¹ Coll, Steve, Ghost Wars (New York: Penguin, 2004)Google Scholar

² McKelvey, Richard, ‘General Conclusions for Global Instransitivities in Formal Voting Models’, Econometrica, 47 (1979), 1085–1112CrossRefGoogle Scholar

Shepsle, Kenneth, ‘Institutional Arrangements and Equilibrium in Multidimensional Voting Models’, American Journal of Political Science, 23 (1979), 27–59CrossRefGoogle Scholar

³ Cunningham, David, ‘Veto Players and Civil War Duration’, American Journal of Political Science, 50 (2006), 875–92CrossRefGoogle Scholar

Davis, David and Moore, Will H., ‘Ethnicity Matters: Transnational Ethnic Alliances and Foreign Policy Behavior’, International Studies Quarterly, 41 (1997), 171–181CrossRefGoogle Scholar

Kydd, Andrew and Walter, Barbara, ‘Sabotaging the Peace: The Politics of Extremist Violence’, International Organization, 56 (2002), 49–80CrossRefGoogle Scholar

⁴ Byman, Daniel, Deadly Connections: States that Sponsor Terrorism (Cambridge: Cambridge University Press, 2005)CrossRefGoogle Scholar

Chailand, Gerard, Terrorism: From Popular Struggle to Media Spectacle (London: Saqi, 1987)Google Scholar

⁵ While the conventional definition of alliances requires a formal contract, we loosen the definition by considering any security co-operation, be it formal or informal, as a type of alliance. This is necessary largely due to the relative lack of codified agreements among non-state actors.

⁶ Fearon, James, ‘Rationalist Explanations for War’, International Organization, 49 (1995), 379–414CrossRefGoogle Scholar

Leeds, Brett Ashley, ‘Domestic Political Institutions, Credible Commitments, and International Cooperation’, American Journal of Political Science, 43 (1999), 979–1002CrossRefGoogle Scholar

Morrow, James D., ‘Alliances: Why Write them Down?’ Annual Review of Political Science, 3 (2000), 63–83CrossRefGoogle Scholar

⁷ Axelrod, Robert, The Evolution of Cooperation (New York: Basic Books, 1984)Google Scholar

Fearon, James, ‘Bargaining, Enforcement, and International Cooperation’, International Organization, 52 (1998), 269–306CrossRefGoogle Scholar

Kydd, Andrew, Trust and Mistrust in International Relations (Princeton, N.J.: Princeton University Press, 2005)CrossRefGoogle Scholar

Sartori, Anne, ‘The Might of the Pen: A Reputational Theory of Communication in International Disputes’, International Organization, 56 (2002), 121–149CrossRefGoogle Scholar

⁸ Cronin, Audrey K., Ending Terrorism: Lessons for Policymakers from the Decline and Demise of Terrorist Groups (London: Routledge, 2009)Google Scholar

Jones, Seth G. and Libicki, Martin C., How Terrorist Groups End: Lessons for Countering al Qaeda (Santa Monica: RAND, 2008)Google Scholar

Kydd and Walter, ‘Sabotaging the Peace: The Politics of Extremist Violence’, pp. 263–296CrossRefGoogle Scholar

Stedman, Stephen, ‘Spoiler Problems in Peace Processes’, International Security, 22 (1997), 5–53CrossRefGoogle Scholar

Walter, Barbara, Committing to Peace: The Successful Settlement of Civil Wars (Princeton, N.J.: Princeton University Press, 2002)CrossRefGoogle Scholar

⁹ Throughout this analysis, we draw parallels between alliance formation amongst militants and alliance formation amongst states. Although one could argue that these are different entities, we believe in applying game theory, there is no qualitative difference between states and non-state actors. Whether there is a difference is an empirical question, which we seek to address in our statistical test.

¹⁰ Byman, Daniel, Chalk, Peter, Hoffman, BruceRosenau, William and Brannan, David, Trends in Outside Support for Insurgent Movements (Santa Monica, Calif.: RAND, 1999)Google Scholar

¹¹ We acknowledge that this model greatly simplifies these conflicts. It is possible that in some cases, there are either more than two groups or more than one state involved. It is also possible that, in a particular case, one of the factions may be a splinter group, or that the groups may co-operate under an umbrella organization. However, to allow the models to develop clear empirical implications, we simplify the model to three players, and assume that each group is distinct from one another. This substantively excludes splinters from the analysis, though the analysis does speak to cases where various groups consider whether to ally under an umbrella organization.

¹² Lichbach, Mark I., The Rebel's Dilemma (Ann Arbor: University of Michigan Press, 1995)CrossRefGoogle Scholar

Zartman, I. William, Elusive Peace: Negotiating an End to Civil Wars 1995–1996 (Washington, D.C.: The Brookings Institution, 1995)Google Scholar

¹³ Conybeare, John A., ‘Arms Versus Alliances: The Capital Structure of Military Enterprise’, Journal of Conflict Resolution, 38 (1994), 215–235CrossRefGoogle Scholar

Lake, David A., Entangling Relations: American Foreign Policy in its Century (Princeton, N.J.: Princeton University Press, 1999)CrossRefGoogle Scholar

Morgan, T. Clifton and Palmer, Glenn, ‘To Protect and Serve: Alliances and Foreign Policy Portfolios’, Journal of Conflict Resolution, 47 (2003), 180–203CrossRefGoogle Scholar

¹⁴ A report by Anand Gopal suggests that while the Quetta Shura favours an Islamic emirate, the Haqqani network's preference appears to be for an Islamic Republic. Reports also indicate that though he was once a member of the Taliban government, Haqqani complained about the heavy-handedness of the Taliban. See Anand Gopal, ‘The most deadly U.S. foe in Afghanistan’, Christian Science Monitor, 31 May 2009, p. 1; http://www.csmonitor.com/World/Asia-South-Central/2009/0601/p10s01-wosc.html

¹⁵ For more information on the Quetta Shura and the Haqqani Network, see David Clark Scott, ‘What is the Quetta Shura Taliban and Why Does it Matter?’ Christian Science Monitor, Global News Blog, available at: http://www.csmonitor.com/World/Global-News/2010/0224/What-s-the-Quetta-Shura-Taliban-and-why-does-it-matter; and Anand Gopal, Mansur Khan Mahsud and Brian Fishman. ‘Inside the Haqqani Network’, Foreign Policy. The AfPak Channel: Inside the War for South Asia. http://afpak.foreignpolicy.com/posts/2010/06/03/inside_the_haqqani_network_0.

¹⁶ Lake, Entangling Relations, pp. 35–77Google Scholar

¹⁷ Arguably, this is the strategy pursued by the United States in the Afghan conflict against the Quetta Shura and the Haqqanis. See Dexter Filkins, ‘Taliban elite, aided by NATO, joint talks for Afghan peace’, 19 October 2010. p. A1. http://www.nytimes.com/2010/10/20/world/asia/20afghan.html?_r =1&pagewanted = 1

¹⁸ We acknowledge that this model greatly simplifies these conflicts. It is possible that in some cases there are either more than two groups or more than one state involved. It is also possible that in a particular case one of the factions may be a splinter group, or that the groups may co-operate under an umbrella organization. However, to allow the models to develop clear empirical implications, we simplify the model to three players, and assume that each group is distinct from one another.

¹⁹ The game assumes complete but imperfect information. Each player's utilities are represented by Von Neumann–Morgenstern utility functions.

²⁰ To allow for generalizability, we assume that the concession may substantively represent the granting of territory to the group, the altering of an offensive government policy, investment in the militant group's home area, cash or any other policy concession the group is seeking.

²¹ Ethan Bueno de, Mesquita, ‘Conciliation, Counterterrorism, and Patterns of Terrorist Violence’, International Organization, 59 (2005), 145–176Google Scholar

²² For a discussion of how the probability that a game continues can serve as a discount factor, please see Evelyn C. Fink, Scott Gates and Brian Humes, Game Theory Topics: Incomplete Information, Repeated Games, and N-Player Games (Thousand Oaks, Calif.: Sage University Papers on Quantitative Applications in the Social Sciences, No. 122, 1998): and/or James D. Morrow, Game Theory for Political Scientists (Princeton, N.J.: Princeton University Press, 1994). These texts indicate that since probabilities and the traditional discount factor are both bounded between 0 and 1 they are equivalent in infinitely repeated games. It is also important to note that the parameter δ in this game is different from the traditional discount factor, and is used here to represent the discount on the payoff for each player if they co-operate.

²³ Bapat, Navin, ‘Insurgency and the Opening of Peace Processes’, Journal of Peace Research, 42 (2005), 699–717CrossRefGoogle Scholar

Cronin, Ending Terrorism, pp. 28–47Google Scholar

Jones and Libicki, How Terrorist Groups End, pp. 9–44Google Scholar

²⁴ Ethan Bueno de, Mesquita and Dickson, Eric, ‘The Propaganda of the Deed: Terrorism, Counterterrorism, and Mobilization’, American Journal of Political Science, 51 (2007), 145–176Google Scholar

²⁵ Since both p and q are probabilities, they must be bounded between (0,1). Therefore, assume that if βp > 1, βp = 1 and if βq > 1, βq = 1. This substantively does not change the result. It indicates that in this situation, if a group is betrayed, the probability that they will be disarmed is equal to 1.

²⁶ This conclusion is very similar to Bueno de Mesquita, ‘Conciliation’.

²⁷ Axelrod, Robert and Keohane, Robert, ‘Achieving Cooperation Under Anarchy: Strategies and Institutions’, World Politics, 38 (1985), pp. 226–254Google Scholar

Lake, Entangling Relations, pp. 35–77Google Scholar

²⁸ Martin, Lisa L., ‘Interests, Power, and Multilateralism’, International Organization, 46 (1992), 765–792CrossRefGoogle Scholar

Koremenos, Barbara, ‘Contracting around International Uncertainty’, American Political Science Review, 99 (2005), 549–565CrossRefGoogle Scholar

Rosendorff, B. Peter and Milner, Helen, ‘The Optimal Design of International Trade Institutions: Uncertainty and Escape’, International Organization, 55 (2001), 829–858CrossRefGoogle Scholar

²⁹ Conybeare, ‘Arms Versus Alliances’, pp. 215–235CrossRefGoogle Scholar

Lake, Entangling Relations, pp. 35–77Google Scholar

Morrow, James D., ‘Alliances, Credibility, and Peacetime Costs’, Journal of Conflict Resolution, 38 (1994), 270–297CrossRefGoogle Scholar

³⁰ Byman, Deadly Connections, pp. 155–186Google Scholar

³¹ This would further suggest that since both the Quetta Shura and the Haqqani Network were initially quite vulnerable to collapse at the hands of the US military following Operation Enduring Freedom it is likely that parts of Pakistan's government assisted the groups in banding together.

³² Byman, Deadly Connections, pp. 155–186Google Scholar

³³ Catanzaro, Raimondo, ‘Enforcers, Entrepreneurs, and Survivors: How the Mafia Has Adapted to Change’, British Journal of Sociology, 36 (1985), 34–57CrossRefGoogle Scholar

³⁴ Victor Asal, Hyun Hee Park, Karl Rathemeyer and Gary Ackerman, ‘With Friends Like These … Why Terrorist Organizations Ally’ (paper presented at the International Studies Association Meeting, 2010).

³⁵ We currently lack a reliable measure of the ideological compatibility of the militant groups. While one strategy might be to examine the manifestos of some of these groups, we encounter two potential problems. First, not all of the groups have public manifestos. Secondly, in forming an alliance, these groups might have incentives to misrepresent their true ideologies in order to make themselves more attractive alliance partners, or to signal that the alliance is stronger than it actually might be. While this is certainly an interesting area of study, we leave it for future research and focus on what we can currently test.

³⁶ Particularly for the ethnically or religiously based conflicts, individual actors could not be identified, but it was clear that there was more than one challenger – examples of the Uppsala/PRIO coding of this are: Sikh insurgents, Palestinian factions or sectarian factions. If there was evidence of some co-operation/co-ordination among these actors, then the entire collective was coded as having formed an alliance. We do not create dyads that allow groups that are fighting separate territorial conflicts to be considered alliance partners. For example, in the case of India, we have dyadic pairings for those groups that participated in the conflict in Kashmir and pairings for groups involved in Naxalite insurgency. However, we do not create dyads that consider the groups in the Kashmir conflict as potential alliance partners with the Naxalites.

³⁷ The eighteen observations that involved major powers as target states were removed from the analysis, though the results are robust even when they remain in the dataset. However, since the major powers’ campaigns involved interventions on foreign territory (United Kingdom–People's Republic of Yemen; France–Algeria; and Soviet Union–Latvia), it is inappropriate to consider these observations to be comparable to cases where the target government was fighting multiple groups within its own territory. The exclusion of the major power observations removed 1.04 per cent of the observations.

³⁸ The full dataset for replicating the analysis is available at: http://www.bsos.umd.edu/gvpt/bond/.

³⁹ Byman, Deadly Connections, pp. 21–52Google Scholar

Gleditsch, KristianSalehyan, Idean and Schultz, Kenneth, ‘Fighting at Home: Fighting Abroad: How Civil Wars Lead to International Disputes’, Journal of Conflict Resolution, 52 (2008), 479–506CrossRefGoogle Scholar

Salehyan, Idean, ‘No Shelter Here: Rebel Sanctuaries and International Conflict’, Journal of Politics, 70 (2007), 1–13Google Scholar

⁴⁰ Thompson, William R., ‘Identifying Rivals and Rivalries in World Politics,’ International Studies Quarterly, 45 (2001), 557–586CrossRefGoogle Scholar

Paul Diehl and Gary Goertz (War and Peace in International Rivalry (Ann Arbor: University of Michigan Press, 2000), pp. 29–34)CrossRefGoogle Scholar

⁴¹ In some cases there are more than two militant groups, which allows the sponsor more than one choice of alliance to sponsor per year. However, since our data focus mainly on the government, these data would essentially repeat. For example, in the case of Israel, four sponsors could intervene to support an alliance between Hamas/Hezbollah, Fatah/Hezbollah, Hamas/PFLP, Fatah/PFLP, Hezbollah/PLFP, etc. This would require repeating data from the years in which each of these groups are active, but the data would not be independent. To keep the observations independent, the sponsorship variable is coded as 1 for this data if the potential sponsor intervenes in any of the potential alliances. We did, however, run our analyses with the repeated observations, and the results are robust to this design.

⁴² However, these states were not excluded from the tests of Hypotheses 1 and 2.

⁴³ Jongman, Albert J. and Alex Peter Schmidt, World Directory of Terrorist and Other Organizations Associated with Guerrilla Warfare, Political Violence and Protest (Edison, N.J.: Transaction, 1988)Google Scholar

Janke, Peter, Guerrilla and Terrorist Organisations: A World Directory and Bibliography (New York: Macmillan, 1983)Google Scholar

⁴⁴ This coding of a sponsor state provides a useful alternative to and/or extension of the UCDP/PRIO coding of ‘secondary parties’ to a conflict. While ‘secondary parties’ (variables SideA2nd and SideB2nd) are identified as state actors that enter the conflict to aid the primary actors, this aid is restricted to the military sort: ‘[S]econdary parties are states that enter a conflict with troops to actively support one of the primary parties’ (UCDP/PRIO Armed Conflict Dataset Codebook, p. 2). Our definition allows for a greater range of sponsorship activities, and includes both economic and tactical support as indicators of outside support for a militant group or alliance. Additionally, where the UCDP/PRIO data restrict the universe of potential secondary parties in a conflict to those countries that ‘share the position of the primary party it is supporting in the incompatibility’, we impose no ideological restrictions on the sponsor state by our definition.

⁴⁵ Cunningham, DavidGleditsch, Kristian and Salehyan, Idean, ‘It Takes Two: A Dyadic Analysis of Civil War Duration and Outcome’, Journal of Conflict Resolution, 53 (2009), 570–597CrossRefGoogle Scholar

⁴⁶ We also created a third version of the weak link variable that simply drops categories 2 and 3, and the results were robust.

⁴⁷ Fearon, James and Laitin, David, ‘Ethnicity, Insurgency, and Civil War’, American Political Science Review, 97 (2003), 75–90CrossRefGoogle Scholar

⁴⁸ Fearon and Laitin, ‘Ethnicity, Insurgency, and Civil War’, p. 80Google Scholar

⁴⁹ Carment, David and James, Patrick, ‘Internal Constraints and Interstate Ethnic Conflict: Toward a Crisis-Based Assessment of Irrendentism’, Journal of Conflict Resolution, 39 (1995), 82–109CrossRefGoogle Scholar

Davis and Moore, ‘Ethnicity Matters’, pp. 171–81Google Scholar

⁵⁰ Fearon and Laitin, ‘Ethnicity, Insurgency, and Civil War’, p. 80Google Scholar

⁵¹ Hegre, Havard and Sambanis, Nicholas, ‘Sensitivity Analysis of Empirical Results on Civil War Onset’, Journal of Conflict Resolution, 50 (2006), 508–535CrossRefGoogle Scholar

Lyall, Jason and III, Isaiah Wilson, ‘Rage against the Machines: Explaining Outcomes in Counter-Insurgency Wars’, International Organization, 63 (2009), 67–106CrossRefGoogle Scholar

⁵² Bennett, D. Scott and Stam, Allan, ‘EUgene: A Conceptual Manual’, International Interactions, 26 (2000), 179–204CrossRefGoogle Scholar

⁵³ Carter, David and Signorino, Curtis, ‘Back to the Future: Modeling Time Dependence in Binary Data’, Political Analysis, 18 (2010), 271–292CrossRefGoogle Scholar

⁵⁴ Manus Midlarsky, Handbook of War Studies II (Ann Arbor: University of Michigan Press), pp. xiii–xxiiGoogle Scholar

⁵⁵ Singer, J. David, ‘Reconstructing the Correlates of War Dataset on the Material Capabilities of States, 1816–1985’, International Interactions, 14 (1987), 115–132CrossRefGoogle Scholar

⁵⁶ Signorino, Curtis and Ritter, Jeffrey M., ‘Tau-b or not Tau-b: Measuring the Political Similarity of Foreign Policy Positions’, International Studies Quarterly, 43(1999),115–144CrossRefGoogle Scholar

⁵⁷ Monty G. Marshall, Keith Jaggers and Ted Robert Gurr, ‘Polity IV Project: Political Regime Characteristics and Transitions, 1800–2010’, available at: http://www.systemicpeace.org/polity/polity4.htm

⁵⁸ Beck, NathanielKatz, Jonathan N. and Tucker, Richard, ‘Taking Time Seriously: Time-Series-Cross-Section Analysis with a Binary Dependent Variable’, American Journal of Political Science, 42 (1998), 1260–1288CrossRefGoogle Scholar

⁵⁹ The results do not change if we use the first weak-link indicator as opposed to the second with the controls.

⁶⁰ As a further robustness check, we re-analysed the models using all of the controls from Fearon and Laitin's 2003 study, ‘Ethnicity, Insurgency, and Civil War’. The results remain consistent and the weak-link variable along with the interaction term remain in the anticipated directions and significant.

⁶¹ The 95 per cent confidence intervals are in the parentheses.

⁶² Byman et al., Trends in Outside Support for Insurgent Movements.