Journal of Peace Research http://jpr.sagepub.

com/

Transnational terrorism, US military aid, and the incentive to misrepresent

Navin A Bapat Journal of Peace Research 2011 48: 303 originally published online 8 April 2011 DOI: 10.1177/0022343310394472 The online version of this article can be found at: http://jpr.sagepub.com/content/48/3/303

Published by:
http://www.sagepublications.com

On behalf of:
Peace Research Institute Oslo

Journal of Peace Research Replication Data

Additional services and information for Journal of Peace Research can be found at: Email Alerts: http://jpr.sagepub.com/cgi/alerts Subscriptions: http://jpr.sagepub.com/subscriptions Reprints: http://www.sagepub.com/journalsReprints.nav Permissions: http://www.sagepub.com/journalsPermissions.nav

>> Version of Record - May 25, 2011 OnlineFirst Version of Record - Apr 8, 2011 What is This?

Downloaded from jpr.sagepub.com at TOURO COLLEGE LIBRARY on January 5, 2014

jour nal of

peace
R
E S E A R C H

Transnational terrorism, US military aid, and the incentive to misrepresent

Navin A Bapat University of North Carolina Chapel Hill

Journal of Peace Research 48(3) 303318 The Author(s) 2011 Reprints and permission: sagepub.co.uk/journalsPermissions.nav DOI: 10.1177/0022343310394472 jpr.sagepub.com

Abstract This study proposes a strategic explanation for the USAs continued provision of military aid to host states with problems of terrorism, despite its poor empirical record. Using a game theoretic model, I demonstrate that US military aid creates a moral hazard problem. If host states are provided with the tools to pacify their territory only if terrorist campaigns are ongoing, but will lose this aid once the problem of terrorism ceases, host states have little incentive to accelerate the demise of terrorist groups. However, the model demonstrates that while military aid does not accomplish the US goal of disarming terrorists, military aid is effective at preventing host states from negotiating with terrorist organizations. The provision of military aid provides a disincentive for host states to reach a negotiated settlement with terrorist groups, and therefore prevents terrorist organizations from altering the status quo that is favorable to the USA. This suggests that while military aid may not be effective at actually disarming terrorist groups, it can be effective at keeping terrorist groups out of power. These hypotheses are tested using the Jones & Libicki (2008) data on terrorism from 1997 to 2006. The empirical results support the conclusions of the theoretical model. Keywords conflict, military aid, terrorism Shortly after the Obama Administration assumed office in 2008, the Interagency Policy Group recommended an increase in military aid to Pakistan to bolster its fight against militant groups.1 The goal of the aid was to give the Pakistani government the tools it needed to disarm Al-Qaeda and Taliban fighters that were responsible for continuing attacks in Afghanistan. This policy follows a larger consensus that military aid to host states from which transnational terrorists operate is a valuable tool in accomplishing the US goal of fighting terrorism (Carothers, 2003; Walt, 2001). However, several studies examining the link between foreign aid and terrorism cast considerable doubt on these conclusions (Azam & Delacroix, 2006; Berrebi, 2003; Kruegar & Maleckova, 2003; Stotsky, 2008). While numerous studies argue that military aid is simply ineffective, others argue that
1

military aid is actually counter-productive. These conclusions raise the question: why, despite theoretical and empirical evidence, do policymakers continue to emphasize the importance of military aid as an instrument of counter-terrorism? This study proposes an explanation for the USAs continued provision of military aid to host states with problems of terrorism, despite its poor empirical record. Using a game theoretic model, I demonstrate that US military aid creates a moral hazard problem. If hosts are provided with the tools to pacify their territory only if terrorist campaigns are ongoing, but will lose this aid once the problem of terrorism ceases, hosts have little incentive to accelerate the demise of terrorist groups. Instead, hosts have incentives to play up the threat of terrorism and demand US military support to keep terrorists at bay. However, the model demonstrates that while military aid does not accomplish the US goal of
Corresponding author: bapat@unc.edu

Council on Foreign Relations. March 2009. White Paper of the Interagency Policy Groups Report on US Policy toward Afghanistan and Pakistan (http://www.whitehouse.gov/assets/ documents/Afghanistan-Pakistan_White_Paper.pdf).

Downloaded from jpr.sagepub.com at TOURO COLLEGE LIBRARY on January 5, 2014

304

journal of PEACE RESEARCH 48(3)

disarming terrorists, military aid provides a disincentive for hosts to negotiate with terrorists. This suggests that while US military aid is unlikely to shorten terrorist campaigns, military aid can be seen as a useful antiterrorist instrument that prevents the groups from altering the status quo that is favorable to the USA. I develop this argument in four steps. First, I outline the literature on foreign aid. Second, I discuss these findings as they relate to the question of how military aid affects the behavior of both hosts and terrorist groups. Third, I develop and solve a game theoretic model of US support for hosts facing transnational terrorists, and derive two testable hypotheses regarding the effect of military aid on terrorist campaigns. I next test these hypotheses by combining data on US disbursements of military aid from the US Overseas Loans and Grants, Obligations and Loan Authorizations Greenbook with Jones & Libickis (2008) data on terrorist campaigns. I then discuss the empirical results and conclude with a general discussion of how the USA uses military aid not necessarily to defeat terrorists, but rather to prevent hosts from accommodating terrorists.

Military aid as an instrument of counterterrorism

According to the US Greenbook, military aid consists of either loans or grants that are intended to purchase US weapons, direct military transfers, or training of the military personnel of the recipient country.2 From the period from 1946 to 2008, the US Greenbook identifies several types of aid granted to states under this program, including counter-narcotics assistance provided to Colombia and the provision of helicopters to Pakistans military. Several policymakers in the United States argue that this type of military aid can be effective in inducing states to adhere to US foreign policy objectives (Alesina & Dollar, 2000; Palmer & Morgan, 2006). In exchange for military equipment or training, the USA could require recipient states to liberalize or support US foreign policy (Cingranelli & Pasquarello, 1985; Meernick, Krueger & Poe, 1998). A classic example of this is the military aid provided to both Egypt and Israel in exchange for their compliance with the Camp David Peace Accords. In exchange for making peace, both states were promised billions of dollars in military aid. These accords led several policymakers and scholars to
USAID Economic Analysis and Data Services (2010) US Overseas Loans and Grants, Obligations and Loan Authorizations Greenbook (http://gbk.eads.usaidallnet.gov/).
2

argue that the use of foreign aid could be an effective substitute for costlier change-seeking policies, such as military force. Although the stated goal of US military aid is often benign, several studies demonstrate that aid typically produces adverse consequences. Numerous studies argue that assistance from abroad creates a disincentive for states to pursue sound economic policy (Schudel, 2008; Svensson, 2000). Other studies demonstrate that foreign aid seems to decrease democratic development (Bueno de Mesquita & Smith, 2007; Cingranelli & Pasquarello, 1985). Additionally, a third consequence frequently identified by scholars of alliance politics is that military aid can create moral hazard (Lake, 1999). Recipient states can use their newfound capability from military aid to become more aggressive in their foreign policy, which may include making greater demands of their rivals. These empirical studies suggest that regimes which receive military aid appear to become less democratic, more prone to repression, and more belligerent toward their neighbors. It is easy to point to the behavior of anti-Communist recipient countries during the Cold War, which often possessed some of the worst human rights records as defined by both Freedom House and Amnesty International (Cingranelli & Pasquarello, 1985; McCormick & Mitchell, 1988; Poe, 1990).3 Applying these insights to the problem of transnational terrorism, we might expect that military aid may minimize terrorism, but might also exacerbate the conditions that led to the formation of terrorist groups in the first place. For example, several empirical studies demonstrate that countries with poor state capacity, heavy corruption, and low per capita GDP scores are likely to experience violent insurgencies (Fearon & Laitin, 2003; Hegre & Sambanis, 2004). These studies suggest that military aid increases the corruption of hosts and creates disincentives for hosts to invest in their population, which in turn may negatively affect economic growth. Similarly, since we see an inverse relationship between democratic development and military aid, one might conclude that military aid stifles democratization. Additionally, numerous studies examining foreign aid more generally argue that external assistance worsens the problem of terrorism. Although foreign aid is intended to alleviate poverty, which is considered a cause of terrorism, studies at the individual level indicate that

See Freedom Houses publication Freedom in the World 1978 2010; Amnesty Internationals annual State of the Worlds Human Rights 19622010.

Downloaded from jpr.sagepub.com at TOURO COLLEGE LIBRARY on January 5, 2014

Bapat

305

terrorists tend to be drawn from relatively wealthier individuals (Kruegar & Maleckova, 2003). Therefore, foreign aid may create an increased number of wealthier citizens while stifling democratic development, which may actually increase terrorist recruiting. These observations raise the question: why do policymakers continue to use military aid if there is such an abundance of evidence that it is unlikely to reduce the problem of terrorism? One possibility may be that scholars have a relatively narrow definition of what constitutes success against terrorism. Typically, scholars view a policy as a successful antiterrorist instrument if it reduces the number of terrorist attacks, decreases the number of groups that are operational, or contributes to the collapse of a group (Azam & Thelen, 2010; Bandyopadhyay, Sandler & Younas, 2009). While these are certainly reasonable metrics, it is possible that an additional goal of military aid is simply to prevent terrorists from accomplishing their objectives. For example, military aid may not be successful in preventing terrorists from engaging in attacks, but may be successful in preventing a government from pursuing conciliatory policies toward these groups. Military aid might also increase the length of time that terrorists must fight in order to accomplish their strategic objectives. Since most groups collapse quickly, this increase in duration may make it impossible for terrorists to accomplish their goals (Bapat, 2005; Cronin, 2009; Jones & Libicki, 2008). If we therefore expand our definition of success so that it is not limited to just a reduction in terrorist attacks, we may see that military aid gives hosts the power to both resist the demands of terrorists and endure the cost of conflict. To illustrate with an example, consider the case of the US/Yemeni relationship. In 2000, Al-Qaeda launched an attack against the USS Cole. This attack, while not particularly crippling, demonstrated that Al-Qaeda was becoming increasingly brazen in its activities. In response, the USA increased its military support for the Yemeni government.4 If we examine the effect of this policy after ten years, the fact that Al-Qaeda still appears to be operational within Yemen might lead to the conclusion that the military aid provided to the Yemeni government was a failure. However, if we consider that Yemen is not a particularly strong state, whereas Al-Qaeda and its affiliates continue to be a relatively stronger group, an alternative policy for Yemen might have been to negotiate a deal with Al-Qaeda in which

it allowed the group to conduct anti-American attacks from its territory. However, with US military support, Yemen continues to profess its support for US policy objectives. In this case, even though Al-Qaeda has not been disarmed, one might consider the fact that Yemen remained loyal to the USA as a policy success. These arguments raise the question: can military aid produce success in terms of (1) preventing hosts from deviating from US policy objectives, and (2) disarming operational terrorists. In the following section, I formally model the US decision to support various hosts from which transnational terrorists operate. The model develops specific hypotheses as to how military aid affects the prospects for US success against terrorists.

Model
Figure 1 presents a stylized model of a transnational, anti-American terrorist campaign that occurs in a foreign host. The conflict is simplified to three players: the US government US , the host H , and the terrorists T that operate within H s territory.5 H begins the game in complete control of his territory, which allows him to distribute the territorys benefits to only his supporters. Additionally, H s control of the territory provides some foreign policy benefit to US , perhaps in the form of trade benefits or a military alliance. On the other hand, T seeks to destabilize H s government and assume sovereignty over the territory for itself, which would allow it to revise H s foreign policy such that it no longer provides this benefit to the US : We can therefore normalize H s utility function such that H receives a payoff of 1 for each round in which H survives, and a payoff of 0 if H is destabilized by T : Similarly, assume that T s utility function is normalized such that T receives a payoff of 0 for each round in which H survives, and receives a payoff of 1 if H destabilizes. In addition to facing opposition from H ; T also faces opposition from US , which seeks to minimize T s control over the territory. Since US benefits from keeping H in power, the primary objective of US is to keep T out of power.6

5 6

US Aid 20102012 Yemen Country Strategy (http://pdf.usaid.gov/ pdf_docs/PDACP572.pdf).

I refer to US as she, H as he, and T as it. A case example of this situation might be the USA/Saudi relationship. The host (the Saudi monarchy) is in control of its territory and distributes the benefits of ruling largely to members of the monarchy. In response, Al-Qaeda and other terrorists organized to destabilize the regime. Since keeping the monarchy in power benefits the United States, American foreign policy seems oriented to keeping Al-Qaeda out of power.

Downloaded from jpr.sagepub.com at TOURO COLLEGE LIBRARY on January 5, 2014

306
US 0 1

journal of PEACE RESEARCH 48(3)

Offer Aid x

T Reject Accept

Negotiate

Offensive

pt/(1-x)+((1-pt)/(1-x))(-a) - x ; pt/(1-x) - cH - ; 1-(pt/(1-x)) - cT)

Defensive

pt/(1-x)+ ((1-pt)/(1-x))(-a) - x; pt/(1-x) - cH - ; 1- (pt/(1-x)) - cT

~Attack Center

T Attack Center

1 - x; 1; 0 US ~Sustain Aid T ~Attack Center Attack Center -a; pt/(1-x); 1-(pt/(1-x)) pt + (1-pt)(-a); pt - cH; 1 - pt - cT Sustain Aid ~Attack Center - - x; pt/(1-x); 1- (pt/(1-x)) pt/(1-x) +((1-pt)/(1-x))(-a) - x ; pt/(1-x) - cH; 1- (pt/(1-x)) - cT T Attack Center pt/(1-x) +((1-pt)/(1-x))(-a) - x ; pt/(1-x) - cH; 1- (pt/(1-x)) - cT

Figure 1. Model of US military support for hosts facing terrorist campaigns

Payoffs are listed (US; H; T)

The game presented in Figure 1 represents a single round of an infinitely repeated game, which continues until one of three outcomes occurs. First, the game ends if both H and T engage in a direct military encounter that destabilizes H and causes him to fall from power. Substantively, the direct military encounter occurs if H engages in an offensive against T , or if T attempts to seize power at H s center. However, if T is defeated following either an offensive or an attempt to seize power at the center, the game terminates with H disarming T and emerging victorious. A third possibility is that both H and T reach a negotiated settlement in which H accommodates the demands of T . In doing so, H re-orients his foreign policy away from US : If any

of these three outcomes (T successfully destabilizes H at the center, H disarms T , or H and T reach a settlement) occurs, the game terminates. Assume that in each round, H survives a direct military encounter with T with probability plt 2 0; 1 and destabilizes in such an encounter with probability 1 plt ; where t represents the number of rounds previously played and l 2 0; L represents a normalizing parameter. This function specifies an assumption that H s probability of survival decreases over time, though the rate of decline slows as l ! L and accelerates as l ! 0: The assumption that H s ability to survive declines with time is based on several studies that conclude that terrorists strengthen as the organization

Downloaded from jpr.sagepub.com at TOURO COLLEGE LIBRARY on January 5, 2014

Bapat

307

persists over time (Bapat, 2005; Cronin, 2009; Jones & Libicki, 2008).

Moves Each round begins with US s decision on whether or not to provide military aid to H to improve his chances of surviving T s attempts at destabilization. US assistance is formally represented by x 2 0; 1, which is assumed to be a continuous and increasing level of assistance to H s regime. US provides increasing military aid as x ! 1; and relatively less as x ! 0: This aid directly increases lt lt H s probability of survival to 1px , where 1px > p so long as x > 0: This indicates that if x 1 plt , US effectively guarantees H s survival in armed conflict against T .7 On the other hand, if US sets x 0; H s lt probability of survival is simply equal to 1p0 plt : We therefore see that US s aid increases the probability that H will survive destabilization, but it is theoretically possible for H to survive without US s support. Once US allocates military aid to H , H has three possible options. First, H may use the military aid provided by US to mount an offensive against T in an effort to disarm it. In this case, H actively attempts to destroy T s organization. Second, H can choose to stay on the defensive against T and simply wait for T to make an attempt to seize power at the center. If H adopts the defensive posture, the game next moves to T , which decides whether or not to attack H at his center. If T attacks at H s center, H disarms T with probability plt plt 1x and T destabilizes H with probability 1 1x : On the other hand, if T does not challenge the center, the game moves to the next round, given that neither H nor T is disarmed. The final option for H is to negotiate with T in an effort to induce the group to cease its violence. Numerous studies in the bargaining and conflict literature indicate that negotiated settlements which reflect the balance of military power should be preferable to conflict (Fearon, 1995; Wagner, 2000). Therefore, T might prefer some negotiated settlement if H makes some offer that reflects the balance of military power between the two sides. With US s military assistance, H could engage in coercive bargaining by threatening to mount an offensive if T does not accept an offer of
Since 1px is a probability that must be bounded between [0,1], the offer x 1 plt represents an upper limit since if x 1 plt ,
7
lt

1 1px , which reflects T s probability of surviving a military encounter, given that US is supporting H :8 We can see that as x ! 1, indicating greater support from US , the value of H s offer to T declines. This indicates that H does not make an offer that reflects the actual balance of power between himself and T , but rather the balance of power created by US s military assistance. The only case in which H makes an offer that does reflect the actual balance of power is if x 0, in which case, H offers 1 1p0 1 plt : Once H makes the offer, T either accepts or rejects it. If T rejects the offer, H engages in an offensive against T in an attempt to disarm the group. On the other hand, if T accepts, T agrees to cease its political violence in exchange for H s concessions, which moves H politically away from US : US then decides whether or not to sustain military aid to H or whether to cut aid, given that H makes concessions to T that are unfavorable to US : Following US s decision of whether or not to continue providing H with military aid, T decides whether or not to abide by the terms of its agreement with H , or whether to defect on the deal and attack H at his center. Should T resume fighting, H survives with probability plt if US does not sustain aid, and survives with probability
plt 1x
lt

lt

if US sustains her military aid.

Payoffs H receives a payoff of 1 for each round in which he both survives as the sole sovereign over the territory and makes no concessions to T . If H negotiates with T , and T terminates its violence, H s payoff reflects the negotiated distribution of benefits. Should H make an offer of lt 1 1px to T , and should T accept, H s payoff would be
If H refuses to negotiate and engages in an offensive against T , H s payoff is equal to his probability of disarming T minus the cost cH 2 0; 1 he must pay to engage in military conflict, and an additional cost r 2 0; 1 that H must pay to arm his forces for an offensive.9 H s payoff for the offensive is therefore equal to 1px 1 1 1px 0 cH 1px cH r: If H adopts defensive tactics, and T responds by attacking
An alternative modeling strategy would be to endogenize the offer made by H to T. Doing so, however, makes the analytical solution more complex without affecting the empirical implications. 9 I assume that there is a cost ci 2 0; 1 for all i 2 H ; T should the players engage in conflict (Fearon, 1995). This cost represents the utility to i for the cost of conflict.
8
lt lt lt

plt 1x :

plt 1x 1.

Downloaded from jpr.sagepub.com at TOURO COLLEGE LIBRARY on January 5, 2014

308

journal of PEACE RESEARCH 48(3)

lt

at the center, H receives a payoff of 1px cH since this outcome also involves a military conflict against T : However, since H does not have to arm his forces for an offensive against T , H does not pay r if he adopts a defensive posture. If T chooses not to attack H s center, H receives a payoff of 1 and pays no cost, since there is no direct military conflict with T : Although both H and US derive utility from keeping T out of power, the utility functions of these players are slightly different. While H s primary objective is to maintain power over the territory, US does not derive utility from keeping H in power, but rather from keeping T out of power. Let us therefore specify that US receives a payoff of 1 if and only if T is kept out of power. On the other hand, let us assume that if T is able to achieve any political gains within H s territory, the US payoff reduces to 0, and US pays some political cost a 2 0; A: This indicates that if the results of the conflict between H and T end up being favorable to T , US will suffer some political punishment for failing to prevent T from achieving some power over the territory.10 Let us therefore assume that if T gains political power, either by destabilizing H or by negotiating with H , the payoff to the US is equal to a: Let us further represent the cost to US for supplying H with military aid as x : If conflict occurs between H and T, US s utility for supporting H can thereforebe represented by the following  expression:
plt 1x 1

misrepresent is often characteristic of terrorist groups, which frequently engage in behaviors intended to signal intense resolve, such as suicide bombings (Arce & Sandler, 2007; Bloom, 2005; Kydd & Walter, 2006; Lapan & Sandler, 1993). Let us therefore assume that US and H are uncertain as to the value of cT , but are both aware that cT * U[0,1], or distributed according to a standard uniform distribution ranging between 0 and 1.

Solution
Since the game is repeated infinitely and involves incomplete information, I solve the game using the Markov Perfect Bayesian Solution concept (MPBE). Stationary MPBE assume that in any subgame, the behavior of each of the players is dependent on only the current value of a particular state variable (Mailath & Samuelson, 2006; Maskin & Tirole, 2001). Let us characterize p as the state variable, or the current likelihood that H will survive a direct military encounter with T : This makes intuitive sense, as we would expect US , H , and T to base their decisions on H s immediate probabilities of surviving. I therefore characterize the solution by the stationary MPBE . The solution can be divided into three categories based on H s behavior.

1 1px a x : On the

lt

other hand, if US provides aid to H , but there is no conflict, the US payoff is equal to 1 x : Finally, if H negotiates with T , and T fulfills the terms of the agreement, the US payoff reduces to a, since H and T alter the status quo from US s preferred position. Let us now consider T s payoffs. T receives a payoff of 0 in each round if H survives, and receives a payoff of 1 if H destabilizes. If a negotiated settlement is reached, T s payoff corresponds to the offer that H makes. Like H and US , T is also assumed to pay a cost cT 2 0; 1 if it fights, which represents the utility of the cost of conflict to T : However, since terrorist groups are typically non-transparent and have incentives to misrepresent their true resolve, let us assume that both US and H have incomplete information as to the value of cT , whereas T has complete and perfect information (Fearon, 1995; Powell, 2002). This ability to

Case 1. Host negotiates x 1 1px Let us first identify the set of conditions under which H chooses to negotiate. In this case, T s final decision of whether or not to attack the center is contingent on whether or not US supports H after negotiation. If US sustains her military aid to H , T attacks at H s center iff lt lt 1 1px cT > 1 1px : We see that due to the costly nature of conflict, this expression can never be true, even if x 0: We can therefore see that if US sustains aid to H following negotiation, T will never renege on the agreement and attack the center. On the other hand, if US cuts her military aid to H , T attacks H s center lt if 1 plt cT > 1 1px : Rearranging terms, this indicates that T is indifferent between maintaining the agreement and attacking at the center plt plt x lt if cT 1x p ; which simplifies to 1x : We can therefore state that T will play Attack Center plt x plt x if cT < 1 x and will *Attack Center if cT  1x : Define
p x H s belief that T will attack at the center as cT 1 x : 0 We can see that x ! 1 as cT ! 1; which demonstrates that T is more likely to attack the center as the level of aid that US provides increases. Substantively, this indicates that the greater the level of military aid provided
0 lt

lt

10

Substantively, an example of this might be if the political price the Obama Administration would pay if the Talibans attempts to destabilize Hamid Karzais government were successful.

Downloaded from jpr.sagepub.com at TOURO COLLEGE LIBRARY on January 5, 2014

Bapat

309

by US , the more likely it is that T will defect from a negotiated settlement, assuming US cuts her aid in response to the negotiated settlement. However, notice that if x 0, the offer made by H is equal to 1 plt : If this is true, the probability that T defects is equal to
plt 0 10

0: In other words, if H makes an offer that reflects the balance of power between H and T without US s military support, T can be considered perfectly credible. This reflects something interesting: the more aid US provides, the less credible any deal is with T , whereas T is perfectly credible if US provides no military aid to H : Intuition would tell us that since US can deter T s attack at H s center with certainty if she continues military aid, US would simply maintain its military aid if H negotiates. However, since US s utility is solely based on keeping T out of power, US receives a payoff of a if H negotiates with T . Therefore, the payoff to US for continuing military aid to H following negotiation with T is equal to a x : On the other hand, if US cuts her aid to H , conflict erupts between H and
lt

1 plt cT  1 1px cT . Again, this condition is always fulfilled, which indicates that these T will accept and subsequently play Attack Center so long as x > 0: If x 0; these T accept and abide by the negotiated settlement. Therefore, H knows that if he plays Negotiate, T will always Accept and US will always *Sustain Aid. Since H is unaware of what type of T he is facing, H s payoff for negotiation can be defined as:   p lt x lt p lt x p cH 1 1: 3 1x 1x We therefore see that negotiation is a risky strategy for H : H is aware that making an offer to T guarantees that it will lose US s military aid, since US is only interested in keeping T out of power. Negotiation therefore requires H to gamble that T is of the type for which
p x if the probability that Nature draws cT  1 x : However,  plt x this type is equal to 1 1 x , we see that it is increaslt

lt

p x T with probability 1 x :US therefore sustains military aid to H after negotiation if:   plt x lt lt a x > p 1 1 p a 1x 1   p lt x a: 1 1x

ingly unlikely as x ! 1: In other words, if H is heavily reliant on US military aid, it is unlikely that negotiation will succeed, and highly likely that T will respond to US s revoking of aid by re-igniting terrorist violence.

This expression simplifies to: x 1 x > p2lt 1 a: 2

Case 2. Host plays defensive If H adopts defensive tactics, H does not mount an offensive against T ; but instead gambles that T will be deterred by US s provision of military aid and its subsequent increase in military power. T chooses to lt attack if 1 1px cT > 0: Re-arranging terms,
T attacks if cT < 1 1px : Let us therefore define the probability that T attacks given that H plays Defensive as cT 1 1px : H s utility for this strategy is therefore equal to:   lt    l t p lt p p 1 cH 1: 4 1x 1x 1x
00 lt lt

Since both x 2 0; 1, this condition cannot be fulfilled. This indicates that if H negotiates with T , US cannot credibly continue providing military aid to H : Both H and T know that US will revoke her aid if a negotiated settlement is reached. This establishes that if H negotiates, H is aware that US will cease its military aid, and T will play Attack Center with 0 probability cT : Given that both players are aware of this, how do H and T behave in negotiation? Interestingly enough, the model demonstrates that both T for which
p x p x cT < 1 x and T for which cT  1x always accept H s offer to negotiate. In the latter case, T will not attack H s center at the end of the game, even if US revokes her military aid. This T therefore accepts any deal if
lt lt

If we compare H s payoff for the defensive posture to H s payoff for negotiation, we quickly see that if x > 0, H s strategy of adopting the defensive posture dominates H s strategy of negotiation. Lemma 1. H strictly prefers to play Defensive over Negotiate if x > 0: Proof. See appendix.11
11

1 1px  1 1px cT , which must always be true. In the former case, where T will attack H s center once US revokes her aid, T accepts if

lt

lt

The appendix is located online at http://www.prio.no/jpr/datasets and http://www.unc.edu/~bapat/JPRAPPENDIX3.30.2011.pdf

Downloaded from jpr.sagepub.com at TOURO COLLEGE LIBRARY on January 5, 2014

310

journal of PEACE RESEARCH 48(3)

If, on the other hand, US sets x 0, H plays Defensive over Negotiate if    l t plt p 1 cH 10 10  lt  p plt 0 lt p cH 1 > 10 10   p lt 0 1; 1 10 which simplifies to 1 plt plt cH plt > plt : This expression is true if plt > cH , meaning that without military aid, only hosts that are likely to resist destabilization will refuse to negotiate, whereas weaker hosts will accommodate terrorists into their governments. Corollary 1. If x 0, H prefers Defensive iff plt > cH and prefers Negotiate otherwise. Proof. See appendix. This yields an interesting insight as to the effect of military aid on H s behavior. If plt is relatively high, H refuses to negotiate with T , despite the lack of US military aid. However, if plt < cH ; H negotiates with T. H s offer would reflect the balance of power between himself and T ; thereby rendering the agreement between the two parties credible. If, on the other hand, US provided military aid, even weaker H for which plt < cH prefer to play Defensive. We therefore see that while military aid does not necessarily eliminate terrorists, military aid discourages hosts from negotiating with terrorists and altering the status quo favored by the United States.

Proof. See appendix. We therefore see a second consequence of military aid, in that it not only discourages H from negotiating, but also discourages H from actually disarming T : This is true because H guarantees himself military losses with the offensive, whereas he instead guarantees that T will be deterred from challenging at the center if H adopts the defensive posture. Empirically, this would suggest that military aid discourages negotiation, which accomplishes US s objective, but discourages H from actually fighting terrorism, which is seemingly antithetical to the interests of US . To illustrate with an example, consider the case of Pakistan following 11 September 2001. During the tenure of Pervez Musharraf, the USA provided substantial military aid to the Pakistani regime. This aid continued despite reports that Musharraf was doing little to purge the Pakistani Inter-Service Intelligence (ISI) of pro-Taliban sympathizers, and despite reports that much of the aid was being directed by Pakistan to bolster their defenses against India. However, given that the Taliban appeared to be gaining in power so rapidly, revoking Pakistans aid could allow the Taliban an incentive to directly challenge Pakistans center. Faced with this possibility, the USA was concerned that the Pakistani government would strike a deal with the Taliban that would allow the group free reign in the north to attack US forces in Afghanistan. Given this scenario, the USA continued to supply military aid to Pakistan, despite reports that Pakistan has been less than cooperative in fighting the Taliban. Although one could argue that this strategy appears to be a US policy failure, defenders of military aid might suggest that simply keeping the Taliban out of power made the provision of military aid worthwhile, despite its costs.

Case 3. Host attempts offensive We see that US s provision of military aid appears to give H a disincentive to negotiate. Ideally, however, US should prefer that H engage in an offensive to eliminate T rather than adopting defensive postures. H s payoff for engaging in an offensive is equal to   plt plt 1 1 0 cH r 1x 1x plt cH r: 1x
Unfortunately for US , this payoff is also dominated by H s strategy of adopting a defensive posture. Lemma 2. H strictly prefers playing Defensive over Offensive.

US strategy We can divide H into two types: one for which plt < cH and one for which plt > cH : In the former case, H responds to a lack of US military aid by reaching some negotiated settlement with T in order to avoid continued violence. However, if US provides military aid while plt < cH , H has a disincentive to negotiate. Since H knows that he is likely to survive conflict with T as a result of US s support, H should prefer keeping T out of power and securing the benefits of sovereignty for himself. We therefore see two interesting features. First, military aid discourages hosts in the immediate term from negotiating with terrorists. Second, weaker H can use the threat of negotiating with T to compel US into sustaining military aid. If US were to set x 0;

Downloaded from jpr.sagepub.com at TOURO COLLEGE LIBRARY on January 5, 2014

Bapat

311
Level of US military aid (x) 1 .8 .6 .4 .2 1 2 3 4 5 6 Terrorist campaign years (t) p =.2, =.5, =1 p =.6, =.5, =1 7 8

H would negotiate, producing a payoff to US of a: If we compare this outcome to any case in which US provides military aid, we see that US strictly prefers aiding H if plt < cH : To illustrate, consider a case where x e: In this case, H will play Defensive since x > 0; which produces a payoff to US of 1 plt plt 1 plt a plt : Simplifying, we see that 1 plt plt 1 plt a plt > a if: plt 2 plt 1 a > 0:
lt

Since p 2 0; 1; this statement is always true. On the other hand, if US sets x 1 plt , which is its maximum value, US payoff simplifies to plt : Since it also must be true that plt > a; we see that for all possible values of x , US prefers setting x > 0 if H will negotiate. Substantively, this indicates that if H threatens to negotiate with T , US will always respond with a level of military aid to effectively bribe H from reaching a negotiated settlement. We can identify the level of aid US her aid  provides  by maximizing   utility  for  military 
1 1px
lt

p =.2, =.8, =1

Figure 2. Level of US military aid provided as a function of time t

plt 1x 1

1 1px a

lt

plt x 1x

1 x

with respect to x :

      lt   lt p plt plt q 1 1px 1x 1 1 1x a 1x 1

0 6

qx Using this expression, we can identify x , or the value of x that maximizes US utility:
1 1233 plt 1 a 18 pp 1 2 l t 18 9p 1 a 3 8 27plt 8a1 a3 q p 1 1 633 9p2lt 1 a 3 8 27plt 8a1 a3 x
1

to

Although the analytical solution is difficult to interpret, Figure 2 graphically present some of the properties of x : First, we know that x has a maximum value of 1 plt :12 If x 1 plt , T is perfectly deterred from challenging H at the center if H plays Defensive. The  probability that  T plays Attack Center is equal
p 1 11 0: plt lt
lt

This

indicates

that

if

which enables the group to improve its internal organization, ties to the population, and general ability to survive. If this is true, and H becomes less able to forcibly disarm T , US will be required to pay an increasing level of military aid to deter T from directly challenging the center. As an example, we can consider the increasing burden presently placed on the USA in Afghanistan. While Hamid Karzais government was seemingly at its peak in terms of legitimacy and international support following the toppling of the Taliban, since that time it has steadily declined in power relative to the group. This requires an increasing burden on the USA to protect Kabul from a Taliban offensive. This pattern is visible in the optimum level of military aid x identified by the model. If plt > cH , H is sufficiently likely to resist destabilization such that he will not negotiate with T , even if he does not receive military aid. In these cases, US can essentially free ride off the efforts of H , since H does not threaten to negotiate with T . However, in cases where a ! 1, it is possible that US will provide military aid even if H plays Defensive. While the full discussion is reserved for the appendix, consider where the US provides the maximum level of aid x 1 plt . In these cases, we see that US prefers to provide military aid if plt > 1 plt plt 1 1 plt a plt : Simplifying, we see that this expression is true if plt : Substantively, this result suggests that even a> 1 plt in cases where H will refuse to negotiate with T , US will supply H with military aid if the political punishment associated with H s destabilization is significantly high. Although this is valuable to H in that it keeps him in

x > 1 p , US sets x 1 p : Substantively, if the level of military aid that maximizes US utility exceeds the maximum allowable level of US military aid, US simply sets the level of military aid to the maximum allowable level. From Figure 2, we see that over time, the level of military aid that maximizes US utility monotonically increases. By playing Defensive, H allows T to survive,
12

lt

Assume that if x*>1plt, x 1plt.

Downloaded from jpr.sagepub.com at TOURO COLLEGE LIBRARY on January 5, 2014

312

journal of PEACE RESEARCH 48(3)

power, there is an interesting consequence to US military aid. Without military aid, T for which cT < 1 plt would attack the center, which would likely end with the disarmament of these groups. Since plt is relatively higher, H would have a better chance of disarming these groups and ending the game than in the previous case where plt < cH : However, because US supplies H with military aid, fewer T will directly challenge H : Instead, T will continue to engage in terrorist violence away from H s center. And, because H profits from keeping T functional, H will do little to stop this activity. We therefore see a perverse effect: US supplies H with military aid if US suffers a substantial political punishment for allowing H to fall, but this act deters T from challenging the center and allows T to remain functional. However, since US risks a large political cost if H were to destabilize, it is preferable to US to supply the military aid and avoid the risk of a confrontation between T and H : This suggests that if even though supplying aid is likely to delay the demise of T , it is preferable for the US to supply the aid to insulate itself against the risk of H s destabilization. We can now characterize the games equilibrium solution: Proposition 1. The following set of strategies constitute a stationary MPE:
1.

Analysis
The model suggests that although military aid appears intended to defeat terrorists, military aid may actually give hosts an incentive not to terminate their conflicts with terrorists. Terminating the conflict would result in a loss of aid, whereas simply using the aid as a deterrent decreases the risk that the terrorist group will directly challenge the center. This insight might explain why military aid appears ineffective against terrorism: it provides hosts with a disincentive to remove their terrorist problem. Relatively weaker hosts can leverage the threat to negotiate with the group to coerce the USA into continuing military aid. On the other hand, states that are relatively stronger, but politically valuable to the USA, may also compel the USA into subsidizing their own defense against terrorists. In both cases, having a terrorism problem benefits the host.
Hypothesis 1: The provision of US military aid decreases the hazard of terrorist collapse.

T : If H plays Defensive, *Attack Center if cT  1 1px and Attack Center otherwise. If H plays Negotiate and US plays *Sustain Aid, play
p x Accept; Attack Center if cT < 1 x and Accept, *Attack Center otherwise.
lt lt

Given that military aid appears to invite blackmail, why would the USA continue to provide it? According to the model, the USA continues to provide military aid if the cost of allowing hosts to negotiate with terrorists are relatively higher. The model predicts that the greater these political punishments are, the longer we should expect governments to provide aid to both prevent hosts from negotiation and inoculate hosts from any risk of destabilization. Giving these states military aid effectively deters terrorists from making any attempts at the center, and preserves these states as loyal US foreign policy allies. Unfortunately, this likely fuels the incentive of hosts to allow anti-US groups to exist within their territory.
Hypothesis 2: Increasing the affinity between a host and the USA decreases the hazard of terrorist collapse.

2.

H: If x > 0, play Defensive. If x 0, play Negotiate if

plt < cH and Defensive otherwise. 3. US : If plt < cH , set x x , where

x 1 18 18  1233 plt 1 a pp 1 9p2lt 1 a 3 8 27plt 8a 1 a3   pq 1 1 633 9p2lt 1 a 3 8 27plt 8a 1 a3
1

Research design
I test the two hypotheses using data from the US Overseas Loans and Grants, Obligations and Loan Authorizations Greenbook (2008).13 Since the hypotheses are specific to anti-terrorism aid, I restrict the analysis to only aid that falls under the category entitled Nonproliferation, Anti-terrorism, Demining, and other related programs (NDR). Part of the NDRs purpose is support programs intended to use the Terrorism Interdiction
Data available at: http://gbk.eads.usaidallnet.gov/. Data for replication are available at www.prio.no/jpr/datasets.
13

if x < 1 plt , and set x 1 plt otherwise. If plt > cH , set x x iff x < 1 plt and a >
p2lt 2x 2plt 1x 1x 2 plt 2plt 2x 2x

, and set x 0 otherwise.

lt

p If x > 1 plt , set x 1 plt if a > 1 and plt x 0 otherwise.

Proof. See appendix.

Downloaded from jpr.sagepub.com at TOURO COLLEGE LIBRARY on January 5, 2014

Bapat

313

Program (TIP) to meet the evolving terrorist threat and improve TIP countries efforts to interdict terrorists.14 From the Greenbook, we are able to get a list of US military aid disbursements to fight terrorists for each year from 1997 to 2006 in constant 2008 US dollars. The theoretical models focus is on the effect of US aid disbursements on the duration of terrorist campaigns. I use the Jones & Libicki (2008) dataset on terrorist campaigns to collect information on these groups. This dataset contains information on 648 groups from 1968 to 2006. Since we are only interested in how the NDR affects the duration of terrorist campaigns, we restrict the analysis to campaigns that were ongoing from 1997 to 2006. Additionally, since the model assumes that the USA is supporting a host that can substantially benefit from aid, I exclude the terrorist campaigns that were operating against major powers. The new dataset therefore consists of 184 terrorist campaigns directed against 48 hosts from 1997 to 2006. The data from Jones & Libicki are time-invariant, meaning that the dataset examines each campaign as a single observation, rather than looking at campaign years. The analysis therefore examines the campaigns as individual units of analysis.

Dependent variable The dependent variable is the duration until terrorists are defeated. Jones & Libicki identify that terrorist campaigns terminate if the group splinters or is defeated by police work.15 I consider both of these cases to be instances in which the host disarms its group. I create a dichotomous variable labeled Terrorist collapse, which is coded as 1 if the campaign terminates from policing or with the group splintering, and 0 otherwise. In the dataset, 58 of the 184 cases (31.52%) terminated with the groups collapse. Of these 58 cases, the mean time until the host prevailed was 1.69 years with a standard deviation of 1.47. Independent variables The key independent variable is the military aid provided by the USA to the host. I measure this in three ways. First, I use a simple dichotomous variable entitled US Aid. This variable is coded as 1 if the USA provided any
US State Department document (http://www.state.gov/ documents/organization/28971.pdf). 15 Splintering in the Jones & Libicki data refers to the collapse of the group. It does not refer to cases where a terrorist organization splits into multiple groups.
14

military aid to the particular host, and 0 otherwise. Second, as is common in the literature, I take the natural log of the maximum level of military aid provided by the USA to the particular host during the course of the terrorist campaign. My expectation is that both variables will decrease the hazard of Terrorist collapse and will have negative coefficients. A key concern of this analysis is the potential for endogeneity. Although the hypotheses predict that military aid lengthens terrorist campaigns, it is conceivable that the USA responds to longer terrorist campaigns with military aid. I therefore create an instrumental variable using the hosts Freedom House score and the hosts logged population score.16 Freedom House divides states into three groups: (0) not free, (1) partially free, (3) free. Using the Freedom House score as an instrument effectively handles the problem of endogeneity, in that the Freedom House measure should be correlated with military aid, but uncorrelated with the length of the terrorist campaign.17 The second variable used to create the instrument is the hosts logged population score. It is reasonable to believe that hosts with relatively larger populations may require more military support to control their territory, but it is unclear if a large population contributes to longer terrorist campaigns. I use a probit model to estimate the US Aid variable with the Freedom House and logged population variables. I then save the predicted probabilities as the variable Instrument US Aid.18 The estimates of US Aid using the Freedom House scores are presented in Table I. The model clusters the standard errors on each hosts respective country code, given that within-host observations are probably not independent. The analysis indicates that both variables are relevant instruments that are correlated with US Aid. As with the dichotomous US Aid variable and the logged military aid variable, my expectation is that the instrument will have a negative coefficient, indicating that US Aid decreases the hazard of Terrorist collapse.
Freedom in the World 19972006 (http://www.freedomhouse.org/ template.cfm?page15). 17 Numerous studies indicate that the presence of terrorists is correlated with regime type (Li, 2005; Sandler, 1995). However, few suggest that the duration of a campaign is a function of regime type. Therefore, while regime type has been shown to be correlated with the presence of terrorists, it has not been theoretically demonstrated to be correlated with duration. 18 I also created an instrument for the logged maximum level of US aid. However, while this instrument yielded results that were consistent with the theoretical hypotheses, this instrument was not effective in removing the possibility of endogeneity.
16

Downloaded from jpr.sagepub.com at TOURO COLLEGE LIBRARY on January 5, 2014

314 Table I. Estimates of US military aid using host Freedom House scores and host logged population Variable Host Freedom House score Ln population Constant N Log likelihood Pr.>chi2
* p < .1, **p < .05, ***p < .01.

journal of PEACE RESEARCH 48(3)

US military aid (probit) b (s.e.) .81 (.27)** .25 (.14)* 3.22 (2.4) 184 105.15 .01

campaign as a second control variable. This variable is also included as an ancillary parameter in the tests that do not use the instrument. Third, I include the hosts logged number of military personnel in the first year of the campaign to control for the militarization of the host. Finally, in the models that do not use the instrumental variable, I include the hosts Freedom House score. Each of these variables is intended to hold the baseline conditions that favor terrorism constant in order to focus on the effect of military aid on the duration of the campaign.

The second hypothesis focuses on the political cost to the USA associated with a particular hosts destabilization. To test this hypothesis, I develop a measure of affinity between the USA and a particular host using information from the US State Departments Voting Practices in the United Nations.19 This publication lists the percentage of times that each state in the system voted in the same manner as the USA on votes in the UN General Assembly. The correlation between UN votes is often used as a proxy for political affinity (Gartzke, 2000). I therefore assume that if a state maintains a close affinity with the USA, as represented by a percentage of votes supported by the USA in the UN General Assembly, the cost to the USA for this states destabilization increases. Therefore, we would predict that since higher scores indicate greater affinity with the USA, the USA is less likely to terminate military aid, thereby increasing the duration of time until a group collapses. I take the natural log of the UN Voting variable to normalize its distribution. My expectation is that like the aid variables, the affinity variable will have a negative coefficient, meaning that increases in affinity with the USA decrease the hazard of Terrorist collapse. The mean logged UN similarity score is equal to 1.34, with a standard deviation of .75, a min of 2.88 and a max of .04. Although the sample is smaller, I include three additional control variables in the model. First, I include the hosts logged per capita GDP score for the first year of the terrorist campaign. Several studies associate higher per capita GDP scores with a more powerful and developed central government, which decreases the ability of militants to operate (Fearon & Laitin, 2003). This variable therefore controls for the hosts ability to suppress its terrorists without help from the USA. Second, I include the logged population score for the host in the first year of the terrorist

Method I test the hypotheses using three Weibull models (Box-Steffensmeier & Zorn, 2002). To mitigate endogeneity, the third model runs the analysis using the instrumental variable (Bijwaard, 2008; Olsen & Farkas, 1988). The statistical models therefore include the military aid variables as ancillary parameters that affect the shape of the hazard rate. My expectation is that the coefficients of the aid variables will be negative, suggesting that each of them decreases the hazard that the terrorist group will collapse. Hypothesis 2, on the other hand, predicts a proportional decrease in the probability that a group is defeated, given the cost to the USA for abandoning a particular host. Therefore, the US Affinity variable is treated only as a variable that proportionately affects the hazard rate, rather than the shape of the hazard rate.20 The expectation here is also that the US Affinity variable will have a negative coefficient that proportionately decreases the probability that a group collapses.

Results
The first column in Table II presents the results using the dichotomous military aid variable, the second column presents the results using the logged military aid expenditure variable, and the third presents the instrumental variable results. In each test, the military aid variables each have negative and statistically significant coefficients, which indicates that increases in each of the military aid indicators significantly increase the length of time until a terrorist group collapses. We further see that the instrumental variable is both negative and significant at the .01 level. This is important given the likelihood that the duration of a campaign may influence the provision of military aid. The correlation between the
20

19

Available at http://www.state.gov/p/io/rls/rpt/.

As a robustness check, each analysis was re-run using the Cox specification instead of the Weibull model with ancillary parameters. The results remain consistent. A table with the Cox results is presented in the appendix.

Downloaded from jpr.sagepub.com at TOURO COLLEGE LIBRARY on January 5, 2014

Bapat

315

Table II. The effect of US military aid on the duration until host government disarms terrorist groups Variable Weibull Ln US Affinity Ln per capita GDP Ln Population Ln Armed force Constant Ancillary Military aid Ln per capita GDP Host Freedom House Ln Population Constant N Log likelihood Pr.>chi2
* p < .1, **p < .05, ***p < .01.

Dichotomous b (s.e.) .73 (.31)** .28 (.22) .41 (.2)** .71 (.27)*** 7.03 (4.02) .78 (.32)** .25 (.07)*** .35 (.14)** .08 (.12) 3.5 (2.2) 174 156.16 .02

Logged b (s.e.) 1.4 (.47)*** .48 (.43) 1.23 (.36)*** 1.71 (.4)*** 1.27 (7.2) .29 (.12)** .26 (.17) .17 (.17) .08 (.12) 7.8 (4) 103 60.12 .00

Instrument b (s.e.) .66 (.32)** .24(.21) .48 (.17)*** .73 (.24)*** 5.6 (3.2) 1.3 (.48)*** .15 (.06)*** 1.96 (.65) 174 164.79 .01

Table III. Predicted effects of US military aid and increasing US affinity on terrorist campaign duration Years until terrorist group is disarmed Variable US Aid (instrument) US Affinity Baseline 4.72 7.52 1 sd 7.52 13.73 % change 59% 83%

error term and the instrument is .04 with a significance level of .56. This demonstrates that the instrument is uncorrelated with the error term. Figure 3 plots the predicted hazard rates both with no US military aid and with military aid using the dichotomous military aid variable and the instrumental variable. We see that without military aid, the probability that terrorists collapse increases throughout the course of the campaign and reaches its highest point by Year 8. However, with US military aid, the probability that terrorists collapse both proportionately drops and decreases throughout the course of the campaign. These results are consistent with the prediction of Hypothesis 1: if the USA provides military aid, then there will be a decrease in the probability that the terrorist campaign will terminate. This supports the models contention that providing military aid to hosts makes the business of fighting terrorists quite profitable. Without military aid, the predicted hazard of terrorist collapse steadily increases over time. Although this result is inconsistent with the theoretical models assumption that the hazard of terrorist collapse declines monotonically with

time, recall that the theoretical model predicts that if the USA were to provide no military aid, the campaign would typically end in the first year with a negotiated settlement if plt < cH : In the alternative cases for which plt > cH , the model predicted resolute terrorists may attack the host at its center if and only if the USA did not provide military aid. In these cases, since plt is relatively higher, the model predicts that the most resolute terrorist groups were indeed likely to collapse quickly following their attacks on the hosts center. The results support these predictions, and also suggest that US policymakers are quite effective in determining where to allocate military aid. The results suggest that US policymakers do not grant military aid in cases where hosts are capable of suppressing their terrorists, and do provide aid when hosts have more difficulty in suppressing terrorists. This result is consistent with the theoretical model. Additionally, if we re-test the results using non-parametric Cox models, the results are consistent with both Hypothesis 1 and the theoretical assumption of a declining hazard of terrorist collapse. These results from the Cox models are presented in the appendix.21 We can therefore conclude that the results are consistent with both Hypothesis 1 and the theoretical models empirical implications.

21

Please see the table marked Supplement 1, located with the appendix at http://www.prio.no/jpr/datasets and http:// www.unc.edu/~bapat/JPRAPPENDIX3.30.2011.pdf

Downloaded from jpr.sagepub.com at TOURO COLLEGE LIBRARY on January 5, 2014

316
US aid (dichotomous) .25 .25

journal of PEACE RESEARCH 48(3)

US aid (instrument)

.2 Terrorist collapse hazard

.2

.15

.15

.1

.1

.05

.05

0 1 2 3 4 5 6 7 Duration of terrorist campaign No military aid 8

0 1 2 3 4 5 6 7 Duration of terrorist campaign No military aid Military aid 8

Military aid

Figure 3. Effect of military aid on the number of years until group collapse

If we use the instrumental variable model to compare the mean duration until a group collapses, we see that when the instrument is equal to .39, signifying no military aid, the predicted duration until a group collapses is 4.69 years. However, when the instrument is increased to its mean of .61, indicating that the USA is providing military assistance, the predicted duration increases to 7.82 years, which is a 67% increase. We therefore see that consistent with Hypothesis 1, US military aid seems to prolong the existence of terrorists, which can be explained by the lucrative nature of having an active terrorist campaign in the post 9/11 era. The results further support Hypothesis 2. Again, we see that each of the coefficients on the US Affinity variable are both negative and significant across each of the models, including the one using the instrumental variable. Figure 4 captures the effect of increasing US Affinity on the hazard of group collapse. We see that increasing affinity proportionately decreases the probability that a terrorist group is defeated. In the model with the instrumental variable, the mean duration until a terrorist group collapses if the USA provides military aid is 7.82 years. A one standard deviation increase in the hosts US Affinity score increases this duration to 13.38 years, which is a 71% increase. This demonstrates that terrorist campaigns in hosts that are relatively more supportive of the USA tend to take longer to terminate. This supports the theoretical models prediction that

increasing the costs to the USA for terrorists entering into the hosts government makes the business of counter-terrorism lucrative and gives hosts no incentive to terminate their terrorist threats. The increase in the hosts military power may deter terrorists from challenging and achieving power at the center, but paradoxically, allows terrorists to survive for longer durations.

Conclusion
The model and empirical test demonstrate that military aid prolongs terrorist campaigns largely because the promise of such aid can make the business of fighting terrorism profitable. However, the model demonstrates that the USA is not completely foolish in continuing to provide military aid despite this set of incentives. The model demonstrates that without military aid, hosts would likely negotiate with terrorists in order to reach an efficient, peaceful solution to their disputes. If the issue in dispute is related to US strategic interests, such a negotiated settlement would likely involve change that is unfavorable to the USA. However, if the USA provides military aid to the host, the gains from military aid give the host no incentive to accommodate the terrorists, and may keep the host loyal to the USA. Therefore, if we evaluate the effectiveness of military aid, we can see that while it is ineffective in that it prolongs campaigns by giving hosts disincentives to disarm their terrorists, it is

Downloaded from jpr.sagepub.com at TOURO COLLEGE LIBRARY on January 5, 2014

Bapat
US aid (dichotomous)
.1 .1

317
US aid (instrument)

Terrorist collapse hazard

.05

0 1 2 3 4 5 6 7 Duration of terrorist campaign Baseline Increased affinity 8

0 1 2 3 4 5 6 Duration of terrorist campaign Baseline 7 8

Increased affinity

Figure 4. Effect of increasing punishment to US for failing to support a host on the number of years until terrorist collapses

effective at preventing terrorists from altering a hosts foreign policy. In that sense, military aid can be effective at accomplishing US objectives.

Replication data
Data for replication are available at www.prio.no/jpr/datasets.

Acknowledgements
I would like to thank Todd Sandler, Andrea Lopez, Ed Kaplan, Shyam Sunder, the participants in the 2010 Terrorism and Policy Conference at the University of Texas at Dallas, the members of the 2010 Daniel Rose Yale UniversityTechnion Initiative, and three anonymous reviewers for their very helpful comments. All remaining errors remain my responsibility. This work was supported by the National Science Foundation (SES-1000328).

References
Alesina, Alberto & David Dollar (2000) Who gives foreign aid to whom and why? Journal of Economic Growth 5(1): 3363. Arce, Daniel G & Todd Sandler (2007) Terrorist signalling and the value of intelligence. British Journal of Political Science 37(4): 573586.

Azam, Jean Paul & Alexandra Delacroix (2006) Aid and the delegated fight against terrorism. Review of Development Economics 10(2): 330344. Azam, Jean Paul & Veronique Thelen (2010) Foreign aid versus military intervention in the war on terror. Journal of Conflict Resolution 54(2): 237261. Bandyopadhyay, Subhayu; Todd Sandler & Javed Younas (2009). Foreign aid as counterterrorism policy. Federal Reserve Bank of St Louis working paper 2009-021A. Bapat, Navin A (2005) Insurgency and the opening of peace processes. Journal of Peace Research 42(6): 699717. Berrebi, Claude (2003) Evidence about the Link Between Education, Poverty, and Terrorism Among Palestinians. Princeton, NJ: Princeton University Press. Bijwaard, Govert E (2008) Instrumental variable estimation for duration data. Tinbergen Institute discussion paper TI 2008-032/4. Bloom, Mia. (2005) Dying to Kill: The Allure of Suicide Terror. New York: Columbia University Press. Box-Steffensmeier, Janet M & Christopher Zorn (2002) Duration models for repeated events. Journal of Politics 64(4): 10691094. Bueno de Mesquita, Bruce & Alastair Smith (2007) Foreign aid and policy concessions. Journal of Conflict Resolution 51(2): 251284.

Downloaded from jpr.sagepub.com at TOURO COLLEGE LIBRARY on January 5, 2014

318

journal of PEACE RESEARCH 48(3)

Carothers, Thomas (2003) Promoting democracy and fighting terror. Foreign Affairs 1: 8497. Cingranelli, David L & Thomas E Pasquarello (1985) Human rights practices and the distribution of US foreign aid to Latin American countries. American Journal of Political Science 29(3): 539563. Cronin, Audrey K (2009) Ending Terrorism: Lessons for Policymakers from the Decline and Demise of Terrorist Groups. London: Routledge. Fearon, James (1995) Rationalist explanations for war. International Organization 49(3): 379 414. Fearon, James D & David Laitin (2003) Ethnicity, insurgency, and civil war. American Political Science Review 97(1): 7590. Gartzke, Erik (2000) Preferences and the democratic peace. International Studies Quarterly 44(2): 191212. Hegre, Hvard & Nicholas Sambanis (2004) Sensitivity analysis of empirical results on civil war onset. Journal of Conflict Resolution 50(4): 508535. Jones, Seth G & Martin C Libicki (2008) How Terrorist Groups End: Lessons for Countering al Qaeda. Santa Monica, CA: RAND. Kruegar, Alan B & Jitka Maleckova (2003) Education, poverty, and terrorism: Is there a causal connection? Journal of Economic Perspectives 17(4): 119144. Kydd, Andrew H & Barbara F Walter (2006) The strategies of terrorism. International Security 31(1): 4980. Lake, David A (1999) Entangling Relations: American Foreign Policy in Its Century. Princeton, NJ: Princeton University Press. Lapan, Harvey & Todd Sandler (1993) Terrorism and signalling. European Journal of Political Economy 9(3): 383397. Li, Quan (2005) Does democracy promote or reduce transnational terrorist incidents? Journal of Conflict Resolution 49(2): 278297. Mailath, George J & Larry Samuelson (2006) Repeated Games and Reputations. Oxford: Oxford University Press. Maskin, Eric & Jean Tirole (2001) Markov perfect equilibrium: I. Observable actions. Journal of Economic Theory 100(1): 191219.

McCormick, James M & Neil Mitchell (1988) Is U.S. aid really linked to human rights in Latin America? American Journal of Political Science 32(1): 231239. Meernik, James; Eric L Krueger & Stephen C Poe (1998) Testing models of U.S. foreign policy: Foreign aid during and after the Cold War. Journal of Politics 60(1): 6385. Olsen, Randall & George Farkas (1988) Endogenous covariates in duration models and the effect of adolescent childbirth on schooling. Journal of Human Resources 24(1): 3953. Palmer, Glenn & T Clifton Morgan (2006) A Theory of Foreign Policy. Princeton, NJ: Princeton University Press. Poe, Steven (1990) Human rights and US foreign aid: A review of quantitative studies and suggestions for future research. Human Rights Quarterly 12: 499512. Powell, Robert (2002) Bargaining theory and international conflict. Annual Review of Political Science 5: 130. Sandler, Todd (1995) On the relationship between democracy and terrorism. Terrorism and Political Violence 7(4): 19. Schudel, Carl Jan Willem (2008) Corruption and bilateral aid: A dyadic approach. Journal of Conflict Resolution 52(4): 507526. Stotsky, Steven (2008) Does foreign aid fuel Palestinian violence? Middle East Quarterly 3: 2330. Svensson, Jakob (2000) Foreign aid and rent-seeking. Journal of International Economics 51(2): 437461. Wagner, R Harrison (2000) Bargaining and war. American Journal of Political Science 44(3): 469484. Walt, Stephen (2001) Beyond bin Laden: Shaping U.S. foreign policy. International Security 26(3): 5689. NAVIN BAPAT, b. 1977, PhD in political science (Rice University, 2004); Assistant Professor at the University of North Carolina Chapel Hill specializing in the study of transnational insurgency, terrorism, international conflict, and economic sanctions. His current research focuses on alliances among militant groups, the spread of internal violence to international wars, and the growth of violent insurgent movements.

Downloaded from jpr.sagepub.com at TOURO COLLEGE LIBRARY on January 5, 2014