How leaders’ experiences and rebellion shape military recruitment during civil war

Abstract

The use of forced recruitment strategies during war can adversely affect military effectiveness and human rights. Given these costs, under what conditions do state leaders adopt coercive recruitment during civil wars? We find that between 1980 and 2009, states changed their recruitment practices 140 times during civil wars – half of which were towards coercive recruitment. Since structuralist explanations focus on factors that remain more or less constant over time, they cannot explain the frequency of these changes. Instead, we focus on individual-level factors and argue that leaders’ dispositions as risk-takers determine their beliefs about using force to solve collective action dilemmas during civil wars. Further, conflict context matters for leaders’ recruitment decisions – when rebel groups engage in coercive recruitment, leaders may also feel more justified in using such strategies. Using the LEAD Dataset and data on recruitment, we find that risk-tolerant leaders, including those who have had careers in the security sector, as well as those who have prior experience as a rebel or revolutionary leader, are more likely to use force to increase recruitment. While we theorize that this effect may be mitigated by combat experience, the evidence is mixed. Lastly, we find that rebels’ use of forced recruitment makes state leaders less likely to use voluntary recruitment.

Keywords

conscription forced recruitment leaders military rebels

Introduction

Government leaders’ implementation of coercive recruitment practices can potentially create principal-agent problems that negatively affect military effectiveness and conflict outcomes (Shils & Morris, 1948). They also increase the likelihood that state militaries will engage in human rights violations, particularly sexual and gender-based violence (Cohen, 2016a). Furthermore, leaders who implement such policies could face domestic and international backlash. Despite these costs, we find that between 1980 and 2009, 50% of the times states changed their recruitment strategies during civil wars, they adopted more coercive strategies. Given the high costs of implementing coercive recruitment strategies, under what conditions do state leaders adopt forced recruitment or press-ganging?

We focus primarily on leaders’ changes to coercive recruitment as opposed to conscription during civil wars. Coercive recruitment, also called forced recruitment, press-ganging, or impressment, is the use of extra-legal measures to recruit members into the military by force. It is distinct from conscription which relies on the threat of punishment backed by legal means. The narrower focus provides us several benefits. Since forced recruitment involves the extra-legal use of force, our study sheds light on the conditions under which states violate particular forms of human rights. While forced recruitment might seem like an outdated practice, Cohen (2013) shows 31% of modern civil wars involve press-ganging by the state.

Previous work analyzing changes in recruitment strategies looks primarily at the influence of inter-state war on these decisions (Asal, Conrad & Toronto, 2017; Margulies, 2018). However, we know little about states’ incentives switching to and away from coercive recruitment. Unlike in inter-state wars, the opportunities for defection may be greater during civil wars. This can establish a heightened competition for recruits, which may create specific dynamics that induce state leaders to switch to coercive recruitment strategies more frequently.

One approach to answer questions about leaders’ decision-making is to focus on leaders’ characteristics and prior experiences (Horowitz & Stam, 2014; Horowitz et al., 2018). However, we believe the context of a civil war also shapes leaders’ decisions. Thus, we follow the approach of Kertzer & Tingley (2018) who argue that two factors influence leaders’ decisions: disposition or leader’s characteristics, and situationality or context. We posit that both leader’s characteristics and the civil-war context can impact a state’s decision to adopt coercive recruitment practices. First, we argue that a leader’s prior experiences function as heuristics which affects risk calculi and subsequent strategies during war. Risk-taking leaders are more likely to adopt coercive measures. We identify this in several ways. A career in security, whether through the police or military, influences leaders’ likelihood of adopting coercive recruitment. That experience is tempered by whether the leader experienced combat – if the leader has not seen combat, they are more risk-accepting than leaders who have. As a result, they may be more optimistic about force as an effective solution to dilemmas during civil wars and possibly more likely to use forced recruitment. We also use other measures to test our risk-acceptance mechanisms, including an analysis of revolutionary leaders, who have been shown to have higher risk-aversion than non-revolutionary leaders (Colgan, 2012, 2013) and therefore, may be more likely to use coercive recruitment too.

However, even risk-averse leaders might be willing to take more risks depending on the civil conflict’s dynamics. We argue that as legitimate international actors, leaders of states might be reluctant to use forced recruitment, as such strategies violate international humanitarian law and human rights norms. Yet, leaders may be more willing to do so when rebels use such strategies. Government leaders may feel justified in implementing rights-violating policies if rebels have already done so. In such cases, states can use rhetoric regarding state security and sovereignty to justify coercive recruitment, and trump concerns about respecting human rights (Jetschke, 2011). Thus, we argue rebels’ use of forced recruitment in ongoing civil wars also makes leaders less likely to recruit using voluntary means.

Using the Leader Experience, Attribute, and Decision (LEAD) data set (Horowitz, Stam & Ellis, 2015), as well as data on recruitment from 1980 to 2009 (Cohen, 2016a), we show that if a leader previously had a career in the security sector, or came to power through a revolution, they were more likely to switch to coercive recruitment strategies during a civil war. Combat experience can mitigate the effects of a leader’s security sector experience, making them less likely to switch to coercive recruitment compared to civilians, though the evidence is mixed. We also find strong evidence that leaders are influenced by the practices of rebel groups they are fighting. When rebel groups engage in forced recruitment, the government is less likely to use voluntary recruitment. More specifically, rebels are often first movers in switching to forced recruitment, and government leaders follow suit. The data show that rebels first switch to forced recruitment 67% of the time, 25% of the time the government acts first, and 8% of the time they act simultaneously.

This article makes three main contributions. First, by focusing on leaders’ characteristics and context, we follow Kertzer & Tingley’s (2018: 10) call for exploring the ‘interaction between situational stakes and dispositional traits’. At the individual level, we fall in line with studies that assess leaders’ past military experiences (Horowitz et al., 2018), and emphasize leaders’ prior experiences as playing a central role in shaping their behavior once they enter office (Jervis, 1976; Byman & Pollack, 2001; Saunders, 2011; Colgan, 2013; Horowitz & Stam, 2014). While previous literature on recruitment strategy takes leadership agency out of the equation, we assume that leaders have that agency. Our aim is not to disregard structural and institutional explanations that influence a state’s recruitment strategies. However, we believe that leaders are ultimately the ones who make decisions about recruitment, especially during civil unrest, and can explain the frequency of changes towards and away from coercive recruitment strategies.

Second, the existing literature looking at changes in a state’s recruitment strategies primarily analyzes changes to and from voluntary recruitment and conscription, finding that states rarely change their strategies (Asal, Conrad & Toronto, 2017; Margulies, 2018). As Peled (1998: 170) points out, ‘the military is a conservative and tradition-bound institution’, and leaders can expect resistance from within the army to manpower innovations. Based on a more nuanced and fine-grained approach to recruitment, we find, however, that states changed their recruitment strategies 140 times during civil war years between 1980 and 2009, and of these, 50% were towards more coercive strategies. Structural-level explanations focus on factors that remain more or less constant over time, and cannot explain the frequency of these changes. On the other hand, both the leaders in power, as well as the rebel groups that they are fighting, change more frequently. This can help us better assess contextual changes during civil wars, mainly changes in recruitment strategy by rebels.

Finally, much of the civil war recruitment literature has largely focused on the determinants of rebel recruitment without much consideration of how the state recruits (Wood, 2003; Weinstein, 2006). Yet, these studies rarely theorize the role of leaders, instead assessing group characteristics. Our study provides insight about risk-taking leaders that could help understand why rebels forcibly recruit. While all rebel leaders may be prone to risk-taking (Colgan, 2012, 2013), we suggest that there is likely variation among how rebel leaders recruit – for example, Boko Haram’s founder, Mohammed Yusuf, and his successor Abubakar Shekau differed in their strategies. Like studies of rebel recruitment, we acknowledge that our study is one-sided as we only assess the government leaders’ motivations. However, the pre-existing literature suggests that rebels’ motivations are often exogenous to government decisions, and instead focus on addressing collective action problems (Wood, 2003; Weinstein, 2006). In short, rebels are often first movers in diversifying recruitment strategies, and we address how/why the government responds.

In the next section, we first highlight costs associated with coercive recruitment strategies. We theorize how leaders’ past experiences in the security sector and combat can impact their decisions while in office. We also look at the effects of situational changes during civil wars, namely the adoption of forced recruitment by rebel groups. We conclude by discussing the implications of our findings for broader conflict processes.

The costs associated with coercive recruitment strategies

We break down government recruitment strategies into three categories: voluntary recruitment, conscription, and forced recruitment/press-ganging/impressment (what we term coercive recruitment). Voluntary recruitment is where individuals enter the military as a matter of choice (Toronto, 2005). Conscription is a practice where citizens (usually male) are legally required to serve. Forced recruitment is the act of taking citizens into military service by compulsion, with or without notice, using extra-legal means. While conscription implicitly or explicitly relies on threats of punishment with legal backing, forced recruitment involves the direct use of violence and is extra-legal. This distinction created by the legal/extra-legal nature of recruitment means that the consequences of conscription are likely to vary from those of forced recruitment. Because coercive recruitment is a clear violation of human rights, whereas conscription is not, we focus on the former.

While the literature has addressed how coercive recruitment practices affect military effectiveness during inter-state wars (Gates, 2002), it may also negatively impact civil war outcomes. The literature on rebel recruitment demonstrates that lower-quality recruits are more likely to engage in indiscriminate killing (Weinstein, 2006). Forced recruits rarely participate in professionalization in academies, resulting in low social cohesion and weak social ties (Cohen, 2013). Cohen (2013, 2016a) shows that soldiers who are forcibly recruited are more likely to perpetrate sexual violence, as wartime rape is often used as a method of socialization to create unit cohesion.

Forced recruitment also has other negative consequences outside the theatre of war. The state may incur large expenses to pay for institutional changes to accommodate forced recruits who need additional training and professionalization. New military academies might have to be built to replace the quantity loss with quality. States may suffer domestic political costs as coercive recruitment would be unpopular and citizens could protest these violations of human rights.

Forced recruitment also violates international humanitarian law, which outlaws a wide range of violence against civilians – including the use of forced methods to gain recruits.¹ The international community can offer incentives to states to outlaw such practices. For instance, during Joseph Kabila’s tenure as the President of the Democratic Republic of the Congo (DRC), he used forced recruitment in order to deal with the pressures of multiple rebel groups’ own use of child soldiers.² However, in 2002, Kabila ratified the Optional Protocol to the Convention on the Rights of the Child, which bans the use of child soldiers, and the Rome Statute of the International Criminal Court in 2002, which includes the war crime of child recruitment (Human Rights Watch, 2007). Consequently, Kabila received aid in excess of $240 million from the World Bank and the International Monetary Fund (IMF) for fulfilling human rights guarantees, one of which was to end recruitment of children under the age of 15.³

Crucially, compared to rebel leaders, only states are party to international human rights treaties and accountable to them. States that violate their citizens’ human rights have more avenues by which they may be punished by the domestic and the international community. Research suggests that states suffer numerous political and economic consequences after getting shamed for human rights violations, which can result in withheld foreign aid, reduced trade, and exclusion from international organizations (Barry & Flynn, 2013; Murdie & Peksen, 2013). For instance, in 1999, Uganda started forcibly recruiting due to internal insurgencies. However, the IMF pressured Uganda to cut military expenditures. When Uganda exceeded this limit, the IMF suspended aid and only resumed after the government ended forced recruitment in 2000 (McCandless & Karbo, 2011). Given the high costs associated with employing coercive recruitment strategies, under what conditions would any leader engage in such practices? We explore both dispositional and contextual explanations next.

Recruitment dilemmas during civil wars

The existing literature on state recruitment strategies focuses overwhelmingly on structural-level factors (Posen, 1993; Asal, Conrad & Toronto, 2017; Margulies, 2018), minimizing individuals’ agency. Structural explanations focus on factors that tend to remain constant over time – while they explain the conditions under which a certain recruitment strategy is likely, they do not necessarily help us understand changes. An agent-focused or leader-centric approach can account for time-variance by acknowledging that switches occur because leaders have agency and can change policies. We are not asserting that individuals matter more than structure in civil wars, rather, that under some circumstances, leader characteristics and experiences also matter.

Initial research on leaders looked at how domestic institutional arrangements shaped or constrained leader choices, instead of focusing on traits or experiences of leaders (Croco, 2011; Weeks, 2012). However, a new wave of recent research has focused on leaders as a unit of observation (Horowitz & Fuhrmann, 2018; Horowitz, Stam & Ellis, 2015; Saunders, 2011; Yarhi-Milo, 2014). This literature recognized that individual leaders have agency and they often make decisions that sometimes go against rational decision-making (Levy, 1997; McDermott, 2001). It also finds leaders’ past experiences determine how risk-accepting they are in developing strategies and tactics during war (Horowitz & Stam, 2014). In line with this work, we argue leaders are critical in explaining recruitment strategies, particularly during civil wars.

One of the main dilemmas leaders face during conflicts is how to resolve their collective action problem (Horowitz & Fuhrmann, 2018) – how do they recruit individuals for the war effort? The collective action problem during a civil conflict differs from related problems that leaders face during inter-state wars. In particular, we argue that relative to inter-state wars, recruitment in civil wars is more of a zero-sum game. At first glance, it may not seem that all sides to a civil conflict recruit evenly across the population. In highly polarized ethnic wars, different sides may lack the willingness or the capacity to recruit geographically or ethnically closer to their opponents. This could impact how recruitment happens in two ways – and both could lead to forced recruitment.

In the first scenario, different sides have specific preferences for recruitment, narrowing the pool even more for any one given side, potentially leading to high-risk behavior by leaders precisely because the pool is smaller. Government (or rebel) leaders may not want to recruit just anyone into their forces. They may be partial to recruiting members of their ethnic group or known supporters. This is because of a ‘Trojan horse dilemma’, which occurs when one side fears that members of the enemy will infiltrate their own side (Peled, 1998). As a result, the pool of recruits during a civil war may be smaller for any one given side than during an inter-state war. With a smaller recruitment pool, government leaders may resort to riskier behavior, including forced recruitment.

In the second scenario, leaders may value courting populations from different ethnic groups because, as Peled (1998) points out, recruitment on purely ethnic grounds can jeopardize the state’s long-term security. On the other hand, recruitment from different ethnic groups is key for preventing conflicts amongst those groups. He argues these soldiers can serve loyally, especially in professional military organizations, where officers, not politicians, are the decision-makers. Indeed, unlike in rebel organizations, professionalization of the security forces is a requirement that could erase or minimize ethnic divisions within the security forces (Samii, 2013). This could reduce these ‘Trojan horse’ concerns about recruits from other ethnic groups. Thus, by seeing all individuals in the country as potential recruits for one side or the other, it could lead to a zero-sum competition.

Even in highly polarized civil wars, competition for recruits may be more zero-sum than inter-state wars because the pool of eligible recruits remains within the same country. Butler & Gates (2009) show that rebel groups purposely provoke the government into attacking civilian populations to boost their own recruitment. This indiscriminate killing of civilians can alter allegiances and increase rebel recruitment (Kalyvas, 2006). Even if a rebel group enjoys low civilian support, civilians may join them if the dangers of remaining a civilian exceed the dangers of fighting (Kalyvas & Kocher, 2007). These dynamics are absent in inter-state wars where citizens do not have to make decisions about whose army to join. Therefore, government forces may be unable to recruit enough troops using just voluntary measures. For instance, in Sudan, Omar Al-Bashir switched to forced recruitment in 1992 when fighting the Sudan People’s Liberation Army (SPLA) in order to increase troop numbers to get an edge over the SPLA (Human Rights Watch, 1995).

Civil wars are also distinct because defection is likely to be higher. There are fewer barriers to entry and exit – one can defect more easily because he or she is within the borders of his or her country and there is greater familiarity with the opposing side’s language and culture. Thus, leaders are not only competing for recruits from a broader pool, but are competing with rebel groups to keep their soldiers. This dynamic is unique to civil wars and demonstrates why leaders may recruit aggressively.

Because recruitment and retention of soldiers is more difficult during civil wars, leaders may make different risk calculi about recruitment. This may particularly be the case because the principle of sovereignty has historically dictated that international actors stay out of the domestic affairs of states and there are fewer checks on leaders’ decisions during a civil war. Leaders may engage in a wider range of strategies, including ones that are harmful to civilians.

We argue that the propensity for leaders to take risks plays an important role in leaders’ decisions. Next, we explain how leaders’ prior experiences and situational factors such as rebel recruitment strategies impact leaders’ likelihood of the adoption of coercive recruitment.

Disposition: Risk-taking leaders

We contend that a leader’s background indicates how risk-taking they may be. In particular, we look at their past history with the use of force, as prior history with violence sometimes makes individuals more prone to use violence in the future (Littman & Paluck, 2015). Government leaders’ experiences with violence usually come from participating in the armed forces or past experience as a revolutionary leader.

We posit that prior security-sector experience (a primary career either in the police or the military) is likely to influence the risks leaders take with respect to recruitment strategy during civil wars. This excludes leaders who have experienced short-term military service, as that can have different socialization effects on behavior, compared to a career in an institution like the police or the military. Previous research has also shown that military socialization has a greater effect the longer one serves and the closer one comes to combat (Krebs, 2004: 91). Short-term military service does not necessarily provide us much information about exposure to risk and potential psychological trauma experienced during combat, whereas actual wartime experience exposes soldiers to, ‘new ideas, technologies, political tactics and forms of social and economic organization’ (Krebs, 2004: 94).

Leaders who have past security-sector experience are more prone to taking risks because they have self-selected into a dangerous career, demonstrating their willingness to take risks. Even though a career in the security sector would likely professionalize leaders, previous literature shows those who embark on careers in the security sector are likely to be risk-takers in the first place (Sookermany, Sand & Breivik, 2019). Thus, prior experience in such careers might just be a proxy for leaders who are generally more risk-acceptant. Moreover, those with military careers are more optimistic about the use of force and coercive threats. Sechser (2004) argues that ties to the military can create parochial interests in favor of using force. Those in the security sector may also exaggerate the ability of force to solve threats and be an effective solution to political problems (Snyder, 1984; Horowitz & Stam, 2014). Military experience thus makes the use of force more viable to leaders.

The risky choice for leaders is whether to use violence in the recruitment process or not, as this would be considered extra-legal in many cases and a violation of human rights. Risk-taking leaders are more likely to keep violence as an option for recruitment because they might be less concerned about the costs associated with coercive recruitment. We do not suggest that risk-taking leaders would automatically choose coercive recruitment, but rather that it remains an option in the set of recruitment strategies – the leader’s risk-tolerance opens the opportunity for the use of coercive recruitment. In sum, the combined pressure of civil war constraints and risk-taking leaders makes the use of coercive recruitment more likely.

While leaders who join the security forces represent risk-takers, so do rebel or revolutionary leaders (Colgan, 2013). If individuals risk their lives for a cause and defend that cause using violence, it demonstrates that they are risk-taking – this propensity may carry over if they enter politics. Thus, another robustness test to our ‘risk’ theory would be to look at revolutionary leaders, those leaders who come to power through a revolution (Colgan, 2012: 453).

Even though a career in the security forces is likely to enhance risk-taking, the influence of having a career in the security sector, we argue, is tempered by whether a leader experienced combat or not. Existing literature shows that those with combat experience are likely to be more cautious about the usefulness of force (Huntington, 1957; Janowitz, 1960; Brunk & Tamashiro, 1990). Those that have had careers in the security sector, but who have not engaged in combat are less familiar with the costs of violence. Cohen (2016b) finds that leaders who have experience in the military, but not in combat, develop beliefs that make them more likely to authorize the use of force than leaders with combat experience. Similarly, they are more likely to initiate militarized disputes (Horowitz & Stam, 2014) and are more likely to issue coercive threats (Horowitz et al., 2018).

In contrast, leaders who have had careers in the security sector and have experienced combat are more attuned to the costs of violence and may be more prone to restraint. Horowitz et al. (2018) show that combat experience tempers leaders’ optimism about the use of force, specifically in terms of leaders’ beliefs about the efficacy of issuing coercive threats to other states. Brunk, Secrest & Tamashiro (1990) find that while combat veterans are more risk-accepting, they are also more restrictive about the situations in which they think the use of force is appropriate. Such individuals may consequently be more wary of using violence to recruit, compared to civilian leaders and leaders with careers in the security forces.

H1a: If leaders have previously had careers in the police or military and have not seen combat, then they are more likely to adopt coercive recruitment strategies.

H1b: If leaders have previously had careers in the police or military and have seen combat, then they are less likely to adopt coercive recruitment strategies.

The foregoing comparison group is civilian leaders – leaders who have careers in the police or military but who have not seen combat are most likely to forcibly recruit compared to civilian leaders. H1b then assesses whether combat experience conditions this relationship. Thus, we offer hypotheses H1a and H1b because H1a specifically tests the risk-taking hypothesis, whereas H1b tests whether risk is tempered by experiences with combat.

Situationality: Responses to rebel recruitment during civil war

A leader’s decision to implement coercive recruitment may also change when the local conflict context shifts. Being at the center of a collective action problem, leaders need to recruit sufficient labor and identify free-riders in order to succeed. Given the comparative difficulty of recruitment during civil wars, leaders may be more likely to change recruitment strategies based on how their opponent is recruiting. Thus, the type of recruitment that rebels use may affect the type of recruitment that state leaders support.

As explained earlier, implementing coercive recruitment strategies can impose domestic and international costs on leaders. International humanitarian law (IHL) outlaws the use of violence against civilians during conflict. However, states may feel more constrained by such laws, as formal mechanisms for enforcing IHL are limited in the case of rebel groups. For instance, conscripting or enlisting children under the age of 15 years into armed forces or groups, or using them to participate in hostilities is outlawed as a war crime.⁴ However, these safeguards, further developed by Protocol II of the Geneva Convention in 1977, are confined to civil wars in which both sides control tracts of territory (Ssenyonjo, 2005).⁵ Thus, rebel groups who do not exercise territorial control cannot be held accountable for non-compliance with recruitment standards.

Further, unlike government armies, rebel groups have greater recruitment and operational discretion and less monitoring. Rebel groups do not face the same constraints and their strategies are less likely to be monitored by domestic and international agencies. For instance, many human rights NGOs, including Human Rights Watch, do not monitor rebel behavior as they mainly seek to hold states accountable, as it may be impossible to lobby rebel groups to change their behavior (Chaudhry, 2019). Without formal procedures for recruitment, rebel leaders have greater freedom to use a wider set of strategies for recruitment, including force. Rebel leaders thus have fewer domestic and international constraints when it comes to recruitment when compared to government leaders. This makes rebel leaders more likely to be first movers when it comes to the use of coercive recruitment.

H2: Compared to state leaders, rebel leaders are more likely to be first movers when it comes to switching toward coercive recruitment during civil wars.

However, government leaders may consider using forced recruitment if rebels use it first. This may be particularly so in civil wars that are not highly polarized or divided along ethnic cleavages and competition for recruits is more acute and from the same pool. Even if the government is aware that rebel groups do not have much civilian support, governments may be tempted to use forced recruitment to prevent their pool of potential recruits from shrinking further. State leaders do not want to be the first to commit human rights violations because they are under higher domestic and international scrutiny compared to non-state actors. However, they may feel justified in the eyes of the public and international community if rebels engage in human rights violations first. Further, coercive recruitment can help increase government troops numbers in order to respond to the increased number of rebel troops. Fielding & Shortland (2012), for example, find that during the Peruvian civil war, each side responded to the recruitment strategies and actions of the other.

To give an example of this dynamic: Joseph Kabila switched recruitment strategies twice, suggesting that the context of the civil war played a part in his decision-making. In 2002, he switched from forced to voluntary recruitment. However, his motivation can be attributed to the Lusaka Accords of 1999, which in addition to providing aid to Kabila’s government, also mandated that all parties, including the rebel groups, agreed to demobilize and stop forcible recruitment.⁶ Shortly thereafter, in 2004, Kabila switched back to forced recruitment because the rebels appeared to have never ended their forced recruitment (Human Rights Watch, 2001, 2007). Notably, he changed methods despite the conditional aid offered to his government. Thus, Kabila initially stopped forced recruitment when rebel groups were forced to demobilize, and he no longer needed additional recruits to maintain an edge over these groups. However, once he realized the rebel groups were violating terms of the peace agreement and forcibly recruiting, it may have persuaded Kabila to do so as well. Importantly, reports show that rebel groups were engaging in such practices before Kabila made the switch.

H3: If rebel groups use coercive recruitment strategies, then it should increase the likelihood of government leaders’ use of coercive recruitment strategies as well.

Research design

This section describes our empirical strategy for determining how leadership characteristics and rebel group behavior influence the likelihood that leaders will employ coercive recruitment strategies. Our unit of analysis is the leader-civil conflict-year, where we employ the data in Cohen (2013) as the basis for civil conflicts, covering 1980 to 2009, based on the Fearon & Laitin (2011) coding.⁷ We use this unit of analysis to account for three factors. First, conflicts can take place over a number of years, during which a leader might change the recruitment strategies that they employ. Second, a leader might have to contend with multiple conflicts taking place in a single year, while potentially utilizing coercive recruitment strategies in some conflicts but not others. Third, leadership can change over the course of a year for a variety of regular (i.e. elections) and irregular (i.e. coups and assassinations) reasons. We draw from a dataset of 1,192 total observations, from which we are able to utilize just a maximum of 970 due to missing data.⁸

While we are interested in leaders and recruitment strategies, we specifically want to measure the change of recruitment strategies over time. In any given conflict-year, we assume that leaders have the ability to decide between employing voluntary recruitment strategies or coercive recruitment. Because we only theorize about leaders with careers in the security forces and/or combat experience, our key variables capture those specific combinations for which we have clear expectations based on our theorizing.⁹

To model this decision, we employ a Cox proportional hazard, measuring which factors are correlated with the voluntary recruitment’s ‘end’. We also provide Kaplan-Meier survival estimates over time to visualize the results, and more directly show the impact of our variables of interest. The choice to utilize a hazard model as opposed to a logistical regression model stems from the difference between the leader’s decision to change the state’s recruitment strategy versus what strategy the leader is employing. With the former, we can demonstrate that the leader is making a conscious choice to move away from one type of recruitment to another, specifically from voluntary to forced. With the latter, we can identify what recruitment strategy a leader employs, but introduces the possibility that the leader either is chosen because of the type of conflict and/or recruitment strategy already in place, thus bringing in a potential source of endogeneity. However, if both models are in line with our predictions, that makes our argument all the stronger; as such, the Online appendix also provides the results from analysis using multinomial logistic regression.¹⁰

As pointed out by Horowitz & Stam (2014), these results could be biased due to endogeneity. Specifically, the entrance of a new leader could be influenced by a predecessor’s recruitment strategy; perhaps that the use of forced recruitment was particularly unpopular, or alternatively, that more drastic measures were needed to fight the insurgency, resulting in a change of leadership. To account for these potential factors, our models include variables such as how long it has been since a change in strategy has taken place and whether a change took place in the previous leader-civil conflict-year. In addition, by separating out conflicts by year and leader, we avoid some issues of endogeneity. We also account for endogeneity in H3 by systematically assessing the side of the first mover to ensure that the government leader is responding to the rebels.

Dependent variable

We use data from Cohen (2013), who identified when states use forced recruitment or press-ganging. The dependent variable is constructed to identify if the state either used a voluntary or coercive recruitment strategy, specifically whether press-ganging was employed. Our Cox proportional hazard model then measures if voluntary recruitment ‘survives’ by being utilized for that leader-civil conflict-year or ‘perishes’ with the use of press-ganging. In some cases, this can take place immediately, as 10% of conflicts start off with forced recruitment, or after some time has passed.

Independent variables

Our independent variables come from the LEAD dataset (Horowitz, Stam & Ellis, 2015) and data from Cohen (2016b) on rebel recruitment strategies. LEAD collects a wide swath of information related to the background and history of a given leader, everything from sex and marital status to careers that the leader had prior to holding office. Cohen’s data similarly looks at leader experiences but in a much more narrow fashion, examining what the leader contends with in a given conflict. In particular, she provides data on whether the rebels that the leader is fighting against are themselves using forced recruitment. For all of these variables, the measures are dichotomous.

The variables that we borrow from Horowitz, Stam & Ellis (2015) and Cohen (2016b) are as follows:

Combat: Did the leader have combat experience?

Security: Did the leader serve in the security sector (police or military)?

Rebel forced: Did the rebels utilize a forced recruitment strategy?¹¹

Since our argument hinges on the idea that security-sector experience translates differently depending on whether or not an individual has been in combat, we have created two variables to cover the relevant combinations: one where a leader with security experience has been in combat (Sec/Combat), and one where a leader with security experience has not been in combat (Sec/No combat). The reference category is a leader without security and/or combat experience, or a ‘civilian’. We do this because even though we have not theorized about the conditional effects of risk-accepting leaders with rebels’ use of forced recruitment, we wanted to be be able to test the interaction.¹²

We also provide a separate test involving an explicit interaction between Security and Combat. Fundamentally, this means that the contrast we are drawing is between leaders who have security experience and those who do not, since both of our key independent variables include an indicator involving a security background.

Importantly, one could consider the baseline as being a ‘civilian’ but the models capture different types of ‘civilians’. This means that Model 1’s reference category is ‘civilian’ leaders who have no security background (excluding those leaders who have combat experience but no security experience), whereas Model 2’s reference category is ‘civilian’ leaders who have no security or combat experiences, and includes the category of leaders who have combat experience but no security experience.

Control variables

We also employ a number of control variables from the LEAD dataset which could also influence the likelihood of forced recruitment: the leader’s level of education, if they had military service of any type, and whether they were themselves a rebel prior to becoming the leader of the state. In addition, we include Colgan’s variable for whether the leader qualifies as ‘revolutionary’, as such leaders have been argued to be more risk-acceptant (Colgan, 2012, 2013), which could make them more willing to utilize forced recruitment. This is a robustness test for our theory, as a positive relationship between revolutionary leader and change to coercive recruitment would indicate that there is general support for our ‘risk’ mechanism.

We also provide controls for aspects of the conflict and state. These include if the conflict involves abductions by the rebels, the goals of the rebel group (specifically ranging from targeting the central government to fighting for autonomy or secession), whether the civil war has an ethnic component, the extent to which the state is ‘failing’, the conflict aims, Polity score as a proxy for regime type, the conflict’s duration, logged population and GDP, overall military expenditures, soldier quality (Cohen, 2016a), whether there is an ongoing interstate conflict (Ghosn, Palmer & Bremer, 2004), and whether the state is a former colony (Hensel & Mitchell, 2007). Finally, we control for prior recruitment methods, specifically how long it has been since a change in recruitment strategy, the state’s original method of recruitment, and whether a change occurred in the previous leader-civil conflict-year.

Results

For H1a, H1b, and H3, our analysis utilizes a step-wise approach with the Cox proportional hazard model, starting off with our main variables of interest, Sec/No combat, Sec/Combat and Rebel forced, and then subsequently adding controls to measure the impact on the variables of interest. The basic comparison that we are making is between those who have security experience, and those who do not, with the latter representing the baseline. The first group of added control variables pertain to the leader’s background or identity and aspects of the conflict itself, and the second group account for time and previous policies. Table I provides the results, while Figures 1, 2, and 3 show the survival graphs for security without combat, security with combat, and rebel forced recruitment, respectively.

The initial model, which only includes our three variables of interest, first shows that security-sector experience, regardless of whether it is tempered by being in combat, is correlated with a higher likelihood of utilizing forced recruitment. This is in line with H1a but not H1b, likely due to the model essentially being misspecified and failing to incorporate the relevant controls. The inclusion of those control variables makes the security/combat variable switch directions, as H1b predicts. Overall, with the additional battery of controls included,

Table I.

Cox proportional hazard model, change to coercive recruitment strategy as DV

VARIABLES	(1)	(2)	(3)
Security/combat	1.584*	0.284^†	0.103*
	(0.368)	(0.213)	(0.109)
Security/no combat	4.321**	5.063*	56.86**
	(0.996)	(4.048)	(47.13)
Rebel forced	1.411^†	1.022	1.714*
	(0.278)	(0.246)	(0.466)
Combat		26.80**	215.9**
		(28.25)	(254.6)
Revolutionary leader		2.746**	4.157**
		(0.809)	(1.571)
Military service		0.161*	0.0267**
		(0.134)	(0.0228)
Education		0.843	0.853
		(0.139)	(0.167)
Rebel		1.173	0.750
		(0.349)	(0.275)
War aims		2.163**	4.263**
		(0.502)	(1.091)
Ethnic war		0.544**	0.591**
		(0.0933)	(0.117)
Mag. fail		0.907	0.898
		(0.102)	(0.128)
Polity		0.886**	0.896**
		(0.0297)	(0.0374)
Duration		0.970*	0.985
		(0.0124)	(0.0173)
Population (logged)		0.457**	0.399**
		(0.0750)	(0.0990)
GDP (logged)		0.376**	0.168**
		(0.0848)	(0.0583)
Ongoing MID		1.985**	2.180**
		(0.513)	(0.596)
Colony		0.788	0.488
		(0.355)	(0.329)
Change time			0.772**
			(0.0355)
Past change			1.117
			(0.283)
Troop quality			1.000
			(1.82e-05)
Original recruitment			2.772*
			(1.340)

Observations	934	874	768

Standard errors in parentheses.

**p < 0.01, *p < 0.05, ^†p < 0.1.

the results provide strong evidence in favor of H1a as well as modest support for H1b.

In addition, our analysis of rebel strategy shows that there is indeed a relationship between the rebels using

Figure 1.

Cox model, security/no combat

Figure 2.

Cox model, security/combat

Figure 3.

Cox model, rebel forced recruitment

forced recruitment strategies and the government choosing to match them (H3). Though the second column shows that under some circumstances the relationship only reaches weak levels of significance, in the full model we find that these results are positive, in line with H3, and statistically significant.

Turning to the Kaplan-Meier graphs, Figure 1 shows that the results are especially pronounced: those leaders who have a background in security without an accompanying experience in combat diverge relatively quickly from those who do not share such a background, resulting in an enormous gap in the survival estimates. In contrast, the results of Figure 2 track far more closely, with only a small separation at the end between those who have both security and combat experience and those who do not. Finally, Figure 3 shows that, similar to Figure 1, there is a large gap between situations where rebels do and do not use forced recruitment, with the former being far more likely to end in the government use of coercive recruitment.

We also provide models showing an interaction between security and combat experience. Shown in Table II, these results demonstrate a high level of statistical significance for all of the categories, but all in the same direction: any combination of security and combat experience is correlated with a higher likelihood of changing to forced recruitment, in comparison to the baseline of no security and no combat. While a security background without combat experience has the largest coefficient, the large standard error makes it indistinguishable from the other categories. There are very few individuals who have experienced combat and do not have a background in the security sector, and they were all part of rebellions prior to their ascendance to leadership.¹³ For this reason, we did not theorize about them. Though they engaged in combat as rebel leaders, the revolutionary leader variable is a better measure for our purposes. As such, we bracket them out of the discussion because they do not fit our theoretical model and it is unclear what they represent.

However, it is still worth examining the Kaplan-Meier estimates based on the full model with the interaction. As shown in Figure 4, while the baseline category (leaders with neither security nor combat background) and leaders who have no security background but do have combat experience are fairly similar to one another, the largest divergence comes from the security/no combat experience as that grouping is the most likely to make a change. Those leaders with both security and combat experience inhabit a middle ground between these two main paths, ultimately having a higher likelihood of

Table II.

Cox model, change to coercive recruitment strategy as DV with interaction

VARIABLES	(1)	(2)	(3)
∼ Sec x combat	1.970	26.80**	215.9**
	(1.182)	(28.25)	(254.6)
Sec x ∼ combat	4.431**	5.063*	56.86**
	(1.031)	(4.048)	(47.13)
Sec x combat	1.631*	7.607*	22.30**
	(0.383)	(6.493)	(20.55)
Rebel forced	1.435^†	1.022	1.714*
	(0.284)	(0.246)	(0.466)
Revolutionary leader		2.746**	4.157**
		(0.809)	(1.571)
Military service		0.161*	0.0267**
		(0.134)	(0.0228)
Education		0.843	0.853
		(0.139)	(0.167)
Rebel		1.173	0.750
		(0.349)	(0.275)
War aims		2.163**	4.263**
		(0.502)	(1.091)
Ethnic war		0.544**	0.591**
		(0.0933)	(0.117)
Mag. failure		0.907	0.898
		(0.102)	(0.128)
Polity		0.886**	0.896**
		(0.0297)	(0.0374)
Duration		0.970*	0.985
		(0.0124)	(0.0173)
Population (logged)		0.457**	0.399**
		(0.0750)	(0.0990)
GDP (logged)		0.376**	0.168**
		(0.0848)	(0.0583)
Ongoing MID		1.985**	2.180**
		(0.513)	(0.596)
Colony		0.788	0.488
		(0.355)	(0.329)
Change time			0.772**
			(0.0355)
Past change			1.117
			(0.283)
Troop quality			1.000
			(1.82e-05)
Original recruitment			2.772*
			(1.340)

Observations	934	874	768

Standard errors in parentheses.

**p < 0.01, *p < 0.05, ^†p < 0.1.

making a change than the baseline or leaders with only combat experience, but less likely than the leaders who have only a background in security. This implies strong support for H1a. Support for H1b is more ambiguous.

Figure 4.

Cox model, security/combat interaction

We find inconsistent evidence about whether leaders with security and combat experience are less likely to engage in forced recruitment than civilian leaders – Table I shows support, but Table II does not corroborate this finding. Thus, we can conclude that leaders with experiences in combat and security are less prone to forced recruitment than leaders with just security experience, but remain agnostic about making comparisons with ‘civilian leaders’.

Looking at the revolutionary leader variable, across all models, this measure is notable for being consistently correlated with a higher likelihood of a change towards forcible recruitment, as well as being statistically significant in all the models where it was incorporated. This indicates that apart from the other variables we include in the model, even whether the leader was previously a rebel, being a revolutionary leader is strongly correlated with forced recruitment in civil conflicts.

Finally, H2 contended that rebel organizations are more likely to initially use forced recruitment while governments then respond with their own forcible tactics. To test this, we identified all cases where one side uses coercive recruitment in a given year where they did not in the previous year, for a total of 87 observations. We also identified all instances where both sides first start using forced recruitment, where in the previous year either one side or both were relying on volunteers, combining for a total of 22 incidents. We identify these as the instances of ‘first movers’ and ‘responders’, respectively.¹⁴

Table III.

First mover and responder tabulations

	First mover	Responder
Gov’t	25%	50%
Rebel	67%	18%
Simultaneous	8%	32%

With the first movers, there are 22 cases where the government initiates using forced recruitment, 58 for the rebels, and 7 cases where both start at the same time. Comparing the three groups against one another using an ANOVA test provides a p-value < 0.00001, demonstrating a distinct difference between them. A t-test between government and rebel first movers returns similar results (p-value < 0.00001). Turning to responders, 11 cases involved the government responding to rebels’ use of forced recruitment, 4 had the rebels reacting, and 7 where they started in the same year.¹⁵ Comparing the three groups against one another using an ANOVA test provides a p-value of 0.081, while a t-test between the government responses against rebel responses alone has a p-value of 0.01. Combined, this provides strong evidence in favor of H2, as rebels not only seem to be first movers more often than governments, but governments responding to rebel action. These results are summarized in Table III.

Discussion and conclusion

The use of coercive recruitment practices by states can negatively affect their military and battlefield performance, as well as generate numerous adverse domestic and international costs. Given these costs, under what conditions do state leaders adopt forced recruitment or press-ganging during civil wars? We argue that leaders’ decisions about coercive recruitment strategies are based on two main factors: leaders’ risk-tolerance based on their prior experiences and the civil war context.

Our findings support the notion that leaders’ prior experiences serve as important heuristics which can affect their risk calculi. In particular, we find that risk-tolerant leaders, measured in multiple ways, including those who had chosen security careers or who became revolutionaries, are more likely to use forced recruitment. Combat experience might mitigate a leaders’ security experience, the evidence of such a tempering effect is mixed.

Dispositional factors may not be alone in explaining leaders’ decisions about forced recruitment; situational factors matter too. Given that recruitment during civil wars has greater potential to be a zero-sum game compared to inter-state wars, leaders may face constraints in increasing troop numbers through purely voluntary means. However, given the numerous domestic and international costs of using forced recruitment, state leaders may not want to initiate such campaigns and may feel more justified in doing so if the rebels act first. We find that government leaders are more likely to mirror rebels’ adoption of coercive recruitment. Moreover, in most cases of the state and rebel organizations using forced recruitment during a civil war, the rebel organization(s) adopt such a strategy first. Our data show that rebels change to forced recruitment before government leaders, reinforcing the conclusions from the statistical analyses. This provides another indication that state leaders are responding to contextual changes during a civil conflict. Our findings are consistent with the idea that both leaders’ dispositions and the contextual dynamics matter for leaders’ decision-making (Kertzer & Tingley, 2018).

Our article provides a new understanding of recruitment decisions by the state and makes several advances over the existing literature. A military’s participation policy exerts powerful influence on politics (Krebs, 2004: 16). Yet, existing literature on state recruitment strategies focuses almost entirely on conscription by looking at structural-level explanations, which cannot explain the frequency of changes in recruitment strategies. Our research shows that government leaders switched recruitment methods 140 times during civil wars between 1980 and 2009 (50% of which are towards more coercive strategies). By using individuals as the main unit of analysis, we break new ground by looking at the conditions under which forced recruitment switches occur during a civil war.

Moreover, we bring attention to government recruitment dilemmas, whereas much of the civil war literature focuses on rebel recruitment. Yet, our research indicates that rebels and governments face different costs when it comes to employing riskier recruitment strategies. Government leaders may face more constraints than rebel leaders and therefore only risk-accepting leaders or those with political cover by rebels’ use of forced recruitment are the ones who are more likely to use coercive practices. More broadly, our results suggests that leaders take cues from rebels in violating human rights, and when rebels engaged in forced recruitment, governments follow suit. This means that the international community might do well to focus their efforts against forced recruitment by rebel groups. If rebel groups do not use forced recruitment, then governments may not either.

Footnotes

Replication data

The dataset and the do-file for the empirical analysis can be found at .

Acknowledgements

We would like to thank Samantha Camilletti for excellent research assistance.

ORCID iD

Suparna Chaudhry

Matt K Scroggs

Notes

References

Asal

Victor

Conrad

Justin

Toronto

Nathan

(2017) I want you! The determinants of military conscription. Journal of Conflict Resolution 61(7): 1456–1481.

Barry

Colin M

Clay

Chad

Flynn

Michael E

(2013) Avoiding the spotlight: Human rights shaming and foreign direct investment. International Studies Quarterly 57(3): 532–544.

Brunk

Gregory G

Secrest

Donald

Tamashiro

Howard

(1990) Military views of morality and war: An empirical study of the attitudes of retired American officers. International Studies Quarterly 34(1): 83–109.

Butler

Christopher

Gates

Scott

(2009) Asymmetry, parity, and (civil) war: Can international theories of power help us understand civil war? International Interactions 35(3): 330–340.

Byman

Daniel L

Pollack

Kenneth M

(2001) Let us now praise great men: Bringing the statesman back in. International Security 25(4): 107–146.

Chaudhry

Suparna

(2019) Bridging the gap: The relationship between INGO activism and human rights indicators. Journal of Human Rights 18(1): 111–133.

Cohen

Dara Kay

(2013) Explaining rape during civil war: Cross-national evidence (1980–2009). American Political Science Review 107(3): 461–477.

Cohen

Dara Kay

(2016a) Rape during Civil War. Ithaca, NY: Cornell University Press.

Cohen

Michael D

(2016b) Live and learn: Availability biases and beliefs about military power. Foreign Policy Analysis 13(4): 968–985.

10.

Colgan

Jeff

(2012) Measuring revolution. Conflict Management and Peace Science 29(4): 444–467.

11.

Colgan

Jeff D

(2013) Domestic revolutionary leaders and international conflict. World Politics 65(4): 656–690.

12.

Croco

Sarah E

(2011) The decider’s dilemma: Leader culpability, war outcomes, and domestic punishment. American Political Science Review 105(3): 457–477.

13.

Fearon

James D

Laitin

David D

(2011) Sons of the soil, migrants, and civil war. World Development 39(2): 199–211.

14.

Fielding

David

Shortland

Anja

(2012) The dynamics of terror during the Peruvian civil war. Journal of Peace Research 49(6): 847–862.

15.

Gates

Scott

(2002) Recruitment and allegiance: The microfoundations of rebellion. Journal of Conflict resolution 46(1): 111–130.

16.

Ghosn

Faten

Palmer

Glenn

Bremer

Stuart A

(2004) The MID3 data set, 1993–2001: Procedures, coding rules, and description. Conflict Management and Peace Science 21(2): 133–154 (https://correlatesofwar.org/data-sets/MIDs).

17.

Hensel

Paul

Mitchell

Sarah

(2007) The issue correlates of war (ICOW) project supplementary data set: Colonial history data set. Harvard Dataverse (https://doi.org/10.7910/DVN/5EMETG).

18.

Horowitz

Michael C

Fuhrmann

Matthew

(2018) Studying leaders and military conflict: Conceptual framework and research agenda. Journal of Conflict Resolution 62(10): 2072–2086.

19.

Horowitz

Michael C

Potter

Philip

Sechser

Todd S

Stam

Allan

(2018) Sizing up the adversary: Leader attributes and coercion in international conflict. Journal of Conflict Resolution 62(10): 2180–2204.

20.

Horowitz

Michael C

Stam

Allan C

(2014) How prior military experience influences the future militarized behavior of leaders. International Organization 68(3): 527–559.

21.

Horowitz

Michael C

Stam

Allan C

Ellis

Cali M

(2015) Why Leaders Fight. New York: Cambridge University Press.

22.

Human Rights Watch (1995) Children in Sudan: States, street children, and child soldiers. Human Rights Watch Children’s Rights Projects (https://www.hrw.org/report/1995/09/01/children-sudan/slaves-street-children-and-child-soldiers).

23.

Human Rights Watch (2001) Congo – reluctant recruits: Children and adults forcibly recruited for military service in North Kivu (https://www.refworld.org/docid/3bd0239ad.html).

24.

Human Rights Watch (2007) Renewed crisis in North Kivu. (https://www.hrw.org/reports/2007/drc1007/) 7.

25.

Huntington

Samuel P

(1957) The Soldier and the State: The Theory and Politics of Civil-Military Relations. Cambridge, MA: Harvard University Press.

26.

Janowitz

Morris

(1960) The Professional Soldier: A Social and Political Portrait. Glencoe, IL: Free.

27.

Jervis

Robert

(1976) Perception and Misperception in International Politics. Princeton, NJ: Princeton University Press.

28.

Jetschke

Anja

(2011) Human Rights and State Security: Indonesia and the Philippines. Philadelphia, PA: University of Pennsylvania Press.

29.

Kalyvas

Stathis N

(2006) The Logic of Violence in Civil War. New York: Cambridge University Press.

30.

Kalyvas

Stathis N

Kocher

Matthew Adam

(2007) How ‘free’ is free riding in civil wars? Violence, insurgency, and the collective action problem. World Politics 59(2): 177–216.

31.

Kertzer

Joshua D

Tingley

Dustin

(2018) Political psychology in international relations: Beyond the paradigms. Annual Review of Political Science 21: 1–23.

32.

Krebs

Ronald R

(2004) A school for the nation? How military service does not build nations, and how it might. International Security 28(4): 85–124.

33.

Levy

Jack S

(1997) Prospect theory, rational choice, and international relations. International Studies Quarterly 41(1): 87–112.

34.

Littman

Rebecca

Paluck

Elizabeth Levy

(2015) The cycle of violence: Understanding individual participation in collective violence. Political Psychology 36(1): 79–99.

35.

Margulies

Max

(2018) Military recruitment in new states. Unpublished. Unpublished PhD dissertation, University of Pennsylvania.

36.

McCandless

Erin

Karbo

Tony

(2011) Peace, Conflict and Development in Africa: A Reader. Geneva: University for Peace.

37.

McDermott

Rose

(2001) Risk-Taking in International Politics: Prospect Theory in American Foreign Policy. Ann Arbor, MI: University of Michigan Press.

38.

Murdie

Amanda

Peksen

Dursun

(2013) The impact of human rights INGO activities on economic sanctions. Review of International Organizations 8(1): 33–53.

39.

Peled

Alon

(1998) A Question of Loyalty–Military Manpower Policy in Multiethnic States. Ithaca, NY: Cornell University Press.

40.

Posen

Barry R

(1993) Nationalism, the mass army, and military power. International security 18(2): 80–124.

41.

Sagan

Scott D

(1994) The perils of proliferation: Organization theory, deterrence theory, and the spread of nuclear weapons. International Security 18(4): 66–107.

42.

Samii

Cyrus

(2013) Perils or promise of ethnic integration? Evidence from a hard case in Burundi. American Political Science Review 107(3): 558–573.

43.

Saunders

Elizabeth N

(2011) Leaders at War: How Presidents Shape Military Interventions. Ithaca, NY: Cornell University Press.

44.

Saunders

Elizabeth N

(2018) Leaders, advisers, and the political origins of elite support for war. Journal of Conflict Resolution 62(10): 2118–2149.

45.

Sechser

Todd S

(2004) Are soldiers less war-prone than statesmen? Journal of Conflict Resolution 48(5): 746–774.

46.

Shils

Edward A

Morris

Janowitz

(1948) Cohesion and disintegration in the Wehrmacht in World War II. Public Opinion Quarterly 12(2): 280–315.

47.

Snyder

Glenn H

(1984) The security dilemma in alliance politics. World Politics 36(4): 461–495.

48.

Sookermany

Sand

Trond Svela

Breivik

Gunnar

(2019) Risk-taking attitudes and behaviors among military personnel in dangerous contexts. Moving Soldiers 1 (https://journals-sagepub-com-s.web.bisu.edu.cn/doi/full/10.1177/2158244018824498).

49.

Ssenyonjo

Manisuli

(2005) Accountability of non-state actors in Uganda for war crimes and human rights violations: Between amnesty and the International Criminal Court. Journal of Conflict and Security Law 10(3): 405–434.

50.

Toronto

Nathan

(2005) Military recruitment data set, version 2005.1. (http://nathantoronto.com/research).

51.

Weeks

Jessica L

(2012) Strongmen and straw men: Authoritarian regimes and the initiation of international conflict. American Political Science Review 106(2): 326–347.

52.

Weinstein

Jeremy M

(2006) Inside rebellion: The Politics of Insurgent Violence. New York: Cambridge University Press.

53.

Wood

Elisabeth Jean

(2003) Insurgent Collective Action and Civil War in El Salvador. New York: Cambridge University Press.

54.

Yarhi-Milo

Keren

(2014) Knowing the Adversary: Leaders, Intelligence, and Assessment of Intentions in International Relations. Princeton, NJ: Princeton University Press.