Armies sometimes use shocking brutality to force reluctant soldiers to fight. In Ukraine, Russia's armed forces and associated paramilitary formations have reportedly used artillery fire against surrendering soldiers;Footnote 1 tortured and mock-executed junior officers;Footnote 2 imprisoned men in open-air pits for disciplinary infractions;Footnote 3 denied medical assistance to wounded soldiers;Footnote 4 executed deserters with sledgehammers;Footnote 5 and threatened to shoot retreating soldiers during “meat storm” frontal assaults.Footnote 6 “They placed barrier troops behind us,” said one soldier, “and they weren't letting us leave our position.”Footnote 7 Another soldier declared, “If we go back, they'll shoot us.”Footnote 8
These anecdotes are not historical outliers. Success in battle, and states’ ability to win wars and engage in coercive diplomacy, rest on individual soldiers’ willingness to fight. Armies turn to fratricidal coercion—the threat or use of violence against reluctant soldiers—to make outside options, like fleeing or surrendering, less appealing.
How fratricidal coercion affects soldiers’ battlefield behavior remains a subject of debate.Footnote 9 Recent quantitative research has exposed the historical ubiquity of this practice, but has not directly explored its behavioral impact on soldiers. Nearly all empirical work has focused on prewar, not wartime, state coercion.Footnote 10 For example, at the cross-national level, Lyall shows that states whose ethnic minorities experienced greater prewar discrimination fared more poorly in battle.Footnote 11 At the sub-national level, Rozenas, Talibova, and Zhukov (hereafter RTZ) show that Soviet soldiers exposed to greater political repression before World War II were more likely to comply with orders, but also less likely to show initiative.Footnote 12 We know surprisingly little about how wartime fratricidal coercion affects battlefield behavior.
Our research note builds on this prior work by incorporating fratricidal coercion into empirical models of combat motivation in conventional, interstate war. We use declassified Soviet army personnel records and divisional histories from World War II (1941–1945) to explore how the presence of People's Commissariat of Internal Affairs (Narodnyy komissariat vnutrennikh del or NKVD) “Special Sections” affected battlefield performance across military units, including casualties, disappearances, and desertion. These tactical and operational dynamics, as we show, are not just important in and of themselves—they are predictive of strategic-level victory and defeat in war. We study how prewar mass repression and officer purges might have amplified or attenuated these wartime incentives. We then consider how unique the Soviet experience was in a cross-national context, using data on blocking units across 526 battles in 75 wars (1939–2011).
Consistent with Lyall's and RTZ's findings on prewar discrimination and repression, we find that fratricidal coercion bolsters soldier compliance but does not improve—and, in key ways, worsens—battlefield performance. Red Army divisions with larger NKVD contingents witnessed less soldier indiscipline (fewer disappearances, desertions, defections, and surrenders) but also higher casualties and fewer medals for valor—an indicator of soldier initiative. We supplement these analyses with matched qualitative comparisons of Soviet divisions and find that larger NKVD sections may have helped drive reluctant soldiers forward, resulting in higher death rates but fewer medals for bravery. We also find that wartime and prewar coercion affected soldiers’ behavior in similar, mutually reinforcing ways. Fratricidal coercion made a greater difference in units whose soldiers were more exposed to prewar repression (a claim RTZ make, but do not test), while wartime coercion may have enhanced the coercive effects of prewar repression. Our cross-national tests yield similar patterns: blocking units are associated with fewer missing soldiers but also higher casualties and worse loss-exchange ratios. We further demonstrate that belligerents employing fratricidal coercion are less likely to win wars.
Fratricidal Coercion and Combat Motivation
We define fratricidal coercion as the threat or use of physical violence by military authorities and their representatives against their own soldiers in wartime. Designed to deter and punish desertion and other behavioral transgressions, fratricidal coercion differs in three ways from other forms of violence within military organizations.
First, fratricidal coercion is intentional, unlike “friendly fire” or (noncoercive) “fratricide,” which denote accidental actions.Footnote 13 Second, fratricidal coercion is top-down (carried out by formal authorities), unlike bottom-up “fragging” actions by troops against commanding officers. Third, it is extrajudicial, sidestepping or replacing routine procedures of military justice.
Fratricidal coercion can include lethal and nonlethal measures, from publicly executing soldiers to forcibly returning them to the front, or simply establishing a visible deterrent presence. In some instances, these coercive actions are the responsibility of specialized “blocking detachments” with separate recruitment procedures. These units station themselves behind front-line positions and patrol rear areas to prevent desertion or surrender. Not all efforts reach this level of sophistication, and desperate commanders frequently use ad hoc measures to push reluctant soldiers forward and cauterize the flow of deserters. We focus here on blocking detachments as the clearest example of fratricidal coercion.
Following Lyall and RTZ, we assume that soldier resolve is variable in wartime.Footnote 14 Resolve depends on intrinsic factors (such as personal duty, honor, or ideology) and extrinsic ones (such as rewards and punishment). Some considerations may predate war, like exposure to state-sanctioned discrimination or repression.Footnote 15 Others stem from wartime fighting conditions. Fratricidal coercion offers a potential means to strengthen soldiers’ extrinsic motivations to fight when intrinsic motivation is low.
We argue that the effect of fratricidal coercion on resolve is not uniform. Expectations of brutality can deter intrinsically reluctant soldiers from fleeing by updating their beliefs about the consequences of failing to fight (“shirkers will be punished”) and the choice they face (“I am better off not shirking”). However, fratricidal coercion may also compel otherwise eager, intrinsically resolved soldiers to dampen their zeal—on the expectation that any perceived deviations from norms of appropriate conduct (including overperformance) could be punished. Coercion, then, makes soldiers’ combat resolve more uniform: malcontents become more compliant with orders, and true believers are less likely to exceed orders.
Observable Implications
We hypothesize that, as fratricidal coercion increases in a unit or army, fewer soldiers will shirk their duties (H1), more soldiers will follow orders (H2), and fewer soldiers will take initiative beyond these orders (H3).
Since we cannot observe these outcomes directly, we follow the literature in viewing different categories of casualties as empirical realizations of resolve.Footnote 16 As a proxy for shirking, we look at how many soldiers went missing in action (MIA), deserted, or surrendered. Such cases should decline as fratricidal coercion forecloses escape opportunities. As a proxy for following orders, we use numbers of killed (KIA) or wounded in action (WIA). Because combat inflicts physical trauma, commitment to one's combat mission implies a tacit willingness to risk life and limb. As a proxy for initiative, we look at how many soldiers received medals for valor. We consider the validity of these measures later.
Finally, we expect the behavioral effects of wartime coercion to be similar to those of prewar repression: both should lead to more homogeneous behavior by actors with variable preferences toward fighting (H4). While RTZ do not study fratricidal coercion, they speculate that “soldiers who had experienced state violence more intimately as civilians may be more responsive to coercive measures on the battlefield.”Footnote 17 We investigate the empirical basis for this claim later.
Empirical Strategy
We adopt a two-part mixed-methods empirical strategy. First, we draw on microlevel data to explain how fratricidal coercion affected battlefield outcomes across combat units in the same army. Second, we assess the cross-national generalizability of our claims by studying how blocking detachments shaped the outcomes of 526 land battles in seventy-five wars since 1939.
Subnational Evidence: The Red Army in World War II
We begin our investigation by creating a new unit-level data set on fratricidal coercion within the Soviet Workers’ and Peasants’ Red Army (RKKA) during World War II. Our data set tracks 1,048 RKKA Divisions over forty-eight months (June 1941 to May 1945), including all active formations that participated in combat, and excluding training and reserve divisions. Rifle Divisions (infantry) represent 78 percent of these observations.Footnote 18
Each division (8,000 to 12,000 troops, on average) reported to an army—a combined arms unit comprising three to five divisions, along with air defense, artillery, reconnaissance and other supporting units. In wartime, armies reported to fronts, each containing three to five armies. These nestings shifted during World War II, with armies reassigned from one front to another, and divisions transferring between armies. Unit designations were not unique, as the high command regularly disbanded, reorganized, renamed, and renumbered its divisions. Given this complexity, we treat each division–army nesting as a separate, unique unit. Since units saw combat at different stages of the war—and virtually none were active for all forty-eight months—our data set is an unbalanced panel of 21,241 division-months. We have information on combat operations for 16,330 division-months (77%).Footnote 19
We measure our dependent variable, battlefield performance, using RTZ's data on 34 million RKKA soldiers who served in World War II. These data integrate 105 million personnel records from the Russian Ministry of Defense's “People's Memory” archive,Footnote 20 including information on promotions and decorations and, central for our purposes, each soldier's fate (discharge, transfer, or death). We have complete personnel records for 8,483,491 soldiers, including unit names, dates, and reasons for discharge, allowing us to match records to specific divisions and months.Footnote 21 We matched soldiers to units and calculated the proportion of each division's monthly losses attributable to death, injury, MIA, capture, desertion, and punishment for misconduct.Footnote 22 To measure initiative in battle, we calculated the proportion of each division's personnel that received a valor decoration each month.Footnote 23 Later in this note, we provide cross-national evidence that these tactical-level dynamics are predictive of strategic-level victory and defeat in war.
To assess the impact of fratricidal coercion on battlefield outcomes, we collected data on NKVD personnel who served in Special Sections (OO) and SMERSH counterintelligence units. These units were embedded in the regular army and had authority to bypass military tribunals to detain and punish deserters and stragglers. OOs were active from the start of the war. Their duties intensified after September 1941, when Stalin ordered that blocking companies be organized in all Rifle regiments. Regular soldiers staffed these companies. OO officers commanded them. Their mission was to patrol rear areas and “liquidate instigators of panic and flight.”Footnote 24
Although most primary sources on the actions of blocking units remain classified,Footnote 25 we can measure the numerical presence of NKVD officers in each division. Our assumption is that units with a larger counterintelligence presence saw more intense monitoring and enforcement. Our data sources are service histories for 41,383 NKVD officers, compiled by Memorial, a Russian human rights NGO.Footnote 26 We identified 25,079 NKVD officers who served in OO or SMERSH during the war, noting the combat units to which they were assigned, and when.
The number of OO/SMERSH personnel per division-month ranged from 0 to 243 (303rd Rifle Division, 7th Guards Army, 2nd Ukrainian Front, November 1943), with a mean of ten officers. This number excludes rank-and-file troops who served in blocking companies under these officers’ command (roughly 100 soldiers each). On average, there was one OO/SMERSH officer for every 1,376 troops.
NKVD officers had other duties in the army beyond fratricidal coercion, like political education and surveillance. Our measure therefore excludes political officers responsible for ideological training (politruk or zampolit) and includes only those tasked with identifying and punishing disloyal soldiers (OO/SMERSH).
The data reveal significant variation in NKVD presence across army units, which does not always correspond to operational tempo. For example, almost three times as many OO/SMERSH officers rotated through the 1st Ukrainian Front as had served in the 3rd Ukrainian Front (a difference of over two standard deviations), although these units participated in a similar number of battles (416 and 339, under 0.33 standard deviations apart) over the same period (see section A1 of the online supplement).
What explains this variation in NKVD presence? Supplementary analyses (section A1) suggest that the NKVD assigned more personnel to units where they expected higher rates of flight. First, there were more NKVD officers in units with more opportunities for contact with the enemy and crossing of front lines (for example, infantry). Second, units with more soldiers from “politically suspect” backgrounds (such as minorities, peasants, or older soldiers with longer exposure to prerevolutionary institutions) had more NKVD officers. Third, NKVD presence grew over time, peaking when the RKKA was deep in Germany in 1945. Finally, the NKVD sent more officers to units whose soldiers were exposed to more prewar repression—as we discuss later.
Statistical Analysis of NKVD Presence and Soviet Performance
Did fratricidal coercion matter for Soviet battlefield performance? Figure 1 reports estimates of the effect of NKVD presence on seven types of battlefield outcomes. Each line reports a coefficient estimate and 95 percent confidence interval from a separate three-way fixed effects model, regressing the percentage of a division's monthly losses (KIA, WIA, MIA, POW, desertion, and punishment) and medals on the number of OO/SMERSH personnel assigned to the unit at that time (see section A2 of the online supplement).Footnote 27 All models account for a unit's average demographics (age, ethnicity, geographic diversity, and urbanization in soldiers’ hometowns) and allow each unit, battle, and month to have a different baseline level of losses.Footnote 28
FIGURE 1. Impact of NKVD presence on Soviet battlefield performance
Figure 1 illustrates several key findings.Footnote 29 First, there is a significant negative relationship between fratricidal coercion and key categories of flight. Doubling OO/SMERSH presence within a division is associated with a 1.4-percentage-point decline in the share of troops reported as MIA in a given month, a 0.2-percentage-point decline in troops reported as prisoners of war, and a 0.06-percentage-point decline in desertions. The magnitude of these shifts is substantively meaningful—equivalent to about two fewer soldiers missing in an average division-month.Footnote 30
The negative result with respect to POWs is striking. Considering that Soviet commanders routinely reported captured troops as MIA (section A1), the negative relationship between NKVD presence and MIAs could indicate either a shift in reporting or a genuine decline in POWs. Since the drop in reported MIAs is not accompanied by a rise in reported POWs, the first scenario seems unlikely.
Second, fratricidal coercion came at the cost of higher fatalities. Doubling OO/SMERSH presence increased the share of troops killed by 0.52 percentage points—about three additional deaths each month for an average division.Footnote 31 So, for every two soldiers the NKVD potentially deterred from fleeing, three stayed and died fighting.
Third, in divisions with more NKVD officers, a significantly smaller share of soldiers received medals for valor (−0.34 percentage points). This pattern challenges the idea that coercion universally increases combat resolve. Had this been true, we would observe not only less flight (and more fatalities), but also more acts of bravery. Instead, fewer soldiers went beyond the call of duty when the NKVD was present.
We conduct a battery of robustness checks to address possible inferential challenges, including placebo tests that randomize allocation of NKVD officers; subsample analyses of officers versus enlisted personnel; time-varying coefficients that capture coercion's changing effects; temporal splines to account for organizational challenges facing newer units; simulations to test for unobserved variation in unit strength; and non-independence between divisions (supplemental section A3). In almost all cases, our results are unchanged. We also consider the plausibility of several alternative interpretations of our findings, like ethnic/class discrimination and reporting biases.
Qualitative Analysis of Matched Soviet Rifle Divisions
We marshal qualitative evidence to detail how NKVD Special Sections generated a credible threat of fratricidal coercion and, in turn, how they affected battlefield behavior. We expect that soldiers and commanders were aware of these units’ presence (or absence) in their divisions; that the threat they posed appeared credible; that OOs had the capacity to close escape routes; and that their presence altered commanders’ tactical decisions, leading to the embrace of mass-casualty frontal assaults.
To detail this causal logic, we draw on a paired comparison of two similar Red Army Rifle Divisions that fought at the Battle of Leningrad (9 July to 26 October 1941, Figure 2a). We used statistical matching to identify pairs of Rifle Divisions that were balanced on almost all observable characteristics, but diverged in NKVD presence (see supplemental section A4). The pair we selected, the 168th and 90th, shared many key traits that might affect battlefield performance, including number of assigned personnel, equipment, organization, and chain of command (the 55th Army). Yet the 168th had fifty-seven NKVD officers assigned to it, while the 90th had just one. We focus on the critical September–October 1941 phase of the Battle of Leningrad, and trace each division's performance using declassified army, division, and regimental war logs, soldiers’ wartime letters and interviews,Footnote 32 newspaper articles from embedded journalists, maps, and RKKA personnel records.Footnote 33
FIGURE 2. The 90th and 168th Rifle Divisions at the Battle of Leningrad
By early September, as the Rifle Divisions reeled under German assaults, they had begun fighting retreats toward Kolpino, a hamlet meant to anchor the defense of Leningrad's southern approaches (Figure 2b). Arriving in mid-September, the divisions found themselves side by side in hastily organized defensive operations to prevent further German advances. Initially ordered to hold at all costs, the units were finally withdrawn from the front line on 25 October.
These eight weeks of combat offer a window into how blocking detachments affected efforts to instill discipline and maintain order amid battlefield chaos. The 168th's sizable Special Section, for example, was tasked with stiffening the resolve of reluctant soldiers through the threat, and actual practice, of shooting deserters. Senior leaders relied on these units to maintain order during fighting retreats, when panicked soldiers, often cut off from their officers, might have set off a cascade of desertion or surrender that could unravel the entire division. As one field-grade officer noted:
[Around 21 September], we received an order from the Supreme Commander—those who abandoned their positions without authorization would be shot … [NKVD] implemented the order immediately and began a merciless struggle against alarmists and deserters. Placing checkpoints near roads was especially useful. Groups of deserters retreating in disarray along the road to Leningrad were stopped by blocking detachments and divisional headquarters staff and sent back to the front. Order and discipline were restored completely.Footnote 34
By contrast, with only a skeleton blocking detachment, the 90th fell apart under German attacks in September, struggling just to get to Kolpino in some semblance of order. Retreats were chaotic, indiscipline high. Orders to counterattack were ignored. “It was very difficult,” one soldier remembered. “There was no leadership, no one knew the situation and, most importantly, there was no communication between officers and their men.” One rifleman recalled a common joke that their officers should be arrested so they could more easily find them.Footnote 35 Without a credible threat of punishment, and with officers unable to monitor soldiers amid the noise and confusion, some soldiers doffed their uniforms to escape German patrols in civilian clothes.Footnote 36 Desertion became more common.Footnote 37 Even when the front lines stabilized in October, soldiers and officers continued to slip away. Those who remained openly broached surrender. “Our situation is without hope,” wrote one soldier.Footnote 38
Where more NKVD officers were present, Special Sections were better able to close escape routes. While some soldiers managed to slip through the cordon—especially during fighting retreats in September—the 168th's blocking detachment managed to clamp down on most escape routes to the rear. Commanders and soldiers recall frequent encounters with Special Sections engaged in this task.Footnote 39 By contrast, we found only a single recorded instance where fleeing soldiers from the 90th came across a Special Section, an encounter they described as unexpected.Footnote 40 Standing orders instructed the 90th's soldiers to avoid roads and move through forests, compounding the NKVD's monitoring problem.Footnote 41 NKVD were too few in number, and the division too scattered, for deterrence to be credible.
Substantial evidence also indicates that NKVD presence encouraged the use of costly frontal assaults to bleed German forces through human wave attacks. The 168th's soldiers, many of whom reported reaching their “breaking point,” were thrown into counteroffensives and forced to advance through fear of punishment.Footnote 42 The division's commanding officer, General Andrey Bondarev, ordered numerous local counterattacks without artillery preparation, even when such fires were available and considered necessary by his subordinates.Footnote 43 Special sections drove reluctant soldiers forward, as journalists wrote glowing reports of their newfound resolve. “Each of the soldiers in Bondarev's division fought literally as ten men,” Krasnaya Zvezda reported, “and they held out without letting the enemy pass.” One German prisoner recounted, “We were scattered and put to flight [at Tosno] by this frightful division that wasn't afraid of artillery or mortar fire … The Russians fought like lions for every meter of ground.”Footnote 44 But these costly assaults introduced new vulnerabilities. Without adequate battlefield preparation, the 168th's command post was repeatedly endangered as it tried to hold fast while other divisions retreated around it.Footnote 45
Despite heavy losses, the 168th was continually engaged in meat-grinder counteroffensives. In early October, the 168th received orders to take up new positions southeast of Kolpino, astride major road and railway networks (Figure 2c). Inching its way forward in bitter, close-quarters fighting, the division drove German forces back at great cost. By 10 October, the 168th had carved out a twelve-to-fifteen-kilometer perimeter that protected the advances to Kolpino and Slutsk.Footnote 46 In rapid succession, the division launched counterattacks at Pushkin (13–14 October), Slutsk (16–17 October), and Putrolovo (21 October), slowing the Germans’ advance to a bloody crawl.
By contrast, the 90th—which lacked the coercive infrastructure to drive its soldiers forward—recorded only a single counterattack over a six-week period from mid-September to late October. Commanders preferred to hunker down in defensive positions, fearful that advances would create new opportunities for soldiers to slip away. For example, in early October, the 90th received orders to fight its way to new defensive positions southwest of Kolpino (Figure 2c). The under-strength division limped its way south, leaving a paper trail of increasingly desperate requests for soldiers (especially officers), weapons, air support, food, and, most notably, reinforcements for its Special Section. Soldiers took this opportunity to desert or go missing in large numbers, many sneaking away at night. The 90th recorded fewer battlefield casualties than the 168th, but higher rates of desertion, POWs, and MIAs. One status report from 10 October, signed by the 90th's commander and political commissar, claimed that only 600 soldiers remained available for duty.Footnote 47 The division's official history mostly passes over this period in silence, noting that “it was a most difficult, hungry, and unpleasant time” and “there was never a worse situation.”Footnote 48
Both units, or what remained of them, were ordered from the front lines on 25 October 1941. Their battlefield performance in September–October 1941 was sharply different (Table 1). In September, the 168th's personnel records reported 30.6 percent KIA and 54 percent MIA, compared to 11.6 percent KIA and 73 percent MIA in the 90th. Disarray and indiscipline were also costly for the 90th: three of its commanding officers were killed in battles in only three months. One survived only two days.Footnote 49 In October 1941, the 168th again reported a much larger share of soldiers KIA (33.9%) than the 90th (18.9%). The trend reverses for prisoners of war: the 90th reported 20.2 percent of their losses as POWs, compared with 3.7 percent from the 168th. The 168th was also able to launch counter-offensives and retreat in reasonably good order, which we ascribe to the presence of blocking detachments. The only category where the 90th outperformed the 168th was in individual medals for valor (5% versus 1.8%). But even this is problematic: it was the breakdown in order within the 90th that created the space for individual initiative, as many soldiers fought tenaciously to force their own escape, rather than to reverse gains by German forces.Footnote 50
TABLE 1. Paired comparison: Battle of Leningrad (9 July to 26 October 1941)
Notes: Battlefield performance indicators are derived from October 1941 declassified personnel records for the 168th (N = 437) and 90th (N = 381) Rifle Divisions. Estimates of divisional strength are drawn from official tables of organization and measured on the eve of the Battle of Leningrad (Askey Reference Askey2016, 526, 548).
Prewar Repression or Wartime Coercion?
Can we distinguish the battlefield effects of wartime coercion from the legacies of prewar repression, like mass terror or officer purges? Following RTZ, we define civilian repression as the number of political arrests that occurred within ten kilometers of a soldier's birthplace in 1936–38, averaged across all soldiers assigned to a unit-month.Footnote 51 We define purges as the number of officers in a given unit whom the NKVD arrested in 1936–38.Footnote 52 We conducted three tests to see how these pressures interact.
First, we look at the relationship between prewar repression and the assignment of NKVD officers to units. Regression estimates indicate that there were significantly more NKVD officers in units whose personnel were exposed to more repression (section A5 in the online supplement). Doubling a unit's exposure to civilian repression (or purges) is associated with an eight (or 3.6) percentage-point increase in NKVD officers assigned to it. It is unsurprising that the NKVD had a larger presence in units they had recently purged. The association with civilian repression is less direct. While the NKVD could observe soldiers’ age, ethnicity, lineage, and party affiliation from personnel records, they lacked the capacity to calculate repression rates in soldiers’ hometowns across thousands of units, or to make assignment decisions on this basis in real time. More plausibly, the same observable socio-demographic factors that led to higher repression in soldiers’ home communities may have led to greater coercion in their units.
Second, we assess whether the effect of prewar repression on behavior still holds when we adjust for fratricidal coercion as a post-treatment mediator. Sequential-g estimatesFootnote 53 suggest that wartime coercion accounts for some, but not all of the observed coercive impact of prewar repression (section A5 in the online supplement). For some outcomes (such as POW and punishment), the average controlled direct effect—the effect of prewar repression or purges on soldiers’ behavior when the mediator, NKVD presence, is held constant—is larger than the effect of prewar repression. For others (such as KIA and medals), the average controlled direct effect is smaller or more uncertain. Notably, estimates for both repression and purges directionally align with RTZ's hypothesized relationships.Footnote 54
Third, we examine how wartime NKVD presence interacted with the legacy of prewar repression on the battlefield. We extend our main model specification with an interaction term between NKVD presence and “high exposure,” which we define as units with an above-median level of prewar repression or purges. Our results provide partial support for RTZ's claim that soldiers are more responsive to wartime coercive incentives if they are more familiar with authorities’ ability and willingness to punish (section A5). For some outcomes, like POWs, this does appear to be the case, with prewar exposure seemingly “activating” the deterrent effect of fratricidal coercion. For most other outcomes, the prewar–wartime interaction effect is more variable. For example, higher exposure to officer purges amplifies the negative impact of NKVD presence on medals, but higher exposure to prewar civilian repression does not.Footnote 55 These findings underscore the need to consider both prewar and wartime treatment of soldiers, and how one may reinforce the other.
Cross-National Evidence
Do our findings generalize beyond the Eastern Front of World War II? And do these battlefield dynamics matter for strategic-level war outcomes?
Battle-Level Outcomes
We answer the first question by merging Project Mars data on blocking detachments with a data set on 526 land battles involving 185 belligerents between 1939 and 2011.Footnote 56 Each observation contains information on losses, including KIA, WIA, MIA, POW, and loss-exchange ratios. With the exception of medals, these measures mirror our RKKA analyses, facilitating a direct comparison.
We regress belligerents’ casualties on blocking-detachment presence, conflict fixed effects, and a battery of belligerent- and battle-level covariates (section A6).Footnote 57 Figure 3 reports average marginal effects estimates from these models, capturing how the presence of blocking detachments affects battlefield performance.Footnote 58 Consistent with our Soviet findings, armies with blocking detachments have fewer MIAs, but more KIAs and WIAs, and higher overall casualties as a proportion of initial strength.
FIGURE 3. Impact of fratricidal coercion across 526 battles, 1939—2011.
The only result inconsistent with our Soviet findings is for POWs, which appears statistically insignificant here. This discrepancy may be due to three factors. First, these differences might arise from variation in how we measure fratricidal coercion. Our cross-national analyses capture the specific impact of blocking detachments, while our Soviet analyses capture the general impact of NKVD officers, who had multiple coercive tools at their disposal. Second, it may reflect reporting differences: Soviet commanders routinely recorded POWs as MIAs for political reasons (see supplemental section A1), a dynamic not present in most armies. Third, the meaning and valence of POW status varies across belligerents: most armies do not view falling into captivity as a disreputable act, provided soldiers fought until they exhausted other options.
Our Soviet and cross-national evidence converge on the same finding: fratricidal coercion increases an army's own casualties. What remains unclear is the effect on relative casualties. Without matching Wehrmacht divisional records, we cannot capture how the interaction of German and Soviet forces might condition the effects of blocking detachments. Coercion might, for example, increase friendly casualties but, by forcing soldiers to stand and fight, inflict even higher casualties on enemy forces.
We evaluate this possibility by running additional cross-national analyses, with loss-exchange ratio (LER) as our dependent variable.Footnote 59 A higher LER indicates greater military effectiveness, in the narrow sense of inflicting higher losses on enemy forces than the enemy inflicts on one's own. We find that armies with blocking units have significantly lower loss-exchange ratios: they suffer more casualties themselves than they inflict on others (Figure 6, last row). While fratricidal coercion might prevent soldiers from fleeing, it does not appear to yield broader tactical advantages. If anything, it tends to make battles deadlier for friendly soldiers than for enemy soldiers.
War-Level Outcomes
Can these battle-level dynamics help explain war outcomes? While space constraints preclude a comprehensive study of how tactical performance aggregates to victory in war, tentative evidence suggests that both higher average battlefield losses and use of blocking detachments are early indicators of strategic-level defeat.
To establish the first result, we linked our cross-national battle dataFootnote 60 to war outcome data from the Correlates of War project, with outcomes defined as “win,” “lose,” and “other.”Footnote 61 We took averages of battlefield losses for each belligerent in each war, and estimated linear probability models, regressing war outcomes on logged average losses by category (KIA, WIA, MIA, POW, LER, and proportion of initial force lost), fixed effects for country and decade, and covariates.Footnote 62 Our estimates suggest that belligerents are more likely to lose wars when their average battle KIA and MIA are high—absolutely, and as a proportion of initial forces (section A6).
Given that blocking detachments are associated with higher casualties and lower MIA rates, the net impact of fratricidal coercion on war outcomes may at first appear ambiguous. A more direct look at the data paints a clearer picture. On average, 27.8 percent of belligerents who used blocking detachments since 1939 won their wars (and 42% lost them), compared to 37.3 percent (30%) for belligerents without blocking units. We find the same pattern in linear probability models, regressing war outcomes on the use of blocking detachments, with the same covariates and fixed effects (section A6). The probability of victory is 0.5 lower for belligerents who employ blocking units.
These estimates are not causal, and it is possible that losing armies sometimes use fratricidal coercion as a gamble for resurrection. Still, the historical track record of belligerents that rely on these methods is not promising, at any level of war. Even in the case of the Soviet Union, which ultimately prevailed in World War II, evidence suggests that fratricidal coercion made victory costlier than it otherwise might have been.
Conclusion
Our microlevel and cross-national evidence point to the same grim conclusion: fratricidal coercion is a powerful determinant of battlefield behavior. Building on earlier analyses of prewar repression,Footnote 63 we find that the presence of NKVD Special Sections in Soviet Rifle Divisions is associated with fewer MIAs, POWs, and desertions, but also higher casualties and fewer medals. Fratricidal coercion's cross-cutting effects—driving reluctant soldiers forward while dampening the initiative of true believers—also appear in cross-national data at multiple levels of analysis. Armies that deploy blocking detachments suffer greater casualties and are more likely to lose wars. For most armies, fratricidal coercion is a gamble than does not pay off.
Our findings suggest several new avenues for the study of wartime coercion. On the theoretical front, there is a pressing need to explore how identity and exposure to state discrimination and repression shape the distribution of resolve within (and across) armies. New theoretical ground could be broken, for example, by exploring how fratricidal coercion interacts with group identities to shape the credibility and effectiveness of deterrent threats, along with nonpunitive motivational strategies like ideological appeals and battlefield spoils. Whether some political regimes and leaders are more likely to resort to such brutality likewise remains an open question.Footnote 64 On the empirical front, our research underscores the need for an ambitious program of data collection on various forms of fratricidal coercion, like extrajudicial executions, penal and labor battalions, and corporal punishment. We lack fine-grained data on the relative size, recruitment, tasks, and lethality of blocking detachments.Footnote 65
Our study also carries policy implications. Russia's war of attrition against Ukraine runs on fratricidal coercion, forcing reluctant soldiers into “meat storms” against entrenched enemy positions. Yet since prevailing frameworks for assessing military effectiveness ignore fratricidal coercion, analysts risk missing its emergence and dismissing its importance. As we have seen, these measures can boost an army's resilience by preventing disintegration, an unwelcome surprise for those who see coercion as a sign of pending collapse. However, the vulnerabilities introduced by fratricidal coercion are real. Militaries and intelligence agencies primed to look for these cross-cutting effects can exploit them. Commanders might, for example, target their adversary's coercive apparatus to create new avenues for disillusioned soldiers to flee, or use information operations to stoke resentment. They might also stand aside, content to watch the enemy kill its own soldiers to hold itself together. Far from a relic of a bygone era, fratricidal coercion remains a persistent feature of the modern battlefield, one that scholars and policymakers would do well to integrate into theories of war and military effectiveness.
Data Availability Statement
Replication files for this research note may be found at <https://doi.org/10.7910/DVN/914AN0>.
Supplementary Material
Supplementary material for this research note is available at <https://doi.org/10.1017/S002081832400033X>.
Acknowledgments
For helpful comments, we thank Jacob Aronson, Avishay Ben Sasson-Gordis, Andrew Bertoli, Stephen Brooks, Stephen Chaudoin, Christian Davenport, Christina Davis, Mark Dincecco, Chris Fariss, Peter Feaver, Margaret Foster, Taylor Fravel, Jeffrey Friedman, Ryan Grauer, Mariya Grinberg, Andy Halterman, Dotan Haim, Paul Huth, Iain Johnston, Jennifer Lind, Christoph Mikulaschek, Nicholas Miller, James Morrow, Emily Myers, Emerson Niou, Roger Petersen, Barry Posen, Katy Powers, Michael Poznansky, Daryl Press, Mara Revkin, Livia Schubiger, David Siegel, Megan Stewart, Roya Talibova, Ben Valentino, and William Wohlforth. We also thank the IO editors and anonymous reviewers for helpful feedback. We also thank audiences where we presented earlier versions at the University of Michigan, Dartmouth College, Duke University, Duke University School of Law, Harvard University, MIT, and the 2020 annual meeting of the American Political Science Association. We thank Kelly Sandefer at Beehive Mapping for our excellent maps.
Open access
