Introduction
On 6 April 1984, members of the Cameroonian Republican Guard, unhappy about personnel reductions facing their paramilitary organization, took up arms against the regime of President Paul Biya in a bloody, unsuccessful coup. Once the rebellion was thwarted, the Republican Guard was disbanded and organizers of the unrest were tried in front of a military tribunal. Not only were defecting officers removed from their posts, but most were executed or sentenced to long prison terms. Following this failed coup, President Biya has managed to stay in office for decades and has not faced anti-government conflict of this nature since purging those responsible for the violence in 1984.
Shortly after signing a peace settlement terminating his country’s long-lasting civil war in 2002, Angolan president Jose Eduardo dos Santos dismissed several top-ranking members of the military, including the army chief of staff. The ruling Popular Movement for the Liberation of Angola (MPLA) party, led by dos Santos, persists in its position of considerable power, while the former rebel group UNITA (National Union for the Total Independence of Angola) coexists peacefully as an opposition party. While the death of UNITA leader Jonas Savimbi dealt a significant blow to the movement at the end of the conflict, the fact that UNITA did not resume fighting as it did in the mid-1990s is intriguing. In this paper, we seek to explore one possible explanation for the continued peace in Angola between MPLA and UNITA, as well as the lack of subsequent violence in Cameroon: by purging powerful military officials after these conflicts ended, these presidents demonstrated the strength of their authority not just to citizens generally, but also and specifically to remaining rebels who might have contemplated a renewed challenge if the government appeared vulnerable in the post-conflict period.
Extant literature on coup-proofing
1 suggests that such activities reduce military effectiveness, yielding decreased capabilities and negative outcomes in interstate conflict (
Pilster and Böhmelt, 2011). Importantly, military ineffectiveness could provide an environment ripe for civil conflict, as rebels are expected to seize upon moments when they are in a position of strength relative to the government. However, this poses an interesting puzzle: if coup-proofing activities are dangerous, why do leaders use such strategies? We argue that a specific type of coup-proofing, military purges (especially the removal of
high-ranking officials), could potentially deter subsequent domestic unrest by demonstrating the capacity of the regime to remove powerful yet undesirable individuals from office. The strategic and intentional nature of purges signals to opponents that the regime is capable of not only identifying its enemies but also eliminating the threat posed by these enemies. Purges should be particularly effective at preventing conflict
recurrence (as compared with onset generally), because during a previous conflict the dictator observed loyalty throughout his military and became more aware of domestic threats. Thus, he can make more informed decisions about which individuals to remove so as to avoid weakening the military in a vulnerable period of recovery.
Why do we observe coup-proofing?
Much of the conventional wisdom regarding coup-proofing suggests leaders will engage in these behaviors when they anticipate that coup risk is high (e.g.
Belkin and Schofer, 2003,
2005;
Quinlivan, 1999;
Stepan, 1971;
Thyne, 2010). Conditions corresponding to poor living conditions, as well as threats to military stability and independence, are understood to motivate the military to take matters into its own hands by way of a coup. In particular, substandard economic performance not only limits resources available to the armed forces, therefore increasing grievances and the desire for a change of leadership, but also promotes coup conditions by compelling the (impoverished) public to provide the crucial support needed for a coup to be carried out (
Galetovic and Sanhueza, 2000;
Thyne and Powell, 2016;
Welch, 1970).
However, perceptions of low-coup-risk environment should not lull leaders into a false sense of security. In fact, contrary to the aforementioned literature, this sort of environment provides the optimal opportunity for a leader to coup-proof.
Sudduth (2015a) argues that engaging in coup-proofing behaviors when coup risk is already high can be counterproductive and serve to hasten a coup. The military will anticipate that, in a high coup risk environment, the leader will attempt to coup-proof, and thus the military is inclined to stage a counter-coup before the regime initiates coup-proofing activities. Building on this logic, Sudduth demonstrates that leaders are actually
more inclined to engage in coup-proofing as the likelihood of overthrow by the military decreases, and the risk of coup is therefore comparatively low. Thus, coup-proofing strategies undertaken during times of strength help the leader avoid being overthrown.
How coup-proofing affects the risk of recurrence
At least with respect to interstate conflict, coup-proofing is generally found to yield negative results for the regime–this is often attributed to the notion that coup-proofing decreases military effectiveness (
Makara, 2013;
Pilster and Böhmelt, 2011,
2012). However, the application of coup-proofing strategies to understanding civil conflict incidence has been understudied thus far.
Powell (2014) employs the measure of effective organizations from
Pilster and Böhmelt (2011) to demonstrate that maintaining more ground-based branches of the military decreases coup risk, but increases the likelihood of civil war onset in sub-Saharan African states. Considering a different coup-proofing strategy,
Roessler (2011) finds that ethnic exclusion has a similar effect on the risk of unrest amongst the same set of countries. We add to this examination of the relationship between coup-proofing and civil conflict by considering how
purges relate to domestic unrest; namely the recurrence of civil conflict globally (not just in sub-Saharan Africa).
One might wonder why we focus on civil conflict recurrence rather than onset. Crucially, in the post-conflict environment we can more reasonably assume that combatants have an improved awareness of each others’ capabilities and resolve, at least in comparison with their awareness prior to the initial onset of violence. Combatants learn about the strength of their opponent(s) during conflict, so perceptions of the enemy should be more informed after war than before it. When former (and latent) rebels observe the leader engaging in targeted removal of military personnel after a conflict ends, rebels can assume that the regime perceives itself as being strong enough to withstand a potential challenge from within its ranks, and, presumably, it has tools to repel confrontation from rebels as well.
There remains the question of whether coup-proofing impairs military effectiveness. Importantly, there could be competing expectations for how leaders anticipate military effectiveness to be impacted by purges. It is possible that purges disrupt cohesion amongst military personnel, encouraging betrayals and paranoia as officers scramble to demonstrate their loyalty (and disloyalty of enemies) to the regime. However, it is also possible that purges are an efficacious way to remove troublemaking individuals who foster the very discord that promotes military ineffectiveness in the first place; the example from Cameroon is a case in point.
If the former is true, and purges impair military effectiveness, we should expect rebels to seize upon the opportunity to (re-)initiate conflict against the government. However, particularly in the post-conflict environment, a leader should be remiss to remove officers if he anticipates this will weaken the military, thereby making the regime an attractive target for further challenges from rebels. Instead, a leader should engage in coup-proofing to strengthen his government in the face of threats, both from inside and outside the regime. We assume purges in a post-conflict environment improve regime security while minimizing risk of military ineffectiveness and subsequent challenge from armed opponents.
We believe this is a reasonable assumption in the post-conflict context because the leader has recently been afforded the opportunity to observe his officers in action during combat. Thus, dictators are able to punish (via purging) those members of the armed forces who proved ineffective, or even disloyal, in combat. Leaders logically will not engage in purges that weaken the military as this makes them vulnerable to recurrent attack. Instead, they should purge officials when this strengthens the armed forces and their loyalty to the leader, and when it signals governmental strength to opposition groups. As a result, former and latent rebels view purges as a sign that the regime is strong and thus opponents will be deterred from renewing their fight. This leads to the following hypothesis.
Hypothesis 1a: Purges decrease the likelihood of civil conflict recurrence.
Furthermore, when high-ranking officers are removed from their posts, they often meet unfortunate fates such as imprisonment or death.
Roessler (2011) warns that when the coup-proofing strategy of ethnic exclusion is employed, civil conflict can ensue as a result of the excluded individuals being motivated to join or form a rebellion as a result of their perceived injustice. In contrast, high-ranking purges leave the targeted enemies of the regime in positions where they cannot easily collaborate with opponents to challenge the government. Thus, we expect that when generals, commanders, or ministers are the targets of a purge, recurrence should be particularly rare. In other words, we have the following hypothesis.
Hypothesis 1b: When high-ranking officers are targeted, purges decrease the likelihood of civil conflict recurrence.
Research design
Our unit of analysis is the country-year, restricted to dictatorships that have previously experienced a civil conflict as defined subsequently.
2 The dependent variable is coded 1 if a post-conflict country experiences the outbreak of a new or renewed civil conflict in a given year, and 0 if not. The population of intrastate conflicts for this study comes from the UCDP Armed Conflict Database (
Themnà and Wallensteen, 2014). To be included, conflicts must involve the government as a combatant as well as at least one organized armed group, and there must be at least 25 battle deaths in a given year. Conflicts end in a year featuring fewer than 25 battle deaths.
Our independent variables come from an original dataset on purges (
Sudduth, 2015b). These data cover all autocracies
3 from 1969–2003. Data on purges were gathered from a variety of news sources, including Keesing’s Record of World Events, LexisNexis news searches, and country studies.
Sudduth (2015b) codes purges when one or more military officers are dismissed, demoted, or arrested for at least one of the following reasons: the officer was popular among other elites and is suspected to threaten the leader’s political survival, the officer had different policy preferences and criticized the dictator’s positions, and/or the officer was (presumed to be) responsible for plans to overthrow the regime.
4
We report results using two alternative specifications of purges. First, we use any purge, indicating whether any sort of purge occurred in a given year, from regular soldiers up to the highest military ranks, coded 1 if yes and 0 if no. There are 416 instances of these events in our dataset.
Second, we employ top purge, examining the purging of high-ranking officials. This includes positions such as army chiefs of staff, commanders of a branch of the military, and defense or interior ministers. We observe 191 cases of top-level purges in our data. We lag both variables by 1 year in order to ensure that the purge event predates conflict recurrence.
We include a number of control variables commonly associated with civil conflict. The outcome of the previous conflict has implications for the post-war environment (
Quinn et al., 2007). In particular, decisive military victories should be less likely to produce renewed unrest. The variable
victory comes from the UCDP Conflict Termination dataset (
Kreutz, 2010) and is coded 1 if the conflict ended in military victory for the rebels or government and 0 otherwise. We also control for the natural log of the
duration of the previous conflict as well as the log of total battle-related deaths (
Lacina and Gleditsch, 2005). Finally, peacekeepers have been found to make recurrence less likely (
Fortna, 2004;
Kreutz, 2010). The variable
peacekeepers is drawn from
Heldt and Wallensteen (2005) and is coded 1 if a peacekeeping operation was present in the country and 0 if not.
Further determinants of conflict recurrence are country-level conditions. We use data from
Cheibub et al. (2010) to control for whether the dictatorship is a
military regime (coded 1 if yes, 0 if civilian-led or a monarchy). We also draw the variable
leader tenure from this dataset to account for how long the head of state has been in office. We control for
population, the natural log of the country’s population, and
GDP per cap, the natural log of per capita gross domestic product from the previous year. Both variables are drawn from
Gleditsch (2002). Data on infant mortality,
IMR, comes from
Abouharb and Kimball (2007), and this variable is also lagged 1 year to ensure that we capture conditions influencing recurrence and not the other way around.
To account for various dynamics unique to the
post Cold War period, we employ a variable indicating whether an observation occurs during that period (coded 0) or after 1989 (coded 1). Finally, we control for the number of years since conflict termination by including
peace years, as well as its squared and cubed polynomials in accordance with
Carter and Signorino (2010), to account for possible temporal dependence amongst observations. Summary statistics of all variables are presented in Table A1 of the online appendix.
Findings
Results are reported in
Table 1. Because the dependent variable is binary, we employ logistic regression to determine the effects of our covariates on the likelihood of civil conflict recurrence.
5 Model 1 assesses the effect of purges at any level of the military on the likelihood of conflict recurrence, while Model 2 restricts consideration of purge events to those that involve the removal of high-ranking officers.
Model 1 suggests that purges in general do not have a statistically significant effect on the likelihood that civil conflict will be renewed. While in some instances it might be true that military purges prevent recurrence, it would appear that there are also some situations in which this has the opposite, or perhaps non-existent, effect. Thus, Hypothesis 1a is not supported.
Turning to the results for Model 2, whose coefficients are plotted in
Figure 1, the effect of removing top military officials appears to reliably decrease the chances of recurrent conflict. The coefficient for
top purge is negative and statistically significant, suggesting that such coup-proofing efforts are an effective strategy for dictators seeking to avoid renewed unrest. Holding other covariates at their mean or median values, a move from no post-conflict purges to a purge of top officials is associated with a 75 per cent decrease in the relative risk of conflict recurrence.
6 This provides support for Hypothesis 1b, which holds that only purges of high-ranking figures systematically deter the recurrence of conflict.
7
With respect to our control variables, we find that autocracies with conflicts that ended in military victory tend to be more peaceful following termination than post-conflict dictatorships where the violence produced a stalemate or settlement. Interestingly, newly installed leaders appear to be less likely to face renewed unrest as compared with their more experienced counterparts. Recurrence is more likely in the post-Cold War period, and peacekeepers are somewhat effective in deterring a return to conflict.
We believe the contrast between our findings regarding recurrence and recent work on civil conflict
onset can best be explained by the level of information dictators are able to employ when coup-proofing in the pre- versus post-conflict contexts. Work by
Roessler (2011) and
Powell (2014) regarding onset suggests that coup-proofing increases the risk of unrest because such activities destabilize and weaken the military. Conversely, we contend that leaders engage in post-conflict coup-proofing, purges in particular, in order to strengthen the war-fighting capacity and loyalty of the armed forces, making the government a more intimidating target for renewed attack. As
Powell (2012: p. 1036) suggests, “coup-proofing can bring increased stability to an otherwise vulnerable country”. We demonstrate that purging high-ranking officials, a coup-proofing strategy not considered previously, also promotes domestic peace, at least in the form of preventing recurrent rebellion.
Conclusion
Despite extant studies suggesting that coup-proofing measures impede military effectiveness and promote conflict, we demonstrate that not all forms of coup-proofing induce unrest. We differ from existing work by highlighting a previously unexplored aspect of coup-proofing: purges. Using new data, we find that purges of high-ranking military officers help prevent further unrest in dictatorships that have experienced civil conflict in the past. This suggests, perhaps perversely, that an effective method to deter remaining challengers from taking up arms is for a dictator to remove from office one or more generals, commanders, or defense-related ministers. In doing so, the leader signals to opponents that he is sufficiently strong so as to be able to withstand any fallout that might ensue following the purge of senior officials.
We suggest one possible explanation for the recurrence-reducing effect of purges is that such actions warn opponents that the regime is able to identify and eliminate its enemies. However, our purge data are not sufficiently nuanced so as to differentiate between alternative mechanisms that might account for this relationship. Future efforts to explore specifically how potential challengers view and react to military purges, likely using micro-level data, would certainly be welcome to further parse out the causal mechanism(s) at play here.
These data on autocratic purges could be employed in many innovative ways. For example, it would be useful to know how officer removal shapes democratization efforts, economic growth, the ability to attract and distribute external assistance, and so on. We offer initial insight into an implication of purges, showing that the removal of high-level officers fosters peace in post-conflict environments, but there is ample opportunity for further investigation of the causes and consequences of this coup-proofing strategy.
Acknowledgments
An earlier version of this paper was presented at the 2014 European Peace Science Association Annual Meeting in Edinburgh, as well as at the 2014 Kobe Sakura Workshop in Brussels. We thank the Graduate School of Law (Kobe University), the CROP-IT Project (JSPS), the Suntory Foundation, and the Daiwa Anglo-Japanese Foundation for their generous support of this workshop. We are grateful to participants at both events for many useful suggestions. We also thank the editors and anonymous reviewer at Research and Politics for their helpful comments. Any remaining errors are our own.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.