Abstract

States frequently employ overt and covert foreign-imposed regime change (FIRC) to pursue their foreign policy interests. Yet there is little scholarship on the question of whether FIRCs improve relations between the states involved. In fact, most FIRCs either fail to reduce—or increase—the likelihood of militarized disputes between interveners and targets. Fundamentally, FIRC entails a principal-agent problem: foreign-imposed leaders rule over states with interests different from those of the intervener. Whereas the intervening state wants the new leader to pursue policies that reflect its interests, once in power, such leaders are focused on ensuring their political survival, a task that is often undermined by implementing the intervener's agenda. Foreign-imposed leaders who carry out the intervener's desired policies attract the ire of domestic actors. These domestic opponents can force the regime to reverse course or may even remove it from power in favor of leaders who are hostile to the intervener; in both cases, the result can be renewed conflict with the intervener. Rwanda's replacement of Mobutu Sese Seko with Laurent-Désiré Kabila in Zaire illustrates this problem.

Introduction

Does foreign-imposed regime change (FIRC) improve relations between the states involved? If a state replaces the political leadership of a rival, will the new regime behave as a friend or foe? Regime change—both covert and overt—is a common foreign policy tool used by the United States and other countries to pursue their interests. Since 1816, states have overthrown more than 100 foreign leaders in overt interventions.1 During the Cold War, the United States alone attempted sixty-three covert FIRCs, twenty-four of which succeeded in bringing new regimes to power.2 In the post–Cold War era, it continues to be one of the key practitioners of FIRC, intervening in Haiti (1994), Afghanistan (2001), Iraq (2003), Libya (2011), and Syria (2012).

The literature on the effects of FIRC and related forms of intervention—such as imposed polities, military occupation, forceful democracy promotion, and nation building—is extensive.3 Most of it, however, concerns internal outcomes in the targeted states—such as democratization and civil war—rather than relations between the interveners and their targets following regime change. The only study that explicitly addresses the effect of FIRC on interstate relations finds that FIRC reduces the likelihood of conflict between interveners and targets.4 According to the study, FIRC enables intervening states to install new leaders with similar policy preferences in target states. In theory, this alignment of interests should resolve disputes between the two states and enhance cooperation. In the best-case scenario, the newly installed regime becomes a reliable client state that promotes the intervener's interests at home and abroad.5

In this article, we conduct a new analysis of FIRC and interstate relations. In contrast to existing accounts, we argue that FIRC generally does not improve relations between interveners and targets. Rather, it often makes them worse. Fundamentally, state interests have deeper roots than the beliefs or policies of any one leader. Changing a state's leader, therefore, is not synonymous with changing its interests. In addition, FIRC entails a principal-agent problem: foreign-imposed leaders rule over states with interests different from those of the intervener. Whereas the intervening state (the principal) wants the new leader (the agent) to pursue policies that reflect its interests, once in power, the new leader is focused on ensuring his or her own political survival, a task that is often undermined by implementing the intervener's agenda.

Foreign-imposed leaders face pressure from two main sources to resist implementing the intervener's desired policies. First, they may promise to pursue the interests of their foreign patron but, once in power, find that they are constrained by geopolitical realities, such as threats from neighboring states. When the interests dictated by their geopolitical environment clash with those of the intervener, conflict between the two states may occur. Second, foreign-imposed leaders who do implement the intervener's desired policies may attract the ire of domestic actors. These domestic opponents can force the regime to reverse course or may even remove it from power in favor of leaders hostile to the intervener; in both cases, the result can be renewed conflict with the intervener. Foreign-imposed leaders therefore face a Catch-22: the more they attempt to implement the intervener's preferred policies, the more likely they are to face resistance from domestic actors. On the other hand, leaders who seek to appease domestic constituencies by defying the intervener risk angering the intervener and provoking interstate conflict with it.

Among overt and covert FIRCs that succeed in replacing targeted leaders, there are three different types that vary in how acutely they trigger this dilemma. First, “leadership FIRCs” replace one foreign leader with another without building or bolstering political institutions within the target state. Leaders installed during leadership FIRCs face the legitimacy problems discussed above most severely. Because these leaders typically lack homegrown support, they are domestically unpopular and rely on their external patron to sustain their rule. In implementing the intervener's policies, however, they further alienate their domestic constituency. These leaders are thus likely to face strong internal pressures to change course or else risk violent resistance; indeed, the majority of these leaders are overthrown and typically replaced by elites hostile to the intervener.6 Whether imposed leaders turn against the intervener on their own initiative or are displaced by other leaders who do so, we argue that the likelihood of intervener-target conflict increases.

Second, “institutional FIRCs” seek to build new political institutions in the target state in addition to removing leaders.7 We contend that institutional FIRCs neither increase nor decrease the likelihood of intervener-target conflict. On the one hand, institutional FIRCs that install repressive institutions enhance the new leader's ability to survive in office, enabling him or her to carry out the intervener's policies and thereby ensuring harmonious relations. Institutional FIRCs that construct democratic institutions promote peaceful relations through a different mechanism: the ability of democracies to resolve their differences without violence.8 Democracy, however, also brings policy more into line with voter preferences in the target state. If these preferences do not align with those of the intervener, the result could be growing rancor (although open conflict is unlikely).9 The bigger problem is that institutional FIRCs do not always succeed in establishing consolidated regimes. When they fail, hostile leaders may take power in target states, leading to further conflict. We argue that these successes and failures cancel each other out, resulting in no systematic relationship between institutional FIRC and interstate conflict.

Third, “restoration FIRCs” reinstate leaders who previously held power in the target state. These individuals already enjoyed some domestic legitimacy and good relations with the intervener. They are thus less likely to be viewed by their domestic constituencies as tools of foreign interests, and restoring them marks a return to a satisfactory status quo for the intervener rather than an ambitious attempt to transform conflictual relations into peaceful relations. We thus hypothesize that restoration FIRCs decrease the likelihood of conflict between targets and interveners.

Thus far our theory applies only to FIRCs that succeed, but regime change operations—particularly covert ones—often fail.10 Although failed covert operations do not suffer from the principal-agent problem outlined above, because these operations are almost always discovered by the target, they often transform a tense relationship into an openly conflictual one, exacerbating the likelihood of conflict between the intervener and the target. We thus argue that failed attempts at covert regime change increase the probability of militarized disputes between the states involved.

This article introduces a new dataset of FIRC that combines data on all overt FIRCs successfully carried out by any country in the world from 1816 to 2000 with data on failed and successful U.S.-backed covert FIRCs during the Cold War. The dataset has three major advantages over existing studies. First, it includes examples of both covert and overt FIRCs, whereas existing studies focus on only overt cases.11 Second, it contains examples of both successful (the target government was replaced) and failed (the target government was not replaced) covert FIRCs launched by the United States during the Cold War, whereas existing studies focus on only successful examples. This expanded sample allows us to evaluate the ramifications of failed covert FIRCs for the states involved. Third, the dataset differentiates between FIRCs that aim to replace the political institutions of the target state and those that aim to replace only its leaders or restore previous leaders to power. This categorization of FIRCs allows us to test whether certain types of regime transformations are more effective than others.

To test our theoretical propositions, we investigate the effect of FIRC on the likelihood of subsequent militarized interstate disputes (MIDs) between intervening and target states.12 Our empirical analysis largely confirms our theoretical expectations. We find, for example, that interstate dyads that experience an overt leadership FIRC are more than twice as likely subsequently to experience a MID; interventions that replace target states' institutions exert no discernable effect on post-FIRC conflict; and FIRCs that restore to power leaders deposed within the last five years reduce the likelihood of conflict. Similarly, covert leadership FIRCs significantly increase the probability of a post-FIRC MID, whereas covert institutional FIRCs do not reliably do so. The effect of covert FIRCs, however, depends on whether or not they succeed in deposing the targeted leader. Successful covert leadership FIRCs, contrary to our expectations, exert no significant effect on the likelihood of a post-FIRC MID. This type becomes strongly positive over the longer term, however, indicating that these FIRCs may eventually backfire. Failed covert FIRCs of any type, by contrast, strongly increase the probability of militarized conflict.

These findings suggest that policymakers have been overly optimistic about the utility of FIRC. Overthrowing a foreign government to replace an unfriendly leader with a more favorable one may appear to be an easy way to improve interstate relations. In many cases, however, the results are counterproductive. The mere replacement of a leader does not resolve the underlying reasons for interstate conflict, leaving open the possibility of renewed conflict.

To illustrate some of the causal mechanisms of our theory, we present an indepth case study of Rwanda's ill-fated leadership FIRC in Zaire in 1997. To eradicate the threat posed by Hutu militants housed in refugee camps in eastern Zaire following the Rwandan genocide, Rwanda's Tutsi regime ordered an invasion of Zaire in October 1996, ending President Mobutu Sese Seko's three decade-long rule. To avoid a recurrence of this threat, the Rwandans installed Laurent-Désiré Kabila in Mobutu's place to rule Zaire, which was then renamed the Democratic Republic of Congo (DRC). Kabila, however, struggled to consolidate domestic control over the DRC because most Congolese viewed him as a Rwandan puppet. Facing severe domestic unrest, Kabila turned on his Rwandan patrons, expelling all Rwandan troops in July 1998. In response, Rwanda launched a second invasion to overthrow its defiant protégé, sparking the Second Congo War (1998–2003).

The remainder of this article proceeds as follows. First, we review the existing literature on FIRC. Second, we present our theory of how FIRC affects relations between interveners and target states and outline our hypotheses. Third, we discuss our research design for testing our hypotheses quantitatively. Fourth, we present our statistical findings. Fifth, we recount Rwanda's replacement of Mobutu with Kabila in Zaire and how it triggered another conflict between the two states. We conclude by considering some of the policy implications of our findings.

Foreign-Imposed Regime Change: The Existing Literature

Scholarly interest in phenomena such as FIRC, military occupation, and nation building has boomed in the wake of the U.S. overthrow of the governments of Afghanistan and Iraq. One strand of this literature analyzes the democratizing potential of such operations. Although some optimists argue that military intervention can foster democratization,13 few empirical studies agree. One group of studies, for example, finds that although foreign military interventions can make target states marginally more democratic, it rarely results in consolidated democracies.14 Other studies maintain that democratization is more likely following foreign military interventions if the interveners invest substantial resources in promoting democracy.15 A third group contends that foreign military intervention triggers democratization only when target states already enjoy conditions favorable to democracy, such as high levels of wealth, ethnic homogeneity, and previous experience with constitutional rule.16

Scholars have also sought to understand when foreign military intervention will spawn internal conflict. Goran Peic and Dan Reiter argue that FIRC, in combination with defeat in interstate war, contributes to the outbreak of civil war by destroying “physical and human infrastructure,” weakening state security forces, and removing competent government officials.17 Alexander Downes also finds that a FIRC enacted simultaneously with defeat in an interstate war can lead to civil conflict, but additionally shows that FIRC has a strong and independent effect on civil war.18

Most relevant to this article is scholarship that assesses the effect of FIRC on the likelihood of international conflict, particularly the chances of a post-FIRC conflict between an intervener and a target. Here, the only existing study finds that regime change promotes peace. Nigel Lo, Barry Hashimoto, and Dan Reiter, examining the effect of FIRC as a war termination strategy, find that peace lasts significantly longer after wars in which the winner imposes FIRC on the loser. The authors argue that by putting a friendly leader in charge—or by transforming a country's political institutions—an intervener changes the target's preferences and thus reduces the likelihood of further conflict between the two states.19

Although Lo, Hashimoto, and Reiter's study suggests that FIRC leads to peace, two aspects of the study leave unresolved the question of the overall effect of FIRC on interstate relations. First, the sample of states in the study includes only interstate war participants. It is unclear, however, whether the effect of FIRC in this limited sample generalizes to all states. Our study widens the lens by estimating the effect of several different types of FIRC on relations among all states whether in peacetime or war.20 Second, Lo and his colleagues use war recurrence as the dependent variable. War is an extreme indicator of interstate relations, however. Interveners and targets could experience conflict-ridden relations after FIRC without going to war. Including more sensitive indicators expands the scope of the analysis regarding the effects of regime change on post-FIRC relations. In this article, we suggest and test one such indicator: militarized interstate disputes.

A Theory of Foreign-Imposed Regime Change and Interstate Relations

This section develops our argument and introduces seven hypotheses about the relationship between FIRC and the likelihood of subsequent militarized conflict between the intervening and target states. We begin by outlining how policymakers expect FIRC to work. Next, we explain why, on average, FIRC does not improve intervener-target relations and, in many cases, makes relations worse. Finally, we describe why different types of FIRCs have different effects on the subsequent likelihood of conflict.

Before continuing, we would like to note that, in general, our hypotheses compare the likelihood of military conflict for interstate dyads that experienced FIRC to those that did not. The statistical results we present below are based on this comparison. However, our hypotheses also apply within dyads that experienced FIRC, comparing the period before to the period after FIRC occurred. In additional tests described below, we find that our hypotheses apply equally well to both types of variation.

HOW POLICYMAKERS EXPECT FIRCS TO WORK

States engage in FIRC to align the policy preferences of another state with their own, thereby reducing the probability of future conflict between the two states. The logic behind their reasoning is simple: if the operation is successful, the newly installed regime and the intervener will have the same interests, suggesting that the former will then act in the latter's interests without having to be bribed or coerced into doing so.21 After the FIRC, disputes between the intervener and the target government are less likely to arise because the target state should voluntarily act in the intervener's interests. This logic suggests that, regardless of the specific source of disagreement during an interstate dispute, FIRC should have a positive influence on intervener-target relations by changing the two states' relationship from one marred by conflicting interests to one characterized by mutual interests.22 Hypothesis 1 reflects this conventional wisdom.

Hypothesis 1: Foreign-imposed regime change decreases the likelihood of militarized conflict between intervening and target states.

WHY FIRC DOES NOT IMPROVE INTERVENER-TARGET RELATIONS

Contrary to proponents' expectations, however, changing another state's preferences and behavior is not a simple task. To begin, states' interests never fully overlap, so international relations are always characterized by some degree of interest divergence. Moreover, changing a state's leader is not synonymous with changing its interests. Instead, replacing the leadership of another state creates a principal-agent problem. Once they take power, foreign-imposed leaders endeavor to hold on to it. Yet because the interests of the intervening and target states are never fully congruent, serving the intervener's interests may upset domestic audiences, which in turn could threaten the new leader's chances of political survival. In some cases, this interest divergence is compounded because interveners also ask the target leader to take actions that make the target more vulnerable to other external threats. Foreign-imposed leaders are thus often “damned if they do and damned if they don't.” If they faithfully implement the intervener's preferred policies, they face possible removal by internal actors or external opponents; if they diverge from the intervener's wishes, however, they increase the chance that the intervener will seek to remove them.

interests and leaders. There are two basic problems with the prevailing view of FIRC. First, states' interests never completely overlap. Even close allies often have major disagreements on important issues. In the post–Cold War era, for example, the United States and its NATO partners have diverged over how to approach the Bosnian conflict, how intensely to bomb Yugoslavia in 1999, whether or not to invade Iraq in 2003, and how to deal with Iran's nuclear program.23 Given that the divergence of state interests can impede cooperation even among the staunchest of allies, this problem is generally much worse between adversaries, and even more severe when one state is contemplating FIRC against another. In pursuing regime change, the intervener has decided that less violent forms of persuasion are unlikely to bring the target state in line with its preferences. Conversely, the choice of the target government to resist rather than acquiesce to the intervener's demands when the conflict first arose—thereby causing the dispute to escalate to the point of regime change—indicates that the target, too, perceived the interests at stake to be substantial, perhaps existential.24

Second, states' interests are rarely those of individual leaders alone. This is not to deny that leaders exert important influence on the direction and details of states' foreign policies. Elizabeth Saunders, for example, documents the divergent approaches to the fight against communism in Vietnam taken by Presidents Dwight D. Eisenhower, John F. Kennedy, and Lyndon B. Johnson.25 Yet despite their different tactics, all three leaders agreed on a broad strategy of containing the Soviet Union from expanding its influence in Western Europe and East Asia. This continuity reflects the fact that states' foreign policy interests are often relatively stable and long-lived, and rarely boil down to the preferences of particular individuals.26

the principal-agent dynamic. FIRC suffers from a principal-agent problem: the intervening state (the principal) installs a new leader (the agent) in the target state, but the principal has limited means to ensure that the agent follows the principal's demands rather than pursuing its own interests.27 One reason why an agent might diverge from the principal's agenda is adverse selection: the agent is able to convince the principal that it shares the latter's views when in fact it does not. This plausible scenario likely explains some of the misalignment of interests that follows many FIRCs. Major misalignments, however, are typically not the result of agents deliberately deceiving principals. Rather, an agent's main interest upon assuming power becomes survival in office, and following the principal's guidance may endanger that goal. Although the agent may have promised to implement the principal's preferred policies, political constraints prevent him or her from doing so after assuming office. In other cases, the agent may be overthrown by domestic forces who object to the intervener's demands.28 Either way, intervener-target relations are unlikely to improve.29

Some might wonder why interveners are unable to solve the principal-agent problem inherent in FIRC through increased monitoring—for instance, by maintaining a full-scale military occupation or by stationing troops in the target state. These measures, however, are unlikely to solve the problem. In a full-scale military occupation, the intervener assumes responsibility for governing the target state, rather than appointing an agent. Furthermore, in the age of nationalism, occupying powers are often plagued by legitimacy problems that undermine their ability to effectively govern the target state.30 Just stationing troops in the target country without taking on governance functions, by contrast, is expensive and does not guarantee that the intervener will be able to influence the agent's actions. Additionally, once the intervener decides to withdraw, the indigenous government left in its place may struggle to consolidate power given the delegitimizing effect the presence of foreign troops has had on its regime.

The experiences of the United States in Afghanistan and Iraq highlight some of the difficulties that principals encounter in influencing their agents. In Afghanistan, relations between the regime of Hamid Karzai and Washington deteriorated to such an extent that the Afghan president accused the United States of conspiring with the Taliban to convince Afghans that violence would increase if U.S. troops left the country.31 In Iraq, Nouri al-Maliki, the prime minister of Iraq from 2006 to 2014, declared that despite the United States' substantial troop commitment to his country and $60 billion in U.S. support, “I'm a friend to the United States, but not America's man in Iraq.”32

external sources of interest divergence. Both before and after FIRCs, political leaders of target states may, for geopolitical reasons, refuse to heed adversaries' demands.33 A weak state targeted for regime change, for example, may reject a stronger state's demand to relinquish some of its military capabilities or take other actions that would jeopardize its future security. Even if the two states can reach an agreement whereby the weaker state cedes some of its capabilities to avoid FIRC, the stronger state would still have to commit not to renege on the agreement and exploit its opponent's weakened position.34 Further complicating the target government's decisionmaking calculus, the intervening state may not be its only external security threat. Consequently, when a powerful state demands that the weaker state take an action that would substantially weaken it further—such as relinquishing nuclear weapons, surrendering territory, forgoing an alliance, or pursuing a less aggressive foreign policy—the target government is unlikely to comply.35 This was the problem confronting Saddam Hussein prior to the 2003 Iraq War, as Neil MacFarquhar explains: “The fatal controversy over whether Iraq was still developing unconventional weapons stemmed in part from Mr. Hussein's desire to convince different audiences of different things…. He wanted the West to believe that he had abandoned the program, which he had. Yet he also wanted to instill fear in enemies like Iran and Israel, plus maintain the esteem of Arabs, by claiming that he possessed the weapons.”36

The presence of other external threats may not only cause target states to resist interveners' demands prior to regime change, but may also lead to disputes with the intervener after FIRC. Experiencing FIRC does not necessarily change a target state's external security environment. As a result, foreign-imposed leaders will likely confront the same security threats as their predecessors. If the intervener requires that the target reduce its military forces or relinquish strategic territory in ways that make it vulnerable to these sources of threat, even leaders imposed by the intervener might resist, sparking further conflict. Conversely, leaders who comply with the intervener's demands could face resistance from domestic actors opposed to these concessions, causing the imposed leader to balk or be overthrown by the domestic opposition. Either of these scenarios could lead to renewed conflict between the target and the intervener.

internal sources of interest divergence. The other source of interest divergence between foreign-imposed leaders and interveners lies in the target state's domestic politics. Because states' interests are not isomorphic, foreign-imposed leaders who pursue policies that primarily benefit their external patrons risk angering influential domestic groups.37 After the U.S.-backed overthrow of Guatemala's Jacobo Arbenz in 1954, for example, one of Arbenz's successors—Miguel Ydígoras Fuentes—permitted the United States to train Cuban exiles in Guatemala to overthrow Fidel Castro. This policy alienated members of the Guatemalan officer corps, some of whom launched an armed rebellion to oust Fuentes, resulting in a thirty-year civil war.38

Previous studies of regime change have found that FIRC increases the likelihood that targets will experience civil war,39 but these studies neglect the link connecting civil conflict, regional instability, and interstate conflict.40 Civil wars, in addition to having disastrous internal effects, can have negative consequences for neighboring states that can lead to interstate disputes, such as refugee flows and humanitarian crises;41 lower rates of economic growth;42 and political instability.43 One study finds that the external consequences of civil wars are the root cause of many interstate conflicts: “Many MIDs indeed do originate in intrastate conflicts. Indeed, about 25% of the events over the period 1993–2001 … involve issues that we would normally think of [sic] internal affairs.”44 These malignant dynamics played out in the Middle East following the 2003 regime change in Iraq. Although U.S. policymakers anticipated that Saddam Hussein's ouster would increase regional stability, the intervention had the opposite effect by destabilizing Iraq and its neighbors, and contributing to subsequent U.S. MIDs with Iran and Syria.45

Imposed leaders may respond to domestic disapproval—and the risk of civil war or violent removal from power—by refusing to implement the policies of the intervener or by breaking completely with it. As described in greater detail below, when the high visibility and influence of ethnic Tutsis (both Rwandan and Congolese) in his regime jeopardized newly installed Laurent Kabila's hold on power in the DRC, Kabila reacted by expelling his Rwandan backers and elevating members of his own ethnic group. In response, Rwanda and Uganda, which only the year before had put Kabila in power, sought his removal.

At a minimum, therefore, we expect that, on average, FIRC does not improve intervener-target relations, and may actually harm them.46 This expectation leads to our hypothesis 2.

Hypothesis 2: Foreign-imposed regime change does not decrease the likelihood of militarized conflict between intervening and target states.

DIFFERENT TYPES OF FIRC, DIFFERENT EFFECTS

Certain types of foreign-imposed regime change may be more harmful to the relationship between intervening and target states than others. In this section, we theorize the impact of three different sets of FIRC operations: (1) FIRCs that introduce new leaders versus those that restore recently deposed leaders versus those that change the target states' political institutions; (2) overt versus covert FIRCs; and (3) covert FIRCs that succeed in replacing the target government versus covert FIRCs that fail.

leadership, restoration, and institutional firc. States launch FIRCs with different objectives in mind. Most FIRCs aim to replace a country's political leader without meddling in its political institutions. Others are undertaken to reinstate a recently deposed foreign leader. Still others attempt to replace a country's political institutions—for instance, by transforming an autocracy into a democracy. We argue that the type of FIRC attempted by the intervener—leadership, restoration, or institutional—influences its subsequent relationship with the target state. Specifically, we argue that leadership FIRCs increase conflict between the intervening and target state; restoration FIRCs decrease conflict; and institutional FIRCs have a mixed effect.

Leadership FIRCs are likely to harm interstate relations because they are most likely to place leaders in a Catch-22. When interveners replace a foreign leader without building any supporting institutions to empower the new regime, the leaders are likely to remain heavily dependent on aid from the intervener to maintain power. Although such puppet leaders are thus more likely to adopt the intervener's interests initially, doing so tends to generate internal opposition, putting the leader's political survival in jeopardy. Furthermore, this kind of imposed leader often lacks a strong domestic base of support, making him or her especially vulnerable to domestic threats. Elites installed in leadership FIRCs are thus subject to powerful pressures to cut ties with their foreign patrons, which may then trigger military conflict between interveners and targets.

Intervener-target relations may also be soured because puppet leaders empowered through a leadership FIRC do not survive long in office. Nearly two-thirds of leaders installed in overt leadership FIRCs and roughly half of leaders installed in covert leadership FIRCs are either assassinated or violently overthrown.47 Members of the opposition who seize power generally resent or despise the intervener for meddling in their internal affairs, which in turn increases the likelihood of conflict between the two states. The overthrow of the shah of Iran is a prime example. In his analysis of the 1953 U.S.-backed coup against Prime Minister Mohammad Mossadegh, historian Malcolm Byrne writes: “Because of its role in the coup, many Iranians came to identify Washington as the shah's all-powerful patron…. This virtually guaranteed that burgeoning hostility toward the shah would also be directed against the United States when the revolutionary Islamic regime came to power in 1979.”48 This discussion suggests our third hypothesis.

Hypothesis 3: Leadership FIRC increases the likelihood of militarized conflict between the intervening and target states.

We hypothesize that, in contrast to leadership FIRCs, restoration FIRCs may improve relations (and decrease the likelihood of conflict) between the intervening and target states, for three principal reasons. First, in almost all restoration FIRCs, reinstalled leaders were previously aligned with the intervener. Thus, these FIRCs mark a return to an acceptable international status quo for both parties rather than an attempt to change state preferences by imposing entirely new leaders. In the Soviet Union's Cold War restoration FIRCs in Hungary (1956) and Czechoslovakia (1968), for example, Moscow was formally allied with its Eastern European neighbors and had ideological confrères in each country committed to close relations with the Soviet Union. Similarly, Austria and Prussia shared monarchic principles of rule with leaders in the small countries where they intervened in the first half of the nineteenth century—for example, Sicily, Tuscany, Baden, and Saxony—to thwart republican revolutions and restore ousted monarchs to the throne. The United States and Great Britain also rehabilitated several democratic regimes in allied countries, such as Belgium, the Netherlands, Norway, and Denmark, that had been toppled by Nazi Germany during World War II. In these cases, the leaders being restored already shared the intervener's policy preferences—if they did not, the intervening state would not have reinstalled them.

Second, in a large proportion of cases, the regimes empowered by restoration FIRCs previously enjoyed substantial domestic legitimacy and therefore were not viewed by their citizens as tools of foreign interests after being brought back to power. In seventeen of the thirty-six restoration FIRC dyads, for instance, the intervener reinstalled leaders who had been ousted by foreign powers, not by their own populations. Fourteen of these dyads involved democracies removing German occupation regimes after the two world wars and restoring democratic governments.49 In cases such as these, citizens were happy to see the reinstitution of their former governments.

Third, in many cases where leaders were toppled by domestic rebellions rather than foreign powers, the rebels were extremely weak compared to the intervener and were largely destroyed as a result. Restored leaders thus faced little resistance in reestablishing close ties with the intervener. For example, the nineteenth-century interventions by Austria and Prussia in Italian and German states, respectively, followed this pattern.50 More recently, French forces reversed what was essentially a palace coup against Léon Mba in Gabon in 1964 in a matter of hours.51 And French intervention in the Comoros in 1995 had to overcome only Bob Denard's band of two or three dozen mercenaries to restore President Said Mohamed Djohar.52 Because the interveners in these cases were able to crush the rebels, restored leaders were less vulnerable than they had been before. With their internal opposition dismantled, restored leaders are able to retain close ties with the intervener at relatively low risk to their political survival. We base our fourth hypothesis on these observations.53

Hypothesis 4: Restoration FIRC decreases the likelihood of militarized conflict between the intervening and target states.

If restoration FIRCs improve interstate relations and leadership FIRCs harm them, institutional FIRCs fall somewhere in the middle. There are two types of institutional FIRCs that mitigate the Catch-22 faced by foreign-imposed leaders. The first increases the repressive capacity of the state, meaning that the imposed leader can implement the intervener's preferred policies without running a substantial risk of being overthrown by domestic opponents. This variety of institutional FIRC is unique to the Soviet Union. After World War II, Joseph Stalin established Soviet-style police states in the Eastern European satellites that helped stifle revolutionary impulses and enabled local communist elites to follow Moscow's directives at relatively little risk to themselves.54

The second type of institutional FIRC promotes representative institutions that make leaders more responsive to their electorates. The logic of these democratizing institutional FIRCs, however, makes contradictory predictions for intervener-target relations. Grounded in the notion of the democratic peace, one strand of this logic holds that when a democratic intervener establishes democracy in a former rival, subsequent relations between the two states will be peaceful.55 A second strand, however, suggests that by making imposed leaders more responsive to their publics, installing democratic forms of government makes such leaders less responsive to the interveners. If such leaders align themselves too closely to the intervener, voters may replace them with elites more oriented to their own needs and desires.56 Thus, whether institutional FIRCs that promote democracy lead to improved intervener-target relations is unclear.

Democracy-promoting institutional FIRCs, however, often do not succeed. In a study of FIRC and democratization in the twentieth century, Alexander Downes and Jonathan Monten found that only five of thirteen attempts to democratize a foreign government via regime change were successful: Japan (1945), West Germany (1945), Grenada (1983), Panama (1989), and Haiti (1994).57 The cases where democratization was attempted after FIRC but failed—the Dominican Republic (1912, 1914, 1916) and Nicaragua (1910, 1926)—as well as the uncertain democratic trajectories of Afghanistan (2001) and Iraq (2003) suggest that institutional FIRCs are more likely to succeed in wealthy, homogeneous countries that have previous experience with democracy.58 Because hostile elites may come to power after failed institutional FIRCs and conflict with the intervener may result, we suggest the following hypothesis.

Hypothesis 5: Institutional FIRC has no effect on the likelihood of militarized conflict between the intervening and target states.59

overt versus covert fircs. Whereas hypotheses 1–5 apply to successful overt and covert FIRCs alike, we now posit two hypotheses about the differences between overt and covert FIRCs, as well as among different types of covert FIRCs. We argue that whether FIRCs are conducted overtly or covertly influences how the intervener and the target interact following the operation. Specifically, we suggest that covert FIRCs are more harmful to the two states' relationship for five reasons. First, although interveners hope that covert conduct will shield them from the negative consequences of their actions, history suggests that such hopes are unjustified.60 For example, one study of the United States' sixty-three covert Cold War interventions found that the target government publicly accused the United States of trying to overthrow its regime in more than 70 percent of the cases.61 Second, many covert FIRCs are explicitly designed to destabilize the political institutions of the target state to create an opportunity for the intervener's preferred leader to topple the existing regime. The effects of these destabilizing efforts, however, often persist after the covert FIRC attempt, regardless of whether the target regime is overthrown.62 Third, studies of state building and military occupation suggest that an intervening state must make a major commitment of time and resources to successfully construct stable political institutions abroad.63 Covert operations, however, are designed to limit the intervener's role and thus are rarely followed by serious investment of resources. Fourth, the United States undertook no covert restoration FIRCs during the Cold War, which we have argued are the type of FIRC most likely to improve interstate relations.64 Fifth, covert conduct allows states to avoid the so-called Pottery Barn rule: “You break it, you own it.” As U.S. Secretary of State Colin Powell famously explained before the 2003 Iraq War, states that launch overt regime change in the modern era are normatively obligated to protect the target state's population and reconstruct its government.65 Although interveners may feel obligated to uphold the “Pottery Barn rule” after overt FIRCs, they are not obliged to do so after covert FIRCs. Consequently, if the two states find themselves at odds following a covert FIRC, the intervener is more likely to respond militarily than it would if it had intervened overtly. These five points suggest the following hypothesis.

Hypothesis 6: Covert FIRC is more likely to lead to militarized conflict between the intervening and target states than overt FIRC.

successful versus failed covert fircs. Most covert FIRC attempts do not succeed, yet failed attempts may be just as damaging to intervener-target relations as some successful FIRCs. For instance, of the sixty-three U.S.-backed covert FIRCs during the Cold War, twenty-four overthrew the target government (38 percent), and thirty-nine missions (62 percent) failed.66 Because the intervening state is generally unable to plausibly deny its role during covert interventions, unsuccessful covert FIRCs represent the worst outcome for the intervener: not only has it failed to oust an unfriendly regime, but its actions are likely to confirm the target government's worst suspicions about its intentions, thus further exacerbating an already fraying relationship. This logic suggests our final hypothesis.

H7: Failed covert FIRC increases the likelihood of militarized conflict between the intervening and target states.

Research Design

This section introduces our quantitative methodology for assessing the likelihood of militarized conflict between interveners and targets after an overt or covert FIRC.

POPULATION OF CASES

We generated a directed-dyad dataset of all states in the international system from 1816 to 2000.67 Thus, any given dyad-year appears twice: once as state A–state B, and once as state B–state A.68 The result is an enormous number of dyads, many of which (say, Burundi-Bolivia) are highly unlikely to experience an international conflict. To reduce the number of observations to those where FIRC and conflict were realistic possibilities, we therefore limited the sample to politically relevant dyads—those pairs that were either territorially contiguous or that contained a major power.69

DEPENDENT VARIABLE

Our dependent variable consists of whether the intervener or the target initiated a militarized interstate dispute against the other party in the years after FIRC. The advantage of using the MID dataset is that it includes information on interstate conflicts below the level of war.70 Specifically states can exhibit one of five “hostility levels” in a militarized dispute: (1) no militarized action; (2) threat to use force; (3) display of force; (4) use of force; or (5) war. We used a dummy variable that detects whether state A in a dyad initiated a militarized dispute of any severity (greater than 1) in a given year. In our dataset, there are 2,230 instances of states initiating MIDs.71 Because our dependent variable is dichotomous, we employ probit regression with robust standard errors clustered by dyad on the assumption that events within dyads are not independent, but events across dyads are.72

INDEPENDENT VARIABLES

We created fourteen dummy variables to capture the effects of all types of overt and covert FIRC.

overt and covert firc. Downes and Monten define overt foreign-imposed regime change as “the forcible or coerced removal of the effective leader of one state—which remains formally sovereign afterward—by the government of another state.”73 Covert FIRC, following Lindsey O'Rourke, consists of “an operation to replace the leadership of another state, where the intervening state does not acknowledge its role publicly. These actions include assassinating foreign leaders; sponsoring coups d'état; manipulating electoral results; and aiding, funding, and arming dissident groups.”74

In this study, overt FIRCs are successful by definition; we include no cases of attempted overt FIRCs that failed. The principal reason for this omission is practical. Identifying all instances of successful FIRCs in the world over a nearly 200-year period is already an enormous task; locating every failed case would entail an even larger research project. Some cases—particularly those that occur during interstate wars, such as the 1956 Suez War or the Iran-Iraq War—are relatively easy to identify, but many others that are not attempted during wartime are far more difficult to identify.75 We believe that the absence of unsuccessful cases of overt FIRC does not bias our analysis because the effect of these failures is probably the same as that of failed covert cases, which is to increase the likelihood of further conflict by creating a desire for revenge in the target state or further attempts by the intervener to unseat the regime. In other words, including these cases would likely strengthen our finding that most FIRCs do not decrease—or even increase—the likelihood of further conflict.

In contrast, we include both failed and successful covert FIRCs. The primary phenomenon we are interested in measuring is the likelihood that an intervening state and a target state engage in militarized conflict after a successful overt or covert FIRC or a failed covert FIRC. In the case of successful FIRCs, which are easy to observe, we simply code a variable to capture a particular period of time—in this case, ten years—beginning in the year after the FIRC.76 In other words, we treat the effect of FIRC as lasting for ten years.

For covert FIRCs, we code three different variables. The first variable—ongoing covert FIRC—is coded 1 for each year that a covert operation is under way. Attempted covert FIRCs should increase tensions and possibly provoke open conflict between interveners and targets, but we cannot be sure that any conflict coded by the MID dataset is not simply a manifestation of the attempted FIRC itself rather than a consequence of it. The best evidence thus comes from the period after a covert FIRC effort has ended. Here, we create two sets of variables: one set for successful covert FIRCs and a second set for covert FIRCs that failed. As with overt FIRCs, we treat the effects of covert FIRCs as lasting for ten years.

Table 1.

Frequency of Different Types of Foreign-Imposed Regime Change (FIRC)

TypeNumber of Cases
Overt FIRC 95 
Overt leadership FIRC 41 
Overt institutional FIRC 18 
Overt restoration FIRC 36 
Covert FIRC 56 
Covert leadership FIRC 45 
Covert institutional FIRC 
Successful covert FIRC 25 
Successful covert leadership FIRC 19 
Successful covert institutional FIRC 
Failed covert FIRC 32 
Failed covert leadership FIRC 26 
Failed covert Institutional FIRC 
TypeNumber of Cases
Overt FIRC 95 
Overt leadership FIRC 41 
Overt institutional FIRC 18 
Overt restoration FIRC 36 
Covert FIRC 56 
Covert leadership FIRC 45 
Covert institutional FIRC 
Successful covert FIRC 25 
Successful covert leadership FIRC 19 
Successful covert institutional FIRC 
Failed covert FIRC 32 
Failed covert leadership FIRC 26 
Failed covert Institutional FIRC 

NOTE: This table lists the number of each type of FIRC in our dataset. The number of directed dyads containing a FIRC, however, is twice each number, given that each FIRC is coded as occurring twice—once in the state A–state B dyad and once in the state B-state A dyad. These figures do not necessarily signify unique FIRCs, given that many regime changes are carried out multilaterally. Because we were unable to determine whether two covert FIRCs (Guyana, 1966–71; and Grenada, 1979) were leadership or institutional, the total number of covert FIRCs, successful covert FIRCs, and failed covert FIRCs does not equal the sum of the two categories.

restoration, leadership, and institutional fircs. We also differentiate among FIRCs that restored recently displaced leaders to power, place new (usually autocratic) leaders in power but leave political institutions untouched, or install new leaders and seek to change institutions as well. We code dummy variables for overt restoration, leadership, and institutional FIRCs. Because there are no cases of covert restoration FIRCs, we code variables for covert leadership and institutional FIRCs only, as well as for covert leadership and institutional FIRCs that succeeded or failed. Table 1 summarizes the frequency of each type of FIRC that appears in our analysis.77

CONTROL VARIABLES

We include the following control variables identified by previous studies as affecting the likelihood of interstate conflict:78 the material capabilities of both states, measured as each state's share of total global capabilities, known as Composite Index of National Capabilities scores; the share of total dyadic capabilities controlled by state A; the weighted global similarity (S) score for the two states; each state's similarity score with the system leader; the logged distance between the two states' capital cities; whether each state in the dyad is a democracy, as well as an interaction term to identify cases where both states were democratic;79 and a variable that counts the number of years between MIDs, as well as three cubic splines, to control for temporal dependence.80

Statistical Results

The first part of this section presents the findings of our statistical analysis. In the second part of the section, we briefly summarize nine robustness tests.81

FIRC AND MILITARIZED CONFLICT

Figures 13 show pairs of bars indicating the predicted probability of a MID between the intervening and target states in each of the ten years following different types of FIRC (on the right side of each pair) compared to the probability of a MID when no FIRC occurred (on the left), along with the 95 percent confidence intervals for each estimate.82 For example, the first pair of bars in figure 1 shows that the chance of a MID occurring after an overt FIRC is 0.0057, whereas the chance of one occurring absent a FIRC is 0.0062. The fully overlapping confidence intervals indicate that this minor difference is not statistically significant. In other words, one state inflicting an overt FIRC on another does not discernably reduce the likelihood that the two states will experience a militarized conflict in the ensuing ten years. This result provides evidence against hypothesis 1, which states that FIRC decreases the probability of intervener-target conflict, and supports hypothesis 2, which argues that FIRC has no effect on the likelihood of such conflict. The second pair of bars in figure 2, which shows that covert FIRCs significantly increase the likelihood of a MID in the ensuing ten years, also provides strong support for hypothesis 2 and none for hypothesis 1.83 A comparison of these two results also suggests that hypothesis 6—that covert FIRCs damage intervener-target relations more than overt ones—is correct.84

Figure 1.

Overt Foreign-Imposed Regime Change (FIRC) and the Probability of Militarized Interstate Disputes (MIDs)

Figure 1.

Overt Foreign-Imposed Regime Change (FIRC) and the Probability of Militarized Interstate Disputes (MIDs)

Figure 2.

Overt Foreign-Imposed Regime Change (FIRC) and the Probability of Militarized Interstate Disputes (MIDs)

Figure 2.

Overt Foreign-Imposed Regime Change (FIRC) and the Probability of Militarized Interstate Disputes (MIDs)

Figure 3.

Successful and Failed Covert Foreign-Imposed Regime Change (FIRC) and the Probability of Militarized Interstate Disputes (MIDs)

Figure 3.

Successful and Failed Covert Foreign-Imposed Regime Change (FIRC) and the Probability of Militarized Interstate Disputes (MIDs)

The quantitative evidence also supports our hypotheses about the effects of different types of FIRC on the probability of a MID. Columns 2–4 in figure 1 and columns 3–4 in figure 2, for example, show that overt and covert leadership FIRCs significantly increase the likelihood of a militarized conflict between the intervening and target state during the ten-year period following intervention.85 Overt leadership FIRCs more than double the likelihood of a MID, whereas covert leadership FIRCs nearly triple the likelihood of militarized conflict. Both of these results support hypothesis 3. In sharp contrast, in accordance with hypothesis 4, restoration FIRCs slash the probability of conflict by almost 90 percent. The effect of institutional FIRC varies depending on whether it is overt or covert. Overt institutional FIRC reduces the chances of a MID about 60 percent, but the uncertainty around this estimate is very broad (produced by a large standard error relative to the coefficient), and it lacks statistical significance. Covert institutional FIRCs, on the other hand, increase the chances of conflict about 250 percent, but the significance of this result is not robust to varying treatment lengths. In short, although overt and covert institutional FIRCs have opposite effects, as predicted by hypothesis 5, neither exerts a reliable effect on conflict.

The evidence for hypotheses 3 and 5 is more mixed when we distinguish between successful and failed covert leadership and institutional FIRCs. Recall that we argued above that successful covert leadership and institutional FIRCs should have the same effects as overt leadership and institutional FIRCs. Column 2 in figure 3 shows that this is not the case: successful covert leadership FIRC increases the likelihood of a MID by nearly 50 percent, but this effect is not significant. Meanwhile, no successful covert institutional FIRC is followed by any type of conflict. The latter divergence is easily explained. For one, there are only five cases of successful covert institutional FIRC.86 Moreover, because all covert FIRCs in our dataset were carried out by the United States, institutional FIRC is synonymous with attempts to democratize the target. All of these FIRCs happened to result in consolidated democracies; thus the likelihood that the United States would engage in militarized conflict with these states should be low.87

The absence of a significant result for successful covert leadership FIRC is more puzzling, but closer examination helps solve the puzzle. For example, several of these operations were carried out in close U.S. allies—such as France, Italy, and Japan—to help defeat Leftist parties in democratic elections during the early Cold War.88 Because of the multitude of factors driving these states into close alliances with the United States during this period, we would expect the likelihood of conflict to be muted. Similarly, in South Vietnam—the site of two successful covert leadership FIRCs (1963 and 1967)—the United States was allied with Saigon in fighting communist subversion, which made conflict between the two states unlikely. Dropping these cases increases the magnitude of the coefficient for successful covert leadership FIRCs, but does not make it significant. Extending the treatment effect to all remaining years in the dyad, however, produces a highly significant positive effect (p < 0.001) without dropping any cases, indicating that successful covert leadership FIRCs produce stability in the short run but backfire in the longer term.

Finally, we turn to the effects of successful versus failed covert FIRCs. Columns 1 and 3 in figure 3 show that, in accordance with hypothesis 7, failed covert FIRCs are at least as bad for intervener-target relations as successful covert FIRCs. Although successful covert FIRCs increase the likelihood of a MID by about 20 percent, this effect is not significant. Failed covert FIRCs, by contrast, result in a more than fourfold jump in the probability of conflict. This difference, however, disappears over time: when the treatment effect is increased to all remaining dyad years, successful and failed covert FIRCs are both positive and significant, and the magnitude of their effects is indistinguishable.89 It is thus unclear whether failed covert FIRCs have a larger effect on the probability of conflict than successful ones. What is clear, though, is that all types of failed covert FIRCs make MIDs significantly more likely (columns 3–5 in figure 3). As shown, although the effect of failed covert institutional FIRC is marginally larger than for failed covert leadership FIRC, there is no significant difference between the two. These results strongly validate hypothesis 7.

ROBUSTNESS TESTS

Although these results provide support for our theoretical expectations, it is possible that they are sensitive to minor changes in variable coding, model specification, the type of model used, or other factors. To probe the stability of our findings, we subjected them to nine robustness tests, none of which resulted in appreciable changes to our findings, except as already discussed:90 varying the length of the treatment effect of FIRC (shortening it to five years and lengthening it to include all remaining years in the dyad); recoding institutional FIRCs to include only cases in which interveners attempted to democratize targets; including a variable for whether the intervener and target were joined in a formal alliance;91 employing alternative versions of the dependent variable (MIDs that escalated to the use of force and MIDs that resulted in war); limiting our analysis of covert FIRCs to dyads including the United States;92 employing methods (spatial lags and cluster robust variance estimation) to correct for potential nonindependence across (as opposed to within) dyads;93 reestimating the models with rare events logit;94 reestimating the models with fixed effects to examine the effect of FIRC within dyads that experienced FIRC as opposed to between dyads that did and did not experience FIRC;95 and using genetic matching to check whether interveners select targets for different types of FIRC with which they are already likely or unlikely to fight.96

In sum, neither overt nor covert FIRCs generally reduce conflict between interveners and targets. Overt regime change improves the prospects for peace only when interveners restore leaders who were already closely aligned with them but were recently overthrown. By contrast, overt FIRCs that change targets' political institutions exert little effect on the probability of intervener-target conflict, whereas those that change only leaders significantly increase the likelihood of such disputes.

Similarly, covert FIRCs either fail to decrease or significantly increase the likelihood of intervener-target conflict. The effect of covert FIRC hinges principally on whether the operation succeeds or fails in displacing the targeted leader. Successful covert leadership FIRCs have little effect on the likelihood of conflict in the immediate aftermath, but significantly worsen relations over the long term. Failed covert leadership or institutional FIRCs are each highly detrimental to intervener-target relations in the short and long terms.

Foreign-Imposed Regime Change and War in Zaire/DRC

Quantitative methods are useful for identifying correlations between variables across a large number of cases. These methods are less adept, however, at validating causal mechanisms by demonstrating that the outcome of any particular case occurred for the reasons identified by the theory. Case studies employing process-tracing methods are well suited to this task. To probe the causal logic of our theory, therefore, we present a case study of the events that followed Rwanda's 1997 leadership FIRC in Zaire, which overthrew Mobutu Sese Seko and replaced him with Laurent-Désiré Kabila.

Four factors guided our case selection strategy. First, because we could not do justice to the many types of foreign-imposed regime change included in our statistical analysis, we chose to focus on the type that not only is the most common but also has the most deleterious consequences for intervener-target relations: leadership FIRC.97 Second, we chose a case that was followed by militarized conflict to allow us to observe whether our proposed causal mechanisms contributed decisively to the outbreak of the conflict. Third, despite the obvious policy relevance of recent FIRCs carried out by the United States and its allies, we selected a case involving two minor powers to highlight the reality that FIRC is a phenomenon not only of U.S. foreign policy. Fourth, the Rwandan FIRC against Zaire—although initial appearances suggest otherwise—is a fairly typical case of a stronger intervener overthrowing a weaker target.98 By the time of Rwanda's invasion, the Rwandan Patriotic Army (RPA) and Zaire's army were roughly the same size, but the former was an experienced, battle-hardened force whereas the latter had by 1996 “been reduced to a mockery of itself.”99 In terms of military power, therefore, Rwanda possessed a considerable advantage over Zaire.100

THE OVERTHROW OF MOBUTU

The origins of the Rwandan-led FIRC in what was known as Zaire in 1997 lie in the Rwandan genocide. Beginning in April 1994, Hutu extremists organized the mass killing of between 500,000 and 800,000 Tutsi (and some moderate Hutu) to avert the implementation of a power-sharing agreement (the Arusha accords) negotiated between the Hutu government of President Juvénal Habyarimana and the Tutsi rebel group, the Rwandan Patriotic Front (RPF). In response to the killings, the RPF attacked and eventually defeated the Rwandan army (Forces Armées Rwandaises, or FAR), ending the genocide.101 As they retreated, the FAR and allied Hutu militia, the interahamwe, forced masses of Hutu civilians to flee alongside them. An enormous exodus from Rwanda took place in July and August 1994: roughly 2 million Hutu fled the country for neighboring states. Between 1.1 and 1.5 million of them ended up in North and South Kivu Provinces, in eastern Zaire, directly bordering Rwanda.102

Members of the ex-FAR and the interahamwe quickly took control of the refugee camps and began to launch armed incursions back into Rwanda as early as October 1994.103 Evidence of the complicity of Zairian President Mobutu and his government—a long-standing ally of the Habyarimana regime—in supporting these armed Hutu elements is overwhelming.104 As Human Rights Watch concluded, “Zairian forces close to president Mobutu Sese Seko have played a pivotal role in facilitating the re-emergence as a powerful military force of those directly implicated in the Rwandan genocide.”105 Unsurprisingly, the RPF regime perceived the presence of the ex-FAR and other Hutu militants in the camps in eastern Zaire as representing an intolerable threat to its security. Filip Reyntjens summarizes the view from Kigali: “The main reason why Uganda, Rwanda and—to a lesser extent—Burundi intervened in the Congo in the Autumn of 1996 was related to their security…. For Rwanda in particular, the presence of massive Hutu refugee camps, which housed hostile armed elements, close to the border constituted a major threat.”106

Using the cover of an ethnic Tutsi rebellion in South Kivu that Rwanda itself had helped foment, the RPA invaded both North and South Kivu in October 1996.107 Refugees fled in all directions, but many converged on the Mugunga camp outside Goma in North Kivu, which held approximately 1 million people by the time Tutsi forces assaulted it on November 13. Within days, hundreds of thousands of refugees began to cross back into Rwanda, while hundreds of thousands of others fled westward, deeper into Zaire.108

Only after the invasion was under way did Laurent Kabila emerge as the leader of the rebel forces. On October 18, the Alliance des Forces Démocratiques pour la Libération du Congo-Zaïre (Alliance of Democratic Forces for the Liberation of Congo-Zaire, AFDL) was formally created from a combination of four opposition groups, none of which had a strong presence in Zaire.109 Kabila had nothing to do with the launching of the uprising or the Rwandan invasion; rather, he was appointed spokesman of the AFDL by the invaders to provide a “Zairian face” for the rebellion.110 To counter Kinshasa's narrative that the war in the Kivus was the result of external aggression, “it was necessary,” according to Reyntjens, “to exhibit leadership, the ‘Zaireness' of which could not be challenged.”111 Gérard Prunier agrees that in the fall of 1996 Kabila “had just become the new local cover for the Rwandese attack on Zaire, in charge of making a foreign invasion look like a national rebellion.”112 It also helped that Kabila was not a Tutsi, which contributed to countering the impression of the rebels as Tutsi dominated.113

It is difficult to determine exactly when the Rwandan leadership—dominated by RPF leader Paul Kagame—decided on regime change in Zaire. Some evidence suggests that it was an objective before the war began.114 Other evidence, however, indicates that the goal developed during the war in response to events on the ground, particularly the ease and speed with which anti-Mobutu forces were able to advance.115 Regardless of when the decision to remove Mobutu was made, it is clear that Rwanda—supported by Angola, Burundi, and Uganda—was the key driver of regime change in Zaire.116 Rather than fight a final battle in the capital, Mobutu fled Kinshasa shortly before the victorious AFDL entered the city on May 17, 1997. Kabila was sworn in as president on May 29 and promptly renamed the country the Democratic Republic of Congo.

REGIME CHANGE IN KINSHASA: NO MAGIC BULLET

The removal of Mobutu and the installation of Kabila was a clear instance of a leadership FIRC. The foreign powers that put Kabila in office made no attempt to build institutions to create support for the country's new leader. Left largely to his own devices, Kabila floundered.117 “Kabila seems to have had only the vaguest notions of what he actually intended to do after overthrowing Mobutu,” writes Prunier. “Everything seemed arbitrary,” and the government was characterized by an “almost complete lack of institutionalization.”118 After a few months in power, Kabila banned all political parties and even purged the AFDL.119 A new constitution issued on May 28, 1997, writes David Van Reybrouck, “essentially placed all power in the hands of the president.”120 The new government was highly authoritarian, repressive, centralized, and personalist, exhibiting many similarities to Mobutu's regime. As Thomas Turner observes, “The ouster of Mobutu created expectations of improvement but Kabila soon revealed himself to be a second Mobutu.”121

Unfortunately for Rwanda, replacing Mobutu with “what looked like the perfect puppet regime” was insufficient to resolve the issues that led to the conflict in the first place, because the Rwandans could not induce their puppet to follow their instructions.122 As noted above, Rwanda had intervened in Zaire to solve its security problem by destroying the refugee camps and eliminating the support base of the ex-FAR and other Hutu militants raiding western Rwanda. As a secondary goal, Kigali aimed to protect the Banyamulenge and other ethnic Tutsis in eastern Zaire from government depredations and attacks by other ethnic groups.

Rwanda's invasion of Zaire, however, failed to improve the security situation and worsened the position of the Banyamulenge by generating resentment against them among other Congolese ethnic groups. Although the RPA succeeded in disbanding the refugee camps in the Kivus, ex-FAR and interahamwe militiamen soon resumed cross-border attacks into Rwanda.123 Indeed, remnants of the ex-FAR and interahamwe came together in late 1996 or early 1997 to form a new Hutu rebel group, the Armée pour la Libération du Rwanda (Army for the Liberation of Rwanda).124 Hutu militants benefited from resentment of the Rwandan and Tutsi presence in Congo by non-Tutsi elements of the new Congolese Army (Forces Armées Congolaises), which allowed the armed Hutu groups to move freely, sometimes even accompanying them to the Rwandan border as they launched deadly raids.125 In addition to cross-border attacks, between 10,000 and 15,000 Hutu fighters had filtered back into Rwanda with the returning refugees. One analyst concludes that, by the end of 1997, “These insurgents sparked the worst fighting the country had seen since the genocide.”126 “Clearly,” concludes Emizet Kisangani, “the issues at stake under Mobutu were not solved during the takeover by Kabila, which presaged another conflict.”127

The Rwandan presence in the DRC was also self-defeating because it generated powerful resentments among the non-Tutsi population against all Tutsis, regardless of whether they were Rwandan or Congolese.128 According to Kisangani, “Many local authorities were killed in the Kivus by the Banyamulenge and Tutsi, and those who escaped the ethnic cleansing fled. These changes in leadership caused deep resentment when the Congolese saw people of Rwandan origin, whom they considered to be foreigners, claim supremacy in Congolese high offices.”129 Reyntjens concurs: “It is not surprising under these circumstances that the anti-Tutsi feelings rapidly became widespread.”130 Indeed, resentment of Tutsi domination and the perception of non-Tutsis in the Kivus that “they were under foreign occupation and their country was being run from Rwanda” sparked the Mai Mai rebellion in July 1997, which killed 3,000 people.131

KABILA'S DILEMMA

Kabila's main problem was that he lacked an independent power base.132 As an exile who had come to power at the point of foreign bayonets, Kabila had to rely on Rwandan and Congolese Tutsi for support, placing him in a Catch-22: his dependence on outside backers was a domestic political liability that undercut his legitimacy and endangered his rule. At the same time, attempting to distance himself from Rwanda risked angering his sponsors in Kigali and raised the possibility that they would seek to replace him, as they had Mobutu.

Kabila's heavy reliance on foreign support to consolidate power in his early days as president of the DRC was obvious. Just as the Tutsi had dominated the AFDL, many foreign Tutsis occupied prominent roles in the new regime.133 Most notable was the new chief of staff of the armed forces, James Kabarebe, who had previously been the commander of the Rwandan Republican Guard. Even Kabila's bodyguards were reputed to be Rwandans.134 “Overall,” concludes François Ngolet, “between October 1996 and January 1998 the Tutsi presence on Kabila's staff was overwhelming.”135

There is also ample evidence that his fellow Congolese viewed Kabila as a Rwandan puppet. As Filip Reyntjens writes, “As soon as he [Kabila] assumed power, he was faced with a serious dilemma. Already during the rebellion, it was clear that his own military and political base was thin and that external forces … carried him to power. Although Kabila was initially well received … this dependency soon became a mortgage in terms of internal legitimacy. The continuing and highly visible presence of foreign troops and officers, particularly those of the RPA, raised accusations that Kabila was but a puppet of Rwanda and, to a lesser extent Uganda.”136 In polls taken in the summer and fall of 1997, a majority of Congolese agreed that Kabila was under the influence of foreigners and that Uganda and Rwanda were colonizing the Congo.137 David Van Reybrouck summarizes Congolese sentiments: “As long as Kabila's court was filled with those hateful foreigners, he could forget about his authority being recognized.”138 The perception that Kabila was a Rwandan puppet had generated anti-Tutsi violence and undermined his regime's legitimacy.

KABILA MAKES A CHOICE

Lacking domestic supporters and legitimacy, many foreign-imposed leaders have little alternative but to comply with the demands of their foreign sponsors, which is one reason why so many FIRCs are followed by civil wars: disgruntled domestic groups resentful of external influence launch rebellions to displace the leader and terminate foreign meddling. As Prunier puts it in the Congolese context, “Did the local population agree to be ruled by a government in Kinshasa not really independent but largely in the hands of a foreign state, that is, Rwanda, which was highly suspected of harboring expansionist views?”139 Kabila faced this situation in mid-1997 when Mai Mai militias sprang up and attacked ethnic Tutsis out of resentment of the latter's newfound prominence.140 Indeed, the situation in the Kivus continued to deteriorate in late 1997 and early 1998, forcing Kabila to travel to Bukavu in January to give a speech in which he railed against the Mai Mai and defended the Banyamulenge. The speech, however, backfired: “Supporting them [the Banyamulenge] in that way,” writes Prunier, “did more harm than good because Kabila looked like a puppet for the Kigali ventriloquists.”141

To neutralize this domestic threat, Kabila sought to appease his local critics by reducing the influence of Tutsis in his government and distancing himself from Rwanda. Over the course of several months, Kabila shifted the balance of power inside his regime away from Tutsis and in favor of officials from his native region of Katanga.142 The government then tried to increase its control over the Tutsis in the army by dispersing its Banyamulenge members across different units throughout the country.143 “In effect,” writes Kevin Dunn, “Kabila began creating a new cabinet and regional governments that ran counter to the team that put him in power.”144

In the first half of 1998, it was becoming increasingly clear in Kigali and Kampala that a change of leadership might be necessary. Little has been written about the deliberations of the Rwandan and Ugandan governments, but one source reports that “the intelligence chiefs of Angola, Rwanda, and Uganda held discussions regarding the desirability of finding an alternative leader for the DRC as early as January 1998.”145 Suspicious that Rwanda and its Congolese Tutsi allies were planning a coup, Kabila sacked James Kabarebe, the Rwandan officer serving as army chief of staff, and ordered the expulsion of all Rwandan soldiers from the DRC in late July.146 On August 2, a new war erupted.

ANOTHER WAR FOR REGIME CHANGE

Kabila's removal of Kabarebe and his ejection of Rwandan troops from the DRC acted as the immediate cause for Rwanda's initiation of the Second Congo War to overthrow its wayward puppet in Kinshasa.147 According to Turner, “‘Africa's world war’ began as a Rwandan attempt to overthrow Kabila.” “The second war,” Turner continues, “was the direct result of Laurent Kabila's expulsion of his Rwandan handlers.”148 Rwandan leaders were not about to accept the loss of influence in Kinshasa that Kabila's reforms implied, especially because they were facing the same security problems they had in 1996, when they decided to overthrow Mobutu. In Kigali's calculations, a friendly, pliable leader in the DRC was required to help contain and eventually eliminate the Hutu militant threat to Rwanda's Tutsi population and to the RPF's continued rule.149 As Georges Nzongola-Ntalaja puts it, “Since Kabila had not lived up to their expectations, Rwanda and Uganda were determined to find a new Congolese puppet.”150

Rwanda's initial plan for removing Kabila apparently consisted of a coup by Rwandan and Congolese Tutsi officers. When this plan failed, a new plan emerged that largely resembled the one used to oust Mobutu: within days of the Rwandan invasion, a new Congolese rebel group—the Rassemblement Congolais pour la Démocratie (Rally for Congolese Democracy)—was cobbled together to put a “Congolese face” on what was essentially another foreign invasion. The members of this coalition, writes Prunier, “were a strange mix of former Mobutists … radical left-wingers … regional barons … UN and NGO figures … and well-known representatives of Rwandese interests.” This “hodgepodge organization” was “so obviously incapable of military or even political autonomy that the whole thing looked more like an invasion than a genuine Congolese uprising.”151 The second war, however, devolved into a bloody and protracted stalemate, as some of Rwanda's allies from the first war—including Angola and Zimbabwe—and a few newcomers—such as Chad, Libya, Namibia, and Sudan—intervened on Kabila's behalf.152 Rwanda's renewed intervention also pushed domestic actors inside the DRC—notably, the Mai Mai militias, ex-FAR, and interahamwe—into Kabila's arms.153 The conflict soon transformed into a war of plunder, as each belligerent exploited the Congolese territory it controlled. Kabila survived the immediate onslaught only to be assassinated by one of his bodyguards in 2001; he was succeeded by his son, Joseph Kabila. Although the interstate phase of the war officially ended in 2002, violence in eastern DRC has continued with an ever-changing constellation of actors.

Conclusion

Contrary to policymakers' expectations and prior scholarship, overt and covert FIRCs do not improve relations between intervening and target states. Often they become worse. An intervener's primary reason for installing a new leader in another state is to get that state to behave in the intervener's interests. Changing another state's behavior is not as simple, however, as changing its leader. Newly installed leaders may intend to pursue policies that promote the intervener's interests, but find themselves constrained by powerful domestic groups or by enduring national interests. Acting contrary to these interests imperils imposed leaders' political survival. Interveners may thus encounter the same problems with the new leader that they had with the old one. Even if interveners manage to find a compliant individual willing to implement their priorities, leaders who act in the intervener's interest at the expense of their own constituents tend to generate domestic opposition. This is one reason why most foreign-imposed leaders are deposed by force. Leaders who come to power after overthrowing a foreign-imposed leader are usually hostile to the intervener, which can lead to conflict between the two states. Exceptions include FIRCs that restore previous leaders to power who already enjoyed close relations with the intervener, and those that seek to install new (usually democratic) institutions in target states. Restoration FIRCs reduce the frequency of intervener-target conflict, whereas institutional FIRCs have no effect because the successful cases of democracy promotion are balanced out by those where democracy fails to take root.

FIRCs that fail, which in our study include only covert FIRCs by the United States, have if anything a more negative effect on intervener-target relations. These operations, which invariably are discovered by the target, make bad relations worse and can lead to overt attempts at regime change.

Our analysis of FIRC yields findings different from those of earlier studies. For example, we find that overt and covert leadership FIRCs increase the likelihood of MIDs between the intervening and target state. Institutional FIRCs (both overt and covert) mostly have no significant effect on conflict, whereas overt restoration FIRCs decrease the probability of conflict. Covert FIRCs tend to have a larger effect on conflict than overt FIRCs, whereas failed covert FIRCs lead to conflict at least as often—and possibly more often—than successful ones.

These findings reflect an almost unremittingly negative assessment of FIRC. Among covert FIRCs, the majority of all operations fail, and failed operations significantly increase the likelihood of conflict. Successful operations, which are relatively rare, do not reduce conflict. Among overt FIRCs, the most common type—leadership FIRC—is again the type that worsens relations the most. Not even promoting democracy in the target guarantees a significant decrease in conflict. Only by restoring governments with which they already had good relations can interveners improve their subsequent relations.

The implication of these findings is that states ought to exercise caution when debating whether to launch a covert or overt FIRC. The Barack Obama administration's recent regime change operations in Libya and Syria illustrate these problems. Despite efforts to limit the U.S. role in each of these missions, both interventions have followed a now familiar script: initial enthusiasm that a regime change could improve U.S. relations with the target country, growing disillusionment with the opposition forces supported by the United States, and, finally, unanticipated negative consequences as a result of the operation.

In March 2011, Secretary of State Hillary Clinton met with Mahmoud Jibril and other Libyan opposition leaders associated with the Transitional National Council to assess their feasibility as a democratic alternative to President Muammar Gadaffi. During these meetings, Assistant Secretary of State for European and Eurasian Affairs Philip Gordon recalls that opposition leaders “said all of the right things about supporting democracy and inclusivity and building Libyan institutions, providing some hope that we might be able to pull this off. They gave us what we wanted to hear.”154 With NATO air support and covert aid, the rebels succeeded in killing Gaddafi and ousting his supporters. Jibril and the Transitional National Council, however, proved incapable of unifying the country. Facing growing resistance from Islamist opponents, Jibril took steps to distance himself from his Western backers and later rejected his “secular” label in an attempt to garner wider domestic support.155 These efforts were in vain. Instead, Libya—which had been governed by a strongman for forty-two years and had no previous experience with democracy—descended into a chaotic civil war. A year after Gaddafi was killed, Islamist militants attacked a U.S. compound in Benghazi and murdered the U.S. ambassador. In December 2012, James Risen reported that U.S.-supplied weapons, originally intended to help moderate rebels topple Gaddafi, had since fallen into the hands of anti-American extremists inside Libya.156 Subsequent reports found similar weapons “turning up in Syria, Tunisia, Algeria, Mali, Niger, Chad, Nigeria, Somalia, Sudan, Egypt and Gaza, often in the hands of terrorists, insurgents or criminals.”157 President Obama has described “failing to plan for the day after” Gaddafi's ouster as the “worst mistake” of his presidency.158 Thus far, the president has resisted calls from members of his administration—including Secretary of Defense Ashton Carter and Chairman of the Joint Chiefs of Staff General Joseph Dunford—to intervene directly in Libya.159 But if the Islamists prevail in the civil war, relations with Washington will worsen, and Libya may again find itself in the crosshairs of the United States.

The Obama administration's support for covert regime change in Syria also has all the hallmarks of a risky policy that is unlikely to result in improved relations with Damascus even if President Bashar al-Assad is eventually ousted. Proponents of regime change argued that overthrowing Assad would diminish Iranian influence in the region, end the ongoing humanitarian disaster, prevent Sunni extremist groups from seizing power, and provide a show of U.S. force that could dissuade other potentially aggressive states from challenging American interests.160 Toward this end, in 2013 the United States initiated a covert program run by the Central Intelligence Agency, code-named Timber Sycamore, whose members worked with intelligence agencies in several Middle Eastern states—most notably, Saudi Arabia—to arm and train “moderate” opposition fighters associated with the Free Syrian Army (FSA).161 From its inception, however, the covert program was hamstrung by an inability to identify reliable opposition forces and prevent U.S.-made weapons from falling into the hands of extremist groups.162 In mid-2013, for instance, it was reported that roughly one-quarter of the FSA's fighters had switched allegiances to an al-Qaida-affiliated group, Jabhat al-Nusra.163 Seven Islamist groups subsequently left the moderate Supreme Military Council associated with the FSA to form a separate Islamic Front, taking 40,000 to 50,000 fighters with them.164

Following the dramatic rise of the Islamic State of Iraq and the Levant (ISIL), U.S. policymakers created a separate Pentagon-run covert program to train Syrian rebels to defeat ISIL, which Washington views as a bigger threat than the Assad regime. Again, however, the U.S.-backed rebels struggled to achieve their objectives. President Obama explained, “I'm the first one to acknowledge that it has not worked the way it was supposed to … when we tried to get them to just focus on ISIL, the response we'd get back is, how can we focus on ISIL when every single day we're having barrel bombs and attacks from the regime?”165 In October 2015, the administration abandoned its $500 million program to covertly train thousands of new opposition fighters, acknowledging that the program only had “four or five” trainees actually fighting within Syria.166 Since then, the administration has expanded its program to defeat ISIL and oust Assad by covertly bolstering the strength of existing Syrian opposition forces and overtly targeting ISIL using U.S.-led air strikes and Special Operation forces.167 The stage is thus set for a prolonged, multifront conflict. The United States could escalate its involvement to seek overt regime change in Syria, but after Afghanistan, Iraq, and Libya, the wisdom of such a policy is highly doubtful.

Acknowledgments

The authors' names are listed in alphabetical order, and each contributed equally to this article. Previous drafts were presented at the Macmillan International Relations Seminar at Yale University, the Security Studies Colloquium at Princeton University, the Dickey Center Post-doctoral Workshop at Dartmouth College, and the 2015 International Studies Association Annual Convention. The authors gratefully acknowledge helpful comments from audience members at these presentations, as well as from Jeffrey Friedman, William Wohlforth, and the anonymous reviewers. For excellent research assistance, the authors thank Amber Diaz, Caitlin Gorback, Alexander Gorin, Michael Joseph, Julia Macdonald, and Paul Zachary. Downes acknowledges support from the Office of Naval Research, U.S. Department of the Navy, Grant No. N00014-09-1-0557. Any opinions, findings, or recommendations expressed in this article are those of the authors and do not necessarily reflect the views of the Office of Naval Research. Replicated data and supplemental materials can be found online at doi:10.7910/DVN/7Y4TD8.

Notes

1. 

Alexander B. Downes and Jonathan Monten, “Forced to Be Free? Why Foreign-Imposed Regime Change Rarely Leads to Democratization,” International Security, Vol. 37, No. 4 (Spring 2013), pp. 90–131.

2. 

Lindsey O'Rourke, “Secrecy and Security: U.S.-Orchestrated Regime Change during the Cold War,” Ph.D. dissertation, University of Chicago, 2013.

3. 

See, for example, Goran Peic and Dan Reiter, “Foreign-Imposed Regime Change, State Power, and Civil War Onset, 1920–2004,” British Journal of Political Science, Vol. 41, No. 3 (July 2011), pp. 453–475; Downes and Monten, “Forced to Be Free”; Alexander B. Downes, “Catastrophic Success: Foreign-Imposed Regime Change and Civil War,” George Washington University, 2016; Andrew J. Enterline and J. Michael Greig, “Perfect Storms? Political Instability in Imposed Polities and the Future of Iraq and Afghanistan,” Journal of Conflict Resolution, Vol. 52, No. 6 (December 2008), pp. 880–915; Andrew J. Enterline and J. Michael Greig, “Against All Odds? The History of Imposed Democracy and the Future of Iraq and Afghanistan,” Foreign Policy Analysis, Vol. 4, No. 4 (October 2008), pp. 321–347; David M. Edelstein, Occupational Hazards: Success and Failure in Military Occupations (Ithaca, N.Y.: Cornell University Press, 2008); Bruce Bueno de Mesquita and George W. Downs, “Intervention and Democracy,” International Organization, Vol. 60, No. 3 (Summer 2006), pp. 627–649; Jeffrey Pickering and Mark Peceny, “Forging Democracy at Gunpoint,” International Studies Quarterly, Vol. 50, No. 3 (September 2006), pp. 539–560; Daniel Berger et al., “Do Superpower Interventions Have Short and Long Term Consequences for Democracy?” Journal of Comparative Economics, Vol. 41, No. 1 (February 2013), pp. 22–34; and James Dobbins et al., America's Role in Nation-Building: From Germany to Iraq (Santa Monica, Calif.: RAND Corporation, 2003).

4. 

Nigel Lo, Barry Hashimoto, and Dan Reiter, “Ensuring Peace: Foreign-Imposed Regime Change and Post-War Peace Duration, 1914–2001,” International Organization, Vol. 62, No. 4 (Fall 2008), pp. 717–736.

5. 

Other studies that articulate a version of this logic include Bueno de Mesquita and Downs, “Intervention and Democracy”; Dan Reiter, How Wars End (Princeton, N.J.: Princeton University Press, 2009); and O'Rourke, “Secrecy and Security.”

6. 

Alexander B. Downes, “Decapitation by FIRC: Foreign-Imposed Regime Change and the Fate of Leaders,” paper presented at the International Studies Association Annual Convention, San Diego, California, April 1–4, 2012, p. 30; and O'Rourke, “Secrecy and Security.”

7. 

Institutional FIRCs can promote democratic or repressive institutions. In our sample, the promotion of democracy is carried out exclusively by democracies; the promotion of repressive institutions is carried out by single-party regimes, particularly the Soviet Union.

8. 

Democracies are also less likely than autocracies to experience coups, meaning that a democratic target backsliding into an autocracy is unlikely to provoke conflict with an intervener. See Clayton L. Thyne and Jonathan M. Powell, “Coup d'État or Coup d'Autocracy? How Coups Impact Democratization, 1950–2008,” Foreign Policy Analysis, Vol. 12, No. 2 (April 2016), pp. 192–213; and Nancy Bermeo, “On Democratic Backsliding,” Journal of Democracy, Vol. 27, No. 1 (January 2016), pp. 5–19.

9. 

Bueno de Mesquita and Downs, “Intervention and Democracy.” On the difficulty of obtaining reliable client states through regime change whether or not it promotes democracy, see Lindsey A. O'Rourke and Alexander B. Downes, “Picking Your Friends: Foreign-Imposed Regime Change and the Quality of Interstate Relations,” Boston College and George Washington University, 2016.

10. 

Of U.S. covert regime-change operations during the Cold War, thirty-nine out of sixty-three (62 percent) were unsuccessful. See O'Rourke, “Secrecy and Security,” p. 300. Although overt regime-change efforts also sometimes fail, we exclude these cases owing to the substantial additional research effort that would be involved in identifying them. We discuss this omission further—and explain why we do not believe it biases our results—in the research design section below.

11. 

Studies suggest that states are more willing to conduct certain types of FIRC covertly—such as missions targeting democratic states, domestically controversial operations, and operations targeting powerful states. See Stephen Van Evera, “The Case against Intervention,” Atlantic Monthly, July 1990, pp. 72–80; Patrick James and Glenn E. Mitchell II, “Targets of Covert Pressure: The Hidden Victims of the Democratic Peace,” International Interactions, Vol. 21, No. 1 (July 1995), pp. 85–107; Sebastian Rosato, “The Flawed Logic of Democratic Peace Theory,” American Political Science Review, Vol. 97, No. 4 (November 2003), pp. 585–602; Alexander B. Downes and Mary Lauren Lilley, “Overt Peace, Covert War? Covert Intervention and the Democratic Peace,” Security Studies, Vol. 19, No. 2 (Summer 2010), pp. 266–306; and O'Rourke, “Secrecy and Security.”

12. 

MIDs are a more sensitive barometer of interstate relations than war recurrence, the variable used in previous studies. See Lo, Hashimoto, and Reiter, “Ensuring Peace.”

13. 

Joshua Muravchik, Exporting Democracy: Fulfilling America's Destiny (Washington, D.C.: American Enterprise Institute, 1992); Charles Krauthammer, “Democratic Realism: An American Foreign Policy for a Unipolar World,” 2004 Irving Kristol Lecture, American Enterprise Institute, Washington, D.C., February 10, 2004; Condoleezza Rice, “The Promise of Democratic Peace: Why Promoting Freedom Is the Only Realistic Path to Security,” Washington Post, December 11, 2005; and Nancy Bermeo, “Armed Conflict and the Durability of Electoral Democracy,” in Elizabeth Kier and Ronald R. Krebs, eds., In War's Wake: International Conflict and the Fate of Liberal Democracy (Cambridge: Cambridge University Press, 2010), pp. 67–94.

14. 

Bueno de Mesquita and Downs, “Intervention and Democracy”; Pickering and Peceny, “Forging Democracy at Gunpoint,” pp. 549–555; Margaret G. Hermann and Charles W. Kegley Jr., “The U.S. Use of Military Intervention to Promote Democracy: Evaluating the Record,” International Interactions, Vol. 24, No. 2 (June 1998), p. 97; Nils Gleditsch, Lene Siljeholm Christiansen, and Håvard Hegre, “Democratic Jihad? Military Intervention and Democracy,” World Bank Policy Research Working Paper 4242 (Washington, D.C.: World Bank, June 2007), p. 39; and Arthur A. Goldsmith, “Making the World Safe for Partial Democracy? Questioning the Premises of Democracy Promotion,” International Security, Vol. 33, No. 2 (Fall 2008), pp. 120–147.

15. 

Hermann and Kegley, “The U.S. Use of Military Intervention to Promote Democracy,” pp. 97–100; Mark Peceny, Democracy at the Point of Bayonets (University Park: Pennsylvania State University Press, 1999), pp. 183–216; Dobbins et al., America's Role in Nation-Building; and James Meernik, “United States Military Intervention and the Promotion of Democracy,” Journal of Peace Research, Vol. 33, No. 4 (November 1996), p. 399.

16. 

Enterline and Greig, “Against All Odds?”; Downes and Monten, “Forced to Be Free”; Minxin Pei and Sara Kasper, “Lessons from the Past: The American Record on Nation-Building” (Washington, D.C.: Carnegie Endowment for International Peace, 2003); Eva Bellin, “The Iraqi Intervention and Democracy in Comparative Historical Perspective,” Political Science Quarterly, Vol. 119, No. 4 (Winter 2004/05), pp. 595–608; and Jason Brownlee, “Can America Nation-Build?” World Politics, Vol. 59, No. 2 (January 2007), pp. 314–340.

17. 

Peic and Reiter, “Foreign-Imposed Regime Change, State Power, and Civil War Onset,” p. 453.

18. 

Downes, “Catastrophic Success.”

19. 

Lo, Hashimoto, and Reiter, “Ensuring Peace.” Indeed, FIRCs by democracies that forcibly democratize a defeated belligerent are a near-sufficient condition for peace. See ibid., p. 732. For a related study on the effect of imposed democracy on regional conflict, see Andrew J. Enterline and J. Michael Greig, “Beacons of Hope? The Impact of Imposed Democracy on Regional Peace, Democracy, and Prosperity,” Journal of Politics, Vol. 67, No. 4 (November 2005), pp. 1075–1098.

20. 

Moreover, one of the data sources that Lo, Hashimoto, and Reiter use in their study includes only thirty-three FIRCs that ended wars during this period, while the other includes only thirty-seven. See, respectively, Suzanne Werner, “Absolute and Limited War: The Possibility of Foreign-Imposed Regime Change,” International Interactions, Vol. 22, No. 1 (1996), pp. 67–88; and Henk E. Goemans, Kristian Skrede Gleditsch, and Giacomo Chiozza, “Introducing Archigos: A Dataset of Political Leaders,” Journal of Peace Research, Vol. 46, No. 2 (March 2009), pp. 269–283. Including FIRCs that occurred outside of interstate wars substantially increases the total. For instance, Downes and Monten count 70 FIRCs in the twentieth century, or roughly double the number in Lo, Hashimoto, and Reiter's study. See Downes and Monten, “Forced to Be Free,” p. 111.

21. 

Lo, Hashimoto, and Reiter, “Ensuring Peace”; Reiter, How Wars End, pp. 26–27; and O'Rourke, “Secrecy and Security,” pp. 64–70.

22. 

For more on how misaligned preferences can lead to conflict, see Robert Axelrod, “Conflict of Interest: An Axiomatic Approach,” Journal of Conflict Resolution, Vol. 11, No. 1 (March 1967), pp. 87–99; and Robert Axelrod and Robert O. Keohane, “Achieving Cooperation under Anarchy: Strategies and Institutions,” World Politics, Vol. 38, No. 1 (October 1985), pp. 226–254.

23. 

In many cases, states are able to cooperate in areas where their interests overlap, such as when facing a common threat, but this cooperation disintegrates when the threat disappears (e.g., the collapse of the U.S. and British alliance with the Soviet Union against Nazi Germany after World War II). Moreover, even seemingly unified alliances—such as the Arab alliance against Israel in 1948—can crumble in the face of adversity. See Benny Morris, Righteous Victims: A History of the Zionist-Arab Conflict, 1881–1999 (New York: Alfred A. Knopf, 1999), pp. 218–222; and Avi Shlaim, Collusion across the Jordan: King Abdullah, the Zionist Movement, and the Partition of Palestine (New York: Columbia University Press, 1988).

24. 

Melissa Willard-Foster, “A Peace Too Costly to Keep: Why States Overthrow Foreign Governments,” Ph.D. dissertation, University of California, Los Angeles, 2011; and Phil Haun, Coercion, Survival, and War: Why Weak States Resist the United States (Stanford, Calif.: Stanford University Press, 2015).

25. 

Elizabeth N. Saunders, Leaders at War: How Presidents Shape Military Interventions (Ithaca, N.Y.: Cornell University Press, 2011).

26. 

Although we stress geopolitical factors here, other variables—such as shared ideology or ethnicity—can also cause state preferences to transcend individual leaders.

27. 

For applications of principal-agent theory in security studies, see George W. Downs and David M. Rocke, “Conflict, Agency, and Gambling for Resurrection: The Principal-Agent Problem Goes to War,” American Journal of Political Science, Vol. 38, No. 2 (May 1994), pp. 362–380; Peter D. Feaver, Armed Servants: Agency, Oversight, and Civil-Military Relations (Cambridge, Mass.: Harvard University Press, 2003); and Idean Salehyan, “The Delegation of War to Rebel Organizations,” Journal of Conflict Resolution, Vol. 54, No. 3 (June 2010), pp. 493–515.

28. 

Bueno de Mesquita and Downs argue that democracies install dictators when they intervene because autocrats—who need not respond to the preferences of their domestic audience—are more likely than democrats to follow external guidance. This argument, however, neglects the reality that authoritarian leaders are accountable to domestic audiences for their policies, though this accountability in enforced with coups, assassinations, and rebellions, rather than at the ballot box. See Bueno de Mesquita and Downs, “Intervention and Democracy.” On authoritarian accountability, see Jessica L.P. Weeks, Dictators at War and Peace (Ithaca, N.Y.: Cornell University Press, 2014).

29. 

Scholarship on U.S. foreign military assistance has found that, because of the principal-agent problem, such missions rarely result in militarily effective units: foreign political elites often undercut training missions to keep their soldiers ineffective in an attempt to secure their own political survival. See Stephen Biddle, Julia Macdonald, and Ryan Baker, “Small Footprint, Small Payoff: The Military Effectiveness of Security Force Assistance,” George Washington University, November 2015.

30. 

See Klaus Knorr, The Power of Nations: The Political Economy of International Relations (New York: Basic Books, 1975); Robert Gilpin, War and Change in World Politics (Cambridge: Cambridge University Press, 1981); Edelstein, Occupational Hazards; Simon Collard-Wexler, “Understanding Resistance to Foreign Occupation,” paper presented at the annual meeting of the American Political Science Association, Chicago, Illinois, August 29–September 1, 2013; and Robert A. Pape, Dying to Win: The Strategic Logic of Suicide Terrorism (New York: Random House, 2005).

31. 

Alissa J. Rubin and Thom Shanker, “Afghan Leader Says U.S. Abets Taliban's Goal,” New York Times, March 10, 2013.

32. 

Adam Taylor, “Karzai Joins a Long List of Leaders Ungrateful for U.S. Support,” Washington Post, September 25, 2014.

33. 

The argument that a state's external security environment dictates its interstate behavior is consistent with realist theory. At the same time, the objective of FIRC—replacing the political leadership of another state—is fundamentally at odds with realism. For more on the relationship between FIRC and realism, see John M. Owen IV, The Clash of Ideas in World Politics: Transnational Networks, States, and Regime Change, 1510–2010 (Princeton, N.J.: Princeton University Press, 2010), p. 8.

34. 

Scott Wolford, Dan Reiter, and Clifford J. Carrubba, “Information, Commitment, and War,” Journal of Conflict Resolution, Vol. 55, No. 4 (August 2011), pp. 556–579.

35. 

O'Rourke, “Secrecy and Security,” pp. 59–60.

36. 

Neil MacFarquhar, “Saddam Hussein, Defiant Dictator Who Ruled Iraq with Violence and Fear, Dies,” New York Times, December 30, 2006.

37. 

There may be cases where an intransigent leader is removed from office for failing to acquiesce on an issue that his or her domestic population viewed as an acceptable concession. This does not, however, negate the external sources of interest divergence discussed above. Nor does it necessarily mean that newly installed elites will comply with the intervener's wishes on other issues.

38. 

Stephen M. Streeter, Managing the Counterrevolution: The United States and Guatemala, 1954–1961 (Athens: Center for International Studies, Ohio University, 2000), pp. 223–224.

39. 

Peic and Reiter, “Foreign-Imposed Regime Change, State Power, and Civil War Onset, 1920–2004”; and Downes, “Catastrophic Success.”

40. 

Giacomo Chiozza, Kristian Skrede Gleditsch, and H.E. Goemans, “Civil War, Tenure, and Interstate Insecurity,” paper presented at the annual meeting of the American Political Science Association, Chicago, Illinois, September 2–5, 2004; Idean Salehyan and Kristian Skrede Gleditsch, “Refugees and the Spread of Civil War,” International Organization, Vol. 60, No. 2 (April 2006), pp. 335–366; and Zaryab Iqbal and Harvey Starr, “Bad Neighbors: Failed States and Their Consequences,” Conflict Management and Peace Science, Vol. 25, No. 4 (September 2008), pp. 315–331.

41. 

Salehyan and Gleditsch, “Refugees and the Spread of Civil War.”

42. 

James C. Murdoch and Todd Sandler, “Civil War and Economic Growth: Spatial Dispersion,” American Journal of Political Science, Vol. 48, No. 1 (January 2004), pp. 138–151.

43. 

Iqbal and Starr, “Bad Neighbors.”

44. 

Chiozza, Gleditsch, and Goemans, “Civil War, Tenure, and Interstate Insecurity,” p. 7.

45. 

These MIDs were a direct by-product of the U.S. war in Iraq and not, as some might argue, the result of the United States' overall strategy in fighting the “war on terror.” See the descriptions of MID #50302 with Iran and MID #50501 with Syria in the Correlates of War Project, “Dispute Narratives, 2002–2010,” MIDv4.0 Project, December 13, 2013, http://www.correlatesofwar.org/data-sets/MIDs.

46. 

One might argue that even if FIRC does not decrease intervener-target conflict, the intervening state may still obtain some other benefit from intervention, such as increased trade or access to natural resources. The existing evidence fails to support this supposition. See Paul Zachary, Kathleen Deloughery, and Alexander B. Downes, “No Business Like FIRC Business: Foreign-Imposed Regime Change and Bilateral Trade,” British Journal of Political Science (forthcoming).

47. 

Downes, “Decapitation by FIRC,” p. 30; and O'Rourke, “Secrecy and Security,” p. 300.

48. 

Malcolm Byrne, “Introduction,” in Mark J. Gasiorowski and Malcolm Byrne, eds., Mohammad Mosaddeq and the 1953 Coup in Iran (Syracuse, N.Y.: Syracuse University Press, 2004), p. xv. Some might argue that although the 1953 coup in Iran contributed to anti-American sentiments during the 1979 revolution, from the U.S. perspective, the coup may still be considered a foreign policy success because it achieved other national security goals—namely, preventing the pro-Soviet Tudeh Party from gaining power and facilitating twenty-six years of cooperative relations with the shah's regime. See James D. Fearon, “Taking the Gamble,” Boston Review, Vol. 36, No. 5 (September/October 2011), http://bostonreview.net/fearon-taking-the-gamble. This interpretation overestimates the benefits of intervention by inflating the threat posed by the Tudeh Party and understimates its costs by minimizing the long-term negative effects of the coup on U.S.-Iranian relations. For accounts that reach similar conclusions, see James A. Bill, The Eagle and the Lion: The Tragedy of American-Iranian Relations (New Haven, Conn.: Yale University Press, 1988); Stephen Kinzer, All the Shah's Men: An American Coup and the Roots of Middle East Terror (Hoboken, N.J.: John Wiley, 2003); and Ervand Abrahamian, The Coup: 1953, the CIA, and the Roots of Modern U.S.-Iranian Relations (New York: New Press, 2013).

49. 

The number is large because both Britain and the United States are coded as interveners.

50. 

On Austria's intervention in the Two Sicilies (1821), see Paul W. Schroeder, The Transformation of European Politics, 1763–1848 (Oxford: Clarendon, 1994), p. 613. On Prussia's interventions in Saxony and Baden, see, respectively, Paul Veit Valentin, 1848: Chapters of German History, trans. Ethel Talbot Scheffauer (Hamden, Conn.: Archon, 1965), p. 402; and Dennis E. Showalter, The Wars of German Unification (London: Arnold, 2004), p. 41.

51. 

Henry Giniger, “Gabon Insurgents Yield as France Rushes In Troops,” New York Times, February 20, 1964.

52. 

One report states that Denard's force numbered only thirty-three individuals. See Marlise Simons, “1,000 French Troops Invade Comoros to Put Down a Coup,” New York Times, October 5, 1995.

53. 

Readers might question why we postulate that a restoration FIRC reduces the likelihood of conflict rather than leaving it unchanged, given that, presumably, the restored leader already had good relations with the intervener in the period preceding FIRC. This is true, but the friendly leader was at some point supplanted by elites who were more hostile to the intervener, thereby increasing the chance of conflict before FIRC occurs.

54. 

Soviet FIRCs in Eastern Europe also empowered leaders who shared the same ideology as their Soviet backers, and who thus wanted their countries to have close relationships with Moscow. On shared values among Marxist regimes as a source of peace, see Mark Peceny, Caroline C. Beer, and Shannon Sanchez-Terry, “Dictatorial Peace?” American Political Science Review, Vol. 96, No. 1 (March 2002), pp. 19–20.

55. 

See, for example, Bruce Russett, Grasping the Democratic Peace: Principles for a Post–Cold War World (Princeton, N.J.: Princeton University Press, 1993).

56. 

Bueno de Mesquita and Downs argue that this is why democratic interveners do not promote democracy. See Bueno de Mesquita and Downs, “Intervention and Democracy.”

57. 

Haiti, however, reverted to autocracy in 1999.

58. 

Downes and Monten, “Forced to Be Free,” pp. 122–128. A sixth case, Costa Rica in 1919, resulted in democracy. Because the country was also a democracy before FIRC, it does not constitute a case of democratic transition.

59. 

Restricting overt institutional FIRCs to those that exclusively promote democracy (to match the covert institutional FIRCs in our sample, which—because they are carried out by the United States—all promote democracy) does not change our theoretical prediction or our empirical results.

60. 

Theoretically, if the intervening state were able to keep its role in installing a foreign leader secret, this would eliminate one of the problems associated with regime change—namely, the legitimacy deficit associated with the new regime having been imposed by a foreign power. We expect such cases to be rare given the frequency with which target regimes accuse interveners of meddling in their domestic affairs. Although states have multiple reasons for intervening covertly, most of these motives impel states to attempt to conceal their role only from outside observers, and thus FIRCs seldom remain a secret from the target regime. See O'Rourke, “Secrecy and Security,” chapter 3.

61. 

Ibid., p. 333. It is possible that a target regime's domestic audience may not deem the regime's allegations of foreign meddling credible, given the regime's incentives to blame outside actors for its own shortcomings. Such a situation is, however, likely rare for two reasons. First, the United States has a well-established reputation for pursuing covert regime change, so foreign audiences are likely to be sympathetic to these allegations. Second, targeted governments can often produce credible evidence of U.S. involvement, such as confiscated U.S.-made weapons and technology, foreign agents willing to testify about their U.S. connections, and, in a few cases, captured U.S. spies and contractors.

62. 

Supporting this line of reasoning, O'Rourke found that states targeted in U.S.-backed covert regime change attempts during the Cold War were more likely to experience civil war or an episode of government-led mass killing within ten years of the intervention compared to similar countries where the United States had not intervened. See Lindsey O'Rourke, “Mission Impossible? The Consequences of Covert Regime Change,” paper presented at the International Studies Association Annual Convention, Atlanta, Georgia, March 16–19, 2016.

63. 

Dobbins et al., America's Role in Nation-Building; and Edelstein, Occupational Hazards.

64. 

One reason why there are no covert restoration FIRCs is that successful covert missions require time. Thus, when states must act quickly to reinstall a recently deposed leader, they are more likely to intervene overtly. See O'Rourke, “Secrecy and Security,” p. 91.

65. 

Powell, quoted in Bob Woodward, Plan of Attack: The Definitive Account of the Decision to Invade Iraq (New York: Simon and Schuster, 2004), p. 150.

66. 

O'Rourke, “Secrecy and Security,” p. 300.

67. 

The dataset was constructed using the EUGene program. See D. Scott Bennett and Allan C. Stam, “EUGene: A Conceptual Manual,” International Interactions, Vol. 26, No. 2 (2000), pp. 179–204.

68. 

We chose a directed-dyad design (rather than one using undirected dyads) because we want to understand not only whether dyads that experience FIRC are more conflict-prone than other dyads, but also to identify which state in a dyad initiated militarized conflict. For overt FIRCs, our theory—and the power imbalance in favor of interveners—implies that post-FIRC conflict in most cases should be initiated by the state that launched the FIRC. This hunch turns out to be correct: nearly 80 percent of MIDs following overt leadership FIRCs, for example, were initiated by the intervener. For covert FIRCs, by contrast, many of which fail, we would expect conflict initiation to be more evenly split: failure could prompt reintervention by the aggressor or retaliation by the target. This expectation is also borne out. For covert leadership FIRCs, interveners initiate only 53 percent of post-FIRC MIDs.

69. 

Inconsistencies in the data sources about when certain states are considered to be members of the international system, however, result in the exclusion of more than a dozen nineteenth-century FIRCs from our analysis. EUGene contains data from a variety of sources that appear to use inconsistent rules to identify states. As a result, eight cases in Central America from 1855 to 1894 as well as FIRCs involving Modena, Parma, Afghanistan, and Peru are omitted. Because several of these cases involved leadership FIRCs followed by conflict, this loss of cases may result in an underestimation of the effect of leadership FIRC. Another reason why our analysis might understate the effect of overt leadership FIRC is that some cases that could have been followed by interstate conflict—such as German FIRCs in World War II—are truncated because the target was occupied by the intervener and disappeared temporarily from the international system. Conflict still occurred in several of these cases (e.g., resistance to German occupation in Greece, Yugoslavia, and elsewhere); it simply took place inside the target rather than between the target and the intervener.

70. 

The MID dataset also has many disadvantages, particularly when used to test theories about phenomena it was not designed for, such as threat effectiveness. See Alexander B. Downes and Todd S. Sechser, “The Illusion of Democratic Credibility,” International Organization, Vol. 66, No. 3 (Summer 2012), pp. 457–489.

71. 

Because some MIDs involve only the threat or display of force, we also tested our models on only those MIDs that escalated to the use of force or war. See the section on robustness tests below and the supplemental materials.

72. 

Methodologists have recently questioned the assumption of statistical independence across dyads. See, for example, Eric Neumayer and Thomas Plumper, “Spatial Effects in Dyadic Data,” International Organization, Vol. 64, No. 1 (Winter 2010), pp. 145–166; Robert S. Erikson, Pablo M. Pinto, and Kelly T. Rader, “Dyadic Analysis in International Relations: A Cautionary Tale,” Political Analysis, Vol. 22, No. 4 (Autumn 2014), pp. 457–463; Peter M. Aranow, Cyrus Samii, and Valentina A. Assenova, “Cluster-Robust Variance Estimation for Dyadic Data,” Political Analysis, Vol. 23, No. 4 (Autumn 2015), pp. 564–577; and Skyler J. Cranmer and Bruce A. Desmarais, “A Critique of Dyadic Design,” International Studies Quarterly, Vol. 60, No. 2 (June 2016), pp. 355–362. For a measured defense of dyads, see Paul Poast, “Dyads Are Dead, Long Live Dyads! The Limits (but Not Rejection) of Dyadic Designs in International Relations Research,” International Studies Quarterly, Vol. 60, No. 2 (June 2016), pp. 369–374. We implement some suggested remedies for this potential problem below in the section on robustness checks.

73. 

Downes and Monten, “Forced to Be Free,” p. 109.

74. 

O'Rourke, “Secrecy and Security,” p. 17. For a full discussion of the many definitional issues associated with covert FIRC, see the supplemental materials.

75. 

On Britain's aim, as Prime Minister Anthony Eden put it, to “bring about the downfall of the present Egyptian Government,” see J.A. Sellers, “Military Lessons: The British Perspective,” in Selwyn Ilan Troen and Moshe Shemesh, eds., The Suez-Sinai Crisis, 1956: Retrospective and Reappraisal (New York: Columbia University Press, 1990), p. 17. For Iranian pronouncements calling for the removal of Saddam Hussein during the Iran-Iraq War, see Dilip Hiro, The Longest War: The Iran-Iraq Military Conflict (New York: Routledge, 1991), pp. 35, 61, 64.

76. 

It is important to code the effect as beginning the year after FIRC because many FIRCs are coded as MIDs. Including the year of FIRC would thus incorrectly code some FIRCs as resulting in MIDs.

77. 

A complete list of FIRCs—and MIDs that follow FIRCs—is available in the supplemental materials.

78. 

See, for example, Jessica L. Weeks, “Strongmen and Straw Men: Authoritarian Regimes and the Initiation of International Conflict,” American Political Science Review, Vol. 106, No. 2 (May 2012), pp. 326–347. Unless otherwise noted, these variables were generated by EUGene.

79. 

This variable is from Carles Boix, Michael Miller, and Sebastian Rosato, “A Complete Data Set of Political Regimes, 1800–2007,” Comparative Political Studies, Vol. 46, No. 12 (December 2013), pp. 1523–1554.

80. 

A correlation matrix of our independent variables, available in the supplementary materials, indicates that multicollinearity is not a problem. Importantly, none of the FIRC variables is highly correlated with any of the control variables.

81. 

Complete regression results are available in appendix A in the supplementary materials. Details on the robustness tests may be found in appendices C through M in the supplementary materials. Here, however, we rely on graphs to summarize our main findings.

82. 

Estimates were generated using CLARIFY. See Michael Tomz, Jason Wittenberg, and Gary King, “CLARIFY: Software for Interpreting and Presenting Statistical Results,” Journal of Statistical Software, Vol. 8 (2003), pp. 1–29. All other variables in the models are held constant at their means or modes.

83. 

The first set of bars in figure 2 indicates that MIDs are also significantly more likely to occur while a covert regime change operation is under way.

84. 

A Wald test indicates that it is extremely unlikely that these two effects are equal (p = 0.0001).

85. 

Careful readers may note that the 95 percent confidence intervals for the presence and absence of overt leadership FIRC overlap, but this does not necessarily mean that the effect of overt leadership FIRC is not significant. Model 2 in table A1 (appendix A) in the supplementary materials shows that overt leadership FIRC is significant at p < 0.05, and the 95 percent confidence intervals of the change in predicted probability calculated by CLARIFY (shown in appendix B) are strictly positive.

86. 

These are the Dominican Republic, 1968; Nicaragua, 1989; Chile, 1989; Poland, 1989; and the Philippines, 1989.

87. 

In other words, the joint democracy strand of the institutional FIRC logic outweighs the interest divergence logic.

88. 

O'Rourke, “Secrecy and Security,” pp. 115–116, 346, 357. Because these covert efforts coincided with large-scale overt efforts to liberalize these countries, such as Marshall Plan aid, a good case can be made that they would be better classified as institutional FIRCs rather than leadership FIRCs. We opted against this coding, however, for two reasons. First, the covert efforts themselves focused on the narrow objective of ensuring that Leftist parties did not win the elections (i.e., determining the leadership of the state), rather than bringing about the elections in the first place (i.e., changing the institutions of the state). Second, as with all coding debates, the question arises as to whether the coding decision biases the evidence for or against the authors' argument. Because we argue that successful leadership FIRCs harm interstate relations, this coding goes against our argument.

89. 

A Wald test indicates that the coefficients are not significantly different.

90. 

The results of these tests appear in appendices C through M in the supplementary materials.

91. 

Our alliance dummy variable—produced by EUGene—is coded 1 if the states in a dyad were joined in any kind of formal alliance (defense pact, nonaggression pact, or neutrality pact) with each other in a given year, and 0 otherwise. Joint alliance membership is never significantly related to MID initiation.

92. 

Given that we lack data on covert interventions by nations other than the United States, including non-U.S. dyads could be inappropriate because it gives undue weight to the U.S. cases. In related work using a slightly different set of covariates, O'Rourke finds that limiting the sample to dyads containing the United States during the Cold War does not eliminate the effect of covert FIRC. See O'Rourke, “Mission Impossible?” pp. 14–18.

93. 

See Eric Neumayer and Thomas Plümper, “Making Spatial Analysis Operational: Commands for Generating Spatial-Effect Variables in Monadic and Dyadic Data,” Stata Journal, Vol. 10, No. 4 (2010), pp. 585–605; and Aranow, Samii, and Assenova, “Cluster-Robust Variance Estimation for Dyadic Data.” The authors thank Peter Aranow and Cyrus Samii for assistance implementing their method.

94. 

Gary King and Langche Zeng, “Logistic Regression in Rare Events Data,” Political Analysis, Vol. 9, No. 2 (Spring 2001), pp. 137–163.

95. 

Fixed effects models also capture unmeasured factors unique to dyads that could influence the likelihood of militarized disputes. In this analysis, overt leadership FIRC slips slightly in significance (p < 0.14). Interestingly, restoration FIRC remains negative and significant, confirming our claim that these FIRCs decrease the likelihood of MIDs even though they restore leaders with whom interveners presumably already had good relations. To further compare dyads' propensity for conflict immediately before and after FIRC, we included a dummy variable coded 1 during the ten years prior to different kinds of FIRCs. Militarized disputes were significantly less likely after all overt types of FIRC than before. Dyads that experienced an overt leadership FIRC, however, continued to be at significant risk of suffering another conflict after FIRC occurred. By contrast, conflict was significantly more likely after most covert FIRCs than it was beforehand.

96. 

Genetic matching produced a set of control cases that was statistically indistinguishable from cases that experienced different types of FIRC. The only notable deviation from our main results is that overt leadership FIRC becomes insignificant after genetic matching. On matching, see Daniel E. Ho et al., “Matching as Nonparametric Preprocessing for Reducing Model Dependence in Parametric Causal Inference,” Political Analysis, Vol. 15, No. 3 (Summer 2007), pp. 199–236; and Alexis Diamond and Jasjeet S. Sekhon, “Genetic Matching for Estimating Causal Effects: A General Multivariate Matching Method for Achieving Balance in Observational Studies,” Review of Economics and Statistics, Vol. 95, No. 3 (July 2013), pp. 932–945. Note that the use of matching to deal with selection bias—although widespread—remains controversial. See Michael K. Miller, “The Uses and Abuses of Matching in Political Science,” George Washington University, n.d.

97. 

Although we code this case as an overt FIRC for the purposes of our statistical analysis, the Rwandan leadership went to great lengths to conceal its military involvement. As we demonstrate below, however, the regime change plot was essentially hatched in Kigali, and Rwandan military forces participated extensively from beginning to end. This Rwandan presence was thus impossible to hide and was widely reported at the time. For examples of this reporting, see Johan Pottier, Re-Imagining Rwanda: Conflict, Survival, and Disinformation in the Late Twentieth Century (Cambridge: Cambridge University Press, 2002), p. 88.

98. 

According to traditional indices of power used in international relations (and our statistical analysis), Rwanda appears far weaker than its enormous neighbor. For example, Rwanda's share of global material capabilities in 1996 was roughly one-fifth that of Zaire's; in addition, Rwanda possessed about one-ninth of Zaire's population and one-eighty-ninth of its land area.

99. 

Jason K. Stearns, Dancing in the Glory of Monsters: The Collapse of the Congo and the Great War of Africa (New York: PublicAffairs, 2011), p. 55.

100. 

This advantage is further accentuated when the capabilities of Rwanda's allies are taken into account. Both the Ugandan People's Defense Forces and the Angolan Armed Forces were much more capable than Kinshasa's military. On state weakness in Zaire in general, see Boaz Atzili, “When Good Fences Make Bad Neighbors: Fixed Borders, State Weakness, and International Conflict,” International Security, Vol. 31, No. 3 (Winter 2006/07), pp. 156–161.

101. 

The literature on the Rwandan genocide is substantial. For exemplars, see Gérard Prunier, The Rwanda Crisis: History of a Genocide (New York: Columbia University Press, 1995); Alison Des Forges, Leave None to Tell the Story: Genocide in Rwanda (New York: Human Rights Watch, 1999); Samantha Power, “A Problem from Hell”: America and the Age of Genocide (New York: Basic Books, 2002), pp. 329–389; and Scott Straus, The Order of Genocide: Race, Power, and War in Rwanda (Ithaca, N.Y.: Cornell University Press, 2006).

102. 

See the figures in Fiona Terry, Condemned to Repeat? The Paradox of Humanitarian Action (Ithaca, N.Y.: Cornell University Press, 2002), p. 158; Sarah Kenyon Lischer, Dangerous Sanctuaries: Refugee Camps, Civil War, and the Dilemmas of Humanitarian Aid (Ithaca, N.Y.: Cornell University Press, 2005), pp. 78–79; Filip Reyntjens, The Great African War: Congo and Regional Geopolitics, 1996–2006 (Cambridge: Cambridge University Press, 2009), p. 45; and Gérard Prunier, Africa's World War: Congo, the Rwandan Genocide, and the Making of a Continental Catastrophe (Oxford: Oxford University Press, 2009), p. 25. Most of the remaining Hutu refugees (about half a million) fled to Tanzania, with a smaller number ending up in Burundi.

103. 

On October 31, 1994, for instance, ex-FAR soldiers killed thirty-six people in Gisenyi in northwest Rwanda. See Prunier, Africa's World War, p. 26; and Stearns, Dancing in the Glory of Monsters, p. 29.

104. 

Mobutu frequently hosted high-ranking leaders of the génocidaire regime. As Gérard Prunier reports, “The former Rwandese leadership had free run of the country.” See Prunier, Africa's World War, p. 28. Zairian military forces also made little attempt to disarm the FAR and other Hutu militants as they entered the country. See Terry, Condemned to Repeat? pp. 160–161; Lischer, Dangerous Sanctuaries, p. 85; and Human Rights Watch, Rwanda/Zaire: Rearming with Impunity—International Support for the Perpetrators of the Rwandan Genocide (New York: Human Rights Watch, May 1995), p. 11. Zairian authorities, moreover, looked the other way as the Hutu leadership imported arms from abroad, and in some cases facilitated these shipments. See Terry, Condemned to Repeat? pp. 161–163; Human Rights Watch, Rwanda/Zaire, pp. 6–15; and Roger Winter, “Lancing the Boil: Rwanda's Agenda in Zaire,” in Howard Adelman and Govind C. Rao, War and Peace in Zaire-Congo: Analyzing and Evaluating Intervention, 1996–1997 (Trenton, N.J.: Africa World Press, 2004), pp. 111–112.

105. 

Human Rights Watch, Rwanda/Zaire, p. 5. See also Prunier, Africa's World War, p. 28.

106. 

Filip Reyntjens, “Briefing: The Second Congo War: More Than a Remake,” African Affairs, April 1999, p. 242. See also Reyntjens, The Great African War, pp. 46–47, 51; Prunier, Africa's World War, p. 67; Timothy Longman, “The Complex Reasons for Rwanda's Engagement in Congo,” in John F. Clark, ed., The African Stakes of the Congo War (New York: Palgrave Macmillan, 2002), pp. 133–134; Georges Nzongola-Ntalaja, The Congo from Leopold to Kabila: A People's History (London: Zed, 2002), p. 225; International Crisis Group, “How Kabila Lost His Way: The Performance of Laurent Désiré Kabila's Government” (Brussels: International Crisis Group, May 21, 1998), p. 4; and William G. Thom, “Congo-Zaire's 1996–97 Civil War in the Context of Evolving Patterns of Military Conflict in Africa in the Era of Independence,” Journal of Conflict Studies, Vol. 19, No. 2 (Autumn 1999), p. 95.

107. 

On the conflicts in the Kivus that endangered ethnic Tutsis, see Prunier, Africa's World War, pp. 46–58; and Reyntjens, The Great African War, pp. 13–23.

108. 

The RPA and the rebels pursued the refugees who fled west, intent on wiping them out. On this murderous pursuit, see Prunier, Africa's World War, pp. 143–148; and Reyntjens, The Great African War, pp. 80–101. Prunier estimates that 300,000 Hutu refugees died. See Prunier, Africa's World War, p. 148.

109. 

The four groups were the Parti de la Révolution du Peuple (Party of the People's Revolution, headed by Kabila), Conseil National de Résistance pour la Démocratie (National Council of Resistance for Democracy), Mouvement Révolutionnaire pour la Libération du Zaïre (Revolutionary Movement for the Liberation of Zaire), and Alliance Démocratique des Peuples (Democratic Alliance of the People).

110. 

Reyntjens, The Great African War, p. 104. Kabila became president of the AFDL in January 1997 after the death in “mysterious circumstances” of the group's initial military commander, André Kisase Ngandu. See Nzongola-Ntalaja, The Congo from Leopold to Kabila, p. 226. Before Kabila became the figurehead of the Rwandan invasion, his main claim to fame was hosting Che Guevara when the Cuban revolutionary came to fight in Congo in 1965. See Prunier, Africa's World War, p. 114. On Kabila as a “Zairian face” for the Rwandan invasion, see Thomas Turner, Congo (Cambridge: Polity, 2013), p. 16; Thomas Turner, The Congo Wars: Conflict, Myth, and Reality (London: Zed, 2007), p. 5; Kevin C. Dunn, “A Survival Guide to Kinshasa: Lessons of the Father, Passed Down to the Son,” in Clark, The African Stakes of the Congo War, p. 56; Peter Rosenblum, “Kabila's Congo,” Current History, May 1998, p. 194; and Stearns, Dancing in the Glory of Monsters, p. 87.

111. 

Reyntjens, The Great African War, p. 107.

112. 

Prunier, Africa's World War, p. 115.

113. 

Kabila was a member of the Luba tribe from Katanga.

114. 

For articles in the Rwandan press that support this interpretation, see Reyntjens, The Great African War, p. 45; and Prunier, Africa's World War, p. 68. Kagame himself acknowledged in a 1997 interview that toppling Mobutu was one goal of the Rwandan invasion. See John Pomfret, “Rwandans Led Revolt in Congo,” Washington Post, July 9, 1997. One author claims that Rwanda intended from the outset to “push on all the way to the capital” and oust Mobutu. See David Van Reybrouck, Congo: The Epic History of a People, trans. Sam Garrett (New York: HarperCollins, 2014), p. 419. Prunier also mentions various discussions among African leaders about toppling Mobutu. See Prunier, Africa's World War, p. 67.

115. 

Thom, “Congo-Zaire's 1996–97 Civil War in the Context of Evolving Patterns of Military Conflict in Africa in the Era of Independence,” p. 116; and International Crisis Group, “How Kabila Lost His Way,” p. 6. Reyntjens also finds no direct evidence from before the war that Rwanda intended to eliminate Mobutu. See Reyntjens, The Great African War, p. 51.

116. 

The principal reason that each of these countries supported or participated in military operations against Mobutu was to eliminate rebel groups that found safe haven (and in some cases material support) in Zaire: the National Union for the Total Independence of Angola for Angola, the National Council for the Defense of Democracy—Forces for the Defense of Democracy for Burundi, and the Alliance of Democratic Forces for Uganda. Other countries that supported the anti-Mobutu coalition included Zambia, Zimbabwe, Ethiopia, and Eritrea. See Reyntjens, “Briefing,” p. 242; Reyntjens, The Great African War, pp. 58–66; and Stearns, Dancing in the Glory of Monsters, pp. 51–54, 285–286.

117. 

Reflecting Kabila's lack of a domestic political base, ten ministers in his new government were recently returned exiles. See International Crisis Group, “How Kabila Lost His Way,” p. 7. Kabila also excluded anyone with ties to the Mobutu regime; almost all leaders of the domestic opposition fell into this category.

118. 

Prunier, Africa's World War, pp. 152, 151, 152, respectively. On the lack of regime institutionalization and the personalization of Kabila's rule, see also Reyntjens, The Great African War, p. 156; Van Reybrouck, Congo, pp. 434–435; and François Ngolet, Crisis in the Congo: The Rise and Fall of Laurent Kabila (New York: Palgrave Macmillan, 2011), pp. 6–7.

119. 

Reyntjens, The Great African War, p. 161.

120. 

Van Reybrouck, Congo, p. 435. See also Reyntjens, The Great African War, pp. 158–159.

121. 

Turner, The Congo Wars, p. 38. Some might contend that Kabila's incompetence or authoritarian nature was to blame for his domestic opposition. We argue below, however, that his poor governing choices and repressive responses to challenges were largely a product of his weak position, which in turn was a consequence of how he came to power. See Stearns, Dancing in the Glory of Monsters, pp. 170–171.

122. 

Prunier, Africa's World War, p. 172.

123. 

Reyntjens, The Great African War, p. 174; and Prunier, Africa's World War, p. 173.

124. 

The Army for the Liberation of Rwanda was itself a further iteration of the Rassemblement pour la Rétour des Réfugiés et la Démocratie au Rwanda (Rally for the Return of the Refugees and Democracy in Rwanda), which was founded in April 1995, and would later evolve into the Forces Démocratiques pour la Libération du Rwanda (Democratic Forces for the Liberation of Rwanda, FDLR). On the relationships among these groups, see Hans Romkema, “Opportunities and Constraints for the Disarmament and Repatriation of Foreign Armed Groups in the Democratic Republic of Congo: The Cases of the FDLR, FNL, and ADF/NALU” (Washington, D.C.: World Bank, 2007), pp. 32–33.

125. 

Reyntjens, The Great African War, p. 146.

126. 

Stearns, Dancing in the Glory of Monsters, p. 181.

127. 

Emizet François Kisangani, Civil Wars in the Democratic Republic of Congo, 1960–2020 (Boulder, Colo.: Lynne Rienner, 2012), p. 134.

128. 

As Van Reybrouck summarizes, “Every Tutsi was seen as Rwandan and every Rwandan as an occupier.” See Van Reybrouck, Congo, p. 438.

129. 

Kisangani, Civil Wars in the Democratic Republic of Congo, 1960–2010, pp. 133–134.

130. 

Reyntjens, The Great African War, p. 148. François Ngolet agrees: “Tutsi hegemony was not well received by the rest of the Zairians. They felt humiliated and frequently accused the Tutsi of a triumphalist and arrogant attitude.” See Ngolet, Crisis in the Congo, p. 12.

131. 

Kisangani, Civil Wars in the Democratic Republic of Congo, 1960–2010, p. 135. “Mai mai” is a generic term that refers to local self-defense militias in Congo.

132. 

Osita Afoaku notes that “Kabila maintained a narrow base of domestic support” that included only two non-Tutsi elements: the so-called kadogos—the child soldiers recruited by the AFDL during its drive to Kinshasa—and natives of Kabila's home province of Katanga. See Afoaku, “Congo's Rebels: Their Origins, Motivations, and Strategies,” in Clark, The African Stakes of the Congo War, p. 113.

133. 

Ngolet, Crisis in the Congo, p. 12; and Idean Salehyan, Rebels without Borders: Transnational Insurgencies in World Politics (Ithaca, N.Y.: Cornell University Press, 2009), p. 152.

134. 

See Rosenblum, “Kabila's Congo,” p. 194.

135. 

Ngolet, Crisis in the Congo, p. 12. Kabila's resistance to the attempts by the United Nations to investigate reports of human rights violations by the RPA in the campaign to overthrow Mobutu also constitutes powerful evidence of his reliance on Rwanda. On this episode, see ibid., pp. 4–8; Dunn, “A Survival Guide to Kinshasa,” p. 58; and Van Reybrouck, Congo, p. 437.

136. 

Reyntjens, The Great African War, p. 167.

137. 

Cited in Reyntjens, The Great African War, p. 167 n. 87.

138. 

Van Reybrouck, Congo, p. 439. For similar assessments, see Ngolet, Crisis in the Congo, p. 12; Afoaku, “Congo's Rebels,” p. 113; Dunn, “A Survival Guide to Kinshasa,” p. 61; Longman, “The Complex Reasons for Rwanda's Engagement in Congo,” p. 138; and Prunier, Africa's World War, pp. 177–178.

139. 

Prunier, Africa's World War, p. 177.

140. 

Kisangani, Civil Wars in the Democratic Republic of Congo, 1960–2010, p. 139.

141. 

Prunier, Africa's World War, p. 176.

142. 

Ngolet, Crisis in the Congo, pp. 11–17.

143. 

This move, however, sparked a mutiny, forcing a reversal of the decision. See Prunier, Africa's World War, p. 176; and Ngolet, Crisis in the Congo, p. 16.

144. 

Dunn, “A Survival Guide to Kinshasa,” p. 62.

145. 

Kisangani, Civil Wars in the Democratic Republic of Congo, 1960–2010, p. 142.

146. 

Kabila expelled more than 10,000 Rwandans and Banyamulenge from the armed forces. See Salehyan, Rebels without Borders, p. 152.

147. 

Longman, “The Complex Reasons for Rwanda's Engagement in Congo,” p. 138.

148. 

Turner, Congo, pp. 17, 54.

149. 

We do not mean to imply that security concerns were the only factors motivating the attack. Some analysts indeed privilege security; see, for example, Afoaku, “Congo's Rebels,” p. 114; and International Crisis Group, “How Kabila Lost His Way,” pp. 2, 21. Others mention additional factors, such as ethnic solidarity with Congolese Tutsi, the domestic unifying effects of a foreign war, economic interests in the DRC, and high levels of confidence among RPF leaders. See Longman, “The Complex Reasons for Rwanda's Engagement in Congo,” pp. 130–133, 134–138.

150. 

Nzongola-Ntalaja, The Congo from Leopold to Kabila, p. 227. See also Dunn, “A Survival Guide to Kinshasa,” p. 62.

151. 

Prunier, Africa's World War, pp. 183–184. Confirming the Rally's lack of independence, Jason Stearns writes, “Major leadership changes were imposed by Kigali, and all military operations were led by Rwandan commanders in the field.” See Stearns, Dancing in the Glory of Monsters, p. 209.

152. 

For mortality estimates, see Benjamin Coghlan et al., “Mortality in the Democratic Republic of Congo: A Nationwide Survey,” Lancet, January 7, 2006, pp. 44–51; and Les Roberts et al., “Mortality in the Democratic Republic of Congo: Results from a Nationwide Survey” (New York: International Rescue Committee, April 2003).

153. 

One study alleges that Kabila struck a deal with the ex-FAR to defend himself against an anticipated Rwandan move to unseat him, which in turn helped precipitate Kigali's attack. See Stearns, Dancing in the Glory of Monsters, p. 183.

154. 

Quoted in Jo Becker and Scott Shane, “Hillary Clinton, ‘Smart Power’ and a Dictator's Fall,” New York Times, February 27, 2016.

155. 

Flavia Krause-Jackson and Caroline Alexander, “Jibril Turns against Foreign Powers That Aided Qaddafi Overthrow,” Bloomberg, November 14, 2011, http://www.bloomberg.com/news/articles/2011-11-14/jibril-turns-against-foreign-powers-that-aided-qaddafi-overthrow; and Scott Shane “As Islamists Gain Influence, Washington Reassesses Who Its Friends Are,” New York Times, July 9, 2012.

156. 

James Risen, Mark Mazzetti, and Michael S. Schmidt, “U.S.-Approved Arms for Libya Rebels Fell into Jihadis' Hands,” New York Times, December 5, 2012.

157. 

Scott Shane and Jo Becker, “A New Libya, with ‘Very Little Time Left,’” New York Times, February 27, 2016.

158. 

Barack Obama, interview with Chris Wallace, “Exclusive: President Barack Obama on ‘Fox News Sunday, ‘” Fox News, April 10, 2016.

159. 

Eric Schmitt, “Obama Is Pressed to Open Military Front against ISIS in Libya,” New York Times, February 4, 2016.

160. 

Michael Doran and Max Boot, “5 Reasons to Intervene in Syria Now,” New York Times, September 26, 2012; Paul Szoldra, “Senior GOP Senators: We Won't Support Military Strikes without ‘Overall Strategy’ to Remove Assad,” Business Insider, August 31, 2013; William C. Martel, “On Syria: Don't Take Regime Change Off the Table,” National Interest, September 9, 2013, http://nationalinterest.org/commentary/syria-dont-take-regime-change-the-table-9017; and Anne-Marie Slaughter, “Stopping Russia Starts in Syria,” Project Syndicate, April 23, 2014, https://www.project-syndicate.org/commentary/anne-marie-slaughter-on-how-us-intervention-in-the-syrian-civil-war-would-alter-vladimir-putin-s-calculus-in-ukraine?barrier=true. Unclassified 2012 State Department emails also reveal Clinton's rationale prior to the intervention: “Victory may not come quickly or easily, but it will come. And the payoff will be substantial. Iran would be strategically isolated, unable to exert its influence in the Middle East. The resulting regime in Syria will see the United States as a friend, not an enemy.” See “Unclassified U.S. Department of State Case No. F-2014-20439, Doc. No. C057944998 Date: 11/30/2015,” WikiLeaks, https://wikileaks.org/clinton-emails/emailid/18328#source.

161. 

Ernesto Londono and Greg Miller, “CIA Begins Weapons Delivery to Syrian Rebels,” Washington Post, September 11, 2013; Mark Mazzetti and Matt Apuzzo, “U.S. Relies Heavily on Saudi Money to Support Syrian Rebels,” New York Times, January 23, 2016; and Mark Mazzetti and Ali Younes, “C.I.A. Arms for Syrian Rebels Supplied Black Market, Officials Say,” New York Times, June 26, 2016.

162. 

Anne Barnard and Hwaida Saad, “Syrian Rebel Infighting Undermines Anti-Assad Effort,” New York Times, July 12, 2013; and Ben Hubbard and Karam Shoumali, “Top Military Body against Syria's Assad Is in Chaos, Undermining Fight,” New York Times, February 23, 2014.

163. 

Mona Mahmood and Ian Black, “Free Syrian Army Rebels Defect to Islamist Group Jabhat al-Nusra,” Guardian, May 8, 2013.

164. 

These forces later seized the Council's warehouses and headquarters, causing the United States to temporarily suspend its supply of aid. See Tim Lister, “Islamic Front in Syria Deals Another Blow to Rebel Alliance,” CNN, December 12, 2013, http://www.cnn.com/2013/12/12/world/syria-islamic-front/; Aron Lund, “Showdown at Bab al-Hawa” (Washington, D.C.: Carnegie Endowment for International Peace, December 12, 2013), http://carnegieendowment.org/syriaincrisis/?fa=53896; and Liz Sly, “U.S. Suspends Aid to Syrian Rebels after Islamists Seize Warehouses,” Washington Post, December 11, 2013.

165. 

White House, “Press Conference by the President” (Washington, D.C.: Office of the Press Secretary, October 2, 2015).

166. 

Michael D. Shear, Helene Cooper, and Eric Schmitt, “Obama Administration Ends Effort to Train Syrians to Combat ISIS,” New York Times, October 9, 2015.

167. 

David E. Sanger and Helene Cooper, “Obama Turns to Diplomacy and Military in Syria, and Is Met with Doubts,” New York Times, November 10, 2015.