Playing with fire: Patronage, personalism, and. counterterrorism aid

Playing with fire: Patronage, personalism, and counterterrorism aid Andrew Boutton Abstract A common strategy pursued by states targeted by international terrorist groups is to provide economic and military assistance to the states that host these groups. However, very little work to date has examined whether foreign aid is actually effective at bringing about favorable counterterrorism outcomes. This paper offers an explanation for why a strategy of foreign aid-for-counterterrorism can be counterproductive in some contexts. Scholars of development aid argue that the political structure of the recipient regime is an important determinant of the effectiveness of aid in boosting economic growth. I apply this logic to the question of counterterrorism aid and argue that not all regimes have the same incentives, and the effects of foreign aid on terrorism will thus be conditional on those incentives. Leaders in personalist dictatorships operate within a unique political environment that makes them heavily dependent on the distribution of rents (patronage) for their survival. While many regimes rely on patronage to some extent to remain in power, personalist regimes typically rely on external sources of revenueforeign aid to fund their patronage networks. Furthermore, they lack viable alternative survival strategies if the patronage network is disrupted and challenges to the regime emerge. Thus, personalist leaders depend on a continual flow of foreign aid in order to buy the loyalty of regime collaborators and remain in power. One well-known and effective means of receiving foreign aid is to have a terrorism problem. While the donor intends for such aid to boost the counterterrorism capacity and cooperation of the recipient, I argue that personalist regimes will play-up terrorist activity in order to continue receiving aid. Using a variety of data on regime type, terrorist attacks, and terrorist group duration, I find that in personalist regimes, U.S. aid significantly increases levels of terrorist activity. This paper contributes to the literature linking foreign aid and terrorism by taking regime type into consideration as an important factor conditioning the effectiveness of foreign aid as a counterterrorism tool. Email: atboutton@gmail.com

Introduction With diplomatic facilities, businesses, and civilian and military personnel in nearly every country in the world, the United States has truly global security interests, which also provide attractive targets for terrorist groups. For day-to-day protection of those interests from instability and terrorism, the US must rely upon the cooperation of the host country governments and their security forces. Paradoxically, the places in which these interests need protection most also often happen to be presided over by governments with a host of other, more pressing priorities (Byman, 2006). Asking governments to radically alter their domestic security policies means reconciling with these other priorities, and foreign aid is often seen as a useful tool to compensate recipient governments for setting aside those priorities in order to crack down on terrorist activity. However, recent history calls into question the efficacy of this practice. During the summer of 2014 for example, many wondered how 30,000 Iraqi soldiers stationed in Mosul could simply collapse as the militant group ISIS advanced on the city with only 1,000 fighters. After all, the Iraqi government had received over $50 billion in foreign assistance since 2006, and the Iraqi army had been trained by US forces specifically to protect against threats such as this one. Militant groups also occupied large swathes of territory in Yemen and Mali in 2011 and 2012, respectively. These governments offered little to no resistance to these advances, despite having received billions of dollars in Western aid and military training specifically to help them combat terrorism. On the other hand, foreign aid has had positive effects in countries such as Indonesia, Bangladesh, and Colombia, whose governments have generally been quite earnest and competent in their counterterrorism efforts. Why is foreign aid such an effective counterterrorism tool in some countries, but so ineffective even counterproductive in others? In this paper, I attempt an explanation for this puzzle. I show that the examples described above represent a systematic trend of certain types of regimes deliberately manipulating the threat from terrorism to receive U.S. foreign assistance. Leaders in patronage-based regimes depend upon the distribution of rents and private benefits to regime insiders and rival factions to remain in power. If access to these rents dwindles, the leader risks being removed from power (or worse). The primary contribution of this paper is to provide a first cut at incorporating the domestic political structure of recipient regimes as a mediating factor in the foreign aid-terrorism relationship. In short, I 2

argue that due to the patronage-dependent nature of personalist rule, these types of regimes have incentive to exacerbate their terrorism problem, rather than eliminate it, in order to keep receiving aid. Thus, aid should increase terrorist activity in personalist regimes. The following section provides an overview of previous literature on the link between foreign aid and terrorism. Next, I present my argument about the role of political institutions in the relationship between foreign aid and terrorism. In the third section, I derive two implications from the argument and test them empirically. Results indicate that higher levels of U.S. foreign aid are associated with a significant increase in the number of anti-american terrorist attacks, as well as with longer terrorist campaigns in recipient countries characterized by personalist rule. In my concluding remarks, I offer some suggestions for future research. Foreign aid & terrorism Since the onset of the so-called war on terror in late 2001, and even during the Cold War years, foreign aid has been the United States preferred counterterrorism instrument. Aid is meant to reduce terrorism through three mechanisms which, while they may operate simultaneously, are very distinct. The first follows from the assumption, commonly heard throughout the war on terror, that poverty and misery drive individuals to join terrorist movements. As various measures of quality of life in recipient countries improve, the thinking goes, fewer people in these countries will resort to terrorism. Some empirical evidence suggests that wealth at the countrylevel is negatively correlated with political violence (e.g., Fearon and Laitin (2003); Collier and Hoeffler (2004); Honaker (2008) Thus, large influxes of aid will indirectly reduce violence at the macro level by improving the standard of living. Both Young and Findley (2011) and Azam and Thelen (2008) find that aid targeted to specific sectors education in particular can reduce transnational terrorism at the country level. However, other research suggests that the connection between socioeconomic variables and terrorism specifically is quite complicated. Both Krueger and Maleckova (2003) and Blair et al. (2013) show that individual-level support for militancy is lowest among the poor in Lebanon and Pakistan, respectively. Bueno de Mesquita (2005)offers a possible theoretical explanation for these findings. Blomberg, Hess and Weerapana (2004) and Li (2005) present cross-national evidence that wealthier and democratic countries experience more terrorism. 3

The former find that periods of economic difficulty are only associated with more terrorism in wealthier democracies. Others argue that aid can increase instability in recipient countries, either through rebel capture (e.g., (Dube and Naidu, 2010)) or by accentuating distributional conflicts (e.g., Tornell and Lane (1999); Grossman (2003)). Nunn and Qian (2014) show that food aid prolongs ongoing civil conflicts. In addition to improving socioeconomic outcomes, the purpose of aid is also often to bolster the capacity of the recipient state, enabling host states to more effectively confront terrorist groups. Piazza (2008) shows that failed states tend to produce more transnational terrorism. However, the connection between state capacity and terrorism is multifaceted. Hendrix and Young (2014) show that increasing the military capacity of a state actually makes that state more vulnerable to terrorism, particularly when bureaucratic capacity is low. In any case, the presumption that aid is effective at building capacity is questionable at best (e.g., Goldsmith (2001); Svensson (2000a); Easterly (2006); Moss and de Walle (2006)). The final mechanism pertains primarily to government actions, and is the one that has received the least attention in arguments linking aid to terrorism. Here, aid is used to increase the willingness of the host state to carry out counterterrorism on behalf of the donor state. Because military intervention by target states is so costly and rarely effective as a counterterrorism policy, the target state prefers to delegate counterterrorism duties to the states that host this activity (Azam and Delacroix, 2006; Bandyopadhyay, Sandler and Younas, 2011; Azam and Thelen, 2010; Fleck and Kilby, 2010; Boutton and Carter, 2014). The 1982 Congressional testimony of then Under Secretary of State James L. Buckley aptly illustrates this thinking: The marginal U.S. dollar loaned under [Foreign Military Sales] to the Thai or Turkish Army or Pakistani Air Force is a dollar that we otherwise would have to spend outright on our own forces to do a job that the Turks and Thais and Pakistanis can do better and at less cost. (quoted in (Siddiqa-Agha, 2001)). The donor state is essentially purchasing the counterterrorism services of the recipient. Even though it has been acknowledged for some time that foreign aid is strongly motivated by donors political and security goals (e.g., Morgenthau (1962); McKinlay and Little (1977)), few have sought to explore whether aid is actually effective at changing the behavior of the recipient. Tessman, Sullivan and Li (2011) show evidence that US military aid actually decreases security cooperation. Only Bapat (2011) and 4

Bandyopadhyay, Sandler and Younas (2011) have examined counterterrorism behavior specifically, with each showing that aid can actually have perverse effects on counterterrorism cooperation. Existing research on the aid-terrorism connection suffers from two major shortcomings. The first is that, with the exception of Bandyopadhyay, Sandler and Younas (2011) and Bapat (2011), none has taken into account the effect of aid on the counterterrorism incentives and behavior of the recipient government. Given that aid is meant in large part to change the behavior of the recipient in some way, this is a crucial omission. Second, existing treatments of the aid-terrorism relationship do not adequately consider variation in political structures across recipients. Scholars of development aid routinely emphasize the importance of recipient politics in determining aid effectiveness (e.g., Boone (1996)), and thus it is reasonable to think that the same logic would hold for counterterrorism aid. 1 These are important oversights that limit our theoretical and empirical understanding of the effect of foreign aid on terrorism, and may help explain the current confusion about the nature of the aid-terrorism nexus. I attempt to address these shortcomings in this paper. Recipient politics & aid effectiveness The overwhelming majority of aid effectiveness research is focused on the question of whether or not aid is effective at promoting economic growth. While there is currently no consensus on this relationship, a common theme throughout this literature is the notion that aid is more likely to have a productive impact in countries with institutionalized checks on government power (Svensson, 1999) or in systems in which the regime is accountable to a larger portion of its citizens (Boone, 1996; Smith, 2008). Burnside and Dollar (2000) concluded that foreign aid is likely to be effective in spurring economic growth only when allocated to regimes which pursue good policies. These institutions ostensibly lessen the probability that recipients will seek to use aid to enrich themselves and their political allies, rather than toward the benefit of the country. Thus, Wright and Winters (2010) rightly advocate for a focus on the politics of recipient states when thinking about aid effectiveness. A democratic leader, whose ability to remain in office is a function of the approval of a relatively 1 Recipient politics play a small role in the model presented in Bandyopadhyay, Sandler and Younas (2011), but the model does not consider variation across recipients. 5

large subset of society, should have greater incentive to use aid constructively. This approval is, in part, conditional upon the leader s performance and provision of public goods, which can include security and economic well-being. In the case of counterterrorism, such leaders should want to channel aid toward efforts to disarm terrorist groups and decrease overall levels of terrorism within the country; by failing to do so, the leader risks appearing either incompetent or deceitful to voters, thus jeopardizing his tenure in office. On the other hand, a dictator who fears removal by other means, and who requires the support of a smaller group in order to remain in office, may use aid to solidify support within this coalition (Bueno de Mesquita et al., 2003). Indeed, Licht (2010) finds that aid increases the long-term durability of authoritarian regimes. According to Bueno de Mesquita and Smith (2009, 312), and in line with Downs (1957), If faced with a contradiction between actions that enhance their own political welfare and actions that advance societal well-being...recipient leaders will select those policies that benefit themselves. Indeed, there is evidence suggesting that rent-seeking states actually reduce public goods provision when provided with aid (Boone, 1996; Svensson, 2000a; Wright, 2009), and that aid may have negative long-term effects on institutional development (Moss and de Walle, 2006). However, all autocracies are not created equally, and institutional variation within this category may be as great as the variation between dictatorships and democracies (Weeks, 2008). The following section examines in greater depth the survival strategies and the nature of political loyalty across dictatorships, and discusses how the institutional landscape in personalist regimes should lead to less effective counterterrorism aid. Personalism, aid, & regime survival Recent literature on authoritarian regimes generally agrees that dictators generate regime support and remain in power through the use of patronage to co-opt various factions by distributing rents; through repression; or through some combination of the two (Acemoglu, Robinson and Verdier, 2004; Gandhi and Przeworski, 2006). The tools available vary widely across regimes, and depend on availability of resources, the quality of institutions, as well as the position of the regime in relation to the military. In a dictator s ideal world, he could keep all rents for himself and simply rely on a competent, 6

docile military to crush any anti-regime movements. Yet Svolik (2012) identifies a fundamental moral hazard in authoritarian regimes between a dictator and his armed forces: If the dictator uses the military to repress domestic opposition, this signals to the military that 1) the leader is vulnerable; and 2) the military is indispensable to him. This knowledge enables the military to demand concessions from the leader that he may not want to grant, with the implicit threat that should he not grant them, he will be overthrown. Furthermore, empowering the armed forces to crush domestic threats also facilitates coordination among the security forces, making it easier for various factions to organize and carry out a successful coup (Powell, 2014). The desire to avoid these dilemmas tempers dictators use of the army for internal security. 2 A civilian dictator seeking to remain in power while using repression sparingly must therefore to mobilize loyalty and support among elites and society through the distribution of rents. This can take the form of clientelism/vote-buying (i.e., the distribution of material goods, such as food, fuel subsidies, or housing, to citizens in exchange for political support) or patronage, which is aimed at securing the support of elites and regime collaborators through promises of employment, lucrative business contracts, cash, or other favors (Stokes et al., 2013). While elites in all regimes engage in rent-seeking in one form or another, the nature of the process, as well as the importance of patronage for regime survival, are qualitatively different in personalist autocracies, and are the points of departure for this argument. By far the most common and durable form of dictatorship in the post-wwii era is the dominantparty autocracy (Geddes, Wright and Frantz, 2014). These dominant parties recruit members early in their careers through grassroots organizations to carry out party-related tasks, while benefits for long-term party membership political influence and material gain are realized by senior party members later in their careers. Thus, party membership is viewed as an investment and loyalty among key factions derives not just from rents and material benefits (which can be distributed 2 Military regimes are an obvious exception here, since it is the military itself that holds power, rather than a civilian leader. These types of dictatorships are therefore much less squeamish about using repression than their civilian counterparts, and have historically displayed the highest levels of it Davenport (2007); Escriba-Folch and Wright (2010). These regimes also differ from civilian-led regimes in the value they place on holding power. Where the primary goal of a civilian leader is to remain in power, the priority of a military junta is to maintain the unity and integrity of the military institution itself at all costs (Nordlinger, 1977). When military unity is jeopardized by continuing to hold power, the junta will often cede power to a civilian government after negotiating a return to the barracks. Thus, periods of military rule that do not evolve into personalist or party dictatorships tend to be significantly shorter than other types of regimes (Geddes, 2003). 7

without a party) but rather the promise of future benefits as a reward for past service and continued loyalty. The durable nature of the party lends this promise additional credibility, giving all members an enduring, rather than fleeting, stake in regime survival (Svolik, 2012). In the event of a crisis, the party provides a credible guarantee to in-groups that their long-term interests are best-served by remaining loyal to the regime (Smith, 2005, 431), thus preventing large-scale elite defection, even if short-term material benefits abate. 3 The continued dominance of Mexico s ruling PRI for nearly 20 years after the Latin American debt crisis is a testament to the loyalty of Mexican elites and voters to the party (Magaloni, 2006). Furthermore, the higher levels of bureaucratic capacity (see Figure??) found in dominant-party regimes enables them to compensate for revenue shortfalls by increasing taxes (Escriba-Folch and Wright, 2010). However, party members remain loyal to the regime for reasons other than immediate material gain; namely, the knowledge that they are better-off working to realize their ambitions within the party than attempting to overthrow it. As a result, the price of retaining the loyalty of key factions is lower for the leader in dominantparty dictatorships, and the consequences of crises and revenue shortfalls such as aid cutoffs or reductions are not fatal for the regime. On the other hand, personalist regimes lack well-developed parties. Political influence is instead concentrated in the hands of an individual, or a small group, who use their power to transfer a large fraction of society s resources to themselves. (Acemoglu, Robinson and Verdier, 2004, 162). Institutions such as parties and the military are marginalized, meaning that political privilege and career advancement depends upon a personal connection to the leader s inner clique. With power limited to this inner circle, and without a party to generate incentives for cooperation, acquiring political power as a member of the opposition entails re-ordering the regime. Politics thus resembles a prisoners dilemma in which the rivalries and struggles of powerful and willful men, rather than impersonal institutions, ideologies, public policies, or class interests, are fundamental in shaping political life (Jackson and Rosberg, 1984, 421). However, even while personalist leaders ( kleptocrats in Acemoglu, Robinson and Verdier (2004) terminology) pursue inefficient and ruinous policies, they are able to forestall challenges to their rule by distributing selective incentives lucrative contracts, privileges, subsidies, and other 3 Brownlee (2004) puts it aptly: [L]oyalty is the salve for today s losses just as it is the currency for future gains. 8

types of patronage to rival factions, making collective action among broad segments of society difficult. Bratton and van de Walle (1997) stress that the survival of a personalist leader in office hinges on his ability to distribute rents and material inducements to political rivals. Likewise, as long as their loyalty to the leader is rewarded with access to patronage, these rival factions have strong reasons not to overthrow the leader, and to mobilize their supporters to back the regime as well. These elites are complicit with the regime s corruption and misrule, and derive their livelihood from being loyal to the regime, and thus have nearly as much to lose as the leader should political order collapse (Bratton and van de Walle, 1997). However, if the leader s ability to reward loyalty with favors diminishes and elites access to patronage dries up, withdrawing their support and mobilizing against the leader becomes a more realistic and attractive strategy. Effects of revenue shortfalls When revenue does wane due to aid cutoffs, sanctions, or some other shock, the effects on stability are most immediate and severe in personalist regimes for several reasons. First, these regimes lack the party infrastructure necessary to generate cooperation among elites. Whereas a leader in a democracy or party-based autocracy can make credible promises of future patronage during crises, personalist leaders cannot. The glue holding personalist regimes together is the fragile network of rents and patronage that gives the ruler s allies and key rival factions access to the spoils of the regime. Should patronage diminish, there is no party or other mechanism to encourage elites to cooperate; any promises by the leader that the flow of patronage will resume in the future are non-credible. Thus, supporting the regime no longer benefits rival factions, and they defect immediately, and often mobilize the masses to do the same Escriba-Folch and Wright (2010). When revenue in Libya dipped during the late 1980s, this precipitated a series of challenges to Qaddafi s regime, including a military mutiny and dozens of assassination attempts (Brownlee, 2002). Along the same lines, Escriba-Folch and Wright (2010) show that the destabilizing effect of economic sanctions is most acute in personalist regimes. Where other types of regimes were able to generate additional revenue from taxes and other sources to compensate for sanctions-induced shortfalls, the lack of institutional capacity in countries ruled by personalist regimes made them unable to do so (see Figure??). In other words, the institutions in non-personalist regimes help lower the price of 9

Figure 1: Mean values of bureaucratic capacity (1984-2007) and logged coup probability (1970-2012) across regime type. I use the normalized bureaucratic capacity measure constructed by Hendrix and Young (2014), while data on coup probability are taken from Ulfelder (2014). loyalty such that a reduction in patronage resources in a personalist regime reduces the leader s ability to remain in power more than the same reduction in any other regime. Second, in the event that a revenue shortage does lead to mass protest or rebellion, using the military for repression is typically a bad strategy for personalist leaders who wish to remain in power. The moral hazard Svolik (2012) describes as inherent between dictators and their military is particularly acute for personalist leaders: Since personalist leaders rule under a constant coup threat (see Figure??), empowering the armed forces to engage in domestic repression also increases the risk of a coup at the hands of a united and emboldened military. While coup-fearing leaders prefer to keep their military divided, under-equipped, and docile, they tend to empower smaller, loyal paramilitary units often headed by family members or loyalists whose job is specifically to detect and foil potential coup plots. In fact, Powell (2014) notes that African leaders have historically refused to deploy these counter-coup units against insurgents, preferring instead to keep them close to the capital. This coup-proofing strategy may be effective at warding off coups, but it renders the conventional forces ineffective and uncooperative in the face of mass uprisings (Roessler, 2011; Powell, 2014). 10

Finally, personalist dictators inability to make credible promises, the absence of institutions, and their usually ruinous and exclusionary policies lead to disproportionately violent exits from office (see Moammar Qaddafi, among others). Escriba-Folch (2013) shows that the political careers of personalist dictators culminate in either imprisonment, exile, or death 61% of the time. By contrast, leaders in single-party regimes and monarchies tend to enjoy the safest post-tenure fate among autocracies. Leaders in military regimes often negotiate leadership changes among a coalition of military commanders within the junta, or simply return to the barracks after negotiating a transition to a civilian regime (Geddes, 2003). Democratic leaders, who are subject to regularized mechanisms for the transfer of power, are also unlikely to be punished after leaving office. Personalist leaders rule thus sits on a razor s edge: it is totally dependent upon a continual source of revenue with which to buy the support of rival factions. Due to the lack of other viable survival strategies, any disruption in the flow of patronage goods can lead to the rapid fall of the regime, which often includes the imprisonment or death of the leader. The primary goal of personalist leaders will therefore be to procure foreign aid; welcoming or deliberately amplifying the presence of terrorism within their borders particularly the sort of terrorism that is of interest to the United States is an effective way of doing this. The following section discusses how donors inability to enforce aid conditions enables recipients to be deceptive about their counterterrorism efforts with relative impunity. Conditionality and donor commitment issues Any argument about foreign aid effectiveness must touch on the issue of aid conditionality. Donors almost always place some set of conditions on aid packages, usually in the form of a specified level of effort or outcome that must be achieved, with the implict threat that aid will be withdrawn or cut if the terms are not met. It is generally agreed upon within the aid literature that credible conditionality is necessary if aid is to achieve donor goals (Easterly, 2006; Dijkstra, 2002; Morrisey, 2004; Svensson, 2000b, 2003; Collier, 2007). Some have argued that states which are heavily dependent upon foreign aid for revenue are more likely to comply with aid conditions, as their political survival is more contingent upon the availability of resources and would be jeopardized by economic punishment (Escriba-Folch and Wright, 2010; Girod and Tobin, 2014). However, others 11

also argue that it is very difficult, if not impossible, for donors to issue credible threats to withdraw aid, either due to informational asymmetries resulting from donors inability to monitor recipient behavior, or if donor s security interest in the recipient state is too great to risk withdrawing aid. The former does apply here, as donor attempts to monitor recipient behavior often carry great risks for personnel, and are usually viewed by the recipient as violations of sovereignty. However, the latter reason is most important for this argument. In cases in which instability or the threat of terrorism, especially anti-us terrorism, is present in a country, the United States will not want to risk withdrawing aid, as this may exacerbate the problem. When US officials debated a possible aid cut-off to Egypt in July 2013, many balked at the idea due to contractual obligations and fears that the instability would spread. Don t think the Egyptian military doesn t know that how we provide the aid constrains us from cutting it off easily, one official was quoted as saying, They know this stuff a hundred times better than us (Londoño and Yeager, 2013). The knowledge that the flow of foreign aid will likely not be jeopardized as long as a terrorist threat is present should negate any fear among personalist leaders that deliberately stoking an insurgency could endanger his grip on power. Personalist regimes often have foreign patrons who have a security interest in stability. If an terrorist threat or insurgency threatens the regime, the foreign power will come to the aid of the regime to prevent its collapse. Reliance upon a foreign patron can in some cases be damaging for a personalist leader s ability to survive an uprising if the donor is squeamish about repression (Brownlee, 2002). But in many cases, they can actually help personalist regimes remain in power, as the US did with Nicaraguan leader Somoza during the 1960s and 70s. US security goals in these countries often depend on the survival of the regime, and thus repression is only lightly punished, if at all. The 2011 anti-regime instability in Yemen offers a recent example. Speaking about the possibility of that regime s collapse, US Secretary of Defense Robert Gates admitted that if that government collapses, or is replaced by one who is dramatically more weak, then I think we d face some additional challenges out of Yemen, there s no question about it. It s a real problem (Presse, 2011). US counterterrorism assistance to Yemen continued to increase even after hundreds of Al-Qaeda suspects escaped from prison between 2006 and 2011, in a series of jailbreaks that many believed to be the work of the Saleh regime itself (Scahill, 2012; Johnsen, 2013). Yemen was lightly punished in 2011: US military aid (but not economic aid) 12

was suspended briefly after a brutal crackdown on civilian protesters, but was resumed shortly thereafter (Lubold and Schactman, 2013). These governments are well aware that the US depends on them for counterterrorism, and thus will be reluctant to stop providing aid even if they are known to be deceptive about their efforts. Hypotheses The argument detailed above points to expectations about the quality and/or quantity of counterterrorism effort undertaken by a recipient of U.S. foreign aid. A direct test of these arguments would necessitate very detailed data on the counterterrorism behavior and capacities of U.S. aid recipients. Unfortunately, the concept of counterterrorism remains ill-defined by scholars, spans across numerous policy areas, and can be quite difficult to observe. Counterterrorism effort could include overt activities such as military and police raids on terrorist hideouts; arrests; assassinations; or passage of anti-terrorism legislation. However, more covert, less observable actions, such as back-channel negotiations with a terrorist group; torture; increases in police budgets; or the freezing of financial assets could also be included, and thus any measure of counterterrorism that relies on observable indicators is inevitably incomplete. Furthermore, observing a government s attempts to deliberately manipulate the level of terrorism present which is something that I argue personalist leaders are doing is very difficult. Perhaps jailbreaks would happen more frequently, as we have seen in Libya, Iraq, Yemen, and Pakistan recently (Reuters, 2013; Johnsen, 2013; Salman and Parker, 2013), although again, systematic data on these events is sparse. While systematic measures of these concepts are not currently available, some implications of this argument can be evaluated empirically. The theory of aid, personalism, and counterterrorism that I discuss points to two implications about terrorism we should expect to see in states receiving foreign aid from the United States. Specifically, the argument implies that states receiving aid for counterterrorism are more likely to use that assistance for counterterrorism if they are not governed by a personalist regime. Given a greater emphasis on counterterrorism in these states, as well as less desperation to continue receiving foreign aid, we can expect better counterterrorism outcomes (from the donor s perspective) from these regimes. Thus, terrorist groups may not to last as long, or be more likely 13

to be disarmed at the hands of the regime in a given year. However, I expect that personalist host regimes will abuse the foreign aid relationship, instead using the external revenue to remain in power by fueling the patronage machine, while playing-up the level of terrorism within their borders in order to keep the aid flowing. Maintaining this level of threat gives personalist leaders a better guarantee of future resources with which to stabilize their regimes. Therefore, we should expect that groups based in these states to last longer. This leads to my first hypothesis: Hypothesis 1. Groups based in states governed by personalist regimes will be less likely to fail by splintering or police force in a given year as US foreign aid increases. My second hypothesis concerns how personalism conditions the effect of foreign aid on the level of activity by terrorist groups in recipient countries. It is reasonable to point out that the goal of this aid may not necessarily be to completely eliminate the groups, but rather to enable the regime to prevent attacks against U.S. citizens and interests. As long as its interests remain safe, the U.S. may be indifferent to the prolonged presence of terrorist groups in recipient countries. However, given their desire for continued U.S. assistance to maintain their patronage networks, some recipient states may seek further assurance of future aid. Fearful that the United States would devote its aid resources elsewhere once it feels its interests are safe from attack, personalist regimes may turn a blind eye toward attacks against U.S. interests, and may even engage in deliberate sponsorship and support of such activity (Byman, 2005). While such a strategy risks incurring the wrath of the United States, it is inherently covert and thus quite difficult for target states to verify (Schultz, 2010). Also, as discussed above, such regimes have reason to be confident that an intolerable level of punishment will not be forthcoming. Therefore, in addition to longer-lasting groups, we might also expect more terrorist attacks against U.S. interests in states under the authority of personalist regimes. This leads to the following hypothesis: Hypothesis 2. Personalist regimes will experience more anti-us attacks in a given year as the amount of US foreign aid increases. In the following section, I outline the data, research design, and methodology used to assess these hypotheses. 14

Research Design Dependent Variables To test my hypotheses about the effect of U.S. aid on a recipient state s incentive to combat insurgent groups, I rely on two types of data. First, I use a modified version of the data used in the Jones and Libicki (2006) study on how terrorist groups end in order to test my hypothesis about group duration. For the second set of hypotheses in which terrorist attacks are the quantity of interest, I take data from the Global Terrorism Database (GTD), which contains information on terrorist attacks between 1970 and 2010. These variables are described in detail below. Group Duration The Jones and Libicki (2006) data identify 648 violent non-state groups that exist or have existed throughout the world between 1968 and 2006. Of these groups, 45 were based inside the United States and are thus not included in the present study. This leaves 603 internationally-hosted, nonstate actors during this time period. I converted the data into binary time-series cross-sectional data, with one observation per year per country that the group was in existence, according to the Jones and Libicki data. This type of arrangement is best to analyze the factors that cause a group to fail or endure, which is what I seek to explain here. In a few cases in the data, groups had legitimate bases in more than one country. When this was the case, I sought confirmation of multiple bases from U.S. Congressional Research reports, the Global Terrorism Database (START, 2012), and from histories of the groups themselves when available. Where appropriate, I coded a group s presence in a second or third country from the date that the secondary or tertiary base was established. With a few exceptions, these groups began operations in the multiple countries more-or-less simultaneously. For example, the Salafist Group for Call and Combat has carried out attacks in Morocco, Algeria, Mali, and Mauritania and is coded accordingly. This coding strategy is appropriate because the United States recognizes that these groups often traverse freely between countries, and its aid patterns reflect this (Ellis, 2004; Boutton and Carter, 2014). Because I am trying to explain action taken by the recipient state against these groups, my 15

dependent variable is the failure of a group in year t due to actions taken by the state. The Jones and Libicki data code five different types of group failure: accomplishment of group goals; entry into politics; defeat by the military forces of an outside state; defeat by police action; and internal dissolution/splintering. Only the latter two are well-suited to my purposes, since these are the result of actions taken by the host state, which is ostensibly what donors have in mind when providing foreign aid to states hosting terrorist groups. Therefore, I combine police failures with splintering to form my dependent variable. This variable takes a value of 1 if the group fails in year t, and a 0 otherwise: 0 if group is active in year t or ends for reasons unrelated to host state action y i,t = 1 if group fails in year t due to internal dissolution or police force When a group fails by splintering or police force, it leaves the data and is not included in any subsequent years; there are 206 failures in the data. Groups which fail for any of the other three reasons remain in the data until they fail according to the Jones and Libicki (2006) data, at which point they drop out of the data without ever having failed in the model. Attacks Data on terrorist attacks are taken from the Global Terrorism Database (START, 2012), which is useful as it contains information on the location, year, and primary target of over 80,000 terrorist attacks from 1970 to 2012. I construct two separate dependent variables based on the primary target of each attack: 1) a count of the number of attacks directed at U.S. interests in a host country in a given year; and 2) attacks directed at non-u.s. targets. It is important to note that GTD classifies terrorist attacks according to the country in which they take place, and not according to the nationality or origin of the perpetrator(s) (START, 2012). For example, a cross-border attack in Nairobi by the Somali group al-shabaab would be associated with Kenya, rather than Somalia. This coding does not present a problem here, since the U.S. responds to terrorism primarily by allocating foreign aid to countries in which terrorist activity takes place (Boutton and Carter, 2014). Descriptive statistics for the three attacks variables are presented in the appendix. 16

Main Independent Variables U.S. Aid The theory outlined above specifies that levels of U.S. foreign aid provided should influence the likelihood of a group failing in year t, conditional upon whether the recipient state is a personalist regime. I construct three separate explanatory aid variables, one each for levels of military, economic, and total aid (the first two categories added together), 4 all of which are taken from the U.S. Agency for International Development s Greenbook (United States Agency for International Development, 2014) and measured in constant 2009 U.S. dollars. I make a couple of adjustments to these variables. First, raw aid variables tend to be highly right-skewed. Values for total U.S. aid, for example, range from $0 to over $17 billion, with a mean of $754 million. To correct for this, I take the natural log of (1 + aid) for each variable, a common approach in past aid literature (Bueno de Mesquita and Smith, 2007; Lai, 2003). This smooths out the distribution of the aid variables, and is also theoretically appropriate: a change in aid from $0 to $10 million is likely to have a much larger impact than an increase from $100 million to $110 million. Taking the natural log of the aid variables thus allows for the decreasing marginal returns to aid. Finally, I lag each aid variable by one year in order to account for the fact that aid received in year t is unlikely to affect the fate of a given group in year t. It is more likely that aid given to a state in year t 1 will be the relevant factor in the fate of a group in year t, as aid must work its way through a bureaucracy and may take time to have an impact in the recipient country. 5 Summaries of the aid variables are presented in the appendix. Personalism To classify personalist regimes, I draw upon data on authoritarian regimes and transitions recently assembled by Geddes, Wright and Frantz (2014), who follow the classification scheme presented in Geddes (2003), updated to 2010. Like other regime types, personalist regimes are classified based 4 Importantly, the argument I make above applies equally to economic, military, and defense aid, since it has been shown that the presence of terrorism leads to increases in all types of U.S. foreign aid (Boutton and Carter, 2014). Therefore, although I model them separately and then together as is common in past aid literature, I do not expect the different types of aid to have substantively different effects on the probability of group failure. 5 I also estimate models using aid normalized by population, or aid per capita, for military, economic, and total aid. The results of these models are available in an appendix. 17

on answers to a certain set of questions about that regime. Questions relevant to personalist regimes are 1) Does the leader lack the support of a party? 2) If there is a support party, was it created after the leaders accession to power? 3) If there is a support party, does the leader choose most of the members of the politburo-equivalent? 4) Does the country specialist literature describe the politburo-equivalent as a rubber stamp for the leader? 5) If there is a support party, is it limited to a few urban areas? 6) Was the successor to the first leader, or is the heir apparent, a member of the same family, clan, tribe, or minority ethnic group as the first leader? 7) Does the leader govern without routine elections? 8) If there are elections, are they essentially plebiscites, i.e., without either internal or external competition? 9) Does access to high office depend on the personal favor of the leader? 10) Has normal military hierarchy been seriously disorganized or overturned? 11) Have dissenting officers or officers from different regions, tribes, religions, or ethnic groups been murdered, imprisoned, or forced into exile? 12) Has the officer corps been marginalized from most decision making? 13) Does the leader personally control the security apparatus? (Geddes, 2003, 227) 6. A regime is given a score for each regime category (personalist, single-party, monarchy, and military regime) equal to the ratio of affirmative answers to total answers, and the typology is assigned according to the highest score. In other words, all regimes have a score for each regime type, but the highest score determines that the regime s designated typology. Some have high scores in multiple categories. For example, Indonesia is classified as a party-based autocracy during Suharto s reign, but the regime also displayed elements of personal and military rule, and thus is also considered a hybrid between the three. I consider military or party-based regimes with tinges of personalism, such as Syria or Egypt, to be quite different from the pure personalist rule of Mobutu or Marcos, and therefore only pure personalist regimes are considered as such in this paper. The non-personalist category thus includes party- and military-based regimes, monarchies, democracies, and non-regime categories such as warlord rule (i.e., Somalia 1992-present), foreignoccupied (i.e., Iraq 2004-2010), and provisional (i.e., Georgia 2004). 7 6 Svolik (2012) likens personalism to a more continuous, dynamic process that occurs in stages. The culmination of the process is what he calls established autocracy, a point at which the dictator holds all power. The view of personalism presented here, while expressed as a discrete regime type, accurately reflects a regime that has passed a number of benchmarks along Svolik s path to established autocracy. While it is certainly true that there is variation even within the personalist category, capturing this is not possible given data limitations. 7 The distribution of the personalism variable in each data set is shown in Table 2 of the appendix. 18

Because I expect the effect of foreign aid to be conditional on personalism, I multiply the personalist indicator by each aid variable to create the interaction terms. In the duration data, there are 6,522 group-years in the duration data across 92 countries and 603 groups between 1968 and 2006, though missing data for countries such as Angola, Mozambique, and Namibia before independence cause me to lose 158 group-years. In the fixed-effects models, 1,110 observations are dropped from cases in which there is no variation on the dependent variable, leaving me with 5,254 observations in the FE models. The attack-count data set contains 5,604 country-year observations, extending from 1970 to 2010, though observations dropped in fixed-effects models bring the number to 5,202. Endogeneity Azam and Delacroix (2006) report a positive relationship between aid and terrorism because donor countries allocate more aid to countries that experience high levels of terrorism, and which are likely to continue experiencing elevated levels of terrorist activity in the future. Any estimate of the effect of aid on terrorist activity that does not account for endogeneity will likely suffer from bias as a result. Most scholars have relied on instrumental variables and two-stage regressions in order to achieve a plausible claim of exogeneity in econometric models (Sovey and Green, 2011). However, this method has been shown to be problematic in non-linear regression (Terza, Basu and Rathouz, 2008). Thus, I follow Terza, Basu and Rathouz (2008), Azam and Delacroix (2006), and Young and Findley (2011) by estimating a series of linear regressions in which a set of exogenous, aid-relevant covariates are used to explain the endogenous variable (aid). The residuals from these reducedform equations are then included in the equations for the attack count models. The first-stage aid regressions are presented in the appendix. Results Aid, personalism, and group duration To analyze the conditional effect of U.S. aid on group duration, I estimate a series of grouped duration models in which the outcome variable is either failure by host state police force or internal 19

dissolution (=1), or not (=0). The model is represented by the following equation: P (F ail = 1) = F (β 1 ln(aid) it 1 +β 2 P ersonalist it 1 +β 3 ln(aid) it 1 P ersonalist it 1 +β 4 X it 1 +γ i +ɛ), (1) where i indicates the country, t the year, F ( ) is the cdf of the logistic distribution, X is a vector of control variables, γ is the set of country fixed-effects, 8 and ɛ is an error term. All time-varying independent variables are lagged. The estimation gives the effect of the independent variables on the probability of group failure in a given year, so a negative (positive) coefficient means that variable makes group failure less (more) likely. The fixed-effects models control for time-invariant, country-specific factors, and estimate within-country variation to ensure that the results are not simply a reflection of cross-national variation. The results of these models are shown in Table 1. 9 Results of the duration models are shown in table 1. In all, Table 1 contains the results of six logit models, two each (with and without fixed effects) for US economic, military, and total aid per capita, estimated separately. Total U.S. aid by itself is positive but insignificant in all but the non-fixed effects model for military aid, in which the effect is negative and significant. This indicates that increases in US military aid per capita to nonpersonalist regimes may lead to slightly longer-lasting terrorist campaigns. The value of primary interest is the coefficient estimate for the aid-personalism interaction term, which gives us the effect of aid on group duration in personalist regimes. In all six of the models, the effect is negative, and is significant in all but the economic aid models (the effect falls just short of significance in the fixed effects model), as anticipated given hypothesis 1. Increases in aid to personalist regimes are associated with lower probabilities of terrorist group failure in these states. The estimates for personalist regimes receiving no aid are negative, though they fall just short of statistical significance in the fixed-effects models groups based in non-aid-receiving personalist regimes are less likely to fail in a given year. Given that, it is still the case that aid to personalist regimes is associated with even lower probabilities of failure than if they received no aid at all. 8 Results of a Hausman test indicate that fixed-effects models outperform random effects models, though I report the results of both here. 9 Results of rare-events logit estimations are identical. 20

Table 1: Aid per capita, personalism, and group duration Total Economic Military Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Intercept -3.84* -3.72* -4.39* [1.88] [1.72] [1.85] ln(aid per capita) t 1 0.02 0.01 0.02 0.01 0.01-0.08* [0.03] [0.02] [0.02] [0.02] [0.02] [0.03] P ersonalist t 1-1.63-0.78* -1.39-0.79* -1.28-1.14* [0.86] [0.37] [0.89] [0.37] [0.82] [0.45] Aid p.c. P ersonal t 1-0.25* -0.08* -0.12-0.05-0.09-0.08* [0.09] [0.04] [0.07] [0.04] [0.06] [0.03] Democracy t 1-0.50-0.62-0.19-0.56 0.26-0.66 [0.51] [0.37] [0.49] [0.37] [0.49] [0.37] Mil. regime t 1-1.62* -1.13-1.20-1.08-1.26-1.13 [0.72] [0.62] [0.69] [0.62] [0.68] [0.61] ln(gdp t 1 ) 0.52 0.39* 0.72* 0.41** 0.70* 0.42* [0.28] [0.12] [0.27] [0.12] [0.27] [0.12] ln(population t 1 ) 2.45* -0.02 1.97* -0.05 2.08* -0.01 [0.99] [0.17] [0.96] [0.16] [0.97] [0.16] ln(number of groups) 0.05 0.01 0.04 0.01 0.04 0.01 [0.03] [0.02] [0.03] [0.02] [0.03] [0.02] Cold W ar 0.71* 0.21 0.78* 0.22 0.79* 0.23 [0.30] [0.30] [0.30] [0.30] [0.30] [0.29] P ost 911-0.24-0.05-0.31-0.09-0.25-0.07 [0.24] [0.22] [0.24] [0.22] [0.25] [0.22] Sponsored -0.30-0.27-0.26-0.26-0.26-0.23 [0.27] [0.18] [0.27] [0.18] [0.27] [0.19] T erritorial 0.44-0.12 0.42* -0.15 0.44-0.11 [0.26] [0.26] [0.26] [0.27] [0.26] [0.26] Regime change 0.42 0.25 0.39 0.25 0.40 0.26 [0.23] [0.18] [0.23] [0.18] [0.23] [0.18] Status quo -0.17-0.89-0.11-0.88-0.16-0.89 [0.67] [0.69] [0.67] [0.68] [0.67] [0.69] Country size -0.02 0.00-0.02 [0.13] [0.15] [0.15] Ruggedness 0.13 0.12 0.11 [0.22] [0.21] [0.21] Africa -1.07* -1.02* -1.11* [0.45] [0.44] [0.44] Asia -0.65-0.62-0.67* [0.34] [0.35] [0.34] Europe Oceania 1.44 1.30 1.39 [0.76] [0.71] [0.75] North America -1.01* -1.02* -1.04* [0.42] [0.45] [0.42] South America -0.20-0.28-0.26 [0.34] [0.38] [0.32] Country fixed effects Yes No Yes No Yes No Log Likelihood -610.79-746.16-613.41-746.84-613.95-745.95 χ 2 190.47 185.21 184.15 N 5,254 6,364 5,254 6,364 5,254 6,364 * indicates significance at p < 0.05 Numbers in brackets are standard errors To save space, the cubic time polynomials are not shown 21

Figure 2: Effect of US aid in non-personalist regimes Figure 3: regimes Effect of US aid in personalist To get a sense of the substantive meaning of these findings, Figures 2 and 3 show what we can expect as US aid increases in personalist and non-personalist regimes. 10 Figure 2 shows that as US aid per capita increases over its full range (i.e., from $0 to $3,276 per capita), the probability of group failure increases slightly, but the effect is nowhere near statistical significance (as indicated by the small and insignificant coefficient estimate for ln(aid per capita) t 1 ). Thus, aid to non-personalist regimes appears to make little difference for the duration of terrorist groups in those states. However, Figure 3 demonstrates that increases in US aid to personalist regimes are associated with significant increases in the longevity of terrorist campaigns in those states (lower probability of group failure). This is in line with my expectation that personalist regimes may seek to prolong terrorist activity as a source of external revenue, and provides further support for Hypothesis 1. Aid, personalism, & terrorist attacks The results in the previous section suggest that U.S. foreign aid in some cases leads to longer terrorist campaigns. It is fair to point out, however, that the United States may have more conservative goals when allocating aid for counterterrorism purposes. Specifically, the purpose may not necessarily be to encourage or bring about the complete elimination of terrorist groups, but may instead be the (arguably more realistic) aim of preventing US interests in the recipient country from being 10 These plots are based on the coefficients in Model 2 of Table 1. 22