Foreign Military Training and Coup Propensity

Foreign Military Training and Coup Propensity

Foreign Military Training and Coup Propensity
Jonathan D. Caverley†
Northwestern University
Jesse Dillon Savage∗
University of Melbourne
November 7, 2014
Abstract
How does aid in the form of military training influence foreign
militaries’ attitudes towards civilian control? The United States has
trained tens of thousands of officers in foreign militaries with the goals
of increasing its security and instilling respect for human rights and
democracy. We argue that training increases the military’s power relative to the regime in a way that other forms of military assistance
do not. While other forms of military assistance are somewhat fungible, allowing the regime to shift resources towards coup-proofing,
human capital is a resource vested solely in the military. Training thus
alters the balance of power between the military and the regime resulting in greater coup propensity. Using data from 189 countries from
1970-2009 we show the number of military officers trained by the US
International Military Education and Training (IMET) and Countering Terrorism Fellowship (CTFP) programs increases the probability
of a military coup.
∗
†
jesse.savage@unimelb.edu.au
j-caverley@northwestern.edu
0
1
Introduction
The historical record of United States training of foreign military officers is
fraught with unpleasant alumni.1 The School of the Americas became infamous for training a number of prominent human rights abusers and strong
men among its graduates. Now renamed the Western Hemisphere Institute
for Security Cooperation, the leaders of the 2009 Honduran coup received
training there through the International Military Education and Training
(IMET) program. More recently, the March 2012 coup in Mali was led by
the IMET-trained officer Amadou Sanogo. The Egyptian military, one of
the largest recipients of US training, recently deposed that country’s democratically elected president. The United States is not the only state to train
such students; Moussau Dadis Camara, the leader of a 2008 coup in Guinea,
had spent four years in Germany as a military trainee (Die Welt 2009).
An Amnesty International report (2010) on the Dadis regime’s massacre of
protesters the following year implicated soldiers trained by France, China,
and the United States.
While for decades the United States has committed considerable resources towards foreign military training (FMT), the number of these programs has blossomed, and funding for existing ones has soared, since the
9/11 attacks. As it winds down its wars in Afghanistan and Iraq, the United
States finds itself increasingly relying on a more indirect approach–variously
1
Authors are listed in reverse alphabetical order. They would like to thank Ariel
Ahram, Steve Biddle, Charles Butcher, Kevin Petit, Charlie Glaser, Saeid Golkar, Daniel
Kebede, Rose Kelanic, Jikon Lai, Timothy Lynch, David Malet, Terry MacDonald, Alex
Montgomery, Fritz Nganje, Will Reno, Drew Stommes, Jordan Tama, and participants at
the George Washington Research in Progress seminar and at the Woodrow Wilson Center
for helpful comments. We thank Drew Stommes for able research assistance.
1
called “security assistance,” “partner capacity-building,” and ”phase zero
operations”–to advance its interests. While it is by far the largest provider
of FMT in the world, the United States is not alone in looking to the training
of militaries as a means of increased influence for the donor and increased
development and political stability for the recipient. Countries as diverse as
France, China, India, Australia, and Morocco have used FMT to advance
their interests. Given the increasing attention to the provision of security
and armed services reform as essential components of development and democratization (Collier, 2008), FMT merits consideration in the larger debate
over the political consequences of foreign aid more generally.
While research on foreign aid’s effects on recipient states’ internal politics
is well-established, little empirical work and even less theory exists on foreign
military training. In this paper we argue that the effects of FMT differ in
important ways from other forms of military (and civilian) assistance. While
military aid in the form of hardware or financing can allow leaders to transfer
their own resources towards coup-proofing, training does not provide such
flexibility. Much of the research to date on military aid shows that it has
remarkably little effect. However, in this paper we theorize and show that
aid in the form of training increases the probability that a state’s military
will intervene in politics. We argue that this increased propensity is due to
the aid’s increasing the military’s human capital, an increase that civilian
leaders have difficulty offsetting by devoting additional resources to coupproofing.
The paper proceeds as follows. First, we discuss the previous literature
addressing the causes of coups, with a focus on external aid. Second, we
2
describe American FMT, in particular the International Military and Education Training Program. Third, we introduce a human capital-based theory
predicting that training will increase military coup propensity. Finally, we
present the results of cross-national statistical analysis demonstrating that
US FMT is associated with higher coup attempt probability.
2
External Influences on Coups
We examine military-backed coups, an illegal replacement of a state’s governmental leadership through its military’s use or threat of violence (Huntington, 2006, 218). Previous research on the military and regime change
suggests that a combination of resource flows and the balance of power between the government and the military drive the military’s willingness to
intervene in politics. How leaders within regimes structure or balance different groups with the capacity for violence can influence the willingness of
the military to engage in coup activity. Leaders are able to “coup-proof”
themselves through the creation of paramilitaries or by weakening the formal military forces (Quinlivan, 1999; Belkin and Schofer, 2003; Pilster and
Böhmelt, 2011, 2012; Talmadge, forthcoming). Alternatively, regime leaders
may buy the support of the military through the provision of resources and
autonomy (Powell, 2012; Svolik, 2012a).
A smaller body of literature has focused on the international influences
of coups. These explanations fall into two broad categories: those that focus
on the leverage over regimes possessed by external actors, and those that
focus on how external resources might influence the military’s capacity to
3
rebel and the regime’s ability to maintain power.
External actors can signal approval or disapproval of coups by limiting
or extending the resources provided to the regime. Goemans and Marinov
(2013) have shown that the willingness of powerful states to tie foreign aid
to democracy in the post-Cold War era has not only reduced the number
of coups, but has increased the probability that coup leaders will reinstate
elections soon afterwards. On the other hand, threats from external actors
can also encourage coups in the right circumstances. Thyne (2010) suggests
that a signal of American hostility toward a government can encourage the
military to intervene to preserve its relationship with the United States.
The second type of explanation looks at direct engagement with foreign
militaries through assistance or arm transfers (Rowe, 1974; Maniruzzaman,
1992; Ruby and Gibler, 2010). These studies have found a positive association between military assistance or arms transfers and coup probability,
although the effects are quite small. Others have shown a relationship between arms transfers and reduced democracy more broadly (De Soysa and
Midford, 2012). Undifferentiated foreign aid can free up resources to allow
regimes to stay in charge by expanding government, consolidating powerful
groups, or repressing the population (Remmer, 2004; Wright, 2008; Licht,
2010). While the research is not unanimous, one recent review concludes
that “a large and sustained volume of aid can have negative effects on the development of good public institutions in low income countries” (Moss et al.,
2006, 18). In short, most military assistance, like other forms of aid, exhibits
a high level of fungibility in the hands of incumbent regimes (Feyzioglu et al.,
1998).
4
2.1
Foreign military training and civil-military relations
While aid from a wide variety of international donors has been associated
with coup propensity, work on military training tends to focus on the United
States. This is not only due to the massive extent of its efforts, but because
the United States is relatively transparent about who it trains.
According to the State Department, programs like IMET have two explicit goals: “increased understanding and defense cooperation between the
United States and foreign countries” and to “increase the ability of foreign
national military and civilian personnel to absorb and maintain basic democratic values and protect internationally recognized human rights,” a goal
which explicitly includes a “greater respect for and understanding of the
principle of civilian control of the military.”2 These two goals can clearly
conflict with each other, and indeed the twin norms of respect for civilians
and for human rights can be in tension for many regimes.
The few studies focusing on US training posit a transmission of democratic norms, with a particular emphasis on civilian control, to foreign militaries. For example Ruby and Gibler (2010) find a negative relationship
between American FMT of relatively high-ranking officers and coup risk
They argue that these officers learn norms of civilian control in class and
absorb them more generally through residing in the United States for an
extended period of time. Atkinson (2006; 2010) shows that American FMT,
broadly construed, is associated with regime transitions toward democracy,
by increasing the probability that a military will refuse to suppress a liber2
http://www.dsca.mil/home/international military education training.htm
5
alizing popular movement. At least some of IMET’s sponsors agree; one US
State Department official testified to Congress in 1999 that “The stability
we saw in military forces around the world during [the] recent radical decrease in defense budgets would have resulted in coups which today never
materialized, in part because of the learned respect for civilian control of
the military” (Pomper, 2000).
We are skeptical that norms transmission exhausts the total effect of
FMT on domestic politics. Examples abound of foreign soldiers appreciating their time in the United States without internalizing such norms. A
young Pakistani major named Mohammad Zia-ul-Haq spent a year at Fort
Leavenworth, where he was “adopted” by a mail carrier named Ed and his
wife, Dollie. Years later, Ed and Dollie were invited to the Reagan administration’s 1982 state dinner for Zia, who installed himself in a coup and
held onto power for over a decade through brutal repression (Powell, 1996,
123). General Abdel-Fattah el-Sissi, who recently ousted the first democratically elected leader in Egypt’s history, is also the first American-, rather
than Soviet- or Russian-, trained Egyptian military chief. Assigned to the
US Army War College, el-Sissi attended Super Bowl and Halloween parties while writing a skeptical (if cursory) thesis on democracy, religion, and
internal conflict in the Middle East (Stewart, 2013).
More theoretically, the effect suggested by previous research requires
a very strong model of normative transformation whereby a few months
of exposure to US instruction is enough to overcome existing and subsequent norms in officers’ home countries and institutions. Substantively, the
same behavior—a military’s decision to side with “the people” over the
6
government—can often be explained equally well from a normative or an
organizational interest standpoint. Indeed, before it refused to repress the
public itself, the Tunisian military failed to step in when internal security
forces put down protests in the south and west of the country prior to the
revolution (Brooks, 2013).3
There is one more reason to be skeptical of the norm transmission component of American FMT: the United States does not appear to assign a
high priority to it.
3
Description of US Foreign Military Training Programs
According to the US State Department, the many different forms of FMT
share the same primary goal: to “further the goal of regional stability
through effective, mutually beneficial military-to-military relations that culminate in increased understanding and defense cooperation between the US
and foreign countries.”4 Most government assessments make it clear that
for these programs, even the flagship International Military Education and
Training Program (IMET), human rights and norm transmission are secondary to this main task.
IMET is far from the only American FMT program, and at $99 million
in 2013, is far from the best funded. These other programs have relatively
3
Tunisia’s under-funded military depended heavily on American funds and training
(Brooks, 2013, 214).
4
State Department web site, http://www.state.gov/t/pm/65533.htm
7
little to do with the promotion of democratic norms.5 The Foreign Military
Financing program (over $5 billion in 2013) contains a training component
to complement the weapons it provides. Indonesia, excluded from IMET
from 1992 to 2005 due to its military’s rampant human rights abuses, was
the largest single beneficiary from the Pentagon’s Regional Defense Counterterrorism Fellowship Program (CTFP), designed to train foreign forces in
counterterrorism techniques through courses such as “Intelligence in Combating Terrorism” and “Student Military Police Prep.” Indonesia has also
benefitted handsomely from the Defense Department’s Global Train and
Equip, or “Section 1206,” program whose allocations from 2006–2012 totaled nearly $1.8 billion (Serafino, 2013), or 2.5 times the entire IMET
budget for the same period.
IMET does remain a large program, and the Defense Security Cooperation Agency (DSCA) refers to it as the “cornerstone of security assistance
training.” Jointly run by the Defense and State Departments since 1976,
IMET trains from 4,000–8,000 foreign military personnel annually.6 In 2013,
its funds supported FMT for personnel from over 130 states, giving a sense
of how widespread this effort has become.
The IMET program is distinctive not only for its longevity and visibility, but also for its specifically targeted trainees. Much of IMET is devoted
to Professional Military Education, which DSCA describes as “progressive
levels of military education that prepares military officers for leadership.”
5
The best overview of these programs is jointly produced by Departments of Defense
and State 2014.
6
Similar programs, authorized under the Foreign Assistance Act of 1961, existed prior
to this year.
8
Members of foreign officer corps enroll in courses, often alongside their American counterparts, ranging from Infantry Officer Basic training, to Command
and General Staff College, and up through the War Colleges (the US military’s capstone programs for “strategic leadership” education). Thus, compared to the other forms of International Security Assistance efforts focused
on lower ranks, IMET alumni are the trainees most likely to lead and influence not only the security policy of a foreign state, but also the decision to
initiate a coup.7
Not surprisingly, while the broad goal of IMET has remained constant
since its inception, many of the program’s details have shifted along with
US government priorities. Since the Cold War’s end, Europe (particularly
former Warsaw Pact states) continues to receive the largest amount of annual
funding as a region. However, in the decade following 2001, funding levels
for the Near East and for South and Central Asia have more than doubled
(GAO, 2011). Training has also grown more expensive. IMET funding in
constant dollars per student rose from nearly $6,100 in fiscal year 2000 to
approximately $15,000 in fiscal year 2010 (GAO, 2011).
While human rights has always been part of the IMET mandate, in 1990
Congress shifted the objectives of the IMET program to “focus on fostering greater understanding of and respect for civilian control of the military,
contributing to responsible defense resource management, and improving
7
While the US government is legally required to track the career progress of its foreign
trainees, if for no other reason than to measure program effectiveness, little has been
done on this front. The DSCA singles out trainees that have “attained a position of
prominence,” but as of June 2011, these trainees account for 1 percent of the nearly
88,000 IMET trainees in the database (978 IMET graduates). Such a tracking system is
of limited value; despite his leadership of a successful coup in Mali, Amadou Sanogo was
not “prominent,” having only the rank of captain.
9
military justice systems and procedures in accordance with internationally
recognized human rights.” A certain percentage of a country’s IMET program must now be selected from approved “Expanded IMET” (E-IMET)
courses which cover these subjects. This percentage varies by country; on
average it is around 20%, but El Salvador had to exclusively use E-IMET
courses in fiscal years 1993–1995 (GAO, 2011).
On the other hand, while respect for human rights standards is mandated
to be a large part of any IMET program, this rarely appears to be the
primary goal. A 2011 US Government Accountability Office (GAO) report
reviewed 29 training plans for countries with poor political and civil freedom
records, and found that only 11 of these plans identified human rights as one
of the key objectives. 7 of the 12 managers interviewed by the GAO said
that human rights was not a priority compared to other IMET objectives
(GAO, 2011). Unlike previous work, we therefore construct a theory linking
FMT to coup propensity that reflects this balance of effort by the United
States.
4
The Uniqueness of Training as Military Aid
To do so, we make a simple theoretical argument based on US FMT being
a form of foreign aid: a delivery of resources to a state by an outside actor.
This aid comes in a very specific form: an increase in the military’s human
capital.
We use “human capital” as an umbrella term to describe a range of
assets–social, instructional, and perhaps even economic–that can be deliv-
10
ered through FMT. Capital is anything that enhances an actor’s ability to
produce something, and is therefore something valuable in and of itself. The
benefits consist in part of professional knowledge, ranging from small unit
tactics to grand strategy, enabling recipients to conduct military operations
more effectively. FMT can also impart a more ineffable form of “social
capital” (Bourdieu, 1986; Coleman, 1988), establishing a powerful network
of prestige, trust, and reciprocal relations that allow privileged actors to
achieve higher status positions.
The myriad forms of human capital provided by FMT share one thing in
common distinguishing them from other forms of foreign (including military)
aid. Unlike other forms of assistance, this human capital is relatively nonfungible. It does not free up indigenous resources that regime leadership
can shift towards counterbalancing through coup-proofing. These benefits
make undertaking a coup more feasible, increase the motivation to do so,
and provide no offsetting benefits to regimes to prevent it. Contrary to more
norms-based arguments, we expect that FMT will result in an increase in
all types of military-assisted coups, not just democratic transitions.
4.1
Military Training as Human Capital
While not using the term “human capital,” Singh (2014, 9) emphasizes the
importance of “soft power” in the prosecution of coups, finding that coups
succeed not because of “differences in hard military power among the parties
but the resources available for setting and coordinating expectations and
making facts.” In this section, we flesh out the ways that FMT can shift
the balance of human capital for small groups of soldiers relative to the rest
11
of the military, as well as for the larger military relative to the government,
leading them to consider a coup.
This addition of human capital, intrinsic to the trainee and military,
has three effects. First, FMT increases the relative competence and social
influence of trainees within the military and consequently the larger military
within the government. Second, it creates a corporate identity among the
recipient military that separates it from the civilian government. Finally, by
tying these trainees into an international network of like-minded American
officers and making them valuable for the execution of US interests, IMET
and similar programs create an alternative source of support for rebellious
military members.
The United States has an enormous capacity to generate and deliver
military human capital. Its operational experience and economies of scale
provide the United States with the same advantages in training it enjoys in
building and selling weapons (Caverley, 2007). The US military’s training
infrastructure is vast, experienced, and combat-focused. It spent $1.2 billion in 2013 on “professional development,” a sum greater than the entire
military budgets of 73 countries (SIPRI). The dizzying logistics and organizational skills required to quickly deploy a massive amount of combat power
anywhere in the world, which the United States has repeatedly done in the
post-Cold War era, is simply beyond any other military. The United States
military has acquired a great deal of experience and interest in counterinsurgency and stabilization operations over the past decade. Finally, the United
States has been training foreign militaries in earnest for well over half a
century.
12
Militaries with a higher number of personnel trained by the United States
(even compared to other, well-established military powers) are therefore
likely to be more competent.8 The IMET training received by Mali’s Sanogo
was basic, tactical, but thorough: English in Texas, intelligence in Arizona,
work with the Marines in Virginia, and finally the Army’s infantry officer
basic training course in Georgia in 2010 (Cavendish, 2012; Nossiter, 2012).
Clearly, instruction in small unit leadership can make someone a more effective leader of coups. More senior officers at the War Colleges receive doses of
IR theory, lessons in strategic planning, and general insight into managing
a large defense bureaucracy. Recipients of such training will not only possess higher levels of professional ability, but will have gained an increased
degree of prestige. They, in turn, are expected to train others upon returning home, increasing the competence of the military more broadly. These
characteristics make the recipients of FMT more capable of identifying and
taking advantage of opportunities to stage a coup.
FMT increases the value of these soldiers to the regime. In doing so,
it makes them harder to punish. Regimes do not want to alienate a valuable and, in the short to medium term, irreplaceable resource. After five
weeks’ incarceration and an apology ceremony following their failed 1981
coup attempt in Thailand, the 52 members of the “Young Turks” (led by
8
France is probably the country with the most active foreign military training effort
after the United States. Yet the entire budget for its Direction de la coopération de
sécurité et de défense (DCSD), about 100 million euros (60% of which is personnel costs),
is roughly the size of IMET’s annual funding level (Gillier, 2014). In contrast the rough
American equivalent to the DCSD, the State Department’s Office of Security Assistance,
directs $6 billion in military grants (Departments of Defense and State, 2014). The United
Kingdom spent roughly $3.6 billion in 2012 on all “Foreign Military Aid” (HM Treasury,
2014). That same year, the United States spent $485.9 billion (USAID).
13
American-trained officers) gained clemency and rejoined the Army (Samutwanit, 1982; Taw, 1994, 64).
Secondly, FMT also creates a sense of professional identity and social
ties (Brooks, 2013). The increased prestige associated with training may
facilitate organization of coups by providing soldiers with greater influence
over their peers. Coups are generally carried about by small groups (Singh,
2014); and FMT can help create a small network of reliable plotters that can
work together over the many months required to plan a coup without fear of
a defector (Nordlinger, 1977, 99). A mere four junior officers instigated the
1994 coup in Gambia, three of whom had all attended officer training in the
United States from 1990-1991 (Hughes and Perfect, 2008).9 The 2008 seizure
in Guinea was known as the “the German coup” because the small number
of low-level officers had been trained by the Bundeswehr, and communicated
in this common language (Heidelberger, 2010).
Beyond allowing for increased trust and communication between plotters, this professional identity does not derive from the officers’ own government. The US military has a unique professional culture, largely adopting Samuel Huntington’s notion of “objective civilian control” (Huntington,
1957). This ideal precludes interference by the military in politics, while generating a strong, separate corporate identity (a “bit of Sparta in the midst of
Babylon” in Huntington’s famous phrase). Ironically, this can backfire. In
states with internal threats or corrupt governments, a “professional” military
may be more likely to step into politics, as Huntington himself recognized
9
There is little record of the fourth plotter’s training, as he was purged and subsequently died within months of the coup. The plotters’ CVs can be found at the Republic
of Gambia 2014’s governmental website. http://www.statehouse.gm/cabinet.html
14
(Huntington, 2006; Stepan, 1986).
The conveyance of norms may actually exacerbate this disconnect between the students and the larger military and government at home. One internal study of the National Defense University cited a significant decrease in
foreign students’ perception of human rights practices in their home country
(Jungdahl and Lambert, 2012). More practically, the increased professionalism of American-trained officers may reassure actors waiting on the sidelines
that bloodshed will be limited, an important consideration of initially neutral leaders and units (Singh, 2014). Thus American norms transmission
may correlate with more support for the trainees involved in a coup.
Finally, FMT delivers a level of connectedness and comfort with the
United States government, particularly its military, that officers considering
a coup may find useful. For example, the strong ties of the Egyptian military
to the United States were apparently an important factor in American support for the original revolution that overthrew Hosni Mubarrak (Nepstad,
2013, 343). Now well-versed in American doctrine and operations, these
leaders facilitate interoperability for joint missions with US forces. The increased value of the soldiers to the United States in terms of capability and
willingness to cooperate, which is after all FMT’s primary goal, may limit
its willingness to punish militaries that intervene in politics.
In practice, it can be difficult to distinguish between all these different effects of training. And indeed, US training of foreign militaries likely
provides a basket of these goods. While future research should attempt to
disentangle these causal threads, for this paper’s purposes, we simply note
that all of these benefits suggest the same effect: increased willingness to
15
consider and ability to prosecute a coup. This contrasts strongly with the
prediction of theories based on norms transmission.
4.2
Fungibility and Other Forms of Aid
We expect to see a stronger relationship between FMT and coups compared
to other forms of aid. Most foreign aid (military or otherwise) is fungible;
corrupt leaders can divert more of the state’s resources to other needs, such
as coup-proofing through buying off elites, creating additional paramilitary
groups, or providing additional public goods. On the other hand, the capital provided by training largely resides in the trainee (and, to a degree, the
military). Moreover, the amount of resources freed up for coup-proofing by
outsourcing some instruction to the United States is likely to be minuscule
in terms of its financial value. Because of the American training advantages
mentioned above, the human capital provided is large relative to the dollar
amount assigned by the United States to it. The Foreign Military Financing budget request, which is designed to provide weapons to aid recipients,
is sixty-three times larger than the IMET budget. Before its coup, FMT
amounted to less than half a percent of the total US aid to Mali (Boswell,
2012).
Moreover, a theory based on adding military human capital makes no
predictions about coups that do not involve the military. On the other hand,
norms-based arguments would suggest otherwise. Non-military advocates
for regime change may recognize that an American-influenced military is
unlikely to side with the autocrat if wide-scale violence is required to maintain power. Thus, the gamble of instigating the overthrow of a regime by
16
civilians may become less costly. Unlike a capital-based argument, a normbased argument about US training would predict increases in non-military
coup attempts.
4.3
Hypotheses
Given our theory of FMT as aid in the form of human capital, we test the
following hypotheses:
H1 More training for a given country will increase the probability of a
military-backed coup attempt.
Clearly, finding a positive relationship between training and coups undermines the alternative theory that norm diffusion reduces coup likelihood.
To begin to flesh out our mechanism of non-fungibility, we test a second
hypothesis comparing FMT to other sources of aid:
H2 The effect of training on the probability of a military-backed coup attempt
differs systematically from other forms of aid.
Finally, to differentiate further our theory from norms-based arguments
we also test:
H3 More training for a given country will increase the probability of coup
attempts without the military’s backing.
Finding support for this third hypothesis would strengthen the norms-based
logic at the expense of our human capital theory. We expect a null finding.
17
4.4
Potential Endogeneity
Coups may be endogenous to FMT. While data show that the United States
is not terribly discriminating about which militaries it trains (about 3 of
every 4 militaries participate to some degree in IMET), it spends a lot of resources on less professional militaries (i.e. ones with a high coup propensity).
If this reverse causation was the case level of US training (our independent
variable) would be co-determined by our dependent variable.
It is, however, not clear that incompetent militaries are more prone to
coups. Indeed, our theory among others suggests the opposite.10 Moreover,
the United States military has historically used ties with FMT alumni to
pressure military non-intervention in politics, especially in the post-Cold
War era and especially when it comes to democracies. We therefore analyzed
the Cold War and post-Cold War eras, and democratic and non-democratic
regimes, separately in some models. The United States is legally required to
cut off IMET when militaries overthrow democratically elected governments;
and often suspends programs, as for Guinea in 2008, even where coups
replace autocrats. Finally, if our mechanism is correct, an autocrat who
believes that the military poses a severe coup threat is unlikely to allow
increased American FMT. We nonetheless include years since the last coup
in our statistical analysis. Importantly, years since last coup is negatively
correlated with Military Backed Coups and positively correlated with FMT
meaning that countries that have not recently experienced a coup event
receive more training (see supplemental section).
10
See also (Egorov and Sonin, 2011).
18
In addition to the state level, there may exist endogeneity at the individual level. Those who receive American training are likely to be relatively
high performers to begin with. Students are selected based on “leadership
potential and likelihood of being assigned, subsequent to IMET participation, to a job relevant to their training for a period of time to warrant the
training expense” (Defense Institute of Security Assistance Management,
2013).11 The cross-national nature of IMET and our analysis mitigates this
problem. IMET does not train the best soldiers in the world, but trains a few
of the best soldiers from each country. That is, a military with more talented
(and conceivably more coup-prone) officers does not get more IMET spots.
The number of positions doled out to each state based on what American
officials believe to be in the United States’ national interests, developed by
the State Department at the regional level in coordination with the Defense
Department. The actual soldiers attending the training are nominated by
the recipient country’s government, which is unlikely to boost the soldiers
it expects to pose the greatest coup threat.
A far more likely possibility (indeed, a near certainty) for endogeneity is that a third, confounding variable simultaneously affects levels of US
military training and coup propensity. For example, as a foreign aid program, IMET is intended not for wealthy countries but rather for developing
states (which are more vulnerable to coups). A more interesting confounding variable is due to the clear US policy of encouraging new democracies
with increased aid. However, democratizing regimes are also quite prone to
11
On the other hand, the Gibler and Ruby data is much more likely to focus on handpicked regime favorites given the seniority of the officers involved.
19
military coups (Svolik, 2012b). Our theory suggests that increased FMT
for new democracies might, ironically, exacerbate coup propensity. While
neglecting to identify and control for a confounding factor, and therefore
introducing omitted variable bias, is always a possibility, we do argue that,
at least for IMET and US foreign policy, the factors that determine the level
of assistance are relatively finite, knowable, and transparent. This makes
possible steps to avoid spurious correlations, and we describe these variables
in detail below.
While we cannot eliminate omitted variable bias, we do argue that endogeneity in the data is more likely to result in a negative correlation between
FMT and coups. Our analysis is therefore a conservative one. As a broader
point, in the empirical section that follows we consistently take a very cautious approach. We aggressively control for possible confounding factors
and employ matching techniques as an additional test of robustness. This
circumspection begins with our choice of explanatory variable.
5
Description of the Data
While many different American training programs exist, we focus primarily
on the IMET program for several reasons. First it is the most transparent,
and receives the largest amount of scrutiny. Human rights and civil-military
relations are explicitly part of the curriculum unlike other forms of American
training. IMET trainees are therefore the population in which we are most
likely to see the effect of norms, i.e. the easy case. If we discover more coups
in countries with a large number of IMET trainees, then programs that are
20
less scrutinized and with priorities that de-emphasize civil-military relations
will likely display an even stronger relationship.
Second, targeted students for IMET tend to be the elite of any given
state’s military. While other programs reach down into the rank-and-file,
the foreign military personnel going through war colleges and similar IMETsupported institutions are the officers that would have the wherewithal to
launch a coup. On the other hand, IMET data does not distinguish rank
among its numbers and thus measures a broader range of trainees compared
to Ruby and Gibler (2010), who focus only on those in the two most senior
grades of War College training. Given that many coup leaders come from
relatively low in the officer corps, this represents both a quantitative and
qualitative improvement on existing data.
Third, IMET is aid; the United States paid for these students’ training.
The large majority of the officers in the Ruby and Gibler (2010) data set
is funded by their home government through the Foreign Military Sales
program. Given our theory’s focus on foreign aid in human capital form,
IMET data best captures this effect.
Our IMET data ranges from years 1970 through 2009 (DSCA, 2012).12
We used three different transformations of the IMET personnel data to test
our argument. First, we used the logged number of students trained in a
year. Second, we used a binary variable measuring if a country’s military
had received any IMET training at all. Finally, similar to Ruby and Gibler
(2010), we used the logged number of a given country’s students trained
12
Note that DSCA still provides “IMET” data prior to 1976, the year IMET was
formally instituted, to reflect the training done before this year through the Foreign Assistance Act of 1961.
21
over the previous five years (this temporal lag also mitigates the effect of
reverse causality somewhat).13
We also tested measurements of IMET spending—the natural log of total
IMET spending for a year, the natural log of a five year sum of spending,
and finally a dummy variable that measured if the state had received any
IMET funding in that year—as a way to describe the “lump sum” of aid in
the form of human capital.14
For the sake of robustness, in some models we incorporate data from the
Regional Defense Combatting Terrorism Fellowship Program (CTFP). In
2013, CTFP trained 3,098 student from 131 countries at a cost of $32 million (Department of Defense, 2014). The purposes and administration of the
CTFP differ considerably from IMET. As the name suggests, the aid is targeted at enhancing “partners’ capacity to combat terrorism.” Unlike IMET,
the program concentrates on relatively senior officers at “the strategic and
operational levels.” While the State Department determines the allotments
of IMET slots and money, CTFP is largely run by the Department of Defense. A state’s development is not directly taken into account, and thus
many countries’ ineligible for IMET–Japan, Australia, the states of Western
Europe–receive CTFP slots. Although CTFP recipients are vetted under
the “Leahy law” for human rights violations, training in human rights and
13
To deal with the large number of observations with zero soldiers trained we added a
small constant of .1. Different constants were substituted, such as 1 and .5, with similar
results. In later regressions we test untransformed versions of the explanatory variable,
thus including the zeroes. See supplemental material.
14
There were 59 country-year observations where the country received funding but no
students were trained, and 18 where students were trained but no funding was listed.
This discrepancy is partly explained by budget drawdowns. Increased per capita IMET
spending might describe the level of human capital imparted to each student, and so we
test these operationalizations as well. See supplemental material.
22
civilian control are not at the core or the program. Countries like Indonesia
received significant CTFP assistance when it was ineligible for IMET, and
China has actually received CTF slots. As a young program established in
2002 there is comparatively little data to work with. Again, we emphasize
that IMET poses the toughest test for our theory. Nonetheless, the roughly
3,000 CTFP students each year is only a third smaller than the IMET program numbers, and not analyzing the combined programs risks biasing our
findings.
5.1
Dependent Variable: Military-Backed Coups
As our primary dependent variable, we adopt coup attempts in which the
military was involved. We derive our set of attempted coups from two
comprehensive data sets released in recent years: the Global Instances of
Coups (GIC, 2011) from 1950–2010 and the Center for System Peace’s Coup
d’Etat events, 1946–2011 (CSP, 2011). To test our military-centric theory,
from these data we constructed a new measure of Military Backed Coups.
As CSP provides a list of coup leaders, we began with their list of coup
events. For transparency and accuracy of measurement we confined our
attention to either successful or attempted coups, excluding those coded as
alleged or plotted coup events. If the leader of a coup was described as a
military officer in the CSP dataset, coups were coded as a Military-Backed
Coup. However, given the complicated coalition politics of authoritarian
regimes, the military need not be the leader or instigator to have played
a vital role in a coup. For this reason we sought out evidence of military
involvement in the coups beyond a leadership role. To do this, we looked for
23
evidence of support for the coup from members of the armed forces in news
articles from two sources: the New York Times and the Times of London. If
involvement of the armed forces was recorded in either of these two sources
the event was coded as a Military Backed Coup. If we failed to find evidence
of military involvement in these two sources, we then expanded our search
to other media. From 1970 until 2009, this provided us with 286 countryyear observations in which Military-Backed Coups occurred, 82% of coup
attempts listed in CSP’s dataset.
We focus on coup attempts rather than their success for the same reason
why bargaining models of war do better at explaining war’s outbreak rather
than its outcome. Obviously, the decision to initiate a coup is likely to be
based in part on whether the plotters believe they are going to succeed. Coup
attempts therefore capture in the words of Jonathan Powell “the disposition
to intervene,” the mechanism linking training to military action against the
regime. Focusing on attempts also provides a larger number of cases to
observe. Finally, failed coup attempts can have profound effects on a state’s
politics, development, and future stability; and thus merit analysis in their
own right.
5.2
Alternate Explanations and Confounding Variables
Our second hypothesis predicts systematic differences between the effects of
FMT and other forms of assistance. We therefore include Military Aid, the
value of US Military Assistance from the USAID Greenbook, deflated to 2005
dollars, and then divided by total GDP from WDI, to determine if training
has a systematically different effect from other security assistance. Including
24
military assistance also allows us to control for the effects of external aid on
military capacity and proxy for any alliance between the US and the country
(Rowe, 1974; Maniruzzaman, 1992).15 We also include non-military aid since
we believe that training differs in its fungibility systematically from almost
all other forms of assistance. Moreover, high levels of aid dependency can
make states more vulnerable to external influence and sanction, reducing the
payoff from staging a coup. Finally, assuming that other states’ FMT effort
correlates to the amount of all aid it provides a certain country, this term
can assuage concerns of bias in our results given that we have no measures
of other countries’ military assistance. We included the variable Total Aid
which measures foreign aid from any source as a percentage of GDP (Tierney
et al., 2011).
We aggressively control for potential confounding variables that would
lead to a spurious relationship between IMET and military backed coups.
For example, the United States may in some cases seek to instigate coups
of regimes it disagrees with politically. We therefore include Erik Gartzke’s
(2006) Affinity of Nations index, which measures the shared interests of
states using the similarity of votes in the UN general assembly of the United
States and another country.
Regimes can take steps to both minimize threats from the armed forces.
We include a measure of Coup-proofing, the degree of fractionalization of the
ground compatible armed forces in the country in a given year (Pilster and
Böhmelt, 2011, 2012). The more fractionalized the armed forces, the more
15
As an additional robustness check models were run using the ATOP measure of
alliances as well, the results are similar to those excluding this variable.
25
the potential for counterbalancing exists, and consequently the more coupproofed a regime. To control for additional institutional factors, we also
included a variable measuring the amount of Spending per Soldier. Higher
levels may increase military loyalty (Powell, 2012; Besley and Robinson,
2010), and is likely to reduce the effect of IMET spending. This variable
was created by dividing total military spending by the total number of armed
personnel taken from the Correlates of Wars (Singer, 1987). The total number of Military Personnel was also included, as military size may dilute the
effect of training a few soldiers and also has an effect on coup propensity
(Powell, 2012).
Political and social development plausibly play roles (Johnson et al.,
1984; McGowan and Johnson, 1984) in both coup propensity and American interest. We include the log of GDP per capita (World Bank, 2013).
We also included controls for Economic Growth, which could lead either to
stability or crises, reducing or increasing the probability of a military intervention (Londregan and Poole, 1990; Alesina et al., 1996). Oil production
may create incentives to stage coups (Moran, 1972) as well as influence military spending, civil military relations (Ross, 2004), and of course American
interest in a country. We therefore include Oil Revenue, taken from Haber
and Menaldo (2011).
Civil war can undermine regime stability and create incentives for the
military to intervene politically. Such instability could also increase the
need for US training. For this reason, a dummy variable was included if
the country was involved in a Civil War during that year (Gleditsch et al.,
2002). We include the overall level of Democracy using Polity data (Gurr
26
et al., 2002); regime type has been shown to have important effects on the
strategies that regime leaders adopt to limit coups and on civil military
relations more generally (Pilster and Böhmelt, 2012). In some models, we
also divide the sample based on regime type. Elites from ethnic groups
that are politically empowered often have incentive to launch coups to avoid
future exclusion and solve an “internal security dilemma”(Roessler, 2011).
To address these issues we included Empowered Ethnic Groups, a count of
politically included ethnic groups taken from the Ethnic Power Relations
dataset (Wimmer et al., 2009).
We included Cold War and post-2001 dummies. The United States
might be more inclined to punish coup leaders in the post-Cold War era
as they are less concerned about alienating allies. It dramatically shifted
IMET’s emphasis after the terrorist attacks of 2001. Because previous work
suggests that the Cold War’s demise had a dramatic effect (Goemans and
Marinov, 2013), we analyze both eras separately in some models. Coups
in the same country are unlikely to be independent events (Londregan and
Poole, 1990); Years Since Last Coup should account for this as well as time
dependence (Roessler, 2011; Powell, 2012).16 Taking time seriously in the
broader sense, we followed the recommendation of Carter and Signorino
(2010) and included a cubic time trend, based on years since the last coup
event.17 In addition to controlling for time dependance, all independent
variables were lagged one time period.
16
Including a count of past coups along with the spell variable does not change the
primary results.
17
Models were also run including cubic splines in place of a cubic time trend. The
results are substantively similar. See supplemental material.
27
5.3
Methods of Analysis
We analyzed our binary dependent variable with logistic regressions, using
multiple imputation(Honaker and King, 2010) to deal with missing data,
rather than excluding from analysis observations with any missing data
(King et al., 2001), which at best results in a loss of efficiency due to discarding observations and information, and if the stringent assumption of
missing completely at random is not met, results in bias.18
For a second round of analysis, we also employed “nearest neighbor”
matching techniques (Dehejia and Wahba, 2002; Smith and Todd, 2005,
153) in case our early results are dependent on our modeling choices: decisions on confounding variables, functional form, or link function (King and
Zeng, 2006, 135). In unbalanced data (where “treated” observations differ
systematically in terms of covariates and number relative to untreated observations), the risk of estimates depending on a particular model increase.
Matching improves the balance of the dataset and the problems related to
extrapolation are less likely to occur (Ho et al., 2007). We do not suggest
that matching eliminates the problems of omitted variable bias. Our supplemental section describes our multiple imputation and matching procedures.
6
Results
We first explore the most conservative models to determine the relationship
between the IMET program (and only the IMET program) and military18
Missing at random assumes that the missingness is random conditional on the imputation model. We analyze the unimpeded data in the appendix, and also report results
for rate events legit (King and Zeng, 2001b,a).
28
backed coup attempts, presented in Table 1. The table’s first three models
look at three different transformations of IMET training, while the final
three show the effect for similar transformations of IMET spending. All 6
versions of our independent variable of IMET training are positively and
significantly associated with a higher likelihood of a military-backed coup.
We then ran an identical set of six models but pooling data on both IMET
and CTFP training. Our results on this second set of analyses are essentially
identical to those of the IMET only analysis. For this reason, we report these
full statistical results for this and subsequent analyses in our appendix.
Because raw coefficients from logistic regressions are difficult to interpret
directly, we depict the differences in predicted probabilities of a coup when
holding our explanatory variable at the 25th and the 75th percentiles (“any
training” or “no training” for dichotomous operationalizations), with all
control variables constant at their mean or mode. Simulations were used
to calculate the predicted probabilities and the confidence intervals.19 The
results show that increasing the such an increase in the number of soldiers
trained increases the predicted probability of a coup by 0.9% (see Figure
1). This represents a roughly 100% increase in the probability of a coup
for the average case. The effect size is .05% when the first differences are
calculated on a smaller change in the independent variable, from the mean
to one standard deviation above the mean, and this effect is significant at
the 5% level. The difference in probability between having no training and
any training is even larger, about 1.1%. Figure 1 also presents the results
19
2000 simulations were run on each imputed dataset. The results were combined to
obtain the estimates.
29
for the pooled IMET and CTFP data. The results are very similar.
Figure 2 shows that when using IMET spending (instead of student
numbers) to measure the amount of added human capital, the effects remain
statistically significant and substantively important. For total spending the
first difference when moving from the 25th percentile to the 75th is a .9%
larger probability of a military backed coup. The changes for both the 5
year sum is similar and the funding dummy shows an effect of approximately
1.2%. Again, the results are robust across this alternative specification of our
explanatory variable. Again, the substantive effects and significance levels
remain remarkably similar even with the CTFP data included. The results
present strong evidence that US military is associated with more militarybacked coups, regardless of whether the training is the counterterror-focused
CTFP or the more norms-oriented IMET version.
[Table 1 and Figures 1 and 2 about here]
We now turn to the results of the matching analysis, which also support
an association between IMET-only as well as IMET and CTFP training
and increased coup propensity. Both the spending and training treatments
produce a large change in the predicted probabilities of a coup relative to the
base-line risk for the average case (see Figure 3). This change in predicted
probabilities is similar to that found for the unmatched data, about 1%,
reducing our concerns regarding model dependence.
[Figure 3 about here]
In a further effort to reduce the potential for biased results, we divided
the IMET data by time period (Cold War and post-Cold War) and present
30
the first differences for both MET students and spending in Figure 4. While
the size of the effect of training during the Cold War is consistently larger
than the post-Cold War effect, although they are distinguishable in terms
of connectional statistical significance. However, even in the post-Cold War
era, IMET training is positively and significantly associated with militarybacked coups.
We performed a similar robustness check, dividing by regime type for
separate analysis.20 Figure 5 presents the first differences. The results remain positive and significant. There does appear to be a consistent difference
in size of the effect; the effect of training appears to be twice as large in autocracies (albeit not significantly different from the effect in democracies).
When limited to democracies, the coefficients remain positive, and roughly
similar in magnitude to the pooled data set, but are not significant in 2 of
the six models. This may be due to the smaller number of observations,
but could also be evidence that IMET more successfully trains students to
respect civilian control in democracies.
There may be common historical or regional differences that encourage
coups. By including regional fixed effects, these slow-moving region specific
factors can be accounted for (McGowan and Johnson, 1984). We ran models
such models and found similar results see Figure 6 (results reported in the
appendix).We also ran the models including only low and middle income
countries as defined by the World Bank. Economic development appears to
matter. The simulated effect is larger in less developed countries see Figure
7.
20
We coded democracies as those regimes that scored above 5 on the Polity IV index.
31
Having established the robustness of our explanatory variable’s link to
military backed coups, we briefly explore the effects of other covariates,
depicted in Table 1. The main body only depicts the results when we analyzed unmatched observations of IMET students, but the results are almost
identical in our other models (contained in the appendix). Our theory and
second hypotheses posited that training works differently compared to other
forms of aid. Regardless of which model we use, US military aid as a percent of GDP has a quite large and but insignificant negative effect on coup
probability. This is of course the opposite effect of military training. This
would support our theory that non-training military aid is fungible and can
be shifted towards coup-proofing. On the other hand, foreign aid has little
effect on coup-propensity, although the coefficients are consistently positive.
As in previous research, wealthier and economically growing states suffer
fewer coups, and those experiencing a recent civil war have more (significant
at conventional levels). While not obtaining significance at p < 0.1, important covariates such as coup-proofing and ethnically empowered groups have
the expected, negative relationship with military-backed coup probability.
[Figure 8 about here]
Finally, to begin testing our third hypothesis, we analyzed non-military
backed coups from the combined CSP and GIC data sets, and report the
results in Figure 8. If IMET was having a positive effect on these sorts of
coups, then our theoretical argument that these results are driven by the
increased human capital invested in the military would have been incorrect.
We do not find any relationship. The lack of any significant correlation
32
suggests why regimes might consent to foreign training despite increasing
the probability of military coups.
These additional models (as well as the ones in the supplemental material) demonstrate that our main finding is robust to different measures of
our outcome and different assumptions about the form of the relationship
between training and coup propensity. Strong reason exists to believe that
there is an association between IMET training and coup-propensity, and
that the effect works differently than other types of aid.
7
Conclusion
Given the nature of aid and the American national interests it serves, the
effect on domestic politics of American foreign military training (FMT) is
unlikely to be limited to a higher respect for human rights. Training imparts valuable resources to a potentially dangerous section of a developing
state’s polity: increasing the trainees’ human capital relative to the rest of
the military, and increasing the military’s capital relative to the government.
This training is likely to increase the military’s resource demands from the
regime, and improve its ability to remove the regime should its demands not
be met. Empirically, we find a robust relationship between American training of foreign militaries and the likelihood of a military-backed coup attempt
taking place, despite limiting our analysis to the International Military Education and Training program (IMET), which explicitly focuses on promoting
norms of civilian control. If the number of soldiers trained moves from the
25th percentile to the 75th, the predicted probability of a coup roughly dou-
33
bles. These findings undermine the idea that such training’s only political
effect is to fundamentally alter the normative beliefs of militaries, at least in
the short-term. The increase in coup activity, despite IMET’s emphasis on
civilian control, demonstrates the limited effect of these normative aspects.
That we see little difference in the effect when we incorporate CTFP data
(which is not as focused on norms transmission) also undermines faith in a
strong mechanism of norm infusion. Additionally, there is no evidence of a
link between training and non-military backed coups, which further undermines the notion that the transference of liberal norms into foreign militaries
can play a strong role in domestic politics. Finally, training’s effect on coup
propensity differs significantly from other forms of military aid in both direction and magnitude, lending more support to our theoretical argument
about the non-fungibility of military human capital.
The theoretical mechanism and correlation identified by this paper point
to further research. Non-fungible military capital may come in other forms.
Peacekeepers deployed to third party countries gain operational experience,
develop ties to other states’ militaries, and receive pay that is often orders
of magnitude greater than their normal wages back home. One Ethiopian
unit deployed to Liberia came home in 2005 only to mutiny in pursuit of
back pay withheld by the government. Analyzing the effects of peacekeeping
experience is a logical extension of this paper’s norms versus human capital
argument. The effects of US training may interact with such operational
experience. The leaders of the 1981 “Young Turks” coup in Thailand were
counterinsurgency experts with extensive pacification experience in Vietnam
and Laos. Both the coup’s leader and spokesperson had also been trained
34
in the United States which, according to one observer, resulted in their
being “more politically minded than the others” (Samutwanit, 1982, 30, 57).
Given the recent, intensive US involvement in training African militaries
for deployed peacekeeping service as well as counterinsurgency operations
against militant groups with ties to terrorism, this combined effect may
become increasingly relevant.
FMT comes in many different forms. Many examples exist of both senior and junior military officers (who received American training commensurate with their rank) overthrowing their civilian members. Distinguishing
between junior- and senior-led coups, as well as the type of instruction received, will help illuminate the causal paths leading from training to regime
change. While they pale in size and scope compared to the American effort,
many other countries train foreign militaries, including France, the United
Kingdom, and China. While data on these programs is hard to come by,
some focused comparisons might compare these states’ different approaches.
Coups are extreme examples of military involvement in domestic politics. Our human capital-based theory suggests more generally that trained
military officers, having resources now exogenously endowed solely in them,
will grow more autonomous from the regime. This can increase inclination
for coups but more broadly means that the military will be less invested
in regime survival more generally (Atkinson, 2006, 2010; Brooks, 2013).
Regime change, particularly democratic regime change, can result in the
army losing access to government funding and other resources (Collier and
Hoeffler, 2002). Providing the military with resources that are not vulnerable to redistribution may mean they are less inclined to repress to pre35
vent regime change in general. In this case, normative and resource-based
mechanisms make similar claims, and it is possible that they may mutually
reinforce each other. The addition of human capital to the military can
force some regimes to co-opt the military (rather than counterbalance), a
mechanism we do not explore here.
We do not argue that training fails to transmit norms. Nor do we suggest that the evidence presented here should lead the United States to stop
training foreign militaries. Coups are rare after all, and these militaries
may serve other interests: producing more capable US allies and creating
more stable regimes in many instances. Perhaps, under some limited circumstances, military coups can lead to democratic consolidation (Goemans
and Marinov, 2013; Thyne and Powell, 2014), although a host of ills, not
least being lower economic growth, seem to accompany coups. Instead, the
central message of this paper is that when an outside patron gives assets to
another state’s military with no countervailing forces, it is likely to produce
important political effects in the recipient country.
In short, research that takes the role of the military in developing states
seriously will contribute to our understanding both of training’s effects and
development’s prospects. Recent work has found evidence that targeted
aid, such as that which focuses on building up a number of robust groups
within civil society, can lead to increased democratization. This is not so
much because liberal norms become more widely adopted within a state,
but because countervailing groups and institutions achieve some semblance
of independence from authoritarian regimes. Our research suggests that
the military should be considered one of these societal groups, and any
36
aid process that does not consider this vital institution may be missing an
opportunity. Whether the coups that may result justify this effort is of
course another matter.
37
References
Alesina, A., Özler, S., Roubini, N., and Swagel, P. (1996). Political instability and economic growth. Journal of Economic growth, 1(2):189–211.
Atkinson, C. (2006). Constructivist implications of material power: Military engagement and the socialization of states, 1972–2000. International
Studies Quarterly, 50(3):509–537.
Atkinson, C. (2010). Does soft power matter? a comparative analysis of
student exchange programs 1980–2006. Foreign Policy Analysis, 6(1):1–
22.
Belkin, A. and Schofer, E. (2003). Toward a structural understanding of
coup risk. Journal of Conflict Resolution, 47(5):594–620.
Besley, T. and Robinson, J. A. (2010). Quis custodiet ipsos custodes? civilian control over the military. Journal of the European Economic Association, 8(2-3):655–663.
Boswell, A. (2012). Leader of mali military coup received u.s. training.
McLatchey Newspapers.
Bourdieu, P. (1986). The forms of capital. In Richardson, J. G., editor,
Handbook of Theory and Research for the Sociology of Education, pages
241–258. Greenwood Press, New York.
Brooks, R. (2013). Abandoned at the palace: Why the tunisian military
defected from the ben ali regime in january 2011. Journal of Strategic
Studies, 36(2):205–220.
38
Carter, D. B. and Signorino, C. S. (2010). Back to the future: Modeling
time dependence in binary data. Political Analysis, 18(3):271–292.
Cavendish, J. (2012). Mali’s coup leader: Interview with an improbable
strongman. Time.
Caverley, J. D. (2007). United states hegemony and the new economics of
defense. Security Studies, 16(4):598–614.
Coleman, J. S. (1988). Social capital in the creation of human capital.
American Journal of Sociology, 94.
Collier, P. (2008). The Bottom Billion: Why the Poorest Countries are
Failing and What Can Be Done About It. Oxford University Press, Oxford.
Collier, P. and Hoeffler, A. (2002). Military expenditure: threats, aid, and
arms races. World Bank Policy Research Working Paper, (2927).
De Soysa, I. and Midford, P. (2012). Enter the dragon! an empirical analysis
of chinese versus us arms transfers to autocrats and violators of human
rights, 1989–20061. International Studies Quarterly, 56(4):843–856.
Defense Institute of Security Assistance Management (2013). The Management of Security Cooperayion. Defense Security Cooperation Agency,
Wright-Patterson Air Force Base, Ohio, 32 edition.
Defense Security Cooperation Agency (2012). Dsca historical facts book.
Dehejia, R. H. and Wahba, S. (2002). Propensity score-matching methods
for nonexperimental causal studies. Review of Economics and statistics,
84(1):151–161.
39
Department of Defense (2014). Regional defense combatting terrorism fellowship program: Fiscal year 2013. Report to congress, Department of
Defense, Washington D.C.
Departments of Defense and State (2014). Foreign military training: Joint
report to congress. Technical report, Department of Defense and the
Department of State, Washington, D.C.
Die Welt (2009). Bundeswehr bildete brutalen junta-chef aus.
Egorov, G. and Sonin, K. (2011). Dictators and their viziers: Endogenizing the loyalty–competence trade-off. Journal of the European Economic
Association, 9(5):903–930.
Feyzioglu, T., Swaroop, V., and Zhu, M. (1998). A panel data analysis of the
fungibility of foreign aid. The World Bank Economic Review, 12(1):29–58.
GAO (2011).
International Military Education and Training: Agencies
Should Emphasize Human Rights Training and Improve Evaluations.
Number GAO-12-123 in Report to Congressional Committees. United
States Government Accountability Office, Washington, D.C.
Gartzke, E. (2006). The affinity of nations: Similarity of state voting positions in the unga.
Gillier, M. (2014). Compte rendu commission de la défense nationale et des
forces armées. Compte rendu 33, Assemblee Nationall, Paris.
Gleditsch, N. P., Wallensteen, P., Eriksson, Sollenberg, M., and Strand, H.
40
(2002). Armed conflict 1946-2001: A new dataset. Journal Of Peace
Research, 39(5).
Goemans, H. and Marinov, N. (2013). Coups and democracy. British Journal
of Political Science, Forthcoming.
Gurr, T. R., Marshall, M. G., and Jaggers, K. (2002). Polity iv project:
Political regime characteristics and transitions, 1800-2002.
Haber, S. and Menaldo, V. (2011). Do natural resources fuel authoritarianism? a reappraisal of the resource curse. American Political Science
Review, 105(1):1–26.
Heidelberger, M. (2010). German military cooperation with african countries
yields mixed results. Deutsche Well.
HM Treasury (2014). Public expenditure statistical analyses 2014. Technical
report.
Ho, D. E., Imai, K., King, G., and Stuart, E. A. (2007). Matching as nonparametric preprocessing for reducing model dependence in parametric
causal inference. Political analysis, 15(3):199–236.
Honaker, J. and King, G. (2010). What to do about missing values in
time-series cross-section data. American Journal of Political Science,
54(2):561–581.
Hughes, A. and Perfect, D. (2008). Historical Dictionary of The Gambia.
Scarecrow Press, Lanham, Maryland, 4 edition.
41
Huntington, S. P. (1957). The Soldier and the State: The Theory and Politics of Civil-military Relations. Harvard University Press, Cambridge,
Mass.
Huntington, S. P. (2006). Political Order in Changing Societies. Yale University Press, New Haven, Conn.
International, A. (2010). Guinea: “You did not want the military, so now
we are going to teach you a lesson”. London.
Johnson, T. H., Slater, R. O., and McGowan, P. (1984). Explaining african
military coups d’etat, 1960-1982. The American Political Science Review,
pages 622–640.
Jungdahl, A. and Lambert, P. (2012). Winning hearts by broadening minds:
Measuring the impact of international military assistance at the national
defense university. The DISAM Annual, 1.
King, G., Honaker, J., Joseph, A., and Scheve, K. (2001). Analyzing incomplete political science data: An alternative algorithm for multiple
imputation. American political science review, 95(1):49–70.
King, G. and Zeng, L. (2001a). Explaining rare events in international
relations. International Organization, 55(3):693–715.
King, G. and Zeng, L. (2001b). Logistic regression in rare events data.
Political analysis, 9(2):137–163.
King, G. and Zeng, L. (2006). The dangers of extreme counterfactuals.
Political Analysis, 14(2):131–159.
42
Licht, A. A. (2010). Coming into money: The impact of foreign aid on leader
survival. Journal of Conflict Resolution, 54(1):58–87.
Londregan, J. B. and Poole, K. T. (1990). Poverty, the coup trap, and the
seizure of executive power. World Politics, 42(2):151–183.
Maniruzzaman, T. (1992). Arms transfers, military coups, and military rule
in developing states. Journal of Conflict Resolution, 36(4):733–755.
Marshall, M. G. and Marshall, D. R. (2011). Coup d’état events, 1946-2011.
Center for Systemic Peace.
McGowan, P. and Johnson, T. H. (1984). African military coups d’etat and
underdevelopment: A quantitative historical analysis. Journal of Modern
african studies, 22(4):633–666.
Moran, T. H. (1972). Coups and costs. Foreign Policy, (8):129–137.
Moss, T. J., Pettersson, G., and van de Walle, N. (2006). An aid-institutions
paradox? a review essay on aid dependency and state building in subsaharan africa.
Nepstad, S. E. (2013). Mutiny and nonviolence in the arab spring exploring
military defections and loyalty in egypt, bahrain, and syria. Journal of
Peace Research, 50(3):337–349.
Nordlinger, E. A. (1977). Soldiers in Politics: Military Coups and Governments. Prenctice-Hall, London.
Nossiter, A. (2012). Leaders of mali’s military coup seem to have uncertain
grasp on power. New York Times.
43
Pilster, U. and Böhmelt, T. (2011). Coup-proofing and military effectiveness
in interstate wars, 1967–99. Conflict Management and Peace Science,
28(4):331–350.
Pilster, U. and Böhmelt, T. (2012). Do democracies engage less in coupproofing? on the relationship between regime type and civil–military relations. Foreign Policy Analysis, 8(4):355–372.
Pomper, M. A. (2000). Battle lines keep shifting over foreign military training. Congressional Quarterly Weekly.
Powell, C. (1996). My American Journey. Ballantine Books, New York.
Powell, J. (2012). Determinants of the attempting and outcome of coups
d’état. Journal of Conflict Resolution, 56(6):1017–1040.
Powell, J. M. and Thyne, C. L. (2011). Global instances of coups from 1950
to 2010 a new dataset. Journal of Peace Research, 48(2):249–259.
Quinlivan, J. T. (1999). Coup-proofing: Its practice and consequences in
the middle east. International Security, 24(2):131–165.
Remmer, K. L. (2004). Does foreign aid promote the expansion of government? American Journal of Political Science, 48(1):77–92.
Republic of Gambia (2014). Republic of the gambia state house online.
Roessler, P. (2011). The enemy within: Personal rule, coups, and civil war
in africa. World Politics, 63(2):300–346.
44
Ross, M. L. (2004). How do natural resources influence civil war? evidence
from thirteen cases. International organization, pages 35–67.
Rowe, E. T. (1974). Aid and coups d’etat: Aspects of the impact of american
military assistance programs in the less developed countries. International
Studies Quarterly, 18(2):239–255.
Ruby, T. Z. and Gibler, D. (2010). Us professional military education and
democratization abroad. European Journal of International Relations,
16(3):339–364.
Samutwanit, C. (1982). The Thai Young Turks. Institute of Southeast Asian
Studies, Singapore.
Serafino, N. M. (2013). Security Assistance Reform: “Section 1206” Background and Issues for Congress. Number RS22855 in CRS Report for
Congress. Congressional Research Service, Washington, D.C.
Singer, J. D. (1987). Reconstructing the correlates of war dataset on material
capabilities of states, 1816-1985. International Interactions, 14:115–32.
Singh, N. (2014). Seizing Power: The Strategic Logic of Military Coups.
Johns Hopkins University Press, Baltimore, M.D.
Smith, J. A. and Todd, P. E. (2005). Does matching overcome lalonde’s
critique of nonexperimental estimators?
Journal of econometrics,
125(1):305–353.
Stepan, A. (1986). The new professionalism of internal warfare and military
45
role expansion. In Lowenthal, A. F. and Fitch, J. S., editors, Armies and
Politics in Latin America, pages 134–150. Holmes and Meier, New York.
Stewart, P. (2013). Insight: In small american town, a window into egyptian
general’s past. Reuters.
Svolik, M. (2012a). The Politics of Authoritarian Rule. Cambridge University Press, London.
Svolik, M. W. (2012b). Which democracies will last? coups, incumbent
takeovers, and the dynamic of democratic consolidation. British Journal
of Political Science, pages 1–24.
Talmadge, C. (forthcoming,). Authoritarian Armies on the Battlefield: Explaining Fighting Power in Non-Democratic Regimes. Cornell University
Press, Ithaca, N.Y.
Taw, J. M. (1994). Thailand and the philippines: Case studies in us imet
training and its role in internal defense and development. Technical report,
DTIC Document.
Thyne, C. and Powell, J. (2014). Coup d’etat or coup d’autocracy? how
coups impact democratization, 1950-2008.” foreign policy analysis. thyne,
clayton, and jonathan powell. 2014. ”coup d’etat or coup d’autocracy?
how coups impact democratization, 1950-2008. Foreign Policy Analysis.
Thyne, C. L. (2010). Supporter of stability or agent of agitation? the effect
of us foreign policy on coups in latin america, 1960—99. Journal of Peace
Research, 47(4):449–461.
46
Tierney, M. J., Nielson, D. L., Hawkins, D. G., Roberts, J. T., Findley,
M. G., Powers, R. M., Parks, B., Wilson, S. E., and Hicks, R. L. (2011).
More dollars than sense: refining our knowledge of development finance
using aiddata. World Development, 39(11):1891–1906.
Wimmer, A., Cederman, L.-E., and Min, B. (2009). Ethnic politics and
armed conflict: a configurational analysis of a new global data set. American Sociological Review, 74(2):316–337.
World Bank (2013). World development indicators.
Wright, J. (2008). To invest or insure?: How authoritarian time horizons
impact foreign aid effectiveness. Comparative Political Studies, 41(7):971–
1000.
47
Any Students
IMET
Annual Students Total
IMET & CTFP
5-year Students Total
-0.01
0
0.01
0.02
0.03
First Difference of Predicted Probabilities
Figure 1: Simulated Effects of IMET and IMET-CTFP Students on
Military-Backed Coups
Points represent first differences of predicted probabilities moving from the
25h percentile of Annual Students or Students (5 year sum) to the 75th
percentile. For Any Students the point represents first difference of receiving
training compared to not receiving training. The lines through the points
represent 95% confidence intervals
48
0.04
Any Spending
IMET
IMET & CTFP
Annual Spending Total
5-year Spending Total
-0.01
0
0.01
0.02
0.03
First Difference of Predicted Probabilities
Figure 2: Simulated Effects of IMET and IMET-CTFP Spending on
Military-Backed Coups
Points represent first differences of predicted probabilities moving from the
25th percentile of Spending or Spending (5 year sum) to the 75th percentile.
For Any Spending the point represents first difference of receiving training
compared to not receiving training. The lines through the points represent
95% confidence intervals
49
0.04
Students
IMET
IMET & CTFP
Spending
-0.01
0
0.01
0.02
0.03
First Difference of Predicted Probabilities
Figure 3: Simulated Effects of any IMET and combined IMET-CTFP Students and Spending on Military-Backed Coups (Matched Data)
Dichotomous treatment variable when any students or money is spent. The
lines through the points represent 95% confidence intervals
50
0.04
Any Students
Annual Students Total
Cold War
Post-Cold War
5-year Students Total
Any Spending
Annual Spending Total
5-year Spending Total
-0.01
0
0.01
0.02
0.03
First Difference of Predicted Probabilities
Figure 4: Simulated Effects of IMET on Military-Backed Coups (Cold War
and Post Cold War periods separated)
Points represent first differences of predicted probabilities moving from the
25th percentile of Students or Spending or Students or Spending (5 year
sum) to the 75th percentile. For Any Students or Spending the point represents first difference of receiving training compared to not receiving training.
The lines through the points represent 95% confidence intervals
51
0.04
Any Students
Annual Students Total
Democracies
Autocracies
5-year Students Total
Any Spending
Annual Spending Total
5-year Spending Total
-0.01
0
0.01
0.02
0.03
First Difference of Predicted Probabilities
Figure 5: Simulated Effects of IMET on Military-Backed Coups (Democracies and Autocracies separated)
Points represent first differences of predicted probabilities moving from the
25th percentile of Students or Spending or Students or Spending (5 year
sum) to the 75th percentile. For Any Students or Spending the point represents first difference of receiving training compared to not receiving training.
The lines through the points represent 95% confidence intervals
52
0.04
Any Students
Annual Students Total
5-year Students Total
Any Spending
Annual Spending Total
5-year Spending Total
-0.01
0
0.01
0.02
First Difference of Predicted Probabilities
Figure 6: Simulated Effects of IMET on Coups (Region Dummies)
Points represent first differences of predicted probabilities moving from the
25th percentile of Students or Spending or Students or Spending (5 year
sum) to the 75th percentile. For Any Students or Spending the point represents first difference of receiving training compared to not receiving training.
The lines through the points represent 95% confidence intervals. Region
variable set to Africa.
53
0.03
Any Students
Annual Students Total
5-year Students Total
Any Spending
Annual Spending Total
5-year Spending Total
-0.01
0
0.01
0.02
0.03
First Difference of Predicted Probabilities
Figure 7: Simulated Effects of IMET on Coups (Developing Countries)
Points represent first differences of predicted probabilities moving from the
25th percentile of Students or Spending or Students or Spending (5 year
sum) to the 75th percentile. For Any Students or Spending the point represents first difference of receiving training compared to not receiving training.
The lines through the points represent 95% confidence intervals.
54
0.04
Any Students
Annual Students Total
5-year Students Total
Any Spending
Annual Spending Total
5-year Spending Total
-0.01
0
0.01
0.02
First Difference of Predicted Probabilities
Figure 8: Simulated Effects of IMET Spending on Non-Military-Backed
Coups
Points represent first differences of predicted probabilities moving from the
25th percentile of Spending or Spending (5 year sum) to the 75th percentile.
For Any Spending the point represents first difference of receiving training
compared to not receiving training. The lines through the points represent
95% confidence intervals
55
0.03
56
-0.366***
(0.087)
-0.076
(0.183)
-5.394
(3.220)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.007
(0.108)
0.577***
(0.149)
-0.022**
(0.011)
-0.011
(0.013)
0.020
(0.572)
-0.040
(0.039)
-0.127
(0.171)
-0.325
(0.256)
-0.149***
(0.048)
0.059
(0.649)
7371
Military Backed Coup
0.667***
(0.141)
-0.385***
(0.090)
-0.121
(0.183)
-6.169
(3.499)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
-0.011
(0.109)
0.524***
(0.150)
-0.022**
(0.011)
-0.011
(0.013)
0.114
(0.564)
-0.036
(0.039)
-0.073
(0.170)
-0.361
(0.256)
-0.153***
(0.048)
0.459
(0.672)
7371
0.108***
(0.023)
Military Backed Coup
-0.389***
(0.104)
0.049
(0.209)
-4.577
(3.096)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
-0.008
(0.117)
0.494***
(0.160)
-0.018*
(0.011)
-0.013
(0.015)
0.208
(0.623)
-0.025
(0.042)
-0.208
(0.176)
-0.313
(0.258)
-0.179***
(0.050)
0.599
(0.784)
6615
0.090***
(0.022)
Military Backed Coup
-0.359***
(0.087)
-0.075
(0.183)
-5.280
(3.149)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.020
(0.108)
0.581***
(0.149)
-0.021**
(0.011)
-0.010
(0.013)
0.065
(0.569)
-0.038
(0.039)
-0.115
(0.171)
-0.333
(0.256)
-0.150***
(0.048)
0.025
(0.649)
7371
0.592***
(0.141)
Military Backed Coup
-0.366***
(0.088)
-0.113
(0.182)
-5.563
(3.224)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.001
(0.109)
0.547***
(0.150)
-0.022**
(0.011)
-0.010
(0.013)
0.109
(0.565)
-0.037
(0.039)
-0.091
(0.170)
-0.346
(0.256)
-0.153***
(0.048)
0.262
(0.655)
7371
0.073***
(0.017)
Military Backed Coup
Standard errors in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01. All models employ lagged independent variables,
a cubic time trend, and multiply imputed data.
Observations
(Intercept)
Years since last coup
Post-9/11
post-Cold War
Ethnic Power Relations
All Foreign Aid
Polity2
Growth
Civil War
Coup Proofing
Oil Revenue
Military Personnel
Military Spending per solider
Military Assistance
Affinity with US
GDP per capita
Spending 5 Year Total
Annual Spending Total
Spending
Students 5 year sum
Annual Students Total
Any Students
Table 1: IMET effects on probability of a military-backed coup
0.065***
(0.017)
-0.372***
(0.102)
0.044
(0.209)
-4.452
(3.290)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.117)
0.505***
(0.159)
-0.018*
(0.011)
-0.012
(0.015)
0.186
(0.624)
-0.026
(0.042)
-0.220
(0.176)
-0.293
(0.258)
-0.178***
(0.050)
0.432
(0.774)
6615
Military Backed Coup
A
Supplemental Information
A.1
Descriptive statistics and correlations
Table 2 presents our list of military-backed coups, while Table 3 shows descriptive statistics covering our variables. Figure 9 shows graphically the
correlation between our explanatory and independent variables. We see little evidence that coup proneness predicts the amount of training. Years
since last coup is, in fact, positively correlated with our explanatory variable, meaning that countries that have not recently experienced a coup event
receive more training.
[Tables 2, 3, and Figure 9 about here]
A.2
Information on multiple imputation and matching procedures
We used the Amelia II package (Honaker and King, 2010) to create five
multiply imputed datasets.We analyzed our binary dependent variable using
logistic regressions via the Zelig package, which combines datasets using
the Rubin method. This method calculates appropriate standard errors for
multiply imputed datasets by taking into both the within imputation and
the between imputation variance.
Matching searches for observations that have a similar probability of receiving the treatment (FMT in this case) given the covariates in the matching model. We matched on the dichotomous measures of any training and
any spending using nearest neighbor matching (Dehejia and Wahba, 2002,
57
153).A logistic regression including all confounding variables described above
was used to generate a predicted probability of receiving training or spending: the propensity score. Each treated observation was then matched with
its “nearest neighbor” based on propensity scores.
The matching was carried out using replacement. When matching without replacement, and there are large differences in propensity scores between
treated and untreated observations, some observations can be matched despite very different propensity scores (Dehejia and Wahba, 2002). Additionally, estimates using nearest neighbor matching without replacement depend
on the order the matching occurs (Smith and Todd, 2005). Matching with
replacement avoids these problems, trading some variance for reduced bias.21
These procedures were carried out on each multiply imputed dataset and
then the data were recombined with Zelig as described above. Figures 10
and 11 (for students) and 12 and 13 (for spending) present the distribution
of the propensity scores after matching.
A.3
Robustness checks
Tables 4-11 present the robustness checks described in the main body of the
paper. Table 12 presents the results where non-military backed coups are
the dependent variable, as described in the main body of the paper.
To explore the robustness of our principal finding, we first used an alternative measures of coups as our dependent variables. We substituted Powell
and Thyne’s measure of coups for our Military Backed Coup variable (see
21
Estimates using matching without replacement are similar to those conducted using
matching with replacement.
58
Table 13). We also used the Marshall and Marshall’s Center for Systemic
Peace coding of attempted and successful coups (see Table 14). Across all
of the models using either the measures of students or spending a significant, positive relationship persists. Across all of the models, using either
the measures of students or spending, a significant relationship persists.
We used six different operationalizations of our dependent variable in the
main body of the paper. Here we include additional transformations. First,
we used the number of students and the 5 year sum of students without
taking the natural log. We also used the total amount of spending without
taking the natural log. The results were mixed. On the full sample, the nonnormalized measure of training was insignificant. These results are being
driven by outliers occurring before the official establishment of the IMET
program, in Cambodia and Laos. Once these seven observations from the
early 1970s are dropped the results are again significant. Finally, we looked
at IMET spending per student for that year. Again we find a significant
association (Table 15).
Table 16 presents the “original” non-imputed data. Table 18 shows the
results when cubic splines are substituted for the cubic time trend.
As mentioned, while the US military provides data on officers trained in
similar programs prior to 1976, the IMET program officially began that year.
Significantly, the nine country-years that experienced the most training occurred during this period (in Cambodia and Laos). We ran models using
only data after 1976. The results, shown in Table 17, remain significant and
in the predicted direction.22
22
The IMET program has undergone a series of institutional changes across time. As
59
The rareness of coups could cause problems with estimation. When
zeros vastly outnumber ones in the data it is possible that probabilities
are estimated incorrectly with quantities of interest sometimes too large or
too small. To address this we reestimate the results using rare events logit
(King and Zeng, 2001b,a). The results, depicted in Table 19 are similar and
significant across all our measures of training or funding.
described above, in each of these there has been a change in how much emphasis has been
placed on training or course related to respect for human rights and civilian control. To
check the findings sensitivity to these changes, we replaced the cold war dummy with
dummies marking these shifts in policy. These dummies were coded to measure the period from 1976 to 1991, which marked the establishment of IMET, and the post-Leahy
amendment era change to E-IMET. Along with this alternative specification, we also fitted models with decade dummies and year dummies. The results with these variables
included are similar to the base-line model.
60
61
Country
Argentina
Bolivia
Cambodia
Congo (Brazzaville)
Haiti
Oman
Syria
Togo
Argentina
Bolivia
Morocco
Sierra Leone
Sudan
Thailand
Turkey
Uganda
Benin
Ecuador
El Salvador
Ghana
Honduras
Morocco
Afghanistan
Chile
Greece
Iraq
Laos
Rwanda
Bolivia
Central African Rep
Cyprus
Ethiopia
Lesotho
Madagascar
Niger
Portugal
Uganda
Bangladesh
Benin
Chad
Ecuador
Honduras
Nigeria
Peru
Portugal
Sudan
Uganda
Argentina
Year
1970
1970
1970
1970
1970
1970
1970
1970
1971
1971
1971
1971
1971
1971
1971
1971
1972
1972
1972
1972
1972
1972
1973
1973
1973
1973
1973
1973
1974
1974
1974
1974
1974
1974
1974
1974
1974
1975
1975
1975
1975
1975
1975
1975
1975
1975
1975
1976
Country
Bangladesh
Burundi
Cambodia
Chad
Ecuador
Lebanon
Nigeria
Sudan
Thailand
Uruguay
Angola
Bangladesh
Chad
Congo (Brazzaville)
Ethiopia
Ghana
Government of Sudan
Honduras
Pakistan
Thailand
Afghanistan
Bolivia
Ghana
Honduras
Mali
Mauritania
Nicaragua
Somalia
Afghanistan
Bolivia
El Salvador
Equatorial Guinea
Ghana
Iraq
South Korea
Spain
Bangladesh
Bolivia
Burkina Faso
GuineaBissau
Iran
Ivory Coast
Liberia
Mauritania
Tanzania
Turkey
Uganda
Zambia
Year
1976
1976
1976
1976
1976
1976
1976
1976
1976
1976
1977
1977
1977
1977
1977
1977
1977
1977
1977
1977
1978
1978
1978
1978
1978
1978
1978
1978
1979
1979
1979
1979
1979
1979
1979
1979
1980
1980
1980
1980
1980
1980
1980
1980
1980
1980
1980
1980
Country
Bangladesh
Bolivia
Central African Rep
Equatorial Guinea
Gambia
Ghana
Mauritania
Poland
Spain
Thailand
Bangladesh
Burkina Faso
Chad
Ghana
Guatemala
Iran
Kenya
Mauritania
Burkina Faso
Central African Rep
Equatorial Guinea
Ghana
Guatemala
Niger
Nigeria
Bolivia
Cameroon
Guinea
Mauritania
Pakistan
Government of Sudan
Guinea
Liberia
Nigeria
Thailand
Uganda
Equatorial Guinea
Lesotho
Phillipines
Togo
Burkina Faso
Burundi
Comoros
Fiji
Phillipines
Tunuisia
Benin
Year
1981
1981
1981
1981
1981
1981
1981
1981
1981
1981
1982
1982
1982
1982
1982
1982
1982
1982
1983
1983
1983
1983
1983
1983
1983
1984
1984
1984
1984
1984
1985
1985
1985
1985
1985
1985
1986
1986
1986
1986
1987
1987
1987
1987
1987
1987
1988
Country
Guatemala
Haiti
Myanmar
Panama
Somalia
Chad
Ethiopia
Government of Sudan
Guatemala
Haiti
Panama
Paraguay
Phillipines
Afghanistan
Argentina
Mauritania
Nigeria
Papua New Guinea
Phillipines
Sudan
Zambia
Chad
Haiti
Lesotho
Mali
Thailand
Togo
Benin
Burundi
Chad
Comoros
Madagascar
Peru
Sierra Leone
Tajikistan
Venezuela
Burundi
Chad
Libya
Nigeria
Azerbaijan
Burundi
Gambia
Lesotho
Liberia
Azerbaijan
Benin
Table 2: Military Backed Coups by Year
Year
1988
1988
1988
1988
1988
1989
1989
1989
1989
1989
1989
1989
1989
1990
1990
1990
1990
1990
1990
1990
1990
1991
1991
1991
1991
1991
1991
1992
1992
1992
1992
1992
1992
1992
1992
1992
1993
1993
1993
1993
1994
1994
1994
1994
1994
1995
1995
Country
Gambia
Iraq
Ivory Coast
Qatar
Sierra Leone
Bangladesh
Burundi
Central African Rep
Guinea
Niger
Paraguay
Sierra Leone
Cambodia
Congo (Brazzaville)
Sierra Leone
Zambia
GuineaBissau
Comoros
GuineaBissau
Honduras
Ivory Coast
Niger
Pakistan
Comoros
Djibouti
Ecuador
Fiji
GuineaBissau
Haiti
Ivory Coast
Paraguay
Burundi
Central African Rep
Ivory Coast
Venezuela
Central African Rep
GuineaBissau
Mauritania
Congo (Kinshasa
Mauritania
Togo
Fiji
Thailand
Bangladesh
Phillipines
Guinea
Mauritania
Year
1995
1995
1995
1995
1995
1996
1996
1996
1996
1996
1996
1996
1997
1997
1997
1997
1998
1999
1999
1999
1999
1999
1999
2000
2000
2000
2000
2000
2000
2000
2000
2001
2001
2001
2002
2003
2003
2003
2004
2005
2005
2006
2006
2007
2007
2008
2008
62
Military Backed Coup
Any Training
Any Spending
Training 5 year sum(logged)
Students (logged)
Spending(logged)
GDP per capita
Affinity
Military Assistance
Oil Revenue
Military Spending
Coup Proofing
Growth
Polity
Total Aid
Ethnic Power Relations
Past Coups
Statistic
7,560
7,560
7,560
6,615
7,560
7,560
6,216
6,348
6,001
5,337
5,449
5,460
6,227
5,632
6,180
5,220
7,560
N
0.031
0.449
0.455
1.355
0.271
1.377
7.603
−0.299
0.001
744.138
18,082.480
1.673
3.686
0.757
.076
1.738
0.753
Mean
0.174
0.497
0.498
3.623
3.014
4.134
1.584
0.458
0.016
3,961.313
45,467.380
0.648
6.385
7.437
.127
1.878
1.506
St. Dev.
Table 3: Descriptive Statistics
0
0
0
−2.303
−2.303
−2.303
3.998
−1.000
0.000
0.000
0.000
1.000
−51.031
−10
0.000
0.000
0
Min
1
1
1
11.093
10.430
9.888
11.591
1.000
0.831
78,588.800
1,722,499.000
5.557
106.280
10
2.223
14.00
9
Max
Years since last coup
ODA
Polity
Civil War
Economic Growth
Polity
ODA
Included Ethnic Groups
Civil War
Military Personnel
Coup Proofing
Economic Growth
post 9/11
Military Personnel
post Cold War
Coup Proofing
post Cold War
Oil Revenue
Military Spending per soldier
post 9/11
Affinity with US
Military Assistance
Oil Revenue
Military Spending per soldier
Affinity with US
Students Total
Spending 5 year sum
Military Assistance
Spending 5 year sum
GDP per capita
GDP per capita
Spending
Spending Total
Spending
Spending Total
Students 5 year sum
Students 5 year sum
Students
Students Total
Students
Military Backed Coup
Military Backed Coup
Included Ethnic Groups
Years since last coup
−1 −0.8 −0.6 −0.4 −0.2
0
0.2
0.4
0.6
0.8
1
Figure 9: Correlation matrix of main explanatory, confounding, and dependent variables
63
Distribution of Propensity Scores
Unmatched Treatment Units
Matched Treatment Units
Matched Control Units
Unmatched Control Units
0.0
0.2
0.4
0.6
0.8
1.0
Propensity Score
Figure 10: Propensity scores for matched dataset (IMET training, dataset
1)
64
0.4
1.5
0.0
Density
1.5
0.0
0.8
0.0
0.4
0.8
Raw Control
Matched Control
0.4
0.8
0.0
Density
0.0
1.0
Propensity Score
1.5
Propensity Score
0.0
Density
Matched Treated
0.0
Density
Raw Treated
0.0
Propensity Score
0.4
0.8
Propensity Score
Figure 11: Propensity scores for matched dataset (IMET training, dataset
1)
65
Distribution of Propensity Scores
Unmatched Treatment Units
Matched Treatment Units
Matched Control Units
Unmatched Control Units
0.0
0.2
0.4
0.6
0.8
1.0
Propensity Score
Figure 12: Propensity scores for matched dataset (IMET spending, dataset
1)
66
0.4
1.5
0.0
Density
1.5
0.0
0.8
0.0
0.4
0.8
Raw Control
Matched Control
Density
0.0
0.4
0.8
0.0 1.0
Propensity Score
1.5
Propensity Score
0.0
Density
Matched Treated
0.0
Density
Raw Treated
0.0
Propensity Score
0.4
0.8
Propensity Score
Figure 13: Propensity scores for matched dataset (IMET spending, dataset
1)
67
68
-0.364***
(0.088)
-0.077
(0.183)
-5.398
(3.221)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.009
(0.110)
0.581***
(0.149)
-0.022**
(0.011)
-0.011
(0.013)
0.040
(0.571)
-0.039
(0.039)
-0.123
(0.171)
-0.354
(0.256)
-0.150***
(0.048)
0.058
(0.653)
7371
-0.384***
(0.090)
-0.121
(0.183)
-6.173
(3.496)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
-0.011
(0.109)
0.526***
(0.150)
-0.022**
(0.011)
-0.012
(0.013)
0.126
(0.563)
-0.036
(0.039)
-0.072
(0.170)
-0.394
(0.256)
-0.154***
(0.048)
0.457
(0.673)
7371
0.106***
(0.023)
(2)
Military Backed Coup
-0.386***
(0.104)
0.047
(0.209)
-4.539
(3.067)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
-0.004
(0.117)
0.494***
(0.160)
-0.018*
(0.011)
-0.013
(0.015)
0.239
(0.621)
-0.024
(0.042)
-0.218
(0.176)
-0.318
(0.258)
-0.181***
(0.050)
0.566
(0.782)
6615
0.084***
(0.022)
(3)
Military Backed Coup
-0.360***
(0.089)
-0.077
(0.183)
-5.316
(3.181)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.017
(0.108)
0.583***
(0.149)
-0.021**
(0.011)
-0.011
(0.013)
0.062
(0.571)
-0.038
(0.039)
-0.116
(0.172)
-0.357
(0.256)
-0.151***
(0.048)
0.040
(0.657)
7371
0.591***
(0.149)
(4)
Military Backed Coup
-0.360***
(0.089)
-0.077
(0.183)
-5.316
(3.181)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.017
(0.108)
0.583***
(0.149)
-0.021**
(0.011)
-0.011
(0.013)
0.062
(0.571)
-0.038
(0.039)
-0.116
(0.172)
-0.357
(0.256)
-0.151***
(0.048)
0.040
(0.657)
7371
0.591***
(0.149)
(5)
Military Backed Coup
Standard errors in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01. All models employ lagged independent variables,
a cubic time trend, and multiply imputed data.
Observations
(Intercept)
Years since last coup
Post-9/11
post-Cold War
Ethnic Power Relations
All Foreign Aid
Polity2
Growth
Civil War
Coup Proofing
Oil Revenue
Military Personnel
Military Spending per solider
Military Assistance
Affinity with US
GDP per capita
Spending 5 Year Total
Annual Spending Total
Spending
Students 5 year sum
Annual Students Total
Any Students
(1)
Military Backed Coups
0.636***
(0.147)
Table 4: Combined IMET and CTFP training’s effect on probability of a military-backed coup
0.071***
(0.017)
-0.366***
(0.088)
-0.113
(0.182)
-5.570
(3.223)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.001
(0.109)
0.548***
(0.150)
-0.022**
(0.011)
-0.010
(0.013)
0.119
(0.564)
-0.036
(0.039)
-0.090
(0.170)
-0.381
(0.256)
-0.154***
(0.048)
0.264
(0.655)
6615
(6)
Military Backed Coup
Table 5: Training and Spending’s effect on probability of a military-backed
coup (matched data)
Any Students
(IMET)
Any Spending
(IMET)
Any Students
(Combined)
Any Spending
(Combined)
GDP per capita
Affinity
Military Assistance
Military Spending per soldier
Military Personnel
Oil Revenue
Coup Proofing
Civil War
Economic Growth
Polity2
Total Foreign Aid
Ethnic Power Relations
post Cold War
post 9/11
Years since last coup
(Intercept)
Military Backed Coup
0.573***
(0.193)
Military Backed Coup
Military Backed Coup
Military Backed Coup
0.583**
(0.215)
0.604***
(0.184)
-0.315***
(0.104)
-0.025
(0.224)
-5.362
(3.413)
0.000
(0.000)
-0.001
(0.001)
-0.000
(0.000)
0.009
(0.119)
0.487***
(0.178)
-0.021*
(0.012)
-0.004
(0.014)
0.092
(0.651)
-0.043
(0.046)
-0.110
(0.186)
-0.285
(0.266)
-0.133**
(0.055)
-0.143
(0.808)
-0.325***
(0.098)
0.016
(0.243)
-6.058
(4.770)
0.000
(0.000)
-0.001
(0.001)
-0.000
(0.000)
0.070
(0.118)
0.496***
(0.179)
-0.019
(0.015)
-0.004
(0.014)
0.101
(0.716)
-0.050
(0.045)
-0.107
(0.193)
-0.289
(0.270)
-0.143**
(0.056)
-0.195
(0.734)
Standard errors in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01.
All models employ lagged independent variables,
a cubic time trend, and multiply imputed data.
69
-0.338***
(0.105)
-0.013
(0.218)
-5.353
(3.576)
0.000
(0.000)
-0.001
(0.001)
-0.000
(0.000)
0.026
(0.120)
0.492***
(0.173)
-0.022
(0.018)
-0.002
(0.014)
-0.088
(0.808)
-0.047
(0.047)
-0.129
(0.199)
-0.304
(0.264)
-0.146***
(0.055)
0.000
(0.847)
0.644***
(0.201)
-0.332***
(0.114)
0.018
(0.212)
-5.466
(3.347)
0.000
(0.000)
-0.001
(0.001)
-0.000
(0.000)
0.049
(0.122)
0.517***
(0.187)
-0.026
(0.016)
-0.004
(0.014)
0.051
(0.840)
-0.052
(0.047)
-0.056
(0.192)
-0.333
(0.269)
-0.143**
(0.058)
-0.174
(0.920)
70
-0.336***
(0.108)
0.084
(0.211)
-6.309
(3.682)
0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
-0.123
(0.146)
0.647***
(0.201)
-0.014
(0.016)
-0.022
(0.018)
0.494
(0.807)
-0.048
(0.048)
-0.401***
(0.113)
0.232
(0.816)
3591
-0.357***
(0.112)
0.023
(0.213)
-7.074
(3.977)
0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
-0.143
(0.147)
0.573***
(0.203)
-0.014
(0.016)
-0.022
(0.017)
0.558
(0.802)
-0.044
(0.048)
-0.410***
(0.113)
0.652
(0.849)
3591
0.109***
(0.028)
(2)
Military Backed Coup
-0.359**
(0.146)
0.244
(0.245)
-5.282
(3.622)
0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
-0.181
(0.168)
0.555**
(0.232)
-0.006
(0.017)
-0.027
(0.022)
0.575
(0.934)
-0.035
(0.055)
-0.460***
(0.121)
0.909
(1.120)
2835
0.093***
(0.028)
(3)
Military Backed Coup
Standard errors in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01.
All models employ lagged independent variables,
a cubic time trend, and multiply imputed data.
Observations
(Intercept)
Years since last coup
Ethnic Power Relations
All Foreign Aid
Polity2
Growth
Civil War
Coup Proofing
Oil Revenue
Military Personnel
Military Spending per solider
Military Assistance
Affinity with US
GDP per capita
Spending 5 Year Total
Annual Spending Total
Spending
Students 5 year sum
Annual Students Total
Any Students
(1)
Military Backed Coup
0.655***
(0.175)
-0.327***
(0.108)
0.089
(0.211)
-6.184
(3.610)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
-0.100
(0.144)
0.652***
(0.201)
-0.014
(0.016)
-0.022
(0.018)
0.552
(0.797)
-0.045
(0.048)
-0.407***
(0.113)
0.181
(0.811)
3591
0.563***
(0.174)
(4)
Military Backed Coup
Table 6: IMET training’s effect on probability of any coup (Cold War)
-0.336***
(0.109)
0.051
(0.211)
-6.476
(3.689)
0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
-0.124
(0.145)
0.607***
(0.202)
-0.015
(0.016)
-0.021
(0.018)
0.585
(0.794)
-0.043
(0.048)
-0.412***
(0.113)
0.427
(0.822)
3591
0.072***
(0.020)
(5)
Military Backed Coup
0.068***
(0.021)
-0.339**
(0.143)
0.256
(0.245)
-5.152
(3.843)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
-0.169
(0.168)
0.569**
(0.232)
-0.006
(0.017)
-0.026
(0.022)
0.560
(0.933)
-0.035
(0.055)
-0.461***
(0.121)
0.717
(1.100)
2835
(6)
Military Backed Coup
71
-0.444***
(0.144)
-1.010*
(0.535)
-22.027
(26.768)
-0.000
(0.000)
0.000
(0.000)
0.000
(0.000)
0.142
(0.159)
0.498**
(0.234)
-0.035**
(0.015)
0.014
(0.020)
-0.328
(0.995)
-0.045
(0.066)
-0.601**
(0.289)
-0.074
(0.072)
-0.415
(1.064)
3780
(1)
Military Backed Coup
0.701***
(0.258)
-0.459***
(0.147)
-0.961*
(0.537)
-23.273
(26.885)
-0.000
(0.000)
0.000
(0.000)
0.000
(0.000)
0.133
(0.160)
0.469**
(0.234)
-0.035**
(0.015)
0.012
(0.021)
-0.193
(0.981)
-0.038
(0.066)
-0.610**
(0.289)
-0.080
(0.072)
0.069
(1.086)
3780
0.106**
(0.044)
(2)
Military Backed Coup
-0.438***
(0.144)
-1.003*
(0.539)
-22.211
(26.402)
-0.000
(0.000)
0.000
(0.000)
0.000
(0.000)
0.143
(0.160)
0.478**
(0.234)
-0.033**
(0.015)
0.012
(0.021)
-0.010
(0.963)
-0.032
(0.066)
-0.628**
(0.289)
-0.072
(0.072)
-0.287
(1.053)
3780
0.084**
(0.042)
(3)
Military Backed Coup
Standard errors in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01.
All models employ lagged independent variables,
a cubic time trend, and multiply imputed data.
Observations
(Intercept)
Years since last coup
Post-9/11
Ethnic Power Relations
All Foreign Aid
Polity2
Growth
Civil War
Coup Proofing
Oil Revenue
Military Personnel
Military Spending per solider
Military Assistance
Affinity with US
GDP per capita
Spending 5 Year Total
Annual Spending Total
Spending
Students 5 year sum
Annual Students Total
Any Students
-0.439***
(0.144)
-1.027*
(0.536)
-21.735
(26.598)
-0.000
(0.000)
0.000
(0.000)
0.000
(0.000)
0.144
(0.159)
0.506**
(0.234)
-0.035**
(0.015)
0.015
(0.020)
-0.302
(0.992)
-0.045
(0.066)
-0.622**
(0.289)
-0.071
(0.072)
-0.441
(1.066)
3780
0.661**
(0.257)
(4)
Military Backed Coup
-0.444***
(0.145)
-1.024*
(0.535)
-22.561
(26.689)
-0.000
(0.000)
0.000
(0.000)
0.000
(0.000)
0.130
(0.160)
0.480**
(0.233)
-0.035**
(0.015)
0.015
(0.021)
-0.231
(0.984)
-0.042
(0.066)
-0.620**
(0.289)
-0.076
(0.072)
-0.143
(1.070)
3780
0.076**
(0.032)
(5)
Military Backed Coup
Table 7: IMET training’s effect on probability of any coup (Post Cold War)
0.061*
(0.032)
-0.424***
(0.142)
-1.046*
(0.537)
-21.802
(26.171)
-0.000
(0.000)
0.000
(0.000)
0.000
(0.000)
0.142
(0.160)
0.487**
(0.233)
-0.033**
(0.015)
0.015
(0.021)
-0.031
(0.961)
-0.034
(0.066)
-0.624**
(0.289)
-0.068
(0.072)
-0.452
(1.050)
3780
(6)
Military Backed Coup
72
-0.328***
(0.099)
0.163
(0.202)
-5.799
(3.474)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.021
(0.128)
0.548***
(0.166)
-0.021*
(0.011)
0.002
(0.019)
0.335
(0.613)
-0.013
(0.045)
-0.079
(0.192)
-0.572*
(0.332)
-0.239***
(0.061)
0.008
(0.705)
2779
(1)
Military Backed Coup
0.583***
(0.153)
-0.379*
(0.207)
-1.309**
(0.572)
-17.501
(29.422)
0.000
(0.000)
0.000
(0.001)
-0.000
(0.000)
-0.174
(0.266)
0.624
(0.412)
-0.024
(0.037)
-0.260*
(0.135)
-1.196
(1.949)
-0.118
(0.098)
-0.124
(0.412)
-0.411
(0.460)
0.021
(0.119)
1.834
(1.728)
2779
0.124**
(0.060)
(2)
Military Backed Coup
-0.406*
(0.214)
-1.378**
(0.667)
-18.327
(32.446)
0.000
(0.000)
-0.000
(0.001)
-0.000
(0.000)
-0.184
(0.261)
0.696
(0.429)
-0.019
(0.038)
-0.224
(0.138)
-0.906
(1.984)
-0.119
(0.101)
-0.089
(0.434)
-0.497
(0.471)
0.023
(0.124)
1.497
(1.848)
2612
0.106
(0.066)
(3)
Military Backed Coup
Standard errors in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01.
All models employ lagged independent variables,
a cubic time trend, and multiply imputed data.
Observations
(Intercept)
Years since last coup
Post-9/11
post-Cold War
Ethnic Power Relations
All Foreign Aid
Polity2
Growth
Civil War
Coup Proofing
Oil Revenue
Military Personnel
Military Spending per solider
Military Assistance
Affinity with US
GDP per capita
Spending 5 Year Total
Annual Spending Total
Spending
Students 5 year sum
Annual Students Total
Any Students
-0.352*
(0.203)
-1.320**
(0.570)
-18.024
(28.815)
0.000
(0.000)
0.000
(0.001)
-0.000
(0.000)
-0.152
(0.263)
0.697*
(0.412)
-0.025
(0.037)
-0.259*
(0.133)
-1.359
(1.990)
-0.132
(0.098)
-0.198
(0.418)
-0.396
(0.462)
0.015
(0.119)
1.302
(1.735)
2779
0.888**
(0.426)
(4)
Military Backed Coup
-0.361*
(0.205)
-1.345**
(0.572)
-19.111
(29.650)
0.000
(0.000)
0.000
(0.001)
-0.000
(0.000)
-0.181
(0.267)
0.614
(0.414)
-0.025
(0.037)
-0.256*
(0.135)
-1.176
(1.968)
-0.125
(0.099)
-0.166
(0.416)
-0.404
(0.460)
0.019
(0.119)
0.262
(0.655)
2779
0.106**
(0.049)
(5)
Military Backed Coup
Table 8: IMET training’s effect on probability of any coup (Democracies only)
0.101*
(0.058)
-0.386*
(0.214)
-1.406**
(0.667)
-21.730
(32.432)
0.000
(0.000)
-0.000
(0.001)
-0.000
(0.000)
-0.196
(0.261)
0.686
(0.430)
-0.021
(0.038)
-0.222
(0.139)
-0.803
(2.018)
-0.123
(0.102)
-0.131
(0.438)
-0.475
(0.472)
0.024
(0.124)
1.214
(1.867)
2612
(6)
Military Backed Coup
73
-0.328***
(0.099)
0.163
(0.202)
-5.799
(3.474)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.021
(0.128)
0.548***
(0.166)
-0.021*
(0.011)
0.002
(0.019)
0.335
(0.613)
-0.013
(0.045)
-0.079
(0.192)
-0.572*
(0.332)
-0.239***
(0.061)
0.008
(0.705)
4683
-0.344***
(0.103)
0.110
(0.203)
-6.448
(3.735)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.015
(0.128)
0.502***
(0.167)
-0.021*
(0.012)
0.002
(0.019)
0.410
(0.606)
-0.012
(0.045)
-0.029
(0.192)
-0.605*
(0.332)
-0.244***
(0.061)
0.351
(0.731)
4683
0.092***
(0.025)
(2)
Military Backed Coup
-0.340***
(0.120)
0.320
(0.228)
-4.580
(3.288)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.017
(0.141)
0.457**
(0.181)
-0.018
(0.011)
-0.005
(0.022)
0.445
(0.674)
0.005
(0.049)
-0.166
(0.197)
-0.530
(0.334)
-0.278***
(0.063)
0.483
(0.874)
4088
0.076***
(0.024)
(3)
Military Backed Coup
Standard errors in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01.
All models employ lagged independent variables,
a cubic time trend, and multiply imputed data.
Observations
(Intercept)
Years since last coup
Post-9/11
post-Cold War
Ethnic Power Relations
All Foreign Aid
Polity2
Growth
Civil War
Coup Proofing
Oil Revenue
Military Personnel
Military Spending per solider
Military Assistance
Affinity with US
GDP per capita
Spending 5 Year Total
Annual Spending Total
Spending
Students 5 year sum
Annual Students Total
Any Students
(1)
Military Backed Coup
0.583***
(0.153)
-0.319***
(0.099)
0.166
(0.202)
-5.678
(3.394)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.034
(0.127)
0.554***
(0.166)
-0.020*
(0.011)
0.002
(0.019)
0.389
(0.608)
-0.011
(0.045)
-0.070
(0.192)
-0.580*
(0.332)
-0.241***
(0.061)
-0.036
(0.705)
4683
0.503***
(0.152)
(4)
Military Backed Coup
-0.325***
(0.100)
0.128
(0.201)
-5.895
(3.461)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.023
(0.128)
0.526***
(0.166)
-0.021*
(0.011)
0.003
(0.019)
0.419
(0.605)
-0.010
(0.045)
-0.048
(0.192)
-0.593*
(0.332)
-0.244***
(0.061)
0.167
(0.712)
4683
0.061***
(0.018)
(5)
Military Backed Coup
Table 9: IMET training’s effect on probability of any coup (Autocracies only)
0.055***
(0.018)
-0.323**
(0.118)
0.328
(0.228)
-4.455
(3.469)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.021
(0.141)
0.467***
(0.180)
-0.018
(0.011)
-0.004
(0.022)
0.433
(0.674)
0.005
(0.049)
-0.176
(0.198)
-0.516
(0.333)
-0.279***
(0.063)
0.333
(0.860)
4088
(6)
Military Backed Coup
74
Any Students
-0.304***
(0.105)
0.043
(0.182)
-4.795
(3.264)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.036
(0.112)
0.581***
(0.151)
-0.023**
(0.011)
-0.002
(0.013)
0.075
(0.577)
-0.049
(0.040)
-0.170
(0.171)
-0.324
(0.255)
-0.148***
(0.048)
0.012
(0.237)
-0.348*
(0.199)
-1.401***
(0.460)
-1.256**
(0.550)
-0.130
(0.731)
7371
-0.311***
(0.107)
0.004
(0.182)
-5.339
(3.460)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.025
(0.113)
0.545***
(0.151)
-0.024**
(0.011)
-0.003
(0.013)
0.147
(0.570)
-0.046
(0.040)
-0.129
(0.171)
-0.354
(0.255)
-0.152***
(0.048)
-0.042
(0.238)
-0.400**
(0.200)
-1.410***
(0.461)
-1.263**
(0.551)
0.167
(0.853)
7371
0.091***
(0.023)
(2)
Military Backed Coup
-0.322**
(0.125)
0.163
(0.209)
-4.038
(3.393)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.034
(0.122)
0.517***
(0.161)
-0.019*
(0.011)
-0.005
(0.015)
0.183
(0.626)
-0.033
(0.042)
-0.244
(0.176)
-0.303
(0.257)
-0.179***
(0.050)
-0.014
(0.264)
-0.330
(0.214)
-1.443***
(0.537)
-0.997*
(0.563)
0.334
(0.711)
6615
0.072***
(0.023)
(3)
Military Backed Coup
-0.297***
(0.105)
0.042
(0.182)
-4.694
(3.197)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.049
(0.112)
0.582***
(0.151)
-0.023**
(0.011)
-0.002
(0.013)
0.122
(0.573)
-0.047
(0.040)
-0.159
(0.171)
-0.332
(0.255)
-0.149***
(0.048)
0.020
(0.237)
-0.341*
(0.198)
-1.421***
(0.460)
-1.270**
(0.550)
-0.161
(0.717)
7371
0.484***
(0.142)
(4)
Military Backed Coup
-0.301***
(0.105)
0.011
(0.182)
-4.899
(3.255)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.035
(0.113)
0.557***
(0.151)
-0.024**
(0.011)
-0.002
(0.013)
0.151
(0.570)
-0.046
(0.040)
-0.141
(0.171)
-0.342
(0.255)
-0.151***
(0.048)
0.006
(0.237)
-0.364*
(0.199)
-1.418***
(0.460)
-1.259**
(0.551)
0.024
(0.846)
7371
0.061***
(0.017)
(5)
Military Backed Coup
6615
0.052***
(0.017)
-0.314**
(0.124)
0.162
(0.209)
-4.035
(3.590)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.041
(0.122)
0.523***
(0.161)
-0.019*
(0.011)
-0.004
(0.015)
0.170
(0.627)
-0.034
(0.042)
-0.251
(0.176)
-0.286
(0.257)
-0.179***
(0.050)
0.029
(0.263)
-0.311
(0.214)
-1.453***
(0.537)
-0.987*
(0.563)
0.225
(6)
Military Backed Coup
Standard errors in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01. All models employ lagged independent variables,
a cubic time trend, and multiply imputed data.
Intercept
(0.712)
Observations
Oceania
Asia
Europe
Americas
Years since last coup
Post 9/11
post Cold War
Empowered Ethnic Groups
All Foreign Aid
Polity
Economic Growth
Civil War
Coup Proofing
Oil Revenue
Military Personnel
Military Spending per soldier
Military Assistance
Affinity with US
GDP per capita
Spending 5 year sum
Annual Spending Total
Any Spending
Students 5 year sum
Annual Student Total
(1)
mi
0.561***
(0.143)
Table 10: Effect of IMET spending on the probability of a coup (Region Fixed Effects)
75
Any Students
-0.301***
(0.098)
-0.013
(0.189)
-5.644
(3.369)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
-0.065
(0.113)
0.536***
(0.150)
-0.021*
(0.011)
0.005
(0.013)
0.450
(0.626)
-0.057
(0.041)
-0.178
(0.176)
-0.380
(0.261)
-0.138***
(0.050)
-0.045
(0.700)
5044
-0.322***
(0.101)
-0.053
(0.190)
-6.261
(3.594)
-0.000
(0.000)
-0.001
(0.000)
-0.000
(0.000)
-0.070
(0.114)
0.497***
(0.151)
-0.021*
(0.011)
0.004
(0.013)
0.513
(0.617)
-0.053
(0.041)
-0.131
(0.175)
-0.406
(0.261)
-0.141***
(0.050)
0.307
(0.729)
5044
0.085***
(0.024)
(2)
Military Backed Coup
-0.322***
(0.114)
0.153
(0.215)
-4.618
(3.085)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
-0.043
(0.122)
0.472***
(0.161)
-0.017
(0.011)
0.001
(0.014)
0.562
(0.679)
-0.033
(0.043)
-0.242
(0.180)
-0.347
(0.263)
-0.162***
(0.051)
0.374
(0.825)
4524
0.070***
(0.023)
(3)
Military Backed Coup
-0.292***
(0.098)
-0.010
(0.189)
-5.529
(3.296)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
-0.053
(0.113)
0.539***
(0.150)
-0.021*
(0.011)
0.006
(0.013)
0.498
(0.620)
-0.056
(0.041)
-0.168
(0.175)
-0.387
(0.260)
-0.139***
(0.050)
-0.085
(0.701)
5044
0.489***
(0.144)
(4)
Military Backed Coup
-0.302***
(0.099)
-0.041
(0.189)
-5.765
(3.356)
-0.000
(0.000)
-0.001
(0.000)
-0.000
(0.000)
-0.062
(0.113)
0.515***
(0.150)
-0.021*
(0.011)
0.006
(0.013)
0.531
(0.615)
-0.054
(0.041)
-0.147
(0.175)
-0.394
(0.260)
-0.141***
(0.050)
0.123
(0.708)
5044
0.057***
(0.017)
(5)
Military Backed Coup
0.051***
(0.017)
-0.303***
(0.112)
0.153
(0.215)
-4.529
(3.234)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
-0.038
(0.121)
0.481***
(0.160)
-0.017
(0.011)
0.002
(0.014)
0.560
(0.679)
-0.034
(0.043)
-0.252
(0.180)
-0.331
(0.262)
-0.161***
(0.051)
0.218
(0.815)
4524
(6)
Military Backed Coup
Standard errors in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01. All models employ lagged independent variables,
a cubic time trend, and multiply imputed data.
Observations
(Intercept)
Years since last coup
Post-9/11
post-Cold War
Ethnic Power Relations
All Foreign Aid
Polity2
Growth
Civil War
Coup Proofing
Oil Revenue
Military Personnel
Military Spending per solider
Military Assistance
Affinity with US
GDP per capita
Spending 5 Year Total
Annual Spending Total
Spending
Students 5 year sum
Annual Students Total
(1)
mi
0.563***
(0.145)
Table 11: IMET training’s effect on probability of a military-backed coup (Developing Countries Only)
76
-0.077
(0.216)
-1.163**
(0.573)
-3.650
(4.905)
-0.000
(0.000)
-0.002
(0.001)
0.000
(0.000)
-0.301
(0.375)
0.678
(0.440)
-0.004
(0.025)
-0.053
(0.033)
-0.357
(2.030)
-0.021
(0.113)
0.425
(0.463)
-0.215
(0.648)
-0.119
(0.129)
-4.107**
(1.631)
7371
0.009
(0.370)
Non-Military Coup
-0.072
(0.215)
-1.172**
(0.573)
-3.706
(4.900)
-0.000
(0.000)
-0.002
(0.001)
0.000
(0.000)
-0.285
(0.377)
0.688
(0.440)
-0.004
(0.025)
-0.051
(0.033)
-0.293
(2.001)
-0.019
(0.113)
0.431
(0.463)
-0.212
(0.649)
-0.120
(0.129)
-4.168**
(1.632)
7371
-0.020
(0.064)
Non-Military Coup
-0.132
(0.232)
-1.805**
(0.708)
-0.073
(4.005)
-0.000
(0.000)
-0.002
(0.001)
0.000
(0.000)
-0.305
(0.436)
0.580
(0.455)
-0.017
(0.025)
-0.045
(0.035)
-1.176
(2.287)
-0.026
(0.113)
0.636
(0.550)
-1.438
(0.945)
-0.357**
(0.158)
-2.807
(1.776)
6615
0.047
(0.068)
Non-Military Coup
-0.076
(0.216)
-1.168**
(0.573)
-3.671
(4.907)
-0.000
(0.000)
-0.002
(0.001)
0.000
(0.000)
-0.296
(0.374)
0.679
(0.440)
-0.004
(0.025)
-0.053
(0.033)
-0.334
(2.020)
-0.020
(0.113)
0.429
(0.463)
-0.215
(0.649)
-0.119
(0.129)
-4.106**
(1.630)
7371
-0.031
(0.369)
Non-Military Coup
-0.075
(0.216)
-1.168**
(0.573)
-3.687
(4.901)
-0.000
(0.000)
-0.002
(0.001)
0.000
(0.000)
-0.293
(0.377)
0.682
(0.440)
-0.004
(0.025)
-0.052
(0.033)
-0.320
(2.011)
-0.020
(0.113)
0.430
(0.463)
-0.213
(0.649)
-0.119
(0.129)
-4.125**
(1.629)
7371
-0.007
(0.046)
Non-Military Coup
0.027
(0.049)
-0.121
(0.234)
-1.805**
(0.709)
-0.113
(4.163)
-0.000
(0.000)
-0.002
(0.001)
-0.000
(0.000)
-0.295
(0.431)
0.594
(0.457)
-0.017
(0.025)
-0.044
(0.035)
-1.152
(2.291)
-0.023
(0.113)
0.633
(0.554)
-1.419
(0.941)
-0.356**
(0.159)
-2.907
(1.767)
6615
Standard errors in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01. All models employ lagged independent variables,
a cubic time trend, and multiply imputed data.
Observations
(Intercept)
Years since last coup
Post-9/11
post-Cold War
Ethnic Power Relations
All Foreign Aid
Polity2
Growth
Civil War
Coup Proofing
Oil Revenue
Military Personnel
Military Spending per solider
Military Assistance
Affinity with US
GDP per capita
Spending 5 Year Total
Annual Spending Total
Spending
Students 5 year sum
Annual Students Total
Any Students
Non-Military Coup
Table 12: IMET effects on probability of non-military-backed coup
77
Any Students
-0.359***
(0.089)
-0.102
(0.199)
-6.358
(4.224)
-0.000
(0.000)
-0.000
(0.000)
0.000
(0.000)
-0.071
(0.120)
0.472***
(0.167)
-0.022*
(0.011)
-0.011
(0.014)
-0.309
(0.687)
-0.041
(0.042)
-0.128
(0.188)
-0.340
(0.295)
-0.239***
(0.060)
0.174
(0.670)
7371
-0.383***
(0.090)
-0.139
(0.199)
-6.947
(4.361)
-0.000
(0.000)
-0.000
(0.000)
0.000
(0.000)
-0.100
(0.121)
0.415**
(0.168)
-0.023**
(0.011)
-0.012
(0.014)
-0.222
(0.687)
-0.039
(0.042)
-0.080
(0.188)
-0.371
(0.295)
-0.242***
(0.060)
0.617
(0.681)
7371
0.115***
(0.025)
(2)
PT Coup
-0.401***
(0.108)
-0.003
(0.232)
-5.903
(4.059)
-0.000
(0.000)
-0.000
(0.000)
0.000
(0.000)
-0.108
(0.133)
0.375**
(0.180)
-0.018
(0.012)
-0.013
(0.015)
-0.235
(0.737)
-0.032
(0.045)
-0.224
(0.193)
-0.341
(0.298)
-0.273***
(0.062)
0.868
(0.826)
6615
0.096***
(0.025)
(3)
PT Coup
-0.353***
(0.089)
-0.100
(0.199)
-6.267
(4.152)
-0.000
(0.000)
-0.000
(0.000)
0.000
(0.000)
-0.060
(0.119)
0.476***
(0.167)
-0.022*
(0.011)
-0.011
(0.014)
-0.264
(0.683)
-0.040
(0.042)
-0.118
(0.188)
-0.350
(0.295)
-0.239***
(0.060)
0.137
(0.671)
7371
0.600***
(0.152)
(4)
PT Coup
-0.362***
(0.089)
-0.135
(0.198)
-6.479
(4.190)
-0.000
(0.000)
-0.000
(0.000)
0.000
(0.000)
-0.084
(0.120)
0.438***
(0.168)
-0.023**
(0.011)
-0.010
(0.014)
-0.229
(0.682)
-0.040
(0.042)
-0.097
(0.188)
-0.363
(0.295)
-0.242***
(0.060)
0.396
(0.674)
7371
0.077***
(0.018)
(5)
PT Coup
Standard errors in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01.
All models employ lagged independent variables,
a cubic time trend, and multiply imputed data.
Observations
(Intercept)
Years since last coup
Post-9/11
post-Cold War
Ethnic Power Relations
All Foreign Aid
Polity2
Growth
Civil War
Coup Proofing
Oil Revenue
Military Personnel
Military Spending per solider
Military Assistance
Affinity with US
GDP per capita
Spending 5 Year Total
Annual Spending Total
Spending
Students 5 year sum
Annual Students Total
(1)
PT Coup
0.660***
(0.153)
0.070***
(0.019)
-0.381***
(0.106)
-0.010
(0.232)
-5.999
(4.226)
-0.000
(0.000)
-0.000
(0.000)
0.000
(0.000)
-0.099
(0.132)
0.386**
(0.179)
-0.018
(0.012)
-0.012
(0.015)
-0.258
(0.737)
-0.033
(0.045)
-0.233
(0.193)
-0.323
(0.298)
-0.274***
(0.062)
0.678
(0.815)
6615
(6)
PT Coup
Table 13: IMET training’s effect on probability of any coup (Powell and Thyne)
78
-0.379***
(0.083)
-0.104
(0.183)
-4.891**
(2.200)
-0.000
(0.000)
-0.001*
(0.000)
-0.000
(0.000)
0.014
(0.111)
0.562***
(0.151)
-0.021*
(0.011)
-0.011
(0.012)
-0.087
(0.604)
-0.019
(0.041)
-0.093
(0.170)
-0.415
(0.256)
-0.158***
(0.049)
0.055
(0.636)
7371
-0.399***
(0.086)
-0.153
(0.183)
-5.676**
(2.390)
-0.000
(0.000)
-0.001**
(0.000)
-0.000
(0.000)
0.001
(0.112)
0.507***
(0.152)
-0.021*
(0.011)
-0.012
(0.012)
0.014
(0.594)
-0.013
(0.041)
-0.037
(0.169)
-0.459*
(0.256)
-0.161***
(0.049)
0.480
(0.664)
7371
0.113***
(0.023)
(2)
MM Coup
-0.408***
(0.093)
0.054
(0.209)
-3.912*
(2.161)
-0.000
(0.000)
-0.001*
(0.000)
-0.000
(0.000)
0.009
(0.120)
0.479***
(0.160)
-0.019
(0.011)
-0.014
(0.014)
0.025
(0.659)
0.000
(0.043)
-0.184
(0.175)
-0.395
(0.258)
-0.189***
(0.050)
0.675
(0.721)
6615
0.096***
(0.023)
(3)
MM Coup
-0.372***
(0.083)
-0.102
(0.184)
-4.802**
(2.170)
-0.000
(0.000)
-0.001*
(0.000)
-0.000
(0.000)
0.028
(0.111)
0.565***
(0.151)
-0.020*
(0.011)
-0.011
(0.012)
-0.041
(0.602)
-0.017
(0.041)
-0.081
(0.170)
-0.421*
(0.256)
-0.159***
(0.049)
0.020
(0.635)
7371
0.651***
(0.144)
(4)
MM Coup
Standard errors in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01.
All models employ lagged independent variables,
a cubic time trend, and multiply imputed data.
Observations
(Intercept)
Years since last coup
Post-9/11
post-Cold War
Ethnic Power Relations
All Foreign Aid
Polity2
Growth
Civil War
Coup Proofing
Oil Revenue
Military Personnel
Military Spending per solider
Military Assistance
Affinity with US
GDP per capita
Spending 5 Year Total
Annual Spending Total
Spending
Students 5 year sum
Annual Students Total
Any Students
(1)
MM Coup
0.728***
(0.145)
-0.380***
(0.084)
-0.143
(0.183)
-5.103**
(2.215)
-0.000
(0.000)
-0.001**
(0.000)
-0.000
(0.000)
0.008
(0.112)
0.529***
(0.151)
-0.021*
(0.011)
-0.010
(0.012)
0.010
(0.597)
-0.015
(0.041)
-0.057
(0.169)
-0.441*
(0.256)
-0.162***
(0.049)
0.282
(0.643)
7371
0.079***
(0.017)
(5)
MM Coup
0.068***
(0.017)
-0.389***
(0.092)
0.047
(0.209)
-3.705
(2.214)
-0.000
(0.000)
-0.001*
(0.000)
-0.000
(0.000)
0.017
(0.120)
0.492***
(0.159)
-0.018
(0.011)
-0.013
(0.014)
0.007
(0.662)
-0.001
(0.043)
-0.196
(0.175)
-0.371
(0.258)
-0.190***
(0.050)
0.492
(0.712)
6615
(6)
MM Coup
Table 14: IMET training’s effect on probability of any coup (Source: Marshall and Marshall)
79
-0.342***
(0.087)
-0.100
(0.181)
-5.035
(3.146)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.028
(0.108)
0.590***
(0.149)
-0.022**
(0.011)
-0.010
(0.013)
0.068
(0.568)
-0.037
(0.039)
-0.100
(0.171)
-0.342
(0.256)
-0.155***
(0.048)
0.194
(0.647)
7371
0.116***
(0.028)
(1)
Military Backed Coup
-0.365***
(0.089)
-0.105
(0.180)
-5.657
(3.236)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.031
(0.108)
0.609***
(0.150)
-0.019*
(0.011)
-0.009
(0.013)
0.031
(0.582)
-0.026
(0.040)
-0.052
(0.171)
-0.360
(0.266)
-0.176***
(0.053)
0.472
(0.669)
7371
(2)
Military Backed Coup
0.145***
(0.039)
-0.304***
(0.091)
-0.147
(0.180)
-4.765
(3.414)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.103
(0.105)
0.561***
(0.154)
-0.020*
(0.010)
-0.007
(0.013)
0.406
(0.559)
0.425
(0.289)
-0.056
(0.170)
-0.350
(0.256)
-0.165***
(0.048)
-0.485
(0.713)
7364
0.000
(0.000)
(3)
Military Backed Coup
-0.320***
(0.092)
-0.166
(0.182)
-4.565
(3.445)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.070
(0.107)
0.530***
(0.152)
-0.021**
(0.010)
-0.008
(0.013)
0.410
(0.559)
0.439
(0.289)
-0.032
(0.171)
-0.364
(0.256)
-0.160***
(0.048)
-0.402
(0.725)
7371
0.001***
(0.000)
(4)
Military Backed Coup
0.000**
(0.000)
-0.299***
(0.091)
-0.163
(0.181)
-5.353
(3.228)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.079
(0.106)
0.552***
(0.150)
-0.020*
(0.010)
-0.007
(0.013)
0.390
(0.546)
0.465
(0.294)
-0.034
(0.171)
-0.354
(0.256)
-0.166***
(0.048)
-0.534
(0.718)
7371
(5)
Military Backed Coup
* p < 0.10, ** p < 0.05, *** p < 0.01
Standard errors in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01. All models employ lagged independent variables,
a cubic time trend, and multiply imputed data. Model 4 drops pre 1975 outliers.
Observations
(Intercept)
Years since last coup
post 9/11
post Cold War
Empowered Ethnic Groups
ODA
Polity
Economic Growth
Civil War
Coup Proofing
Oil Revenue
Military Personnel
Military Spending per soldier
Military Assistance
Affinity with US
GDP per capita
Spending Total
Students Total
Students/Military Personnel
(logged)
Spending/Students
(logged)
Students Total
Table 15: Robustness checks for IMET training’s effect on probability of a military-backed coup
80
-0.245**
(0.109)
0.058
(0.225)
-26.470
(25.720)
-0.000**
(0.000)
-0.001**
(0.001)
0.000
(0.000)
-0.051
(0.135)
0.621***
(0.181)
-0.018
(0.013)
-0.003
(0.015)
-0.174
(0.854)
-0.037
(0.051)
-0.328
(0.206)
-0.411
(0.321)
-0.241***
(0.068)
0.315
(0.800)
3285
-0.263**
(0.111)
0.052
(0.226)
-31.540
(27.550)
-0.000**
(0.000)
-0.001**
(0.001)
0.000
(0.000)
-0.062
(0.136)
0.597***
(0.181)
-0.019
(0.013)
-0.003
(0.015)
-0.075
(0.844)
-0.034
(0.051)
-0.298
(0.206)
-0.433
(0.321)
-0.242***
(0.068)
0.571
(0.810)
3285
0.061**
(0.030)
(2)
Military Backed Coup
-0.300**
(0.119)
0.103
(0.261)
-48.800
(37.770)
-0.000*
(0.000)
-0.001*
(0.001)
0.000
(0.000)
-0.052
(0.140)
0.647***
(0.190)
-0.019
(0.014)
-0.001
(0.016)
-0.060
(0.884)
-0.010
(0.053)
-0.390*
(0.211)
-0.410
(0.324)
-0.240***
(0.070)
0.726
(0.868)
3097
0.052**
(0.024)
(3)
Military Backed Coup
-0.238**
(0.109)
0.065
(0.225)
-23.970
(24.810)
-0.000**
(0.000)
-0.001**
(0.001)
0.000
(0.000)
-0.037
(0.135)
0.618***
(0.181)
-0.017
(0.013)
-0.002
(0.015)
-0.122
(0.849)
-0.036
(0.052)
-0.321
(0.206)
-0.421
(0.321)
-0.242***
(0.068)
0.309
(0.798)
3285
0.270
(0.182)
(4)
Military Backed Coup
-0.246**
(0.110)
0.052
(0.225)
-27.230
(26.130)
-0.000**
(0.000)
-0.001**
(0.001)
0.000
(0.000)
-0.049
(0.136)
0.606***
(0.180)
-0.018
(0.013)
-0.002
(0.015)
-0.094
(0.846)
-0.035
(0.052)
-0.309
(0.206)
-0.425
(0.321)
-0.243***
(0.068)
0.439
(0.802)
3285
0.037*
(0.022)
(5)
Military Backed Coup
Standard errors in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01. All models employ lagged independent variables,
a cubic time trend, and multiply imputed data.
Observations
(Intercept)
Years since last coup
Post-9/11
post-Cold War
Ethnic Power Relations
All Foreign Aid
Polity2
Growth
Civil War
Coup Proofing
Oil Revenue
Military Personnel
Military Spending per solider
Military Assistance
Affinity with US
GDP per capita
Spending 5 Year Total
Annual Spending Total
Spending
Students 5 year sum
Annual Students Total
Any Students
(1)
Military Backed Coups
0.364**
(0.184)
Table 16: IMET training’s effect on probability of a military-backed coup (non-imputed data)
0.050**
(0.023)
-0.289**
(0.118)
0.096
(0.261)
-48.690
(37.790)
-0.000*
(0.000)
-0.001*
(0.001)
0.000
(0.000)
-0.050
(0.140)
0.650***
(0.190)
-0.019
(0.014)
0.001
(0.016)
-0.073
(0.884)
-0.012
(0.053)
-0.390*
(0.211)
-0.404
(0.324)
-0.241***
(0.070)
0.549
(0.861)
3077
(6)
Military Backed Coup
81
-0.371***
(0.102)
0.076
(0.209)
-4.508
(3.756)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.012
(0.115)
0.543***
(0.159)
-0.018*
(0.011)
-0.012
(0.014)
-0.025
(0.635)
-0.028
(0.042)
-0.199
(0.175)
-0.277
(0.258)
-0.182***
(0.050)
0.361
(0.773)
6426
-0.386***
(0.104)
0.050
(0.209)
-4.921
(3.699)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.116)
0.503***
(0.160)
-0.019*
(0.011)
-0.013
(0.014)
0.088
(0.628)
-0.025
(0.042)
-0.155
(0.175)
-0.305
(0.258)
-0.186***
(0.050)
0.709
(0.786)
6426
0.094***
(0.025)
(2)
Military Backed Coup
-0.389***
(0.104)
0.049
(0.209)
-4.577
(3.096)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
-0.008
(0.117)
0.494***
(0.160)
-0.018*
(0.011)
-0.013
(0.015)
0.208
(0.623)
-0.025
(0.042)
-0.208
(0.176)
-0.313
(0.258)
-0.179***
(0.050)
0.599
(0.784)
6426
0.090***
(0.022)
(3)
Military Backed Coup
-0.366***
(0.102)
0.074
(0.209)
-4.464
(3.735)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.021
(0.115)
0.546***
(0.159)
-0.018*
(0.011)
-0.012
(0.014)
0.008
(0.632)
-0.028
(0.042)
-0.191
(0.175)
-0.286
(0.258)
-0.181***
(0.050)
0.334
(0.772)
6426
0.536***
(0.152)
(4)
Military Backed Coup
-0.372***
(0.102)
0.044
(0.208)
-4.685
(3.717)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.005
(0.116)
0.516***
(0.159)
-0.019*
(0.011)
-0.012
(0.014)
0.061
(0.629)
-0.027
(0.042)
-0.169
(0.175)
-0.297
(0.258)
-0.184***
(0.050)
0.544
(0.775)
6426
0.066***
(0.018)
(5)
Military Backed Coup
Standard errors in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01. All models employ lagged independent variables,
a cubic time trend, and multiply imputed data.
Observations
(Intercept)
Years since last coup
Post-9/11
post-Cold War
Ethnic Power Relations
All Foreign Aid
Polity2
Growth
Civil War
Coup Proofing
Oil Revenue
Military Personnel
Military Spending per solider
Military Assistance
Affinity with US
GDP per capita
Spending 5 Year Total
Annual Spending Total
Spending
Students 5 year sum
Annual Students Total
Any Students
(1)
Military Backed Coups
0.584***
(0.152)
Table 17: IMET training’s effect on probability of a military-backed coup (post-1976)
0.065***
(0.017)
-0.372***
(0.102)
0.044
(0.209)
-4.452
(3.290)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.117)
0.505***
(0.159)
-0.018*
(0.011)
-0.012
(0.015)
0.186
(0.624)
-0.026
(0.042)
-0.220
(0.176)
-0.293
(0.258)
-0.178***
(0.050)
0.432
(0.774)
6615
(6)
Military Backed Coup
82
-0.384***
(0.089)
0.002
(0.184)
-5.398
(3.198)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.006
(0.108)
0.537***
(0.151)
-0.019*
(0.011)
-0.009
(0.013)
-0.028
(0.581)
-0.042
(0.040)
-0.123
(0.170)
-0.308
(0.251)
-0.089***
(0.018)
0.531
(0.681)
7371
-0.403***
(0.092)
-0.040
(0.185)
-6.191
(3.466)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
-0.012
(0.109)
0.484***
(0.151)
-0.020*
(0.011)
-0.010
(0.013)
0.065
(0.572)
-0.038
(0.040)
-0.070
(0.170)
-0.345
(0.252)
-0.089***
(0.018)
0.930
(0.702)
7371
0.107***
(0.023)
(2)
Military Backed Coup
-0.405***
(0.105)
0.097
(0.211)
-4.751
(3.225)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
-0.008
(0.116)
0.475***
(0.160)
-0.016
(0.011)
-0.012
(0.015)
0.189
(0.625)
-0.027
(0.042)
-0.186
(0.175)
-0.309
(0.254)
-0.091***
(0.019)
0.790
(0.796)
6615
0.091***
(0.022)
(3)
Military Backed Coup
-0.377***
(0.089)
0.002
(0.184)
-5.284
(3.128)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.020
(0.107)
0.542***
(0.151)
-0.019*
(0.011)
-0.009
(0.013)
0.020
(0.577)
-0.040
(0.040)
-0.111
(0.170)
-0.316
(0.251)
-0.089***
(0.019)
0.490
(0.680)
7371
0.577***
(0.141)
(4)
Military Backed Coup
-0.384***
(0.090)
-0.033
(0.184)
-5.575
(3.201)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.000
(0.108)
0.506***
(0.151)
-0.020*
(0.011)
-0.009
(0.013)
0.062
(0.573)
-0.039
(0.040)
-0.087
(0.170)
-0.331
(0.251)
-0.089***
(0.018)
0.728
(0.685)
7371
0.072***
(0.017)
(5)
Military Backed Coup
Standard errors in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01. All models employ lagged independent variables,
a cubic splines, and multiply imputed data.
Observations
(Intercept)
Years since last coup
Post-9/11
post-Cold War
Ethnic Power Relations
All Foreign Aid
Polity2
Growth
Civil War
Coup Proofing
Oil Revenue
Military Personnel
Military Spending per solider
Military Assistance
Affinity with US
GDP per capita
Spending 5 Year Total
Annual Spending Total
Spending
Students 5 year sum
Annual Students Total
Any Students
(1)
Military Backed Coups
0.655***
(0.142)
Table 18: IMET training’s effect on probability of a military-backed coup (cubic splines)
0.066***
(0.017)
-0.388***
(0.103)
0.091
(0.211)
-4.631
(3.423)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.116)
0.486***
(0.160)
-0.016
(0.011)
-0.011
(0.015)
0.168
(0.627)
-0.029
(0.042)
-0.197
(0.175)
-0.289
(0.253)
-0.092***
(0.019)
0.790
(0.796)
6615
(6)
Military Backed Coup
83
-0.362***
(0.088)
-0.071
(0.182)
-5.356*
(2.923)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.006
(0.108)
0.576***
(0.149)
-0.022**
(0.011)
-0.011
(0.013)
0.092
(0.574)
-0.038
(0.039)
-0.127
(0.170)
-0.306
(0.255)
-0.145***
(0.048)
0.044
(0.651)
7371
-0.380***
(0.091)
-0.116
(0.183)
-6.154*
(3.265)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
-0.011
(0.109)
0.523***
(0.150)
-0.023**
(0.011)
-0.011
(0.013)
0.180
(0.565)
-0.034
(0.039)
-0.074
(0.170)
-0.341
(0.255)
-0.149***
(0.048)
0.439
(0.673)
7371
0.107***
(0.023)
(2)
Military Backed Coup
-0.384***
(0.103)
0.053
(0.208)
-4.434
(2.793)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
-0.010
(0.116)
0.492***
(0.159)
-0.018*
(0.011)
-0.013
(0.014)
0.277
(0.620)
-0.023
(0.042)
-0.209
(0.175)
-0.295
(0.257)
-0.173***
(0.050)
0.572
(0.780)
6615
0.090***
(0.022)
(3)
Military Backed Coup
-0.355***
(0.088)
-0.071
(0.182)
-5.238*
(2.849)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.019
(0.108)
0.580***
(0.149)
-0.022**
(0.011)
-0.010
(0.013)
0.134
(0.570)
-0.037
(0.039)
-0.116
(0.170)
-0.314
(0.255)
-0.146***
(0.048)
0.011
(0.651)
7371
0.585***
(0.140)
(4)
Military Backed Coup
-0.362***
(0.088)
-0.108
(0.182)
-5.524*
(2.945)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
0.000
(0.109)
0.546***
(0.149)
-0.023**
(0.011)
-0.010
(0.013)
0.177
(0.566)
-0.035
(0.039)
-0.092
(0.170)
-0.327
(0.255)
-0.148***
(0.048)
0.244
(0.656)
7371
0.072***
(0.016)
(5)
Military Backed Coup
Standard errors in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01. All models employ lagged independent variables,
a cubic time trend, and multiply imputed data.
Observations
(Intercept)
Years since last coup
Post-9/11
post-Cold War
Ethnic Power Relations
All Foreign Aid
Polity2
Growth
Civil War
Coup Proofing
Oil Revenue
Military Personnel
Military Spending per solider
Military Assistance
Affinity with US
GDP per capita
Spending 5 Year Total
Annual Spending Total
Spending
Students 5 year sum
Annual Students Total
Any Students
(1)
Military Backed Coups
0.658***
(0.141)
Table 19: IMET training’s effect on probability of a military-backed coup (rare-events logit)
0.065***
(0.017)
-0.367***
(0.102)
0.048
(0.208)
-4.435
(3.041)
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
-0.002
(0.116)
0.505***
(0.159)
-0.018*
(0.011)
-0.012
(0.014)
0.253
(0.622)
-0.024
(0.042)
-0.220
(0.175)
-0.274
(0.257)
-0.173***
(0.050)
0.406
(0.771)
6615
(6)
Military Backed Coup