Kicking the Nuclear Ladder The Effects of U S Economic
Statecraft on Nuclear Reversal
This essay examines whether U.S. economic statecraft is effective in dissuading nuclear
aspirants from going nuclear. Conventional wisdom holds that nuclear nonproliferation is
not strongly affected by economic sanctions. This article posits, however, that U.S.
economic statecraft has exhibited a greater effectiveness than previously explained in the
literature. This study finds that when U.S. policy objective is to extract the perfect
compliance with nonproliferation from a nuclear aspiring state, U.S. negative sanctions have
no significant impact on the state’s nuclear reversal. However, U.S. economic statecraft is
moderately successful in extracting suboptimal policy concessions of nonproliferation-
“nuclear power without nuclear weapons proliferation.” The empirical evidence uses a
statistical analysis of 18 countries from 1970 to 2004. Analysis shows that U.S. sanctions
through international institutions and U.S. positive inducements (e.g., foreign assistance)
are effective at least in driving nuclear aspirants into remaining in a nuclear latency
status.
2.1 Introduction
On July 14, 2015, Iran and the P5+1 nations (i.e., the United States, the United Kingdom,
France, China, and Russia, plus Germany) reached a momentous agreement that limited
Iran’s nuclear development in exchange for the lifting of economic sanctions. Media
commentators and policy analysts consider this landmark Iranian nuclear deal as a victory
for international sanctions (Cassidy 2015; Rosofsky 2015). According to the deal called the
7
“Joint Comprehensive Plan of Action (JCPOA),” Tehran is requested to significantly reduce its
capacity to develop nuclear bombs for more than a decade by eliminating 98 percent of its
enriched uranium stockpile. Why did Iran decide to limit its nuclear weapons development?
Did U.S. economic statecraft force/induce Iran and/or other (previous) nuclear proliferators
to make decisions to reverse their nuclear weapons development?
The existing research on nuclear proliferation primarily concentrates on why states go
nuclear. Since Sagan (1996) began his theoretical and empirical work on nuclear
proliferation, many international relations scholars have conducted quantitative
examinations to determine the causes of nuclear proliferation (e.g., Bleek and Lorber 2014;
Brown and Kaplow 2014; Fuhrmann 2009; Fuhrmann and Horowitz 2015; Jo and Gartzke
2007; Kroenig 2009; Monteiro and Debs 2014; Reiter 2014; Singh and Way 2004; Way and
Weeks 2014). Proliferation studies emphasize a state’s motivation or intention to pursue
nuclear development specifically by looking into domestic politics (Hymans 2011; Way and
Weeks 2014), the nuclear umbrella or security guarantees (Bleek and Lorber 2014; Monteiro
and Debs 2014; Reiter 2014), a state’s openness to the international market (Solingen 2009),
a state’s ratification of the nonproliferation treaty (Potter 2010), and political leaders’ prior
experiences (Fuhrmann and Horowitz 2015). Recent scholarship, though, takes into account
the supply-side approach to nuclear proliferation. A nuclear aspiring state is more likely to
seek nuclear weapons development when it has outside groups that would supply nuclear
weapons materials or nuclear technology (e.g., Brown and Kaplow 2014; Fuhrmann 2009;
Kroenig 2009). Despite the increase in quantitative studies on determinants of nuclear
proliferation, much less attention has been paid to reversal behaviors of nuclear
proliferators. In particular, the effect of U.S. economic statecraft on a state’s nuclear rollback
have not, with a few exceptions (e.g., Early 2012; Miller 2014), been examined systemically
or thoroughly.
Since the 1970s, the United States has taken great efforts to prevent the proliferation of
nuclear weapons, seeking close cooperation from other countries in the international
community.The United States has implemented several foreign policy instruments to
strengthen nonproliferation efforts, using for instance economic statecraft-including
positive inducements and negative sanctions. Here, this paper pays attention to evaluating
the role of U.S. economic statecraft in counterproliferation. I will ask whether U.S. positive
inducements and/or negative sanctions have worked at persuading or forcing a state to
reverse its nuclear weapons program that it had already initiated and/or pursued. The answer
to this puzzle produces a systematic and quantitative examination of the impacts of U.S.
economic statecraft on nuclear reversal-an area little investigated by the quantitative
studies of nuclear proliferation.
Relying on updated data of nuclear proliferation from 1970 to 2004, this paper tests
quantitatively whether U.S. economic statecraft has either forced or induced nuclear
proliferators to rollback their nuclear weapons development. The study assumes that U.S.
economic statecraft brings political consequences to target regimes. U.S. economic
statecraft is likely to generate political externalities, which may threaten leaders’ political
survival in target countries. Political leaders in a sanctioned country are constrained by their
domestic political audiences or members of a winning coalition when they are targeted by
U.S. economic statecraft. Leaders’ political careers may be in danger when their policy
performance (e.g., economic hardship generated by foreign sanctions) is poor. This paper
9
contends that U.S. economic statecraft affects, in a target state, the domestic
constituency’s assessment of its leaders’ policy competence. Political and/economic
distributional consequences of both positive incentives and negative sanctions may impact
a sanctioned state’s decisions regarding its nuclear weapons development. This study finds
that U.S. economic statecraft results in a greater probability of nuclear latency status, a
suboptimal outcome of nuclear reversals for a sanctioning state such as the United States.
Both U.S. economic aid and negative sanctions do little in the way of leading to a nuclear
aspirant’s perfect compliance with nonproliferation-stable non-nuclear status. But when
nuclear development paths are disaggregated into three outcomes-i.e., stable non-nuclear
status, nuclear latency status, and nuclear pursuit/acquisition-, U.S. economic statecraft
has a substantial impact on a target’s decision to keep the latent capability. More U.S. foreign
aid and negative sanctions through international organizations make a nuclear aspiring state
move down from nuclear pursuit to nuclear latency status, though not to perfect
nonproliferation. The main contribution of this paper to the existing literature is in revealing
the suboptimal impacts of U.S. economic statecraft on nuclear nonproliferation, bringing
nuclear latency status into the paths of nuclear development. Specifically, I argue that while
nuclear aspiring states tend to be reluctant to suspend the ongoing nuclear development,
U.S. sanctions through international organizations induce them to reach a less undesirable
nuclear deal with the U.S.-‘nuclear power without nuclear bombs.’
This chapter assesses how earlier research conceptualizes nuclear weapons
proliferation and identifies the driving forces that induce states to go nuclear. It then
examines some conceptual and theoretical issues and builds a theoretical argument about
the role U.S. economic statecraft plays in leading to nuclear reversal of proliferators. The
chapter then proposes testable hypotheses and lays out a research design. Finally, the
results of logistic and multinomial logistic regressions are presented and a
discussion/conclusion put forward.
2.2 Existing Research on Nuclear (Non-)Proliferation
Existing literature on nuclear proliferation primarily focuses on causes of the nuclear
weapons proliferation. Scholars find causal factors of nuclear proliferation through
qualitative works and test the statistical correlations using the large-N quantitative research.
Here I discuss the concept of nuclear proliferation and briefly mention the current studies
on nuclear proliferation because it helps define nuclear reversal and understand plausible
explanations for nuclear reversal.
2.2.1 The Concept of Nuclear Proliferation
Nuclear proliferation can be defined as “the spread of nuclear weapons” (Sagan and Waltz
2003). Despite this simple definition, the existing and inadequate concept of nuclear
proliferation may prevent scholars from carrying out rigorous studies (Ogilvie-White 1996).
At the same time, few quantitative studies have conceptually defined or measured a
relatively-overlooked feature of nuclear proliferation-nuclear rollback as the opposite
direction of the spread of nuclear weapons. Some critics suggest that the current research
on nuclear proliferation tends to focus too much on developing bombs by relatively
11
disregarding the aspect of nuclear restraints or reversal (Levite 2002; Solingen 1994). At the
beginning of his article, Levite (2002) described the following:
A serious gap exists in scholarly understanding of nuclear proliferation. The gap
derives from inadequate attention to the phenomena of nuclear reversal and
nuclear restraint as well as insufficient awareness of the biases and limitations
inherent in the empirical data employed to study proliferation.(59)
How do the empirical studies conceptualize nuclear proliferation? What proliferation
means is closely related to whether it is a static dichotomy or a dynamic continuum. Some
focus on one dimension of proliferation related to the extent of proliferation: stages or
phases of nuclear proliferation. For example, Singh and Way (2004, 861) explicitly
emphasized proliferation as “continuum instead of a dichotomy,” defining “degrees of
nuclearness,” which consist of nuclear acquisition, the pursuit of nuclear weapons
programs, the exploration of the nuclear weapons option, and no interest in nuclear arms.
By the same token, Jo and Gartzke (2007, 167-168) disaggregated nuclear proliferation into
“the presence of a nuclear weapons production program” and “nuclear weapons
possession.” Most quantitative approaches to nuclear proliferation choose either type of
nuclear proliferation paths as the outcome of research interests. However, few quantitative
studies have evaluated the causes of nuclear reversal as another dimension of nuclear
weapons policy. In order to obtain a complete understanding of nuclear proliferation,
scholarship on nuclear proliferation should be interested in conditions under which nuclear
proliferators stop and/or reverse their ongoing nuclear weapons programs.
2.2.2 Explanations of Nuclear Weapons Proliferation
Why do countries go nuclear? To answer the question, many scholars suggest a variety of
explanations. Sagan (1996) first presented the systematic analysis of nuclear proliferation by
offering the three models of security, domestic politics, and norms. First, the security model
posits that nuclear weapons development is associated with the maximization of national
security against security threats. Second, the domestic politics model stresses the political
interests of actors within the state in regard to nuclear pursuit. Lastly, the norms model refers
to norms and shared beliefs about whether nuclear pursuit is “legitimate” and “appropriate.”
Since Sagan (1996) suggested his three theoretical models of nuclear proliferation were
suggested, many international relations scholars have conducted systematic and
quantitative examinations of causes of proliferation to find causal factors that drive nuclear
weapons proliferation.
This chapter presents a review of the existing literature on nuclear proliferation as it
pertains demand and supply frameworks (Sagan 2011). The demand-side explanation refers
to the motivation or intention to acquire nuclear weapons while the supply-side one is
associated with the available resources for nuclear weapons development. The demand-
side factors include, for instance, security threats, the security guarantee/ defensive alliance
from a nuclear-armed state, and domestic political constraints. The supply-side factors
include economic conditions, nuclear technology, and foreign nuclear
cooperation/assistance.
The Demand-Side Explanation
13
This section lays out how the demand-side conditions influence states’ nuclear
behaviors. In particular, how do security threats and domestic political concerns influence
states’ decisions regarding going nuclear?
First, states attempt to develop or obtain nuclear weapons for national security
purposes. A large proportion of existing proliferation literature relies on the security model
(Bowen and Kidd 2004; Ogilvie-White 1996; Paul 2000; Sagan 1996). States seek nuclear
bombs when they face a significant conventional or nuclear threat to their security, which
cannot be met by existing conventional forces (Sagan 1996). Such a security-based
explanation for nuclear arms has been strongly supported by the realist/neorealist tradition
in international relations. For realists/neorealists, the nuclear weapons possession is
regarded as the rational behavior to protect national interests-state survival-against
potential threats in the international system (Jo and Gartzke 2007; Ogilvie-White 1996; Singh
and Way 2004). Any state that seeks to maintain its national security must come up with a
counterbalance to any rival state that develops nuclear weapons; the counterbalance
involves gaining access to a nuclear deterrent. In this respect, the acquisition of nuclear
weapons plays a role in deterring the adversaries. Within the international system, the
serious concerns about national survival lead states to maximize power and to try to become
the most powerful nation in the regional or world (Mearsheimer 2003). In this context, states
seek nuclear weapons to maximize national security and deter foreign attacks.
Second, forming a defense pact with a nuclear weapons state or with the dominant state
reduces a state’s motivation to acquire nuclear arms (Bleek and Lorber 2014; Jo and Gartzke
2007; Singh and Way 2004). This is not only because a state does not have an incentive to
possess nuclear weapons due to nuclear umbrella but also because the powerful nuclear
weapons state discourages a state from developing a nuclear weapons program. For
example, Singh and Way (2004, 863) argued that “a credible security guarantee from a
powerful state can dull the desire for nuclear weapons.” Jo and Gartzke (2007, 170) claimed
that “states with security commitments from patrons with nuclear weapons may be less
likely to proliferate.” In contrast, Reiter (2014) found empirical evidence that neither alliances
nor foreign-deployed troops reduces an alliance-partner’s motive to develop nuclear
weapons. However, the variable of foreign-deployed nuclear weapons is the strong predictor
of decreasing the proliferation intention. A nuclear umbrella can provide a non-nuclear
alliance partner with the extended nuclear deterrence. In a similar logic, Gerzhoy (2015)
developed the logic of alliance coercion and argued, based on it, that a nuclear aspiring ally
is compelled to withdraw its nuclear weapons program due to its nuclear patron’s threats of
military abandonment. However, even if in history alliance restraints on a state’s nuclear
weapons program is substantial, it is just effective on a state within defensive alliances and
during the Cold War.After the end of Cold War, most nuclear arms seekers were states
without alliance and so-called rogue states-Iran and North Korea, for instance. Alliance
factor may have a limited influence on nuclear proliferation in regard to country
characteristics and changes in international structure.
Third, domestic-political explanations assume that the decision to go nuclear depends
upon the strategic games over policy makings among domestic political actors (Jo and
Gartzke 2007; Paul 2000; Sagan 1996; Singh and Way 2004). Domestic political factors do
not necessarily affect a state’s nuclear ambition in one direction. Domestic politics may
encourage or discourage a state’s nuclear arms development (Sagan 1996, 63). For instance,
15
according to Gray (1999), nuclear weapons programs are essentially based on their domestic
process. Sathasivam (2003) argued that, in the case of Pakistan, the domestic power interest
group forced the Pakistani government to build and test nuclear arms in response to India’s
acquisition of nuclear weapons. In the case of Japan, the domestic constitution, legislation,
and non-nuclear principles restrain Japan from building nuclear weapons (Hughes 2007). In
particular, three non-nuclear principles include ”not manufacturing, possessing, or
importing nuclear weapons” (Hughes 2007, 85).
The Supply-Side Explanation
The supply-side model is associated with the resources necessary to obtain nuclear
weapons. It includes such factors economic/industrial capacity, nuclear technology, and
foreign nuclear assistance. Recent studies have shown the importance of nuclear diffusion
through the outside assistance (e.g., Kroenig 2009) and proliferation rings or networks (e.g.,
Chestnut 2007).
First, economic development simplifies a nation’s acquisition of nuclear weapons by
providing sufficient resources (Jo and Gartzke 2007; Singh and Way 2004). For instance,
Singh and Way (2004, 862) argued that states “may achieve the capability to assemble
nuclear weapons by an explicit intentional effort or as an implicit by-product of economic
and industrial development.”
Second, nuclear technology can provide states with the opportunity to build nuclear
weapons (Jo and Gartzke 2007; Singh and Way 2004). According to Jo and Gartzke (2007,
169), nuclear technology is not considered easily obtainable. For instance, Hymans (2012,
8) discussed the common claims about the role of nuclear technology in the sense that
“states that have more prior experience with nuclear technology might be able to make the
bomb more quickly after deciding to do so.” Nonetheless, nuclear technology is required to
build nuclear weapons, regardless of how it is obtained-through national research or outside
assistance and nuclear networks.
Third, nuclear assistance refers to the international spread of nuclear arms through
providing nuclear transfers to non-nuclear powers and nuclear proliferation networks (Braun
and Chyba 2004; Chestnut 2007; Kroenig 2009; Montgomery 2005). Most of the proliferation
literature depends on the demand-side approach. Kroenig (2009), however, by criticizing the
current overdependence on the demand-side explanation, showed the importance of the
supply-side model of nuclear proliferation. He argued that the patterns of sensitive nuclear
assistance to non-nuclear states may be explained by the strategic characteristics of the
suppliers. Fuhrmann (2009) also found that civilian nuclear assistance is more likely to lead
to some features of nuclear proliferation such as the initiation and acquisition of nuclear
weapons due to its lowering nuclear technology barriers. According to Braun and Chyba
(2004), proliferation rings-e.g., a Pakistani scientist Khan’s network-lead to nuclear
technology transfers and challenge global nonproliferation efforts. Thus, foreign nuclear
assistance can be one of the main determinants of nuclear weapons proliferation.
In the next section, I will present a theoretical discussion about possible impacts of U.S.
economic statecraft on the nuclear reversal commitment by looking into U.S. positive
inducements and negative economic sanctions.
17
2.3 Theory: U.S. Economic Statecraft and Nuclear Reversal
A nuclear aspiring state considers its distributional effects of international pressures on
its regime survival in regard to the policy concession of nonproliferation. Since embarking on
a nuclear weapons program requires strong support from key political actors or ruling elites,
state leaders rarely reverse their nuclear paths due to anticipated political costs compared
to benefits arising from the nuclear policy change. Nonetheless, since 1970s, key members
of the international nonproliferation community have experienced some cases of nuclear
rollbacks. What then drives nuclear reversal? Does U.S. economic statecraft induce nuclear
aspirants to give up ongoing nuclear development programs? First, let us consider how
nuclear proliferation scholars define nuclear reversal. Levite (2002, 67) stated that nuclear
reversal is
a governmental decision to slow or stop altogether an officially sanctioned
nuclear weapons program. At the core this definition is the distinction between
states that have launched (indigenously or with external assistance) a nuclear
weapons program and then abandoned it and those that never had such a
program in the first place. Nuclear reversal excludes both termination of
unauthorized nuclear weapons-related activity within a government and private-
sector research and development in a nuclear weapons-related field (e.g.,
nuclear fuel-cycle technologies) if the latter was not formally pursued as part of
an effort either to create a bomb or at least to acquire standby status.
This definition is commonly mentioned in existing literature on nuclear (non-)proliferation
studies, which can be appropriately linked with my theoretical argument. However, the
dichotomous classification of nuclear proliferation/reversal may overlook an important path
of nuclear proliferation. So in theoretical and practical senses nuclear reversal might be
disaggregated into a few outcomes. For example, Lodgaard (2010, 116) argued that nuclear
proliferation and rollback include “rejection, hedging, restraint, active pursuit and
acquisition” among which the former three terms are considered nuclear rollback. Similarly,
recent studies on nuclear proliferation pay attention to the feature of nuclear latency
(Fuhrmann and Tkach 2015; Sagan 2010). They contend that current scholarship of nuclear
proliferation relatively overlook the importance of nuclear latency in accounting for the
causes of nuclear weapons development. Meyer (1986) also posited that a state is said to
“have a latent capacity when it has sufficient technical, industrial, material, and financial
resources to support a wholly indigenous weapons program.” Fuhrmann and Tkach (2015, 2)
built the dataset focusing on “the development of enrichment and reprocessing facilities”
which “provide countries with the ability to produce fissile material - weapon-grade highly
enriched uranium or plutonium.” Nuclear reversal outcomes may include “stable non-
nuclear status” or states with “a certain preparedness for going nuclear if changing
circumstances so suggest” (Lodgaard 2010, 115). In accordance with the significance of a
state’s nuclear latent capabilities in nuclear proliferation, nuclear development paths are
here disaggregated into three outcomes: stable non-nuclear status, nuclear latency status,
and nuclear pursuit/acquisition.
Before moving on to the theoretical argument, I first delineate the theoretical background
of the political survival approach to international relations to identify the role of U.S.
economic statecraft on nuclear reversal. First, I assume that state leaders make foreign
policy choices with the purpose of maximizing the likelihood of their political survival (Bueno
de Mesquita et al. 2003; Schultz 2001). A variety of research groups using the domestic level
(of analysis) in international relations posit that political leaders concern about their
19
domestic political survival as the sine qua non of foreign policy decisions (Bueno de
Mesquita et al. 2003; Fearon 1994; Mintz 1993). Fearon (1994), for instance, argued that
political leaders generate domestic audience costs that they pay when they back down after
having made a public threat during the international crisis. Smith (1998) emphasized the
domestic constituencies’ evaluation of their leaders’ policy performance. In particular,
national leaders revealed to have low competence in foreign policy tasks are more likely to
be removed through re-election, coup, or assassination. Leaders who want to retain power
in office should adopt policies that help them achieve that goal. By the same token, leaders
in the nuclear aspiring state tend to reveal a noncompensatory decision rule for eliminating
in the first place policy alternatives that might harm their political survival (Mintz 1993). Thus,
political leaders hardly roll back a nuclear program in progress, given that it typically requires
an enormous financial and technological investment that should have a broad base of
support among key domestic constituencies (Fuhrmann and Sechser 2014).
Since the domestic political constraints make leaders in the nuclear aspirant reluctant
to withdraw the existing (and/or ongoing) nuclear weapons (program), major determinants of
nuclear proliferation may play a minor role, at most, of inducing nuclear reversal. In this case,
international factors such as major world powers’ influence may affect a state’s domestic
political/policy decisions (e.g., Gourevitch 1978; Pevehouse 2002). American statecraft is
one of the critical factors to influence the behaviors of other countries. The United States,
after all, is a country that possesses sufficient resources to pursue its national interest
throughout the world (Ross 2007). Therefore, in explaining nuclear reversal outcomes, this
study focuses on the role of the United States and its economic statecraft, in particular. The
argument put forward is that American economic statecraft influences nuclear rollback
behaviors of nuclear proliferators that have already pursued and even acquired nuclear
bombs. The next section provides possible theoretical explanations regarding the impacts of
American negative sanctions and positive inducements on nuclear proliferating states’
reversal behaviors.
2.3.1 The Effects of U.S. Negative Sanctions on Nuclear Reversal
The United States has employed negative sanctions as a foreign policy instrument of
coercive diplomacy to maintain international stability and/or resolve international crises.
Sanctions are used to “presuppose the sender country’s willingness to interfere in the
decision-making process of another sovereign government, but in a measured way that
supplements diplomatic reproach without the immediate introduction of military
force”(Hufbauer et al. 2007, 5). If the United States imposes sanctions against the target
state, that state’s political leaders suffer domestic audience costs that increase as
sanctions produce a negative externality such as international isolation, economic
depression/hardship, and higher inflation. In particular, economic sanctions threaten
incumbent leaders’ political survival because they are likely to generate political costs for
the targeted leaders through economic damage, political disintegration among domestic
societal groups, and even political protests or violence (Allen 2008b; Marinov 2005;
McGillivray and Stam 2004). The domestic political costs from sanctions can lead a targeted
government to make a policy concession to the sanctioning state. But, the sanctioning state
can extract the target’s compliance only if sanction costs for noncompliance outweigh the
21
benefits for noncompliance. In nuclear rollback decisions, target leaders should calculate
the costs and benefits of reversing the nuclear development that has already been pursued.
However, sanctions rarely extract policy concessions from the target nation. The target
state
is likely to resist the demands of the sanctioning state when it has an issue of high salience,
such as nuclear weapons development, that involves the state’s national security concerns
(Adrian and Peksen 2007). Since nuclear weapons programs are financially expensive and
politically important to a target state, the state’s leaders face high audience costs if they
back down from going nuclear (Gartzke and Jo 2009). These costs force target leaders to
resist making the concessions to the sanctioning state’s demands for nonproliferation.
Instead, the targeted nuclear aspirant seeks to avoid political and economic damage from
imposed sanctions. At the same time, countries facing critical security threats will seek to
build their own nuclear bombs if they have no credible security guarantees that a nuclear
states will provide, according to the security model of nuclear proliferation. Thus, U.S.
international sanctions decrease the likelihood that the target government will reverse the
nuclear weapons development due to its expectation of a potential conflict with the
sanctioning state and its coalition countries.
A nuclear aspirant targeted by negative sanctions would expect potential military conflict
with the U.S. and its coalition states. This conflict expectation may lead the target state to
accelerate its nuclear weapons program rather than suspend it. In this sense, U.S. sanction
imposition is indirectly linked with the external security threats to which the target state may
respond by building nuclear bombs or strengthening its military capabilities. Negative
sanctions generate the costly
target state. Economic blockade and trade disruptions make the target economy suffer negative externalities,
such as a higher rate of inflation and quite heavy job losses. The target state, however, searches for alternative
trading partners to reduce negative distributional effects of sanctions. Under this circumstance, some third-
party countries play a sanction-busting role of spoiling sanction effectiveness by offering the opportunistic
trade relations to and/or investing in targeted countries (Drezner 2000; Early 2009, 2011; Lektzian and Biglaiser
2013; McLean and Whang 2010). In addition to the busting behaviors of third-party countries, target leaders
use political manipulation to diminish the negative impacts of U.S. economic sanctions on their economy. It is
widely accepted in the sanction literature that negative sanctions cause a “rally-round-the-flag effect” through
which a targeted state’s leaders generate a nationalistic sentiment within the society, which would make
sanctions less successful (Verdier and Woo 2011). The sanctioned governments are likely to frame the U.S.
demands for nuclear withdrawal and its sanction imposition as foreign threats to the national sovereignty
(Drezner 2001).
signaling effects so the sender is likely to initiate military conflicts against a target after
sanction imposition (Drezner 1998). When a target’s leaders face sanction imposition of the
sender, they are likely to realize the sender’s resolve because of its costly signaling role. For
instance, Lektzian and Souva (2007, 416) argued that economic sanctions are positively
associated with the use of military force by functioning “a costly signal of a state’s
commitment to have a dispute resolved.” Drezner (1998) also discussed the notion that
sanctioning countries are likely to impose sanctions as an alternative when future conflict is
anticipated. The Bush administration began putting more pressure for the North Korean
regime for its nuclear weapons development than did its predecessor the Clinton
administration (Pardo 2014). With the U.S. slow commitments to the 1994 Agreed Framework
and its more hawkish policies toward North Korea, the Pyongyang government perceived the
U.S. tough sanctions and other measures as significant threats to its regime survival. Yet
rather than give a concession, North Korea accelerated its nuclear weapons development
and even quickly began nuclear testing (Byun and Snyder 2007). Therefore, North Korea,
perceiving itself as a U.S. adversary, was strongly reluctant to make any concessions to the
U.S. and its allies (Drezner 2001).
23
The discussion described above may lead one to believe that a target state’s chances of
terminating its nuclear weapons program is low. Political leaders in the sanctioned country
expect potential conflicts with the U.S. and its coalition countries when they become a target
of U.S. sanctions. Thus, when the United States imposes negative sanctions against a
nuclear aspiring state, the target state will be likely to resist America’s nonproliferation
demands, even trying to accelerate nuclear weapons development.
Hypothesis 2.1 Nuclear aspiring states targeted by U.S. negative sanctions have a lower
likelihood of reversing their nuclear pursuits.
2.3.2 The Effects of U.S. Multilateral IO Sanctions on Nuclear Reversal
Another factor that might influence nuclear reversal behavior of the target state is U.S.
sanction imposition through international institutions. The United States has implemented
both multilateral sanctions and sanctions through international organizations. This section
includes as another main independent variable U.S. sanctions through international
institutions rather than multilateral sanctions. As discussed above, negative sanctions are
considered ineffective in part because of third-parties’ sanction-busting and backsliding
behaviors. Scholars of economic sanctions argue that multilateral sanctions, especially
through international institutions, are more effective in extracting policy concessions from
the target state than are other sanctions (Bapat and Morgan 2009; Bapat et al. 2013; Drezner
2000; Drury 1998). Sanctions through international organizations should resolve
enforcement problems such as sanction-busting behaviors of third parties to a considerable
extent. Drezner (2000), for instance, contends that “sanctions with reasonably high levels of
international cooperation should impose greater costs on the target country because of the
inability to find alternative markets and suppliers.” International institutions possess
enforcement powers to prevent other countries from defecting and to reduce the probability
of backsliding.
Thus, political leaders in a target state need to recalculate the anticipated benefits for
pursuing nuclear weapons programs and costs of U.S. sanction through international
organizations. If they choose to pursue nuclear weapons development, a target state’
leaders may gain political benefits. Leaders in a target state improves its military power
deterring potential security threats from attacking it and strategically symbolizes nuclear
bombs as a national pride.
However, when an international organization is in the U.S. sanction coalition, a target
country cannot help but to encounter potential threats from overseas as well as
domestically. A target state faces domestically much greater economic and political costs
of sanctions in which the U.S. and other members in an international organization would
support for the U.S.-led sanctions against nuclear proliferating state. In this case, it is hard
for a target state to find alternative trading routes and investors in order to avoid negative
externalities of sanctions. Resisting nonproliferation demands from the U.S. and the
international community is not a good choice for a sanctioned nuclear aspirant. Economic
damages from sanctions may be sufficient enough to put the political survival of target
leaders in danger.
For example, in response to North Korea’s nuclear and ballistic missile tests on January 6,
2016, the United States recently led the tightened sanctions through the United Nations
Security
25
Council against Pyongyang and its government officials/agencies and companies (Morello
2016)
It was said that the sanctions would be “more robust than any imposed in decades” because
the UN sanctions require member states to inspect all cargo going to or coming from North
Korea to ensure it does not contain anything that would further the country’s nuclear or
missile programs.”
The US-led multilateral sanctions through international organization (esp. the UN) may
improve the effectiveness of sanctions to extract some policy concessions because the
United States is one of the only countries in the world with an extensive defense alliance
system. Using intra alliance politics, the U.S. could induce its third party allies to join
sanctions agains the target countries (Early 2012). When the U.S. forms asymmetric alliance
with less powerful countries, it can receive those countries’ support for US-led multilateral
sanctions. There is a tradeoff of security and policy concession between the U.S. and its
weaker allies (Morrow 1991). For example, the recent nuclear deal between Iran and the US
and its coalition states shows that Iranian sanctions primarily implemented by the U.S. were
complemented by the strong cooperation of the EU countries and member states of the UN
in 2006, forcing Iran to change nuclear policy (Borhani 2015). In addition, South Korea, a U.S.
ally, joined the US-led Iranian sanctions under the pressure of the US-Korean alliance, adding
“the names of 102 Iranian firms and 24 people to the blacklist of those with whom South
Koreans cannot do business and also promised to inspect cargo from Iran more diligently
and hold back on investment in oil and gas enterprises.” Like South Korea, the EU, Canada
and Japan have also joined the U.S. sanctions against Iran (Kirk 2010).
Bapat and Morgan (2009) also found that “multilateral sanctions do appear to work more
frequently than do unilateral sanctions” because they have “the potential to create more
coercive power than unilateral sanctions.” In their recent study, Early and Spice (2015)
argued that economic sanctions through smaller international institutions can extract
secondary sanctioners’ deeper commitments not to spoil sanction effectiveness than those
through larger international institutions. This indicates that despite the variation in sanction
effectiveness between the different size of international institutions, sanction imposition
through international organizations is still effective at inducing the target’s compliance.
According to Palkki and Smith (2012), for an illustration, regardless of Libya’s poor economic
performance the actual impact of unilateral U.S. sanctions between 1986 and 1992 was
insignificant due to Libya’s capacity to continue looking for alternative markets and foreign
petroleum investors. Yet the United States was able to impose substantial costs to Qaddafi’s
regime using the international institution sanctions, in particular through the United Nations
(Jentleson and Whytock 2005).
Hypothesis 2.2 Nuclear aspiring states targeted by U.S. sanctions through international
organizations have a higher likelihood of reversing their nuclear pursuit.
2.3.3 The Effects of U.S. Positive Inducements on Nuclear Reversal
Reward dimension of coercive diplomacy refers to a strategy that “can use positive
inducements and assurances” to influence an adversary. According to George (1991), “the
magnitude and significance of the carrot can range from a seemingly concessions that bring
about a settlement of the crisis through a genuine, balanced quid pro quo” (10). Positive
27
inducements may have an influence on the policy change of a target state in two regards. As
foreign rewards or assistance increase, a sanctioned state becomes more accountable to
the sender country than to its domestic constituencies. So, when a target state is more
dependent upon foreign positive inducements, it is likely to make a policy concession to the
sanctioning state in exchange for material benefits. Like negative sanctions, on the other
hand, positive inducements also have a costly signaling effect. In contrast to negative
sanctions, foreign positive rewards may signal to the target the sender’s positive attitude
toward the target state.
Fist, U.S. positive inducements may lead to the target state’s policy change through ‘aid-
forpolicy’ deal. Target leaders tend to be accountable to the aid provider (Bueno de Mesquita
and Smith 2009b). Based on this logic, a target nation may reverse the nuclear development
programs responding to the aid donor’s nonproliferation request. A target state’s leaders
benefit their ruling supporters using foreign aid in exchange for withdrawing the existing
nuclear development. If target leaders believe that the policy concession does not result in
regime change or breakdown, positive inducements may strengthen the power of the
moderates within the incumbent government. Positive sanctions have distributional
consequences in the target state in the sense that they may improve and/or strengthen the
target government’s political legitimacy through the distribution of economic benefits from
them. On the other hand, when the aid-recipient country does not make a concession, it is
likely to become a sanction-target and lose economic benefits that it has been receiving. In
this case, the target nation should be loss-averse. For example, South Korea, at one point
attempted to build nuclear weapons for security purposes but then made a decision to
withdraw its nuclear weapons program because of the U.S. pressure, i.e., the threat of cutting
its assistance and security guarantees (Hersman and Peters 2006). Indeed, South Korea was
a major recipient of U.S. economic and military aid in the post-Korean War period. Egypt also
initiated a nuclear program in the 1960s but gave it up in the 1970s because Anwar Sadat, a
successor of Gamal Abdel Nasser, sought to avoid giving much power to the existing ruling
elites but preferred market-oriented reforms (Nincic 2011). At the same time, Egyptian
government needed to withdraw its nuclear exploration in order not to lose economic aid
from the US.
Second, U.S. positive inducements retain a significant effect on a target country and
third-party countries as well. Positive rewards such as foreign aid are a positive signal to a
recipient country itself and other countries because a sender spends a considerable amount
of its resources in the aid allocation. For instance, Garriga and Phillips (2014) argued that
U.S. aid allocation is prone to follow geo-strategic concerns. Especially during the Cold War,
U.S. foreign assistance to a recipient country played a role in signaling that it was
strategically important to the U.S. or at least the U.S. may not implement hostile policies.
Hypothesis 2.3 Nuclear aspiring states receiving U.S. positive inducements have a higher
likelihood of reversing their nuclear pursuit.
2.4 Research Design
This study evaluates the testable hypotheses of a theory of U.S. economic statecraft and
nuclear reversal, using a time-series cross-sectional dataset that include 18 countries and
the time period lasting from 1970 through 2004. The test period is primarily based on a
29
theoretical reason. Since the birth of nuclear age, the first nuclear club-i.e., the US,
USSR/Russia, the United Kingdom, France, and China-has attempted to deter nuclear
proliferation. In particular, the U.S. government has adopted a nonproliferation policy since
the establishment of the 1970 Nuclear Proliferation Treaty (NPT) (Dunn 2006). Thus, the first
club nations are excluded from this dataset. Also excluded are countries such as Japan and
Sweden that withdrew their nuclear weapons programs before 1970. This doesn’t
necessarily cause a sampling problem because the U.S. nuclear nonproliferation activity
does not apply to them as targets. This dataset is well documented by previous studies
(Singh and Way 2004; Jo and Gartzke 2007). It does, however, need to be modified for testing
the hypotheses of the nuclear reversal based on updated information (Levite 2002;
Cirincione and Rajkumar 2005).
2.4.1 Dependent Variables
The outcome variable of interest is whether a nuclear-proliferating state that has explored
and/or pursued nuclear weapons programs reverses or keeps pursuing, in a given year, its
nuclear weapons development. To be included in the dataset, the country must have at least
begun an initial exploration and/or pursuit of nuclear weapons, or engaged in nuclear
weapons activity.
Existing quantitative studies of nuclear proliferation tend to place the variable of nuclear
latent capabilities on the side of the independent variables. In reality, many nuclear
proliferating countries have retained nuclear latent capacity in the process of nuclear
pursuit. Some countries removed the entire infrastructure and technology with latent
capacity after renouncing nuclear weapons programs. Some countries have maintained
nuclear latent capabilities even though they have suspended existing nuclear weapons
development.
The measurement of nuclear reversal conceptually relies on Lodgaard’s discussion on
nuclear rollback. According to Lodgaard (2010, 115), nuclear reversal “is a process in the
opposite direction (of nuclear proliferation), reversing intentions and/or capabilities to
acquire nuclear arms. Some states have rolled back to a stable non-nuclear status. Others
have kept a certain preparedness for going nuclear if changing circumstances so suggest.
Yet others have rolled back and forth between different degrees of interest in the nuclear
option and different degrees of material preparedness to exercise it.” Thus, I regard a state in
stable non-nuclear status when it discarded both the intention and capability to construct
nuclear bombs. I also consider a state in nuclear latency status when it still maintains some
degrees of nuclear latent capacity to build nuclear weapons within some time period if it
wants to do while it removes the willingness to make nuclear bombs. To code dates of
nuclear reversal paths for each state, I discuss coding rules from four different dataset: Bleek
and Lorber (2014), Kroenig (2016), Levite (2002), and Way and Weeks (2014).
Based on the conceptualization of the paths of nuclear weapons development, I use the
dependent variable, which consists of two measures. The first dependent variable is a binary
measurement. A dichotomous outcome variable, Nuclear reversal, is coded 1 if a nuclear
proliferating state makes a decision to reverse its nuclear weapons development in a given
year and 0 if not.
The second dependent variable has value 0 for all country years for no nuclear pursuit
(Nonnuclear status), value 1 for Nuclear pursuit, and value 2 for Nuclear latency. Data for the
31
dependent variable are collected by Way and Weeks (2014) for nuclear paths with nuclear
latency from Fuhrmann and Tkach (2015). A state having an ENR plant in operation is
considered a state with nuclear latency capacity (Fuhrmann and Tkach 2015). But, nuclear
latency status as a category of nuclear reversal outcomes is coded if the state ever acquires
and possesses nuclear latent capability in spite of its renouncing nuclear weapons
development. I code a state as in a non-nuclear status if a state not only has no intention to
pursue its nuclear weapons development but also lacks nuclear latent capacity to quickly
restart nuclear weapons programs when circumstances suggest. The data are a recent
updated version, which includes cases ignored by past studies.
2.4.2 Independent Variables
The data on independent variables primarily depend on the HSEO dataset (Hufbauer et
al. 2007). To estimate the impact of U.S. sanctions on the targeted state’s nuclear reversal,
U.S. negative sanctions are measured in a dichotomous way in terms that they are coded 1
for each year that the HSEO dataset records all sanctions imposed by the United States, and
0 otherwise. Another sanction variable is U.S. economic sanctions through international
organizations. The US multilateral IO sanctions is coded as 1 for each year that the United
States has imposed negative sanctions through international organizations. I also analyze
the effect of U.S. sanctions on a variation in nuclear reversal behaviors using US sanction
duration and US nuclear sanctions. Including U.S. sanction duration in the equation reflects
a possibility that sanctions are more likely to have the cumulative effect over time rather than
the immediate effects. The variable of U.S. nuclear sanctions is also measured in a
dichotomous way. The aim of U.S. nuclear sanction is specifically to prevent nuclear
proliferation.
The US aid (% GDP) variable as a proxy for U.S. positive inducements is a country’s is a
country’s annual inflows of U.S. total aid as a percentage of domestic product (GDP). It
appropriately captures a country’s aid-dependence on the U.S., which indicates that the
country has more incentives to give policy concession to the U.S. in exchange for U.S. aid
allocation. Alternatively, the US aid variable includes U.S. economic aid and military aid
flowing into targeted countries. It is measured as the natural logarithm of U.S. total aid in a
given year (the result is reported in the section of Appendix.). It is argued that the the increase
in superpower aid (e.g., U.S. aid) by the aid-recipient country is regarded as the actual
increase of the superpower’s support to the recipient country no matter how big the
recipient’s economic size is (Mintz and Heo 2014).
2.4.3 Control Variables
For controlling for other factors to a proliferating state’s affect nuclear reversal behaviors,
the study includes several relevant variables including Sensitive nuclear assistance, Civil
nuclear cooperation, NPT ratification, US security alliance, Trade openness, GDP per capita
(log), Polity score, Disputes, and Rivalry.
First, then, in line with the recent studies on the supply-side of nuclear proliferation, I
include both foreign sensitive nuclear assistance (Kroenig 2009) and civil nuclear
cooperation (Fuhrmann 2009). Sensitive nuclear assistance is a dichotomous variable
coded 1 if a nuclear proliferating state obtained sensitive nuclear materials or technologies
33
from abroad, and 0 otherwise. Sensitive nuclear transfers enable nuclear aspiring states to
overcome technical obstacles in pursuing nuclear weapons development. At the same time,
purchasing foreign sensitive nuclear technology is said to be less expensive than the
indigenous nuclear weapons development in part because of nuclear suppliers’ strategic
interest in helping nuclear aspirants to obtain sensitive nuclear technology (Kroenig 2009).
However, it still should not be always less expensive to obtain foreign sensitive nuclear
assistance given that international community works hard to take measures to deter the
spread of nuclear weapons technology. In that situation, nuclear aspiring states need to pay
both political and economic costs as it becomes more difficult for them to obtain the access
to sensitive nuclear technology from abroad. For example, according to a news report, North
Korean government “bribed top military officials” in Pakistan “to obtain access to sensitive
nuclear technology in the later 1990s” through the secretive deal with transferring more than
$3 million in payments (Smith 2011). Thus, it is expected that Sensitive nuclear assistance is
negatively associated with a state’s nuclear reversal (Bleek and Lorber 2014; Kroenig 2009;
Reiter 2014).
Civil nuclear cooperation is a variable, which counts the aggregated number of bilateral
civil-
ian nuclear agreements that a proliferating state has signed in a given year (Fuhrmann 2009).
Civil nuclear cooperation is said a key determinant of nuclear proliferation because a such
cooperation enables a state to obtain the capacity to pursue nuclear weapons programs. The
probability of nuclear proliferation is said to be increased by the number of civil nuclear
agreement (Fuhrmann 2009). Some scholars, however, argued that civil nuclear agreements
often used by the international community may play a role in dissuading nuclear aspiring
states from obtaining access to sensitive nuclear technology (Bluth et al. 2010, 190). Thus,
the number of civil nuclear agreements may also decrease a proliferating state’s incentive to
pursue nuclear weapons development.
Second, a state’s nuclear behaviors tend to be affected by domestic variables. Previous
research has found that political factors have no significant impact on nuclear proliferation
(Jo and Gartzke 2007). However, some recent studies has founded that domestic political
variables have a significant influence on a state’s nuclear development commitments
(Fuhrmann 2009; Kroenig 2009). For instance, Way and Weeks (2014) found that
“personalistic” dictatorships are more likely to pursue nuclear weapons than are other
political regimes. It is expected that more democratized countries are more likely to reverse
nuclear weapons development than less democratized countries
(Polity score). Polity score reflects a proliferating state’s regime type based on the 21-point
scale (Marshall and Gurr 2011). This variable ranges from -10 to +10, with higher values
representing greater levels of democracy and with lower values indicating greater levels of
autocracy.
The Nonproliferation Treaty membership (NPT ratification) is a dichotomous variable
measured 1 if a proliferating state has ratified the NPT in a given year and 0 if not. A state’s
ratification of the Nonproliferation Treaty (NPT) tends to signal its willingness to conform to
international rules/norms over the nonproliferation commitments. It is expected that a
state’s NPT ratification is more likely to increase the probability of its nuclear rollback
commitments (Bleek and Lorber 2014; Fuhrmann 2009).
35
Third, several international factors may influence a state’s nuclear reversal behaviors.
Security guarantees through defense pacts are likely to reduce a state’s motivation to go
nuclear. It is expected then that security alliance with the U.S. (US security alliance) is more
likely to increase the probability that a proliferating state withdraws its nuclear weapons
development in accordance with the existing findings (Bleek and Lorber 2014; Gerzhoy 2015;
Kroenig 2009; Reiter 2014). A state’s involvement in international disputes (Disputes),
however, may decrease its incentive to withdraw the ongoing nuclear development (Brown
and Kaplow 2014; Fuhrmann 2009; Reiter 2014). In a similar vein, having an enduring rivalry
(Rivalry) may decrease a state’s intention to reverse its ongoing nuclear pursuit due to
security concerns. The next international factor is a state’s openness to trade (Trade
openness). When a state is more open to the international economy and economically more
interdependent, it is more likely to stop nuclear development and rollback its nuclear
programs because it does not want such risky behaviors to hurt its economy and ruling elites
(Solingen 1994, 2012). It is expected that the more open a state’s economy is, the higher the
probability that it reverses its nuclear weapons development (Brown and Kaplow 2014;
Kroenig 2009).
Lastly, I takes into account, using a method suggested by Carter and Signorino (2010), the
temporal dependence in the data. The study includes a variable that counts the number of
years that pass with a nuclear pursuit or non-nuclear-pursuit (Years) with its square (Years2)
and its cube (Years3). All independent and control variables are measured as the lagged one
due to possible lag-effect of sanctions on nuclear behaviors.
2.5 Methods and Results
To test this paper’s arguments, the study estimates a series of binary logit and
multinomial logit model (MNLM). However, the analysis produced the warning message of
“... observation completely determined. Standard errors questionable,” when running a
standard logit regression. This results from problems of complete- or quasi-separation in the
small-sample data. Some studies suggest that researchers should implement penalized
likelihood logistic regression to cope with the problems of separation and finite sample
biases (Allison 2004; Brown and Kaplow 2014; Firth 1993; Zorn 2005). In this case, one or
more of my independent variable are excellent predictors of nuclear reversal (Brown and
Kaplow 2014; Zorn 2005). To run a penalized likelihood logistic model, I use the command
“firthlogit” in Stata.
It begins with the binary dependent variable-nuclear pursuit vs. nuclear reversal. Table
2.1 shows the results of five logistic regressions that test whether U.S. economic statecraft
is associated with a change in the likelihood of a nuclear aspirant’s reversing a nuclear
weapons development. Regarding the effects of U.S. economic statecraft on nuclear
reversal, Models 1 and 2 include U.S. negative sanctions, U.S. sanction duration, and U.S.
aid (% GDP) plus control variables while Models 3 and 4 include U.S. multilateral IO
sanctions and U.S. non-IO sanctions plus control variables. First, the coefficient for U.S.
negative sanctions is statistically insignificant and negative, which is, however, in the
expected direction. The coefficient for U.S. sanction duration is also insignificant but
positive. The result in Model 3, on the other hand, shows that the variable of U.S. non-IO
sanctions decreases the probability that a nuclear-aspiring state reverses its nuclear
37
pursuit. For instance, the logit coefficient on U.S. non-IO sanctions is -1.990 and its odd ratio
is
0.137. This means that, other things being equal, the odds of a nuclear proliferating state
reversing its nuclear pursuit are almost 0.14 times as likely when under U.S. non-IO
sanctions than when under no such sanctions. In fact, becoming a target of U.S. non-IO
sanctions decreases the odds of reversing nuclear pursuit by over 86 percent. However, U.S.
sanctions with IO involvement have a positive sign (direction toward nuclear reversal) but are
not statistically significant in affecting a state’s nuclear reversal behavior. The coefficient for
U.S. multilateral IO sanctions is, for instance, 2.141 in Model 4. The odds ratio for the variable
is 8.5. This means that, other things being equal, the odds of reversing nuclear weapons
development increase by over 750 percent. Regarding the impacts of U.S. positive
inducements, U.S. aid (% GDP) has a insignificant but positive effect on nuclear reversal in
all five models (Models 1-5). Overall, these findings show that, in a binary choice setting
(nuclear reversal vs. pursuit), neither U.S. negative sanctions nor U.S. nuclear sanctions are
not influential factors in extracting policy concessions of nuclear nonproliferation from a
target state. But, U.S. multilateral IO sanctions are effective in inducing a state’s reversal
behavior.
Let us now turn to the determinants of each of a trichotomous nuclear behavior, using a
multinomial logit model (MNLM) to examine the likelihood that a specific outcome is
determined in nuclear development paths. In the multinomial logit models, the study
estimates the effects of U.S. economic statecraft for stable non-nuclear status (outcome 0),
nuclear pursuit/acquisition (outcome 1), and nuclear latency status (outcome 2). The
reference category of the dependent variable is non-nuclear status. Model 6 in Table 2.2
shows the effects of U.S. negative sanctions on each outcome of a state’s nuclear
proliferation commitments. U.S. negative sanction imposition lead to a positive effect on
nuclear pursuit. However, they are not statistically significant in pursuit and latency
outcomes. U.S. sanction duration in Model 7 is also insignificant in affecting both nuclear
pursuit and latency outcomes. Model 8 shows that, when the U.S. imposes economic
Table 2.1: Logit Analysis of Nuclear Reversal, 1970-2004
Model 1
Model
2
Model
3
Model4
Model
5
US
negative
sanctions
-0.593
US
sanction
duration
0.014
2
0.026
2
(0.036
0)
(0.032
7)
US
multilater
al IO
sanctions
1.72
4
2.141
US non-IO
sanctions
-
1.990
US
nuclear
sanctions
-
1.057
-
1.439
US aid (%
GDP)
0.138
0.12
1
0.16
4
0.111
0.060
8
(0.127)
(0.12
1)
(0.12
0)
(0.134)
(0.16
1)
Polity
score
0.128
0.134
∗
(0.0559)
(0.055
5)
(0.062
7)
(0.0623)
(0.056
6)
Trade
openness
(log)
1.312
1.211
+
(0.779)
(0.70
6)
(0.72
7)
(0.727)
(0.70
1)
Sensitive
nuclear
assistanc
e
-0.117
-
0.183
0.093
4
0.247
0.13
4
(0.730)
(0.68
7)
(0.62
3)
(0.635)
(0.67
8)
Civil
nuclear
cooperati
on
0.0733
0.066
5+
0.128
∗
0.125∗
0.070
9∗
(0.0362)
(0.034
9)
(0.050
0)
(0.0501)
(0.033
4)
GDP per
capita
(log)
-0.716
-
0.650
-
1.244
∗
-1.231+
-
0.630
(0.459)
(0.45
3)
(0.62
1)
(0.630)
(0.47
0)
NPT
ratificatio
n
3.048
2.812
∗∗
4.159
∗∗
4.160∗∗
2.973
∗∗
(1.051)
(0.97
8)
(1.26
6)
(1.267)
(1.00
9)
US
security
alliance
0.556
0.66
3
0.56
5
0.643
0.80
0
(0.627)
(0.61
8)
(0.67
4)
(0.668)
(0.64
3)
Disputes
-0.102
-
0.152
0.12
4
0.164
0.075
4
(0.360)
(0.28
2)
(0.12
8)
(0.117)
(0.19
0)
Rivalry
-0.430
-
0.460
-
0.612
(0.365)
(0.34
1)
(0.33
1)
(0.333)
(0.37
4)
Proliferati
on years
0.396
0.43
0
0.37
6
0.391
0.44
4
41
(0.313)
(0.32
9)
(0.33
2)
(0.314)
(0.32
0)
Non-
proliferati
on years
6.885∗∗
7.367
∗∗
(2.163)
(2.40
5)
(2.28
2)
(2.081)
(2.07
5)
Constant
-4.809
-
5.071
-
3.217
-3.764
-
6.323
(4.188)
(4.04
5)
(4.22
6)
(4.297)
(4.47
7)
Observati
ons
528
528
528
528
528
49.84
44.7
8
44.6
2
44.89
43.2
3
126.7
121.
6
125.
7
130.3
124.
3
-6.918
-
4.391
-
3.310
-2.444
-
2.615
Note: Penalized likelihood coefficients with standard errors in parentheses.
All time-variant explanatory variables are lagged at t-1.37
Squared and cubed terms for temporal dependence are not reported.
+ p < 0.10, ∗ p < 0.05, ∗∗ p < .01, ∗∗∗ p < .001
sanctions against a nuclear aspirant through international organizations, it can extract
suboptimal and/or partial compliance i.e., a nuclear latency outcome. This result indicates
that when it comes to America’s higher goals (a target state’s complete and/or optimal
concessions of nuclear nonproliferation), U.S. economic sanctions are generally ineffective.
However, with lower or sub-optimal goals (nuclear power with nuclear disarmament; (Miller
and Sagan 2009), U.S. economic statecraft is moderately successful and effective in
nonproliferation outcomes.
In regard to the substantive interpretation of the results (Table 2.2), this study interprets
the results using the relative risk ratios (rrr), similar to odd ratios in the logistic estimation. In
Model 8, for example, for the category of nuclear latency, the rrr value for the U.S. IO
sanctions is 37.6. This means that the relative risk of choosing nuclear latency over non-
nuclear status is approximately 38 times for a state targeted by U.S. sanctions involving
international organizations relative to a state under no such sanctions. The rrr value for the
U.S. multilateral IO sanctions is 64 in Model 9. U.S. aid (% GDP) is, in Models 8 and 9, both
positively and significantly associated with nuclear latency outcome, relative to non-nuclear
status. For instance, the rrr value for U.S. aid (% GDP) is 1.35 for the category of nuclear
latency in Model 8. This represents the effect of a change of one percent of U.S. aid/GDP in
changing the odds of selecting either nuclear latency or non-nuclear status. The odds of 1.35
means that for every percent of U.S. aid/GDP, the odds of a state adopting nuclear latency
over non-nuclear status increase by almost 1.5 times.
Figures 2.1 and 2.2 plot the predicted probabilities of key independent variables (i.e., U.S.
multilateral IO sanctions and U.S. aid (% GDP) for each category of nuclear development
outcomes. In Figure 2.1, plots the predicted probability for U.S. IO sanctions based on Model
9. The predicted probability of continuing nuclear pursuit is around 0.9 and is highest (near
1) but drops to near 0.5 when the U.S. imposes sanctions through an international
organization. However, it is
38
Table 2.2: Multinomial Logit Analysis of Nuclear Reversal, 1970-2004
Model
6
Model 7
Model 8
Model
9
Model
10
Base
outcome
(Non-
nuclear)
1.123
(1.167
)
43
Nuclear
pursuit
US
negative
sanctions
US
sanction
duration
0.00528
(0.0528)
US
multilatera
l IO
sanctions
-0.338
0.0047
0
(1.062)
(1.032
)
US non-IO
sanctions
3.039
3.086
(2.157)
(2.174
)
US nuclear
sanctions
-0.641
0.050
3
(0.922
)
(0.881
)
US aid
(%GDP)
-
0.0919
-0.0771
-0.113
-0.138
-
0.0964
(0.142
)
(0.126)
(0.207)
(0.251
)
(0.128
)
Sensitive
nuclear
assistance
0.602
0.454
0.460
0.479
0.329
(1.187
)
(1.078)
(1.244)
(1.177
)
(0.973
)
Polity
-
0.128∗
-0.120∗
-0.135∗
-
0.136∗
-
0.119∗
(0.062
0)
(0.0584)
(0.0596
)
(0.060
2)
(0.059
8)
Trade
openness
(log)
-
3.551∗∗
-3.372∗∗
-
4.017∗∗∗
-
3.928∗∗
∗
-
3.286∗∗
(1.126
)
(1.254)
(1.141)
(1.051
)
(1.226
)
Civil
nuclear
cooperatio
n
-
0.139∗∗
∗
-0.130∗∗
-0.192∗∗
-
0.191∗∗
-
0.124∗∗
∗
(0.040
3)
(0.0406)
(0.0587
)
(0.059
4)
(0.037
1)
GDP per
capita
(log)
0.473
0.300
0.894
0.897
0.342
(0.608
)
(0.560)
(0.774)
(0.815
)
(0.605
)
NPT
ratification
-2.146
-1.737
-2.983
-3.028
-1.876
(1.499
)
(1.341)
(1.950)
(2.110
)
(1.410
)
US
security
alliance
-0.155
-0.476
-0.0261
-
0.0126
-0.493
(0.685
)
(0.557)
(0.729)
(0.707
)
(0.494
)
Disputes
0.046
6
0.0837
-0.0325
-
0.0110
0.053
2
(0.177
)
(0.156)
(0.139)
(0.132
)
(0.135
)
Rivalry
0.817∗
0.782+
0.913∗
0.917∗
0.782+
(0.417
)
(0.408)
(0.438)
(0.438
)
(0.411
)
Proliferatio
n years
0.769∗
∗
0.887∗∗∗
0.624∗
0.640∗
0.880∗∗
∗
(0.252
)
(0.200)
(0.301)
(0.305
)
(0.194
)
Constant
11.38+
12.00+
10.67
10.27
11.45
(6.889
)
(7.090)
(6.915)
(7.118
)
(7.193
)
Model 6
Model 7
Model 8
Model 9
Model
10
45
Base
outcome
(Non-
nuclear)
Nuclear
Latency
US negative
sanctions
0.0610
US sanction
duration
0.0188
(0.0380)
US aid
(%GDP)
0.307∗∗∗
0.298∗∗
(0.0907)
(0.0923)
(0.0661)
(0.0840)
(0.104)
Sensitive
nuclear
assistance
1.090
1.045
1.317
1.342
1.132
(1.191)
(1.138)
(1.166)
(1.153)
(1.189)
Polity
0.0637
0.0526
0.0806
0.0769
0.0641
(0.0631)
(0.0618)
(0.0754)
(0.0742)
(0.0668)
Trade
openness
(log)
-3.123
-3.137∗
(1.469)
(1.498)
(1.500)
(1.501)
(1.492)
Civil
nuclear
cooperation
-0.0567
-0.0541
-0.0566
-0.0552
-0.0568
(0.0440)
(0.0429)
(0.0484)
(0.0478)
(0.0434)
47
GDP per
capita (log)
-0.726
-0.596
-0.984
-1.055
-0.888
(0.505)
(0.516)
(0.789)
(0.778)
(0.555)
NPT
ratification
2.139
2.095
2.267
2.319
2.274
(1.850)
(1.822)
(1.907)
(1.883)
(1.815)
US security
alliance
1.383
1.292
1.759
(0.904)
(0.892)
(0.796)
(0.807)
(0.844)
Disputes
-0.0464
-0.0668
-0.0416
0.0161
0.0287
(0.170)
(0.174)
(0.164)
(0.137)
(0.138)
Rivalry
0.0547
0.0709
0.0263
-0.0336
-0.0393
(0.326)
(0.324)
(0.325)
(0.373)
(0.351)
Proliferation
years
-0.783
-0.780
-0.865
-0.845
-0.775
(0.543)
(0.504)
(0.686)
(0.686)
(0.521)
Constant
16.65
15.62+
(8.135)
(8.211)
(11.03)
(10.82)
(8.410)
Observatio
ns
528
528
528
528
528
399.0
402.3
374.4
373.1
398.7
471.6
474.8
447.0
445.7
471.3
-182.5
-184.1
-170.2
-169.6
-182.4
Note: Robust standard errors adjusted for clustering over country appear in parentheses.
All time-variant explanatory variables are lagged at t-1.
Squared and cubed terms for temporal dependence are not reported.
+ p < 0.10, ∗ p < 0.05, ∗∗ p < .01, ∗∗∗ p < .00140
Figure 2.1: U.S. IO sanctions and Predicted Probabilities of Nuclear Reversal Outcomes
49
not statistically significant. The probability of adopting nuclear latency status climbs slightly
to near 0.5 from near zero when a state becomes a target of U.S. multilateral IO sanctions. In
Figure 2.2, the predicted probability of continuing pursuit is highest (almost 1.0) at the lowest
US aid (% GDP) and drops to a low level (near 0) at the highest level of US aid (% GDP) in a
target state.
The probability of opting for nuclear latency is the lowest (almost 0) at the lowest level of US
aid (% GDP), and climb to near 1, at the highest level of US aid (% GDP).
Models 11-15 in Table 2.3 investigates whether U.S. IO sanctions and U.S. aid (% GDP) lead
to
50
Figure 2.2: U.S. aid (% GDP) and Predicted Probabilities of Nuclear Reversal Outcomes
the transition from nuclear pursuit to nuclear latency status. The coefficients for U.S.
multilateral IO sanctions have a statistically significant and positive effect on a state’s
nuclear latency behavior in Models 13 and 14. U.S. aid (% GDP) has a significant impact in
all models (Models 11-15). In contrast, Models 16-19 investigate the onset of nuclear reversal
outcomes (i.e., from nuclear pursuit to nuclear latency or to non-nuclear) only, excluding
decisions to continue nuclear reversal. In the model of the transition from pursuit to latency
(Models 16 and 17), the presence of U.S.
51
IO sanctions has a positive and significant effect on the probability of transition in nuclear
latency status. U.S. aid (% GDP) has , on the other hand, an insignificant impact in Models
16 and 19 (i.e., nuclear latency and non-nuclear status).
The remaining section of the results deals, based on the existing findings, with the
influence of control variables in explaining nuclear reversal outcomes. First, a state’s NPT
ratification has a consistent and positive relationship with its commitment to nuclear
reversal in almost all models in logit analysis of nuclear reversal (Models 1-5 in Table 2.1).
This finding confirms the evidence gathered from existing studies (Bleek and Lorber 2014;
Fuhrmann 2009). But, the variable has no significant effect on nuclear latency and nuclear
pursuit in all models in multinomial logit regression in Table 2.2 in spite of its expected
direction.
Second, the previous research emphasizes the deterrent effect of U.S. security alliances
in preventing the spread of nuclear weapons. The result of analysis offers up mixed findings
with inconsistent significance-in Models 8-10, US security alliance has a significant and
positive effect on nuclear latency outcome. U.S. security alliance also has a positive and
significant effect on the transition from nuclear pursuit to latency (Models 11-15 in Table 2.3)
or to latency onset (Models 15 and 17 in Table 2.4).
Third, some scholars have focused on the role of a state’s openness to the world on
nuclear proliferation (Solingen 1994, 2009). The findings in the logit regression are consistent
with the results of the existing quantitative studies in that an increase in trade openness
opens up more incentives for a state to reverse its nuclear weapons development (Models 1-
5). Trade openness variable also yields similar findings in the multinomial logistic models.
Rivalry variable is negatively associated with a proliferating state’s nuclear reversal in logit
analysis (Models 3-4) while it has a positive impact only on nuclear pursuit status in
52
multinomial logit analysis (Models 6-10). The results confirm previous findings (Brown and
Kaplow 2014; Fuhrmann and Horowitz 2015).
Disputes variable has, however, no significant relationship with nuclear reversal in spite of
the expected direction.
Finally, according to the existing literature (e.g., Fuhrmann 2009; Kroenig 2009), the
supplyside determinants to nuclear proliferation are expected to influence nuclear reversal
outcome(s). On one hand, sensitive nuclear assistance is anticipated to encourage a state
to develop its nuclear weapons (Kroenig 2009). It has no significant influence on a
proliferating state’s nuclear policy change in logit and multinomial logit models in Tables 2.1
and 2.2. Such a finding does not confirm the previous findings (Bleek and Lorber 2014;
Kroenig 2009; Reiter 2014). However, the variable is statistically and positively significant in
logit analysis-transition from pursuit to latency in Models 11-15. In contrast, civil nuclear
cooperation have a significant relationship with nuclear reversal in both logit and
multinomial logit regression. These findings contradict previous findings in that the
international community may use civil nuclear agreements to dissuade a state from
obtaining access nuclear weapons technology (Bluth et al. 2010, 190).
2.5.1 Robustness Analysis
Overall, the findings in the previous section are consistent with this study’s hypotheses.
To avoid the spuriousness of statistical results, a few robustness checks were conducted on
the empirical analysis to show the validity of the inference from the data. The detailed results
can be found in Appendix 1. First, one of the most problematic issues in nuclear proliferation
literature, however, is the measurement of nuclear behavior because nuclear programs are
53
usually implemented in secrecy (Montgomery and Sagan. 2009; Sagan 2010). Recent
quantitative works on nuclear proliferation have constructed several new datasets with
nuclear proliferation dates by updating and/or
Table 2.3: Logit Analysis of U.S. Economic Statecraft and Nuclear Reversal: from pursuit to
latency
Model 11
Model 12
Model 13
Model 14
Model
15
US negative sanctions
-0.282
(0.590)
US sanction duration
0.000713
0.00753
(0.0244)
(0.0246)
US multilateral IO
sanctions
2.655*
2.646*
(1.058)
(1.040)
US non-IO sanctions
-3.435**
-3.419**
(1.251)
(1.208)
US nuclear sanctions
0.266
-1.456
(0.941)
(1.059)
US aid (% GDP)
0.337***
0.332***
0.338***
0.339***
0.315***
(0.0879)
(0.0878)
(0.0942)
(0.0955)
(0.0898)
Sensitive nuclear
assistance
1.019*
1.022*
1.252**
1.250**
1.123*
(0.435)
(0.437)
(0.444)
(0.444)
(0.436)
Polity
0.0975**
0.0946**
0.100**
0.100**
0.104**
(0.0343)
(0.0334)
(0.0373)
(0.0374)
(0.0355)
Trade openness (log)
-1.892***
-1.893***
-1.945***
-1.948***
-1.729***
(0.471)
(0.484)
(0.520)
(0.527)
(0.496)
Civil nuclear cooperation
-0.0350*
-0.0350*
-0.0376*
-0.0377*
-0.0355*
(0.0157)
(0.0158)
(0.0168)
(0.0169)
(0.0159)
GDP per capita (log)
-0.802*
-0.763*
-1.080*
-1.077**
-0.966*
(0.391)
(0.383)
(0.420)
(0.416)
(0.418)
NPT ratification
1.982**
1.895**
2.073**
2.067**
1.896**
(0.707)
(0.717)
(0.693)
(0.690)
(0.735)
US security alliance
1.428**
1.440**
1.819**
1.818**
1.637**
55
(0.496)
(0.497)
(0.555)
(0.553)
(0.525)
Disputes
-0.386+
-0.399+
-0.350+
-0.354+
-0.349+
(0.207)
(0.205)
(0.205)
(0.206)
(0.210)
Rivalry
-0.0786
-0.0701
-0.103
-0.101
-0.150
(0.312)
(0.312)
(0.317)
(0.316)
(0.322)
Proliferation years
-0.931**
-0.942**
-0.836**
-0.831**
-0.920**
(0.343)
(0.341)
(0.323)
(0.322)
(0.342)
Non-proliferation years
0.699*
0.704*
0.795*
0.799*
0.712*
(0.353)
(0.351)
(0.372)
(0.372)
(0.352)
Constant
10.77**
10.44**
12.74***
12.73***
11.46**
(3.635)
(3.625)
(3.720)
(3.685)
(3.750)
Observations
528
528
528
528
528
AIC
149.9
143.8
137.0
138.8
143.0
BIC
226.8
220.7
218.1
224.1
224.2
Log likelihood
-56.96
-53.92
-49.51
-49.38
-52.52
Note: Penalized likelihood coefficients with standard errors in
parentheses.45 All time-variant explanatory variables are lagged at t-1.
Squared and cubed terms for temporal dependence are not reported.
+ p < 0.10, * p < 0.05, ** p < .01, *** p < .001
Table 2.4: Logit Analysis of Nuclear Reversal Onset
pursuit-to-latency pursuit-to-non nuclear
Model 16 Model 17 Model 18 Model 19
US
multilateral
IO sanctions
2.582∗
2.589∗
-
0.0364
0.128
(1.172)
(1.245)
(0.957)
(1.242)
US non-IO
sanctions
-2.010
-1.892
-2.508
(2.082)
(2.060)
(1.220)
(1.294)
US nuclear
sanctions
-0.500
-0.312
(1.138)
(1.564)
US aid (%
GDP)
0.165
0.132
-5.471
-5.450
56
(0.472)
(0.548)
(5.361)
(5.213)
Sensitive
nuclear
assistance
0.250
0.349
-1.533
-1.492
(1.226)
(1.181)
(1.414)
(1.541)
Polity
0.0827
0.0790
-0.155
(0.111)
(0.115)
(0.0855
)
(0.0835
)
Trade
openness
(log)
0.312
0.418
2.234
(1.011)
(1.104)
(1.059)
(1.105)
Civil nuclear
cooperation
0.0463
0.0455
0.127
(0.0693)
(0.0705
)
(0.0458
)
(0.0446
)
GDP per
capita (log)
-0.768
-0.741
-0.937
-0.909
57
(0.845)
(0.841)
(0.659)
(0.734)
NPT
ratification
3.680
3.618
1.982
(2.521)
(2.598)
(0.818)
(0.825)
US security
alliance
3.385∗∗
3.368∗∗
(1.183)
(1.198)
(1.053)
(1.077)
Disputes
-0.0684
-
0.0483
0.174
0.189
(0.329)
(0.344)
(0.170)
(0.133)
Rivalry
-0.290
-0.317
-0.998
-0.994
58
(0.486)
(0.510)
(0.859)
(0.864)
Proliferation
years
0.143
0.149
0.484
0.488
(0.611)
(0.604)
(0.407)
(0.420)
Constant
-3.916
-4.452
-6.665
-7.097
(9.500)
(9.338)
(7.118)
(8.249)
Observation
s
399
399
448
448
95.41
97.29
120.8
120.7
159.2
165.1
186.5
186.4
-31.70
-31.64
-44.39
-44.37
Note: Robust standard errors adjusted for clustering over country appear in parentheses.
All time-variant explanatory variables are lagged at t-1.
Squared and cubed terms for temporal dependence are not reported.
+ p < 0.10, ∗ p < 0.05, ∗∗ p < .01, ∗∗∗ p < .001
revising the existing periods of nuclear proliferation-the updated datasets of Jo and Gartzke
(2007) and Kroenig (2016). Second, the empirical findings are robust to alternative measures
of U.S. positive inducements-US aid (log). It is argued that the the increase in superpower aid
(e.g., U.S. aid) by the aid-recipient country is regarded as the actual increase of the
superpower’s support to the recipient country no matter how big the recipient’s economic
size is (Mintz and Heo 2014). The substantive results are unchanged. Third, as another
robustness check, the study also subset the data by excluding non-NPT nuclear weapons
states that possessed nuclear weapons-India, Israel, and Pakistan. The substantive results
are still robust. Fourth, the Cold War and the Post-Cold War were separatged because for
many countries in the world the dissolution of the Soviet Union led to a decrease in the
efficacy of possession nuclear weapons (Fuhrmann and Sechser 2014; Sagan 1996). Despite
adding the Cold war variable, the results still support the main hypotheses. Lastly, I add the
variable of Liberalization for a state’s openness with the international economy, in addition
to trade openness (Solingen 2009). The finding remains unaltered. All these findings of
robustness tests are reported in the appendix.
59
2.6 Conclusion
This paper tests whether U.S. economic statecraft influences nuclear reversal. The
empirical evidence suggests that U.S. negative sanctions have a detrimental effect on
counterproliferation in both the logistic and multinomial logistic estimations. In other words,
when states that have initiated or pursued nuclear development come under U.S. sanction
imposition, they are less likely to reverse their nuclear pursuit. The findings indicate that with
regard to nonproliferation U.S. sanctions result in counterproductive outcomes. U.S.
sanctions produce negative political and economic externalities, as sanctioned states may
anticipate the U.S. using of military force against them (Lektzian and Souva 2007).
In contrast, when nuclear reversal outcomes are disaggregated into non-nuclear status,
nuclear latency, and nuclear latency, U.S. positive inducements including US aid (% GDP)
have a positive impact on a proliferating state’s decision to maintain nuclear latent capacity
in exchange for renouncing nuclear weapons development. These findings suggest that U.S.
aid allocation provides a costly signal to recipient countries; they are sufficiently important
geo-strategically that their political survival will not be hurt and that they are not potential
targets for U.S. military attack (Garriga and Phillips 2014). U.S. sanctions through
international institutions leads to a state’s decision to choose nuclear latency over nuclear
pursuit.
These findings offer several important contributions to the current literature on nuclear
(non)proliferation. They confirm the detrimental effects of U.S. negative sanctions on
unintended products such as a state’s continuing to pursue nuclear weapons development
due to potential imminent threats from the U.S. and its coalition countries However, the
results indicate that if the U.S. were to adopt a lower standard of success in its economic
statecraft (e.g., U.S. sanctions through international institutions, U.S. positive inducements),
it could extract from a proliferating state a sub-optimal policy concession like “nuclear
latency status” or “nuclear power without nuclear proliferation.” At the same time, when the
60
U.S. induces more international cooperation using its negative sanctions, it may lead to a
partial nonproliferation outcome. The 1994 Agreed Framework between the U.S. and DPRK
illustrates, despite its later breakdown, the effectiveness of positive inducement in
nonproliferation and/or nuclear reversal. The recent negotiation and the conditional
agreement over nuclear withdrawal between the West (in particular the U.S.) and Iran imply
that U.S. sanctions through international institutions and U.S. positive inducements (e.g.,
lifting existing sanctions) can be effective enough to extract a nuclear-aspiring state’s
compliance with nonproliferation.
61
2.7 Appendix I
2.7.1 Robustness Analysis
• Alternative nuclear datasets (Models A1-A2). As I discussed in the paper, it is
sometimes difficult to identify whether states are pursuing the bomb in light of the
secrecy that often shrouds nuclear weapons programs. To address this issue, I recode
the dependent variable using two alternative nuclear behavior datasets coded or
produced by (Jo and Gartzke 2007) and (Kroenig 2016). The results show that U.S. IO
sanctions remains closely associated with a state’s nuclear latency behavior when I
use the Jo and Gartzke (2007) and Kroenig (2016) dataset to construct my dependent
variable.
• Alternative measure of U.S. positive inducements (Models A3-A4). Alternatively, the US
aid variable includes U.S. economic aid and military aid flowing into targeted
countries. It is measured as the natural logarithm of U.S. total aid in a given year. It is
argued that the the increase in superpower aid (e.g., U.S. aid) by the aid-recipient
country is regarded as the actual increase of the superpower’s support to the recipient
country no matter how big the recipient’s economic size is (Mintz and Heo 2014). The
results show that U.S. aid (log) is positively associated with the category of nuclear
latency status indicating that a state receiving the higher level of U.S. total aid,
regardless of its economic size, adopts nuclear latency status.
• Excluding non-NPT nuclear weapons states (Models A5-A6). The study also subsets
the data by excluding non-NPT nuclear weapons states (i.e., India, Pakistan, and Israel)
that acquired nuclear weapons One could argue that de facto non-NPT nuclear
weapons states have already acquired nuclear bombs and lack the intention to reverse
their existing nuclear weapons development. Thus, those countries are unlikely to ever
62
reverse nuclear pursuit. Yet, the findings are similar when the study exclude country-
year observations representing three de facto nuclear weapons states and even
coefficients for main independent variablesU.S. IO sanctions and U.S. aid (% GDP)
have a much stronger explanatory power in determining nuclear latency status. Unlike
previous findings, U.S. sanction duration has a positive and significant relationship
with nuclear latency outcome when the study exclude non-NPT nuclear weapons
states.
• Including Cold War dummy (Model A7). The change of international structure
influences a state’s behavior according to neo-realist argument. The collapse of the
Soviet Union might decrease a nuclear proliferating state’s incentive to keep pursuing
nuclear weapons development in part due to the reduction of security threats. My
findings remain largely unchanged when I control for the Cold War period.
• Superpower alliance (Model A8). This study argued that security alliance with the U.S.
tends to discourage proliferating states to keep pursuing nuclear weapons program. It
is said that alliances with nuclear-armed superpowers (i.e., the U.S. and the
USSR/Russia) can play a deterrent role in dissuading a state from pursuing nuclear
weapons development and also provide it with nuclear umbrella (Fuhrmann and
Horowitz 2015). A robustness analysis shows that the findings are similar to previous
ones.
• Adding liberalization variable (Model A9). My initial empirical tests include, based on
Solingen (2009) argument, trade openness representing a state’s integration with the
international economic system. For another robustness check, the study also include
the variable
63
of a state’s liberalization measuring 5-year changes in trade openness over time (Bleek
and Lorber 2014). Model 8 shows, however that both U.S. IO sanctions and U.S. aid (%
GDP) remain statistically significant when I add another factor that might influence a
state’s reversal behavior.
64
2.7.2 Summary Statistics, Nuclear Proliferation Period,US Economic
Sanctions, 1970-2004, Operationalizations of Control Variables
Table 2.5: Summary statistics
Variable
Mean
Std. Dev.
Min.
Max.
N
Nuclear reversal
0.35
0.477
0
1
555
Nuclear reversal outcomes
0.884
0.651
0
2
630
US negative sanctions
0.395
0.489
0
1
612
US nuclear sanctions
0.208
0.406
0
1
612
US sanction duration
5.835
10.538
0
54
612
US multilateral IO
sanctions
0.157
0.364
0
1
612
US non-IO sanctions
0.284
0.451
0
1
612
US aid (% GDP)
0.647
2.155
0
22.5
612
Sensitive nuclear
assistance
0.363
0.481
0
1
612
Polity2
-1.017
7.599
-10
10
597
Trade openness (log)
3.68
0.72
0.6
5.9
612
Civil nuclear cooperation
14.871
13.372
0
63
612
GDP per capita (log)
8.640
0.823
6.9
10.6
599
NPT ratification
0.629
0.483
0
1
612
US security alliance
0.33
0.471
0
1
612
Disputes
1.376
2.071
0
27
612
Rivalry
0.763
1.183
0
5
612
65
Table 2.7: Nuclear Proliferation Period (updated)
Singh and Way (2004)
Jo and Gartzke (2007)
Explore Pursue Acquire
Program Acquire
66
Algeria
1983-
Argentina
1968-1977
1978-1990
1978-1990
Australia
1956-1960
1961-1973
Brazil
1953-1977
1978-1990
1978-1990
Egypt
1960-1964
1965-1974
India
1954-1963
1964-19787
1988-
1964-1987
1988-
Iran
1976-1984
1985-
1974-
Iraq
1976-1982
1982-1995
1973-2002
Israel
1949-1957
1958-1968
1955-
1965
1966-
Korea, North
1965-1979
1980-2005
2006-
1982-2005
2006
Korea, South
1959-1969
1970-1978
1971-1975
Libya
1970-2003
1970-2003
1970-2003
Pakistan
1972-1986
1987-
1972-1986
1987-
Romania
1985-1990
1981-1989
South Africa
1969-1973
1974-1978
1979-
1991
1971-1978
1979-
1990
Syria
2000-2009
Taiwan (1)
1967-1977
1967-1976
Taiwan (2)
1987-1988
Yugoslavia
1974-1988
1982-1987
Source: Way and Weeks (2014)
Table 2.8: US Economic Sanctions, 1970-2004
country
Sanction Periods
Sanction Goals
Argentina
1977-1983
Improve human rights
1978-1982
Adhere to nuclear safeguards
Brazil
1977-1984
Improve human rights
1978-1981
Adhere to nuclear safeguards
India
1971
Cease fighting in East Pakistan (Bangladesh)
67
1978-1982
Adhere to nuclear safeguards
1998-2001
Retaliate for nuclear test
Constrain nuclear program
Iran
1979-1981
Release hostages
Settle expropriation claims
1984-
Terminate support for international terrorism
End war with Iraq & Renounce WMD
Iraq
1980-2003
Terminate support for international terrorism
Renounce WMD
1990-1991
Withdraw from Kuwait & Release hostages
1991-2003
Renounce WMD & Destabilize Huessein government
Israel
1970-1983
Withdraw from Sinai & Implement UN Resolution 242
Push Palestinian autonomy talks
Korea, North
1970-
Impair military potential
Destabilize communist government
1993-1994 / 2002-
Renounce nuclear weapons
Korea, South
1973-1977
Improve human rights
1975-1976
Forgo nuclear reprocessing
Libya
1978-2004
Terminate support for international terrorism
Destabilize Gadhafi government
Stop pursuit of chemical, nuclear weapons
Pakistan
1971
Cease fighting in East Pakistan (Bangladesh)
1979-2001
Adhere to nuclear safeguards & stop pursuit of nuclear
weapons
1999-2001
Restore democracy
Romania
1983-1989 / 1990-1993
Improve human rights & Ease restrictions on emigration
Establish democracy, election
South Africa
1975-1982
Adhere to nuclear safeguards
Avert explosion of nuclear device
1985-1991
End apartheid
Syria
1986-
Terminate support for international terrorism
Taiwan
1976-1977
Forgo nuclear reprocessing
Yugoslavia
1991-2001
End civil war in Bosnia, Croatia
1998-2001
Stop aggression in Kosovo & Destabilize Milosevic
Source: Hufbauer, Schott, Elliott and Oegg (2007, 20-33)
Table 2.9: Multinomial Logit Analysis of Nuclear Reversal, 1970-2004
Jo & Gartzke (2007)
Kroenig (2016)
Model A1
Model A2
68
Nuclear Pursuit Nuclear Latency
Nuclear Pursuit Nuclear Latency
Robust standard errors adjusted for clustering over country appear in parentheses.
All time-variant explanatory variables are lagged at t-1.
+ p < 0.10, ∗ p < 0.05, ∗∗ p < .01, ∗∗∗ p < .001
70
Chapter 3
71
Another Aid Curse?: U.S. Foreign Aid, Autocratic Institutions, and
Political Violence against Americans, 1970-2007
Abstract. This paper examines the determinants of political violence against Americans
abroad, situating such violence in the context of U.S. aid policies. Despite a growing number
of studies on anti-Americanism (Chiozza 2010; Katzenstein and Keohane 2007), few studies
pay attention, in a systematic or empirical way, to the violence dimension of anti-
Americanism. The political violence literature has paid little attention to exploring the
phenomena of political incidents against specific subjects, in particular U.S. citizens and
properties. When it comes to anti-American violence, the literature offers an important
rationale. Anti-government groups in a country tend to exercise violent behaviors against U.S.
citizens and/or properties when the U.S. backs up for their repressive dictators by providing
them with free resources including U.S. foreign assistance. I hypothesize here that a
recipient’s constituencies can consider U.S. aid as bad money because it can facilitate the
lengthening of its autocrat’s tenure and record of repression. The role of U.S. aid varies
depending on a recipient’s political institutions. Especially, when a recipient state is an
autocratic regime but is bounded to nominal-democratic institutions, it has a lower
likelihood of experiencing anti-U.S. violence.
3.1 Introduction
Since the end of World War II, the United States has become the object of both hatred
and love. People living outside the U.S. have, in a variety of ways, expressed their discontent
with the country and its foreign policies. Anti-American sentiments are seen in published
criticism, non-violent demonstrations and deadly terrorist attacks. Protesters are often seen
in front of U.S. embassies burning American flags. Others detonate explosives on U.S.-
72
owned properties such as local branches of U.S. banks, multinational corporations, and
even American churches. Such violent behaviors are often a reaction to the U.S.’s support of
dictators. In the late 1960s, the United States did back the Greek military junta and in the
1950s began supporting the Franco regime of Spain. In Africa, the United States offered its
backing for decades to Mobutu Sese Seko who coercively governed the Democratic Republic
of Congo. In Latin America, the United States provided material and political support to the
Pinochet regime of Chile after its coup in 1973. In Asia, U.S. government became an
important sponsor of President Ferdinand Marcos of the Philippines (The Economist, Feb.
19, 2005). During the Cold War, many anti-American incidents took place in all of those
countries.
Anti-American resentment and violence persists even today. During the recent “Arab
Spring,” protesters in Cairo accused the United States of protecting the Mubarak regime for
30 years by assisting it with a large amounts of aid while relatively ignoring the political
liberalization for Egyptian people (England 2011). In 2012, the U.S. Ambassador to Libya was
killed in Benghazi due to terrorist attacks on the U.S. Special Mission Compound and Annex.
With anti-Americanism’s resurgence in many third world, scholars have revisited anti-
Americanism research primarily focusing on public opinion/attitude (e.g., Blaydes and Linzer
2012; Bush and Jamal 2015; Chiozza 2010; Katzenstein and Keohane 2007) and terrorism
(e.g., Krieger and Meierrieks 2015; Neumayer and Plümper 2011). Despite these academic
efforts, few have systematically investigated antiAmericanism in the context of political
violence looking into causal mechanisms through which U.S. foreign aid leads to political
actions against Americans in a recipient state.
A growing number of studies have focused on the externalities of foreign aid in recipient
countries. Proponents of foreign aid posit that aid flows are (conditionally and) positively
associated with boosting economic activity and subsequently lead to economic growth
(Bearce and Tirone 2010; Karras 2006; Minoiu and Reddy 2010). At the same time, recipient
73
governments increase their investment in public goods and improve the welfare conditions
of the general public (Alvi and Senbeta 2012). Pessimists of aid effectiveness point to the
counterproductive impacts of aid on the political economy of authoritarian recipients. For
example, foreign aid helps autocrats stay in office and lowers the risk of ouster by
revolutionary threats (Ahmed 2012; Bueno de Mesquita and Smith 2009b; Licht 2010; Smith
2008) and decreases the likelihood of democratization or regime change (Kono and
Montinola 2013; Kono, Montinola and Verbon 2015; Morrison 2012; Wright 2009).
This essay empirically investigates the effect of U.S. aid on the frequency of anti-
American incidents by proposing a theory that accounts for how U.S. foreign aid leads
political groups outside the winning coalition to resort to violent behaviors against
Americans in dictatorships. The study focuses on the negative externalities of U.S. aid to
domestic politics in autocracies. As U.S. foreign aid influences autocratic recipients’
domestic policies and/or governance, it may also increase or decrease extreme opposition
groups’ incentives to behave aggressively toward Americans in their national territory. The
study argues that U.S. foreign aid flowing to an authoritarian recipient gives rise to more
political incidents against Americans than it does when flowing to a democracy. Variation in
autocratic institutions has an intervening effect on the relationship between U.S. foreign
assistance and anti-Americanism. To test this argument, the study employs a series of
statistical estimations to deal with count data for 117 developing countries from 1970 to
2007. The results confirm the positive impact of U.S. aid on the frequency of anti-American
incidents in autocracies compared to in democracies when the study uses U.S. aid as a share
of gross domestic product (GDP) of a recipient country. the study also finds that when an
autocratic recipient has politics institutions, the effect of U.S. aid on anti-American incidents
in that country is significant. A substantive interpretation of the results indicates that
autocracies having institutional constraints are predicted to experience fewer anti-American
incidents than either democracies or autocracies without such constraints.
74
This research makes a significant contribution to both existing anti-Americanism and
American foreign policy studies. Earlier studies have tended to focus on the attitudinal
aspect of antiAmericanism. However, previous work pays less attention to the behavioral
dimension of antiAmericanism across countries. The research also adds to the growing body
of literature that looks at the adverse effects of U.S. foreign policy as represented by foreign
aid in recipient countries. It offers theoretical and empirical findings that a dictatorship with
a more inclusive institutional process is particularly prone to experience fewer anti-
American incidents.
The next section reviews the literature regarding the concepts and the determinants of
antiAmericanism. The study then offers a theory regarding the condition on which U.S.
foreign aid increases the frequency of anti-American incidents in recipient states. To
evaluate the hypotheses, the section on the research design consists of measurement and
the methodological techniques. The study relies on a series of statistical models using zero-
inflated negative binomial (ZINB) regression to account for both the overdispersion and
excessive zero in the data. Results and implications are discussed at the end.
3.2 Literature Review: Defining and Analyzing Anti-Americanism
Scholars have proposed the concepts about anti-Americanism focusing on two
dimensions: (1) anti-Americanism is associated with attitudes/views toward the United
States, U.S. society, culture, and foreign policies, and (2) anti-Americanism is about
behaviors/actions toward Americans and U.S. foreign policies.
A growing number of studies emphasize the attitudinal approach of anti-Americanism.
Katzenstein and Keohane (2007, 12) regarded “anti-Americanism as a psychological
tendency to hold negative views of the United States and of American society in general.”
75
Anti-Americanism may be viewed as an attitude. Oh and Arrington (2007, 331) posited that
anti-American sentiment represented “the negative attitudes of foreigners toward the US,
her policies, and/or the consequences of those policies” in evaluating anti-Americanism in
South Korea. Naghmi (1982, 508) defined anti-Americanism as “an unfavorable, or hostile
attitude toward the American people, government, symbols, or policy” when looking to
Pakistan’s public attitude toward the United States. Friedman (2012, 5) posited that “the
term anti-Americanism is variously defined as an ideology, a cultural prejudice, a form of
resistance, a threat, or as opposition to democracy, the rejection of modernity, or neurotic
envy of American success.”
On the other hand, some studies pay more attention to the behavioral dimension of
antiAmericanism. Scholars like Tai, Peterson and Gurr (1973) and Rubinstein and Smith
(1988) described anti-Americanism in the sense of actions or statements against U.S. policy,
society, and values. For example, Rubinstein and Smith (1988, 36) defined anti-Americanism
as “any hostile action or expression that becomes part and parcel of an undifferentiated
attack on the foreign policy, society, culture, and values of the United States.”
Studies on anti-Americanism propose various typologies of anti-Americanism regarding
the causes of anti-American sentiment. Some factors are common across studies. Recently,
international relations scholars revisited anti-Americanism not only focusing on a
conceptual discussion but also empirically analyzing anti-Americanism (e.g., Chiozza 2010;
Katzenstein and Keohane 2007). In particular, Katzenstein and Keohane (2007, 35-41)
broadly categorized antiAmericanism as following one of four approaches: liberal, social,
sovereign-nationalist, and radical anti-Americanism. First, liberal anti-Americanism is
common in advanced industrialized countries. This kind of identity tends to share a liberal
ideology such as Western values. Second, social anti-Americanism is associated with
democracies embedded in social welfare policies and the acceptance of various types of
76
democracy. Third, sovereign-nationalist anti-Americanism is mainly related with
nationalism, national sovereignty, and people’s perceptions of power relations in the world.
For instance, with their nations penetrated by foreign economic and military presences,
indigenous citizens are hostile to the United States out of their tendency to protect
themselves and their identity (Tai, Peterson and Gurr 1973, 458). Lastly, radical anti-
Americanism is strongly associated with violent actions against Americans.
Liberal anti-Americanism posits that people in advanced democracies share similar ideals
and political freedoms as do the United States citizens. Yet they criticize U.S. foreign
policies and/or the hypocrisy of U.S. behaviors. Since the 9/11 terrorist attacks, the U.S.
wars on terror has stirred up a considerable amount of anti-American sentiment around
the world. This is partly because U.S. foreign policy has been perceived as being relatively
unilateral. Unlike the widespread support for U.S. military campaigns against the Taliban in
Afghanistan in 2001, American military intervention in Iraq in 2003 led to an outbreak of
severe anti-Americanism in Europe. A large number of people in Europe stood against U.S.
unilateralism. They expressed their sentiments against
U.S. foreign policy through mass demonstrations (Fabbrini 2010). Some scholars have noted
that U.S. post 9/11 foreign policy has given rise to a new wave of anti-Americanism,
“provoking the widespread public expression of antipathy toward the United States” (Singh
2007, 26).
The second approach, social anti-Americanism, “derives from a set of political
institutions that embed liberal values in a broader set of social and political arrangements
that help define market processes and outcomes left more autonomous in the U.S."
(Katzenstein and Keohane 2007, 31). It is closely associated with the conflicts of social
values that both Americans and people in other countries regard as salient. Such conflicts
over social values include the negative perceptions of the market-driven American value
77
system, the death penalty, social protection, and so on. In practice, social anti-Americanism
is not stronger than other forms of anti-Americanism (Katzenstein and Keohane 2007).
The third, sovereign-nationalist anti-Americanism, can be found in most of the developing
world. According to Katzenstein and Keohane (2007, 32), nationalists tend to emphasize
“sovereignty” and collective “nationalism.” Nationalist identities tend to generate the potential
for anti-American sentiments while sovereignty “becomes a shield against unwanted intrusions
from America.” The concept of sovereignty is critical to nationalists in developing countries.
Members of national-
ist groups may consider foreign interventions and penetrations as a threat to national
sovereignty. Sovereignty and nationalism are linked with some factors that may create
potential anti-American sentiments within a country: U.S. economic/military presence and
U.S. foreign direct investment.
American economic and military presence faced strong resistance from a hosting
country because nationalist-revolutionary groups often regarded the foreign direct
investment (FDI) as a symbol of economic exploitation or the foreign military deployment as
an infringement on their national sovereignty in the developing world (Rubinstein and Smith
1988; Tai, Peterson and Gurr 1973). Rubinstein and Smith (1988, 41) contended that
revisionist or revolutionary groups tend to resist the spread of American capitalism since
they identify it with “the perpetuation of traditional values and institutions” that they want to
overthrow. In their study on terrorism, Krieger and Meierrieks (2015) evaluated the impact of
capitalism on anti-American terrorism. They found that antimarket political groups
intentionally attack the U.S. as a primary supporter of market-capitalism and globalization.
Especially, during the transition from clientalism to market-capitalism, traditional power
groups that have usually benefited from “the pre-market clientalist-traditionalist order” in
the country may strategically target the United States “to effectively voice dissent and
rollback pro-market developments” (Krieger and Meierrieks 2015, 59).
78
Nationalists regard foreign military deployment as an infringement on their sovereignty
(Rubinstein and Smith 1988; Tai, Peterson and Gurr 1973). A foreign political/military
presence in a nation may directly motivate nationalists to resist it. In particular, the
asymmetry of a security alliance between the U.S. and less powerful countries can cause
ambivalent responses from a hosting country. Its people welcome the military security while
feeling discontented about the decrease of national autonomy.
Lastly, the fourth possible approach to anti-Americanism is radical anti-Americanism,
which renounces all U.S. economic policies and institutions in the world (Katzenstein and
Keohane 2007). For revolutionary groups, their incumbent governments embrace U.S.
institutional features and values that these groups strongly disapprove of. In extreme
cases, this type of anti-Americanism can be fomented into revolutionary movements that
attack both their own governments and Americans. This is similar to “revolutionary” anti-
Americanism as defined by Rubinstein and Smith (1988). They argued that
revolutionary/opposition groups seek to “overthrow regime closely identified with the
United States,” and such an attempt also includes violent attacks against the U.S.
government and Americans (Rubinstein and Smith 1988, 42).
Of these aforementioned types of anti-Americanism, the latter two are what most often
produce anti-U.S. violence. The next section explores a theoretical argument about how U.S.
foreign aid influences anti-American incidents.
3.3 Theoretical Argument: U.S. Aid and Anti-American Violence
In pursuing its national interests, every country employs several foreign policy
instruments including military intervention or use of force, economic sanctions, foreign aid,
and public diplomacy (Goldstein and Pevehouse 2014). The available range of those tools
depends on a country’s available resources. The more resourceful a country is, the more
79
extensive her policy instruments are. In this respect, the United States has been one of the
most resourceful countries, as it has available to it many policy tools. Among such tools,
foreign aid has been leveraged to maximize U.S. national interest. The Unites States has can
operate as a donor government and buy from recipient countries policy concessions or
cooperation; leaders of the recipient countries gain additional and unearned income that
they can freely spend to promote economic development and improve public goods and
services.
Critics contend, however, that foreign aid is ineffective at economic development and/or
the poverty reduction in a recipient country. Its ineffectiveness is due to its fungibility. In using
foreign aid, leaders in a recipient country can freely convert it to their own use, for example,
to maximize their leadership tenure. Thus, the theoretical arguments begin with the general
assumption of the existing literature. What the leader of a country prioritizes is political
survival (Bueno de Mesquita et al. 2003). Autocratic leaders in the developing world tend to
utilize available resources and/or tools to stay in power.
In estimating anti-Americanism, this study proposes theoretical explanations for anti-
American incidents. The study discusses why U.S. foreign aid can increase the frequency of
anti-American events in the developing world. It then accounts for the intervening effect of
political institutions on the relationship between U.S. aid and anti-American incidents.
Existing studies on aid-survival relationships suggest that the negative externality of foreign
aid might be of increasing the dissident’s opportunities to express their discontent to both
the home government and the donor country (esp. the United States) (Ahmed 2012; Kono
and Montinola 2009; Licht 2010; Bueno de Mesquita and Smith 2007, 2009b; Smith 2008).
Why and how does U.S. aid raise the number of anti-U.S. violence in recipient countries.
This section identifies two theoretical explanations-the “accountability” hypothesis and
“aidfungibility” hypothesis. The accountability hypothesis assumes that unlike a democracy,
80
an autocratic aid-recipient state becomes accountable to ruling coalition groups, a small
segment of the entire population, while relatively ignoring the general public when its
resource is more dependent upon foreign aid. The aid fungibility hypothesis, on the other
hand, posits that since political leaders may divert foreign aid to resources so as to maximize
their political survival (e.g., through co-optation and repression), opposition groups outside
the winning coalition have an incentive to express, through violent actions, their discontent
with the U.S.
The first causal link is that U.S. foreign aid may generate a negative externality through
socalled aid-for-policy deals. Such deals create some policy conditionality in that recipient
leaders make policy concessions; they exchange their policies for donors’ aid (Bueno de
Mesquita and Smith 2009b; Morrison 2012). The policy concessions may suit the donor’s
policy preferences but are often at odds with the preferences of domestic political groups
outside the ruling coalition. Thus, policy concessions, since they make their incumbent
leaders much less accountable to other segments of social groups in the country, may
increase the political opportunity of such groups to express their discontent with their
government as well as with the United States. In particular, policy concessions to donors
may lead to domestic opposition/discontents from their constituencies, particularly
nationalists and revolutionaries. Those nationalists and/or revolutionary groups consider the
aid-for-policy deals as a loss of sovereignty/autonomy; they condemn the United State as the
enemy of the national autonomy and pride. Political leaders in the developing world have
taken an anti-communist stance in deference to U.S. foreign policy interests and in return
have received U.S. economic and military assistance. As a way of preventing the spread of
communism, the United States has given economic aid to authoritarian countries that
border communist countries, such as Pakistan and Honduras (Meernik and Poe 1998).
According to French (1997), Zaire was strongly supported by the United States during the
Cold War because of its anti-communist frontline in the Central African region. (Askin and
81
Collins 1993) described U.S. support for the Mobutu regime as follows: “In 1983, after Zaire
sent troops to defend U.S.-backed Chadian President Hissen Habre, President Reagan
praised Mobutu’s courageous action” and rewarded him with a request to Congress for the
doubling of U.S. foreign aid” (79).
These kinds of deals provide a recipient state with foreign aid as unearned revenues and
this makes them less to domestic groups outside the ruling coalition (Knack 2004; Smith
2008). As their dependence on foreign aid increases, autocratic leaders rely much less on
domestic ordinary citizens and thus have little incentive to expand public goods and/or invest
in good governance for the general public than doe their democratic counterparts (Bueno de
Mesquita and Smith 2007). Instead, autocratic leaders tend to spend aid money on their
winning coalition by increasing private goods. Autocratic countries use foreign aid for
economic development and social welfare to a much lesser degree than do democratic
countries. Aid in fact tends to increase inequality as it may be spent on projects that have no
productive value or are for the public goods and services. For the general public in a recipient
country (esp. an autocratic country), foreign economic assistance seems to improve their
social welfare very little and to increase the inequality between the ruling elites and ordinary
people outside the ruling coalition. Like oil wealth, foreign aid may lead to another source of
rent-seeking among ruling elites. Political corruption resulting from ruling coalition’s rent-
seeking behaviors may increase the level of grievances among political groups outside the
ruling coalition. In their use of foreign aid, authoritarian leaders tend to implement patronage
politics rather than carry out productive policies. Unless the institutional process take into
account their demands or grievances, political actors left out from this rent-seeking have an
incentive to resort to political violence. In this context, political groups outside the ruling
groups may either rely on direct violence to the incumbent government or strategically take
out their grievances on a third-party target-especially, the aid-donor country (e.g., the United
States) that the incumbent government cares strongly about. According to some studies, a
82
large number of Muslim publics supported, with such political reasons, for political violence
against Americans (Berger 2014; Tessler and Robbins 2007). Islamist radicals and extreme
groups opposing their governments often committed violent attack on Americans, taking
advantage of those support from the publics.
Secondly, the aid fungibility argument holds that foreign aid has a detrimental impact on
the socio-political conditions of recipient countries, as it augments the autocratic surviving
capacity. Autocratic rulers try to stay in power as long as possible using co-optation and
repression in general. Foreign aid seems like bad money because it might help dictators
survive longer and effectively prevent revolutionary threats and opposition movements. As
free resources, incumbent autocrats discretionally allocate dollars from U.S. foreign aid
through private goods so as to survive politically. Dictators invest free money to repress
revolutionary movements, for instance increasing defense spending and strengthening
security forces. Democratic recipients tend to redistribute it to the general public as public
goods (Bueno de Mesquita and Smith 2010). Bader and Faust (2014, 576) mentioned that
“autocratic recipient governments use foreign aid at least partly for their survival, be it by
redistributing additional rents to strategic groups or by financing repression.”
To maintain a regime, dictators must handle only a small winning coalition, which permits
them to relatively ignore the general public and its welfare. To keep their power, they merely
distribute private goods and great privilege to the ruling elites. Autocrats employ a repressive
strategy of suppressing political dissidents and deterring revolutionary threats. Aid inflows
from foreign donors tend to increase the resources available to repress dissidents.
The fungibility of foreign aid posits that aid may often be converted as military
expenditures, which can contribute to the development of its military capability to defend
the country from foreign invasions or to deter/repress dissident movement in autocracies.
Kono and Montinola (2013), in their study on the relationship between aid and domestic
unrest, argued that foreign aid expands dictators’ repressive capacity to deter dissidents,
83
to fund repressions. In particular, they found that foreign aid helped autocrats reduce
domestic unrest by using the repressive strategy and by increasing military spending to
deter opposition movements. In response, opposition/revolutionary leaders in autocratic
regimes have a strong incentive to manipulate and utilize anti-American sentiments in
order to boost mass support and to compensate for their vulnerability to incumbent
dictators. In this context, Americans and U.S. properties abroad can be very useful targets
of scapegoating for threatened political groups. For opposition groups, the United States is
the primary donor country to support authoritarian leaders but is also one of the most
visible and powerful countries (Neumayer and Plümper 2011; Tai, Peterson and Gurr 1973).
At the same time, the United States has, since the end of World War II, established and
maintained a U.S.-led international system along with its Western allies. Revolutionary
groups strategically blame the U.S. for all of their internal political struggles between
themselves and incumbent governments. So, they may attain the internal cohesion and
anticipate political support from the ordinary public by attacking the U.S., especially the
U.S. embassies and the properties of U.S.-owned multinational corporations. In this
respect, revolutionary-opposition groups take advantage of anti-Americanism as an
instrumental purpose (Rubinstein and Smith 1988). Blaydes and Linzer (2012) and Rubin
(2002) seek to find the source of Islamic anti-Americanism and maintain that elites and
radical
Islamists tend to receive great political benefits for popular anti-American appeals. In
particular, “political elites need to persuade mass public to feel it possible to spit on the
United States” (Rubin 2002, 83). According to the discussion of Ratner (2009), political
challengers often implement anti-Americanism to weaken and also delegitimize incumbent
dictators.
84
Lai and Morey (2006, 388-389) also explained how U.S. aid for anti-communist power was
transferred to boost autocrats’ repression capacity during the Cold War. They discussed the
Iranian case where the United States “provided aid and training for the national police and
intelligence services of the Shah (in Iran), which shared responsibility for internal security.
The United States security assistance programs greatly improved the effectiveness of Iran’s
security forces and therefore substantially increased the Iranian state’s ability to use
repression.” Kono and Montinola (2013) also argued that authoritarian rulers primarily rely on
the military or coercive forces in suppressing public discontent. To pay for this costly
coercive capability, autocrats sometimes rely on foreign aid. For instance, “with no strong
state or party as his beck and call, Marcos (of Philippines) had little choice but to rely on the
military” (Slater 2010, 177). As an important U.S. ally during the Cold War, the Philippines
has, since 1970, received about 80% of its military aid from the United States government
(Lee 2008). This misuse of foreign aid might help autocrats stay in power and even develop
their repressive capacity to deter revolutionary threats. This will generate greater sources of
anti-American sentiments to dissidents and/or discontented groups and further strengthen
the incentive to resort sometimes to violent behaviors against Americans. This suggests the
following hypothesis:
Hypothesis 3.1 As U.S. aid increases, autocratic regimes experience more anti-American
incidents than democracies.
The Role of Psuedo-Democratic Institutions: Two Competing Explanations
Do anti-American incidents vary across autocratic countries? The existing literature on
authoritarianism point out that dictators in maximizing their political survival use their main
instruments such as “co-optation” and “repression” (Frantz and Kendall-Taylor 2014; Gandhi
85
and Przeworski 2007; Magaloni 2008). The political survival of autocrats is primarily based
on the support of their winning coalition. The maximization of regime survival may rely on
autocrats’ governance and/or policies regarding both public and private goods. Recent
scholarship demonstrates that authoritarian countries exhibit a considerable variation in
their governance patterns. What makes autocrats’ governance so varied are “nominal-
democratic institutions” (e.g., political party and/or legislature). Autocratic institutions play
the role of deterring potential revolutionary threats from both inside and outside the winning
coalition. Through a co-optation strategy, autocrats respond to the demands of ruling elites
by distributing to them private goods and material benefits. In other words, autocratic
countries with institutions tend to be more accountable to the winning coalition groups than
do other dictatorships with no institutions and/or a relatively small winning coalition. This
variation in accountability implies that autocratic rulers have different incentives in
distributing unearned foreign incomes (i.e., foreign aid and oil rent). Foreign aid inflows can
be an enormous source of both economic and political benefits to members of the winning
coalition and/or sometimes even to the broader segment of population in a country. Aid
inflows in institutionalized autocracies create positive externalities to public goods such as
trade openness, foreign direct investment and economic development. Compared to other
types of autocracies, dictatorships with pseudo-democratic institutions are likely to spend
non-tax revenues from foreign aid inflows to deliver the better governance. In particular, more
institutionalized autocracies utilize foreign aid not only to benefit their patronage but also to
buy off potential political challengers outside the winning coalition. Hence, the relatively
better governance in institutionalized dictatorships may reduce the potential challengers’
incentive to resort to anti-American behaviors.
Autocratic leaders with pseudo-democratic institutions tend to pursue more extensive
and inclusive policies to co-opt ruling elites and a broader swath of society; domestic
constituencies have little incentive to divert their discontent with their governments that are
86
economically and militarily supported by the United States. In contrast, autocrats without
those kinds of institutions tend to rely more on co-optation of ruling coalition groups and
more repression of potential challengers and dissent groups. Gandhi and Przeworski (2007)
argued that “seemingly democratic institutions” such as legislature play an instrumental role
in making more extensive policy concessions by inducing cooperation from a larger segment
of population in a society. Specifically, institutionalized autocracies tend to focus more on
public goods such as trade openness (Hankla and Kuthy 2013), investment (Gehlbach and
Keefer 2011, 2012), and economic growth (Gandhi 2008; Wright 2008). Hankla and Kuthy
(2013) found that autocrats with more institutional constraints tend to rely on an open
economy, which may lead to economic development because of their longer time horizons
and relative regime stability. In addition, Wright (2008) maintains that dictators with more
binding institutions experience higher economic growth rates and domestic investment. He
also investigates the impact of foreign aid on economic growth in recipient countries. Wright
(2008) found that dictators with longer time horizons have an incentive to invest foreign aid
in public goods, which is strongly associated with good economic performance. Kim and
Gandhi (2010) demonstrated that autocratic regimes retaining various institutions face lower
levels of labor protests than military dictatorship because of the trade-off between material
benefits and workers’ cooperation with the regime. Wright (2008) argued that foreign aid
tended to promote economic growth in dictatorships with long time horizons, which led to
stable authoritarian regimes.
Again, autocrats buy off opposition groups to prevent or deter potential challenges or
revolutionary threats. Autocratic regimes with pseudo democratic institutions can effectively
co-opt potential challengers through this institutional channel. At the same time,
institutional autocracies tend to extract public support by investing more in public goods
such as investment, trade, and social spending. Those kinds of relatively inclusive policies
87
may improve regime legitimacy and instead, weaken the political power of some opposition
groups and/or revolutionaries that are excluded from the power-sharing political process.
As free resources such as foreign aid increase, institutional autocracies are more likely to
use them for the long-term economic performance and expansion of public goods because
of their tenures’ long time horizons. Consequently, institutional dictators can even gather
strong support from domestic constituencies. As revolutionaries or dissidents lose their
political legitimacy to oppose the autocratic incumbents, they tend to boost nationalist
sentiments by targeting foreigners or foreign countries. When autocrats are more dependent
on foreign aid, the revolutionaries target donor countries. Or they target foreigners or
Americans.
Some studies have posited that foreign aid helps incumbents win the elections, even in
autocracies. Gandhi and Lust-Okar (2009) discussed autocrats’ manipulation of elections to
ensure their political survival. They emphasized the cooptation of opposition groups in
autocratic countries. For example, autocrats take advantage of elections by “provid(ing)
mixed incentives to opposition parties, who may oppose the current dictatorship but also
want to benefit from the spoils of government” (Gandhi and Lust-Okar 2009, 405). Hence,
dictators who even allow the elections and multiple parties take advantage of foreign aid to
win elections, compared to opposition parties or groups. Incumbent autocrats possess
enormous resources in part from aid flows in the elections. For instance, Jablonski (2014)
examined the relationship between foreign aid and electoral outcomes by exploring how the
incumbent diverted foreign aid to draw political support from voters in the elections. This
may make some segment of opposition groups weak in terms of political power. Even when
incumbent dictators allow some opposition groups to join the existing political process but
exclude others, the excluded groups of the opposition can become radicalized.
88
Within a political system, according to Lust-Okar (2004), there are two types of groups-
the loyalists and the illegal opposition. The opposition can be divided into moderate and
radical groups. Moderate groups can be coopted by incumbent dictators through quasi-
democratic institutions. Incumbent dictators may allow moderates to participate in political
institutions and also distribute economic benefits and political privilege to them. As foreign
aid inflows increase, autocratic incumbents can strategically use this unearned income to
buy off the opposition moderates in part through power-sharing arrangements. However,
dictators still exclude other segments of the opposition, such as radical or extremist groups,
from the winning coalition. The selective co-optation and repression strategy in institutional
autocracies make the radicals or dissidents within opposition groups lose their public
support and become weak. The power-sharing argument posits that power sharing between
incumbent autocrats and their supporting elites through institutions tends to deter potential
rebellion. After all, institutions reduce autocrats’ commitment problems in terms of their
power-sharing promises to them. This power sharing even expands the size of the winning
coalition through coopting opposition groups (Boix and Svolik 2013; Svolik 2009).
This kind of institutional autocrats’ governance may make a certain opposition groups
radicalized. In other words, radicalized groups within the opposition cannot help, given their
weakened power, but resort to political violence to express their grievances or discontents.
In particular, when U.S. foreign aid inflows increase and help autocrats effectively co-opt
moderates within opposition groups, radical groups are excluded and have a weak political
stance. Thus, they are more likely to rely on violent behaviors against foreigners or a foreign
country in part to boost nationalism and a rally-round-the-flag. From the above two
competing arguments, I draw the following hypotheses:
89
Hypothesis 3.2 As U.S. foreign aid increases, autocratic countries with pseudo-democratic
institutions experience fewer anti-American incidents than those with no such institutions or
democra-
cies.
Hypothesis 3.3 As U.S. foreign aid increases, autocratic regimes with nominal democratic
institutions experience more anti-American incidents than do other political systems.
3.4 Research Design and Data
This section examines the systematic impact of U.S. aid on anti-American incidents. The
dependent variable is the count of anti-American incidents, which capture both violent and
nonviolent behaviors against Americans. They include protests, demonstrations, terrorist
activities and/or violence such as bombing, physical attacking, and kidnapping. The data for
analysis are based on a composite indicator using multiple sources rather than one separate
source. AntiAmerican incidents data come from three sources including the Global Terrorism
Database (GTD) of the Center for the Study of Terrorism and Responses to Terrorism, the
RAND Database of Worldwide Terrorism Incidents (RDWTI) of the RAND Corporation and the
reports of Political Violence against Americans released by the U.S. Department of State. For
the validation of measurement, the study adopts following strategies that cover 117
developing countries from 1970 to 2007, broken down into country-year observations. First,
the study takes into account only incidents that target Americans and/or U.S. owned
properties on purpose. On April 22, 1999, for instance, guerrillas with the Colombian
National Liberation Army (ELN) hijacked an airplane that carried 46 passengers including
some U.S. citizens. I exclude cases in which international terrorist groups that were not
bound in the country. I also exclude the cases in which international terrorist groups involve
90
with the multiple bases of their operations. For example, many religion-based terrorists such
as al-Qaeda in the Middle East have operated within a large number of countries from the
Middle East, Asia, Africa, and even Europe. Certainly, incidents against Americans carried
out by those kinds of terrorists should not be considered as anti-American violence due to
their relative lack of domestic political consideration in a given country’s territory. Regarding
coding strategies, I do not include the cases in which an attacker group’s nationality and/or
origins is different from that of an incident’s location and then reconfirm it after checking the
description of the incident in the GTD database, in particular. RDWTI dataset and PVAA
reports tend to describe the incidents’ information in detail and thus I can figure out the
incidents’ full stories. Third, I check the overlapped incidents cross datasets by taking into
account incidents’ time, location, targets, attackers, and incidents’ contents. For example,
when a similar incident occurs at same location with the same target and the same content
but different reported date, I consider is as the same incident.
Different reported data should be within less than a 2-day range. Practically, most of the
incidents across datasets share the same dates but are at least within a 1 or 2-day time
range.
3.4.1 Independent Variables
The purpose of this analysis is to evaluate the impacts of the U.S. foreign aid on anti-
American incidents in the developing world. The independent variable is U.S. foreign aid
representing U.S. aid (% GDP). U.S. aid/GDP (%) is a country’s annual inflows of U.S.
economic assistance as a percentage of gross domestic product (GDP). The data for U.S. aid
are collected by the AidData (Tierney et al. 2011). The study expresses U.S. aid as a
percentage of GDP because it not only is commonly used in aid studies but it also does
capture a recipient country’s aid dependence (Kono and Montinola 2009; Licht 2010).
91
In measuring the institutionalization of political regimes, the study makes use of the
regime indicator taken from Cheibub, Gandhi and Vreeland (2010)’s the Democracy and
Dictatorship (DD) dataset. The variable Institutions is measured as a trichotomous indicator,
which is coded 0 if a non-democratic regime does not have pseudo-institutions, 1 if a non-
democratic regime allows for nominal-democratic institutions such as legislatures and at
least one political party, and 2 as democracy (Kim and Gandhi 2010).
3.4.2 Control Variables
Previous literature has discussed potential determinants of violence against Americans.
First, economic sanctions may influence the frequency of anti-American political violence.
Economic sanctions is positively associated with the political violence against Americans
because they tend to negatively affect the socio-economic and political conditions of
ordinary people in a target nation. For the variable of U.S. economic sanctions (US sanction),
the study treats it as a dichotomy based on the fact that a state is either subject to sanctions
in a given country-year or not. The indicator is coded as 1 for every year that the sanctions
were in place and 0 otherwise. But, the variable is also one-year lagged because it can be
assumed that the last-year sanctions have realistic impacts on the target in current year. The
study uses the dataset of economic sanctions compiled by Hufbauer, Schott, Elliott and
Oegg (2007). Secondly, to measure Regime duration, the study uses “agereg” in the CGV
dataset that measures the total number of years that the current regime type has existed in
the country.
Another control is alliance between U.S. and weaker counterparts. The alliance should
also be a potential indicator to increase the motivations and incentives of a ally’s people to
express antiAmerican sentiment primarily based on the nationalist assumption (Tai,
Peterson and Gurr 1973). An asymmetric alliance (US defense alliance) can lead to anti-
92
American violence as it involves a more powerful country and a less powerful one. One
important distinction about an asymmetric alliance is that powerful nations are strong
enough to provide weaker counterparts with security guarantees. The study codes a country
as 1 if it is in an alliance with the U.S. and 0 otherwise. None of the authoritarian countries
are signatories of a symmetrical alliance with the U.S. The alliance data come from Alliance
Treaty Obligations and Provisions (ATOP) created by Leeds, Ritter, Mitchell and Long (2002).
Besides these, the study also includes a set of control variables widely used in explaining
political actions such as domestic protests or unrest and terrorism. First, the study uses GDP
per capita (log) to account for the impact of economic development on anti-American
terrorism. The data of GDP per capita are taken from the World Bank’s World Development
Indicators (WDI). In addition, population that also comes from WDI, is logged. Political
violence is measured as the “total summed magnitudes of all societal” conflicts including
civil violence, civil war, ethnic violence, and ethnic war. The data come from ‘Major episodes
of political violence and conflict regions, 1946-2012’ (the Center for Systemic Peace). To
control the effects of regional variation and the change of international structure, the study
uses region-fixed dummies from the World Bank and a dummy of Cold War.
The study also includes two different sources of foreign aid that may influence the
occurrence of anti-American incidents. When a country receives more foreign aid from other
countries or international institutions than from the United States, it may not experience
anti-U.S. sentiment. To test the impacts of other aid sources on anti-American violence, the
study include non-U.S. aid amount divided by the GPD of the recipient country (Non-US
aid/GDP (%)). Non-US aid measure is also taken from AidData (Tierney et al. 2011). Another
source of foreign aid is multilateral aid (Multilateral aid/GDP (%)). This is the total official
gross disbursements from multilateral aid agencies controlled by Development Assistance
Committee (DAC) in OECD (Kersting and Kilby 2014).
93
3.5 Methods and Result
The study examines whether U.S. foreign aid determines the frequency of anti-American
political incidents using a series of count-data estimation. Its dependent variable is a
nonnegative integer-counts of anti-American incidents. Ordinary Least Squares (OLS) is
appropriate if the dependent variable is independently and identically distributed. However,
the use of OLS for count outcomes can result in inefficient, inconsistent and biased
estimates if one or more OLS assumptions go unmet. Thus, statistical techniques other than
OLS regression have been used to deal with the count data. This study considers two
possible regressions-the Poisson regression and the negative binomial regression models.
First of all, the Poisson model assumes that the variance of Y is equal to the means. But, in
the study’s incident data there is high probability that the count variable has a variance
greater than the mean, which is called over-dispersion. In this case, the Poisson model is not
appropriate so it is necessary to employ the negative binomial regression model.
To begin with, the study checks whether or not the data has a Poisson distribution. If the
mean of the dependent variable (i.e., political incidents against Americans) is surely less
than the variance, then the candidate model will be a negative binomial model. The
descriptive statistics shows that the variance(5.943) is higher than the mean (0.494). For the
next step, I run the negative binomial regression model and then compare it with the Poisson
model using the Vuong test for the model selection. The test rejects the null hypothesis that
two models are not different. Thus, the negative binomial regression model-an alternative to
the Poisson model- is preferred. However, the study also considers the zero-inflated negative
binomial regression to deal with the excessive zeros because the large portion of the
dependent variable is zero counts (e.g., > 80 percent). For the model selection, an additional
Vuong tests is performed along with AIC/BIC post-estimation statistics to check the
comparative appropriateness of the two models. The test statistics show both Vuong
94
(z=7.07***) and AIC/BIC tests prefer the zero-inflated model to the standard negative
binomial model with the statistical significance. In addition, the hurdle model can be another
estimation to deal with many zeros. In the zero-inflated model, the zero values include two
different sources. The first source consists of countries that always have zero counts of anti-
U.S. violence (Long 1997; Zorn 1998). For instance, some countries had undergone anti-
American violence but not in a given year. Other countries were entirely free of anti-U.S.
incidents. That is, their experiencing anti-American violence was assumed to be on a
negative binomial distribution that includes both zero and non-zero counts. However, the
hurdle model assumes that the zero observations can come from only countries that always
have non-zero counts. The Vuong tests consistently prefer the zero-inflated negative
binomial model to the hurdle model (z=1.607*) (See Table 3.4 in the Appendix II).
The lagged values for all independent and control variables were used for two reasons.
On one hand, regressors are going to take some time to influence anti-American incidents.
Many studies on domestic violent behaviors (e.g., protests, demonstration, unrest, and
terrorism) tend to use lagged regressors for that reason (Murdie and Bhasin 2011; Pierskalla
and Hollenbach 2013). Secondly, to lag all explanatory variables help avoid potential
endogeneity problem and reverse causality (Garriga and Phillips 2014; Krieger and Meierrieks
2015).
Table 3.1 presents the main results of zero-inflated negative binomial (ZINB) models with
standard errors clustered on country. First, the study introduces the base-line analysis.
Model 1 represents the comparison between a democracy and an autocracy regarding the
outcome of antiAmerican violence. In the count equation, the predictor US aid/GDP (%) has
a coefficient of 0.251 with a statistical significance at 99.9% confidence level. This indicates
that for one-unit increase
Table 3.1: ZINB Regression of Anti-US Political Violence, 1970-2007: Full Sample
95
Count
Binary (zero)
Count
Binary (zero)
Count
Binary (zero)
US aid/GDP (%)
0.251***
0.245
0.028
-0.159
0.005
-0.424
(0.077)
(0.154)
(0.041)
(0.168)
(0.072)
(0.393)
Democracy
0.354*
0.795*
-0.176
-1.057
-0.157
-0.152
(0.211)
(0.462)
(0.229)
(0.683)
(0.254)
(0.771)
Institutional autocracy
-
0.518***
-2.045**
-
0.717***
-1.364
(0.182)
(0.840)
(0.228)
(1.025)
Democracy x US aid/GDP (%)
–
0.244***
-0.401**
-0.157
-0.152
(0.090)
(0.163)
(0.072)
(0.388)
Institutional autocracy x US aid/GDP
(%)
0.422***
1.136*
(0.155)
(0.595)
Regime duration
0.006
0.014
0.011**
0.046**
0.011*
0.029
(0.007)
(0.018)
(0.006)
(0.019)
(0.006)
(0.021)
Openness with US (log)
0.167
0.221
0.213***
0.314
0.214**
0.381
(0.111)
(0.337)
(0.080)
(0.244)
(0.096)
(0.260)
US defense alliance
1.722***
2.316*
1.692***
2.849**
1.809***
2.484*
(0.241)
(1.296)
(0.283)
(1.186)
(0.281)
(1.287)
US sanctions
0.260
-0.337
0.157
-0.875
0.208
-0.346
(0.174)
(0.571)
(0.171)
(0.706)
(0.182)
(0.600)
Political violence
0.042
-0.992
0.050
-2.496
0.059
-0.536*
(0.040)
(0.791)
(0.037)
(2.423)
(0.043)
(0.322)
GDP per capita (log)
-0.013
-0.510
-0.165
-0.849**
-0.084
-0.591
(0.169)
(0.380)
(0.134)
(0.391)
(0.177)
(0.379)
Population (log)
0.014
-0.953***
-0.010
-1.386***
-0.027
-1.143***
(0.106)
(0.277)
(0.087)
(0.333)
(0.101)
(0.295)
Cold war
0.239
-0.564
0.277**
-1.027**
0.282*
-0.435
(0.151)
(0.417)
(0.136)
(0.476)
(0.151)
(0.353)
Latin American & Caribbean
0.501
-1.507
0.509
-1.654
0.162
-2.361
(0.460)
(1.578)
(0.440)
(1.768)
(0.478)
(1.760)
Sub-Saharan Africa
0.920**
2.360*
0.655*
2.750
0.757*
2.004**
(0.389)
(1.254)
(0.368)
(1.699)
(0.429)
(0.920)
Middle East & North Africa
0.750
-0.033
0.415
-15.996***
0.157
-2.362
(0.591)
(1.819)
(0.487)
(2.290)
(0.576)
(2.679)
Asia
0.403
0.456
0.256
1.146
0.060
-0.443
96
(0.520)
(1.723)
(0.432)
(1.928)
(0.540)
(1.682)
Past anti-US violence
0.123***
-2.358***
0.128***
-2.008***
0.117***
-2.154***
(0.026)
(0.539)
(0.025)
(0.506)
(0.026)
(0.457)
Constant
-3.770**
10.419***
-2.321*
17.243***
-2.507
12.816***
(1.596)
(3.722)
(1.368)
(4.881)
(1.627)
(3.958)
Observations
3,282
3,282
3,282
Alpha
1.507
1.307
1.593
(0.200
(0.149)
(0.198)
lnalpha
0.410****
0.495***
0.369***
(0.132)
(0.121)
(0.133)
AIC
3722.135
3291.502
3717.696
BIC
3935.503
3437.383
3955.448
Robust standard errors in parentheses. Reference region is West.
*** p<0.01, ** p<0.05, * p<0.1
in U.S. aid/GDP, the expected log count of the number of anti-U.S. incidents increased by
0.251, given the other variables are held constant. In other words, for every additional
percentage in U.S. aid/GDP, an anti-American incident’s mean increases approximately by
28.6 %, holding all other predictors constant. Democracy variable shows that a democracy
experiences more anti-U.S. violence than an autocracy and is statistically significant. But,
the coefficients on the constituent variables of interaction terms should be conditionally
interpreted in the context of interactions. Regarding the interaction effect of U.S. aid/GDP
and democracy variable, Figure 3.1 shows that as U.S. aid increases, an autocracy is more
likely than a democracy to experience “No” anti-American violence in the zero-inflation
equation. In other words, an autocracy has a lower likelihood of experiencing anti-American
incidents than a democracy when it receives U.S. foreign aid. But, of countries experiencing
anti-U.S. incidents, an autocracy is expected to have a higher number of anti-American
incidents than a democracy as it is more dependent on U.S. aid (Figure 3.2).
Model 2 in Table 3.1 introduces the diverse effects of U.S. aid/GDP on anti-U.S. violence
in different political systems-democracy, institutionalized autocracy, and non-
institutionalized autocracy. First, in the count equation, regarding the effect of U.S. aid/GDP,
97
for a standard deviation increase in the percentage of U.S. aid/GDP, the expected count of
anti-U.S. violence increases by 4.8 %, holding all other variables constant. The coefficient,
however, is not statistically significant. In regard to the effects of different political systems,
being an institutional autocracy (compared to a non-institutional autocracy) multiplies the
expected count of anti-U.S. violence by 0.518, holding other explanatory variables constant.
In other words, an institutionalized autocracy decreases the number of anti-American
incidents by 40.4 % compared to a non-institutionalized autocracy.
Model 3 shows the interaction effect of U.S. aid and a recipient country’s political system.
The coefficients on the interaction terms between U.S. aid/GDP and a recipient’s political
system are positive and statistically significant at the level of 0.01 in the count model and 0.1
in the inflate model, respectively. First, in the count equation, the positive coefficient
indicates that when a recipient country is an autocracy under nominal democratic
institutions (institutionalized autocracy=1), an increase in the percentage of U.S. aid/GDP
leads to an increase in the expected count of anti-American violence. However, the
substantive interpretation of the interaction terms is complicated. Hence, the study shows
the marginal effects of US aid/GDP on the expected values of anti-U.S. incidents conditioned
by a recipient’s political institutions. Figure 3.2. displays the difference in the predicted
values of anti-U.S. violence against three different types of political institutions. When the
percentage of U.S. aid/GDP is zero, the expected value of anti-U.S. violence in institutional
autocracies is a little higher than both democracy and non-institutional autocracy. However,
as the U.S. aid/GDP percent increases, the expected count of anti-American violence
increases until U.S. aid/GDP is around 2 %. Once past this point, the predicted count of
antiAmerican incidents decreases. The expected value reaches almost zero as U.S. aid/GDP
becomes around 15 %. For non-institutionalized autocracies, the expected value of anti-U.S.
incidents increases as the percentage of U.S. aid/GDP grows.
98
Models 4-6 in the Table 3.2 show the results using the reduced sample in which the study
conducted several outlier tests such as Bonferroni-adjusted outlier tests. Based on those
tests, Columbia, Argentina, Chile, South Korea and Philippines were excluded from the
analysis. The main findings of Table 3.2 are similar to those of Table 3.1 with regard to the
interacted effect of U.S. aid/GDP and political systems.
Figure 3.1: Plot for ZINB regression: Inflation model
Figure 3.2: Plot for ZINB regression: Count model
99
Figure 3.3: Plot for ZINB regression: Inflation model
Figure 3.4: Plot for ZINB regression: Count model
100
Table 3.2: ZINB Regression of Anti-US Political Violence, 1970-2007: Reduced
Sample
Model 4: Baseline Model 5: Non-US Aid Model 6: Multilateral Aid
Count Inflate Count Inflate Count Inflate
US aid/GDP
(%)
-0.0002
-0.203**
-0.018
-0.216*
0.0004
-0.175
(0.042)
(0.101)
(0.058)
(0.127)
(0.048)
(0.116)
Institutional
autocracy
-0.494**
-1.980
-0.562**
-1.244
-0.644**
-1.782
(0.233)
(1.223)
(0.243)
(0.914)
(0.257)
(1.604)
Democracy
0.171
1.089
-0.117
-0.441
0.079
0.022
(0.270)
(0.899)
(0.298)
(0.793)
(0.284)
(0.900)
Institutional
autocracy x
US aid/GDP
(%)
0.269**
1.244***
0.422***
0.998**
0.410**
1.111**
(0.106)
(0.453)
(0.149)
(0.411)
(0.165)
(0.532)
Democracy x
US aid/GDP
(%)
-0.056
-0.903*
0.319*
0.395
-0.086
-0.759
(0.064)
(0.528)
(0.170)
(0.287)
(0.085)
(0.580)
Non-US
aid/GDP (%)
0.033
0.075*
(0.049)
(0.044)
101
Regime
duration
0.010
0.051**
0.007
0.019
0.007
0.027
(0.006)
(0.024)
(0.007)
(0.020)
(0.007)
(0.023)
Trade
openness
with US
0.177*
0.159
0.210*
0.363
0.248**
0.440
(0.101)
(0.369)
(0.119)
(0.333)
(0.119)
(0.358)
US defense
alliance
1.675***
3.278*
1.742***
2.589**
1.639***
2.696
(0.246)
(1.807)
(0.257)
(1.066)
(0.292)
(1.881)
US sanctions
0.159
-0.657
0.209
-0.221
0.216
-0.107
(0.184)
(0.769)
(0.187)
(0.635)
(0.203)
(0.770)
Political
violence
0.039
-1.861*
0.050
-0.595*
0.051
-0.646
(0.039)
(1.035)
(0.044)
(0.339)
(0.048)
(0.593)
GDP per
capita (log)
-0.095
-0.486
-0.020
-0.330
-0.113
-0.533
(0.153)
(0.369)
(0.196)
(0.355)
(0.202)
(0.447)
Population
(log)
0.006
-1.240***
-0.014
-
1.091***
-0.064
-1.231***
(0.114)
(0.359)
(0.126)
(0.365)
(0.138)
(0.387)
Cold war
0.221
-0.805
0.253*
-0.323
0.272*
-0.301
(0.142)
(0.534)
(0.144)
(0.375)
(0.154)
(0.438)
102
Latin
American &
Caribbean
0.356
-1.987
0.083
-2.346
-0.036
-2.963
(0.444)
(1.617)
(0.525)
(1.807)
(0.499)
(2.257)
Sub-Saharan
Africa
0.767*
3.369
0.813**
2.254**
0.728*
2.332
(0.429)
(2.295)
(0.405)
(1.118)
(0.425)
(1.770)
Middle East
& North
Africa
0.385
-6.432*
0.362
-1.899
0.229
-2.737
(0.531)
(3.454)
(0.600)
(2.551)
(0.606)
(2.911)
Asia
0.354
1.092
0.162
-0.378
0.042
-0.726
(0.424)
(1.831)
(0.543)
(1.554)
(0.547)
(2.036)
Past anti-US
violence
0.179***
-2.027***
0.151***
-
2.607***
0.158***
-2.604***
(0.033)
(0.507)
(0.033)
(0.478)
(0.039)
(0.653)
Constant
-2.890*
12.874***
-3.251
10.259**
-2.182
12.758***
(1.746)
(4.462)
(2.019)
(4.177)
(2.161)
(4.556)
Observations
3,093
3,093
3,024
Alpha
1.730
1.494
1.537
(0.251)
(0.250)
(0.324)
lnalpha
0.548***
0.401**
0.430**
103
(0.145)
(0.167)
(0.211)
AIC
3036.479
3040.19
BIC
3271.918
3287.703
Wald χ2
15.39***
14.91***
11.03**
Robust standard errors clustered on country in parentheses. Reference region is
West. *** p<0.01, ** p<0.05, * p<0.1
Regarding the impacts of control variables on the dependent variable, the study includes
three regressors that may influence anti-American incidents: U.S. sanctions, U.S. security
alliance, and U.S. trade openness. Those control variables may directly affect domestic
politics and socioeconomic conditions in US targeted countries or U.S. allies. Only U.S.
alliance is positively and significantly associated with anti-American violence across all
models. In the inflate equation, a country under the U.S. alliance system have a higher
likelihood of experiencing “no” violence against Americans. Of those experiencing anti-U.S.
violence, an increase in anti-American incidents in a hosting country results from a U.S.
security alliance. A country’s trade openness with the
U.S. also leads to the increase in anti-U.S. incidents in the count equation of Models 2 and 3.
But U.S. sanctions variable is not significant even if the causal direction is positive in the
count equation and negative in the inflate equation, respectively. This indicates that U.S.
sanctions are positively correlated with anti-American incidents. Regarding the effect of
general political violence, the political violence variable has a positive and significant impact
on anti-American incidents only in the zero equation of Model 3. In regard to regional effects,
countries in the Middle East and North Africa, for example, have a lower chance of
experiencing “always zero” incidents of anti-U.S. violence only in Model 2. The impact of
international structure (cold war dummy) is also significant. In Models 2 and 3, countries
have more expected count of anti-American incidents during the Cold War than during the
Post Cold War. The results reflects the fact that anti-U.S. violence frequently occurred during
the Cold War when the U.S. was often offering economic and political supports to dictators
in exchange for policy concessions (i.e., anti-communist policy).
104
3.6 Robustness Checks
To confirm whether my analysis is robust, a series of additional regressions were
conducted regarding both alternative specifications and methodological issues. First, the
study estimates the analysis using alternative measures of the dependent variables. The
analysis disaggregates antiAmerican violence into two separate incidents-anti-U.S.
government violence and anti-U.S. civilian violence (Models A3-A6). According to some
studies on anti-Americanism, this sentiment may hold an ambivalent attitude toward the
United States (Katzenstein and Keohane 2007; Diven 2007). People around the world tend to
criticize U.S. foreign policy and international influence while they accept U.S. culture and like
buying products made in U.S. (Diven 2007). So, once anti-American violence is said to have
resulted from the U.S. foreign policy (bilateral aid) impacting domestic politics, U.S. foreign
aid should have more of an impact on the anti-U.S. government violence than on anti-U.S.
civilian violence. To test this alternative explanation, the study measured anti-
U.S. government incidents as incidents in which the targets were U.S. government properties
(e.g., U.S. Embassy, U.S. culture center, etc.) and U.S. government officials. The study also
measured anti-American civilian violence as violent incidents in which the targets were
American civilian citizens and local branches/offices of U.S.-based multinational
corporations. The results show that an increase in U.S. aid leads to a decrease in expected
violence against U.S. government properties and/or officials in autocratic regimes having
institutions, compared to non-institutionalized autocracies. However, when a rise in U.S. aid
results in an increase in anti-U.S. civilian incidents in autocracies having institutions, it would
seem to indicate that revolutionary groups or extremist groups tend to target non-
government Americans such as U.S.-owned multi-corporations (e.g., McDonalds, Citibank
branches, etc.) rather than U.S. government properties because American civilian targets are
relatively easy to attack.
105
The study estimated zero-inflated models using civilization dummies instead of the
simple regional geographic dummies. At the same time, the study also took into account the
outliers’ influence by using both full sample and reduced sample excluding outlier states
(Berger 2014; Henderson and Tucker 2001; Huntington 1993). The findings in Models A7-A12
overall support the hypotheses. Regarding the effects of different civilizations, Islamic
countries have a lower probability of experiencing “no” anti-US violence at 0.1 confidence
level. Developing countries in the Western world also have a lower likelihood of experiencing
“no” anti-American incidents at 0.05 or 0.001 levels. However, among countries experiencing
anti-American incidents, countries in the West have fewer expected count of anti-American
violence.
Second, the study used two alternative indicators of U.S. aid to confirm the previous
results in Models A13-18 (i.e., U.S. aid/GNI (%) and U.S. economic aid/GDP (%). The overall
findings support the main results.
Finally, in addition to model specifications with lagged dependent variables (LDVs), the
study ran the models without LDVs for a robustness check for my statistical findings (Models
19-21). Some political methodologists raised doubts about the utility of using LDVs in the
model. They believe the inclusion of LDVs might increase the risk of making inaccurate
inference leading other regressors to lose their explanatory power (Achen 2000). The results
of count data models without LDVs shows that coefficients for the interaction term between
U.S. foreign aid and autocracy is still statistically significant and have more explanatory
power than models with LDVs in zero-inflated negative binomial regression models.
3.7 Conclusion
This study has investigated whether and/or how U.S. foreign aid leads to anti-American
incidents in a recipient country. Using the composite data based on several sources for 117
developing countries for the period between 1970 and 2007, the study has found the
106
evidence suggesting that anti-American incidents are incited by U.S. foreign aid conditioned
by autocrats’ survival strategy through their institutional channels. Compared to
democracies and non-institutionalized autocracies, in other words, institutions in
autocracies play an moderating role in influencing the frequency of anti-American incidents.
However, as the percentage of U.S. aid/GDP rises, so does the expected count of anti-U.S.
violence, though it soon decreases in autocracies with pseudodemocratic institutions.
The study found, however, that U.S. foreign aid itself does not affect the likelihood of
antiAmerican incidents regardless of its expected direction. Of those that do experience anti-
Americans incidents, U.S. foreign aid is expected to increase their frequency in some model
specifications. Nominal democratic institutions in autocracies influence the variation in
autocrats’ use of U.S. foreign aid. The number of anti-American incidents is affected by an
interaction between U.S. aid and autocratic institutions in recipient countries. There are
some policy implications in the findings. Pseudo-democratic institutions have a mediating
effect on the causal nexus between the harmful role of U.S. aid and anti-American violence.
The results may create several avenues for further research. First, to deal with the
endogeneity problem appropriate instrumental variables are needed. Even if this study tried
to cope with reverse causality or endogenous issues, it is not sufficient enough to check
reverse causality through the regression estimation on U.S. foreign aid/GDP (%). Also, the
lagged variable on the right side of the model equation partially, at most, dealt with the
endogeneity problem.
Second, despite its broader range of anti-American incidents, this project needs to add
more information on demonstrations and protests against U.S. government properties (e.g.
demonstrations in front of U.S. Embassies in recipient countries). Further study could focus
on collecting data on anti-American demonstrations and/or protests by looking into a variety
sources including media coverage (e.g., New York Times, BBC News, etc.)
107
Lastly, another future study could reflect the recent phenomena in the Middle East and
Africa. During the Arab Spring, people in several countries expressed anti-American
sentiments on the street using nonviolent and sometimes violent protests. So it may be
interesting to investigate the relationship between democratization or regime change and
anti-American incidents.
108
3.8 Appendix II
3.8.1 List of Countries in the Sample (117)
Albania, Algeria, Angola, Argentina, Armenia, Azerbaijan, Bahrain, Bangladesh, Belarus,
Benin,
Bhutan, Bolivia, Botswana, Brazil, Burkina Faso, Burundi, Cambodia, Cameroon, Central
African
Republic, Chad, Chile, China, Colombia, Comoros, Congo, Costa Rica, Croatia, Cyprus,
Czech
Republic, Congo, Djibouti, Dominican Republic, Ecuador, Egypt, El Salvador, Eritrea,
Ethiopia, Fiji, Gabon, Gambia, Georgia, Ghana, Guatemala, Guinea, Guinea-Bissau,
Guyana, Haiti, Honduras, India, Indonesia, Iran, Israel, Ivory Coast, Jamaica, Jordan,
Kazakhstan, Kenya, Kuwait, Laos, Lebanon, Lesotho, Liberia, Libya, Macedonia,
Madagascar, Malawi, Malaysia, Mali, Mauritania, Mauritius, Mexico, Moldova, Morocco,
Mozambique, Namibia, Nepal, Nicaragua, Niger, Nigeria, Pakistan, Panama, Papua New
Guinea, Paraguay, Peru, Philippines, Romania, Russia,
Rwanda, Saudi Arabia, Senegal, Sierra Leone, Slovakia, Slovenia, Solomon Islands, South
Africa,
South Korea, Sri Lanka, Sudan, Suriname, Swaziland, Syria, Tajikistan, Tanzania, Thailand,
Togo,
Trinidad and Tobago , Tunisia, Uganda, Ukraine, Uruguay, Uzbekistan, Venezuela, Vietnam,
Yemen, Zambia, Zimbabwe
3.8.2 Summary Statistics
Table 3.3: Summary statistics
Variable
Mean
Std. Dev.
Min.
Max.
N
Anti-US political violence
0.531
2.534
0
62
4490
US aid/GDP (%)
0.821
1.971
0
34.141
3763
Non-US aid/GDP (%)
3.123
5.308
0
78.804
3763
109
Multilateral aid/GDP (%)
3.225
5.012
0
86.173
3717
Regime type
0.89
0.74
0
2
4271
Regime duration
17.863
15.062
1
109
4271
Trade openness with US
5.466
2.418
0
12.881
4137
US defense alliance
0.263
0.441
0
1
4340
US sanctions
0.165
0.371
0
1
4229
Political violence
0.909
1.856
0
10
4088
GDP per capita (log)
7.165
1.198
3.913
10.49
3772
Population (log)
8.797
1.727
4.477
14.086
4262
Cold war
0.487
0.5
0
1
4490
110
3.8.3 Robustness Checks
1. Table 3.4: Comparison of ZINB and Hurdle Models
2. Table 3.5.: Alternative Dependent Variables: Anti-US Government and Anti-US Civilians
3. Table 3.5: Using civilization dummies (full sample)
4. Table 3.6: Using civilization dummies (reduced sample)
5. Table 3.7: Using an alternative measure of US aid - US Aid/GNI (%)
6. Table 3.8: Using an alternative measure of US aid - US Economic Aid/GDP (%)
7. Table 3.8: Without Lagged Dependent Variables
111
Table 3.4: ZINB Regression of Anti-US Political Violence, 1970-2007: Full Sample
Model A1: ZINB Model A2: Hurdle Count Inflate
(zero) Count Logit
US aid/GDP (%)
0.002
-0.424
-0.013
0.028
(0.069)
(0.359)
(0.030)
(0.030)
Institutional autocracy
-
0.708***
-1.293
-0.482*
-0.539***
(0.230)
(1.008)
(0.287)
(0.209)
Democracy
-0.179
-0.234
-0.116
-0.292
(0.259)
(0.829)
(0.280)
(0.211)
Institutional autocracy x US aid/GDP (%)
0.421***
1.110**
0.272**
0.029
(0.150)
(0.550)
(0.113)
(0.042)
Democracy x US aid/GDP (%)
0.010
0.284
0.005
0.018
(0.069)
(0.357)
(0.019)
(0.035)
Regime duration
0.012*
0.030
0.003
0.002
(0.006)
(0.021)
(0.007)
(0.005)
Trade openness with US
0.204**
0.385
0.121*
0.171*
(0.100)
(0.255)
(0.065)
(0.103)
US defense alliance
1.804***
2.405**
1.536***
1.254***
(0.278)
(1.166)
(0.267)
(0.297)
US sanctions
0.205
-0.390
0.326*
0.292
(0.186)
(0.616)
(0.176)
(0.192)
Political violence
0.061
-0.512*
0.168***
0.061
(0.043)
(0.280)
(0.038)
(0.047)
GDP per capita (log)
-0.084
-0.595
-0.021
-0.138
(0.180)
(0.382)
(0.120)
(0.146)
Population (log)
-0.029
-1.128***
0.278***
0.153
(0.098)
(0.286)
(0.080)
(0.107)
Cold war
0.274*
-0.469
0.459***
0.399**
(0.148)
(0.348)
(0.154)
(0.194)
Latin American & Caribbean
0.170
-2.275
-0.183
11.299***
(0.479)
(1.668)
(0.511)
(0.757)
Sub-Saharan Africa
0.747*
1.914**
-0.211
11.057***
112
(0.443)
(0.882)
(0.496)
(0.664)
Middle East & North Africa
0.123
-2.477
0.560
11.475***
(0.588)
(2.828)
(0.521)
(0.670)
Asia
0.074
-0.457
-0.516
10.695***
(0.546)
(1.594)
(0.496)
(0.755)
Past anti-US violence
0.118***
-2.132***
0.453***
0.113***
(0.026)
(0.449)
(0.065)
(0.024)
Constant
-2.420
12.766***
-6.046***
-14.460***
(1.607)
(3.957)
(1.247)
(1.465)
Observations
3,295
3,481
Alpha
1.319
(0.211)
lnalpha
0.373***
1.047***
(0.131)
(0.333)
AIC
3759.89
4125.987
BIC
Vuong z-statistic: 1.607* ZINB > Hurdle
3997.796
4366.035
Robust standard errors clustered on country in parentheses. Reference region is the West.
*** p<0.01, ** p<0.05, * p<0.1
113
Table 3.6: ZINB Regression of Anti-US Political Violence, 1970-2007: Full Sample
Model A7: Baseline Model A8: Non-US Aid Model A9: Multilateral Aid
Count Inflate Count Inflate Count Inflate
US aid/GDP (%)
-0.037
-0.547**
0.087
-0.131
-0.087
-0.664*
(0.062)
(0.260)
(0.107)
(0.170)
(0.088)
(0.394)
Institutional autocracy
-0.593**
-0.635
-0.488**
-0.247
-0.585**
-0.677
(0.235)
(0.679)
(0.246)
(0.704)
(0.236)
(0.673)
Democracy
-0.157
0.080
-0.057
0.355
-0.162
0.011
(0.234)
(0.628)
(0.255)
(0.667)
(0.234)
(0.617)
Institutional autocracy x US aid/GDP (%)
0.351***
0.725**
0.253
0.314
0.399***
0.846**
114
(0.119)
(0.281)
(0.163)
(0.203)
(0.134)
(0.384)
Democracy x US aid/GDP (%)
0.036
0.358
-0.075
-0.006
0.083
0.490
(0.059)
(0.259)
(0.118)
(0.205)
(0.082)
(0.368)
Non-US aid/GDP (%)
0.043
0.086
(0.080)
(0.074)
Multilateral aid/GDP (%)
0.050
0.019
(0.042)
(0.037)
Regime duration
0.013**
0.026**
0.012**
0.022*
0.013**
0.026*
(0.005)
(0.013)
(0.006)
(0.014)
(0.005)
(0.014)
Trade openness with US
0.157
0.177
0.153
0.167
0.154
0.181
(0.117)
(0.270)
(0.112)
(0.257)
(0.126)
(0.299)
US defense alliance
1.253***
0.089
1.304***
0.144
1.257***
0.066
(0.360)
(0.881)
(0.418)
(1.045)
(0.375)
(0.936)
US sanctions
0.143
-0.614
0.181
-0.602
0.179
-0.568
(0.201)
(0.575)
(0.203)
(0.612)
(0.206)
(0.588)
Political violence
0.061
-0.347*
0.064
-0.366
0.071
-0.328*
(0.046)
(0.198)
(0.051)
(0.235)
(0.048)
(0.172)
GDP per capita (log)
-0.199
-0.976**
-0.101
-0.731**
-0.123
-0.991**
(0.183)
(0.428)
(0.226)
(0.368)
(0.203)
(0.450)
Population (log)
-0.018
-0.811**
0.012
-0.731**
0.014
-0.824**
(0.129)
(0.363)
(0.133)
(0.316)
(0.141)
(0.371)
Cold war
0.254
-0.409
0.299*
-0.238
0.274
-0.455
(0.159)
(0.399)
(0.169)
(0.413)
(0.169)
(0.416)
Buddhist
-
1.187***
-1.163
-1.132***
-1.018
-1.078***
-1.139
(0.385)
(0.938)
(0.426)
(1.041)
(0.407)
(0.979)
Hindu
-0.427
-0.689
-0.304
-0.498
-0.335
-0.819
(0.540)
(4.401)
(0.477)
(3.065)
(0.482)
(3.038)
Islamic
-0.005
-1.175*
0.052
-1.086*
0.117
-1.084
(0.302)
(0.654)
(0.313)
(0.660)
(0.338)
(0.734)
Latin America
0.289
-0.846
0.280
-0.791
0.377
-0.727
(0.514)
(1.002)
(0.568)
(1.133)
(0.557)
(1.094)
115
Orthodox
-0.822*
-0.859
-0.749*
-0.702
-0.650
-0.857
(0.423)
(1.002)
(0.449)
(0.956)
(0.423)
(1.005)
Sinic
0.551
2.113**
0.418
1.980*
0.613
2.222**
(0.829)
(1.020)
(0.927)
(1.068)
(0.854)
(0.999)
Western
-1.335**
-21.416***
-1.235
-4.472
-1.078*
-17.114***
(0.643)
(0.944)
(0.987)
(3.441)
(0.648)
(0.988)
Others
-0.005
-0.725
0.020
-0.671
0.104
-0.698
(0.520)
(0.891)
(0.577)
(1.074)
(0.546)
(0.944)
Past anti-US violence
0.112***
-2.043***
0.111***
-2.224***
0.110***
-1.981***
(0.025)
(0.472)
(0.026)
(0.584)
(0.026)
(0.430)
Constant
-0.771
15.240***
-2.023
12.127***
-1.851
15.346***
(1.505)
(4.771)
(2.041)
(3.632)
(1.980)
(4.889)
Observations
3,295
3,295
3,268
Alpha
1.319
1.892
1.317
(0.211)
(0.271)
(0.200)
lnalpha
0.277*
0.283*
0.275*
(0.160)
(0.162)
(0.152)
AIC
3765.222
4071.151
BIC
4051.93
4266.357
Robust standard errors clustered on country in parentheses. Reference civilization is
African. *** p<0.01, ** p<0.05, * p<0.1
Table 3.7: ZINB Regression of Anti-US Political Violence, 1970-2007: Reduced
Sample
Model A10: Baseline Model A11: Non-US Aid Model A12: Multilateral Aid
Count Inflate Count Inflate Count Inflate
US aid/GDP (%)
-0.020
-0.484*
0.007
-0.447
-0.005
-0.425
(0.063)
(0.251)
(0.121)
(0.369)
(0.088)
(0.406)
Institutional autocracy
-0.540**
-0.501
-0.528*
-0.502
-0.602**
-0.495
(0.269)
(0.746)
(0.271)
(0.753)
(0.262)
(0.705)
Democracy
-0.120
0.161
-0.109
0.153
-0.107
0.228
(0.273)
(0.732)
(0.281)
(0.680)
(0.259)
(0.656)
Institutional autocracy x US aid/GDP (%)
0.329***
0.677**
0.296
0.625*
0.318*
0.620
(0.124)
(0.283)
(0.210)
(0.380)
(0.191)
(0.450)
Democracy x US aid/GDP (%)
0.031
0.298
0.014
0.253
0.014
0.236
116
(0.063)
(0.251)
(0.101)
(0.362)
(0.083)
(0.385)
Non-US aid/GDP (%)
-0.023
0.011
(0.049)
(0.062)
Multilateral aid/GDP (%)
-0.033
-0.013
(0.058)
(0.087)
Regime duration
0.013**
0.022
0.013**
0.020
0.012**
0.021
(0.006)
(0.016)
(0.006)
(0.015)
(0.006)
(0.016)
Trade openness with US
0.230*
0.254
0.229*
0.263
0.276***
0.343
(0.118)
(0.293)
(0.117)
(0.308)
(0.103)
(0.260)
US defense alliance
1.101***
0.373
1.077***
0.350
1.064***
0.322
(0.241)
(0.655)
(0.243)
(0.686)
(0.215)
(0.676)
US sanctions
0.283
-0.341
0.277
-0.354
0.273
-0.369
(0.223)
(0.528)
(0.221)
(0.587)
(0.216)
(0.522)
Political violence
0.041
-0.392
0.032
-0.436
0.041
-0.402
(0.047)
(0.344)
(0.065)
(0.587)
(0.051)
(0.277)
GDP per capita (log)
-0.294
-0.977**
-0.315
-0.955**
-0.396*
-1.113***
(0.192)
(0.475)
(0.196)
(0.445)
(0.209)
(0.429)
Population (log)
-0.077
-0.834*
-0.078
-0.821**
-0.145
-0.931***
(0.135)
(0.453)
(0.132)
(0.396)
(0.124)
(0.343)
Cold war
0.147
-0.493
0.137
-0.453
0.118
-0.512
(0.182)
(0.493)
(0.168)
(0.460)
(0.162)
(0.415)
Buddhist
-
1.017***
-1.278
-0.965*
-1.207
-1.078***
-1.296
(0.327)
(0.799)
(0.567)
(1.183)
(0.349)
(0.806)
Hindu
-0.365
-0.517
-0.375
-0.598
-0.459
-0.796
(0.973)
(7.781)
(0.645)
(5.993)
(0.568)
(3.621)
Islamic
0.046
-1.189*
0.070
-1.161*
-0.020
-1.264**
(0.323)
(0.627)
(0.483)
(0.628)
(0.320)
(0.591)
Latin America
0.292
-1.264
0.320
-1.220
0.206
-1.363*
(0.383)
(0.821)
(0.501)
(0.870)
(0.383)
(0.816)
Orthodox
-0.752
-0.947
-0.726
-0.832
-0.653
-0.861
(0.516)
(0.934)
(0.747)
(1.104)
(0.449)
(0.887)
117
Sinic
-2.077**
-3.219
-2.091**
-3.490
-2.094***
-3.320
(0.816)
(16.370)
(0.946)
(21.014)
(0.607)
(13.887)
Western
-1.278**
-16.467***
-1.195
-15.803***
-1.339**
-16.005***
(0.587)
(1.053)
(0.772)
(1.714)
(0.531)
(1.088)
Others
-0.014
-0.852
0.027
-0.772
-0.134
-1.027
(0.531)
(0.845)
(0.657)
(1.031)
(0.487)
(0.824)
Past anti-US violence
0.127***
-2.167***
0.129***
-2.191***
0.132***
-2.264***
(0.037)
(0.561)
(0.043)
(0.630)
(0.032)
(0.615)
Constant
0.002
15.080**
0.195
14.711***
1.269
16.460***
(1.726)
(6.125)
(1.838)
(5.222)
(2.080)
(4.850)
Observations
3,128
3,128
3,059
Alpha
1.212
1.220
1.195
(0.370)
(0.406)
(0.200)
lnalpha
0.192
0.199*
0.178
(0.305)
(0.332)
(0.285)
AIC
2956.39
2956.792
BIC
3240.653
3247.103
Robust standard errors clustered on country in parentheses. Reference civilization is
African. *** p<0.01, ** p<0.05, * p<0.1
Table 3.8: ZINB Regression of Anti-US Political Violence, 1970-2007: US Aid/GNI
Model A13: Baseline Model A14: Non-US Aid Model A15: Multilateral Aid
Count Inflate Count Inflate Count Inflate
US aid/GNI (%)
0.978
10.052
0.048
3.164
0.085
3.847
(1.297)
(7.606)
(1.444)
(5.857)
(1.690)
(6.689)
Institutional autocracy
-0.668***
-1.779*
-0.696***
-0.852
-0.703***
-1.020
(0.200)
(0.997)
(0.257)
(1.329)
(0.215)
(0.968)
Democracy
-0.044
0.050
-0.139
0.239
0.001
0.362
(0.211)
(0.711)
(0.330)
(1.182)
(0.226)
(0.742)
Institutional autocracy x US aid/GNI (%)
9.947***
13.566
14.384***
19.388*
14.558***
20.616**
(3.167)
(11.093)
(3.095)
(10.371)
(3.471)
(9.603)
Democracy x US aid/GNI (%)
-1.821
-34.988
-0.707
-12.848
-1.068
-11.408
(1.909)
(21.771)
(1.720)
(11.605)
(1.706)
(10.469)
118
Non-US aid/GDP (%)
0.059
0.102**
(0.049)
(0.050)
Multilateral aid/GDP (%)
0.040
0.045
(0.073)
(0.076)
Regime duration
0.009**
0.034
0.011
0.025
0.009**
0.030
(0.005)
(0.022)
(0.007)
(0.022)
(0.005)
(0.019)
Trade openness with US
0.252***
0.454
0.206*
0.347
0.269***
0.465
(0.076)
(0.362)
(0.108)
(0.322)
(0.088)
(0.298)
US defense alliance
1.917***
2.610***
1.812***
2.354*
1.839***
2.452
(0.235)
(0.999)
(0.272)
(1.250)
(0.315)
(1.650)
US sanctions
0.256*
-0.392
0.224
-0.448
0.339*
-0.076
(0.155)
(0.595)
(0.214)
(0.834)
(0.200)
(0.530)
Political violence
0.054
-1.949*
0.079
-0.503
0.079*
-0.478
(0.034)
(1.172)
(0.051)
(0.476)
(0.047)
(0.326)
GDP per capita (log)
-0.157
-0.768
0.001
-0.279
-0.085
-0.451
(0.119)
(0.473)
(0.208)
(0.404)
(0.168)
(0.337)
Population (log)
-0.027
-1.287***
-0.019
-1.047***
-0.058
-1.170***
(0.080)
(0.347)
(0.104)
(0.318)
(0.100)
(0.318)
Cold war
0.338**
-0.655
0.303**
-0.294
0.364**
-0.334
(0.150)
(0.501)
(0.153)
(0.395)
(0.162)
(0.391)
Latin American & Caribbean
0.033
-2.551*
0.181
-1.888
-0.044
-2.605
(0.420)
(1.494)
(0.490)
(1.801)
(0.471)
(2.047)
Sub-Saharan Africa
0.570
2.309
0.769
2.348*
0.711*
2.117**
(0.376)
(1.436)
(0.475)
(1.296)
(0.381)
(1.078)
Middle East & North Africa
0.315
-3.259
0.138
-1.889
0.199
-2.669
(0.398)
(2.311)
(0.687)
(3.638)
(0.450)
(2.488)
Asia
-0.045
-0.034
0.109
0.006
-0.085
-0.659
(0.421)
(1.528)
(0.546)
(1.720)
(0.508)
(1.876)
Past anti-US violence
0.127***
-1.937***
0.116***
-2.156***
0.115***
-2.096***
(0.021)
(0.440)
(0.027)
(0.544)
(0.025)
(0.495)
Constant
-2.380**
14.812***
-3.324*
8.888**
-2.724
10.860***
(1.113)
(4.973)
(1.816)
(4.159)
(1.689)
(3.524)
119
Observations
3,482
3,280
3,309
Alpha
1.570
1.442
1.419
(0.0.151)
(0.228)
(0.148)
lnalpha
0.451***
0.366**
0.350**
(0.096)
(0.158)
(0.148)
Robust standard errors clustered on country in parentheses. Reference region is the
West. *** p<0.01, ** p<0.05, * p<0.1
Table 3.9: ZINB Regression of Anti-US Political Violence, 1970-2007: US Economic
Aid
Model A16: Baseline Model A17: Non-US Aid Model A18: Multilateral Aid
Count Inflate Count Inflate Count Inflate
US economic aid/GDP (%)
0.037
0.931
0.018
0.283
0.002
-0.025
(0.047)
(1.105)
(0.025)
(0.231)
(0.026)
(0.067)
Institutional autocracy
-0.514***
-1.379*
-0.649***
-1.935**
-
0.683***
-0.960
(0.188)
(0.832)
(0.200)
(0.911)
(0.219)
(0.998)
Democracy
-0.057
-0.267
-0.198
-0.730
-0.044
0.200
(0.209)
(1.065)
(0.228)
(0.670)
(0.223)
(0.640)
Institutional autocracy x US economic aid/GDP
(%)
0.149
-1.078
0.326**
0.523
0.602***
0.902**
(0.107)
(1.204)
(0.135)
(0.390)
(0.203)
(0.447)
Democracy x US economic aid/GDP (%)
-0.030
-1.413
-0.004
-0.396
0.001
-0.128
(0.053)
(2.164)
(0.024)
(0.400)
(0.020)
(0.329)
Non-US aid/GDP (%)
-0.019
-0.069
(0.023)
(0.048)
Multilateral aid/GDP (%)
0.024
0.024
(0.071)
(0.080)
Regime duration
0.010**
0.041**
0.013**
0.043**
0.009*
0.029
(0.004)
(0.019)
(0.006)
(0.019)
(0.005)
(0.022)
Trade openness with US
0.222***
0.254
0.203***
0.258
0.256***
0.422
(0.071)
(0.285)
(0.077)
(0.235)
(0.087)
(0.282)
US defense alliance
1.902***
2.829***
1.744***
2.655***
1.836***
2.328
(0.246)
(0.926)
(0.246)
(0.913)
(0.329)
(1.787)
US sanctions
0.231
-0.424
0.185
-0.714
0.298
-0.169
120
(0.155)
(0.708)
(0.171)
(0.623)
(0.200)
(0.531)
Political violence
0.046
-3.410
0.045
-2.018**
0.072
-0.468*
(0.037)
(2.341)
(0.038)
(0.965)
(0.045)
(0.244)
GDP per capita (log)
-0.190*
-0.668
-0.167
-0.696*
-0.118
-0.512
(0.111)
(0.513)
(0.134)
(0.392)
(0.183)
(0.370)
Population (log)
0.001
-1.322***
-0.029
-1.260***
-0.035
-1.088***
(0.078)
(0.415)
(0.086)
(0.316)
(0.099)
(0.308)
Cold war
0.311**
-1.098**
0.212
-1.063**
0.334**
-0.366
(0.139)
(0.466)
(0.143)
(0.461)
(0.161)
(0.358)
Latin American & Caribbean
0.296
-1.549
0.310
-2.000
0.019
-2.374
(0.413)
(1.391)
(0.395)
(1.346)
(0.472)
(2.058)
Sub-Saharan Africa
0.661*
3.042**
0.563
2.437
0.683*
1.878*
(0.361)
(1.536)
(0.367)
(1.583)
(0.367)
(0.973)
Middle East & North Africa
0.574
-5.892***
0.183
-5.471**
0.280
-2.603
(0.411)
(2.084)
(0.490)
(2.629)
(0.454)
(2.641)
Asia
0.153
1.330
0.165
0.826
-0.070
-0.624
(0.415)
(1.588)
(0.400)
(1.579)
(0.505)
(1.823)
Past anti-US violence
0.129***
-2.090***
0.128***
-1.931***
0.115***
-2.193**
(0.022)
(0.634)
(0.024)
(0.441)
(0.025)
(0.895)
Constant
-2.417**
15.002**
-1.827
15.630***
-2.562
11.078***
(1.104)
(6.054)
(1.434)
(4.827)
(1.820)
(3.840)
Observations
3,481
3,305
3,308
Alpha
1.420
1.615
1.622
(0.213)
(0.479)
(0.1888)
lnalpha
0.484***
0.479***
0.351**
(0.096)
(0.110)
(0.150)
Robust standard errors clustered on country in parentheses. Reference region is the
West. *** p<0.01, ** p<0.05, * p<0.1
Table 3.10: ZINB Regression of Anti-US Political Violence, 1970-2007: With No
LDVs
Model A19: Baseline Model A20: Non-US Aid Model A21: Multilateral Aid
Count Inflate Count Inflate Count Inflate
US aid/GDP (%)
-0.038
-1.766***
-0.042
-1.788***
-0.041
-1.763***
(0.030)
(0.634)
(0.031)
(0.588)
(0.032)
(0.582)
121
Institutional autocracy
-
0.696***
-4.537**
-
0.712***
-4.391**
-0.677***
-4.450**
(0.225)
(2.302)
(0.218)
(1.720)
(0.220)
(1.999)
Democracy
-0.007
-1.673
-0.107
-1.723*
0.046
-1.400
(0.247)
(1.033)
(0.256)
(1.042)
(0.251)
(0.993)
Institutional autocracy x US aid/GDP (%)
0.295***
2.441***
0.306***
2.449***
0.307***
2.514***
(0.051)
(0.918)
(0.057)
(0.749)
(0.051)
(0.795)
Democracy x US aid/GDP (%)
0.009
0.465
0.016
0.382
0.010
0.455
(0.025)
(1.307)
(0.025)
(1.190)
(0.027)
(1.251)
Non-US aid/GDP (%)
-0.010
-0.0009
(0.026)
(0.076)
Multilateral aid/GDP (%)
-0.032
-0.413*
(0.0225)
(0.234)
Regime duration
0.006
0.059**
0.008
0.059*
0.007
0.067**
(0.007)
(0.029)
(0.008)
(0.032)
(0.007)
(0.034)
Trade openness with US
0.188*
0.353
0.138
0.356
0.171
0.155
(0.100)
(0.344)
(0.105)
(0.285)
(0.105)
(0.433)
US defense alliance
2.135***
3.247
1.959***
2.538
2.045***
2.919
(0.267)
(2.458)
(0.291)
(2.320)
(0.289)
(2.896)
US sanctions
0.311
-1.675
0.269
-1.701
0.341
-1.573
(0.222)
(1.349)
(0.218)
(1.274)
(0.223)
(1.252)
Political violence
0.114**
-5.497*
0.113**
-5.526**
0.119***
-5.949**
(0.046)
(3.239)
(0.047)
(2.715)
(0.046)
(2.924)
GDP per capita (log)
-0.026
-0.971*
0.083
-1.062*
0.010
-0.778
(0.135)
(0.520)
(0.151)
(0.641)
(0.140)
(0.528)
Population (log)
0.117
-1.884***
0.135
-1.936***
0.116
-1.753***
(0.114)
(0.521)
(0.131)
(0.530)
(0.115)
(0.515)
Cold war
0.474***
-1.123*
0.423**
-1.115*
0.466***
-1.124*
(0.177)
(0.622)
(0.178)
(0.592)
(0.176)
(0.594)
Latin American & Caribbean
0.673
-1.854
0.898*
-1.032
0.819
-0.486
(0.517)
(1.634)
(0.526)
(2.513)
(0.515)
(2.234)
Sub-Saharan Africa
0.701
3.958**
0.732*
3.922*
0.775*
4.812**
122
(0.433)
(1.832)
(0.420)
(2.161)
(0.428)
(2.016)
Middle East & North Africa
0.558
-21.593***
0.504
-20.491***
0.637
-28.086***
(0.468)
(4.890)
(0.512)
(3.951)
(0.468)
(5.219)
Asia
0.046
1.645
0.265
2.190
0.182
3.036
(0.511)
(1.429)
(0.521)
(2.326)
(0.514)
(2.100)
Constant
-
4.645***
22.352***
-
5.234***
23.432***
-4.895***
19.776***
(1.604)
(7.044)
(1.915)
(8.419)
(1.700)
(7.128)
Observations
3,472
3,333
3,365
Alpha
1.420
1.615
1.622
(0.213)
(0.479)
(0.1888)
lnalpha
0.865***
0.895***
0.847***
(0.157)
(0.150)
(0.156)
Robust standard errors clustered on country in parentheses. Reference region is the
West. *** p<0.01, ** p<0.05, * p<0.1
Chapter 4
Making Dictators’ Pockets Empty: How U.S. Sanctions Influence
Social Policies in Autocratic Countries?
Abstract. This work examines how U.S. economic sanctions affect social welfare spending in
authoritarian countries. U.S. economic sanctions play a role of inducing autocratic targets to
change social policy through two theoretical channels. First, U.S. economic sanctions may
reduce autocrats’ resources to buy off supports from ruling elite groups and so force
autocrats to reallocate government expenditure in favor of their supporting groups.
Consequently, autocrats facing longer U.S. sanctions are likely to cut spending on public
goods and services, especially on education and health care spending. Second, the impacts
of U.S. sanction duration on social spending varies according to political variables such as
autocrats’ pseudo-democratic institutions. The empirical findings show that, even when U.S.
123
sanctions last a long time, autocrats under nominal democratic institutions cut spending on
education and health to a lesser degree than do autocrats with no such
institutions.
4.1 Introduction
Do U.S. economic sanctions have substantial effects on social welfare policies in
authoritarian targets? The conventional wisdom demonstrates that “sanctions often
produce unintended and undesirable consequences” for the society of sanctioned countries
(Haass 1997). According to the literature on sanctions, economic sanctions reduce a
targeted state’s available resources for public goods and services (Wood 2008). Especially,
authoritarian countries seem to be worse off after sanctions than before (Brooks 2002;
Haass 1997). Are all autocracies equally bad at maintaining social welfare policies when they
face sanctions? This study offers a new theoretical and empirical study on the impact of U.S.
economic sanctions on social welfare spending by looking into the variations among
targeted autocratic states. The existing literature on sanction studies pays attention to the
counterproductive effects of sanctions on target states. Scholars find that economic
sanctions fail to extract policy concessions from a target country while producing unwanted
negative externalities (Allen and Lektzian 2013; Brooks 2002; Escribà-Folch and Wright 2010;
Peksen 2009, 2011; Peksen and Drury 2009, 2010; Wood 2008). For instance, Brooks (2002)
demonstrated that sanctions retain “distributional implications” for the social welfare of
different political actors within a target nation while they have “adverse macroeconomic
effects” on its entire population. Sanction costs hurt the general population, who tend to feel
deprived of basic needs (e.g., consumption goods including food, water, etc.) and
consequently, are more likely to act out against the government (Allen 2008b). Scholars also
place an emphasis on the conditional effects of political institutions on the costs of
sanctions. Democratic targets are easily affected by the counterproductive costs of
124
sanctions due to their high sensitivity to “domestic audience costs” (Allen 2008a). So,
democratic leaders are more likely than authoritarian rulers to concede the demands of a
sender, as sanctions aggravate the macroeconomy and thus hurt the general population. As
a result, a democratic target may be less likely to suffer economic sanctions while
experiencing shorter duration of sanctions (Lektzian and Souva 45; McGillivray and Stam
2004). In contrast, autocrats often resist economic sanctions because of their relative lack
of political constraints. Making concessions to sanctioning states may not only weaken
authoritarian leadership but threaten their political survival as well. Therefore, autocrats’
resistance to economic sanctions forces the ordinary citizens to suffer sanction costs (Allen
2008a; Lektzian and Souva 2007).
A growing number of sanction studies seek to evaluate the negative externality of
sanctions against a target country, mostly looking into its variation according to regime type-
democracy versus autocracy. However, scholars of international politics have recently
focused on authoritarian countries adopting a variety of research topics including
international trade, foreign direct investment, and public policies. They find that autocratic
regimes show significant variation in their political and economic performances. By the
same token, we might expect that not all autocracies implement equally well social welfare
policies as they respond to U.S. economic sanctions. Second, the majority of sanction
targets are non-democratic countries. According to Escribà-Folch and Wright (2010, 336)
and Kaempfer, Lowenberg and Mertens (2004), 85% of U.S. sanctions targets are “not-fully-
democratized” countries. One reason for this pattern is that democratic leaders are more
likely “to accommodate the sender’s demands” because they are more vulnerable than their
authoritarian counterparts to economic costs (Hufbauer et al. 2007, 166-167). In this sense,
democratic regimes are more likely to make concessions to the demands of the senders after
sanction imposition or even under sanction threats to prevent and/or lower the negative
economic and political consequences of sanctions.
125
As noted above, autocrats tend to experience and resist economic sanctions for a longer
time. Authoritarian rulers can make use of foreign sanctions for a longer tenure by
scapegoating the sender. In this respect, autocrats rely heavily on the use of nationalistic
sentiments and “rallyround-the-flag” events. They also have a strong incentive to reallocate
their government resources in favor of satisfying their core supporting groups and/or
repressing the domestic public who may have grievances over lower living standards that are
a result of economic sanctions. In the end, economic sanctions can considerably affect
autocrats’ social welfare policies. The effect is varied, however, according to political
institutions.
This study shows that a dictatorship’s expenditure on social welfare goods and services
depends on U.S. sanction duration and its political institutions. When dictators suffer
economic sanctions that are sustained for a long time, the rulers have no choice but to
reallocate their government revenues in public expenditure to ameliorate the negative
impacts on their political survival. For example, a trade embargo imposed by the U.S. may
worsen a national economy in a non-democracy. The economic devastation results in huge
socioeconomic problems such as skyrocketing unemployment and higher costs for basic
needs. Thus, an authoritarian government will face political pressure to reallocate its
spending on social security, education, and public health. Dictators may lose their free
sources because of a specific sanction (i.e., the suspension of foreign aid). Authoritarian
regimes in the third world produce weak economies and inefficient governance. In those
conditions, losing free income from abroad makes it difficult for autocrats to provide enough
private goods to members of their ruling coalition. Dictators cut their expenditures on public
goods to compensate for the huge losses of private goods to their ruling elites. It may thus be
expected
126
commitments to public goods and services that may be aggravated by foreign economic sanctions (Bueno de
Mesquita et al. 2003; Fearon 1994) Thus, democratic regimes become the sanction targets during a shorter
time compared to autocracies.
that an autocrat under U.S. economic coercion will reduce spending on social welfare goods
and services, expecting and/or experiencing economic hardships from sanctions.
However, the diversity of a political system within authoritarian countries leads to diverse
policies of social welfare. Autocrats ruling with “some degree of (political) institutional
constraints” and autocrats free of such constraints will implement quite different policies.
Some dictators may strategically create political institutions to mimic ones in democracies.
Those quasidemocratic institutions (e.g., political party and legislature) enable autocrats to
survive longer in power. The institutions do this by providing the society relatively higher
levels of public goods and by strengthening regime stability through the institutionalization
of leadership change. Even under economic hardship, autocrats having institutions may
expand the proportion of social welfare spending over government expenditure, comparing
with autocrats having no such institutions. At the least, pseudo-democratic regimes refrain
from sharply decreasing social spending, in part because they care about the public
discontent and support for the revolutionary movement. It is thus hypothesized that U.S.
sanction duration is positively associated with the changes in social expenditure when
autocracies are constrained by pseudo-democratic institutions.
To test this argument, a series of econometric estimations are conducted to evaluate the
impacts of U.S. sanction duration on changes in social welfare spending of autocratic
governments. This study collects social spending data of 95 countries under dictatorial rule
from 1970 to 2007. As concrete evidence and to minimize missing value problems, the study
fills gaps in the data by using multiple imputations. A matching estimation is also conducted
to improve the causal inference of the model by incorporating the potential outcomes in
quasi-experimental settings. The empirical evidence confirms the diverse impacts of U.S.
127
economic sanction duration on considerable variation in social spending among autocracies
even with several controls.
This essay is divided into four sections. The first section briefly reviews of the negative
externalities of economic sanctions against a target country. The second section discusses
theoretical arguments regarding the impacts of sanctions on social welfare policies in
autocratic targets. It also puts forward testable hypotheses. The third section lays out the
research design and explains the measurement and the methodological techniques. As a
statistical model, the study employs a panel error-correction regression with fixed effects
using a multiple imputation to avoid inference problems from a large proportion of missing
values. The section presents the regression results and their substantive interpretation using
a matched dataset. Finally, I conclude this paper with a discussion of policy implications of
the analysis and recommendations for the future research.
4.2 Literature Review
A growing number of quantitative studies have paid attention to the negative externalities
of economic sanctions against target nations. Previous scholarship on economic sanctions
has mainly focused on sanction effectiveness or sanction outcomes, asking whether
economic sanctions extract targets’ policy concessions. In contrast, recent studies on the
political-economic consequences of economic sanctions have found that sanctions
negatively influence the subjects outside winning coalitions in authoritarian
countries.Existing studies on sanction initiation show that “democratic targets are more
likely to be held responsible for economic failings and are more likely to view the threatened
costs of sanctions as sufficiently severe. So, while democracies use sanctions more
frequently, they tend not to use them against other democracies as often as they use them
against autocracies” (Lektzian and Souva 2007, 856).
128
Recent studies has maintained that economic sanctions inadvertently harm the
socioeconomic
and political conditions of a target nation. Such studies may be divided into two broad topics
of interest: (1) sanction effects on social welfare conditions in targeted countries; and (2)
sanction effects on political environments of targeted countries. On one hand, scholars have
found that economic sanctions worsen public health conditions-such as infant or child
mortality, and immunization (Peksen 2011)- and government commitments to public goods
in the sanctioned states (Allen and Lektzian 2013). The second group of scholars have
attempted to test the impacts of sanctions on political freedom (Peksen 2009; Peksen and
Drury 2009), political repression/terrorism (Choi and Luo 2013; Escribà-Folch 2012; Wood
2008), and democratization (Peksen and Drury 2010). Most such studies have found that with
regard to the aforementioned socio-political outcomes, economic sanctions have brought
devastating consequences to the public in sanctioned states.
Arguments on sanctions’ devastating effects demonstrate that economic sanctions
worsen the social welfare environment for the public, weakening a target state’s public goods
and services. Peksen (2011) evaluated how economic sanctions affected the health
conditions of civilians using the child mortality rate as an indicator. Economic hardships
resulting from sanctions increase unemployment, inflation and poverty. Those unhealthy
conditions of a target’s economy “reduce people’s ability to afford health-care services to
maintain a healthy life and standard of living” and force “the government to cut health-care
services” (Peksen 2011, 240). Peksen (2011) found that U.S. sanctions and sanction costs
significantly and consistently increase child mortality rates, confirming his hypothesis of
sanctions’ detrimental effect on public health conditions. Another study attempted to
analyze more indicators of public health conditions including several immunization rates, life
expectancy, and food availability (Allen and Lektzian 2013). Allen and Lektzian (2013, 123)
maintained that reduced resources caused by economic sanctions inevitably led to
129
“allocation decisions” that negatively and indirectly influenced health outcomes in targeted
countries. However, their study produced relatively mixed findings in different model
specifications, which can neither confirm nor disconfirm the negative impacts of sanctions
on public health.
Other scholars have focused on the reallocation of government expenditures on social
welfare responding to economic sanctions. For example, Allen and Lektzian (2013)
estimated the impact of sanctions on government spending on public health, finding a
negative externality of sanctions. Interestingly, they took a sample of year-cases under
military conflict and evaluated whether sanctions reduced government spending on public
health. The result showed that high-cost and lowcost sanctions both significantly led
governments to decrease their health expenditures. They also found that democracy has a
positive effect on government health spending across several model specifications.
Focusing mainly on authoritarian responses to economic sanctions, Escribà-Folch (2012)
suggested that economic sanctions might affect autocratic governments’ decisions to
redistribute their public goods as they confront a sanctions-induced scarcity of resources.
Escribà-Folch (2012) found that in all autocratic types the total government expenditure
decreased under sanction imposition while single-party regimes increased government
spending even under economic sanctions to a greater degree than did personalist regimes
(the reference-regime in the model). He interpreted the result as single-party and military
regimes “benefit[ting] their main support groups and thus negate(ing) the destabilizing effect
of an increase in the price of loyalty brought on by sanctions” (Escribà-Folch 2012, 699).
Despite his findings, the analysis with many missing values lost a considerable amount of
information in the dataset, increasing the risk of inefficient and biased estimation (King et al.
2001; Honaker and King 2010).
Sanction studies also have an interest in the effect of economic sanctions on political
environments in a target nation. Those works primarily focus on political freedom,
130
democratization and political repression. Scholars of sanctions have evaluated whether and
how economic sanctions worsen political freedom and increase state repression within the
targeted countries (Peksen 2009; Peksen and Drury 2009; Wood 2008). Wood (2008)
estimated, for the first time, the relationship between economic sanctions and state
repression, suggesting that economic sanctions tended to increase incentives for target
incumbents to repress potential challengers and public discontents. In other words,
incumbent leaders in a sanctioned country take advantage of using sticks against opposition
groups and civilians so as to stabilize the country and set their core supporters at ease under
gloomy effect of sanctions. He finds that “the results of the interaction terms for sanctions
imposed on democracies provide partial support for (the hypothesis) on the mitigating effect
of democratic institutions” in part because “democratic and autocratic states respond
differently to sanctions events” (Wood 2008, 504). In addition, democracies tend to
decrease, compared with autocratic regimes, their use of political repression even under
economic sanctions imposed by the U.N. In a similar study, Peksen (2009) found that
sanctions led to an increase in human rights abuses represented by physical integrity rights
including disappearances, extrajudicial killings, political imprisonment, and torture. The
difference between Wood (2008) and Peksen (2009) is the fact that the latter makes use of
disaggregated indicators of human rights abuses in addition to two composite indicators
such as physical integrity index and political terror scale; the former uses only political terror
scale as his outcome variable. Peksen (2009) found that economic sanctions increased the
human rights abuses in targeted nations across various model specifications. Unlike Wood
(2008), however, he does not look into the conditional effect of economic sanctions on
human rights given the variation in states’ political systems.
Despite numerous studies on the negative externalities of sanctions, relatively few have
paid attention to variation in autocrats’ commitments to public goods under economic
sanctions. Most studies focus on variation in government policies between democracy and
131
autocracy. Previous findings have shown that democracies tend to improve public goods for
the ordinary population and increase expenditures for them and a high portion of economic
sanctions have been imposed against non-democratic regimes. Thus, existing empirical
results do little to provide new and interesting policy implications (King, Keohane and Verba
1994). For these reasons, the project focuses on the cases of non-democratic countries and
their policy commitments under the influence of economic sanctions.
4.3 Theory: U.S. Sanction Duration and Social Welfare In
Autocratic Targets
Under what conditions does U.S. sanction duration influence variation in social welfare
policy within autocratic countries? As found in the literature, autocratic regimes provide
ordinary civilians with relatively lower levels of public goods and services. Regarding social
services, nondemocracies seem to lack the capacity to reduce poverty, infant mortality, and
illiteracy. When the sender imposes economic sanctions to a target state, target leaders
should decide whether to make a concession to the sender’s demands. If conceding to the
sender’s demands, a target leader can get more rewards from the sender through the direct
assistance or the withdrawal of trade sanctions. If not, sanction costs will be influential.
Sanctions aggravate the social-economic conditions of ordinary people rather than political
elites in targeted countries (Allen 2008b; Lektzian and Souva 2007). For example, economic
sanctions may lead some authoritarian rulers to bring hardship to their people by raising
taxes and more repression (Escribà-Folch and Wright 2010).
Marinov (2005, 571-572) found that democratic leaders were more likely to be susceptible to
economic sanctions than autocratic leaders. After all, “domestic publics if they could,
replace their leaders” if sanction were more costly to the target state. It should be noted that
a large proportion of sanctioned states are in fact non-democracies and less developed
132
countries. It is said that autocratic countries tend to spend less on social welfare services
and public goods than democratic ones. When targeted with sanctions, authoritarian rulers
can, with a relatively small winning coalition, change social policies easier than can their
democratic counterparts (Bueno de Mesquita et al. 2003). When economic sanctions are
imposed against targeted autocracies, those autocracies are supposed to decide how to
allocate governments’ resources in order to decrease both political and economic costs
caused by sanctions. Those kinds of sanction costs are closely associated with dictators’
political calculation in regard to political survival. This section presents the theoretical
mechanisms by disaggregating the effects of economic sanctions on social policies in
autocracies.
Economic sanctions play a devastating role, through a variety of channels, in hurting
economic wealth and changing political environments. First, a foreign embargo on import
and export generate a negative externality to national economy. That is, a trade embargo on
export reduces the export of industries in authoritarian states. Bans on exports tend to give
rise to skyrocketeting unemployment rates and lead to the bankruptcy of whole industries.
The negative impacts of a trade embargo would be worse to economies containing less
competitive industries. The subsequent increase in unemployment leads to a reduction in
household income, which gives rise to growing poverty and the consequent malnutrition of
children in poor households. In particular, the bankruptcy of domestic industries and the
consequent high unemployment carry a heavy impact for many unskilled workers (specially
women) who work mostly in labor-intensive industries in developing countries. For example,
the U.S. trade embargo on Myanmar worsened the economic conditions of garment workers,
who “earn as little as 30 cents a day” and forced many girls and young women to “turn to
prostitution or other marginal means to make a living” (Welsh 2003). In response to those
kinds of economic difficulties, governments must decide whether to increase social
spending to minimize the disastrous impacts on the public of a trade embargo on the public.
133
However, autocrats are more concerned about maintaining power than responding to the
costs of sanctions that influence the life of the general population. As the duration of
economic sanctions persists, dictators have a strong incentive to reallocate limited
resources at least to compensate-by reducing government expenditure on public goods-for
the loss of private goods for their coalition members. In addition, autocrats manipulate
information and propaganda as they appease a discontented public, by diverting sanction
costs to sanctioning states and at the same time boosting nationalism (Brooks 2002; Byman
and Lind 2010).
Another theoretical rationale comes from aid reduction or suspension. Since the end of
the World War II, U.S. foreign assistance has been ranked number one in foreign aid. Cutting
off U.S. economic/military aid robs dictators of their “free resources” that mostly flow into
sources of private goods for winning coalition groups (Escribà-Folch and Wright 2010). To
compensate for the loss of foreign assistance, dictators have a strong incentive to reallocate
sources of public goods to private goods by reducing education and health expenditures of
government consumption. Without free resources, dictators can no longer offer enough
financial sources to allow their supporting groups to buy and enjoy imported products
including luxury commodities. In this context, ruling elites give much weight to potential
losses to their existing benefits from economic sanctions. So to buy their support for the
regime, they are likely to force incumbent autocrats to reallocate government resources to
compensate for the lack of free resources. The above logic outlines the general allocation
decision autocrats make when they face foreign economic sanctions, suggesting the
following hypothesis.
134
Hypothesis 4.1 Main effect of U.S. sanction duration: An increase of U.S. sanction duration
will be negatively associated with a change in social spending (social security/protection,
education, and health.
I expect that there might some diversity in social welfare policy changes as reactions to
foreign economic sanctions. Policy changes vary in terms of the political system of
sanctioned governments. The allocation of social spending in autocracies may be influenced
to the extent to which they rely on institutions. Institutions in authoritarian governments are
created for the purpose of political survival. In other words, autocrats’ political institutions
play a role in preventing potential revolutionary threats from the inside and/or even outside
the winning coalition. For doing so, autocratic rulers sometimes co-opt their potential
challengers or repress them. Autocrats with institutional constraints tend to increase public
goods that benefit a much broader size of the winning coalition to deter revolutionary or
opposition movements.
When an autocrat becomes a target of U.S. sanctions, the commitment(s) to social
welfare policy depends on the autocrat’s political institutions. As sanctions persist,
autocrats may face both economic costs and political risks. Economic sanctions result in
economic hardships and poor economic performance in targeted countries. Worsened
economic circumstances may increase public discontent with and grievances against the
incumbent autocracy. The mass public will have a strong incentive to support opposition
groups that used to be relatively weak (Wood 2008, 494). Potential challengers or the
opposition represent imminent threats to a dictatorship because they can mobilize a protest
on behalf of regime change or revolution through the public’s economic grievances (Haggard
and Kaufman 1995). Those groups prefer revolution or rebellion against autocracies. To deter
this revolutionary threat, autocrats need to handle public grievances and prevent the
opposition from becoming stronger. Thus, some autocrats have strong incentives to favor
135
more public goods than do some other autocrats. When dictators have nominal-democratic
institutions including political parties and a legislature, they tend to adopt a more inclusive
policy on public goods when economic sanctions may threaten their political survival by way
of triggering a potential revolution or a mass movement.
When facing economic difficulties from sanctions, autocrats with institutions implement
a more extensive policy on public goods than those without institutions. Existing literature on
powersharing argument provides some theoretical backgrounds to solve this puzzle.
According to the power-sharing argument, autocracies with pseudo-democratic institutions
tend to be less sensitive to regime breakdown than non-institutionalized autocracies. For
instance, Boix and Svolik (2013, 301) argued that quasi-democratic dictatorships have
relative transparency in power because of “regular interaction between the dictator and his
allies” within the party and legislature. Even if the political process within institutionalized
autocracies is still less transparent than in democracies, it is relatively transparent
comparing to autocracies without institutions. Autocratic regimes with seemingly-
democratic institutions are advantageous in terms of regime resilience due to their
transparency and commitment problem solutions. Hence, the institutional dictator, once
established, is less likely to experience the threat of a coup (Svolik 2009). In addition, power-
sharing institutionalization between autocrats and ruling elites can make a ruling coalition
more stable, even under unfavorable conditions (Boix and Svolik 2013, 301). Co-optation
through nominaldemocratic institutions is closely associated with the long-term political
careers within political system because members of the winning coalition care about more
than just the immediate material benefits such as cash and subsidies (Svolik 2012).
When economic sanctions last longer and devastate the economy, ruling elites calculate
the cost-benefits of defecting from the incumbent autocrats. Deciding to defect relies on the
equation of the probability of success and the guarantee of privileges/benefits that at least
reach the level of the old regime. Such uncertainty may lead ruling elites to prefer the status
136
quo to defection. At the same time, public spending on education and health play a signaling
role of caring about the broader segment of the population and improving the long-term
prospects of them because those public goods are closely associated with promoting
human capital (Tenorio 2014). As sanctions are imposed for a longer time, autocrats with
institutional constraints seek to deter potential mass movement connected with the
opposition. They do so by showing their costly signals to the public-
i.e, providing at least moderate social welfare.
Hypothesis 4.2 Interaction effect of U.S. sanction duration: An increase of in the duration of
U.S. sanctions duration will be positively associated with a change in social spending (social
security/protection, education, and health) when an autocrat has pseudo-democratic
institutions.
4.4 Data and Methodology
This examines whether U.S. economic sanctions affect social welfare spending based on
a country-year panel data with 95 authoritarian countries between 1970 and 2007. To
measure social welfare expenditures, the study uses data of “expenses by function of
government” in Government Finance Statistics (GFS) yearbook (1977-2011) published by the
International Monetary Fund (IMF). Public spending on social security/protection, education
and health is calculated as percentage of total government expenditure.
137
4.4.1 Dependent Variable
The outcome variables in this essay are the amount of social welfare expenditures as the
percentage of total government spending. The social security/protection expenditure
represents public spending on social security and welfare/services including labor
interventions, pensions, and unemployment assistance. The education expenditure is used,
as it indicates a government’s expenses on education that include spending for “pre-primary
and primary education, secondary education” and so on, for instance. The health
expenditure includes spending on “services provided to individual persons and services
provided on a collective basis” referring to “hospital services and public health services”
(IMF 2001, 97-101).
This study covers 95 authoritarian countries from 1970 to 2007, thus including both the
Cold War and post-Cold War. I adopts Przeworski and Limongi (2000, 28-29)’s definition of
dictatorship. They classified a country as a dictatorship, which does not satisfy all the
following conditions: (1) the government should be elected; (2) the legislature should be
elected; (3) there should be more than two parties; and (4) there should be alternation which
provides a real chance of opposition’s taking power in the future.
4.4.2 Independent Variables
The main explanatory variable for the analysis is the duration of the U.S. sanctions. The
U.S. sanction duration is measured as the cumulative years of sanctions imposed by the U.S.
This indicates that as economic sanctions are imposed, their effect on political leaders and
civilians in each state is in the long term rather than the short term.
Secondly, in measuring the institutionalization of autocratic regimes, this study makes
use of an indicator taken from Cheibub, Gandhi and Vreeland (2010)’s the Democracy and
Dictatorship (DD) dataset. The institution variable is measured as a dichotomous indicator,
138
which is coded as 1 if a dictatorship allows for nominal-democratic institutions such as
legislatures and at least one political party, and 0 otherwise (Kim and Gandhi 2010).
4.4.3 Control Variables
Previous literature provides alternative explanations for social welfare spending. To
include additional determinants to social spending is likely to reduce the risk of a spurious
relationship between the outcome and explanatory variables. So several alternative
variables are included for control: GDP per capita, trade openness, aid/GDP, oil income,
government consumption, population, and civil war.
First, I include some economic variables as controls. The income variable (GDP per
capita), indicating economic development, is measured as the natural log of GDP per capita.
The GDP per capita comes from the Penn World Tables (PWT). Trade openness is logged and
measured as a percentage of exports plus imports divided by GDP at 2005 constant prices.
It also comes from the PWT. Trade openness is logged and measured as a percentage of
exports plus imports divided by GDP at 2005 constant prices. It also comes from the PWT. It
is widely used as a driving force to increase government revenues that “collected at the
border are among the least difficult to obtain. Because they have to pass through a few
checkpoints to leave or enter a territory legally, imports and exports form a base upon which
governments may impose a tax with relative ease. For this reason, foreign trade taxes have
tended to represent the main source of government income” (Cheibub 1998; Escribà-Folch
and Wright 2010).
To measure free resources for dictators, both foreign aid and oil-revenues are used. Aid
indicator refers to foreign aid as a percentage of GDP (Aid/GDP). The foreign assistance data
come from World Bank’s World Development Indicator (WDI). Another free resource is
generated from oilproduction. This variable (Oil income) is measured as the total oil income
139
per capita provided by (Haber and Menaldo 2011). Government spending is measured with
the percent share of GDP and the data come from PWT. The increase in government
expenditure might flow into social spending. Population, which also comes from PWT, is
logged.
Lastly, civil war, a dichotomous variable, is coded as 1 if a country has experienced an
intrastate conflict with at least 1,000 battle deaths in a given country-year and coded as 0
otherwise. The data come from Version 4 - 2012 of the UCDP/PRIO Armed Conflict Dataset
(Gleditsch 2002).
4.5 Estimation & Results
4.5.1 Methodology
Since the data are characterized as the pooled time-series, I takes into account the
methodological issues of both time-series and cross-sectional features. In other words, the
study needs to control for the possibility of autocorrelation and heteroskedsticity. First, to
control for autocorrelation, it is recommended to include a lagged dependent variable (LDV)
an independent variable (Beck and Katz 2011). Some political scientists raise questions
about the effectiveness of using LDVs in the model because the inclusion of LDVs might
increase the risk of making inaccurate inference leading other independent variables to lose
their explanatory power (Achen 2000). However, Keele and Kelly (2006, 17-18) suggested
some criteria for applied researchers to use LDVs:
(1) under the dynamic situation, without LDVs, linear regression can lead to a biased
estimation, (2) an Autoregressive-Moving-Average (ARMA) model is recommendable even
under the ‘weakly dynamic’ the data generating process, and (3) under the dynamic process,
regression with LDVs produces a better model than the alternatives. For deciding whether to
use LDVs and/or to take into account the dynamic situation in this research question, the
140
study employs an appropriate method to estimate the impact of U.S. sanctions on the
‘change’ of social welfare spending in autocratic countries. It is expected that U.S. economic
sanctions have dynamic effects on social spending. For this, I use error-correction model
(ECM) to account for dynamic processes, which is modeled in the following way (De Boef and
Keele 2008).
∆ Yt =α0 +α1Yt−1 +β1∆ Xt +β2Xt−1 +εi
This model becomes popular among political scientists who primarily focus on the
dynamic change of their dependent variables that are affected by independent variables.
ECM is recommendable for both non-stationary and stationary data to primarily analyze
the dynamic change of the outcome variables (De Boef and Keele 2008; Keele and Webb
2016).
Another issue to be considered in data analysis here is missing values in the data. There
is some degree of missing values in this data. In particular, missing values in social spending
data of authoritarian countries may cause serious problems in the analysis (Honaker and
King 2010). The study estimated the regressions using the imputed dataset. To increase the
validity of the statistical inference from the multiple imputation, the study generated five
imputed dataset (as the default of Amelia II). The imputation procedure was set up to deal
with time-series crosssectional datasets (Honaker and Blackwell 2011). The imputation
includes all variables used in the main analysis along with some relevant economic
indicators. It is recommended that analysts should add more information to make
imputation more plausible (Honaker and Blackwell 2011). After doing the imputation, post-
imputation diagnostics were carried out using plots to compare observed with imputed
values (See Appendix). Then, I estimated five regression analyses using five imputed
datasets to obtain the mean value of the coefficients and standard errors with significance
141
levels. The Stata automatically calculated them based on Rubin’s rule (Honaker and
Blackwell 2011). Table 4.1 shows the results of the error-correction model using the imputed
datasets. Many explanatory variables for the immediate effects are not significant except for
a few variables with statistical significance. But, with regard to the long-term effects, the
interaction effects of U.S. sanction duration and autocratic institutions are statistically
significant, consistent with the main findings except for the health spending model.
Alternatively, the study ran each separate regression using all the imputed data. The finding
of each model is more consistent with the study’s main results and supports the hypothesis.
A third methodological concern is the causality of the empirical findings. Unlike
experimental methods, quantitative analysis using observational data is assumed to be
weak in the sense of controlling extraneous variables (Lijphart 1971). The important concern
in the analysis is whether sanctioned states will tend to reduce spending on social welfare
policies compared with ‘similar’ countries without U.S. sanctions. To improve the effect of
cause, empirical studies need to consider the counterfactual or ‘potential outcomes’ of the
model (Morgan and Harding 2006). In other words, empirical studies should take into
account both observed and unobserved outcomes. For a quasi-experimental setting, I use
propensity score matching to compare authoritarian countries that were targets of U.S.
economic sanctions to very similar autocracies that experience no U.S. sanction
impositions. To make appropriate cases, I matched autocratic countries on GDP per capita,
governmental expenditure, trade openness, oil income, population, foreign aid over GDP, and
civil war. This allows for a comparison of closely matched pairs without additional controls
for variables represented as alternative explanations of the level of social spending in
autocratic regimes. In the analysis, the study treated an autocratic country with U.S. sanction
imposition as a treated group and the most similar country except the sanction treatment is
included in the control group to estimate the causal effect of U.S. economic sanctions. For
conducting matching, the study employs Matchit program using R (Ho et al. 2011).
142
Lastly, I used a matching algorithm to construct a sample for both treatment groups (i.e.,
sanction-targeted countries) and control groups (i.e., non-sanctioned countries) that are
similar in several characteristics (see Appendix for balance statistics). For the matching
procedure, the study applied a genetic algorithm (as mentioned above). First, I estimate the
analysis combining all five matched datasets and then produce the point estimates using
Rubin’s rule. At the sample, I attempt to calculate the average values of all five imputed data
and then collect the matched data. After this, I find which country is included in the matched
dataset. Based on the selected sample, I get the matched data using only selected countries
from the previous procedure.
4.5.2 Results
In this section, I present the main findings of estimating the impacts of U.S. sanction
duration on social security/protection, education, and health spending, by looking into both
the short-term and the long-run effects. Using the ECM method, I can identify changes in the
outcome variable by looking into both the immediate effect of and a long-run effect of
regressors on the dependent
variables.
∆ Social Spendingt =a0 +a1Social Spendingt−1 +b1US Sanction Durationi,t−1
+b2∆US Sanction Duration + b3Institutionsi,t−1 +b4US Sanction Duration x Institutionsi,t−1
+b5∆US sanction duration x Institutionsi,t−1 +b6Controlsi,t−1 +b7∆Controls
The analysis begins in Table 4.1 with six models of social security, education, and health
spending, using five imputed datasets. The estimation results are based on the Driscoll and
Kraay robust standard errors estimation that controls for heteroskedasticity, autocorrelation,
and cross-sectional dependence (Driscoll and Kraay 1998; Hoechle 2007). The first model
143
specification in each outcome variable (Models 1, 3 and 5) in Table 4.1 simply conducts a
baseline estimation (the main effects of U.S. sanction duration) with no interaction variables.
In regard to the short-term effects, the study finds that the analysis does not support for my
hypothesis that the impacts of U.S. sanction duration on change in social welfare spending
(social security, education, and health). For instance, the result in Model 1 shows that the
changes in U.S. sanction duration have a positive but insignificant impact on changes in
social security spending. Regarding education and health spending, the changes in U.S.
sanction duration also have negative, though not statistically significant, impacts.
In the interaction models (Models 2, 4 and 6), the effects of U.S. sanction duration on
social spending depend on autocracies’ political institutions. For example, the outcome in
Model 2 shows that the changes in U.S. sanction duration are negatively associated with the
changes in social security expenditure when autocracies have institutional constraints. But,
the coefficient for the variable is not statistically significant. Regarding spending on
education and health, interaction effects of changes in U.S. sanction duration and
institutions are also not significant even though the causal direction is positive.
With regard to the long-term effects, the level of U.S. sanction duration have a positive
and significant influence on the change in social security expenditure (Model 1). The result
indicates that autocratic leaders tend to increase government spending on social
security/protection in order to compensate for the loss of income for their coalition members
in part due to economic sanctions. This result is consistent with some literature on social
security policy that tends to be restricted to the small segment of population in the
developing world. When autocrats experience U.S. sanctions, they co-opt regime supporting
groups including the military, civil servants, and privileged
