Untitled

Why (Not) Arrest? Explaining State (Non)Compliance with the

International Criminal Tribunal for Rwanda

Mark Berlin [email protected]

Ph.D. Student Department of Political Science University of California, Irvine

Abstract: The international community is increasingly turning to international criminal tribunals (ICTs) to address grave violations of international human rights law and the laws of war. But because ICTs lack their own police force or sanctioning authority, they must rely on the willing cooperation of states to pursue and arrest indictees. Even though states are usually legally obligated pursue suspects in their territories, some states nonetheless resist. Why do some states comply with their legal obligations to aid ICTs while others do not? Existing research on this question has mostly focused on transitional states, for which noncompliance is mostly taken for granted. But a more comprehensive understanding of ICT compliance requires looking to a wider range of cases, including those cases, such as third-party states, for which the motivations for noncompliance are less obvious. To this end, this study examines state compliance with the International Criminal Tribunal for Rwanda (ICTR), which, owing to the numerous third-party states involved in arresting indictees, provides an excellent testing ground for extant theories of ICT compliance. Using crisp set qualitative comparative analysis, I offer the first test of the generalizability of these theories. I also develop and test new hypotheses to explain those cases of noncompliance for which these theories fail to account. I argue that existing instrumental and dispositional theories can explain much of the positive cases – that is, compliance successes – but that factors such as rent extraction, noncompliance constituencies, and national security are needed to account for instances of noncompliance. An event history model offers some additional support for this argument. Together these findings suggest that a generalizable model of ICT compliance – and international legal compliance more generally – requires explanations for negative cases – that is, compliance failures – that differ from those that explain successes.

Presented at the second biennial Dean Maxwell and Isle Cohen Seminar at McGill University, June 15-16, 2012

2

“Our tribunal is like a giant who has no arms and no legs. To walk and work, he needs artificial limbs. These artificial limbs are the State authorities; without their help, the Tribunal cannot operate.”

(Antonio Cassese, President of the International Criminal Tribunal for the Former Yugoslavia, address to the UN General Assembly, 7 November 1995)

The international community is increasingly turning to international courts to address grave

violations of international human rights law and the laws of war. In the early 1990s, the

formation of international criminal tribunals for Rwanda (ICTR) and the former Yugoslavia

(ICTY) signaled a renewed interest in bringing to justice government officials and other high-

level individuals responsible for mass atrocities. Following these and other “ad hoc” tribunals,

the International Criminal Court (ICC) was established in 1998 as the world’s first permanent

international court tasked with prosecuting perpetrators of genocide, war crimes, and crimes

against humanity. Since its founding, the ICC has indicted 28 individuals from seven countries,

including one former and two sitting heads of state.1 And just recently the court was injected into

international efforts to resolve the civil war in Libya.2 To its advocates, the ICC holds the

promise of facilitating conflict resolution and ending of the “culture of impunity” that has

traditionally shielded the world’s worst human rights violators from accountability.

But the increasing salience of international criminal justice in world politics has also

brought attention to its frailties. For years the highest ranking Serbian and Croatian war criminals

enjoyed protection before their governments finally succumbed to European pressure to arrest

and transfer them to the ICTY. Sudanese President, Omar al-Bashir, currently under indictment

by the ICC for genocide and other crimes, has frustrated the court by traveling to numerous

African states without arrest. Meanwhile, the ICC has been struggling to convince the new

1 The two indicted sitting heads of state were Sudanese president Omar al-Bashir (still at large) and Libyan leader Muammar Gaddafi (now deceased). The one indicted former head of state was former Ivory Coast president Laurent Gbagbo. See the website of the International Criminal Court, http://www.icc-cpi.int/. 2 “Security Council Call for War Crimes Inquiry in Libya,” New York Times 26 February 2011.

3

Libyan government that the two surviving Libyans indicted by the court – Muammar Gaddafi’s

son, Saif Gaddafi, and former intelligence chief, Abdullah Senussi – should be handed over and

tried in The Hague.

These examples underscore an important reality: when it comes to gaining custody of the

accused, international courts are at the mercy of states. ICTs lack their own police forces or

sanctioning authority, so they must depend on states to pursue, arrest, and transfer indictees. But

states’ interests do not always coincide with those of international courts. Even though states are

legally obligated to arrest indicted suspects at large in their territories, they often fail to do so.

The International Criminal Tribunal for Rwanda (ICTR) is a case in point. Of the almost 30

states where indictees were believed to be at large, about one third initially defied their legal

obligations to cooperate with the court in pursuing suspects. Why then do some states comply

with their international legal obligations to pursue indicted suspects while others do not? Recent

studies have shown that international criminal prosecutions for past atrocities may have positive

effects on peacebuilding3 and the protection of human rights4 in post-conflict societies. But if

ICTs are unable to gain custody of suspects, they largely cannot function.5 Therefore, in light of

ICTs’ capacity for affecting outcomes, it is important to understand why some states support

their work while others do not.

A growing body of empirical research has sought to make sense of state compliance with

ICTs. On the one hand, instrumental approaches focus on the particular mix of international and

3 Akhavan 2001. 4 Kim and Sikkink 2010. 5 In absentia trials are generally discouraged by ICTs. The Rome Statute of the ICC explicitly prohibits trials in absentia. See Schabas 2011, 304–308. Other tribunals are less explicit in their statutes, but practice has strongly tended towards avoiding trials in the absence of defendants. See Schabas 2006, 419–422. The Special Tribunal for Lebanon, a “hybrid” tribunal that has both international and domestic legal features and which operates under Lebanese law, is unique in that it explicitly allows for trials in absentia. See the Statute of the Special Tribunal for Lebanon, UN Doc. S/2006/893, Attachment, Art. 22.

4

domestic pressures states face.6 But these studies have mostly focused on explaining changes in

behavior for states where initial noncompliance is treated as given – particularly transitional

states that are the explicit target of ICTs and that are obligated to turn over their own citizens.7

Instrumental studies thus have had little to say about why states resist compliance in the first

place. For other, third-party states, such as the states where all ICTR indictees had fled,

motivations for noncompliance are less obvious, and thus demand closer examination. And as

demonstrated by a number of recent high-profile cases – such as those involving Sudanese

president Omar al-Bashir, former Liberian president Charles Taylor, and former Chadian

president Hissène Habré – the willingness of such third-party states to arrest wanted war

criminals is often crucial for their prosecution. Failure to examine these types of cases obscures

the full range of motivations that may drive states to harbor fugitives and thus limits scholars’

ability to construct a generalizable model of ICT compliance.

On the other hand, dispositional explanations attribute compliance behavior to deep-

seated normative commitments to the values of international justice. Strong domestic norms

relating to the respect for civil and political rights and the rule of law make compliance with

obligations of international justice the “appropriate” thing to do.8 Yet, as with instrumental

approaches, when it comes to explaining noncompliance, dispositional approaches remain

undertheorized. In this case it is unclear why the absence of liberal domestic norms in some

states, particularly third-party states, should lead to noncompliance in the first place. Further,

these studies offer little explanation for why states that lack these pro-compliance attributes still

6 Peskin 2008; Subotic 2009; Grodsky 2011; Meernik 2008; McClendon 2009. 7 In this study, I refer to those states over whose conflicts ICTs exercise jurisdiction as “target” or “transitional” states. These include, for example, Serbia, Croatia, and Bosnia (in the case of the ICTY), and Rwanda (in the case of the ICTR). I use the term “third-party state” to refer to any state other than these target states where indictees were believed to be at large. 8 Bass 2000; Kelley 2007.

5

sometimes comply, while some liberal states that otherwise exhibit strong commitments to

human rights nevertheless defy their obligations. Thus, both instrumental and dispositional

approaches (1) focus too much on monocausal explanations at the expense at leaving important

variation unexplained, (2) don’t adequately explain the phenomenon of noncompliance, and (3)

may have difficulty generalizing to third-party states.

I address the last problem by focusing on the case of the ICTR, which has so far required

the arrest warrant cooperation of at least 25 states (all third-parties), and thus provides an

excellent testing ground for the generalizability of extant explanations. I address the second

problem by developing a series of hypotheses that specifically theorize state noncompliance and

which supplement existing positive case explanations. And I address the first problem by

employing qualitative comparative analysis (QCA), a method that is designed to account for

equifinality and which can perform formal, systematic cross-case comparisons of small and

medium-N samples.9 Thus, this study engages in both theory testing (do existing explanations

account for the variation in compliance in this population?) and theory development (how can

existing theory be improved to better account for the unexplained variation?) The core of my

argument is that while either incentives or norms have the potential to motivate different types of

states to comply with their legal obligations, considerations of rent extraction, domestic politics,

and national security, can drive states to resist compliance, despite the presence of otherwise

encouraging factors. My argument suggests that compliance with ICTs is more complex than

monocausal or symmetric explanations imply.10 That is, on the one hand, more than one path to

9 Rihoux and Ragin 2008. 10 “Causal symmetry” is the assumption that “the same model explains both occurrence and nonoccurrence” of an outcome. Goertz and Mahoney forthcoming, 63.

6

compliance exists, and, on the other, noncompliance cannot necessarily be explained solely in

terms of the absence of these positive case conditions.11

Beyond the main argument, this study makes two broader contributions. First, it

demonstrates that to understand international legal compliance more broadly, scholars must

move beyond monocausal and symmetric models. The debate over compliance has often been

characterized by competing “schools” that appeal to a single meta-factor – whether incentives or

norms – in explaining behavior. Instead, scholars must be open to the possibilities that (1)

multiple paths to compliance may exist and (2) failures of compliance may require separate

explanations than successes. A study of compliance with the ICTR is especially useful for

examining the more general phenomenon of international legal compliance. Unlike with treaties,

states did not voluntarily assent to their ICTR obligations, but were obligated to comply by virtue

of UN Security Council resolution 955. Therefore, in this case, endogeneity is less of a

methodological obstacle than for so many other compliance studies that instead focus on treaty

commitments.12 Second, this paper also represents the first systematic study of not only the

ICTR, but also third-party ICT arrest warrant compliance in general. While a growing collection

of studies has brought social scientific methods to bear on explaining the politics of ICTY arrests

and transfers,13 the ICTR has received relatively scant social scientific attention,14 as has the

phenomenon of third-party ICT compliance in general. This is surprising considering the unique

opportunities for theory building and testing offered by a case with such an uncharacteristically

large population and high variation.

11 While parsimony is desirable in social science, it should not be maintained at the expense of good theory. As King, Keohane, and Verba comment, “theory should be just as complicated as all our evidence suggest.” King, Keohane, and Verba 1994, 20. 12 For a discussion of the endogeneity problem in compliance research, see Downs, Rocke, and Barsoom 1996. 13 Grodsky 2011; Lamont 2010; McClendon 2009; Meernik 2008; Peskin 2008; and Subotic 2009. 14 Notable exceptions include McClendon 2009; and Peskin 2008.

7

This study proceeds as follows. First I review the existing literature on ICT compliance,

while also placing it in the context of the scholarship on international legal compliance more

generally. My goal is to draw attention to the inadequacies of both for establishing a

generalizable model of ICT compliance. I then develop a series of negative case hypotheses that

are meant to account for cases of noncompliance unexplained by extant theory. Next I construct

two models – one that includes only the positive case hypotheses and the other that combines

them with negative case hypotheses – which I test using crisp set QCA. To provide more support

for my QCA findings, I also test my negative case hypotheses using an event history analysis

model. A conclusion summarizes and discusses the findings in light of existing theory.

Why do states (not) comply with ICTs?

For the purpose of this study, the general notion of “compliance” refers to behavior that

conforms to legally binding prescriptions. Likewise, “noncompliance” occurs when an actor’s

behavior does not conform to his or her legal obligations.15 This definition emphasizes positive

legal obligation and brackets behavior vis-à-vis international “soft” law or non-legal social

norms. International criminal tribunals (ICTs) are courts established and operated as a result of

multilateral efforts and which prosecute individuals for criminal violations of international law.16

Most ICTs create specific, legally binding obligations on states. These obligations include 15 These definitions are adapted from Oran Young’s seminal work on compliance in international relations. See, Young 1979, 4-5. In reality, compliance is not a dichotomous variable; it is usually a matter of degree, and these degrees can be difficult to measure. See Kapiszewski and Taylor forthcoming. Nevertheless, ICT compliance is relatively more amenable to dichotomization for the purposes of systematic analysis than other types of compliance, as states for the most part either are or are not attempting to arrest indictees in their territories. For a contrary view, see Rajkovic 2012. 16 I include in this universe of cases courts established by and deriving their authority from UN Security Council resolutions, such as the ICTR, ICTY, and the Special Court for Sierra Leone (SCSL), as well as the treaty-based International Criminal Court (ICC). In the former type, all UN member states, by virtue of Security Council resolutions, are under legally binding obligations to comply, while in the latter type, only ICC member states are legally obliged to do so. States that are not members of the ICC, however, can be obligated to cooperate with the court in a situation under which the Security Council has referred a case to the court and opts, via Security Council resolution, to impose such an obligation on non-party states. See Akande 2009, 343-344.

8

cooperation in the transfer of witnesses and other logistical support. One of the central and most

demanding of states’ obligations is to cooperate with ICTs in the pursuit, arrest, and transfer of

indicted suspects. ICTs lack their own police force, so they must rely on domestic law

enforcement agencies to pursue and arrest suspects.17 For the purpose of this study, “compliance

with ICTs” will refer to a state’s international legal obligations to pursue, arrest, and transfer

indicted suspects believed to be at large in its territory. While states are subject to these

obligations, ICTs’ lack of formal enforcement or sanctioning capacity means they must rely on

other tools they do have at their disposal. I touch upon some of these resources below, but the

larger point to emphasize is that while a growing body of research has focused on the efforts of

ICTs and the international community to alter the behavior of noncompliant states, very little

work attempts to explain why states initially do or do not comply in the first place.

Scholarship on explaining state compliance with ICTs has coalesced around two broad

approaches.18 Instrumental approaches point to the particular mix of domestic and international

pressures to explain why governments ultimately comply with their legal obligations. In his book

on the ICTY and ICTR, Victor Peskin details how both the international community and the

courts themselves employed various strategies – from coercion to inducements to public shaming

– to undermine government resistance and compel compliance.19 Jelena Subotic also examines

the record of the ICTY in Serbia, Croatia, and Bosnia and similarly concludes that state

compliance is a function of strategic decisions by governments and other influential domestic

17 NATO forces in the former Yugoslavia did eventually pursue and arrest some indicted ICTY suspects, but these were stabilization forces already stationed in the former Yugoslavian territories, and not deployed specifically for this purpose. See Kerr 2004, Ch. 7. Thus, the involvement of foreign or international forces in the arrest of suspects is the exception, not the norm. 18 While these approaches are not necessarily competing, studies tend favor the explanatory power of one over the other. 19 Peskin 2008. In this instance, compliance with the ICTR by Rwanda refers to that state’s obligations to provide assistance to the court in various forms permitting investigators to enter the country and allowing witnesses to travel to the court in Tanzania. No ICTR indictees were actually believed to be at large in Rwanda.

9

actors in response to international and domestic incentives.20 Both of these studies note that

incentives can be either material – such as either coercive threats in the form of sanctions or

withholding of aid, or even conciliatory inducements in the form of offers to negotiate over

particular indictments – or symbolic – such as promises of international organization

membership or avoidance of public shaming.21 Other studies, using survival analysis models,

have found that reductions in aid to or threats of sanctions against governments increase the

likelihood that suspects in their territories will be apprehended faster.22 Jack Goldsmith makes

what is perhaps the most simplistic version of the instrumentalist argument when he asserts that

such economic coercion was the only reason why former Yugoslavian states complied with the

ICTR.23 Nevertheless, whatever forms pressures take, these studies share a common analytic

insight: that is, states that initially defy their obligations to ICTs will likely change their behavior

when pressures become too costly and compliance thus becomes beneficial.24

But because these authors explicitly focus on compliance by transitional states, the

question of why states behave the way they do in the first place is largely unaddressed. In these

transitional cases, where governments are obligated to turn over members of their own in-groups,

noncompliance is largely taken for granted. Accordingly, these authors focus on explaining the

success or failure of efforts to change the behavior of initially noncompliant states. Thus

instrumental studies have little to say about the initial behavior of other, third-party states, for

which noncompliance can not be so easily assumed and which often play a critical role in ICT

success.

20 Subotic 2009. 21 Peskin 2008, 9-16; Subotic 2009, 30-32. 22 Meernik 2008; and McClendon 2009. 23 Goldsmith 2003. 24 For another example of an instrumentalist argument, see also Grodsky 2011.

10

One set of the broader literature on international legal compliance – the so-called

“enforcement school” – assumes that states usually find noncompliance advantageous, but

nevertheless comply when the costs of defying legal obligations outweigh the benefits.25 But

how do third-party states benefit from harboring indicted fugitives? What, for example, does a

state like Nigeria gain by refusing to turn over an indictee like Charles Taylor to the Special

Court for Sierra Leone (SCSL)?26 Due to narrow case selection, this question has been largely

ignored by the instrumental literature on ICTs. I argue that explaining ICT compliance requires

better explanations for the phenomenon of noncompliance. That is, why do states initially resist

their international legal obligations in the first place? Furthermore, instrumental approaches rely

on a single meta-cause (incentives), but the logic of incentives does not tell us why many third-

party states comply from the beginning despite a lack of explicit pressure, material or otherwise.

On the other hand, dispositional approaches to ICT compliance say more about states’

initial behavior, but nevertheless also have trouble explaining important variation. These studies

link state behavior to deep-seated normative commitments that condition governments’ attitudes

towards legal obligations prior to instrumental concerns. In his well-known historical survey of

ICTs, Gary Bass concludes that “liberal” states – that is, “states that respect civil and political

rights” – owing to their internalized commitments to these rights and the associated principles of

“legalism,” are more likely to support the creation and operation of ICTs than illiberal ones.27

Bass’s study is part of a larger tradition in international law scholarship that sees democracies as

more likely to comply with their international legal obligations than nondemocracies.28

25 Downs, Rocke, and Barsoom 1996. 26 For a discussion of the politics surrounding Charles Taylor’s indictment by the SCSL, exile from Liberia, and arrest in Nigeria, see Tejan-Cole 2009. 27 Bass 2000, 20. “Legalism” is “the ethical attitude that holds moral conduct to be a matter of rule following, and moral relationships to be consist of duties and rights determined by rules.” Shklar 1986, 1. 28 Slaughter 1995

11

Similarly, Judith Kelley finds that states characterized by a strong rule of law or an affinity for

the ICC were more likely to uphold their treaty commitments and refuse to sign bilateral

nonsurrender agreements with the United States, even in the face of explicit material

incentives.29 Neither of these authors’ arguments requires that leaders genuinely internalize pro-

norm behavior (though that is possible). But at minimum they posit that the high salience of

norms relating to rights and the rule of law in states’ domestic politics makes it particularly

difficult for these leaders to justify to domestic audiences any course of action other than

compliance.30

But while dispositional approaches provide a compelling explanation for why so many

liberal or rule of law states comply with ICTs from the start, these studies lack a theoretical

account for the inverse proposition: that is, why should an absence of these values necessarily

dispose governments against compliance? Once again, it is rather intuitive that transitional states

under obligation to turn over their own citizens would initially prefer noncompliance, but it is not

clear why illiberal or weak rule of law third-party states would necessarily be disposed to

tolerate indicted war criminals in their territories. Further, these studies’ reliance on dispositional

factors leaves them unable to explain important variance: that is, why do some states that lack

commitment to these values nevertheless still comply with ICTs, while some liberal democracies

resist compliance?

The so-called “management school” of international legal compliance does attempt to

connect assumptions regarding the innate pro-compliance dispositions of states to explicit

29 Kelley 2007. “Affinity for the ICC” is measured by whether a country was a member of the so-called “like-minded group,” a group of states that identified themselves during the Rome Statute negotiations as the staunchest supporters of the court. 30 Cortell and Davis 1996.

12

hypotheses for noncompliance.31 This literature identifies three main explanations for

noncompliance: ambiguous rules, the need for gradual implementation over time, and weak state

capacity. The first two explanations do not travel well to the question of states’ initial willingness

to comply with ICTs. The third factor, weak state capacity, may seem an intuitive explanation for

states’ failures to pursue indicted suspects. Nevertheless, it is my (as well as other authors’)

contention that the empirical record suggests capacity is usually not a dominant factor in ICT

noncompliance.32 This is not to say that capacity is irrelevant, but it is often the case that once

previously noncompliant states decide to pursue suspects, they arrest them rather quickly. Also, I

attempt to draw a distinction between unwillingness and inability to arrest suspects, and I

maintain that the operationalization of my dependent variable (see below) captures the former

without being biased by the latter.

In sum, both instrumental and dispositional hypotheses suffer similar shortcomings that

leave them inadequate for developing a generalizable model of ICT compliance. First, both focus

too much on monocausal explanations at the expense of leaving important variation unexplained.

If incentives produce compliance, why do so many states initially comply despite an absence of

international pressure? Similarly, if a commitment to liberal values explains compliance, why do

many illiberal and repressive states comply without being the target of international pressure,

while some liberal democracies nonetheless defy their obligations? It is tempting to attribute the

unexplained variation in each approach to the other, but without systematic testing – which has

yet to be done – we cannot say for sure. Jonas Tallberg shows that, in the case of the European

Union, state compliance with supranational rules is best explained by a combination of incentive-

31 Chayes and Chayes 1993. 32 See Bass 2000; Moghalu 2005; Peskin 2008; and Subotic 2009.

13

based and capacity-based approaches.33 Similarly, one purpose of my study is to assess the

potential complementarity of contrasting instrumental and dispositional hypotheses in explaining

ICT compliance.

Second, neither approach adequately explains the phenomenon of noncompliance. Both

approaches are characterized by an assumption of “causal symmetry,” whereby negative cases –

cases of noncompliance – can be explained by reference to (the absence of) the same factors as

those that explain positive ones. But as Goertz and Mahoney note, “Ultimately, whether a

relationship is symmetrical or asymmetrical is an empirical question.”34 Many social phenomena

exhibit asymmetric properties whereby “[t]he causal model for failure is often different from the

explanation for success,”35 and thus negative cases must be examined for their own sake.

Understanding why states violate their legal obligations might require different explanations than

those that address why they comply. Beyond merely an important object of examination on their

own, negative cases are also valuable for accounting for unexplained variance in positive cases,

and therefore improving existing models. As Charli Carpenter notes, “exploring negative social

outcomes can provide important clues as to the causal factors underlying outcomes that do

occur.”36 Sonia Cardenas argues that the literature on international legal compliance has been

dominated by a focus on explanations for positive cases. My study is consistent with her

recommendation that in order to grasp a full picture of how international law affects state

behavior, “we may need to bring violations back into the study of compliance.”37 Thus, to assess

the potential asymmetry of ICT compliance dynamics, I develop a series of negative case

hypotheses and test them against, and in conjunction with, positive case hypotheses.

33 Tallberg 2002. 34 Goertz and Mahoney forthcoming. 35 Ibid., 67. 36 Carpenter 2007, 646. 37 Cardenas 2007, 9.

14

Finally, both approaches have difficulty generalizing beyond transitional states. As

demonstrated by a number of recent high profile cases – such as those involving Sudanese

president Omar al-Bashir, former Liberian president Charles Taylor, and former Chadian

president Hissène Habré – third-party states often play a critical role in the arrest of indicted war

criminals and human rights violators. Failure to examine these cases prevents scholars from

developing generalizable models of ICT compliance. This is especially important in light of the

ICC’s increasing caseload and the court’s dependence on member states in arresting suspects. I

focus on the case of the ICTR, which, owing to the uncharacteristically high number of states

where indictees were at large and variation in state behavior, provides an excellent testing

ground for the generalizability of ICT compliance hypotheses.

The ICTR was established in 1994 to try the worst perpetrators of the Rwandan

genocide.38 Of the 91 individuals indicted between 1995 and 2005 for crimes committed during

the genocide, all 82 who have been arrested were apprehended in one of 25 African and

European states (plus one arrested in the U.S.).39 Unlike the indictees of the ICTY or the SCSL,

where the majority of suspects were arrested in target states, all these captured ICTR indictees

were arrested in third-party states. Most of these states demonstrated an initial willingness to

pursue suspects believed to be at large in their territories. But roughly a third of these states

initially resisted compliance with their obligations to some extent. Though almost all states that

resisted compliance eventually cooperated to some degree in arresting suspects, the question

remains: why did these states initially defy the tribunal in the first place? 38 For background on the politics of the courts founding and operation, see Akhavan 1996; Cruvellier 2006; Magnarella 2000; and Peskin 2008. 39 Two individuals were indicted for crimes not related to the actual genocide. Léonidas Nshogoza, a defense investigator, was indicted for contempt and surrendered to the court in Tanzania. See http://www.unictr.org/tabid/155/Default.aspx?ID=43. (Accessed 2 Jun 2012). A witness, know by pseudonym “GAA,” was indicted for providing false testimony to the court and arrested in Rwanda. See http://www.unictr.org/tabid/155/Default.aspx?ID=64. (Accessed 2 Jun 2012). I exclude these two indictees from my analysis.

15

The large sample size and variation in compliance behavior vis-à-vis the ICTR provide

an excellent opportunity to address the three shortcomings in the existing literature outlined

above. But the ICTR also provides a unique opportunity to study international legal compliance

more generally. Most studies of international legal compliance focus on states behavior in

relation to treaties, which are essentially contracts to which states voluntarily accede. Thus, these

studies are plagued by questions of endogeneity, as it is difficult to separate out states’ prior

willingness to comply from the subsequent effect of the treaties on their behavior.40 However, in

the case of the ICTR, all UN member states, by virtue of Security Council resolution, were

obliged to arrest suspects in their territories.41 Security Council resolutions are not agreements to

which states voluntarily commit (other than those states that, at the time, are members of the

Security Council and which vote in favor of the relevant resolution).42 All UN member states are

bound to Security Council resolutions whether they like them or not, and thus their legal

obligations are not endogenous to their preferences. Therefore, examining ICT compliance

provides a clearer portrait of the effects of international law on state behavior.

In the next section I draw on broader scholarship in International Relations (IR) and

comparative politics to develop a series of negative case hypotheses meant to supplement these

existing approaches by explaining noncompliance as a phenomenon unto itself. I then test these

hypotheses alongside positive case ones to assess their potential complementarity.

Negative case hypotheses: motivations for noncompliance

National security

40 Downs, Rocke, and Barsoom 1996. 41 UN Security Resolution 955. 42 Of the 27 states where indictees were believed to be located, only two – the U.S. and U.K. – were on the Security Council that approved resolution 955, which established the ICTR. Both states voted in favor.

16

In realist IR theory, national security is a state’s paramount concern.43 While under some

circumstances states may have a strategic interest in conforming to international law and norms,

ultimately, if their security is threatened, the logic of survival will motivate states to defy their

legal obligations. Benjamin Valentino, Paul Huth, and Sarah Croco argue that, regardless of

regime type or treaty commitments, states will target civilian noncombatants in war in violation

of international law if it serves their larger strategic objectives.44 Scholarship on state compliance

with human rights norms has shown analogous findings. Steven Poe and Neal Tate find a

positive relationship between participation in either domestic or international conflict and

government violations of personal integrity rights.45 Likewise, Sonia Cardenas, in her book-

length study of state compliance with human rights norms, finds that domestic and international

pressures to change human rights practices are less effective when a target government perceives

itself to have a national security interest in continuing transgressive practices.46

A government may find it advantageous to defy its international obligations and tolerate,

harbor, or even exploit ICT indictees in its territory if it perceives such individuals as beneficial

for its security objectives. These benefits may take different forms. For instance, the government

in question may share an adversary with an ethnic or political group to which indictees belong.

Therefore the government may recruit indictees (who, typically, though not always, are military

actors themselves) or other members of their relevant group to participate in a conflict. Even if

the government and group in question are not formally cooperating, their common cause against

the same adversary would make it advantageous for the government to not agitate or weaken its

ally by pursuing indicted members in its territory. This was the case with Zairian president

43 Mearsheimer 1994. 44 Valentino, Huth, and Croco 2006. 45 Poe and Tate 1994. See also, Poe, Tate, and Keith 1999. 46 Cardenas 2007.

17

Mobutu Sese Seko and the Rwandan Hutu rebels who took refuge in eastern Zaire. Mobutu and

the Hutus shared a common enemy in the post-genocide Tutsi-led Rwandan government, and so

Mobutu armed and supported the rebels, tolerating their use of his country as a staging ground

from which to launch incursions into Rwanda and refraining from arresting suspected Hutu

genocidaires located among them and elsewhere in the country.47 But national security

motivations do not have to be connected to ongoing violence. US efforts following World War II

to shield some Nazis deemed to possess high intelligence value demonstrates that governments

may also perceive knowledgeable noncombatants as a valuable national security resource.48

Thus, the national security hypothesis for ICT noncooperation is as follows:

HI: To the extent that the presence or protection of indictees offers benefits for its national

security objectives, a government will be less likely to comply with its legal obligations to ICTs.

Rent-seeking

Political economy approaches to foreign policy attribute a leader’s decision calculus to a “logic

of political survival,” whereby the goal of regime security is paramount.49 In democracies,

regime survival means winning elections. But in autocratic states, such as most of the African

states to which ICTR indictees fled, regime security is based on a different dynamic. Scholars of

African politics often characterize African autocratic regimes as based on “personal” or

“neopatrimonial” rule. Neopatrimonial rule entails autocratic leaders who rely not on the

approval of publics to maintain their regime, but on personalized patronage networks of

supporters and clients. These rulers are not constrained by institutions, which are weak, nor is

47 Prunier 2009, 28; Human Rights Watch 1995. 48 See Gimbel 1986; and Ryan 1983. 49 Bueno de Mesquita et al. 2003.

18

their authority derived from rational-legal legitimacy. Rather, a leader’s power is a function of

her ability to maintain and exploit these clientelistic relations as well as use force to subdue

opposition actors.50 Both elements – patronage and coercion – require vast economic resources.

Therefore, neopatrimonial regimes are often maintained by the extraction of rents from various

domestic sources, most typically natural resources, but also wealthy elites and other privileged

stakeholders, such as firms and political parties.

Governments characterized by such highly clientelistic, informal institutions will be more

likely view ICT indictees in their territories through the lens of potential extortion and rent-

extraction than more institutionalized, rational-legal ones. Specifically, these governments or

their clients may be willing to offer protection to indicted war criminals in exchange for rents.

ICTs typically do not target low-level actors for indictment, but focus mostly on higher-level

commanders and enablers. These types of actors are often considerably wealthy in their own

right, and thus are well suited to offer rents in exchange for protection. For example, one of the

few ICTR indictees to remain at large is Felicien Kabuga, a wealthy Rwandan businessman who

was a major financier of the Hutu genocidaires. Despite a $5 million reward offered by the FBI

for information leading to his arrest, Kabuga has reportedly been able to use his wealth and high-

level official connections to buy protection in Kenya.51 Under the reign of former president

Daniel Arap Moi, that country was a textbook case of kleptocratic, personalized rule.52 Thus, the

rent-seeking hypothesis for noncooperation is as follows:

50 Jackson and Rosberg 1982; Bratton and Van de Walle 1994; and Hyden 2006. 51 Moghalu 2005, 168-171; and “Africa’s Pol Pot,” The Sunday Times 1 April 2007. 52 Barkan 2004. According to Barkan, under Moi “[c]orruption became the principal mechanism for regime maintenance.” Barkan 2004, 89.

19

H2: Governments that are characterized by neopatrimonial rule will be more likely to see

indictees as a source of rents and thus be less likely to comply with their legal obligations to

ICTs.

Noncompliance constituencies

Liberal theories of International Relations view a state’s foreign policy preferences as being

shaped by the competing pressures of domestic society. As Andrew Moravcsik writes: “The state

is not an actor but a representative institution constantly subject to capture and recapture,

construction and reconstruction by coalitions of social actors.”53 One strand of the literature on

international legal compliance has focused on this link between domestic interests and state

preferences. For example, the existence of influential “compliance constituencies” can raise the

political costs for governments to defy their obligations, especially in democracies.54 Xinyuan

Dai develops a model of international legal compliance that makes central the role of competing

domestic interest groups. To the extent that pro-compliance constituencies posses an

“informational endowment” and “electoral leverage,” (democratic) governments that value

reelection will be responsive to these groups’ preferences.55 If compliance constituencies can

successfully pressure a government to obey international agreements, then the inverse is also

likely. That is, “noncompliance constituencies,” if sufficiently influential, can motivate

governments to defy their international obligations. Sonia Cardenas argues that “pro-violation

constituencies” that have an interest in violating human rights norms can, if sufficiently

influential, compel their governments to resist compliance.56 Cardenas hypothesizes that these

53 Moravcsik 1997, 518. 54 Kahler 2000. 55 Dai 2005. 56 Cardenas 2007.

20

types of noncompliance constituencies will be more influential in nondemocracies – specifically,

she has in mind either economic elites or “members of the coercive apparatus” who perceive

benefits in continuing to violate human rights norms.57 However, in a study of compliance with

the global “counter-terrorism regime,” Beth Elise Whitaker argues that different types of

noncompliance constituencies – this is, civil society and other domestic interest groups – can be

effective in influencing their governments in democracies as well.58

A government will likely resist pursuing ICT indictees in its territory when organized

domestic interests that oppose compliance are sufficiently influential. Regarding ICT indictees,

such noncompliance constituencies would likely come in the form of identity-based interest

groups, such as those formed around religious, ethnic, or political identities, which often have

strong transnational ties. If indictees share an identity with influential domestic groups, than

these groups may view the indictees through the lens of in-group solidarity. In turn, these groups

may mobilize to pressure their own government to refrain from pursuing and/or transferring

indictees. If the interest group is sufficiently influential, the mere potential for their response,

especially in democracies, might initially dispose a government against compliance, despite prior

deep-seated affinities for international justice. In Italy, the Catholic Church actively sheltered a

Rwandan Catholic priest, Athanase Seromba, indicted by the ICTR. The Italian government was

reportedly under pressure from the Vatican to not pursue Seromba, and initially, the government

obliged the powerful institution, defying its international obligations and leading to a public

standoff with the tribunal prosecutor.59 This was in spite of Italy’s normally strong support for

57 Ibid., 27-28. 58 Whitaker 2010. 59 Del Ponte 2009, 189-190; “Italy Balks on War Crimes Extradition Despite Calls from Prosecutor, Human Rights Groups.” Associated Press 17 July 2001.

21

international justice institutions.60 While the Italian government eventually changed its position,

it nonetheless initially sided with the Church and defied the authority of the court. Thus, the

noncompliance constituency hypothesis is as follows:

H3: To the extent that an interest group that identifies with indictees wields significant influence

on government policy, a government will be less likely to comply with its legal obligations to

ICTs.

Data and Methods

Qualitative comparative analysis

To test the generalizability and complementarity of both positive and negative case hypotheses, I

employ crisp set qualitative comparative analysis (csQCA). Qualitative comparative analysis

(QCA) is a method developed by Charles Ragin that uses set-theoretic logic and Boolean

operations to infer necessary and sufficient conditions associated with an outcome under study.61

The crisp set version classifies cases (in my case, states) dichotomously in terms of their

membership or nonmembership in various “sets,” which correspond to each hypothesized

“condition” (or what statistical methods call “variables”). The analysis in csQCA consists of

identifying patterns of set memberships (that is, the presence or absence of conditions) that

correspond to the presence or absence of an outcome, thus producing a “solution” in the

language of necessary and sufficient conditions.62

60 Italy has been a major financial contributor to both the ICC and the ICTR. See Ford 2011. 61 Ragin 1987. 62 Rihoux and De Meur 2009. Because most explanations for social phenomena are actually probabilistic, scholars using QCA often qualify the deterministic nature of set-theoretic language by adopting terms such as “probabilistically” necessary or “usually” sufficient. See, for example, Mahoney 2003.

22

QCA is especially well-suited to address my research question for at least three reasons.

First, QCA can perform systematic cross-case comparisons of small- and medium-N samples.63

Indeed, the small size of my sample (n=26) would be problematic for conventional statistical

analyses. Second, QCA is designed to handle both equifinality and conjunctural causation.

Equifinality is when multiple casual paths may produce the same outcome, and conjunctural

causation means that those paths are made up of configurations of more than one condition.

Conventional regression methods, on the other hand, assume “causal additvity,” whereby each

independent variable contributes the same net effect towards the outcome across every case. In

contrast, QCA takes seriously the possibility that different configurations of variables may have

different effects on an outcome; some conditions may lead to either the occurrence or

nonoccurrence of a phenomenon depending on the presence or absence of other conditions.

Finally, unlike conventional regression methods, QCA does not assume “causal symmetry,”

whereby the same model that explains the presence of an outcome also explains its absence.64

Instead, QCA can infer different explanations for positive and negative cases, which, I have

argued, is particularly important for understanding ICT compliance.

The procedure for conducting csQCA is as follows.65 After identifying the population of

cases (see below), conditions are selected, operationalized, and dichotomously coded. Then cases

are arranged in a truth table, which contains one row for each logically possible combination of

conditions found in the sample. Finally, the analysis proceeds by way of Boolean minimization,

which produces a “solution” to the specified model in the language of necessary and sufficient

conditions.

63 Rihoux and Ragin 2009. 64 Berg-Schlosser et al. 2009, 9; Grofman and Schneider 2009, 666-667. 65 This procedure is detailed in Rihoux and De Meur 2009.

23

Case selection

In general, the case selection process in QCA is an iterative one, and requires researchers to use

their in-depth case knowledge to make judgments regarding what counts as a case of the

phenomenon under study.66 A “case” in my model is a state that had the potential to arrest ICTR

indictees in its territory.67 My case selection aims to account for the entire population of states in

which the 91 ICTR indictees were or are currently at large. But because of the limits of existing

data, identifying this population is difficult. Based on official ICTR reports, I can identify which

states actually arrested one or more of the 82 captured indictees (a total of 25 states).68 Many of

these states willingly complied with their legal obligations form the start, while some others

initially defied the court, only arresting indictees after succumbing to pressure. But relying on

where indictees were arrested is an imperfect measure of which states had the potential to

comply, as suspects may have traveled to various states before being arrested in the final one.

Ideally, I would include every state where an indictee was located, regardless of whether he or

she was arrested there, but limited data means that I am largely unable to identify these

intermediate states as well as the states in which remaining indictees are still at large, thus

excluding these cases from the analysis. Indictees typically do not announce where they are

hiding, and to the extent that media reports have shed light on their locations, these reports have

mostly focused on the most high profile cases, such as Kenya and the Democratic Republic of

Congo (DRC), which attracted most of the court’s attention. This limitation means I will

potentially undercount the number of “noncompliers” in the population, and thus my sample may

be less representative than I would like. Nevertheless, based on official ICTR reports and limited

66 Berg-Schlosser and De Meur 2009. 67 I uses states instead of indictee-states as my unit of analysis because the data is not fine-grained enough to make distinctions between a state’s behavior towards different indictees within its territory. Below, I discuss in more detail the limitations of my data. 68 Data are taken from the ICTR website: http://www.unictr.org/Cases/StatusofDetainees/tabid/202/Default.aspx.

24

media accounts, I can still identify a couple “cases” that do not fit the coding rule of actually

having arrested indictees.

Specifically, I include in my sample two states where no arrests occurred: Italy and

Zimbabwe. Even though no suspects were arrested in Italy, one indictee, Athanase Seromba, did

take refuge there. The Italian government publically resisted arresting him (thus Italy is coded as

initially noncompliant), but eventually relented under public pressure from the tribunal to

transfer him to the court. However, despite the Italian government’s change in position, Seromba

apparently flew to Tanzania voluntarily to surrender himself.69 Thus, even though Seromba

surrendered in Tanzania, I still include Italy as a case of (non)compliance. No arrests have

occurred in Zimbabwe either, but it is widely believed that one of the few indictees still at large,

Protais Mpiranya, has been living in that country while the government continues to neglect

pursuing him.70 Thus, I also include Zimbabwe as a case of noncompliance. Finally, I remove

one case from my sample, Angola, because of conflicting reports regarding whether the single

indictee reportedly arrested there was actually residing and/or apprehended in Angolan or DRC

territory.71 The number of cases in my sample thus totals 26 countries.

Operationalization of variables

My dependent variable is the initial behavior of a government regarding its legal obligation to

pursue indicted ICTR suspects in its territory, coded as either compliance or noncompliance. By

“initial” behavior, I mean the behavior of a government at the time that its legal obligation was

first activated, that is, when the first indictment for a suspect residing in the state was issued. But

the initial behavior of these governments is difficult to observe. Indictments were often sealed,

69 “Rwandan Priest Turns Himself In,” The Guardian 8 Feb 2002. 70 Jallow 2010, 2011b; “Talks Between ICTR and Zimbabwe on Fugitives,” Hirondelle News Agency 1 July 2011. 71 See Prunier 2009, 444n97.

25

and efforts to pursue indictees often proceeded inconspicuously. Access to internal documents is

also not available for such recent events. Therefore, I rely on another source: the subjective

assessments of public officials, NGOs, scholars and journalists regarding which governments

were or were not actively pursuing suspects believed to be at large in their territories. Subjective

assessments are an imperfect indicator, but the sources from which I draw consist of either the

views of insiders or those with expert investigative knowledge. (A full list of the sources on

which I drew is contained in Appendix A).72 One strength of QCA is that it presumes in-depth

knowledge of cases. Therefore, QCA permits researchers to make quantitative classifications of

qualitative indicators as long as the coding process is transparent, justifiable, and replicable.73

My coding procedure was as follows: First I assembled as many data sources as I could

find that referred specifically to the behavior of third-party states. My data consisted of reports

and public statements by ICTR officials, NGOs, scholars and journalists that make explicit

claims regarding which states were initially complying and which were not. I then used these

sources to code states’ initial behavior.74 I analyzed these sources inductively, looking for

descriptions and assessments of states’ behavior regarding their willingness to arrest and transfer

indictees. For example, assessments that I regarded as indicating noncompliance made references

to states “harboring,” or “providing protection” or “refuge,” to indictees. Similarly, I regarded

states as noncompliant when they were described as, for example, “hampering” or “obstructing” 72 Since my study focuses on the international legal obligations of states vis-à-vis ICTs, it is important when analyzing these reports to distinguish between compliance in pursuing indicted suspects and willingness to pursue other individuals suspected of crimes but not specifically under indictment. France and Belgium were often accused by NGOs of not pursuing suspected Rwandan war criminals residing in the territories, but these NGOs were critical of these governments unwillingness to prosecute these individuals themselves domestically under the principle of universal jurisdiction. The behavior of these governments shows that they dutifully pursued any suspects who were actually under indictment, even if they tolerated the presence of other suspected war criminals in their countries. 73 Berg-Schlosser et al. 2009. 74 The list of sources in Appendix A may seem small, but very few sources exist that directly speak to the behavior of third-party states in complying with the ICTR. Nevertheless, Appendix A represents my best effort to locate data on these cases. However, some high profile cases, such as Kenya and DRC, received more attention than others. As such, I omitted sources from the list that were effectively redundant for my purposes, that is, sources – such as additional media accounts – that only confirmed data that I had already collected from numerous other sources.

26

efforts to obtain custody of indictees, or requiring “pressure” in order to act on arrest or transfer

requests. These examples are not exhaustive of indicators for noncompliance, but they should

demonstrate the types of references that I regarded as relevant. I coded a state as initially

noncompliant if my sources used this or similar language to single out the government as

willingly resisting its obligation to pursue or transfer suspects. My dependent variable is thus

coded dichotomously, as either denoting an initial behavior of compliance or noncompliance.75

One observable implication that follows from the logic of my classification is that

countries that initially resisted compliance should take longer to arrest suspects in their

territories. As a check on the validity of my coding, I collected data on each ICTR suspect

regarding the duration of time between when an indictment for him was issued by the court and

when the suspect was arrested.76 If I find a statistically significant difference in the durations of

arrest warrants across these two groups (“compliers” and “noncompliers”), then I can proceed

with greater confidence in the coding of my dependent variable.77 Figure 1 displays the Kaplan-

Meier survival curves for the two groups. A Wilcoxon rank test of the survival functions of these

two groups is significant with a p-value of 0.03, indicating my coding decisions are robust.

[Figure 1]

The operationalization of my first independent variable, national security interest, is

meant to get at whether a government would view indictees in its territory as potentially

75 The particular assignment of a “1” or a “0” for these different outcomes changes depending on whether compliance or noncompliance is the explanandum of the model being tested. 76 Data were assembled from various sources, including the ICTR website, the ICTR Basic Documents and Case Law database (http://www.ictrcaselaw.org/home.aspx), and a dataset constructed by James Meernik (Meernik 2008). 77 These data only pertain to suspects who were indicted prior to arrest, which amount to just over half of the eventual arrested indictees (45 out of 83). The rest of the indictees were already in custody, either with a state or the tribunal, at the time their indictments were issued.

27

beneficial to its national security objectives. Because all the ICTR indictees were aligned with

the former Hutu government in Rwanda, the governments that would stand to gain the most

national security benefit from harboring indictees would be those that were adversaries of the

subsequent post-genocide Tutsi government. Therefore, my coding of national security interest

refers to whether a state has been involved in a military conflict against the post-genocide

Rwandan government. Data come from the Correlates of War Dyadic Militarized Interstate

Dispute v3.10 dataset. I code each case a “1” if the state has participated in a conflict against the

post-genocide Rwandan that has reached the level of “war,” the highest category on the dataset’s

five-point “hostility level” scale.78

The operationalization of my second independent variable, rent-seeking, is meant to get

at the potential for a government to view indictees in its territory as a source of rents. Ideal data

for this variable would consist of each indictee’s relative wealth as well as some measure of his

embeddedness in clientelistic networks within the state to which he fled. However, because of

the unavailability of this type of fine-grained data, I use a measure of corruption from the World

Bank’s Worldwide Governance Indicators project as a proxy for the potential for rent-seeking by

a government.79 Clientelism and bribery are integral to corrupt governance, and so the former

phenomena should be collinear with the latter. Therefore, states characterized by higher

corruption should be more likely to extract rents from indictees for protection, and thus less

likely to comply. The World Bank measure uses expert surveys of institutes, think-tanks, NGOs,

and international organizations to construct a 5 point scale that “[r]eflects perceptions of the

extent to which public power is exercised for private gain, including both petty and grand forms

of corruption, as well as ‘capture’ of the state by elites and private interests.” Scores range from

78 Ghosn and Bennett 2003. I code the presence of such a conflict as positive only if the particular conflict occurred previous to the year of the first indictment for a suspect eventually arrested in that state’s territory. 79 Kaufmann, Kraay, and Mastruzzi 2010, 4.

28

-2.5 (weak control of corruption) to 2.5 (strong control). These scores are scaled to ensure a

population mean of 0 and standard deviation of 1,80 making the zero-point a natural break around

which to dichotomize cases into either “high” or “low” categories. I code scores above 0 as “0,”

indicating low potential for rent-seeking, and below 0 as “1,” indicating a high potential. Data for

each state is taken from the year of the first indictment for a suspect eventually arrested in its

territory.81

The operationalization for my third independent variable, noncompliance constituency, is

meant to get at the potential for interest groups with transnational linkages to possess enough

power resources to influence a government’s compliance behavior. For the case of the ICTR, two

transnational group traits are relevant. The first is ethnicity. All 91 Rwandans indicted by the

ICTR were Hutus. Therefore, we would expect to find noncompliance constituencies in states

that contain large Hutu populations. Traditionally, significant permanent Hutu populations only

exist in Rwanda, Burundi, and the DRC. Of these three states, only the DRC is in my sample.

Nevertheless, the Hutu population is the DRC is traditionally small, less than 1% according to

the Minorities at Risk dataset,82 and thus I do not code DRC as a case of a potentially influential

noncompliance constituency.

The second relevant trait is religion. Seven ICTR indictees were Christian clergymen of

various confessions. Therefore, I examine the seven states in which these indictees were arrested

(DRC, France, Italy, Kenya, Switzerland, Uganda, and the U.S.). If an indictee’s associated

confession exercises significant influence over policymaking in the state where he had taken

refuge, then that government should be less likely to comply with the ICTR. Operationalizing

80 Ibid., 9. 81 From 1996 to 2002, the dataset only contains data for even-numbered years. For cases from odd-numbered years within this range, I use the score from one year prior. Measurement error arising from this limitation should be minimal, as variation from year to year tends to be very small. 82 Minorities at Risk Project 2009.

29

this variable requires me to treat democracies and nondemocracies differently. Noncompliance

constituencies in democracies exert their influence through democratic institutions. Thus, for the

four democracies (France, Italy, Switzerland, and the U.S.), I assign a “1” if the ruling coalition

at the time of the first indictment for a suspect arrested in its territory included a Christian party

that explicitly identifies itself with the indictee’s particular denomination. Only Italy fits this

criterion.

In nondemocracies, noncompliance constituencies are only influential to the extent that

they constitute or have access to ruling elites.83 In order to assess their potential influence, I rely

on the scholarship of area experts to examine the degree to which the relevant Christian

denominations are connected to centers of political power in these states. The single indicted

clergy member arrested in Uganda, Jean-Bosco Uwinkindi, was a Pentecostal pastor. Uganda has

a long history of a robust Christian civil society, but to the extent that these groups carry political

influence, the Anglican church has been far more connected to centers of power than either the

Catholic or Pentecostal churches.84 Though the Pentecostal church is growing in numbers and

visibility, it is still far smaller and carries less influence than either the Catholic or Anglican

churches,85 thus making it a weak case for a potential noncompliance constituency. The single

indicted clergy member arrested in Cameroon, Hormisdas Nsengimana, was a Catholic priest.

According to Paul Gifford, a scholar of religion in Africa, “[T]he Catholic Church is easily the

biggest organization in the country apart from the state.”86 About 30% of population define

themselves as Catholics in a country where Christians are the largest religious group, constituting

83 Cardenas 2007, 27-28. 84 Gifford 1998, Ch. 4. 85 “Slain by the Spirit: The Rise of Christian Fundamentalism in the Horn of Africa,” The Economist 1 July 2010. 86 Gifford 1998, 270.

30

just over half the population.87 Nevertheless, the autocratic Biya regime has traditionally viewed

the church as more of a threat than an ally or source of political power. Thus, the Cameroonian

government and ruling elite have traditionally been hostile to the Catholic church, seeking to

keep it weak vis-à-vis the state.88 Finally, Samuel Musabyimana was an Anglican bishop

indicted by the ICTR and arrested in Kenya. Like Cameroon, the Kenyan government under the

one-party rule of Daniel Arap Moi was traditionally hostile to the Church. While the Anglican

church had been an important civil society institution and a central force in Kenya’s slow

transition to democracy, its position vis-à-vis Moi and his KANU party had been distinctly one

of opposition, not partnership.89 Because the relevant Christian groups in Uganda, Cameroon,

and Kenya all appear to have been too weak to influence their government’s foreign policy

decision-making, I code all three as “0” to indicate the weak power of noncompliance

constituencies.

I also include a series of independent variables, drawn from the existing literatures on

ICT and international legal compliance and meant to test the utility of extant “positive case”

hypotheses. Instrumental approaches attribute ICT compliance to incentives. My focus here is on

initial behavior, not change in behavior over time, and thus it is difficult to discern whether

threats or inducements were made prior to or at the outset of an indictment such that a

government’s initial behavior would be influenced. Therefore I reason that if the instrumental

thesis holds, then states that a priori had the most to lose from resisting compliance would be the

least likely to do so. The most common “stick” used by external actors to compel ICT

compliance is economic coercion, usually in the form of the withholding of aid. States that are

dependent on foreign aid would thus have the most to fear from defying the will of the

87 Ibid., 251-253. 88 Ibid., Ch. 6. 89 Sabar-Friedman 1995; Sabar-Friedman 1997.

31

international community. Therefore to operationalize this variable, I focus on foreign aid

dependency, measured as the combined net official development assistance and official aid as a

percent of GDP.90 But not just any amount of foreign aid is enough to classify a state as

“dependent.” Therefore I follow Deborah Bräutigam, who uses a threshold of 10 percent of GNP

as a proxy for the point at which “a country cannot perform many of the core functions of

government, such as operations and maintenance, or the delivery of basic public services,

without foreign aid funding and expertise.”91 Since net aid can fluctuate from year to year, I use

a more robust two-year average taken from the year of the first indictment for a suspect

eventually arrested in that state’s territory, as well as one year before. States that maintain a two-

year average of 10 percent or higher are deemed “dependent” and are coded as a “1,” while states

below 10 percent are coded as “0.”

I also include two variables meant to test dispositional explanations. Existing literature in

this tradition has identified two explanatory variables that have been shown to predict various

instances of norm compliance. The first dispositional variable, regime type, is meant to capture

the central explanatory factor in Bass’s well-known study. As mentioned above, Bass attributes

pro-ICT behavior to liberal states’ affinity for legalism.92 I use regime type (democracy versus

non-democracy) as a proxy for adherence to legalism. I operationalize this variable using the

Polity 2 scores from the Polity IV dataset.93 Following the co-authors’ recommended

categorization, I code a state that scores 6 or higher as a democracy (“1”) and states below as

nondemocracies (“0”). Data are taken from the year of the first indictment for a suspect

eventually arrested in the state’s territory. The second dispositional variable comes directly from

90 Data are taken from the World Bank. World Bank 2011. 91 Bräutigam 2000, 2. Arthur Goldsmith also uses this measure. Goldsmith 2001. 92 Bass 2000. 93 Marshall and Jaggers 2010.

32

Kelley’s study of states’ willingness to sign bilateral nonsurrender agreements with the U.S.

Kelley finds that states characterized by a stronger rule of law were less likely to defect from

their ICC treaty obligations.94 Following Kelley, I use a measure of rule of law provided by the

World Bank Group.95 This measure is a five-point scale ranging from -2.5 to 2.5. As with the

corruption dataset discussed above, scores are scaled to ensure a population mean of 0,96 making

this a natural break around which to dichotomize scores. I classify states above 0 as characterized

by “strong” rule of law (“1”) and below 0 as “weak” (“0”). The cases’ raw scores and their

corresponding crisp set memberships are summarized in Tables 1 and 2, respectively.

[Tables 1 and 2]

Analysis and discussion

My QCA analysis proceeds in two steps. First, I run a csQCA analysis using only the positive

case conditions drawn from the extant literature.97 (The truth table for Model 1 is contained in

Table 3). Second, I combine these conditions with my hypothesized negative case conditions to

construct a more comprehensive analysis. Model 1, which contains only positive case conditions,

yields the following solution:

REGIME + DEPENDENCY → COMPLIANCE

94 Kelley 2007. 95 Kaufmann, Kraay, and Mastruzzi 2010. 96 Ibid., 9. 97 csQCA analyses were performed using the fs/QCA 2.0 software. Ragin, Drass, and Davey 2006.

33

In QCA language, the symbol + indicates the logical “or,” and the capitalization of a condition

indicates its presence. In plain language, this means that either a democratic regime or aid

dependency is sufficient for compliance with the ICTR. However, one qualification is necessary.

The REGIME condition is not strictly sufficient. While a high number of democratic states did

initially comply, the solution is only 93 percent consistent for these cases. That is, democracy is

only likely sufficient for compliance. Italy is the once case of a democracy that exhibited an

initial resisted compliance. Nevertheless, the DEPENDENCY solution is 100 percent consistent,

suggesting that aid dependency is strictly sufficient for compliance. These two solutions cover

100% of the positive cases. In other words, every state that initially complied with the ICTR was

either a democracy or highly dependent on foreign aid. In contrast to Bass’s theory, Model 1

suggests that liberalism is not a necessary condition for compliance with ICTs. In contrast to

Kelley, rule of law is not a good predictor of ICTR compliance. And in contrast to Goldsmith,

economic leverage is not a necessary condition for compliance. Instead, there appear to be two

paths to compliance with the ICTR, one dispositional and one instrumental. Democracies, owing

to their commitment to liberal values like human rights and legalism, are very likely to comply.

Nondemocracies that are highly dependent on foreign aid, given their reluctance to jeopardize

valuable assistance, are also likely to comply. These findings suggest that the generalizability of

monocausal explanations is limited; appealing to a single explanatory factor leaves much

variation unexplained.

[Table 3]

34

But this is not the end of the story. While in mathematical terms these findings seem

adequate, in theoretical terms they are lacking. Neither the instrumental nor dispositional

hypotheses offer a theoretical account for noncompliance. Allowing indictees to roam free is an

embarrassment for most countries and violators face the possibility of sanctions or other threats.

It is unclear from the positive case explanations what benefits states might gain from tolerating

or encouraging such a situation. Therefore, in order to identify a more theoretically satisfactory

explanation, Model 2 incorporates the negative case hypotheses developed above.98 (The truth

table for Model 2 is presented in Table 4). Model 2 thus yields the following result:

dependency * [(REGIME * CONSTITUENCY) + (RENT * law * regime)] → NONCOMPLIANCE

where the symbol * indicates the logical “and,” and a condition in lowercase indicates its

absence. In plain language, the above solution means that a lack of aid dependency combined

with either the joint presence of democracy and a noncompliance constituency or the joint

absence of rule of law and democracy along with high potential for rent-seeking are sufficient for

noncompliance. In other words, either noncompliance constituencies within a democracy or high

potential for rent-seeking within an autocracy are compelling motivations for noncompliance.

Both solutions have a consistency of 100% for their respective subsets of cases, indicating strict

sufficiency, and together these solutions cover 100% of the negative cases.

These solutions suggest that the benefits of noncompliance may be either interest group

support (in democracies) or rents (in nondemocracies with weak rule of law). The results also

suggest that the absence of aid dependency is a necessary condition for noncompliance. In other

words, it appears that governments that are tempted by the benefits of noncompliance will 98 The outcome in Model 2 is recoded to produce a solution for noncompliance.

35

calculate that once they reach a critical level of aid dependence (in this model, 10 percent of

GDP), it is no longer worth jeopardizing that support by defying their obligations. This finding is

consistent with a body of research that establishes a link between a state’s aid dependence and

the convergence of its foreign policies with that of donor countries.99 In sum, there appear to be

at least two distinct paths to noncompliance – one that exploits formal, democratic institutions

and one that exploits informal, clientelistic ones.

[Table 4]

One limitation of my model is that the national security condition is washed out by the

rent-seeking condition. That is, every case of a nondemocratic state that scores positive on

national security interest also scores high on corruption (DRC and Zimbabwe). Therefore the

csQCA analysis does not adequately account for the influence of national security motivations.

Since the coding rule I use for national security interest only accounts for conflicts with Rwanda,

it may also fail to account for instances where national security interests are in play, but don’t

specifically implicate a conflict with Rwanda. For example, there are reports that Hutu militias

fought in support of embattled Congo-Brazzaville president Denis Sassou Nguesso in that

country’s civil war,100 which might have contributed to that government’s initial noncompliance.

Nevertheless, an event history analysis model presented in the next section does offer some

support for the national security hypothesis.

99 See, for example, Moon 1983. 100 Prunier 2009, 201; and “Curbing their Enthusiasm,” Africa Confidential 12 June 2009.

36

Observable implications: event history analysis

Though my csQCA model appears robust, the relatively small size of my sample and the

difficulty of operationalizing my key concepts are reasons to remain cautious about the validity

of my findings. King, Keohane, and Verba suggest that in order to strengthen findings derived

from a small sample, one should seek to increase the number of observations. One way to do this

is to identify the observable implications of a theory, that is, other effects one would expect to

see should the hypothesized relationships indeed be valid. The key is to identify an alternative

dependent variable that would be influenced by the same explanatory variables as the original

dependent variable and for which the researcher can obtain a greater number of observations.101

A confirmatory test of the relationship between the original explanatory variables and this

second dependent variable should increase confidence in one’s original inferences.102 If my

hypotheses are valid, then one place I would expect to see auxiliary effects is on the time it takes

indictees to be arrested. As mentioned above, even most initially noncompliant states eventually

arrested at least some indictees in their territories. The difference between these states and

initially compliant states was that in the case of the former, various types of pressure were

required in order to compel their compliance. This dynamic should be reflected in the time it

takes these states to arrest indictees in their territories as compared to initially compliant states.

Therefore I construct an event history model with indictees as the unit of analysis. The

dependent variable is measured as the number of days from when an indictment was issued for a

suspect to when that suspect was arrested.103 Specifically, I employ a Cox proportional hazards

model, because this model makes no assumptions regarding the direction or shape of the

101 King, Keohane, and Verba 1994, 223. 102 This technique is analogous to the notion of “auxiliary outcome causal process observations” developed in the literature on process tracing methodology. See Mahoney 2010, 129–131. 103 See supra note 71 for data sources. The dependent variable ranges from 0 to 3463 days, with a mean of 756.8 and a median of 203.

37

underlying baseline hazard rate. For the purposes of comparability, I limit my sample to those

individuals who were indicted while at large and then subsequently arrested. Of the 82 total

captured indictees, 46 fit these criteria. The remaining captured indictees were already in

custody, either with the arresting state or the court, at the time their indictments were issued. For

the independent variables, I use the same country-level measures from which I derived the crisp

set scores above, except that, when applicable, I retain their continuous form instead of

dichotomizing them. The country-level covariates correspond to the territory where each indictee

was arrested and are taken from the year of each suspect’s indictment.104 Because of limited

variation on the noncompliance constituency variable (only a single case scores a “1” on this

variable) I will not be able to test it in the event history model. One of the most interesting

findings from my csQCA model is that a combination of minimal aid dependency and high

potential for rent-seeking makes a state more likely to be noncompliant. To model this

conjunctural relationship, I include an interaction term for the rent-seeking and aid dependency

variables. Finally, because of high correlation between regime type and rule of law (r = 0.83), I

first include these variables in separate models, and then combine them into a single one.

It is also necessary to check whether the event history model’s proportionality

assumption holds. That is, the effect of each covariate should be constant over time. If not, the

model’s proportionality assumption is violated, and additional measures may be needed to adjust

for this dynamic effect.105 Following the suggestions of Box-Steffensmeier and Jones, I

calculated Harrell’s rho statistics based on Schoenfeld residuals for each model’s individual

104 My assumption is that each indictee was only located in one state throughout the duration of his arrest warrant. Empirical accounts suggest this some indictees actually traveled to multiple states before being arrested. Nevertheless, because of limited availability of data, I must make this assumption. 105 Box-Steffensmeier and Jones 2004, 131–137.

38

covariates.106 I identified any covariate that returned a p-value of 0.10 or lower on this test as

exerting a nonproportional effect. Then for each such variable I included in the relevant model an

interaction between the variable and the log of time, as recommended by Box-Steffensmeier and

Jones.107

[Table 5]

The results of the event history analysis offer some additional support for the findings of

my negative case csQCA model (see Table 5). Three of the five models (Models 2, 3, and 5)

indicate a significant relationship between the corruption-aid dependency interaction and arrest

warrant duration. In other words, states characterized by a combination of low aid dependency

and a high potential for rent-seeking tend to take longer to arrest indicted suspects in their

territories. Nevertheless, this interaction is only significant in models that contain a measure of

rule of law (Models 2, 3, and 5), but not those models that include regime type in lieu of rule of

law (Models 1 and 4). Also, recall that in the csQCA model, the national security factor was

washed out by other negative case factors. In contrast, the event history model allows me to test

the effect of the national security variable independently of these other factors. To this end, the

event history model offers some evidence that states that have fought a war against the post-

genocide Tutsi regime take longer to arrest indicted Hutus in their territories, suggesting that

these states might derive some geostrategic benefit from providing refuge to these indictees.

Nevertheless, similar to the corruption-aid dependency interaction, the national security factor is

significant only in the model that controls for rule of law (Model 5), but not regime type (Model

106 Calculations were performed using Stata 12.0. 107 Box-Steffensmeier and Jones 2004, 136.

39

4). While these analyses do not constitute definitive proof that the csQCA model is valid, they do

nonetheless lend greater confidence to its findings.

Conclusion

I have sought to test existing theories of ICT and international legal compliance using the

patterns of state compliance with the ICTR. My findings suggest that a generalizable model of

ICT compliance is more complicated than extant research has implied. No single explanation or

theoretical perspective accounts for why states complied with the ICTR. Instead, instrumental

and dispositional approaches appear to explain different types of cases. The former applies well

to states that comply despite lacking the deep-seated normative values of liberal states, and the

latter applies to states that comply despite having little tangible incentive to do so. In other

words, democracies comply with their international obligations because they have already

internalized – or at least, by virtue of the domestic salience of liberal norms, are restrained by –

values of international justice in their foreign policy decision-making. Nondemocracies comply

because they fear punishment in the form of withholding of vital foreign aid. Nondemocracies

that are not dependent on foreign aid are the least likely to comply.

Nevertheless, these positive case models do not necessarily tell us why some states defy

their obligations. To answer that equally important question, I have developed new hypotheses

that speak directly to the causes of noncompliance. To that end, the csQCA analysis suggests at

least two different factors may motivate states to resist compliance: the potential to extract rents

and the influence of a noncompliance constituency. Event history analysis offers support for a

third motivation, national security. There are benefits to be had from defying ICTs, though

benefits are not large enough to justify sacrificing critical foreign aid. States that are heavily

40

dependent on foreign aid, regardless of other factors, will likely comply with ICTs despite these

other countervailing motivations. Further research should employ qualitative case studies to

asses whether these types of motivations are indeed at work in cases of third-party

noncompliance with the ICTR or other ICTs. The nature of this phenomenon, however, is such

that states’ motivations cannot easily be observed and limited availability of data makes process-

tracing difficult. One solution could be to conduct interviews with ICT officials who have

worked on the front lines of efforts to secure compliance from states. Elite interviews have

indeed proved to be rich source of data for a number of excellent qualitative case studies of

compliance with ICTs by transitional states.108

My findings also speak to the growing body of scholarship on the determinants of

international legal compliance more broadly. For one, they suggest that scholars should be more

attuned to the potential for causal asymmetry. It is not enough to explain why states comply with

their legal obligations; scholars must also investigate the sources of noncompliance and not

merely assume that it can be explained by the absence of pro-compliance factors. Further, the

ICC’s increasing caseload has drawn particular attention to the importance of third-party ICT

compliance. Since his indictment in March 2009, Sudanese president Omar al-Bashir has

traveled to a number of ICC member states without arrest despite these states’ treaty obligations

to apprehend him. Extant research on ICT compliance has focused mostly on how noncompliant

states are compelled to change their behavior. My study suggests that the appropriate strategies

for pressuring these states should take into account the motivations of particular governments in

resisting compliance in the first place. As the ICC and its supporters attempt to exert pressure on

these states, they should understand that the willingness of governments to defy their ICC

commitments does not necessarily result from a lack of support for the norms of international 108 See, for example, Peskin 2008; and Subotic 2009.

41

justice, but may be the result of a variety of different factors. Thus, attention should be paid

countervailing benefits states may be receiving from harboring indicted suspects, and how to

neutralize these particular sources of resistance. Otherwise, efforts to secure ICC compliance will

be misguided and ineffective. Put simply, the cure must fit the disease.

Appendix A: Sources on initial preferences

Official statements • Jallow 2008, 2009, 2010, 2011a, 2011b • United Nations Security Council resolution 1503

NGOs

• International Crisis Group 2001 Scholarly works

• Magnarella 2000 • Peskin 2008 • Prunier 1997

Insider accounts

• Del Ponte 2009 • Moghalu 2005

Journalism

• “Cameroon Snubs Rwanda Genocide Tribunal Official” Reuters 2 Oct 1996 • Cruvellier 2006 • “Judge Urges World to Heed Rwanda Crisis” New York Times 31 Dec 1995 • Off 2000

References

Akande, Dapo. 2009. The Legal Nature of Security Council Referrals to the ICC and Its Impact

on Al Bashir’s Immunities. Journal of International Criminal Justice 7 (2): 333–352.

Akhavan, Payam. 1996. The International Criminal Tribunal for Rwanda: The Politics and

Pragmatics of Punishment. The American Journal of International Law 90 (3): 501-510.

Akhavan, Payam. 2001. Beyond Impunity: Can International Criminal Justice Prevent Future

Atrocities? The American Journal of International Law 95 (1): 7-31.

42

Asal, Victor, Amy Pate and Jonathan Wilkenfeld. 2008. Minorities at Risk Organizational

Behavior Data and Codebook, Version 9/2008. http://www.cidcm.umd.edu/mar/data.asp/.

http://www.cidcm.umd.edu/mar/data.aspBarkan, Joel D. 2004. Kenya after Moi. Foreign Affairs

83 (1): 87-100.

Bass, Gary Jonathan. 2000. Stay the Hand of Vengeance: The Politics of War Crimes Tribunals.

Princeton: Princeton University Press.

Berg-Schlosser, Dirk, and Gisele De Meur. 2009. Comparative Research Design: Case and

Variable Selection. In Configurational Comparative Methods: Qualitative Comparative

Analysis (QCA) and Related Techniques, edited by Benoît Rihoux and Charles C. Ragin, 19-

32. Los Angeles: Sage.

Berg-Schlosser, Dirk, Gisele De Meur, Benoit Rihoux, and Charles C. Ragin. 2009. Qualitative

Comparative Analysis (QCA) as an Approach. In Configurational Comparative Methods:

Qualitative Comparative Analysis (QCA) and Related Techniques, edited by Benoît Rihoux

and Charles C. Ragin, 1-18. Los Angeles: Sage.

Bratton, Michael, and Nicolas Van de Walle. 1994. Neopatrimonial Regimes and Political

Transitions in Africa. World Politics 46 (4): 453-489.

Box-Steffensmeier, Janet M., and Bradford S. Jones. 2004. Event History Modeling: A Guide for

Social Scientists. Cambridge: Cambridge University Press.

Bräutigam, Deborah. 2000. Aid Dependence and Governance. Stockholm: Almqvist & Wiksell

International.

Cardenas, Sonia. 2007. Conflict and Compliance: State Responses to International Human Rights

Pressure. Philadelphia: University of Pennsylvania Press.

Carpenter, R. Charli. 2007. Studying Issue (Non)-Adoption in Transnational Advocacy

Networks. International Organization 61 (03): 643-667.

Chayes, Abram, and Antonia Handler Chayes. 1993. On Compliance. International Organization

47 (2): 175-205.

Cortell, Andrew P., and James W. Davis. 1996. How Do International Institutions Matter? The

Domestic Impact of International Rules and Norms. International Studies Quarterly 40 (4):

451–478Dai, Xinyuan. 2005. Why Comply? The Domestic Constituency Mechanism.

International Organization 59 (2): 363-398.

43

Cruvellier, Thierry. 2006. Court of Remorse: Inside the International Criminal Tribunal for

Rwanda. Translated by Chari Voss. Madison: University of Wisconsin Press.

Downs, George W., David M. Rocke, and Peter N. Barsoom. 1996. Is the Good News About

Compliance Good News About Cooperation? International Organization 50 (3): 379-406.

Ford, Stuart. 2011. How Leadership in International Criminal Law is Shifting from the U.S. To

Europe and Asia: An Analysis of Spending on and Contributions to International Criminal

Courts. Saint Louis University Law Journal 55: 953–1000.

Ghosn, Faten, and Scott Bennett. 2003. Codebook for the Dyadic Militarized Interstate Dispute

Data, Version 3.10. Online: http://correlatesofwar.org.

Gifford, Paul. 1998. African Christianity: Its Public Role. Bloomington: Indiana University

Press.

Gimbel, John. 1986. U.S. Policy and German Scientists: The Early Cold War. Political Science

Quarterly 101 (3): 433-451.

Goertz, Gary, and James Mahoney. Forthcoming. A Tale of Two Cultures: Contrasting

Qualitative and Quantitative Paradigms. Princeton: Princeton University Press.

Goldsmith, Arthur A. 2001. Foreign Aid and Statehood in Africa. International Organization 55

(01): 123-148.

Goldsmith, Jack. 2003. The Self-Defeating International Criminal Court. University of Chicago

Law Review 70: 89-104.

Grodsky, Brian. 2011. Exploring the Schelling Conjecture in Reverse: “International

Constraints” and Cooperation with the International Criminal Tribunal for the Former

Yugoslavia. European Journal of International Relations 17 (1): 121-143.

Grofman, Bernard, and Carsten Q. Schneider. 2009. An Introduction to Crisp Set QCA, with a

Comparison to Binary Logistic Regression. Political Research Quarterly 62 (4): 662 -672.

Human Rights Watch. 1995. Rwanda/Zaire: Rearming with Impunity  : International Support for

the Perpetrators of the Rwandan Genocide. New York: Human Rights Watch.

Hyden, Goran. 2006. African Politics in Comparative Perspective. Cambridge: Cambridge

University Press.

International Crisis Group. 2001. International Criminal Tribunal for Rwanda: Justice Delayed.

Nairobi: International Crisis Group.

Jackson, Robert H., and Carl Gustav Rosberg. 1982. Personal Rule in Black Africa: Prince,

44

Autocrat, Prophet, Tyrant. Berkeley: University of California Press.

Jallow, Hassan B. 2008. Statement by Justice Hassan B. Jallow, Prosecutor of the ICTR, to the

United Nations Security Council. New York: 6 June 2008. Available from

<http://www.unictr.org/>. Accessed 27 December 2011.




Jallow, Hassan B. 2010a. Statement by Justice Hassan B. Jallow, Prosecutor of the ICTR, to the



Jallow, Hassan B. 2010b. Statement by Justice Hassan B. Jallow, Prosecutor of the ICTR, to the

United Nations Security Council. New York: 6 December 2010. Available from





Kahler, Miles. 2000. Conclusion: The Causes and Consequences of Legalization. International

Organization 54 (3): 661-683.

Kapiszewski, Diana, and Matthew M. Taylor. Forthcoming. Compliance: Conceptualizing,

Measuring, and Explaining Adherence to Judicial Rulings. Law & Social Inquiry.

Kaufmann, Daniel, Aart Kraay, and Massimo Mastruzzi. 2010. The Worldwide Governance

Indicators: Methodology and Analytical Issues. World Bank Policy Research Working

Paper No. 5430. Available from

<http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1682130>. Accessed 13 March 2011.

Kelley, Judith. 2007. Who Keeps International Commitments and Why? The International

Criminal Court and Bilateral Nonsurrender Agreements. American Political Science Review

101 (3): 573-589.

Kerr, Rachel. 2004. The International Criminal Tribunal for the Former Yugoslavia: An Exercise

in Law, Politics, and Diplomacy. Oxford: Oxford University Press.

Kim, Hunjoon, and Kathryn Sikkink. 2010. Explaining the Deterrence Effect of Human Rights

Prosecutions for Transitional Countries. International Studies Quarterly 54 (4): 939-963.

45

King, Gary, Robert O. Keohane, and Sidney Verba. 1994. Designing Social Inquiry: Scientific

Inference in Qualitative Research. Princeton: Princeton University Press.

Lamont, Christopher K. 2010. International Criminal Justice and the Politics of Compliance.

Surrey: Ashgate.

Magnarella, Paul J. 2000. Justice in Africa: Rwanda’s Genocide, Its Courts and UN Criminal

Tribunal. Aldershot: Ashgate.

Mahoney, James. 2003. Long‐Run Development and the Legacy of Colonialism in Spanish

America. American Journal of Sociology 109 (1): 50–106.

Mahoney, James. 2010. After KKV: The New Methodology of Qualitative Research. World

Politics 62 (1): 120–147.

McClendon, Gwyneth C. 2009. Building the Rule of International Criminal Law: The Role of

Judges and Prosecutors in the Apprehension of War Criminals. Human Rights Review 10

(3): 349-372.

Mearsheimer, John J. 1994. The False Promise of International Institutions. International

Security 19 (3): 5-49.

Meernik, James. 2008. It’s Time to Stop Running: A Model of the Apprehension of Suspected

War Criminals. International Studies Perspectives 9 (2): 165-182.

Mesquita, Bruce Bueno de, Alastair Smith, Randolph M. Siverson, and James D. Morrow. 2004.

The Logic of Political Survival. Cambridge: MIT Press.

Minorities at Risk Project. 2009. Minorities at Risk Dataset. College Park, MD: Center for

International Development and Conflict Management. http://www.cidcm.umd.edu/mar/.

Moghalu, Kingsley. 2005. Rwanda’s Genocide: The Politics of Global Justice. New York:

Palgrave Macmillan.

Moon, Bruce E. 1983. The Foreign Policy of the Dependent State. International Studies

Quarterly 27 (3): 315–340.

Moravcsik, Andrew. 1997. Taking Preferences Seriously: A Liberal Theory of International

Politics. International Organization 51 (4): 513-553.

Off, Carol. 2000. The Lion, the Fox and the Eagle. Toronto: Random House Canada.

Peskin, Victor. 2008. International Justice in Rwanda and the Balkans: Virtual Trials and the

Struggle for State Cooperation. Cambridge: Cambridge University Press.

Poe, Steven C, C. Neal Tate, and Linda Camp Keith. 1999. Repression of the Human Right to

46

Personal Integrity Revisited: A Global Cross‐National Study Covering the Years 1976–

1993. International Studies Quarterly 43 (2): 291-313.

Poe, Steven C., and C. Neal Tate. 1994. Repression of Human Rights to Personal Integrity in the

1980s: A Global Analysis. The American Political Science Review 88 (4): 853-872.

Del Ponte, Carla. 2009. Madame Prosecutor: Confrontations with Humanity’s Worst Criminals

and the Culture of Impunity  : A Memoir. New York: Other Press, LLC.

Prunier, Gerard. 2009. Africa’s World War: Congo, the Rwandan Genocide, and the Making of a

Continental Catastrophe. Oxford: Oxford University Press.

Ragin, Charles C. 1987. The Comparative Method: Moving Beyond Qualitative and Quantitative

Strategies. Berkeley: University of California Press.

Ragin, Charles, Kriss A. Drass, and Sean Davey. 2006. Fuzzy-Set/Qualitative Comparative

Analysis 2.0. Department of Sociology, University of Arizona, Tuscon, AZ.

Rajkovic, Nikolas M. 2012. The Politics of International Law and Compliance: Serbia, Croatia

and The Hague Tribunal. London: Routledge.

Rihoux, Benoit, and Gisele De Meur. 2009. Crisp-Set Qualitative Comparative Analysis

(csQCA). In Configurational Comparative Methods: Qualitative Comparative Analysis

(QCA) and Related Techniques, edited by Benoît Rihoux and Charles C. Ragin, 33-68. Los

Angeles: Sage.

Rihoux, Benoît, and Charles C. Ragin, eds. 2009. Configurational Comparative Methods:

Qualitative Comparative Analysis (QCA) and Related Techniques. Los Angeles: Sage.

Ryan, Allan A. 1983. Klaus Barbie and the United States Government: A Report to the Attorney

General of the United States. Washington, D.C.: United States Department of Justice.

Sabar-Friedman, Galia. 1995. “Politics” and “Power” in the Kenyan Public Discourse and Recent

Events: The Church of the Province of Kenya (CPK). Canadian Journal of African Studies /

Revue Canadienne des Études Africaines 29 (3): 429-453.

Sabar-Friedman, Galia. 1997. Church and State in Kenya, 1986–1992: The Churches’

Involvement in the “Game of Change.” African Affairs 96 (382): 25-52.

Schabas, William A. 2006. The UN International Criminal Tribunals: The Former Yugoslavia,

Rwanda and Sierra Leone. Cambridge: Cambridge University Press.

Schabas, William. 2011. An Introduction to the International Criminal Court. 4th ed. Cambridge:

Cambridge University Press.

47

Shklar, Judith N. 1986. Legalism: Law, Morals, and Political Trials. Cambridge: Harvard

University Press.

Slaughter, Anne-Marie. 1995. International Law in a World of Liberal States. European Journal

of International Law 6 (1): 503-538.

Subotic, Jelena. 2009. Hijacked Justice: Dealing With the Past in the Balkans. Ithaca: Cornell

University Press.

Tejan-Cole, Abdul. 2009. A Big Man in a Small Cell: Charles Taylor and the Special Court for

Sierra Leone. In Prosecuting Heads of State, edited by Ellen L. Lutz and Caitlin Reiger,

205–232. Cambridge: Cambridge University Press.

Valentino, Benjamin, Paul Huth, and Sarah Croco. 2006. Covenants Without the Sword:

International Law and the Protection of Civilians in Times of War. World Politics 58 (3):

339-377.

Whitaker, Beth Elise. 2010. Compliance Among Weak States: Africa and the Counter-Terrorism

Regime. Review of International Studies 36 (3): 639-662.

World Bank. 2011. World Development Indicators. Washington, D.C.: World Bank.

Young, Oran R. 2010. Compliance & Public Authority: A Theory With International

Applications. Baltimore: Johns Hopkins University Press.

48

Figure 1: Comparison of arrest warrants durations across coding of “compliers” and “noncompliers”

0.00

0.25

0.50

0.75

1.00

0 1000 2000 3000 4000days from indictment to arrest

Noncompliers Compliers

Kaplan-Meier survival estimates

49

Table 1: State compliance with the ICTR (raw scores)

Positive Case Conditions Negative Case Conditions

Country Initial Behavior Year Dependency

(Aid/GDP) Regime (Polity2)

Rule of Law (WB)

Noncomp. Constituency

Rent-seeking (WB)

National Security (COW)

Cameroon N 1996 0.05 -4 -1.44 0 -1.16 0 Gabon N 2005 0.01 -4 -0.47 0 -0.55 0 Italy N 2001 0.00 10 0.80 1 0.76 0 Kenya N 1995 0.07 -5 -0.98 0 -1.03 0 Congo-Brazzaville N 2000 0.03 -6 -1.43 0 -0.96 0 DRC N 1997 0.02 0 -1.95 0 -2.06 1 Zimbabwe N 2000 0.03 -3 -1.25 0 -0.94 1 Belgium C 1996 0.00 10 1.35 0 1.30 0 Benin C 1998 0.09 6 -0.13 0 -0.60 0 Burkina Faso C 1997 0.15 -4 -0.85 0 0.47 0 Denmark C 2000 0.00 10 1.81 0 2.51 0 France C 1999 0.00 9 1.36 0 1.40 0 Germany C 1999 0.00 10 1.62 0 2.16 0 Ivory Coast C 1996 0.08 -6 -0.81 0 0.20 0 Mali C 2000 0.13 6 -0.47 0 -0.67 0 Namibia C 1998 0.05 6 0.27 0 0.59 1 Netherlands C 2001 0.00 10 1.73 0 2.33 0 Senegal C 2002 0.08 8 -0.04 0 0.29 0 South Africa C 1995 0.00 9 -0.01 0 0.76 0 Switzerland C 1996 0.00 10 1.95 0 2.10 0 Tanzania C 1995 0.17 -1 -0.21 0 -1.03 0 Togo C 1997 0.09 -2 -0.69 0 -0.76 0 Uganda C 2000 0.13 -4 -0.75 0 -0.86 0 UK C 2000 0.00 10 1.64 0 2.17 0 US C 1996 0.00 10 1.48 0 1.56 0 Zambia C 1995 0.33 6 -0.63 0 -1.03 0

C = Initial complier; N = Initial noncomplier

50

Table 2: State compliance with the ICTR (crisp set memberships)

Positive Case Conditions Negative Case Conditions

Country Initial Behavior Dependency Regime Rule of Law Noncomp.

Constituency Rent-

Seeking National Security

Cameroon N 0 0 0 0 1 0 Gabon N 0 0 0 0 1 0 Italy N 0 1 1 1 0 0 Kenya N 0 0 0 0 1 0 Congo-Brazzaville N 0 0 0 0 1 0 DRC N 0 0 0 0 1 1 Zimbabwe N 0 0 0 0 1 1 Belgium C 0 1 1 0 0 0 Benin C 1 1 0 0 1 0 Burkina Faso C 1 0 0 0 0 0 Denmark C 0 1 1 0 0 0 France C 0 1 1 0 0 0 Germany C 0 1 1 0 0 0 Ivory Coast C 0 0 0 0 0 0 Mali C 1 1 0 0 1 0 Namibia C 0 1 1 0 0 1 Netherlands C 0 1 1 0 0 0 Senegal C 0 1 0 0 0 0 South Africa C 0 1 0 0 0 0 Switzerland C 0 1 1 0 0 0 Tanzania C 1 0 0 0 1 0 Togo C 1 0 0 0 1 0 Uganda C 1 0 0 0 1 0 UK C 0 1 1 0 0 0 US C 0 1 1 0 0 0 Zambia C 1 1 0 0 1 0 C = Initial complier; N = Initial noncomplier

51

Table 3: Model 1 truth table (with positive case conditions only)

Dependency Regime Law N Compliance Consistency Cases

0 1 0 2 C 1 Senegal, South Africa 1 1 0 3 C 1 Benin, Mali, Zambia

1 0 0 4 C 1 Burkina Faso, Tanzania, Togo, Uganda

0 1 1 10 C 0.9

Belgium, Denmark, France, Germany, Italy, Namibia, Netherlands, Switzerland, UK, US

0 0 0 7 N 0.14

Cameroon, Congo-Brazzaville, Gabon, Ivory Coast, Kenya, Zaire/DRC, Zimbabwe

C = Initial complier; N = Initial noncomplier Table 4: Model 2 truth table (with positive and negative case conditions)

Dependency Regime Law Security Rent Constituency N Compliance Consistency Cases

0 1 1 0 0 1 1 N 1 Italy

0 0 0 0 1 0 4 N 1

Cameroon, Gabon, Kenya, Congo-Brazzaville

0 0 0 1 1 0 2 N 1 DRC, Zimbabwe

1 0 0 0 1 0 3 C 0

Tanzania, Togo, Uganda

1 1 0 0 1 0 3 C 0 Benin, Mali, Zambia

0 1 0 0 0 0 2 C 0 Senegal, South Africa

0 1 1 1 0 0 1 C 0 Namibia 0 0 0 0 0 0 1 C 0 Ivory Coast

0 1 1 0 0 0 8 C 0

Belgium, Denmark, France, Germany, Netherlands, Switzerland, UK, US

1 0 0 0 0 0 1 C 0 Burkina Faso C = Initial complier; N = Initial noncomplier

52

Table 5: Event history analysis of arrest warrant durations

Model 1 Model 2 Model 3 Model 4 Model 5 Regime 0.986 0.941 1.008 (0.726) (0.261) (0.841) Rule of law 424.932 800.689 3206.085 (0.055) (0.033) (0.007) Aid dependency 44.48 1005323 117722 10.68 679285 (0.283) (0.016) (0.025) (0.424) (0.011) Control of corruption 4.316 292875 0.023 1.217 0.008 (0.011) (0.238) (0.182) (0.556) (0.097) Aid dependency * control of corruption 75.715 981200 91056.4 27.717 237036.5 (0.136) (0.010) (0.023) (0.312) (0.026) National security 0.515 0.119 (0.161) (0.009) Control of corruption * log of time1 0.851 1.689 1.816 2.184 (0.066) (0.222) (0.158) (0.059) Rule of law * log of time1 0.431 0.391 0.282

(0.066) (0.041) (0.004)

Number of indictees 46 46 46 46 46 Wald 19.42 31.850 33.86 21.36 42.04 (0.002) (0.000) (0.000) (0.001) (0.000) Log pseudolikelihood -125.6 -122.6 -122.0 -127.6 -117.4 p (Grambsch and Therneau global proportionality test) 0.979 0.829 0.579 0.548 0.771 Cox proportional hazard model with robust standard errors. Table displays hazard ratios with p-values in parentheses. Ratios greater than 1 indicate a positive effect on the likelihood of faster arrests. Rations less than 1 indicate a negative effect on the likelihood of faster arrests. Bold indicates p < .10 1. Interaction term included to correct for nonproportionality

Untitled

Documents

configurational

iccs increasing

security council

national security

national security

gisele de

international

event history