WPS6179 Policy Research Working Paper 6179 Why Follow the Leader? Collective Action, Credible Commitment and Conflict Philip Keefer The World Bank Development Research Group Macroeconomics and Growth Team August 2012 Policy Research Working Paper 6179 Abstract Most analyses of conflict assume that conflicting groups civilians: leaders who prefer to limit military organization act in a unitary fashion. This assumption is often are more likely to pursue these tactics. Leader decisions violated: to reduce their risk of replacement, group regarding collective action also introduce an unexplored leaders prevent both group members and soldiers from mechanism by which state capacity is created and a acting collectively, making it difficult for leaders to make specific reason to regard state capacity as endogenous credible commitments to them. Lifting the assumption to conflict risk. This focus, finally, suggests that that groups are unitary shifts the analysis of a wide range interventions to reduce conflict risk, such as safety net of conflict issues. The effects of income shocks and rents payments or service delivery, are likely to be most difficult on conflict risk become contingent on collective action. to deliver precisely where leaders are most reluctant to Leader decisions regarding collective action explain the allow collective action and where, therefore, conflict risk forcible recruitment of child soldiers and predation on is highest. This paper is a product of the Macroeconomics and Growth Team, Development Research Group. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at pkeefer@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Why follow the leader? Collective action, credible commitment and conflict Philip Keefer Prepared for the Oxford Handbook of the Economics of Peace and Conflict Michelle Garfinkel and Stergios Skaperdas, editors. JEL codes: D70, D72, D74 Key words: Conflict, civil war, credible commitment, collective action Sector Board: Public Sector Governance (PSM) Acknowledgements: This paper has benefited greatly from the generous and insightful comments of Gary Milante, Jacob Shapiro and the editors. Contact information: Development Research Group The World Bank 1818 H St. NW Washington, DC 20433 pkeefer@worldbank.org 2 Credible commitment is at the heart of the study of conflict. However, most of the literature is concerned with the credibility of assurances by opposing groups not to take up arms against each other: when each group is unwilling to trust the other to lay down its arms or to refrain from a preemptive attack, conflict is more likely (Azam 1995 and Fearon 1995). Less attention has been paid to commitments between group leaders and members, although the credibility of these commitments influences every aspect of conflict, from a group‟s ability to mobilize military capacity to its ability to make agreements with other groups. This paper focuses on these commitments. The central argument here is that leader commitments are credible to the extent that group members and armed forces can act collectively to enforce them. However, to insulate themselves from the risk of expulsion by their own supporters or armed forces, leaders impose limits on collective action. These limits make it more difficult for them to win the hearts and minds of citizens and to field an effective military, each of which increases a group‟s vulnerability to attack by opposing groups. Fearon (2008) describes the modal civil war in the last 60 years as persistent, small and relying on guerrilla tactics, rather than as a conflict between large conventional armies, as in the United States Civil War. However, restrictions on collective action also seem to distinguish the modal conflict from the U.S. Civil War. The Confederacy allowed substantially more collective action by (white) citizens in its territory than did the Tamil Tigers in northern Sri Lanka, the Shining Path in the Peruvian Andes, or the Democratic Forces for the Liberation of Rwanda in the eastern part of the Democratic Republic of Congo (DRC). Union forces were not comprised of a well-funded Presidential Guard and a poorly-funded regular army, as in the DRC, or closely directed by the president, as in Sudan. The discussion below therefore examines the decisions of leaders to allow citizens and soldiers to act collectively. Reviews of qualitative and quantitative evidence, particularly related to political parties, indicate that citizens in countries vulnerable to conflict exhibit less ability to act collectively. Ample qualitative evidence also demonstrates significant differences in the extent to which governments allow militaries to organize collectively: inter-unit cooperation, information transmission within the military, and military control over promotions and honors are all more heavily restricted in some countries than others. The dynamics of within-group collective action have significant implications for debates across the conflict literature. Recent research concludes that income shocks encourage conflict by changing the costs of conflict (the wages of combatants) more than the rents at stake in conflict. The discussion here concludes that these arguments apply most strongly in precisely those countries where armed forces cannot act collectively to enforce a claim on future rents or where citizens cannot act collectively to prevent the armed forces from making such a claim. The discussion is also relevant to particularly disturbing features of conflict, such as predation on civilians and coerced recruitment of child soldiers. The first could be a direct consequence of leader efforts to undermine organization and discipline within the armed forces in order to defuse coup threats. The latter is usefully seen as part of a class of measures that leaders take to reduce coup threats, since adolescents are less likely than adults to organize a rebellion against them. Scholars have highlighted the role of state capacity in shaping the vulnerability of countries to conflict. That research generally focuses on the fiscal costs of capacity-building and only recently has recognized that capacity is not an exogenous determinant of conflict 3 risk, but one that is determined, jointly with conflict risk, by other factors. The discussion here emphasizes that capacity is a function of organizational choices that leaders make, which are independent of financing. The constraint on capacity is not budgetary, in this case, but the threat that capacity (a better organized public administration or military) might pose to leader tenure. The conjectures developed here justify a closer examination of current strategies to minimize conflict risk. For example, in view of their finding that rainfall (and, therefore, income) shocks trigger conflict in Africa, Miguel et al. (2004) propose the adoption of crop insurance and safety net payments. The donor community has begun to focus on improved service delivery in post-conflict settings as a way to reduce tensions and build up state legitimacy. However, if conflict is most likely in countries where leaders are reluctant to allow citizen collective action, these strategies may be least likely to succeed precisely where they are most needed. Leaders who do not allow collective action are less likely to provide services effectively, to target safety net payments appropriately, or to enforce insurance contracts. Credible commitment, collective action and conflict Most of the conflict literature examines the effects of political institutions on conflict from the perspective of the credibility of inter-group agreements. As Garfinkel and Skaperdas (2007) observe, scholars tend to assume that antagonists are unitary actors and abstract from problems of intra-group collective action, regardless of whether they use bargaining models that focus on the inability of antagonists to make credible commitments, or contest functions that examine the sacrifices that groups make in terms of productive activity in the pursuit of conquest. There are exceptions to this rule. Kaplan (2010) uses evidence from Colombia to support his argument that villages with greater ability to organize collectively are better able to resist the pressures placed on them by armed forces (insurgent or government). Garfinkel (2004) analyzes decisions governing group size. Larger groups are more likely to prevail in conflict, but individuals in larger groups have to devote more resources to protect their share from other group members. Larger groups are therefore more likely to emerge when institutions of conflict management, exogenously determined, are better established within a group, attenuating intra-group conflict. The focus here is: To what degree do leaders of insurgent or government forces encourage collective by citizens and armed forces, and how do the efforts of group leaders to remain in power shape group institutions? In particular, the next section describes leader decisions to allow collective action by supporters and examines the consequences of these decisions for the ability of leaders to wage the battle for “hearts and minds�. The section following considers the decision to limit collective action by armed forces, even if this degrades military effectiveness. The final sections of the paper then outline the likely implications of variation in intra-group institutional arrangements for the role of income and income shocks in precipitating conflict, predation on civilians, reliance on child soldiers, state capacity, and improved service delivery as a strategy for defusing conflict. Collective action and the battle for hearts and minds Berman, Shapiro and Felter (2009a) point to a consensus among prominent practitioners of insurgency and counterinsurgency, from Mao Tse-Tung to David Petraeus, 4 on the importance of gaining popular support in conflict outcomes. According to practitioners, and scholars such as Kalyvas (2006) and Fearon (2008), the key strategic advantage from gaining the support of non-combatants is the provision of information about the armed forces of the opposition. Problems of within-group credible commitment (such as whether citizens can rely on leaders to pursue their interests in the event that they prevail in the conflict) are not the central focus of these analyses. Nagl (2002) places great weight on the importance for counterinsurgency strategy of gaining the trust of local actors. He focuses most on how trust matters and less on how it emerges. One way it can emerge, however, as is implicit in Kaplan‟s (2010) analysis, is if local actors are able to act collectively, allowing them to impose a larger sanction on leaders who renege on their commitments. The literature portrays the battle for hearts and minds in two ways; collective action by citizens is relevant to both. One is the struggle by competing groups to make credible commitments to civilian populations to reward them for cooperation. Another is their struggle to deliver the benefits of good government to these populations, such as public security, social services, or predictable and non-predatory taxation. Collective action by civilian populations plays a central role in both cases: collectively organized civilians can more easily punish groups that renege on their commitments or who fail to provide the benefits of good government. The most-studied institutional arrangement through which citizens can hold leaders accountable is voting. In fact, the absence of competitive elections is a notable feature of conflict countries. Systematic data supports this claim in the case of governments; casual observation offers little indication that insurgent-controlled territories are any different. The Sambanis (2004) conflict database records 71 conflicts from 1975 – 2000 for which data on competitive elections (from the Database of Political Institutions) are available. In only 15 of these cases were countries governed by competitively elected leaders in the year before the conflict started. In the remainder, governments were not competitively elected or not elected at all. Indeed, this empirical regularity is as striking as the association of conflict with income: of these same 71 conflicts, again only 15 (not the same 15) occurred in the richest 50 percent of countries. Although the paucity of elections in conflict countries is an indication that leaders in conflict countries impose limits on collective action by citizens, elections are not a sufficient condition for effective collective action. Even citizens with unfettered voting rights are handicapped in their ability to hold leaders to account when they are constrained by imperfect information (the inability to observe leader actions or the effects of those actions on their welfare). More importantly for the analysis here, though unfettered voting rights lower the individual costs of holding leaders accountable, they do not guarantee that challengers will emerge who can credibly commit to pursuing different policies than those of the incumbent. Electoral accountability of leaders depends on the existence of such challengers. Ferejohn (1986) and Persson and Tabellini (2000) examine the case where credible challengers do not exist. Their absence substantially attenuates incumbent incentives to pursue the public interest (e.g., to provide public goods). These results extend immediately to the conflict setting: such leaders, even if elected, have limited incentives to win hearts and minds through the provision of public services. 5 An important open question is how political competitors develop the capacity to make credible promises. A likely answer, though, is rooted in the ability of citizens to act collectively to choose and support candidates, working especially through programmatic political parties that are organized to represent the interests of like-minded citizens (Keefer 2007b). Such parties allow members to coordinate with each other; offer them fora at which they can select candidates; require potential new members or candidates to invest in costly signals that they share the goals of the existing members (e.g., as in Snyder and Ting 2002); and exhibit procedures to expel members who do not pursue the group‟s collective interests. Challengers chosen in such a process are better able to make credible commitments because the existence of the party ensures the continuing ability of citizens to act collectively to choose a different candidate if challengers renege on those commitments. Other research extends this logic to non-democracies. Even if citizens cannot vote, autocrats can nevertheless decide to admit a certain fraction of the society into the ruling party and to permit members of the ruling party to act collectively. Gehlbach and Keefer (2009), for example, argue that when autocrats allow ruling party members to share information about autocrat behavior that is not available to others, members can act collectively to punish leader expropriation of their investments. By using ruling party institutionalization to limit their rents from expropriation, autocrats can attract greater private investment and can elicit greater effort from party members in the pursuit of leader objectives. These objectives could include insurgency (for leaders of rebel groups) or counter-insurgency (for government leaders). This discussion suggests, then, that whether leaders try to win hearts and minds depends not only on the institutions that affect inter-group contracting and that are the focus of the conflict literature, such as elections, electoral rules (proportional representation or plurality elections, for example), or political checks and balances. They also depend on the intra-group institutional arrangements that allow leaders to credibly commit to pursue the interests of supporters. These arrangements are those that allow large groups of supporters to act collectively. This logic predicts that governments should confront a greater risk of insurgency by opposing groups when they are reluctant to allow their own supporters to organize, which in turn is most likely to be the case when competition for political office is not mediated by institutionalized parties (parties, for example, that facilitate collective action by members regarding the selection of party candidates). In the absence of such parties, the probability of insurgency success rises since government ability to win the battle for hearts and minds through the provision of public services is attenuated. Keefer (2008) presents evidence to this effect. In that analysis, each of the following proxies for the ability of citizens to act collectively has a significant effect on conflict risk: the continuous years of competitive elections (capturing the ability of political competitors to make broadly credible commitments to citizens, as in Keefer 2007); the degree to which parties convey a programmatic stance to citizens (since such a stance is only credible if parties have organizational arrangements that force out leaders whose actions are inconsistent with the party program); and the age of the ruling party relative to the years that the leader has been in office (ruling parties created by leaders are both likely to be younger than the leader‟s years in office and under the control of the leader). When values of any of these variables are lower, conflict is significantly more likely, particularly in the poorest 50 percent of countries 6 where more than 75 percent of all conflicts occur. Every additional year by which the age of the governing party exceeds the years a ruler has been in office reduces the odds of conflict by about 2 percent a year. The Democratic Republic of Congo provides a specific illustration of the association of weak collective action with conflict. In the 2006 elections, the first after the cessation of widespread conflict (although guerrilla activity persisted in the eastern provinces), 213 parties competed for parliamentary seats. Following the elections, 14 parties were represented in the government and 70 in the National Assembly. Such fragmentation is consistent with the inability of parties to make credible commitments to serve the interests of broad groups of citizens. It is also consistent with the fact that, except for the brief periods from 1960-1967 and 1990-1997, political parties had been banned in the DRC, offering no historical basis for party-based political competition in 2006 (International Crisis Group 2008 p. 15). If parties are based on personal relationships and clientelist ties, as in the DRC, leadership changes should have a dramatic effect on party stability. Consistent with this, when Jean-Pierre Bemba, the leader of the opposition Movement for the Liberation of Congo, left the country, a significant number of the party‟s legislators crossed to the government coalition (Oxford Analytica May 9, 2009; “Congo-Kinshasa: Kabila gains from opposition erosion.�). Finally, if clientelist ties are the foundation of partisan organization and party members are not able to act collectively, legislators from a party confront little electoral risk in casting votes that are inconsistent with the preferences of the party‟s voters. Consistent with this, after the 2006 elections, in nearly all provinces legislators elected governors from the president‟s coalition, despite the fact that in many provinces, the president‟s coalition won only a small fraction of the vote (“Congo-Kinshasa: Clashes highlight post-poll challenges,� OA, March 26, 2007). This discussion points to a strong association between conflict and the ability of citizens to act collectively. It has not touched on the question of why citizens are organized for collective political action in some countries, but not in others. This is again a largely open area in the literature. In an early effort, Keefer and Vlaicu (2008) argue that politicians compare the costs of building political credibility to the political advantages of being able to make broadly credible commitments to them. When costs are high, politicians choose to make credible pre-electoral commitments only to narrow groups. Such politicians then favor low public good provision; high rates of private good provision to those narrow groups; and to engage in high rates of rent-seeking – all contrary to the objective of winning hearts and minds. They are particularly likely to make this decision where, as in many conflict countries, patron-client relationships are deeply-rooted in society, making narrow appeals cheap relative to broad-based appeals. In many parts of Afghanistan or Iraq, for example, it is easier for political competitors to build support by making commitments to patrons, who have no interest in broad public good provision, than to invest in the ability to make credible commitments to broad groups of voters. Alternatively, though, the choice for politicians may not be how much to invest in building their credibility. Instead, it may be whether to form a party in which members, rather than they themselves, are allowed to choose candidates. The less sure they are that they will have the collective support of party members, the less likely they are to organize a party in which members have free rein to act collectively. This tradeoff re-emerges in the 7 next section: leaders are more likely to allow the armed forces to act collectively, increasing both their military effectiveness and their ability to launch a coup, the lower the risk that the armed forces will want to launch a coup.1 Collective action and military effectiveness Military capacity is naturally a central issue in conflict, but most analyses abstract from the leader‟s dilemma that armed forces able to undertake collective action, and therefore more effective militarily, can also demand a higher share of rents from leaders and threaten the leader with replacement. This section points to the advantages that well- organized security forces offer to leaders; documents decisions by leaders that intentionally impede the efficacy of their armed forces; and traces these decisions to leaders‟ fear of overthrow by well-organized troops. It is easy to see that the ability of a leader‟s armed forces to act in a coordinated manner – collectively – is essential to effective performance; examples below show how leaders in some countries impose large barriers to joint exercises by their country‟s own air and ground forces, with obvious implications for military readiness. Another advantage, discussed in Weinstein (2005) and Keefer (2008), relates to the contracting of soldiers. Leaders who cannot credibly promise future rewards to their soldiers (government or insurgent) must pay them in the form of spot payments that fully cover the reservation wage of the fighters.2 Their ability to recruit a fighting force depends entirely on the rents that they control during the conflict period. Since conflict itself tends to degrade the productive capacity of countries, leaders who cannot make credible commitments to their soldiers regarding future compensation are more likely to be compelled to rely on natural resource rents and external funding. One of the most successful insurgent groups, the Viet Cong, recognized this: long after the Vietnam War, veterans of the North Vietnamese army and their children continued to receive privileged access to jobs and housing.3 Keefer (2008) argues further that even if leaders have sizeable rents at their disposal, they might still be reluctant to rely on spot contracts with soldiers. Effective armed forces are rarely constructed on the basis of spot contracts because military effort is difficult to observe and bravery hard to purchase. One way that leaders can promote effort is to credibly commit to pursuing national goals that are compatible with those of the armed forces.4 Another is to offer ex post rewards for successful military action, whether medals or promotions. However, these promises must be credible if they are to have an effect. 1 A similar tradeoff is central to the democratization literature. Under what conditions do elites allow elections (e.g., as in Acemoglu and Robinson 2006 and Boix 2003)? The decision is difficult because of a well-known double-edged commitment problem: elites cannot promise to refrain from expropriating non-elites if they do not allow elections; non-elites cannot promise to refrain from establishing high tax rates at the expense of elites if elections are allowed, prompting elites to resist elections. 2 The logic here is similar to that in Schultz and Weingast (2003), who argue that leaders who can credibly commit to repay loans have an advantage in conflicts with countries whose leaders cannot. 3 Author interview, Hanoi, February 24, 2006. This could reflect the possibility that former Viet Cong enjoy “insider status� and privileged access to rents; the examples given suggest that even former soldiers with little influence on government decisions enjoy these privileges. 4 Weinstein (2005) and Humphreys and Weinstein (2006) emphasize that pecuniary compensation matters less when soldiers and leaders share the same ideological objectives. Even here, however, soldiers require some 8 Although organized militaries, capable of collective action, are more effective in conflicts with regime opponents, they also pose greater threats to leaders themselves. Leaders who are more fearful of coup threats than of attacks by regime opponents are therefore more likely to make decisions that lead to military disorganization. These decisions can take many forms. One is to starve the armed forces of resources. Previous research has emphasized access to resources as the key to military effectiveness, to meet the usual requirements of salaries, supplies and weapons. Consistent with this emphasis, military dysfunction and low salaries seem to go together. In the Democratic Republic of Congo, for example, salaries of top officers in the army, known for its ineffectiveness, were approximately US$80 per month (International Crisis Group 2006, pp. 5, 11). However, under-funding may not signal the leader‟s lack of resources, as the literature emphasizes. Instead, it could result from the leader‟s reluctance to fund a potential coup threat. In the DRC, although resources were not available for army offices, ministers‟ salaries approached US$4,000 per month. Leaders concerned about coup threats can also create separate and competing security forces and prevent them from cooperating with one another. De Atkine (1999) reports that joint commands and exercises are rare in many Middle Eastern militaries and efforts by one branch of the military to secure the assistance of another (for example, if an army wishes to obtain aircraft from an air force for army airborne training), must be coordinated by heads of services at the ministry level, and likely requires presidential approval. Sadaam Hussein‟s management of the Iraqi armed forces provides the most extreme example of this. Hashim (2003) summarizes his discussion of Hussein‟s policy throughout his rule to rotate officers, to replace even successful officers with incompetent loyalists, and to establish competing armed forces, even at substantial cost to military readiness. “Even though Iraq faced a higher threat environment in the 1990s vis à vis both external and internal threats, in [Saddam Hussein‟s] mind it was the most proximate force to him, the Iraqi military, which continued to represent the gravest threat� (p. 14). The DRC military was also divided. In 2006, the army was 140,000 strong and poorly funded, but the Presidential Guard, with the sole purpose of protecting the president, had 12,000 soldiers and was relatively well-funded. In July 2006, for example, 42 tanks and armored vehicles were officially delivered to the Congolese army, but only the Presidential Guard had forces trained to use them (Oxford Analytica (2006); “Congo-Kinshasa: Historic poll not end of transition,� October 27). Another way to inhibit intra-military coordination is to prohibit information-sharing across military units.5 Focusing on armies in the Arabic countries of the Middle East, De Atkine (1999) indicates that within-military flows of information are tightly restricted, another indication of limitations on collective action. In particular, promotions, transfers, names of unit commanders, and unit designations are all frequently classified. Observers also find that leaders in the region prevent independent collective action within the military assurance that leaders will not pursue goals that are incompatible with those objectives; collective organization is again important. 5 As Gehlbach and Keefer, 2009 demonstrate, limits on information flows are a significant obstacle to collective action within ruling parties 9 by limiting the delegation of authority: soldiers and officers who are denied the discretion to undertake independent action can less easily initiate collective action against the leader. Gehlbach and Keefer (2010) contrast the decisions of some unelected leaders, who make coordination within the military difficult to minimize coup threats, even at the expense of military effectiveness, with the experience of Indonesia under Suharto. After defeating the Communist insurgency, Suharto made significant efforts to unify a fragmented military. In the immediate post-insurgency era, security threats loomed larger than coup threats. Consequently, Suharto introduced a joint command and permitted strategies that gave substantial discretion to lower level officers to collaborate with local civilian authorities. These decisions were well-suited to combating the re-emergence of insurgent threats, at the cost of improving the military‟s coup capability. In the later years of his regime, however, Suharto changed strategies, concerned that he had more to fear from a military from which his peers had retired and less to fear from insurgents after years of fast growth. He inserted himself into the military‟s promotion decisions, placing close associates in high positions; he more overtly “played powerful generals off one another, probably resulting in multiple informal chains of command that led only to Suharto� (Callahan, p. 13). Discipline broke down among the lower ranks of soldiers “and was non-existent in many regions� (Callahan, p. 15).6 The effects of de- institutionalization were evident in May 1998, when the military played little role during the popular uprisings that brought the regime down (Callahan, p. 15). These examples come from the government side, but the logic applies as strongly to insurgent armed forces: coup threats constrain insurgent leaders‟ decisions about the organization of insurgent military forces. For example, in 2002, an ethnic Tutsi, Laurent Nkunda, commanded the 7th Brigade of the Congolese National Army – the rebel army fighting the government of Laurent Kabila in the DRC. After a failed coup attempt by non- Tutsi members of the brigade, he purged the brigade of non-Tutsis, executing at least 160, reducing the coup threat at the expense of his capacity to project force (Oxford Analytica, October 24, 2007, “Congo-Kinshasa: Nkunda key to eastern security�). The fear of coup threats (against either government or insurgent leaders) also provides a complementary explanation of the organizational decisions of rebel leaders that Weinstein (2005) analyzes. He focuses on Uganda‟s National Resistance Army and Mozambique‟s Renamo. The first began with an ethnically homogeneous leadership that initially recruited co-ethnic (Banyaloke) fighters – “loyal� troops, in the context of the model, to whom leaders could most easily make credible commitments and which were least likely to engage in coup activity. The NRA later expanded its forces by recruiting non- Banyalokes, but focused primarily on the Baganda, the group with which Banyaloke leaders had developed strong ties (e.g., the capacity to make credible commitments). Renamo, in contrast, began with a highly ethnically fragmented leadership drawn from the ethnically- mixed exile community; this leadership group recruited from many ethnic groups and areas of Mozambique. Compared to the NRA leaders, they had a more limited ability to make credible commitments to their troops. 6 The motivation for deinstitutionalization is not documented, but is plausibly one of the following: a decline in the risk of insurgency (reducing the value of an institutionalized military); an increase in internal threats to the regime from the military; and, related to this, a potential decline in the willingness of the regime to share rents with the military. 10 Weinstein (2005) emphasizes rents as the explanation for the different recruitment strategies. The NRA had no access to rents and could only compensate soldiers with promises of future compensation. It recruited soldiers with whom the leaders had ethnic ties that made promises of future compensation more credible. Renamo, in contrast, initially received significant support from Rhodesia and did not need to rely on promises of future payment. When Rhodesia fell and support declined Renamo then turned to coercion to ensure troop performance.7 The arguments here are consistent with this explanation, but they also suggest that the NRA and Renamo might have pursued similar strategies even if they had had equal access to rents. Only the NRA had an ethnically homogeneous leadership. This gave it the option of reducing the barriers to collective action among its soldiers with an ethnically homogeneous recruitment strategy. Moreover, when it sought to expand the rebel force, it focused on only one group, the Baganda, ensuring that new rebels were similarly able to act collectively to enforce agreements. In contrast, the more ethnically fragmented early leaders of Renamo could not easily pursue this strategy. Disproportionate recruitment of any one ethnic group would have put leaders from any other ethnic group at a disadvantage, since they would be more subject to a coup threat than the co-ethnic leaders. Renamo was correspondingly unable to make credible commitments to soldiers, neither during the period when they had access to ample external assistance, nor during the period when they did not. They relied instead on high current payments when they had access to external assistance and resorted to coerced recruitment – not the recruitment of co-ethnics – when they did not. In general, conflict models abstract from coup threats and leader decisions regarding the organization of the military organization. The qualitative evidence suggests that this is a potentially important omission. Garfinkel and Skaperdas (2007) review a series of conflict technologies. All share the feature that the probability that one party wins a conflict increases in its own material inputs (guns) and decreases in the guns deployed by the other party. The efficacy with which guns are deployed on each side is a parameter of these models, not a choice variable. Furthermore, in most models, all rents are available for leaders to compensate military effort, whether or not leaders control them. This implies that leaders can credibly commit to share rents with troops in the event of victory. A key determinant of both efficacy and the credibility of compensation commitments is the degree to which soldiers can act collectively. One straightforward way to formally introduce the organization of the military into standard conflict analyses is to allow leaders to choose to rely on two separate armed forces. One of the forces might be personally loyal to the leader (such as a presidential guard). Leaders can rely on these personally loyal soldiers not to undertake a coup and to defend them if a coup attempt by other soldiers occurs. While they confront no coup risk if all of their armed forces are drawn from this group, loyal troops are less effective in conflict. In addition, the supply of loyal troops is likely to be limited. 7 An outside group, such as Rhodesia, that wants to exert strong influence over one group in a conflict might even encourage ethnic diversity, since the less credible are leader commitments to rebels, the more that leaders must rely on outside financing. 11 The probability of a successful coup against the leader is then given by p c G , G L   g G , G L , p c  0 for G  0 . The probability increases in G (the size of the contingent of “professional� soldiers), falls in GL, and the cross derivative is negative: an increase in G raises the probability of coup by less the larger is the number of soldiers personally loyal to and willing to defend the leader, GL. The probability that group i will prevail in conflict against group j is then given by the usual function f G i , G Li  p i G 1 , G 2 , G L 1 , G L 2   ,    f G i , G Lj  f G j , G Lj  where f is the conflict function that translates material inputs into military effectiveness and is assumed identical for both groups. The probability of group i‟s success rises with its own conflict effectiveness, f G i , G Li  , and falls with group j‟s, f G j , G Lj  . Unlike the coup function g, the conflict function f increases in both G and GL, though increments to G contribute more to the probability of conflict success than funding for loyal soldiers GL f f ( > for G = GL). Again the cross-derivative is negative: additional soldiers G G L contribute less to victory the larger is a leader‟s reliance on two, independently organized groups of soldiers. In the literature, antagonists typically choose G subject to the foregone rents or production losses they incur by transferring resources to appropriative activities, as in the models reviewed by Garfinkel and Skaperdas (2007). Coup risk introduces a second constraint. The greater is coup risk, the more reluctant are leaders to improve their capacity to conduct insurgency or counter-insurgency. To see this most simply, it is useful to assume that all conflict activities are financed out of exogenous rents. The leader of group i chooses G and GL to maximize the returns from conflict, V i G1 , G 2 , G L1 , G L 2    p i  p ci R  G i  G Li , where expected leader rents are a function of the probability of prevailing in conflict, p i , less the probability of being expelled by a coup, p ci . This yields first order conditions (holding constant the response of group j),    p  p ci  (1)  i  R  1  0   G i G   i   p i  p ci  (2)   G  G R  1  0 .   Li Li  From (1), the greater the threat that additional regular forces pose to the leader,  p ci , the fewer of them the leader will recruit. From (2), the larger the contribution that G i  p ci additional expenditures on personally loyal soldiers make to coup prevention, , the  G Li more of them the leader hires, even though their addition to the leader‟s armed forces contributes less to conflict success than additional expenditures on regular forces. 12 Coup threats also attenuate leader responses to rents. Rewriting (1) as p i  p ci 1 p i  p ci   and under the usual assumptions that is declining and increasing G i G i R G i G i in G i , it follows immediately that coup risk suppresses leader investments to pursue rents, holding constant the response of the other group. The leader can only incompletely offset this effect by investing in a parallel security service, since this investment contributes less to overall military effectiveness than investments in a unitary military staffed by regular troops.8 This simple setup makes two substantial assumptions about rents. First, coup threats are exogenous. If they were determined within the model, they would likely rise with the rents at stake. This would tighten the coup constraint confronting the leader and make him more likely to deviate from organizational choices that optimize military effectiveness. Second, leaders‟ ability to finance military expenditures is assumed to be limited by the total rents at stake, not the rents that the leader controls. This is reasonable if the leader can make credible commitments to pay out of future rents in the event that his group prevails. However, future commitments to the “professional� soldiers, G , become less credible as the leader relies more on “loyal� soldiers, GL . The cost of relying on professional soldiers is therefore not constant, but rising in the fraction of loyal soldiers that the leader hires. The model abstracts from another important characteristic of real world insurgencies to which future research might turn: coup threats and the cost in terms of military effectiveness of relying on loyal troops (that is, the function f ) are heterogeneous across groups. The NRA could rely on an ethnic strategy as a basis for credible commitment between leaders and troop; Renamo could not. Laurent Nkunda could rely on the ethnic strategy (many of his relatives were killed in anti-Tutsi uprisings, making his appeals to Tutsi soldiers particularly credible), but also on religion. “In Masisi [in 2004] Nkunda – a long- standing „born-again‟ Christian – began to receive heavenly visions informing him that he was to be a savior for all Congolese Tutsis. His men – believing they were on a „divine mission‟ – entered Bukavu in June 2004 following Nkunda‟s claim that a „genocide‟ of Tutsis was taking place there. . .� (Oxford Analytica, October 24, 2007, “Congo-Kinshasa: Nkunda key to eastern security�). In none of these cases could government leaders rely on similar appeals to expand their stock of loyal troops. These considerations are particularly important for the empirical analysis of conflict, which tends to rely on measures of financial strength or access to rents to compare army and insurgent capacity. However, a small, poorly funded group of soldiers that believes the commitments of its leaders and presents little coup risk may have greater military capacity than a large and well-funded group for which these conditions are not met. Credible commitment, investment and income shocks Leader decisions to allow collective action by citizens and security forces influence a number of debates in the literature. One of these concerns the roles of income and rents in conflict. The most well-established empirical regularity in the study of conflict is that poor countries are more likely to experience conflict than rich countries: 75 percent of conflicts 8 The model in the previous section abstracts from strategic considerations by coup-plotters. However, rents raise the payoffs to coups just as they do in the case of insurgency, exacerbating the reluctance of leaders to allow their armed forces to act collectively. 13 from 1975 – 2000 occurred in the poorest 50 percent of countries. Negative income shocks are also frequently associated with a higher risk of conflict. Similarly, natural resource rents seem to be associated with greater risk of conflict, though significant outliers (e.g., Norway) raise persistent issues about the mechanisms through which this occurs.9 The income-conflict association is not easy to explain. Fearon (2008) observes that in poor countries, the costs of building military capacity (the reservation wages of potential recruits) are low, but so also are the potential rewards from conflict. Several explanations have emerged to grapple with this paradox, either by pointing out that income shocks need not fall proportionately across all sectors (labor- and capital-intensive) of the economy, or that temporary income shocks have a larger effect on the costs of waging conflict than on the future rents earned by conflict success. All abstract from intra-group collective action. Dal Bó and dal Bó (forthcoming) observe that income shocks may occur in the capital- or labor-intensive sectors of the economy. Because conflict is labor-intensive, income shocks to capital-intensive sectors have a larger effect on the stakes in conflict; shocks to the labor-intensive sectors have a larger effect on the costs of waging conflict. Dube and Vargas (2007) find evidence for this logic in Colombia. Coffee is labor-intensive, and violence fell in coffee-producing regions when the international price of coffee increased; oil is capital-intensive, and violence increased in petroleum-rich regions of the country when the price of oil rose. Besley and Persson (2009a) make a similar argument and show that increases in the GDP-weighted prices of countries‟ commodity exports increase the probability of conflict. These analyses are focused on income shocks, rather than the large average differences in conflict incidence across rich and poor countries. However, to the extent that poor countries are more reliant on commodity production than rich countries, they help to explain the higher frequency of conflict in poor countries. Differences across countries in the ability of citizens and armed forces to act collectively can also explain these results, however. The decision by leaders to limit the organization of citizens and armed forces has four effects. Three of these arise because citizens who cannot act collectively are more exposed to opportunistic behavior by leaders. First, citizens are more reluctant to invest, so incomes are lower. Second, they are more likely to engage in labor-intensive production for the domestic market, which is less vulnerable to expropriation. Third, any capital-intensive production that occurs is likely to be concentrated in high-return activities, offsetting expropriation risk, such as copper or gold extraction where the copper or gold are closer to the surface or the ores are particularly rich. The fourth effect of leaders‟ restrictions on collective organization is clear from the earlier discussion: it reduces military readiness (making insurgency less costly) and forces the leader to rely on current rents to sustain military effort (hence the association of rents and conflict). Taken together, these could explain both the income-conflict association (countries where collective action by citizens and armed forces is more difficult are both poorer and more vulnerable to insurgency), and the association of income shocks with conflict (countries where collective action is difficult are likely to have larger labor-intensive sectors sitting side- by-side with high rent commodity exports). 9 The micro evidence on income and violence is less clear, however. Berman, et al. (2009b) find robust evidence of a negative relationship between unemployment and attacks on government forces in both and the Philippines. 14 Some evidence is consistent with this alternative view. One is that weak constraints on leader‟s actions towards their own group are characteristic of conflict countries. Data already cited show that conflicts occur disproportionately in countries without competitive elections. Subjective measures of opportunistic behavior by governments (the rule of law and corruption indicators from International Country Risk Guide, available for 39 conflict countries) tell a similar story: they stood at 2.1 and 2.6 the year before conflict, compared to 3.8 and 3.5 for all non-conflict country/years (higher scores indicate greater rule of law and less corruption; the maximum is six). Chassang and Padró i Miquel (2009) observe that a negative productivity shock reduces the reservation wages of potential military recruits. However, because shocks are transitory, they have a negligible effect on the value of future production around which conflicts are fought. Two groups each choose a subset of their members to fight the other group. Non-fighters continue to be productive, paying wages to fighters sufficient to cover their opportunity costs of foregoing productive activity. A negative economic shock reduces this opportunity cost more than it does the future rents from productive activity, triggering conflict. However, this asymmetric effect depends on another, implicit assumption: that armed citizens refrain from using their military advantage to expropriate a share of all future rents. The greater the share of future rents that armed members of the group can capture, the less it is the case that temporary income shocks reduce the costs of conflict more than the stakes of conflict. This implicit assumption holds precisely in countries where armed forces are disorganized and less able to enforce their claims on future rents. Among countries where this is likely to be the case, the observed link between income shocks and conflict is predictably strong. Miguel, et al. (2004) show that negative income shocks, instrumented by rainfall, have a large, positive effect on the probability of civil war in Sub- Saharan Africa. They emphasize that this region is appropriate for their test because it relies heavily on rain-fed agriculture. It is also appropriate, however, because in this region, armed groups are particularly unable to enforce claims to future rents. In settings where they can, income shocks would be expected to have a smaller effect. In OECD countries, the armed forces are exceptionally well-organized and yet do not make large claims on economic rents. The conditions under which armed forces exercise restraint in predating on the population are another area where more research is needed. However, one promising explanation for this restraint is the level of organization of unarmed citizens. In these countries, the capacity for collective action by citizens, whether through well-established political parties or local governments, allows them to more easily resist military efforts to extract a share of future rents. It also allows them to punish leaders who tolerate undisciplined armed forces, however, an explanation for high levels of organization of security forces. The evidence in Besley and Persson (2009) can be interpreted in a similar light. They demonstrate that higher prices for a country‟s commodity exports increase the risk of conflict only in countries with low scores on the Polity IV index of executive constraints. They take this as support for their argument that rents cause conflict when political checks and balances are absent that would compel governments to share rents with the opposition. The alternative interpretation the emerges from the discussion here is that weak executive constraints are an indication that citizens cannot act collectively to restrain leaders, so that leaders refrain from sharing rents with everyone, and not only the opposition. In such 15 countries, conflict risk increases following a rise in commodity rents because leaders who do not allow collective action are more likely to have disorganized militaries. Insurgents confront lower costs of violently pursuing higher rents in these countries than in countries that exhibit significant constraints on the executive. The validity of this alternative interpretation depends on what, exactly, the Polity measure of executive constraints captures. Keefer (2010b) shows that the subjective executive constraints measure is highly correlated with objective measures of political checks and balances, consistent with the interpretation traditionally given the variable in the literature. However, even after controlling for checks and balances, it is just as strongly associated with the age of the largest government party and the years of continuous competitive elections, both of which relate to the degree to which citizens can act collectively to hinder opportunistic behavior by leaders. Other implications: Predatory behavior, child soldiers, state capacity, ethnicity and service delivery Leaders‟ decisions to allow collective action affect numerous other issues in the study of conflict. Two are the problems of predation on civilian populations and the forced recruitment of child soldiers. Others relate to arguments that state capacity or ethnic fragmentation raise conflict risk. A fifth is the degree to which external agencies should focus on improved service delivery as a way to defuse the risk of conflict recurrence in post- conflict countries. This section briefly reviews these. Predation Predation by both soldiers and rebels on the local population is well-documented in many conflict settings, including the DRC.10 Gates (2002) argues that predation is more likely when leaders cannot easily control shirking by their soldiers in their conflict effort, making them more tolerant of looting.11 Humphreys and Weinstein (2006) argue that organization within fighting units (rather than the ability of leaders to supervise fighting units) drives predation. The evidence they assemble from Sierra Leone shows that fighting units that exhibited less discipline (e.g., more drunkenness or within-unit conflict) were more likely to predate on civilians. Kalyvas (2006) argues that violence against civilians could also be spurred by private efforts to settle scores. In all three approaches, predation is the consequence of the inability of leaders to prevent individual fighting units or soldiers from taking private actions that undermine, or at least do not contribute to, overall conflict success. The earlier discussion about collective action bears directly on the contractual relationship between leaders and soldiers. If predation hurts conflict success, and if leaders can credibly promise future rewards based on conflict success, soldiers are less likely to predate. They are also more likely to support leader punishment of soldiers who predate, since such soldiers threaten the rewards of all soldiers. However, leaders (either of the entire group or of fighting units within the group) who fear a coup threat are likely to discourage 10 Predation here is understood as non-strategic violence undertaken by fighting units or individual soldiers in pursuit of their own private objectives. 11 Padró i Miquel and Yared (2009) examine a related issue in deriving the optimal strategy of a government interested in quelling local disturbances through the use of local actors who can do this at lower cost. 16 the collective organization by soldiers that is essential for the credibility of leader commitments. Predation, therefore, is likely to be more common where leaders have a greater fear of overthrow. More generally, where militaries are not organized for collective action, military indiscipline and predation are more likely; where citizens cannot act collectively, leaders bear fewer political costs from allowing predation. The organizational arrangements that allow a military to curb deviant behavior by its members are less effective to the extent that leaders intervene in military decision making in order to subvert potential collective action by the military against them. For example, leaders are unlikely to discipline personally loyal troops, on whom they depend for their own security, for predating on civilians. At the same time, in countries where citizens are unable to act collectively, they can impose fewer political penalties on leaders who do not insist on disciplined security forces.12 The connection between organizational dysfunction, a lack of military discipline and predation is evident in the DRC: just prior to the hotly-contested 2006 elections, observers estimate that officers embezzled as much as one-half of their soldiers‟ salaries (International Crisis Group 2006, p. 10). Child soldiers The problem of collective action can also explain reliance on child soldiers. Beber and Blattman (2008) examine the coerced recruitment of child soldiers by the Lord‟s Resistance Army in Uganda, concluding that when leaders lack material resources and local popularity, they are more likely to rely on coercion as a way to build military capacity. They argue that adolescents are the ideal target for coercive recruitment because they were not too old to be disloyal (they find that loyalty falls with age) or too young to be competent with weapons (their evidence indicates that the likelihood of receiving a firearm rose with age). However, recruitment of child soldiers can also be placed in a broader class of responses that leaders make to the threat of coup and the forced sharing of rents: the greater this threat, the more likely they are to sacrifice military readiness in favor of security from internal threats. In the case of coerced recruitment, leaders make tradeoffs between military effectiveness and the ease of controlling collective action (rebellion) by abducted troops. Children, though less effective fighters, are also less likely to be able to act collectively against these leaders. State capacity State capacity is a frequent explanation of conflict: where capacity is weak, insurgents are more likely to challenge the government. Fearon (2008) observes that external assistance to counter-insurgencies has a weak effect in poor countries, where state capacity to use the funds is weak, implying that state capacity is the ability of the state to build up organizations capable of conducting counter-insurgencies. Fearon and Laitin (2003) also argue that internal conflicts since 1945 have occurred mostly in poor countries because these governments have weaker counter-insurgency capacity. Empirically, Fearon and Laitin (2003) interpret income as a proxy for state capacity. 12 Humphreys and Weinstein (2006) make a related point in passing. They suggest that leaders resort to fragmented fighting forces to minimize the risks of organized defection when rebellion is driven by pecuniary objectives and not by a shared sense of grievance. 17 Besley and Persson (2009) endogenize the acquisition of legal capacity (the probability that a group‟s rights to credit repayment are enforced) and fiscal capacity (the ability to limit tax evasion), assuming that it requires large investments by governments. They argue that higher wealth raises the payoff to state capacity (e.g., because the value of assets protected is higher); the greater the demand for public goods (such as national security); the greater is political stability (which ensures that governments that make investments in capacity will reap the benefits); and the more representative is the political system of all interests in society. In Besley and Persson (2010), they extend these insights to conflict, arguing that state capacity and conflict are both determined by other factors, such as natural resource rents. The arguments regarding collective action contribute to the analysis of state capacity in several ways. First, they emphasize that state capacity is a function of both organization and financing, and that the two need not go together. Though under-funding can hurt organization, abundant funding is no guarantee of it. Second, like Besley and Persson (2009), the arguments here emphasize that state capacity is a decision of leaders that is driven by circumstances that also influence conflict risk. However, these circumstances are not those related to the fiscal costs of state capacity, as in Besley and Persson, but to leaders‟ fear that well-organized officials in the public administration or military can act collectively against them. That is, the purely organizational decisions that leaders make have a large effect on state capacity, independent of their cost. Finally, the discussion in this paper explains why low state capacity and low income go hand-in-hand with the absence of restraints on opportunistic behavior by governments in conflict countries. Leaders concerned about coup threats by members of their group are more likely to limit collective action by citizens, the military, and also by the public administration. The second and third decisions have a direct negative impact on state capacity. The first decision prevents citizens from acting collectively to repel predatory behavior by the leader. All of the decisions undermine incentives to invest. The data reported earlier, that indicators of corruption and the rule of law are significantly lower in conflict countries the year before conflict compared to all other countries is consistent with this. Ethnicity Ethnicity is a frequent theme in the conflict literature. In their careful analysis, for example, Montalvo and Reynal-Querol (2005) argue that ethnic polarization (rather than fragmentation) promotes conflict. The nature of collective action in a society is likely to influence the extent of this association. Fearon and Laitin (1996) provide one reason for this. Inter-ethnic violence is less likely when agreements between ethnic groups are more credible. However, credibility depends on the ability of each ethnic group to police violations of the agreements by its members. When ethnic groups are better organized to punish deviations, inter-ethnic violence is less likely. Intra-group enforcement arrangements, like intra-military arrangements to maintain discipline in the armed forces, are an essential component of intra-group collective action. The analysis in Keefer and Vlaicu (2008) points to another potential link between collective action and ethnic conflict. They conclude that precisely in environments where political competitors cannot make broadly credible commitments to voters, they will rely on appeals to narrow groups, such as clientelist promises. To the degree that patron-client 18 networks are more ethnically homogeneous than the population at large, however, this implies that politics could look “ethnic�, even though the strategy of political mobilization is clientelist. Reliance on such narrow promises is more likely when institutionalized political parties, in which like-minded citizens can act collectively to nominate and support political challengers, do not exist. Conflict is more likely in these settings. In the Democratic Republic of Congo, for example, voting in the 2006 elections clearly followed ethnic lines. Lingala speakers voted for Jean-Pierre Bemba (e.g., Bemba received 97 percent of the second round votes from his home province of Equateur) and Swahili speakers voted for Joseph Kabila (e.g., Kabila received 98 percent of the second round votes in his home province of South Kivu) (Weiss 2008). In principle, ethnic voting could simply be a reflection of citizens‟ confidence that co-ethnic candidates will support the interests of their ethnic group. However, ethnic groups are not politically organized: parties are fragmented (even those appealing to a single ethnic group) and organized around personalities rather than the pursuit of ethnic interests. This is more generally true across Sub-Saharan Africa, where conflict and ethnicity are most often linked. Keefer (2010) uses Afrobarometer survey data from 16 Sub-Saharan African countries and compares the partisan behavior of two kinds of respondents: one belongs to ethnic groups that cluster their political support on only some parties; the other kind is no more likely to support any particular party than the population at large. Respondents from the first group should be less likely than those in the second group to express partisan indifference if clustered partisan support exhibited by the first group is the product of credible appeals by ethnic parties to serve the interests of the entire ethnic group (since support for the party benefits them more than support for no party). In most countries, this is not the case: respondents from ethnic groups that do not cluster their support are no more likely to express partisan indifference than those from ethnic groups that do cluster. Service delivery The provision of services to citizens is frequently viewed, particularly by practitioners, as a way to win hearts and minds. For example, service delivery purchases cooperation from locals, making it easier to drive out insurgents. In their study of the conflict in Iraq, Berman, Shapiro and Felter (2009a) offer the first systematic evidence of the proposition that service delivery reduces violence. The discussion here suggests that the efficacy of that exchange depends significantly on collective action. First, the price, in terms of service delivery, that leaders pay for civilian assistance in an armed conflict depends on civilian expectations that the armed forces with which they are dealing will be able to block reprisals by the other side. An armed group that is poorly organized and ineffective, and therefore less likely to succeed, would therefore have to pay a higher price to elicit cooperation. A given amount of service delivery would have a smaller effect on violence. Berman et al. (2009a) focus on assistance given by US Army commanders directly to communities in Iraq. The less-organized Iraqi army, offering similar levels of service delivery, would have likely elicited less cooperation. Second, better-organized militaries, which are more effective at fighting, may also be better organized to provide service delivery. In this case, even if service delivery itself contributes nothing to violence, it is a marker for military effectiveness, which would explain lower violence. Evidence from Iraq supports the notion that the organization of the “service provider� varies substantially. Berman, et al. (2009a) observe that other programs in 19 Iraq (those administered by the Iraqi government or large aid programs contracted out to local subcontractors) were beset by significant corruption. Finally, third, the provision of service delivery could be the product of pressure by well-organized citizens (perhaps along the lines of the Colombian villages in Kaplan 2010). Citizens who are better able to act collectively are better able to extract service delivery, but also to resist violent incursions by armed groups. Again, even if service delivery itself did not influence violent outcomes, it would be a marker for the ability of citizens to act collectively to repel violent groups. These observations suggest that service delivery itself may have a limited impact on conflict risk. Instead, underlying factors that drive both conflict risk and service delivery, particularly the ability of citizens to act collectively vis à vis the leader, may matter more. Service delivery strategies could be designed to address obstacles to collective action, however, by including features that favor the collective organization of beneficiaries. How this might best be done, however, is an area that also demands further research. Conclusion Most analyses of conflict assume that leaders can make credible commitments to members of their groups and to their armed forces. Critical to credibility is the degree to which groups can act collectively, allowing group members to more easily sanction leaders who renege on their commitments. However, a wide range of evidence indicates that the assumption that groups in conflict settings are unitary and have overcome barriers to collective action is frequently violated, particularly in countries that experience conflict. Leaders impede collective action by their militaries by prohibiting coordination among units and the circulation of information, and by establishing personally loyal presidential guards that are independent of, and better funded than, the regular army. An importance vehicle through which citizens organize for collective action to influence political decision making is the institutionalized political party; conflicts are systematically more likely in countries where institutionalized parties are missing. Examining the implications of breakdowns in intra-group collective action is a significant area of future conflict research. The discussion offers conjectures about a number of important conclusions that such research might reach. Income shocks and rents should matter most when citizens and armed forces cannot act collectively; forced recruitment of child soldiers and predation on civilians should be more likely; and state capacity should be lower. Each of these conclusions should be subjected to further investigation. The discussion indicates ways in which policy recommendations based on earlier analyses might need to be adjusted. For example, income support to countries that have experienced a substantial negative income shock may have a limited effect on conflict, since it is precisely in countries where collective action is weak that income support is least likely to be effectively targeted. 20 References Acemoglu, Daron and James A. Robinson (2006). Economic origins of dictatorship and democracy. New York: Cambridge University Press. Azam, Jean-Paul (1995). “How to Pay for the Peace? A Theoretical Framework with References to African Countries.� Public Choice 83:1-2, 173-84 (April). Bennett, Richard M. (2001). “The Syrian Military: A Primer.� Middle East Intelligence Bulletin. August/September. www.meib.org. Berman, Eli, Jacob N. Shapiro and Joseph H. Felter (2009a). “Can Hearts and Minds be Bought? The Economics of Counterinsurgency in Iraq.� Mimeo, Department of Economics, UC San Diego. ____________ (2009b) “Do Working Men Rebel? Insurgency and Unemployment in Iraq and the Philippines.� National Bureau of Economic Research Working Paper 15547 (November). Besley, Timothy and Torsten Persson (2009a). “The Logic of Political Violence.� London School of Economics and IIES, Stockholm University, Mimeo (November). _____ (2009b). “The Origins of State Capacity: Property Rights, Taxation, and Politics.� American Economic Review 99:4, 1218-44. _____ (2010). “State Capacity, Conflict and Development.� Econometrica 78:1, 1-34. Boix, Carles (2003). Democracy and Redistribution. New York: Cambridge University Press. Callahan, Mary P. (1999). “Civil-Military Relations in Indonesia: Reformasi and Beyond.� Occasional Paper #4: The Center for Civil-Military Relations, Naval Postgraduate School, Monterey, California (September). Chassang, Sylvain and Gerard Padró i Miquel (2009). “Economic Shocks and Civil War.� Mimeo, Princeton University. Dal Bó, Ernesto and Pedro Dal Bó (forthcoming). “Workers, Warriors, and Criminals: Social Conflict in General Equilibrium.� Journal of the European Economics Association. De Atkine, Norvell B. (1999) “Why Arabs Lose Wars�. Middle East Quarterly (December). Dube, Oeindrila and Juan F. Vargas (2007). “Commodity Price Shocks and Civil Conflict: Evidence from Colombia.� Mimeo, Harvard University and UCLA. Fearon, James D (1995). Rationalist Explanations for War. International Organization 49 (Summer): 379-414. _____ (2008). “Economic Development, Insurgency, and Civil War.� In Elhanan Helpman, ed., Institutions and Economic Performance (Cambridge, MA: Harvard University Press, 21 2008). _____ and David D. Laitin (1996). “Explaining Interethnic Cooperation.� American Political Science Review 90:4, 715-735 (December). _____ (2003). “Ethnicity, Insurgency, and Civil War.� American Political Science Review 97:75- 90. Ferejohn, John (1986). “Incumbent performance and electoral control.� Public Choice 50: 5- 26 Garfinkel, Michelle (2004). “On the Stability of Group Formation: Managing the Conflict Within.� Conflict Management and Peace Science 21:1, 43-68 _____ and Stergios Skaperdas (2007). “Economics of Conflict: An Overview.� Chapter 3 in Todd Sandler and Keith Hartley, eds., Handbook of Defense Economics, vol. 2, forthcoming. Gates, Scott (2002). “Recruitment and Allegiance: The Microfoundations of Rebellion.� Journal of Conflict Resolution 46:1, 111-30. Gehlbach, Scott and Philip Keefer (2009). “Investment Without Democracy: Ruling-Party Institutionalization and Credible Commitment in Autocracies.� Working Paper, University of Wisconsin Department of Political Science. _____ (2010). “Private Investment and the Institutionalization of Collective Action in Autocracies: Ruling Parties, Militaries, and Bureaucracies.� Mimeo, The World Bank Development Research Group. Hashim, Ahmed (2003). “Saddam Husayn and civil-military relations in Iraq: The quest for legitimacy and power.� Middle East Journal 57:1 (Winter) 9-41. Humphreys, Macartan and Jeremy Weinstein (2006). “Handling and Manhandling Civilians in Civil War.� American Political Science Review 100:3, 429-447 (August). International Crisis Group (2006). “Escaping the Conflict Trap: Promoting Good Governance in the Congo.� Africa Report No. 14 (July 20). Kalyvas, Stathis (2006). The Logic of Violence in Civil War. Cambridge: Cambridge University Press. Kaplan, Oliver (2010). Civilian Autonomy in Civil War. Ph.D. dissertation, Department of Political Science, Stanford University. Keefer, Philip (2007). “Clientelism, credibility and the policy choices of young democracies.� American Journal of Political Science 51:4, 804-821 (October). _____ (2007b). “Programmatic parties: Where do they come from and do they matter?� Mimeo, The World Bank Development Research Group. 22 _____ (2008). “Insurgency and credible commitment in autocracies and democracies.� World Bank Economic Review. 22:1, 33-61 _____ (2010). “The ethnicity distraction? Political credibility and partisan preferences in Africa.� World Bank Policy Research Working Paper 5236 (March). _____(2010b). “Interpreting Polity: Collective action and political credibility as defining characteristics of Polity measures of political regimes.� Mimeo, The World Bank Development Research Group. _____ and Razvan Vlaicu (2008). “Democracy, Credibility and Clientelism.� Journal of Law, Economics and Organization 24:2, 371-406 (October). Miguel, Edward, Shanker Satyanath and Ernest Sergenti (2004). “Economic Shocks and Civil Conflict: An Instrumental Variables Approach.� Journal of Political Economy 112:4, 725 – 753. Montalvo, José G. and Marta Reynal-Querol (2005). “Ethnic Polarization, Potential Conflict, and Civil Wars.� American Economic Review 95:3, 796-816. Nagl, John A. (2002). Learning to Eat Soup with a Knife: Counterinsurgency Lessons from Malaya and Vietnam. Chicago: University of Chicago Press. Padró i Miquel, Gerard and Pierre Yared (2009). “The Political Economy of Indirect Control.� Mimeo, London School of Economics. Persson, Thorsten and Guido Tabellini (2000). Political Economics: Explaining Public Policy. Cambridge, MA: The MIT Press Sambanis, Nicholas (2004). “What is Civil War? Conceptual and Empirical Complexities.� Journal of Conflict Resolution 48:6, 814-858. Schultz, Kenneth A. and Barry R. Weingast (2003). “The Democratic Advantage: Institutional Foundations of Financial Power in International Competition.� International Organization 57:1, 3-42. Snyder Jr., James M., and Michael M. Ting (2002). “An Informational Rationale for Political Parties.� American Journal of Political Science 46 (1):90-110. Weinstein, Jeremy M. (2005). “ Resources and the Information Problem in Rebel Recruitment.� The Journal of Conflict Resolution, 49:4, 598-624 (August).