Capital gains: Revising the impact of coup proofing on tactical military effectiveness

Abstract

This paper investigates how counterbalancing shapes the relationship between distance from the capital and government troops' military effectiveness. It argues that as counterbalancing intensifies, the negative effect of distance on battlefield performance becomes more pronounced. This argument rests on three mechanisms: the concentration of elite units near the capital, intensified coordination and logistical challenges in peripheral areas, and the compounding of these constraints as fragmentation increases. To test these claims, I combine battle-level data with measures of counterbalancing and estimate fractional regression models. The results show that the erosion of effectiveness with distance is greater at higher levels of counterbalancing.

Keywords

Civil war conflict coup prevention military effectiveness

For all rivers, even though they be impassable at a distance from their sources, become passable, without even wetting your knees, as you approach toward the sources.

(Xenophon, Anabasis, Book III, Chapter II)

Introduction

Current scholarship finds that counterbalancing—the creation of armed counterweight units to offset the army—helps leaders mitigate coup risk but diminishes state security forces’ effectiveness (Böhmelt and Pilster, 2015; De Atkine, 1999; Hoekstra, 2020; Marcum and Brown, 2017; Pollack, 2004; Pilster and Böhmelt, 2011; Quinlivan, 1999). Notable exceptions exist: from the Eastern Roman Tághmata and Varangian Guard to Syria's Republican Guard 104th Brigade, elite units tasked with regime protection have repeatedly mounted fierce resistance to enemy forces (D’Amato, 2012; Phillips, 2011; Waters, 2018).

This article addresses the resulting tension between theoretical expectations and anecdotal evidence by analyzing engagement outcomes between government forces and violent non-state actors (VNSAs). I argue that the location of combat mediates the relationship between counterbalancing and government troops’ tactical military effectiveness (GTME). Specifically, I contend that counterbalancing efforts increase the extent to which distance from the regime center (Distance) erodes GTME.

I ground this claim in three considerations. First, leaders who engage in counterbalancing tend to keep counterweight units—which are often the best equipped, best trained, and most motivated formations in the armed forces—near the capital, while stationing other forces in the periphery (Asher and Hammel, 1987; Brooks, 2007; Pollack, 2004; Powell, 2014). As a result, in counterbalanced states, engagements near the capital are more likely to involve these elite formations, whereas distant engagements are more likely to be fought by peripheral units—until, and unless, the leadership deploys regime-protection forces to the front.

Second, counterinsurgency in remote areas offers some advantages but poses several substantial challenges. Although political and operational constraints on the use of force may loosen as Distance increases, supply lines lengthen and become more critical;¹ coordination and intelligence collection become more difficult; and communication barriers as well as information asymmetries grow (Black, 2021; Boisvert, 2016; Boulding, 1963; Drea, 2016; Eccles, 1989; Foxton, 1994; Gropman, 1997; Kane, 2001; Moore et al., 1999; Muckensturm and Longhorn, 2019; Nagl et al., 2008; US Army and US Marine Corps, 2014; Skoglund et al., 2022; Van Creveld, 1977, 1985). Counterbalancing reduces opportunities to exploit these advantages and compounds these distance-induced challenges by reducing inter-unit and inter-service coordination (Böhmelt and Pilster, 2015; Pilster and Böhmelt, 2011), and by depressing morale and skill accumulation among units outside privileged formations (Brooks, 2007; Dahl, 2016). Combined, these mechanisms reduce the capacity of counterbalanced security forces to sustain effective operations as Distance grows.

Finally, as states increase the number of counterweight units, these mechanisms are likely to increase. The more leaders engage in counterbalancing, the more pronounced the concentration of high-quality forces near the capital becomes, morale problems in non-privileged units are likely to increase, and coordination frictions intensify (Böhmelt and Pilster, 2015; Buchanan and Tullock, 1965; Chamberlin, 1974; Olson, 1971; Pilster and Böhmelt, 2011).

To test this argument, I combine data on armed counterweights from the State Security Forces Dataset (SSFD) (De Bruin, 2021) with battle events from the Uppsala Conflict Data Program Georeferenced Event Dataset (UCDP GED) (Sundberg and Melander, 2013) to conduct a battle-level analysis of 115 active government–VNSA conflicts across 36 countries from 1991 to 2010. This study offers several contributions. First, improving tactical performance in counterinsurgency has the potential to reduce both human and material costs. Second, because GTME constitutes a core dimension of military power (Brooks, 2007) and contributes to shape post-conflict patterns (Binetti and Steinwand, 2025), identifying its determinants enhances our understanding of state capacity, state-building, and conflict dynamics. In this sense, the article speaks to a growing body of recent scholarship examining leader survival, armed conflict, and post-conflict development (Binetti and Steinwand, 2025; Hidaka et al., 2025; Hlatky and Landry, 2025). Third, by linking counterbalancing to battlefield outcomes, the study extends debates on how institutional coup proofing shapes regime survival beyond its effects on coup risk and war onset (Belkin and Schofer, 2003, 2005; Böhmelt and Pilster, 2015; Powell, 2019; Sudduth, 2017a). Fourth, it expands current research on the effect that battle location has on structural dimensions of military effectiveness (Binetti, 2024; 2025). Finally, it extends the research of Pilster and Böhmelt (2011) by incorporating Boulding's (1963) “loss-of-strength gradient” at the tactical level and by shifting the focus from interstate to intrastate conflict.

Coup-proofing and counterweights

Leaders have a range of strategies to deter or defeat coup plotters. They may redirect national resources toward the military (Albrecht, 2015; Albrecht and Eibl, 2018; Bueno de Mesquita and Smith, 2017; Powell et al., 2018), purge disloyal actors (Bueno de Mesquita and Smith, 2017; Easton and Siverson, 2018; Oswald et al., 2024; Sudduth, 2017b), incorporate military leaders into national politics (Albrecht, 2015), cultivate regime legitimacy (Roberts, 1975; Esen and Gumuscu, 2017; Rabinowitz and Jargowsky, 2018; Rozenas and Zeigler, 2019), or engage in ethnic stacking—recruiting and promoting coethnics within the armed forces (Enloe, 1975; Goldsworthy, 1981; Harkness, 2022). Yet, counterbalancing—the creation or empowerment of armed units outside the regular army's chain of command to offset the army—often remains the centerpiece of coup proofing (Belkin and Schofer, 2003, 2005; Böhmelt and Pilster, 2015; De Bruin, 2020, 2021; Geddes et al., 2018; Marcum and Brown, 2017; Quinlivan, 1999; Sudduth, 2017a).

Counterbalancing operates through four primary mechanisms. First, it fosters divergent interests within the security apparatus and increases the number of actors willing to resist a coup (De Bruin, 2018; Marcum and Brown, 2017). Second, it degrades coordination and communication within the armed forces by multiplying chains of command and constraining opportunities for joint inter-unit and inter-service training (Albrecht and Eibl, 2018; Böhmelt and Pilster, 2015; Pascal et al., 1979; Pilster and Böhmelt, 2011). Third, it creates forces specifically tasked with protecting the leader and monitoring other security organizations, and equips and trains them accordingly (De Bruin, 2018, 2021; Marcum and Brown, 2017; Quinlivan, 1999). Fourth, it enables leaders to staff counterweight units with loyalists, thereby strengthening regime control over the security forces (De Bruin, 2018).

Military effectiveness and its determinants

In this article, tactical military effectiveness is defined as an actor's ability to inflict damage on an adversary while minimizing losses to its own forces in discrete engagements (Biddle, 2010; Dupuy, 1987; Millett et al., 1986).² It rests on three core conditions: access to adequate high-quality equipment, willingness to fight, and the organizational capacity to convert these inputs into combat power (Brooks, 2007; Hart, 1996; Jones, 1994; Millett et al., 1986; Reese, 2011; Thoral, 2011; Watson, 2008).

Several factors shape these conditions. In intrastate wars, the government's ability to mobilize civilian support or, at minimum, deny it to VNSAs, plays a crucial role (Berman and Matanock, 2015; Galula, 2006; Kalyvas, 2006; Petersen, 2001; Wood, 2003). States may pursue this objective through hearts-and-minds strategies—despite their practical difficulties (Berman and Matanock, 2015; de Bruin, 2022; Dixon, 2009; Egnell, 2010; Galula, 2006; Kheng, 1991; Lyall and Wilson, 2009; Nagl et al., 2008; Sexton and Zürcher, 2024)—or through coercive approaches (Downes, 2007; Kalyvas and Kocher, 2009; Kocher et al., 2011; Böhmelt et al., 2020). Other relevant influences include regime type and institutional characteristics (Binetti, 2024; Brooks, 2007; Gentil-Fernandes and Otto, 2024; Posen, 1986; Lyall, 2010, 2020; Reiter and Stam, 1998),³ troop cohesion and esprit de corps (Jones, 1994; Millett et al., 1986), economic development (Beckley, 2010), reliance on private military companies (Petersohn, 2017; Dunigan, 2020), and patterns of aid distribution in conflict zones (Polman, 2010a, 2010b; Binetti, 2025).

Research on ruler–military relations identifies further conditions affecting military effectiveness. First, when rulers cannot remove poorly performing commanders without incurring political risks, military performance is likely to suffer (Reiter and Wagstaff, 2018). Second, purges weaken military institutions and degrade intelligence collection (Eck, 2015; Glantz, 1989; Reese, 1992). Third, weak civil–military coordination hampers accurate assessment of strategic developments (Brooks, 2007). Fourth, ethnic stacking can depress morale and reduce skills among soldiers and officers (Brooks, 2006, 2007). Finally, the creation of armed counterweights disrupts inter-unit coordination (Böhmelt and Pilster, 2015; Pilster and Böhmelt, 2011), increases defection risks, and is associated with declines in officer quality, training standards, morale, and intelligence effectiveness—especially within regular-army units excluded from privileged counterweight formations (Brooks, 2007; Dahl, 2016; Dworschak, 2020; Lutscher, 2016; Pollack, 2004; Talmadge, 2015).

Counterweights, distances, and military effectiveness

Building on this literature, I posit that the more counterweight units a country fields, the more Distance erodes GTME. This expectation derives from three considerations.

First, leaders who adopt counterbalancing typically hollow out the regular army while concentrating the best-trained, best-equipped, and most motivated personnel in counterweight units oriented toward regime defense and deployed in the proximity of the capital (Asher and Hammel, 1987; Brooks, 2007; De Bruin, 2018, 2020; Pilster and Böhmelt, 2011; Pollack, 2004; Powell, 2014; Scott, 2006; Warrington, 1998). Consequently, in counterbalanced states, GTME is likely to decline as Distance increases because clashes further from the capital are less likely to pit rebels against units that are well equipped, well trained, and highly motivated to fight.

Second, operating in remote theaters can offer occasional advantages but typically imposes substantial challenges. On the one hand, fighting in sparsely populated areas or politically hostile localities can relax political and operational constraints on the use of firepower, enabling government troops to limit their own losses and impose high costs on VNSAs—as illustrated by Operation Anaconda (2002) and the Second Battle of Fallujah (2004) (Head, 2013; Kugler et al., 2009). On the other hand, as Distance increases, counterinsurgency becomes more demanding. Lines of communication lengthen and become more vulnerable to infiltration and guerrilla attack (Black, 2021; Drea, 2016; Muckensturm and Longhorn, 2019). Cultural barriers and manipulation by local actors make intelligence collection more difficult in peripheral and contested regions (Boisvert, 2016; Nagl et al., 2008; US Army and US Marine Corps, 2014). Finally, the inter-service and inter-unit coordination required for rapid, lethal, long-range operations is harder to sustain as operational complexity increases with Distance (Van Creveld, 1985).

Counterbalancing both limits any ability to exploit the occasional advantages of remote-area warfare and intensifies the frictions such operations entail. First, it degrades inter-unit and inter-service coordination, which are central both to rapid joint operations and to protecting extended lines of communication (Bensahel and Byman, 2004; House, 1984; Pilster and Böhmelt, 2011). Second, it keeps elite formations near the capital, leaving distant fronts disproportionately staffed by less capable personnel (Asher and Hammel, 1987; Brooks, 2007; Pollack, 2004; Powell, 2014, 2019; Scott, 2006). Third, it depresses morale and slows skill accumulation in non-counterweight units (Brooks, 2007; Dahl, 2016)—the formations most likely to operate far from the capital. Fourth, it reallocates resources toward internal threat monitoring (Quinlivan, 1999), which can crowd out intelligence and surveillance directed at VNSAs. Together, these dynamics reduce the ability of counterbalanced security forces to secure lines of communication, concentrate combat power, and execute the synchronized operations needed to fix, encircle, and defeat enemy forces as Distance increases.

Third, if these mechanisms operate as argued, the marginal effect of Distance on GTME should increase as counterbalancing intensifies. The more counterweight units leaders field, the more pronounced the concentration of high-quality forces near the capital becomes, morale problems in non-privileged units are likely to deepen, and coordination frictions increase (Buchanan and Tullock, 1965; Chamberlin, 1974; Olson, 1971; Pilster and Böhmelt, 2015). From this argument, therefore, it is possible to derive the following testable hypothesis:

H1: The adverse impact of Distance on GTME becomes more pronounced as the number of counterweight units increases.

Notably, the theory is agnostic about two issues. First, it does not specify how government forces in counterbalanced states perform relative to those in non-counterbalanced states when Distance is low. Second, it does not make a definitive claim about the effect of Distance on GTME in the absence of counterbalancing.

On the one hand, when Distance is small, counterbalancing efforts could be associated with higher GTME. When engagements occur close to the regime center, support and command-and-control requirements are easier to meet. Under such conditions, the advantages of concentrating multiple elite units around the capital may outweigh the costs of counterbalancing discussed above. In this scenario, counterbalanced forces may not only outperform VNSAs but may also perform more effectively than forces from non-counterbalanced states.

On the other hand, the negative effects of counterbalancing on GTME may be too substantial to be offset, even under favorable conditions. If so, even in proximity to the capital, armed forces in counterbalanced states would be less effective—not only relative to their adversaries but also in comparison with forces from non-counterbalanced states.

The expected association between Distance and GTME in non-counterbalanced states is likewise ex ante indeterminate. As discussed, operations in remote areas can simultaneously relax some tactical constraints while intensifying logistical, intelligence, and coordination demands; the direction of the Distance–GTME relationship therefore depends on which of these dynamics dominates. Non-counterbalanced militaries may be well positioned for the advantages of fighting in remote areas to outweigh the disadvantages because they are less likely to experience the morale erosion, coordination degradation, uneven training, and equipment skew that counterbalancing generates. In other words, while counterbalancing is likely to preclude a positive Distance–GTME gradient, in non-counterbalanced states the gradient could plausibly be positive—or at least non-negative.

Counterbalancing and its consequences on the field

A comparison of Rhodesia and Libya in their counterinsurgency efforts illustrates how intensive counterbalancing steepens the negative relationship between Distance and GTME. Rhodesia shows that a relatively small, well-trained force can sustain tactically effective operations across a large territory when institutions promote inter-unit and inter-service coordination. In contrast, the early phase of the 2011 Libyan civil war indicates that heavily counterbalanced security forces were able to suppress entrenched rebels near the capital yet struggled to halt enemy advances and inflict decisive losses in more distant theaters.

Our mutual comrade: Counterinsurgency in Rhodesia and Operation Dingo (1965–1977)

On the eve of the Unilateral Declaration of Independence, police-led operations—supported by the army, air force, and intelligence services—had largely neutralized insurgent threats to Rhodesian urban areas (Mills and Wilson, 2007; Wood, 2009). Nevertheless, the Rhodesian security forces confronted a daunting set of challenges: conducting counterinsurgency against a geographically dispersed adversary across vast, sparsely populated, and often impassable terrain, while being outnumbered by at least three to one and operating under stringent resource constraints imposed by international economic sanctions (Mills and Wilson, 2007; Wood, 2009).

To meet these challenges, the Rhodesian government moved early to strengthen coordination across the security forces (Wood, 2009). This integration, reinforced through continuous joint training, sustained high levels of operational and tactical performance over the course of the war and facilitated adaptation to shifting constraints and insurgent countermeasures (Baxter, 2014; Mills and Wilson, 2007). Between 1972 and 1978, the kill ratio between insurgents and government forces was roughly 6:1 in favor of the Rhodesian security forces, and in some engagements it reached 80:1 (Arbuckle, 1979; Wood, 2009).

Inter-service cooperation typically followed a standardized sequence under a unified command structure. Special forces and intelligence personnel would identify infiltrating rebel groups, their routes, logistical networks, and base areas by extracting information from villagers, detainees, aerial reconnaissance, and insurgent intermediaries (Arbuckle, 1979; Wood, 2009). Once a target was fixed, airborne rapid-reaction forces—supported by close air support and, where necessary, ground elements—moved to interdict or destroy it (Arbuckle, 1979; Wood, 2009).

The 1977 strike on the ZANLA camp at New Farm effectively illustrates how high levels of integration and coordination enabled the Rhodesian security forces to inflict heavy losses while sustaining minimal casualties, even at considerable distance from the capital. In late 1976 and early 1977, Rhodesian intelligence identified two major rebel base complexes just across the Mozambican border: New Farm, roughly 300 km southeast of Salisbury, and Tembué, approximately 381 km northeast (Baxter, 2014; Wood, 2009). These installations served as logistical hubs and command centers, as well as reception, holding, and training facilities (Baxter, 2014).

Striking the camps posed several formidable challenges. First, both sites featured layered fieldworks, dug-in anti-aircraft positions, and early-warning systems (Baxter, 2014; Wood, 2009), and benefited from support provided by Mozambican armored units (Wood, 2009). Second, the camps’ distance from the Rhodesian border required the operation to be conducted entirely by air, including the extraction of troops at its conclusion (Wood, 2009). Third, mounting attrition within the Rhodesian Air Force constrained the extent to which airpower could be relied upon to deliver a decisive effect (Wood, 2009). Finally, insurgents enjoyed numerical superiority and had dispersed their facilities across a wide area, complicating efforts at containment and encirclement (Wood, 2009).

In the second half of 1977, the plan to strike the camps took clearer shape as Operation Dingo. The concept called for a joint air force–airborne assault against New Farm (Zulu I), followed approximately one day later by a similar operation against Tembué (Zulu II) (Wood, 2009). To overcome the constraints associated with long-range raids on fortified camps, the plan depended on intensive inter-service coordination and on close coupling of logistical planning with tactical implementation.

As Wood (2009) reports, Zulu I was designed to begin with a diversionary overflight of New Farm by a civilian aircraft to mask the sound of approaching strike aircraft and helicopters. Approximately 30 s after the diversionary overflight, fixed-wing strike aircraft and helicopters were to initiate attacks on the camp, suppressing air defenses, disrupting insurgent formations, and striking structures and infrastructure. Roughly 2 minutes after the diversionary flight, paratroopers were to begin their drops on the camp's periphery to seal it from the southwest and southeast. Approximately 3 minutes later, helicopters were to join the assault, inserting troops to secure key high ground and close the camp on a third side. The remaining open side of the encirclement was to be covered by helicopter fire to prevent escape. Ground elements would then converge toward the center of the camp under the direction of a command helicopter, which coordinated the movement of troops and the employment of air and rotary-wing fire to reduce friendly-fire risk, suppress insurgent resistance, and channel fleeing ZANLA fighters toward blocking positions. Approximately 15 minutes after the opening strikes, helicopter-borne elements were tasked with establishing an improvised forward administrative area roughly 20 km from New Farm. Helicopters would insert technical personnel alongside medical staff and intelligence teams to set up the site, after which fuel supplies would be parachute-dropped. This forward support node was designed to sustain continuous helicopter coverage that would otherwise have been infeasible given the distance from Rhodesian bases, enable rapid ammunition resupply, facilitate casualty evacuation, and allow intelligence to be collected and processed quickly.

Zulu I was launched on 23 November 1977, and Rhodesian forces had returned to base by the evening of the following day (Wood, 2009). The assault force confronted approximately 10,000 ZANLA personnel, including roughly 4500 armed fighters, with fewer than 200 men (Baxter, 2014; Mills and Wilson, 2007; Wood, 2009). The operation destroyed the New Farm complex and inflicted losses in the thousands killed and wounded, while Rhodesian casualties totaled two killed and 12 wounded (Baxter, 2014; Mills and Wilson, 2007; Wood, 2009).

A tale of two fronts: Counterinsurgency in Tripolitania and Cyrenaica in 2011

At the outset of the uprising, Libya's regular armed forces were small, poorly trained, and poorly equipped. The force structure was largely limited to brigade- and battalion-level formations; command arrangements were politicized and often duplicated; and the regime relied primarily on loyalist elite paramilitary units as the backbone of its security forces (Bell and Witter, 2011; Cordesman and Nerguizian, 2009; Pollack, 2004; Quinlivan, 1999). The most prominent of these formations was the 32nd Reinforced Brigade, which combined heavily armed infantry with mechanized elements and artillery (Bell and Witter, 2011). As the protests that began in Benghazi on 17 February escalated into a nationwide insurrection, a pronounced disparity emerged between loyalist performance in the west—close to Tripoli—and in the east, far from the capital.

In Cyrenaica, regime control eroded rapidly. On 18 February, protesters in Benghazi—the principal city of Cyrenaica (approximately 1018 km east of Tripoli along the coastal highway)—began besieging the military barracks known as the Katiba (Bell and Witter, 2011). Two days later they were joined by defecting military units and compelled the isolated elements of the 32nd Reinforced Brigade that Tripoli had deployed there days earlier to withdraw under a safe-passage arrangement (Bell and Witter, 2011).

The fall of the Katiba, together with accelerating defections across Cyrenaica, enabled insurgents to seize weapons and advance westward toward Brega (around 779 km east of Tripoli and roughly 241 km west of Benghazi), capturing the town (Bell and Witter, 2011). Loyalist forces mounted no immediate, effective response, and the first major counterattack occurred only on 2 March, when regime troops attempted to retake Brega (Bell and Witter, 2011). Yet, although they initially recaptured portions of the town, they failed to hold the terrain and soon withdrew demoralized units after proving unable to combine suppressive artillery fire and air support effectively enough to halt rebel pressure (Bell and Witter, 2011).⁴

Events in Tripolitania unfolded differently. By 18 February, demonstrations had erupted in Zawiyah (about 50 km west of Tripoli), Libya's fourth-largest city (Bell and Witter, 2011). By 20 February, protests had reached Tripoli, and outnumbered security forces had ceded control of Sabratha (roughly 77 km west of Tripoli) and Zawiyah (Bell and Witter, 2011). On 21 February, the regime prepared and launched a counteroffensive around the capital (Bell and Witter, 2011). By the end of the day, paramilitary units had suppressed protests in Tripoli, and by 22 February they had reoccupied the capital's main epicenters of unrest (Bell and Witter, 2011). On 23 February, after securing Tripoli, loyalist forces moved to attack Sabratha, a step intended to isolate Zawiyah from Zuwarah (115 km west of Tripoli) and prevent the formation of a unified front of resistance immediately west of the capital (Bell and Witter, 2011).

Over the following days, pro-regime forces retained the initiative in the west. On 24 February, a small loyalist contingent attempted to push into Zawiyah (Bell and Witter, 2011). When this effort failed, the regime deployed elite counterweight forces under Major General Khweldi Hamedi to retake the city (Bell and Witter, 2011). Hamedi began to employ combined-arms tactics against protesters and defectors led by Colonel Hussein Darbouk, who had entrenched in the city and obtained small arms, anti-tank and anti-aircraft weapons, technicals, armored vehicles, and tanks (Bell and Witter, 2011). By 1 March, loyalist forces had retaken Sabratha and initiated operations to encircle Zawiyah (Bell and Witter, 2011).

This cross-front disparity persisted into early March. In the east, rebels expelled loyalist troops from Ras Lanuf (about 644 km east of Tripoli) on 4 March and pushed them out of Bin Jawad (roughly 604 km east of Tripoli) the next day, before returning again to Ras Lanuf (Bell and Witter, 2011). In Tripolitania over the same period, in contrast, elements of the 32nd Brigade—operating from a Tripoli now firmly under regime control—moved against Zawiyah, completed the city's encirclement, disrupted a surprise breakout attempt led by Colonel Darbouk, and initiated a coordinated pincer maneuver on the city (Bell and Witter, 2011).

Developments after 5 March eventually reversed momentum in the east, but they continued to underscore the gap in loyalist capabilities between Cyrenaica and Tripolitania. On 6 March, loyalist forces achieved their first tactical success in the east when, with the support of the local population, they ambushed rebels re-entering Bin Jawad (Bell and Witter, 2011). Yet the gains remained limited: despite access to aircraft and helicopters, loyalist units failed to encircle and destroy retreating rebel forces (Bell and Witter, 2011). Constraints were similarly evident on 7 March, when loyalist troops attacked Ras Lanuf (Bell and Witter, 2011). Rather than maneuvering to envelop the town and employing combined-arms tactics to annihilate rebel units, they relied primarily on bombardment to dislodge defenders before ground forces advanced (Bell and Witter, 2011). Rebels consequently suffered limited casualties and withdrew to Brega by 10 March (Bell and Witter, 2011). Between 13 and 14 March, loyalist forces again used preparatory bombardment before advancing to dislodge rebels from Brega—an approach that again allowed rebel forces to withdraw eastward and reorganize (Bell and Witter, 2011).

By contrast, on the western front, loyalist forces executed a large, coordinated assault on Zawiyah on 8 March, employing as many as 50 tanks and 120 light trucks along three axes of advance (Bell and Witter, 2011).⁵ The offensive broke through rebels’ static defenses and inflicted substantial damage.⁶ By 11 March, loyalist units had completed their pincer maneuver and captured Zawiyah through synchronized artillery strikes and infantry advances (Bell and Witter, 2011). This success enabled Hamedi's forces to move on Zuwarah and recapture it on 14 March, while elements of the 32nd Brigade began redeploying to the eastern front (Bell and Witter, 2011).

That redeployment produced a decisive shift in the east. On 15 March, the arrival of 32nd Brigade elements enabled loyalist forces to encircle Ajdabiya—where rebels had regrouped after withdrawing from Brega the previous day—and to initiate an assault while maneuvering to bypass defensive positions and threaten Benghazi (Bell and Witter, 2011). By 18 March, loyalist forces had secured Ajdabiya and appeared poised to enter Benghazi (Bell and Witter, 2011). On 19 March, however, French air strikes—marking the onset of international air operations following United Nations Security Council Resolution 1973—halted the advance, prevented the fall of Benghazi, and brought the war's first phase to a close (Bell and Witter, 2011).

Method

Sample selection

I test H1 by estimating a series of fractional regression models using global battle-level observations from 1991 to 2010. The analysis draws on the UCDP GED (Sundberg and Melander, 2013) for data on battles' location and outcome. Following the dataset, a battle is defined as an incident involving armed force between government forces and a VNSA at a specific location on a specific date that results in at least one direct death (Croicu and Sundberg, 2015; Sundberg and Melander, 2013).

Battle-level data might be imprecise in terms of geolocation and fatalities count (Weidmann, 2015; Vesco et al., 2026). To ensure data reliability, I apply filters using the variables event_clarity, where_prec, and date_prec from the UCDP GED (Sundberg and Melander, 2013). Specifically, I include only those events where the original sources provide sufficient detail to identify individual incidents, the event's date is known within a two- to six-day range, and the location is specified at least down to the second-order administrative division (Croicu and Sundberg, 2015).

Counterweight units are mostly ground units and comparing their performance with that of non-ground units would be inappropriate. Therefore, to include only direct ground-based clashes and exclude airstrikes and long-distance bombardments, all events in the dataset with words such as “airstrike”, “bomb”, “mortar”, “rocket”, “air”, “artillery”, “blast”, and “plane” in the variable source_article are dropped, and only those containing words like “retake”, “battle”, “fight”, “push”, “assault”, “shooting”, “clash”, “capture”, “defend”, “kill”, “repel”, “hold”, “gunman”, “unrest”, “victory”, “defeat”, “repulse”, “trap”, “ambush”, “sniper”, “tank”, “gunfire”, “defense”, “ground”, “advance”, “take”, “attack”, “liberate”, “commando”, “offensive” and “storm” are retained. This selection is further restricted and relaxed in additional specifications discussed in the robustness checks section. Figure 1 shows the global distribution of events in the sample used to estimate Model Main in the next section.

Figure 1.

Number of battles per country.

It is important to note that the UCDP GED (Sundberg and Melander, 2013) identifies the government and VNSAs involved in each battle but does not distinguish among specific government units. For example, for a battle occurring in Syria in 2011, the dataset cannot determine whether the engaged government forces belonged to one of the Republican Guard's armored brigades or to one of the independent light infantry brigades active at the time. This limitation necessitates an empirical assumption: the closer a battle occurs to the capital of a counterbalanced regime, the more likely it is that counterweight units—rather than regular forces—are involved, and vice versa as Distance increases. Historical evidence supports this assumption. In Iraq and Syria, counterweight formations often enjoyed privileged access to the capital and were held in reserve nearby to secure the regime, even when frontline demands were pressing (Asher and Hammel, 1987; Brooks, 2007; Pollack, 2004; Scott, 2006).

Measuring tactical military effectiveness

I operationalize GTME in a given battle b as:

G T M E_{b} = \frac{r e b_{b}}{r e b_{b} + g o v_{b}}

(1)

for each b where

r e b_{b} + g o v_{b} > 0

(2)

where reb_b represents the number of fatalities that VNSAs suffered in battle b, and gov_b the number of fatalities that government troops suffered in the same encounter.

This operationalization offers several advantages. First, similar to the loss-exchange ratio, it provides a continuous indicator that is not reliant on subjective coding and corresponds closely with qualitative assessments of military effectiveness (Beckley, 2010; Biddle, 2010; Biddle and Long, 2004; Desch, 2002; Pilster and Böhmelt, 2011). Second, it captures outcomes at the tactical level by disaggregating broader military operations into discrete engagements. Third, it offers a measure of tactical gains for contexts where territorial control is an unreliable indicator—as during civil wars with porous frontlines. Fourth, by compressing extreme values, the bounded nature of this operationalization mitigates the influence of outliers resulting from miscoding or incomplete reporting in the source material. This is a critical advantage given that, unlike traditional datasets on interstate warfare, the UCDP GED includes highly granular event data, often involving relatively low fatality counts.

In addition to these considerations, it is worth recalling that treating “state and nonstate military methods as autonomous, mutually exclusive categories” is an unfortunately common misconception in conflict studies (Biddle, 2022, chapter 2). Although vast open-field battles between VNSAs and regular armies are quite unlikely, still large-scale ambushes (as occurred during the war in Chechnya; Askerov, 2015), assaults on military bases (as occurred in Myanmar),⁷ sieges and urban warfare (like the battles for the control of Mosul in 2014 and 2017 and around Aleppo between 2012 and 2016), are not uncommon at all. In this sense, body counts provide an instrument to compare and rank the outcome of different tactical developments from the least to the most favorable for the warring factions involved.

Finally, it is important to acknowledge that, despite its advantages, this approach to capturing GTME is not without limitations (Gartner and Myers, 1995). Relying exclusively on casualty figures to measure combat effectiveness risks overlooking other dimensions, such as mission objectives and the broader operational and strategic context. In some cases, high casualty rates may reflect deliberate strategic trade-offs aimed at achieving longer-term goals. As such, favorable aggregated GTME outcomes do not necessarily imply success at the strategic or operational level: despite the effectiveness of the Rhodesian armed forces, today's capital of Zimbabwe is Harare.

Nevertheless, this underscores a key contribution of the present study: its emphasis on the often-neglected tactical dimension of military performance, rather than on broader strategic or operational results. Using the observations from the Model Main, Figure 2 reports the cross-country distribution of the sample means for GTME.

Figure 2.

Mean government troops’ tactical military effectiveness (GTME) across countries.

Measuring counterbalancing efforts and battles’ location

To measure counterbalancing efforts and Distance, I rely on N. H.A. and Distance. The former is defined as the number of distinct heavily armed counterweight units a country fields in a given year according to the SSFD (De Bruin, 2021)—lagged by one year. The latter is defined as the natural logarithm of one plus the distance between the battle location and the capital of the country in which the battle occurred, measured in tens of kilometers. Using the observations included in Model Main, Figures 1A and 2A in the Online Appendix report the cross-country distributions of the sample means for both variables.

Model covariates

To mitigate omitted-variable bias while limiting post-treatment bias and multicollinearity concerns, the models used to test the hypotheses include, beyond country fixed effects, only a parsimonious set of covariates (Dworschak, 2024; Schrodt, 2014) including: the number of government–VNSA clashes within 50 km of the battle site during the month preceding the engagement, to capture localized conflict intensity; distance to the nearest city with a population of at least 100,000, as a proxy for the urban–rural character of the battlefield and associated logistical and operational constraints; the two-year-lagged annual average level of GTME; and the logarithms of GDP per capita and population, lagged by one year and drawn from V-Dem (Coppedge et al., 2021), to capture baseline military effectiveness and broader sources of state capacity.

Including lagged average GTME also helps address endogeneity concerns, particularly the possibility that prior battlefield performance influences leaders’ decisions to create heavily armed counterweights. The model's temporal structure preserves causal ordering: annual average GTME measured between t − 3 and t − 2 precedes N. H.A. measured at t − 1, which in turn precedes battle-level GTME measured at t. Using the sample employed to estimate Model Main, Figure 3A in the Online Appendix reports the correlation matrix for the dependent variable and covariates. Descriptive statistics are reported in Table 1A in the Online Appendix.

Figure 3.

Predicted values of government troops’ tactical military effectiveness (GTME) over Distance and N. H.A.—Model Main.

Results

Table 1 reports estimates from the models used to test H1 with standard errors clustered at the government–VNSA dyad level. Model Baseline includes only Distance, N. H.A., and country fixed effects and serves as the baseline specification. Model Naïve extends Model Baseline by adding the interaction between the two variables. Model Main extends Model Naïve by including the full set of covariates discussed in the previous section. The remaining specifications—Model Intensity, Model City, Model GDP, Model Population, and Model Past GTME—each extend Model Naïve by adding one covariate at a time, as each model's name indicates.

Table 1.

The effect of distance and counterbalancing on GTME.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Model Baseline	Model Naïve	Model Intensity	Model City	Model GDP	Model Population	Model Past GTME	Model Main
Variables	GTME	GTME	GTME	GTME	GTME	GTME	GTME	GTME
Distance	−0.024	0.103***	0.112***	0.074*	0.088**	0.105**	0.102***	0.078*
	(0.032)	(0.040)	(0.043)	(0.042)	(0.037)	(0.043)	(0.038)	(0.046)
N. H.A.	−0.108	0.368***	0.348***	0.383***	0.380***	0.454***	0.357***	0.390***
	(0.109)	(0.109)	(0.110)	(0.115)	(0.127)	(0.149)	(0.114)	(0.139)
N. H.A. × Distance		−0.123***	−0.121***	−0.125***	−0.116***	−0.145***	−0.122***	−0.125***
		(0.025)	(0.026)	(0.026)	(0.025)	(0.032)	(0.025)	(0.028)
Local Intensity			0.018***					0.016***
			(0.007)					(0.005)
City				0.097**				0.079*
				(0.046)				(0.042)
GDP					0.799*			0.473
					(0.424)			(0.345)
Population						1.119*		0.280
						(0.611)		(0.551)
Past GTME							−0.086	−0.008
							(0.341)	(0.201)
Constant	0.209	−0.277	−0.344*	−0.371*	−0.867**	−11.645*	−0.219	−3.604
	(0.163)	(0.174)	(0.188)	(0.204)	(0.431)	(6.282)	(0.250)	(5.460)

Observations	9666	9666	9666	9666	9666	9666	9666	9666
Cluster Standard Errors	YES	YES	YES	YES	YES	YES	YES	YES
Country Fixed Effects	YES	YES	YES	YES	YES	YES	YES	YES

Robust standard errors in parentheses. GTME, government troops’ tactical military effectiveness.

*** p < 0.01, ** p < 0.05, * p < 0.1.

The coefficients for Distance and N. H.A. are negative and not statistically significant in Model Baseline, but positive and statistically significant—at least at the 10% level—in all subsequent models. In contrast, the coefficient on the interaction between these two variables is consistently negative and statistically significant at the 1% level across specifications. This pattern accords with the theoretical expectations. Notably, the sign reversal for Distance and N. H.A., together with the fact that the interaction term in Model Main is significant at the 1% level, supports the interpretation that the negative interaction between N. H.A. and Distance is substantively meaningful rather than an artifact of sampling idiosyncrasies (Pepinsky, 2018).

To move the analysis beyond the regression coefficients, Figure 3 visualizes the joint effect of N. H.A. and Distance on GTME by plotting predicted conditional means of GTME across Distance at four representative values of N. H.A. (0, 1, 2, and 4) using estimates from Model Main. Shaded bands denote 95% confidence intervals, and the dashed horizontal line at 0.5 marks where government forces inflict and suffer equal losses. In each panel, the vertical dashed line denotes Distance = 1.8 (approximately 50 km from the capital), and the vertical dotted line denotes Distance = 4.1 (approximately 600 km). As a reference, these points correspond roughly to the distances between Tripoli and Zawiyah and between Tripoli and Bin Jawad, respectively. At the same time, Figure 4 reports the average marginal effect of Distance across the 10 observed levels of N. H.A. As in Figure 3, shaded bands denote 95% confidence intervals. The dashed horizontal line at 0 serves as the reference.

Figure 4.

Predicted values of the average marginal effects of government troops’ tactical military effectiveness (GTME) over Distance and N. H.A.—Model Main.

Looking at Figure 3, when N. H.A. = 0, predicted GTME is modest and weakly increasing, rising from approximately 0.58 at the capital to about 0.74 at Distance = 6. The 95% confidence interval intersects the 0.50 parity threshold only at very low distances (up to roughly 0.75) and remains above parity thereafter, implying that—over most of the support—government forces are expected to inflict at least as many casualties as they sustain. When N. H.A. = 1, the association becomes mildly negative: predicted GTME decreases from about 0.71 near the capital to roughly 0.62 at the upper end of the distance distribution, yet remains statistically above parity throughout. In substantive terms, limited counterbalancing is associated with some degradation in effectiveness with distance, but not sufficient to eliminate the government's loss-exchange advantage.

The pattern shifts markedly as counterbalancing intensifies. At N. H.A. = 2, predicted GTME declines from roughly 0.82 near the capital to about 0.50 at Distance = 6, becoming statistically indistinguishable from parity at approximately Distance ≈ 4.75. At N. H.A. = 4, the decline is strongest: predicted GTME falls from around 0.95 close to the capital to approximately 0.25 at Distance = 6, crossing parity between roughly 3.25 and 5.25 and remaining clearly below it thereafter. These patterns provide direct support for H1. In counterbalanced states, GTME decreases as Distance increases, and the magnitude of this effect grows as the number of heavily armed counterweights rises.

Figure 4 presents the same implication in terms of marginal effects. As N. H.A. increases, the estimated average marginal effect of Distance on predicted GTME becomes progressively more negative, indicating that distance imposes increasing performance penalties as counterbalancing efforts strengthen. For non-counterbalanced countries, the estimated average marginal effect is positive but not statistically distinguishable from zero. For counterbalanced countries, the marginal effect is negative across the observed range and becomes statistically different from zero once the number of counterweights exceeds two. The magnitude of these estimates is substantively consequential: the average marginal effect at N. H.A. = 2 (−0.057) is substantially smaller than at N. H.A. = 4 (approximately −0.27), implying that comparable increases in distance correspond to markedly larger expected declines in GTME in more heavily counterbalanced settings.

Finally, the figures also provide suggestive inductive evidence on two relationships that the theory treats as indeterminate ex ante. Predicted GTME near the capital is higher when multiple counterweight units are present. This is consistent with the possibility that multiple elite, capital-oriented formations increase the effectiveness of government forces where logistical and command-and-control constraints are least binding. Conversely, the slightly positive Distance–GTME association in non-counterbalanced countries is consistent with the view that, absent counterbalancing-induced distortions, some forces can exploit the tactical opportunities associated with remote-area engagements.

To clarify the support in the data across Distance and N. H.A. and facilitate assessment of where the model's inferences are most credible, Figures 5 and 6 display, respectively, the distribution of sample observations over Distance and over N. H.A. In Figure 5, the solid line traces the distribution of observations for the full sample; the dotted line for non-counterbalanced countries; the dashed line for counterbalanced countries, and the dotted-dashed line for countries with more than one counterweight unit.

Figure 5.

Distribution of observations in the sample over Distance—Model Main.

Figure 6.

Distribution of observations in the sample over N. H.A—Model Main.

Robustness checks

To corroborate the analysis, I conduct a series of robustness checks. First, I assess sensitivity to unobserved confounding using the framework developed by Cinelli and Hazlett (2020) and Cinelli et al. (2024), implemented with sensemakr. Because sensemakr does not support fractional response models, I re-estimate Model Main using OLS, with government–VNSA dyad–clustered standard errors obtained via bootstrap, and use this linear specification to compute the sensitivity metrics.

The procedure asks how strong an omitted variable, Z, would have to be—measured by its partial R² with the treatment (conditional on covariates) and with the outcome (conditional on the treatment and covariates)—to reduce the estimated interaction effect to zero. The results, reported in the Online Appendix (Figure 4A), indicate that a confounder would need to be roughly 10 times as powerful as Past GTME—that is, explain about 10 times more of the relevant residual variation in both the treatment and the outcome—to nullify the estimated interaction effect.

This is a demanding benchmark, because the comparison is based on residual variation after partialling out the included controls, and past battlefield performance is plausibly among the strongest predictors of future performance. Importantly, this benchmarking exercise does not assert that such a confounder exists; rather, it provides a transparent yardstick for assessing the plausibility of the results—namely, whether any omitted factor is credibly more predictive than Past GTME in both the treatment and outcome equations. For further details and limits of the test, see Cinelli and Hazlett (2020) and Cinelli et al. (2024).

As a second robustness check, I assess sensitivity to sample idiosyncrasies. Specifically, I re-estimate Model Main 2000 times using a bootstrap that resamples observations at the government–VNSA dyad level. The results are consistent with the main findings and are reported in the Online Appendix. As a third robustness exercise addressing potential sample- and specification-specific idiosyncrasies simultaneously, I re-estimate Model Naïve 2000 times using bootstrapped samples clustered at the government–VNSA dyad level, and—at each iteration—augment it with a random subset of covariates (one to fifteen) drawn from the pool described in section Model covariates and the additional variables listed in Table 2. Results from this analysis are in line with the main findings and fully reported in the Online Appendix.

Table 2.

Pool of additional candidate covariates used to estimate coefficient distributions for the third robustness check.

Variable name	Description	Source
CINC	Composite Index of National Capability. One year lag	National Military Capacity dataset (Singer, 1988)
Military expenditure per soldier	Military expenditure per soldier. One year lag	National Military Capacity dataset (Singer, 1988)
Appointment	Index measuring the extent to which armed-forces appointments are driven by personal or political connections rather than skills and merit. One year lag	V-Dem (Coppedge et al., 2021)
Time since coup	Number of years since the country's last successful military coup—starting from 1946—along with its squared and cubic terms (Carter and Signorino, 2010). Two year lag	Colpus dataset (Chin et al., 2021)
Duration	Number of continuous years of activity of a dyad, along with their squared and cubic terms (Carter and Signorino, 2011)	UCDP Dyadic Dataset (Harbom et al., 2008)
Begin Dyad	Number of years since a dyad first became active	UCDP Dyadic Dataset (Harbom et al., 2008)
Border	Dummy variable indicating whether a country has at least one military unit dedicated to border patrol. One year lag	SSFD (De Bruin, 2021)
CBI index	Number of equally powerful military organizations in a country. One year lag	Pilster and Böhmelt (2011; 2015).
C.W. count	Number of units—heavily armed and not—with specific counterweight tasks. One year lag	SSFD (De Bruin, 2021)
Intel	Dummy variable indicating whether a country has at least one military unit dedicated to internal surveillance. One year lag	SSFD (De Bruin, 2021)
Gov. support	Count of foreign states providing assistance to the government involved in the battle. One year lag	External Support Dataset (Meier et al., 2023 )
VNSA support	Count of foreign states providing assistance to the VNSA involved in the battle. One year lag	External Support Dataset (Meier et al., 2023 )
Corruption	Regime corruption index. One year lag	V-Dem (Coppedge et al., 2021)
Democracy	Liberal Democracy Index. One year lag	V-Dem (Coppedge et al., 2021)
Number of Actors	Number of active VNSAs in the country where the battle occurred. One month lag.	UCDP GED (Sundberg and Melander, 2013)
Country Intensity	Number of battles occurred nationwide. One month lag	UCDP GED (Sundberg and Melander, 2013)
Coast distance	Distance from the battle location to the country's nearest coastline
Border distance	Distance from the battle location to the country's nearest border
Island	Dummy variable assuming value 1 if the battle occurred on an island and 0 otherwise
Elevation	Elevation of the battle

SSFD, State Security Forces Dataset; UCDP GED, Uppsala Conflict Data Program Georeferenced Event Dataset; VNSA, violent non-state actor.

As a fourth step, I re-estimate Model Main using nine alternative specifications and sampling criteria (Models Main 1–9). These models, described in full and reported in the Online Appendix (Table 2A), yield results consistent with the baseline estimates. As a final robustness exercise, I re-estimate Model Main by adding to it the additional covariates from Table 2 in thematically coherent blocks and by estimating specifications that include government–VNSA dyad fixed effects and year fixed effects, as well as a specification that omits country fixed effects (Models A–O). These estimates likewise remain consistent with the main findings and are reported in the Online Appendix (Table 3A).

Conclusion

In this article, I argue that as counterbalancing intensifies, the negative effect of distance from the capital on GTME becomes more pronounced. This argument rests on three considerations. First, leaders who engage in counterbalancing concentrate the best trained, best equipped, and most motivated personnel in elite formations near the capital. Consequently, battles fought farther from the capital are more likely to involve units with limited capacity to generate combat power. Second, distance increases the demands of counterinsurgency, while counterbalancing exacerbates these challenges by undermining coordination, keeping elite formations near the regime center, and diverting attention toward internal threat monitoring. Third, as counterbalancing intensifies, these mechanisms compound, producing a steeper spatial gradient in combat performance.

To test these claims, I combine data from the SSFD (De Bruin, 2021) with data from the UCDP GED (Sundberg and Melander, 2013) and estimate fractional regression models on battle-level observations in intrastate conflicts between 1991 and 2010. Across specifications, the extent to which distance from the capital erodes GTME increases with higher levels of counterbalancing. Counterbalancing may, however, confer limited advantages in engagements near the capital. Robustness checks support these findings.

These results have several implications. First, they introduce a spatial gradient linking institutional coup proofing to battlefield outcomes. Second, they provide microfoundations—loyalty, selective training, and morale—that connect elite-survival strategies to performance against VNSAs. In other words, they show how leaders may secure resilience at the center, but typically at the cost of steeper capability decline in the periphery.

This analysis also has limitations. It relies on the assumption that battles closer to the capital are more likely to involve counterweight units, and casualty-based measures cannot fully capture operational success. These constraints point to the need for future research using unit-level deployment data and more direct measures of force employment.

Supplemental Material

sj-docx-1-cmp-10.1177_07388942261447909 - Supplemental material for Capital gains: Revising the impact of coup proofing on tactical military effectiveness

Supplemental material, sj-docx-1-cmp-10.1177_07388942261447909 for Capital gains: Revising the impact of coup proofing on tactical military effectiveness by Marco Nicola Binetti in Conflict Management and Peace Science

Supplemental Material

sj-do-2-cmp-10.1177_07388942261447909 - Supplemental material for Capital gains: Revising the impact of coup proofing on tactical military effectiveness

Supplemental material, sj-do-2-cmp-10.1177_07388942261447909 for Capital gains: Revising the impact of coup proofing on tactical military effectiveness by Marco Nicola Binetti in Conflict Management and Peace Science

Supplemental Material

sj-dta-3-cmp-10.1177_07388942261447909 - Supplemental material for Capital gains: Revising the impact of coup proofing on tactical military effectiveness

Supplemental material, sj-dta-3-cmp-10.1177_07388942261447909 for Capital gains: Revising the impact of coup proofing on tactical military effectiveness by Marco Nicola Binetti in Conflict Management and Peace Science

Supplemental Material

sj-dta-4-cmp-10.1177_07388942261447909 - Supplemental material for Capital gains: Revising the impact of coup proofing on tactical military effectiveness

Supplemental material, sj-dta-4-cmp-10.1177_07388942261447909 for Capital gains: Revising the impact of coup proofing on tactical military effectiveness by Marco Nicola Binetti in Conflict Management and Peace Science

Supplemental Material

sj-csv-5-cmp-10.1177_07388942261447909 - Supplemental material for Capital gains: Revising the impact of coup proofing on tactical military effectiveness

Supplemental material, sj-csv-5-cmp-10.1177_07388942261447909 for Capital gains: Revising the impact of coup proofing on tactical military effectiveness by Marco Nicola Binetti in Conflict Management and Peace Science

Supplemental Material

sj-rmd-6-cmp-10.1177_07388942261447909 - Supplemental material for Capital gains: Revising the impact of coup proofing on tactical military effectiveness

Supplemental material, sj-rmd-6-cmp-10.1177_07388942261447909 for Capital gains: Revising the impact of coup proofing on tactical military effectiveness by Marco Nicola Binetti in Conflict Management and Peace Science

Supplemental Material

sj-txt-7-cmp-10.1177_07388942261447909 - Supplemental material for Capital gains: Revising the impact of coup proofing on tactical military effectiveness

Supplemental material, sj-txt-7-cmp-10.1177_07388942261447909 for Capital gains: Revising the impact of coup proofing on tactical military effectiveness by Marco Nicola Binetti in Conflict Management and Peace Science

Footnotes

Acknowledgement

The author would like to thank the reviewers and editors for their constructive comments and guidance, which helped improve the paper. Parts of this research were conducted while the author was a researcher at the Center for Crisis Early Warning (Kompetenzzentrum Krisenfrüherkennung). The Center for Crisis Early Warning is funded by the German Federal Ministry of Defense and the German Federal Foreign Office. The views and opinions expressed in this article are those of the author and do not necessarily reflect the official policy or position of any agency of the German government.

ORCID iD

Marco Nicola Binetti

Funding

The author received no financial support for the research, authorship, and/or publication of this article.

Supplemental material

Supplemental material for this article is available online.

Notes

References

Albrecht

(2015) Does coup-proofing work? Political–military relations in authoritarian regimes amid the arab uprisings. Mediterranean Politics 20(1): 36–54.

Albrecht

Eibl

(2018) How to keep officers in the barracks: Causes, agents, and types of military coups. International Studies Quarterly 62(2): 315–328.

Arbuckle

(1979) Rhodesian bush war strategies and tactics: An assessment. The RUSI Journal 124(4): 27–33.

Asher

Hammel

(1987) Duel for the Golan: The 100-Hour Battle that Saved Israel. New York, NY: William Morrow.

Askerov

(2015) Historical Dictionary of the Chechen Conflict. Lanham, MD: Rowman & Littlefield.

Baxter

(2014) Bush War Rhodesia: 1966-1980. Solihull: Helion & Company Limited.

Beckley

(2010) Economic development and military effectiveness. The Journal of Strategic Studies 33(1): 43–79.

Belkin

Schofer

(2003) Toward a structural understanding of coup risk. Journal of Conflict Resolution 47(5): 594–620.

Belkin

Schofer

(2005) Coup risk, counterbalancing, and international conflict. Security Studies 14(1): 140–177.

10.

Bell

Witter

(2011 The Libyan Revolution—Roots of Rebellion. Washington, D.C: Institute for the Study of War. https://understandingwar.org/research/middle-east/the-libyan-revolution-part-i-roots-of-rebellion/ .

11.

Bensahel

Byman

(2004) The Future Security Environment in the Middle East: Conflict, Stability, and Political Change. Santa Monica, CA: Rand Corporation.

12.

Berman

Matanock

(2015) The empiricists’ insurgency. Annual Review of Political Science 18: 443–464.

13.

Biddle

(2010) Military Power. Princeton, NJ: Princeton University Press.

14.

Biddle

(2022) Nonstate Warfare: The Military Methods of Guerillas, Warlords, and Militias. Princeton, NJ: Princeton University Press.

15.

Biddle

Long

(2004) Democracy and military effectiveness: A deeper Look. Journal of Conflict Resolution 48(4): 525–546.

16.

Binetti

(2024) Far from home: The impact of corruption on tactical military effectiveness. Journal of Global Security Studies 9(4): ogae044.

17.

Binetti

(2025) Assessing the impact of non-military aid on tactical military effectiveness: The case of Iraq between 2004 and 2014. International Interactions 51(5): 822–849.

18.

Binetti

Steinwand

(2025) Aid targeting in post-conflict Nepal. Conflict Management and Peace Science: 1–21.

19.

Black

(2021) Logistics: The Key to Victory. Barnsley: Pen; Sword Military.

20.

Böhmelt

Dworschak

Pilster

, et al. (2020) A cross-national analysis of forced population resettlement in counterinsurgency campaigns. Peace Economics, Peace Science and Public Policy 26(1): 20190022.

21.

Böhmelt

Pilster

(2015) The impact of institutional coup-proofing on coup attempts and coup outcomes. International Interactions 41(1): 158–182.

22.

Boisvert

M-A

(2016) Forces armées maliennes, une lente reconstruction. Afrique Contemporaine 260(4): 87–90.

23.

Boulding

(1963) Conflict and Defense: A General Theory. New York, NY: Harper.

24.

Brooks

(2006) An autocracy at war: Explaining Egypt’s military effectiveness, 1967 and 1973. Security Studies 15(3): 396–430.

25.

Brooks

(2007) Creating Military Power: The Sources of Military Effectiveness. Stanford, CA: Stanford University Press.

26.

Buchanan

Tullock

(1965) The Calculus of Consent: Logical Foundations of Constitutional Democracy. Vol. 100. Ann Arbor, MA: University of Michigan Press.

27.

Bueno de Mesquita

Smith

(2017) Political succession: A model of coups, revolution, purges, and everyday politics. Journal of Conflict Resolution 61(4): 707–743.

28.

Carter

Signorino

(2010) Back to the future: Modeling time dependence in binary data. Political Analysis 18(3): 271–292.

29.

Chamberlin

(1974) Provision of collective goods as a function of group size. American Political Science Review 68(2): 707–716.

30.

Chin

Carter

Wright

(2021) The varieties of coups d’état: Introducing the Colpus dataset. International Studies Quarterly 65(4): 1040–1051.

31.

Cinelli

Ferwerda

Hazlett

(2024) Sensemakr: Sensitivity analysis tools for OLS in R and Stata. Observational Studies 10(2): 93–127.

32.

Cinelli

Hazlett

(2020) Making sense of sensitivity: Extending omitted variable bias. Journal of the Royal Statistical Society Series B: Statistical Methodology 82(1): 39–67.

33.

Coppedge

Gerring

Knutsen

, et al. (2021) V-Dem Dataset V11. 1 - https://www.v-dem.net/data/the-v-dem-dataset/.

34.

Cordesman

Nerguizian

(2009) The North African Military Balance: Force Developments in the Maghreb. Vol. 31. Washington, DC: CSIS.

35.

Croicu

Sundberg

(2015) UCDP Georeferenced event dataset codebook version 4.0. Journal of Peace Research 50(4): 523–532.

36.

Dahl

(2016) Military defection. PRIO Policy Brief 18(1): 1–4.

37.

D’Amato

(2012) Byzantine Imperial Guardsmen 925–1025: The Tághmata and Imperial Guard. Vol. 187. Oxford: Osprey Publishing.

38.

De Atkine

(1999) Why Arabs lose wars. Middle East Quarterly 6(1): 1–2.

39.

De Bruin

(2018) Preventing coups d’état: How counterbalancing works. Journal of Conflict Resolution 62(7): 1433–1458.

40.

De Bruin

(2020) How to Prevent Coups d’État: Counterbalancing and Regime Survival. Ithaca, NY: Cornell University Press.

41.

De Bruin

(2021) Mapping coercive institutions: The state security forces dataset, 1960–2010. Journal of Peace Research 58(2): 315–325.

42.

De Bruin

(2022) Policing insurgency: Are more militarized police more effective? Small Wars & Insurgencies 33(4–5): 742–766.

43.

Desch

(2002) Democracy and victory: Why regime type hardly matters. International Security 27(2): 5–47.

44.

Dixon

(2009) ‘Hearts and minds?’ British counter-insurgency from Malaya to Iraq. Journal of Strategic Studies 32(3): 353–381.

45.

Downes

(2007) Draining the sea by filling the graves: Investigating the effectiveness of indiscriminate violence as a counterinsurgency strategy. Civil Wars 9(4): 420–444.

46.

Drea

(2016) Japan’s Imperial Army: Its Rise and Fall, 1853–1945. Lawrence, KA: University Press of Kansas.

47.

Dunigan

(2020) Victory for Hire: Private Security Companies’ Impact on Military Effectiveness. Redwood City, CA: Stanford University Press.

48.

Dupuy

(1987) Understanding War: History and a Theory of Combat. New York, NY: Paragon.

49.

Dworschak

(2020) Jumping on the bandwagon: Differentiation and security defection during conflict. Journal of Conflict Resolution 64(7–8): 1335–1357.

50.

Dworschak

(2024) Bias mitigation in empirical peace and conflict studies: A short primer on posttreatment variables. Journal of Peace Research 61(3): 462–476.

51.

Easton

Siverson

(2018) Leader survival and purges after a failed coup d’état. Journal of Peace Research 55(5): 596–608.

52.

Eccles

(1989) Logistics in the National Defense. Washington, DC: US Marine Corps.

53.

Eck

(2015) Repression by proxy: How military purges and insurgency impact the delegation of coercion. Journal of Conflict Resolution 59(5): 924–946.

54.

Egnell

(2010) Winning ‘hearts and minds?’ A critical analysis of counter-insurgency operations in Afghanistan. Civil Wars 12(3): 282–303.

55.

Enloe

(1975) The military uses of ethnicity. Millennium 4(3): 220–234.

56.

Esen

Gumuscu

(2017) Turkey: How the coup failed. Journal of Democracy 28(1): 59–73.

57.

Foxton

(1994) Powering War: Modern Land Force Logistics. Vol. 11. London: Brassey’s.

58.

Galula

(2006) Counterinsurgency Warfare: Theory and Practice. Westport, CT: Praeger Security International.

59.

Gartner

Myers

(1995) Body counts and “success” in the Vietnam and Korean wars. The Journal of Interdisciplinary History 25(3): 377–395.

60.

Geddes

Wright

, et al. (2018) How Dictatorships Work: Power, Personalization, and Collapse. Cambridge: Cambridge University Press.

61.

Gentil-Fernandes

Otto

(2024) Corrupting the battlefield: How corruption influences belligerents’ battlefield performance. International Interactions 50(5): 809–838.

62.

Glantz

(1989) Attache assessments of the impact of the 1930s purges on the Red Army. The Journal of Soviet Military Studies 2(3): 417–438.

63.

Goldsworthy

(1981) Civilian control of the military in Black Africa. African Affairs 80(318): 49–74.

64.

Gropman

(1997) The BIG’l’american Logistics in World War II. Washington: National Defense University, Institute for National Strategic Studies.

65.

Harbom

Melander

Wallensteen

(2008) Dyadic dimensions of armed conflict, 1946—2007. Journal of Peace Research 45(5): 697–710.

66.

Harkness

(2022) The ethnic stacking in Africa dataset: When leaders use ascriptive identity to build military loyalty. Conflict Management and Peace Science 39(5): 609–632.

67.

Hart

(1996) Montgomery, morale, casualty conservation and ‘colossal cracks’: 21st army group’s operational technique in north-west Europe, 1944–45. Journal of Strategic Studies 19(4): 132–153.

68.

Head

(2013) The battles of al-Fallujah: Urban warfare and the growth of air power. Air Power History 60(4): 32–51.

69.

Hidaka

Kubota

Yukawa

(2025) Life after exile: Introducing a new dataset on post-exile fate. Conflict Management and Peace Science 42(6): 726–741.

70.

Hlatky

Landry

(2025) Ethnic preferences, domestic audiences and military coalition formation. Conflict Management and Peace Science 42(1): 32–55.

71.

Hoekstra

(2020) How Mosul fell: The role of coup-proofing in the 2014 partial collapse of the Iraqi security forces. International Politics 57(4): 684–703.

72.

House

(1984) Towards Combined Arms Warfare: A Survey of Tactics, Doctrine, and Organization in the 20th Century. Leavenworth, KS: U.S. Army; Command and General Staff College.

73.

Jones

(1994) Military Psychiatry: Preparing in Peace for War. Vol. 1. Washington, DC: Office of the Surgeon General, US Department of the Army.

74.

Kalyvas

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

75.

Kalyvas

Kocher

(2009) The dynamics of violence in Vietnam: An analysis of the hamlet evaluation system (HES). Journal of Peace Research 46(3): 335–355.

76.

Kane

(2001) Military Logistics and Strategic Performance. Vol. 1. Abingdon: Routledge.

77.

Kheng

(1991) Malaysia—Hearts and minds in guerrilla warfare: The Malayan Emergency, 1948–1960. By Richard Stubbs. Singapore: Oxford University Press, 1989. Pp. Xiv, 286. map, bibliography, index. Journal of Southeast Asian Studies 22(2): 427–430.

78.

Kocher

Pepinsky

Kalyvas

(2011) Aerial bombing and counterinsurgency in the Vietnam war. American Journal of Political Science 55(2): 201–218.

79.

Kugler

Baranick

Binnendijk

(2009) Operation Anaconda: Lessons for Joint Operations. Washington, DC: Center for Technology, National Security Policy, National Defense University.

80.

Lutscher

(2016) The more fragmented the better?—The impact of armed forces structure on defection during nonviolent popular uprisings. International Interactions 42(2): 350–375.

81.

Lyall

(2010) Do democracies make inferior counterinsurgents? Reassessing democracy’s impact on war outcomes and duration. International Organization 64(1): 167–192.

82.

Lyall

(2020) Divided Armies: Inequality and Battlefield Performance in Modern War. Princeton, NJ: Princeton University Press.

83.

Lyall

Wilson

(2009) Rage against the machines: Explaining outcomes in counterinsurgency wars. International Organization 63(1): 67–106.

84.

Marcum

Brown

(2017) Is more institutional coup-proofing better or worse for regime protection? Evidence from the Philippines, 1986–1987. Asian Journal of Political Science 25(1): 109–129.

85.

Meier

Karlén

Pettersson

, et al. (2023) External support in armed conflicts: Introducing the UCDP External Support Dataset (ESD), 1975–2017. Journal of Peace Research 60(3): 545–554.

86.

Millett

Murray

Watman

(1986) The effectiveness of military organizations. International Security 11(1): 37–71.

87.

Mills

Wilson

(2007) Who dares loses? Assessing Rhodesia’s counter-insurgency experience. The RUSI Journal 152(6): 22–31.

88.

Moore

DMB

Antill

Peter

(1999) The role of logistics in modern warfare: Operations Desert Shield and Desert Storm. RMCS Research Paper 14.

89.

Muckensturm

Longhorn

(2019) Assessing the vulnerability of military theater distribution routes. Journal of Defense Analytics and Logistics 3(1): 60–82.

90.

Nagl

Amos

Sewall

, et al. (2008) The US Army/Marine Corps Counterinsurgency Field Manual. Chicago, IL: University of Chicago Press.

91.

Olson

Jr (1971) The Logic of Collective Action: Public Goods and the Theory of Groups, with a New Preface and Appendix. Vol. 124. Cambridge, MA: Harvard University Press.

92.

Oswald

Walterskirchen

Häffner

, et al. (2024) The night before or the morning after: Coup risk, failed coup attempts, and minister removals in autocracies. OSF: 1–12.

93.

Pascal

Kennedy

Rosen

, et al. (1979) Men and Arms in the Middle East: The Human Factor in Military Modernization. Santa Monica, CA: RAND Corporation.

94.

Pepinsky

(2018) Visual heuristics for marginal effects plots. Research & Politics 5(1): 2053168018756668.

95.

Petersen

(2001) Resistance and Rebellion: Lessons from Eastern Europe. New York, NY: Cambridge University Press.

96.

Petersohn

(2017) Private military and security companies (PMSCs), military effectiveness, and conflict severity in weak states, 1990–2007. Journal of Conflict Resolution 61(5): 1046–1072.

97.

Phillips

(2011) The Fourth Crusade and the Sack of Constantinople. London: Pimlico.

98.

Pilster

Böhmelt

(2011) Coup-Proofing and military effectiveness in interstate wars, 1967–99. Conflict Management and Peace Science 28(4): 331–350.

99.

Pollack

(2004) Arabs at War: Military Effectiveness, 1948–1991. Lincoln, NB: University of Nebraska Press.

100.

Polman

(2010a) War Games: The Story of Aid and War in Modern Times. London: Penguin.

101.

Polman

(2010b) The Crisis Caravan: What’s Wrong with Humanitarian Aid? New York, NY: Metropolitan Books.

102.

Posen

(1986) The Sources of Military Doctrine: France, Britain, and Germany Between the World Wars. Ithaca, NY: Cornell University Press.

103.

Powell

(2019) Leader survival strategies and the onset of civil conflict: A coup-proofing paradox. Armed Forces & Society 45(1): 27–44.

104.

Powell

Faulkner

Dean

, et al. (2018) Give them toys? Military allocations and regime stability in transitional democracies. Democratization 25(7): 1153–1172.

105.

Powell

(2014) Trading coups for civil war: The strategic logic of tolerating rebellion. African Security Review 23(4): 329–338.

106.

Quinlivan

(1999) Coup-proofing: Its practice and consequences in the Middle East. International Security 24(2): 131–165.

107.

Rabinowitz

Jargowsky

(2018) Rethinking coup risk: Rural coalitions and coup-proofing in sub-saharan Africa. Armed Forces & Society 44(2): 322–346.

108.

Reese

(1992) The impact of the Great Purge on the Red Army: Wrestling with hard numbers. The Soviet and Post-Soviet Review 19(1): 71–90.

109.

Reese

(2011) Why Stalin’s Soldiers Fought: The Red Army’s Military Effectiveness in World War II. Lawrence, KS: University Press of Kansas.

110.

Reiter

Stam

III (1998) Democracy and battlefield military effectiveness. Journal of Conflict Resolution 42(3): 259–277.

111.

Reiter

Wagstaff

(2018) Leadership and military effectiveness. Foreign Policy Analysis 14(4): 490–511.

112.

Roberts

(1975) Civil resistance to military coups. Journal of Peace Research 12(1): 19–36.

113.

Rozenas

Zeigler

(2019) From ballot-boxes to barracks: Votes, institutions, and post-election coups. Journal of Peace Research 56(2): 175–189.

114.

Schrodt

(2014) Seven deadly sins of contemporary quantitative political analysis. Journal of Peace Research 51(2): 287–300.

115.

Scott

(2006) The Battle for the Golan Heights in the Yom Kippur War of 1973: A battle analysis. https://trace.tennessee.edu/handle/20.500.14382/9223 .

116.

Sexton

Zürcher

(2024) Aid, attitudes, and insurgency: Evidence from development projects in northern Afghanistan. American Journal of Political Science 68(3): 1168–1182.

117.

Singer

(1988) Reconstructing the Correlates of War dataset on material capabilities of states, 1816–1985. International Interactions 14(2): 115–132.

118.

Skoglund

Listou

Ekström

(2022) Russian logistics in the Ukrainian war: Can operational failures be attributed to logistics? Scandinavian Journal of Military Studies 5(1): 99–110.

119.

Sudduth

(2017a) Coup risk, coup-proofing and leader survival. Journal of Peace Research 54(1): 3–15.

120.

Sudduth

(2017b) Strategic logic of elite purges in dictatorships. Comparative Political Studies 50(13): 1768–1801.

121.

Sundberg

Melander

(2013) Introducing the UCDP georeferenced event dataset. Journal of Peace Research 50(4): 523–532.

122.

Talmadge

(2015) The Dictator’s Army: Battlefield Effectiveness in Authoritarian Regimes. Ithaca, NY: Cornell University Press.

123.

Thoral

M-C

(2011) From Valmy to Waterloo: France at War, 1792–1815. New York: Palgrave Macmillan, 99–120.

124.

US Army and US Marine Corps (2014) Field Manual: FM 3-24/MCWP 3-33.5. Washington, DC: Marine Corps Warfighting Publications.

125.

Van Creveld

(1977) “Supplying War.” Logistics from Wallenstein to Patton. Cambridge: Cambridge University Press.

126.

Van Creveld

(1985) Command in War. Cambridge, MA: Harvard University Press.

127.

Vesco

Randahl

Hegre

, et al. (2026) The underreported death toll of wars: A probabilistic reassessment from a survey with UCDP coders. Journal of Conflict Resolution: 1–23.

128.

Warrington

(1998) ‘Standing to arms in Lilliput’—The armed forces, external relations and domestic politics in a micro-state: Malta, 1965–1997. Public Administration and Development: The International Journal of Management Research and Practice 18(2): 185–199.

129.

Waters

(2018) Syria’s Republican Guard. Washington, DC: The Middle East Institute.

130.

Watson

(2008) Enduring the Great War: Combat, Morale and Collapse in the German and British Armies, 1914–1918. New York, NY: Cambridge University Press.

131.

Weidmann

(2015) On the accuracy of Media-based conflict event data. Journal of Conflict Resolution 59(6): 1129–1149.

132.

Wood

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

133.

Wood

JRT

(2009) Counter-strike from the Sky: The Rhodesian All-arms Fireforce in the War in the Bush 1974–1980. Solihull: Helion & Company Limited.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

13.61 MB

0.03 MB

0.02 MB

50.86 MB

0.00 MB

86.17 MB

1.88 MB

0.00 MB