transfer technology, innovate, and commit
production facilities to produce the embargoed good at home
(Hausmann and Hidalgo 2011). A good example of such an
adjustment was the apartheid government’s Sasol initiative in
South Africa. The regime’s advanced production facilities in
other areas of petrochemical production were key to (par-
tially) offsetting the shortages induced by the international oil
embargo as these facilities were retrofitted to produce oil from
coal (Kaempfer and Lowenberg 1988). We hypothesize that
countries with diverse export portfolios are more capable of
internal substitution. Economists have shown that more de-
veloped countries and those that are able to produce more
complex products tend to export a wider variety of goods in
the global market (Hidalgo et al. 2007; Saviotti and Frenken
2008). Moreover, data on exports tend to be classified con-
sistently across countries. For these reasons, we believe that
diversity of exports is a good indicator of domestic production
capabilities.
6
This argument anticipates our next hypothesis.
H3. The greater the number of commodities exported
by the target state, the lower the likelihood of sanctions
success.
Next, using arguments about contracting costs and en-
forcement capability, we hypothesize that countries with
more concentrated export portfolios will be easier to sanc-
tion effectively. When a target concentrates on a few exports,
sanctions should be easier to design and enforce, resulting in
less sanction-busting, which is important for sanction suc-
cess (Early 2009).
7
H4. The more concentrated the export portfolio of
the target state, the higher the likelihood of sanctions
success.
Finally, given that sanctions are a process, there is a
question of the stage at which they will be effective. States
may actually have to implement sanctions and inflict pain
even if there are some observable measures of their capacity
at the onset of a crisis. First, if an actor underestimates the
cost of sanctions or the other side’s determination to impose
them, then imposing sanctions may be necessary to correct
this information problem (Hovi, Huseby, and Sprinz 2005).
Second, Slantchev (2003) shows, in an analogous scenario,
that costly conflict can be an equilibrium even between fully
informed states. In this case, states prefer sanctions to certain
unacceptable settlements, and one side will opt to settle only
when its capacity to hurt the opponent or to bear costs has
been sufficiently compromised. Drawing on these ideas, we
present our last hypothesis.
H5. The cost associated with a particular sanction
will matter more in the imposition stage than in the
threat stage.
RESEARCH DESIGN
Our hypotheses concern the effects of economic power on
the efficacy of sanctions. Thus, our main dependent variable
in our initial analysis is the success or failure of sanctions,
whether threatened or imposed. We draw the sanctions data
from Morgan et al.’s (2014) TIES data set. Later we analyze
these two stages (threat and imposition) separately. Our unit
of analysis is the sanction episode, which includes any in-
teraction during which one or more states imposed or
threatened to impose sanctions on another. While the TIES
data range from 1945 to 2005, the availability of our key
independent variable restricts our analysis to the years 1962–
2000. We limit our set of observations to those sanctions
episodes in which export or import sanctions or both were
threatened or imposed, as these constitute the set of relevant
cases for our theory; we do not expect market power to play a
role in the outcome of noneconomic sanctions. Empirically,
we do this by omitting sanction episodes in which the threats
fall solely within categories 7–10 in the TIES data.
8
The main dependent variable in our analysis is binary: we
code episodes in which the target capitulates or a negotiated
settlement is reached as successful, and we code episodes
in which the sender capitulates or there is a stalemate as
failures. In coding this variable, we note that some sanctions
do not definitively terminate but simply fizzle out. Thus, we
are missing data on episodes that are technically still ongoing
but have effectively ended without changing target behavior.
6. Consistent with this idea, the number of commodities exported by
South Africa in 1983 was 71, more than 1 standard deviation above the
world average of 56 that year.
7. Our theory contends that the target may consider hurting back with
countersanctions. Following the same line of logic, we may also posit that the
sender’s industrial setupand its level of export concentration may mitigatethe
costs imposed by sanctions. We choose not to incorporate these additional
two hypotheses for the sake of brevity. Calculating these two variables for
sanctioning coalitions is also problematic: various transaction costs may
hinder product substitution across sender countries. Nonetheless, additional
tests in the appendix show that controlling for the breadth of the sender’s
export portfolio does not change our main findings.
8. Categories 7–10 include, respectively, asset freezes, termination of
foreign aid, travel bans, and suspension of economic agreements. Our results
are robust to including category 10.
170878.proof.3d 5 04/29/20 21:13Achorn International
Volume 82 Number 3 July 2020 / 000
If a sanction has not explicitly terminated but has not been
mentioned in 10 years or more, we assume that the sanction
has failed. As the TIES data set ends in 2013, any sanction
episode listed as ongoing as of 2003 or earlier is coded as a
failure. We show below that our results are robust to using
only those sanction episodes that definitively terminated. They
are also robust to using a three-category dependent variable
(win, lose, and draw) analyzed with an ordered logit estimator.
To test hypothesis 5, we first focus on “threat effective-
ness”: whether a sanction threat altered target behavior with-
out escalating to the imposition stage. If sanctions were
implemented, we count this as an ineffective or failed threat.
Finally, we analyze imposed sanctions separately. This anal-
ysis is the same as the original except that threats that were
never implemented are excluded (regardless of outcome)
from the sample.
Measuring cost of sanctions
To test our hypotheses, we employ four key independent
variables. The operationalization of the variables for hy-
potheses 3 and 4 is relatively straightforward. The breadth of
the target’s export portfolio (hypothesis 3) is simply a measure
of the number of goods that a country exports in a given year.
We use Feenstra et al.’s (2005) commodity trade data set,
at the two-digit level (approximately 100 commodity types),
and count the number of different commodities traded by
each country in each year. This value ranges from a min-
imum of nine export items (Cambodia in 1988 and Rwanda
in 2000) to a maximum of 79 (the Netherlands in 1986 and
China in 1997). We expect higher values of this variable to
be related to greater target resilience.
9
The concentration
variable (hypothesis 4) calculates the value concentration of
these commodities, regardless of the absolute number of the
goods traded by the target country. We create this variable
by calculating the Herfindahl-Hirschman (HH) index of the
target country’s trade portfolio in a given year, in terms of
dollars.
10
The higher this value, the more concentrated a
target’s export portfolio is, and the more likely that sanctions
should succeed.
Our measures of relative market power (hypotheses 1 and
2) are more complex. We begin by calculating, in each year,
for each pair of states, iand j, the market power of state iover
state j. This measure is a function of market size and com-
parative advantage, across various different commodities.
In particular, we conceptualize i’s market power over jin
year tas a weighted measure of dependence:
Dijt po
mXm
ijt
Mjt
#CAm
it ;ð1Þ
where Xm
ijt represents the volume of i’s exports of commodity
mto country jin year t,
11
M
jt
represents state j’s total imports
in year t, and CAm
it is state i’s comparative advantage in
commodity mduring year t. The ratio of state j’s imports of
commodity mfrom state ito its total imports provides a
measure of dependence that accounts for total market size.
12
We obtain the value of dyadic commodity exports from the
Feenstra et al. (2005) data set.
We operationalize comparative advantage as a country’s
relative revealed comparative advantage (RRCA). Revealed
comparative advantage (RCA), originally developed by
Balassa (1965, 103), is widely used in international economics
to measure the degree of production/price advantage a state
has in exporting a specific good (French 2017). We create our
RRCA measure by first calculating RCA for each commodity-
country-year, using Balassa’s formula. A state’s RCA is ef-
fectively the ratio of a state’s exports in a given commodity to
9. The correlation between the number of exports and the total gross
domestic product (GDP) generated from industrial endeavors (in logged
dollars) for a country in a given year is 0.64. This correlation suggests
more breadth in an export portfolio indicates a stronger industrial and
technological base, which can be geared toward the production of different
commodities in the targeted state.
10. For a country trading ndifferent commodities in a given year, the
HH index is calculated as HH pon
ip1T2
i,whereT
i
is the value of trade
in commodity i. HH will be bounded by 1/nand 1.
11. Our measure includes all commodities in each state’s export portfolio,
even though most sanctions and threats do not include all trade between the
parties. As mentioned above, this is because neither party can be sure which
goods will ultimately be restricted. Indeed,more than half of threats that made
no mention of export sanctions ultimately restricted exports to the target
when implemented. Additionally, the target can retaliate by restricting any
export it wishes. Recent sanctions against Russia illustrate these phenomena.
Even after sanctions were initially implemented, EU officials emphasized the
possibility of broadeningthe sanctions unless Russia changed its policy. At the
same time, Russia threatened symmetrical countermeasures (see, e.g., Reuters
2014). For these reasons, both sides must consider their costs and benefits
from a trade disruption of all commodities. This approach is similar to
academics’use of information on a country’s total military and economic
resources to measure its power during a crisis, even though very few conflicts
require that countries employ all of their national resources in battle.
12. Our measure of market power is somewhat similar to a measure of
trade elasticity. Senders with greater power can be compared to suppliers of
inelastic goods: they are able to inflict more pain on the target, and the target
can do little to ease that pain. We thank a reviewer for pointing this out.
However, our measure has an important advantage in that it is global in
nature. Therefore, itcan account for situationsin which an important trading
partner can be replaced by alternatives.The grain embargo against the Soviets,
mentioned above, is a good example of this phenomenon. Even though the
United States was the USSR’s largest supplier of grain, dyadic trade was highly
elastic because the Soviets were able to replace the United States as a supplier
with ease, and the sanctions had little impact.
170878.proof.3d 6 04/29/20 21:13Achorn International
000 / Effectiveness of International Sanctions Kerim Can Kavaklı, J. Tyson Chatagnier, and Emre Hatipoğlu
its total exports, divided by the global mean. Thus, in year t,
for each state, i, and each commodity m:
RCAm
it
pXm
it
on≠mXn
it oj≠iXm
jt
oj≠ion≠mXn
jt ;ð2Þ
where Xm
it is the value of state i’s total exports of commodity
mto all other states in year t.
13
We then scale the result, to get
a relative measure. We do this by finding the maximum (fi-
nite) value of RCA for each commodity-year and dividing by
that value. This returns a value, RRCAm
it ∈½0;1, which we
use to proxy for CAm
it in equation (1).
There are two possible cases in which the value of RCAm
it
can be infinite: if the state only exports one commodity (so
that the denominator of the numerator in eq. [2] is zero) or
if the state is a monopolist in a given year (in which case
the denominator in eq. [2] is zero). In both cases, we simply
set the value for CAm
it to one, indicating maximum com-
parative advantage.
After calculating equation (1), we are left with a weighted
aggregate dependence value for every directed-dyad-year. Ex-
porters with high market power in a given year are those that
make up a large proportion of their partner’s imports, by sell-
ing commodities for which they have acomparative advantage.
Given that 27% of sanction threats and impositions in our
sample come from a coalition of states (Morgan et al. 2014),
we use the weighted aggregate dependence values for each
dyad to create a measure of the total market power of the
sanctioning coalition (defined below) on the target. We do
this by simply aggregating the values for all coalition mem-
bers. Since each state participates in the sanctions, the target
state should lose the combined value of all members. Thus,
an additive measure of value is appropriate.
14
Similarly, we
measure target power over a sanctioning coalition by sum-
ming the dyadic measure over each member of the sanc-
tioning coalition. We define a sanctioning coalition as the
group of all individual sender countries listed in the TIES data
set for a given sanction episode. If sanctions are initiated
through an international institution, we include all of that
institution’s members.
15
Data
We begin by examining the initial relationship between our
dependent variable and our variables of interest. Table 1 q2dis-
plays the mean levels of our four main independent variables
for successful and failed sanctions. In two cases the differences
are significant at the p!:01 level and in the expected direc-
tion.
16
A third variable, target market power, shows a large
effect and nearly attains statistical significance (p!:11). In
general, successful sanctions are more common when target
export variety and market power are low and sender market
power is high. In this bivariate analysis, levels of target export
concentration do not seem to influence sanction outcomes.
These basic results are consistent with the predictions of three
of our hypotheses above.
The results in table 1 are encouraging. However, this is
only a first look. Our regression analysis includes a number of
controls drawn from the literature.
17
Our first set of control
variables accounts for target characteristics. Sanctions should
be less effective on powerful states, as such countries are both
more self-sufficient and better able to deploy countermea-
sures against sanctions. We account for this with measures of
both economic (the log of the target’s total and per capita
GDP; Gleditsch 2002) and military power (measured by the
target’s composite index of national capability [CINC] score;
Table 1. Average Values of Our Main Variables
by Sanction Outcome
Target
Export
Variety
Target
Portfolio
Concentration
Target
Market
Power
Sender
Market
Power
Successful
sanctions 63.27 .10 2.09 3.31
Failed
sanctions 66.93 .10 3.35 2.35
Difference 23.66*** .004 21.26 .95***
Note. All significance tests are two-tailed.
*p!.10.
** p!.05.
*** p!.01.
13. As discussed below, some values are infinite. Within the data, the
finite values of RCA range from a minimum of zero, for any case in which
a state did not export a particular commodity in that year, to a maximum
of 179,085.
14. Our measure of market power enters into our model linearly.
However, given its skewness and the possibility of diminishing marginal
effects, we have examined several logarithmic specifications. Our general
substantive findings remain unaffected.
15. We include sanctions initiated by the EU but exclude sanctions
where the EU is the target. The reason for this is that variables such as
target democracy and target’s military capabilities are not well defined for
the EU, which has supranational authority over its member states in some
issues but has not traditionally exercised it in others, including interna-
tional security.
16. There are 606 complete observations for each variable in table 1.
As discussed above, non-trade-related sanction episodes are deleted from
our sample.
17. All data come from the TIES data set, unless otherwise indicated.
170878.proof.3d 7 04/29/20 21:13Achorn International
Volume 82 Number 3 July 2020 / 000
Singer 1988). To account for the fact that democratic states
may be more vulnerable to sanctions, we also include the tar-
get’s Polity 2 score (Marshall and Jaggers 2002), which varies
between 210 and 10. Additionally, we include the number of
states involved in the sanctioning coalition. Large sanction-
ing coalitions should be more likely to face collective action
problems and are expected to be associated with lower levels
of sanction success.
We also consider the context in which sanctions are im-
posed. Most importantly, we expect different dynamics of
sanction imposition and sanction-busting in the Cold War
and post–Cold War periods (Jentleson 2000). In addition, we
control for security-related sanctions, as target states are likely
to show greater resolve in such issues and less likely to give in
(Morgan and Schwebach 1997).
The challenge of strategic interaction
and nonparametric estimation
The most straightforward way to analyze our data would be
with a simple logit model, which is how many previous
studies have proceeded (e.g., Ang and Peksen 2007; Bapat
and Morgan 2009). So that it can be more easily compared
with previous research, we begin our analysis with a logit
model. However, this technique neglects two possible sources
of bias: selection and strategy. Nooruddin (2002) points out
that sanctions are a matter of choice and that senders and
receivers select themselves into sanctions (see also Lektzian
and Souva 2007). Because the selection stage is potentially
correlated with the overall outcome, failure to account for this
process can result in selection bias, which is a type of speci-
fication error (Heckman 1979).
Other scholars argue that there is a specific structural re-
lationship between sanction imposition and success. These
authors posit that the decision to impose sanctions is based on
the sender’s beliefs about the target’s likelihood of concession,
given that sanctions are inflicted. Thus, the process is not
simply one of selection but is fully strategic in nature, and
failure to account for the strategic aspect will bias estimates
(McLean and Whang 2010; Whang 2010; Whang, McLean,
and Kuberski 2013).
We remain skeptical of both approaches. First, a properly
specified model for either the selection or the strategic pro-
cess requires that the analyst have the full universe of cases on
which selection or strategic interaction could occur. In this
case, this means that we must have data on all of the episodes
for which one state might have threatened or imposed sanc-
tions but opted not to do so. Without these data, we can-
not correctly estimate either type of model. Second, while
a structural statistical model that matches the true data-
generating process would be ideal, applying a model with an
inappropriate structure would result in the very bias that we
hope to avoid.
For this reason, we choose an alternative model specifi-
cation. Both selection and strategic misspecification are tan-
tamount to omitted variable problems, which can be con-
ceptualized as model specification problems (see Signorino
and Yilmaz 2003). For situations in which it is appropriate to
treat these issues as nuisances rather than substantively in-
teresting features in themselves, we can obtain unbiased esti-
mates of the other coefficients if we use methods that “allow
for the estimation of functions of unknown form”(Kenkel
and Signorino 2012, 2), such as fully nonparametric estima-
tion. In this case, we are most concerned with the bias that
specification error might induce with respect to our variables
of interest. Thus, rather than attempt to use a structural es-
timator that may not be appropriate for our data, we opt for
flexibility, choosing to implement local likelihood logistic
regression (Fan, Heckman, and Wand 1995; see also Cha-
tagnier 2014; Frölich 2006). This is a local smoothing tech-
nique that makes no assumptions about global functional
form and can model any number of processes.
18
Thus, our
flexible, nonparametric estimator can account for bias from
either selection or strategic interaction, without forcing us to
apply one particular specification.
19
Estimation is similar to
that of a parametric logit at each point within the data set, but
the contributions of other observations are “weighted”ac-
cording to their distance from the current data point. The size
of the smoothing window is constant across all points and is
established beforehand, using leave-one-out cross-validation.
Rather than a single set of coefficients, the local logit estimator
returns ndifferent sets of coefficients. For this reason, we
cannot present a simple table of parameter estimates and must
instead present results graphically.
RESULTS
We begin by estimating a series of conventional logit models
that examine the effects of our key independent variables,
controlling for other factors.
Logit analysis of sanctions success
The results from the parametric logit models appear in ta-
ble 2. The first three models include our primary dependent
variable: sanctions success for sanctions that have terminated.
18. This includes the parametric logit. When the smoothing window
becomes sufficiently large to include all observations, the local logit estimates
converge to the parametric estimates. Thus, the local logit actually subsumes
the conventional logit.
19. The trade-off that we must make is increased variance; our estimates
will have wide confidence intervals.
170878.proof.3d 8 04/29/20 21:13Achorn International
000 / Effectiveness of International Sanctions Kerim Can Kavaklı, J. Tyson Chatagnier, and Emre Hatipoğlu
In columns 4 and 5 we analyze the threat and imposition
stages separately. Column 1 provides results from a baseline
model that uses target trade dependence as the key indepen-
dent variable.
20
Columns 2–5 show the results of regressions
using the variables that capture a state’s ability to inflict or
withstand pain from sanctions. The results in model 1 show a
positive and significant effect for target trade dependence,
suggesting that a simple measure of trade dependence is in-
deed associated with sanctions success. However, when look-
ing at measures of goodness of fit, we see that the trade de-
pendence model has significantly less explanatory power. The
20. Target trade dependence is defined as the target’s total trade with
sender coalition over target’s total GDP.
Table 2. Logit Analysis of Sanctions Success
Trade
Dependence
Power
to Hurt
Interaction
Term
Threat
Effectiveness
a
Only Imposed
Sanctions
b
(1) (2) (3) (4) (5)
Trade dependence 6.797*
(4.006)
Sender market power over target .126** .128** 2.019 .228***
(.050) (.050) (.044) (.065)
Target market power over sender 2.021* 2.020* 2.023* 2.037**
(.011) (.011) (.013) (.015)
Target export variety 2.051*** 2.054*** 2.009 2.084***
(.018) (.019) (.016) (.029)
Target portfolio concentration 22.057 2.544 1.072 22.421
(.710) (.772) (.821) (1.556)
Target portfolio concentration #
target democracy .287*** .433*** .014
(.102) (.109) (.189)
Target’s total GDP 2.145 .022 .088 .015 .037
(.114) (.131) (.137) (.101) (.180)
Target’s GDP per capita 2.015 .174 .291** .418*** .059
(.147) (.128) (.139) (.130) (.204)
CINC 4.970** 2.880 1.975 4.250 3.551
(2.376) (2.628) (2.692) (2.598) (3.838)
Target democracy 2.021 2.043** 2.076*** 2.101*** .005
(.019) (.018) (.020) (.023) (.033)
Cold War 2.119 .040 .050 .344 .335
(.199) (.222) (.225) (.239) (.343)
Security issue 2.461** 2.301 2.231 2.431 2.100
(.220) (.225) (.240) (.291) (.378)
Coalition size .004 2.003 2.004 .007 .009
(.004) (.007) (.007) (.007) (.009)
Constant 2.883 1.414 2.695 24.403** 3.511
(1.849) (1.881) (2.023) (1.851) (2.777)
N594 594 594 573 312
Log likelihood 2401.249 2387.939 2384.342 2306.418 2182.157
AIC 820.498 799.878 794.684 638.835 390.314
Area under ROC curve .62 .67 .68 .65 .74
Note. Robust standard errors clustered on target in parentheses. GDP pgross domestic product; CINC pcomposite index of national capability; AIC p
Akaike information criterion; ROC preceiver operating characteristic. All significance tests are two-tailed.
a
Dependent variable is whether a sanction threat succeeds. Sanctions imposed without a threat are excluded.
b
Dependent variable is whether an imposed sanction succeeds. Sanctions that were threatened but not imposed are excluded.
*p!.1.
** p!.05.
*** p!.01.
170878.proof.3d 9 04/29/20 21:13Achorn International
Volume 82 Number 3 July 2020 / 000
Akaike information criterion (AIC) for model 1 is signifi-
cantly higher than that of the other models, and the within-
sample predictive success—as measured by the receiver op-
erating characteristic (ROC) curve—is markedly lower.
21
The
greater explanatory power for our measures is noteworthy,
but not wholly unexpected. Unlike the aggregate trade mea-
sure, our variables incorporate information about the target’s
domestic production capabilities and each side’s comparative
advantages. They also disaggregate trade volume in a way that
reveals its contribution to both sender and target market
power, rather than trying to estimate an ambiguous net effect.
Turning to the second set of models, we first look at
sender and target market power variables. As expected, sender
power has a positive and target power a negative impact on
sanction success, suggesting that our measures capture states’
coercive abilities. These findings support both hypotheses 1
and 2. Next, we look at the variety of the target state’s export
portfolio. Consistent with hypothesis 3, the estimated effect is
negative and significant, indicating that, all else equal, the
greater the number of unique target exports, the less likely
sanctions are to succeed.
Surprisingly, while we expected target portfolio concen-
tration to have a positive effect, the estimate is negative and
statistically significant at the p!:01 level. To understand
this estimate better we took a closer look at “deviant cases”
(i.e., sanction failures against targets with highly concen-
trated export portfolios). We find that almost all of those
targets are full autocracies. To be sure, most countries with
highly concentrated export portfolios are autocratic, but the
few democracies among them give in to sanctions.
22
Based on
this observation, model 3 includes an interaction between
target portfolio concentration and target democracy, to test
whether the effect of a target’s export concentration depends
on its regime type. This interaction term is statistically sig-
nificant, and we interpret the substantive effects graphically
in figure 1. Model comparisons show that including this
interaction improves model fit. We also checked whether
interactions between target democracy and any other eco-
nomic variables should be included, but these additional
interactions did not produce significant effects, and AIC
values suggested that they did not fit the data as well.
Before we turn to substantive effects, we investigate hy-
pothesis 5 by examining threat effectiveness and imposed
sanction effectiveness separately. According to models 4
and 5, target market power over sender has similar effects in
each stage, and target portfolio concentration (conditional
on a target’s democracy level) matters primarily in the threat
stage. But, sender market power over target and target export
variety matter solely at the imposition stage. These two
variables measure a target’s ability to substitute for imports
from the sender coalition. Consistent with hypothesis 5, this
pattern suggests that negotiations at the threat stage are
not sufficiently informative about the costliness of conflict
to compel acquiescence from the target. Sometimes actors
must escalate and actually impose economic pain in order to
demonstrate their bargaining power.
Moving to the substantive interpretation of our results,
we provide depictions of table 2 model 3 in figure 1. We plot
the results with continuous variables set to their means, and
discrete variables set to their medians. We then vary the
factor of interest. We set dichotomous variables to zero.
23
The plotted lines represent the predicted probability of sanc-
tions success, while the shaded regions are 90% confidence
intervals. The substantive results are strong. As the sender’s
market power increases from our empirical minimum to our
empirical maximum, the probability of sanction success more
than doubles. Meanwhile, varying target power from its min-
imum to maximum value reduces the probability of success
by about 90%.
24
Portfolio size has a similarly powerful effect.
Whereas sanctions against a state that exports only a few goods
(the minimum value in our data set is nine) are virtually as-
sured of success, sanctions against a country with an especially
broad trading portfolio (the maximum in our data is 79) face
about a 75% chance of failing. Finally, when examining the
effect of portfolio concentration, we see the role of regime
type. For democratic states, we obtain the expected pattern:
sanctions against targets with highly concentrated trade port-
folios are almost guaranteed to succeed (although the confi-
dence interval is large at the high end), while those against
targets with diverse portfolios are more likely to fail. For au-
tocracies, the effect is reversed. Sanctions against autocratic
targets with heavily concentrated portfolios are almost certain
to fail, while those against targets with diverse portfolios have
a high likelihood of success. Note that these findings regarding
portfolio concentration are not driven by oil exporters and are
21. The results are similar for models that use total aggregate trade,
rather than trade dependence, or that use separate measures of aggregate
imports and exports.
22. For instance, there are 29 sanction target countries with a concen-
tration higher than 0.5. Of those 29 sanctions, 16 succeeded and 13 failed. Of
the 13 targets that refused to give in to sanctions, only 1 has a Polity score
higher than 0 (Mali 1999). By contrast, 5 of the 16 targets that gave in to
sanctions had scores above 0.
23. To account for the interaction effect, we plot portfolio concen-
tration results using Polity 2 scores at both 28and8.
24. As discussed in the appendix, the effects of sender and target power
are similar to those of target democracy and intergovernmental organization
involvement.
170878.proof.3d 10 04/29/20 21:13Achorn International
000 / Effectiveness of International Sanctions Kerim Can Kavaklı, J. Tyson Chatagnier, and Emre Hatipoğlu
robust to controlling for a target’s oil exports and whether the
target is an OPEC member.
The finding that among targets with a more concentrated
portfolio autocracies have an advantage is consistent with
previous work (Lektzian and Souva 2003, 2007). When a
target with a concentrated economy faces deep trade cuts,
authoritarian governments with stronger control over the
economy may more easily weather the shock. In contrast,
a broader economy is harder to control by force. When a
country with a diverse economy is targeted by sanctions,
some amount of popular participation and democratic le-
gitimacy may be necessary to mobilize the public.
Figures 2 and 3 depict the effects seen in models 4 and 5
of table 2, separating out threatened and imposed sanctions,
respectively. In general, these plots suggest that the effects
that we see in figure 1 are primarily driven by imposition.
For sanction threats, the results are mostly flat, although
we see minor effects for target power. The role of portfolio
concentration appears to be similar to that seen in figure 1.
Moving to actual sanctions (fig. 3), we see strong predicted
effects that look similar to the substantive results presented
above. The one difference is portfolio concentration, par-
ticularly among democracies. For autocracies, we see an
effect of target portfolio concentration that is roughly sim-
ilar to the predictions in figure 1. Turning to sanctions im-
posed on democracies, however, we see an effect that is nearly
identical to that of autocracies, although estimated imprecisely.
This may be because there are relatively few cases in which
sanctions were imposed on democracies with high levels of
portfolio concentration. Ultimately, the substantive results
in figures 2 and 3 suggest that the realization of the power to
hurt economically is crucial for sanction effectiveness.
As a complement to our statistical evidence, we can illus-
trate our arguments with real-world cases of sanctions. Take,
for instance, the 1969 Football War between Honduras and El
Salvador. In July 1969, the two countries fought a brief war,
during which the Salvadoran army invaded Honduras. When
Salvadoran troops proved reluctant to depart after the hos-
tilities ended, the Organization of American States (OAS) had
to threaten El Salvador with economic sanctions unless it
withdrew. Within our data, this is a case in which the sender
coalition had significant market power over the target—target
vulnerability was within the top 5% of observations—and its
threat should have compelled the target to comply. Indeed,
according to James (1990, 61), “it took a tacit threat of [OAS]
sanctions to persuade El Salvador to withdraw.”
We can contrast this case with the unsuccessful sanction
threat that Egypt and its allies issued against the United States
in early 1970 in an effort to curb the latter’s material support
to Israel. The United States refused to alter its policy, and
the sanctions were never implemented. This is unsurpris-
ing, as the coalition—which was not composed of all OPEC
members, notably lacking actors such as Saudi Arabia, Iran,
and Libya—possessed relatively little market power over the
United States, falling just outside of the lowest quartile within
our data. Meanwhile, the United States held significant power
Figure 1. Logit predicted probabilities
170878.proof.3d 11 04/29/20 21:13Achorn International
Volume 82 Number 3 July 2020 / 000
over the coalition, falling well within the top decile of the
data set. Thus, the fact that the threat failed is consistent with
our arguments.
In addition to the results here, we conduct a number of
robustness checks, provided in the appendix. These include
models that use alternative specifications, alternative defini-
tions of both the dependent variable (sanction success) and
the sanctioning coalition, and additional checks on raw mea-
sures of trade dependence. In most cases, our results are sub-
stantively similar to what we report above. As a further ro-
bustness check, we estimate an auxiliary model that includes
three additional variables: whether sanctions were brought to
Figure 3. Imposed sanctions
Figure 2. Sanction threats
170878.proof.3d 12 04/29/20 21:13Achorn International
000 / Effectiveness of International Sanctions Kerim Can Kavaklı, J. Tyson Chatagnier, and Emre Hatipoğlu
bear under the aegis of an international governmental orga-
nization, whether “smart”sanctions (asset freezes, travel bans,
etc.) were used or threatened, and whether the United States
was among the primary senders in the coalition.
We also run our model on various subsamples of the
data. We first restrict our sample to sanctions involving high-
politics issues, finding that our results remainrobust. We then
analyze different types of sanction threats (import vs. export
restrictions) separately. Our main analysis treats each sanc-
tion as threatening the total trade between countries because
sanctions can (and often do) grow as a crisis evolves, meaning
that neither side can be certain which commodities can con-
tinue to be traded freely. However, we may wonder whether
different economic factors matter when senders threaten im-
port restrictions versus export restrictions. We find that this is
not the case; all of our variables have similar effects across
sanction threat types, which supports the validity of our re-
search design. Next, we divide the data temporally to explore
whether the end of the Cold War and the growing use of smart
sanctions in recent decades have changed the effects of our
variables. Our substantive results hold across both periods,
but the effect of target export variety was stronger during the
Cold War than after.
As with the results in table 1, these findings support our
hypotheses, but they are also tentative. Estimating a standard
logit model raises the possibility of specification error due to
the strategic nature of the sanctioning decision. For this rea-
son, we now turn to an analysis using the nonparametric
approach described above.
Local logit analysis of strategic interaction
The local logit estimator provides separate estimated effects
at each profile of values for our independent variables. As
mentioned above, this means that we cannot report a single
table of coefficient estimates, as we did with the parametric
estimator. Instead, we must select a profile of values and
estimate effects locally, given a predetermined set of smooth-
ing parameters.
25
To illustrate our results, we use the same
variable profiles used in figure 1.
Figure 4 shows the results from our nonparametric esti-
mation. As before, we plot predicted probabilities of sanctions
success as we vary each of the relevant independent vari-
ables from its minimum to maximum values, along with 90%
confidence intervals. The results are encouragingly similar to
those in figure 1.
26
As before, the results for three of our
variables of interest are supportive of our hypotheses. The
results for sender power are virtually unaffected, but we see
25. We choose optimal bandwidth parameters using leave-one-out
cross-validation.
26. These specification here is analogous to model 3, which pools
threatened and imposed sanctions. In the interest of space, we place local
logit results from the differentiation of the two in the appendix. Their
relation to figs. 2 and 3 is similar to the relationship between figs. 1 and 4.
Figure 4. Local logit estimation
170878.proof.3d 13 04/29/20 21:13Achorn International
Volume 82 Number 3 July 2020 / 000
greater levels of concavity for both target power and portfolio
breadth. In the case of target power, the result is even non-
monotonic: at very low levels of target market power, sanc-
tions are (slightly) more likely to succeed as power increases.
However, we caution against reading too much into the result,
as the nonmonotonicity occurs entirely within the 90% con-
fidence region.
The results relating to the portfolio concentration of tar-
get exports once again do not quite conform to our expec-
tations. In this instance, the estimated effects are less linear,
and in the case of democracies, nonsignificant. As mentioned
above, there are several cases in the data in which sanctions
failed against autocracies with heavily concentrated trade
portfolios (largely OPEC members). Meanwhile, there are few
cases of democracies with highly concentrated portfolios.
Despite the fact that we observe a substantial increase in the
predicted probability of success as we vary from the lowest to
highest level of portfolio concentration (from approximately
0.48 to 0.66), our confidence interval is too large to draw any
firm conclusions.
The primary result that follows from both sets of analyses
is that the power to hurt and to bear harm from trade dis-
ruption matters for sanction success. Sanctions are most
successful when senders have more market power over targets
than vice versa and when targets have relatively narrow ex-
port portfolios. Moreover, it appears that states must use their
power in order for it to be effective. The strong effects for
sender and target power arelimited to cases of imposition and
not threat.
CONCLUSION
This article studied the question of when trade sanctions
can be effective, by proposing a framework based on the
Schelling maxim that the power to hurt (and to bear hurt) is a
source of bargaining power. This idea implies that inter-
national crises (economic or military) are decided by each
side’s ability to inflict costs on the adversary and to nullify
the other side’s attempts to impose costs in return. Using
disaggregated trade data, we operationalize the two ways by
which a state can minimize the costs of trade disruption:
switching to alternative partners (external substitution) and
adapting domestic industries to new needs (internal substi-
tution). We also test whether the level of concentration of a
target’s export portfolio has a positive effect on sanction
success, because a more concentrated portfolio means fewer
goods to monitor in order to block most of a target’sex-
ports. Finally, we examine whether these factors matter at
the threat stage, the imposition stage, or both.
Our empirical results mostly bear out our theoretical
expectations. Sender market power increases sanction suc-
cess, while target market power and portfolio diversity are
negatively related to success. Our findings on portfolio con-
centration prove to be conditional and, where significant,
run contrary to our hypothesis. Export portfolio concen-
tration makes autocracies more resilient to sanctions and
has no effect on democracies.
27
We believe that this finding
warrants further study. We also find that these results tend
to hold for the imposition stage but not for the threat stage,
suggesting that states may need to exercise their power to
hurt and deny their opponent the same power in order to
reach an agreement. In addition to using parametric esti-
mators, we implement a flexible nonparametric estimator to
account for the effects of potential model specification issues.
Our results from this latter analysis are broadly consistent
with the parametric results.
Our findings carry interesting policy implications for
sanction design. All else equal, our results suggest that em-
ploying smart sanctions may be more effective against ad-
vanced economies, because these targets can more easily find
alternative export markets and convert domestic industries
to replace banned imports. Another implication for sanction
design is that primary sanctioners should strategically select
their coalition partners to maximize their comparative ad-
vantage in the traded goods. This involves looking beyond
the target’s current trade partners, focusing on whom the
target could turn to, and blocking those channels of exter-
nal substitution.
Finally, this study has several implications for future re-
search. First, our conception of market power and the related
empirical measure provide scholars with an opportunity to
test important hypotheses about the politics of economic
sanctions. For example, within the TIES data, approximately
30% of trade-related sanctions episodes involve multiple
senders. We might suspect that senders with relatively little
market power would be more likely to form coalitions and
would generally form larger coalitions.
28
Our measure would
allow researchers to examine this directly. Second, scholars
have argued that states maintain peaceful relations with their
trade partners and engage in third-party interventions in
order to avoid costly trade disruption (Crescenzi et al. 2011;
Russett and Oneal 2001). If this is true, then states that can
manage the costs of trade disruption should be more war
prone and less willing to intervene in other countries’
disputes. Our measures allow researchers to test these ideas
more directly. Third, we can explore whether a country’s
27. An interesting recent example is the resilience of Qatar, an au-
thoritarian monarchy whose economy is highly reliant on petroleum and
natural gas, in the face of a trade embargo by its neighbors.
28. We thank an anonymous reviewer for pointing this out.
170878.proof.3d 14 04/29/20 21:13Achorn International
000 / Effectiveness of International Sanctions Kerim Can Kavaklı, J. Tyson Chatagnier, and Emre Hatipoğlu
ability to manage the costs of trade sanctions is related to its
domestic political setup. We have already shown that the
effect of one factor—the target’s export concentration—
depends on the target’s democracy level. Future work should
take additional steps in integrating the domestic politics and
international interaction perspectives on sanctions by com-
bining our measures with nuanced typologies of domestic
institutions. Finally, financial sanctions are outside of this
article’s scope, but future work should study how targets of
financial sanctions minimize their pain and develop em-
pirical measures to capture these mechanisms.
ACKNOWLEDGMENTS
We are grateful for feedback on previous versions of this ar-
ticle from Songying Fang, Soo Yeon Kim, Katja Kleinberg,
Ashley Leeds, Elena McLean, and Cliff Morgan, in addition to
the editor and two anonymous reviewers. Previous versions
of this manuscript were presented at the 2014 American Po-
litical Science Association conference, the 2014 European Po-
litical Science Association conference, and the International
Relations Workshop at Rice University.
REFERENCES
Ang, Adrian U-Jin, and Dursun Peksen. 2007. “When Do Economic Sanc-
tions Work? Asymmetric Perceptions, Issue Salience, and Outcomes.”
Political Research Quarterly 60 (1): 135–45.
Balassa, Bela. 1965. “Trade Liberalisation and ‘Revealed’Comparative Ad-
vantage.”Manchester School 33 (2): 99–123.
Bapat, Navin A., and Bo Ram Kwon. 2015. “When Are Sanctions Effective?
A Bargaining and Enforcement Framework.”International Organization
69 (1): 131–62.
Bapat, Navin A., and T. Clifton Morgan. 2009. “Multilateral versus Unilateral
Sanctions Reconsidered: A Test Using New Data.”International Studies
Quarterly 53 (4): 1075–94.
Chapman, Steve. 2014. “Sanctions on Russia Won’t Work.”Chicago Tribune,
March 9.
Chatagnier, J. Tyson. 2014. “Teachingthe Enemy: The Empirical Implications
of Bargaining under Observation.”Journal of Conflict Resolution 58 (6):
673–95.
Chatagnier, J. Tyson, and Kerim Can Kavakli. 2017. “From Economic
Competition to Military Combat: Export Similarity and International
Conflict.”Journal of Conflict Resolution 61 (7): 1510–36.
Chellaney, Brahma. 2017. “China’s Weaponization of Trade.”Project Syndi-
cate, July 26.
Chikuhwa, Jacob Wilson. 2006. Zimbabwe at the Crossroads. Bloomington,
IN: Author House.
Crescenzi, Mark J. C., Kelly M. Kadera, Sara McLaughlin Mitchell, and
Clayton L. Thyne. 2011. “A Supply Side Theory of Mediation.”Interna-
tional Studies Quarterly 55 (4): 1069–94.
Dombey, Daniel, Harvey Morris, and James Blitz. 2009. “Iran Faces Global
Push towards More Sanctions.”Financial Times, December 15.
Early, Bryan R. 2009. “Sleeping with Your Friends’Enemies: An Explanation
of Sanctions-Busting Trade.”International Studies Quarterly 53 (1): 49–
71.
Fan, Jianqing, Nancy E. Heckman, and M. P. Wand. 1995. “Local Polynomial
Kernel Regression for Generalized Linear Models and Quasi-Likelihood
Functions.”Journal of the American Statistical Association 90 (429): 141–
50.
Fattah, Zainab. 2017. “Why Punishing Qatar Could Also Hurt Its Gulf
Rivals.”Bloomberg, July 12.
Feenstra, Robert C., Robert E. Lipsey, Haiyan Deng, Alyson C. Ma, and
Hengyong Mo. 2005. “World Trade Flows, 1962–2000.”NBER Working
paper, National Bureau of Economic Research, Cambridge, MA.
French, Scott. 2017. “Revealed Comparative Advantage: What Is It Good
For?”Journal of International Economics 106:83–103.
Frölich, Markus. 2006. “Non-parametric Regression for Binary Dependent
Variables.”Econometrics Journal 9(3):511–40.
Galtung, Johan. 1967. “On the Effects of International Economic Sanctions,
with Examples from the Case of Rhodesia.”World Politics 19 (3): 378–
416.
General Accounting Office. 1992. Economic Sanctions: Effectiveness as Tools
of Foreign Policy. Washington, DC: US General Accounting Office.
Girit, Selin. 2016. “Turkey Faces Big Losses as Russia Sanctions Bite.”BBC
News, January 2.
Gleditsch, Kristian Skrede. 2002. “Expanded Trade and GDP Data.”Journal
of Conflict Resolution 46 (5): 712–24.
Hausmann, Ricardo, and Cesar A. Hidalgo. 2011. “The Network Structure
of Economic Output.”Journal of Economic Growth 16 (4): 309–42.
Heckman, James J. 1979. “Sample Selection Bias as a Specification Error.”
Econometrica: Journal of the Econometric Society 47 (1): 153–61.
Hidalgo, César A., Bailey Klinger, Albert-László Barabási, and Ricardo
Hausmann. 2007. “The Product Space Conditions the Development of
Nations.”Science 317 (5837): 482–87.
Hovi, Jon, Robert Huseby, and Detlef F. Sprinz. 2005. “When Do (Imposed)
Economic Sanctions Work?”World Politics 57 (4): 479–99.
Hufbauer, Gary Clyde, Jeffrey J. Schott, Kimberly Ann Elliott, and Oegg
Barbara. 2007. Economic Sanctions Reconsidered: History and Current
Policy. 3rd ed., vol. 3. Washington, DC: Peterson Institute.
James, Alan. 1990. Peacekeeping in International Politics. New York: Palgrave
Macmillan.
Jentleson, Bruce W. 2000. “Economic Sanctions and Post–Cold War
Conflicts: Challenges for Theory and Conflict.”In Paul C. Stern and
Daniel Druckman, eds., International Conflict Resolution after the Cold
War. Washington, DC: National Academy, 123–77.
Kaempfer, William H., and Anton D. Lowenberg. 1988. “The Theory of In-
ternational Economic Sanctions: A Public Choice Approach.”American
Economic Review 78 (4): 786–93.
Kaempfer, William H., and Anton D. Lowenberg. 2007. “The Political
Economy of Economic Sanctions.”In Todd Sandler and Keith Hartley,
eds., Handbook of Defense Economics. Amsterdam: Elsevier, 868–911.
Kenkel, Brenton, and Curtis S. Signorino. 2012. “An Alternative Solution to
the Heckman Selection Problem: Selection Bias as Functional Form
Misspecification.”Presented at the Annual Meeting of the Midwest
Political Science Association, Chicago.
Kertzer, Joshua D. 2016. Resolve in International Politics. Princeton, NJ:
Princeton University Press.
Kobayashi, Yoshihaur. 2017. “Economic Sanction as Foreign Policy.”
Oxford Research Encyclopedia of Politics. https://oxfordre.com/politics
/view/10.1093/acrefore/9780190228637.001.0001/acrefore-9780190228637
-e-477.
Lektzian, David, and Mark Souva. 2003. “The Economic Peace between
Democracies: Economic Sanctions and Domestic Institutions.”Journal of
Peace Research 40 (6): 641–60.
Lektzian, David, and Mark Souva. 2007. “An Institutional Theory of Sanc-
tions Onset and Success.”Journal of Conflict Resolution 51 (6): 848–71.
Li, Quan, and Rafael Reuveny. 2011. “Does Trade Prevent or Promote In-
terstate Conflict Initiation?”Journal of Peace Research 48 (4): 437–53.
170878.proof.3d 15 04/29/20 21:13Achorn International
Volume 82 Number 3 July 2020 / 000
Marshall, Monty G., and Keith Jaggers. 2002. Polity IV Project: Political
Regime Characteristics and Transitions, 1800–2007.http://www.systemic
peace.org/polity/polity4.htm.
McLean, Elena V., and TaeheeWhang. 2010. “Friends or Foes? Major Trading
Partners and the Success of Economic Sanctions.”International Studies
Quarterly 54 (2): 427–47.
Morgan, T. Clifton, Navin Bapat, and Yoshi Kobayashi. 2014. “The Threat
and Imposition of Economic Sanctions, 1971–2000.”Conflict Manage-
ment and Peace Science 31 (5): 541–58.
Morgan, T. Clifton, and Valerie L. Schwebach. 1997. “Fools Suffer Gladly:
The Use of Economic Sanctions in International Crises.”International
Studies Quarterly 41 (1): 27–50.
Nooruddin, Irfan. 2002. “Modeling Selection Bias in Studies of Sanctions
Efficacy.”International Interactions 28 (1): 59–75.
Paarlberg, Robert L. 1980. “Lessons of the Grain Embargo.”Foreign Affairs
59 (1): 144–62.
Peksen, Dursun, and Timothy M. Peterson. 2016. “Sanctions and Alternate
Markets: How Trade and Alliances Affect the Onset of Economic Coer-
cion.”Political Research Quarterly 69 (1): 4–16.
Reuters. 2014. “Germany Warns of ‘Massive’Political, Economic Damage if
Russia Stays on Course in Ukraine.”ABC News Australia, March 13.
Russett, Bruce, and John Oneal. 2001. Triangulating Peace: Democracy, In-
terdependence, and International Organizations. New York: Norton.
Saunders, Paul J. 2014. “When Sanctions Lead to War.”New York Times,
August 21.
Saviotti, Pier Paolo, and Koen Frenken. 2008. “Export Variety and the
Economic Performance of Countries.”Journal of Evolutionary Economics
18 (2): 201–18.
Schelling, Thomas C. 1966. Arms and Influence. New Haven, CT: Yale Uni-
versity Press.
Sher, Gerson. 2014. “Russia, Ukraine and Sanctions:A Double-Edged Sword.”
International Business Times, March 6.
Signorino, Curtis S., and Kuzey Yilmaz. 2003. “Strategic Misspecification
in Regression Models.”American Journal of Political Science 47 (3): 551–
66.
Singer, J. David. 1988. “Reconstructing the Correlates of War Dataset on
Material Capabilities of States, 1816–1985.”International Interactions
14 (2): 115–32.
Slantchev, Branislav L. 2003. “The Power to Hurt: Costly Conflict with
Completely Informed States.”American Political Science Review 97 (1):
123–33.
Stojek, Szymon M., and Mwita Chacha. 2015. “Adding Trade to the Equation:
Multilevel Modeling of Biased Civil War Interventions.”Journal of Peace
Research 52 (2): 228–42.
Van de Graaf, Thijs. 2013. “The ‘Oil Weapon’Reversed? Sanctions
against Iran and US-EU Structural Power.”Middle East Policy 20 (3):
145–63.
Whang, Taehee. 2010. “Structural Estimation of Economic Sanctions:
From Initiation to Outcomes.”Journal of Peace Research 47 (5): 561–
73.
Whang, Taehee. 2011. “Playing to the Home Crowd? Symbolic Use of Eco-
nomic Sanctions in the United States.”International Studies Quarterly
55 (3): 787–801.
Whang, Taehee, Elena V. McLean, and Douglas W. Kuberski. 2013. “Coer-
cion, Information, and the Success of Sanction Threats.”American
Journal of Political Science 57 (1): 65–81.
170878.proof.3d 16 04/29/20 21:13Achorn International
000 / Effectiveness of International Sanctions Kerim Can Kavaklı, J. Tyson Chatagnier, and Emre Hatipoğlu
QUERIES TO THE AUTHOR
Q1. Au: Your article has been edited for grammar, clarity, consistency, and conformity to journal style. Please read the article to
make sure that your meaning has been retained. Note that we may be unable to make revisions that conflict with journal style or
create grammatical problems. Thank you.
Q2. Au: Table 1: Can an additional, third, decimal precision be added to the second column's .10 values to better match to the
.004 difference reported? Similarly, can the .95 difference in the final column be revised to .96 (or one of the values above it be
revised so that the difference between them is .95)?
170878.proof.3d 17 04/29/20 21:13Achorn International