External intelligence assistance and the recipient government’s violence against civilians

Abstract

During a civil war, does external intelligence assistance reduce violence perpetrated by the recipient government against civilians? I contend that intelligence assistance reduces violence against civilians by facilitating identification problems and adopting a “winning-hearts-and-minds strategy,” which enhances the recipient government's legitimacy and intelligence potential. Enhanced intelligence capability solves the recipient government's identification problems. I examined this logic using a dataset on external support and one-sided violence between 1990 and 2008. The empirical findings show that external intelligence assistance reduces the recipient government's violence against civilians.

Keywords

External intelligence assistance counterinsurgency government violence

Introduction

The fight against rebel groups is becoming increasingly complex worldwide as several countries have active rebel groups (Walter, 2019) that form alliances (Christa, 2012) to spread and instrumentalize terrorism (Fortna, 2015). Often, governments alone find it unfeasible to combat security threats without cooperation at both the national and international levels. External intelligence assistance is one form of international cooperation that functions to support governments in dealing with threats posed by rebel groups. For instance, to disband escalating insurgency in Mozambique, the USA is looking to supply intelligence assistance to the Mozambican government, and US soldiers are already training Mozambican troops.¹

Does external intelligence assistance influence conflicting parties’ power struggles? Intelligence serves two purposes: to provide policy information and to support operations that ensure state security (Bruneau, 2008). To protect and advance a state's interests, intelligence systems collect, analyze and disseminate information (Sims, 2006).² Overall, intelligence is focused on the policies, intentions and capabilities of foreign governments but increasingly on the plans and activities of non-state actors that threaten the state (George, 2013). As a form of external intelligence assistance, information on non-state actors is collected, analyzed, and provided. Foreign intelligence assistance includes providing maps and relaying enemy positions, information on troop capability, and/or data on the whereabouts of rebel leaders (Högbladh et al., 2011). Intelligence supports military action involving minimum violence through effective tactics that isolate revolutionary elites from their followers (Joes, 2004).

However, does external intelligence assistance to governments reduce the violence levels experienced by civilians? This question is germane to contemporary security issues, as governments seeking international cooperation and intervention (with minimum violence), aimed at appeasing security threats posed by rebel groups, do so with certain expectations about their effectiveness under dire circumstances. This question is pertinent as the outcomes of external support are pivotal to understanding international cooperation, which could influence policymakers’ decision making.

I examine external intelligence assistance from 1990 to 2008 and its impact on the levels of violence perpetrated by recipient governments against civilians. I argue that intelligence assistance reduces the recipient government's violence against civilians by mitigating identification problems and supplementing the winning-hearts-and-minds strategy. Intelligence assistance is followed by the adoption of such a strategy as it enhances reliable information from the local population. Intelligence has long-term political objectives to enhance the target government's legitimacy and accompanies democratic control. Such a strategy endogenously mitigates identification problems by enhancing the recipient government's intelligence capability. To test this logic, I investigate the violence levels against civilians by the government after receiving external intelligence. This empirical analysis illustrates that a government that receives external intelligence assistance reduces violence levels against civilians.

This study has important theoretical and policy implications. First, while earlier studies emphasized the winning-hearts-and-minds approach of intelligence assistance (Charters, 2012; Segell, 2009; Smith, 2001), the existing literature has overlooked the empirical evaluation of the effectiveness of intelligence assistance in minimizing civilian victimization. The study findings imply that intelligence cooperation during civil wars reduces civilian targeting by the recipient government, indicating the efficiency of the winning-hearts-and-minds approach. Thus, the interventions influence victimization of civilians by altering the balance of power (Wood et al., 2012) and also supplement the counterinsurgent strategy. Second, while previous studies relied on the assumption that interveners are decisive in structuring the good governance approach, as pointed by Hazelton (2017), I develop a theory of interaction between intelligence assistance, the recipient government, and a good governance approach, namely, winning-hearts-and-minds strategy, with empirical evaluations. This has practical implications as the counterinsurgent government's willingness to adopt the good governance approach may not be adequate to implement the strategy if the intelligence capability is low. However, with enhanced intelligence capability, external actors could supplement strategy implementation by avoiding civilian targeting. This study, therefore, contributes toward developing understanding on conflict management through external support during civil wars.

Literature review

Determinants of violence against civilians during civil wars

Why do governments inflict violence against civilians during civil wars? A strand of research argues that governments inflict violence against civilians owing to identification problems. When belligerent groups lack the information necessary to target their adversaries selectively, indiscriminate violence against civilians is resorted to (Kalyvas, 2006). Domestic structural factors represent one of several factors that influence such information shortages. Structural factors, such as ethnically homogenous districts, offer fewer cooperative informants and less reliable intelligence, increasing indiscriminate violence (Siroky and Dzutsati, 2015). Moreover, a structural factor, such as state capacity, also influences identification problems. When security forces lack the intelligence capacity to collect sufficient private information and rely on civilian elites for information, the number of victims increases owing to the widened targeting criteria through false information (Winward, 2021). Another factor is the involvement of domestic actors during conflicts. For instance, local militias enable greater access to local knowledge and enhance the government's ability to target opponents selectively (Clayton and Thomson, 2016).

Alternatively, governments target civilians as a part of their strategy to battle their opponents (Valentino et al., 2004). During the insurgency, guerrilla forces depend on the civilian population's support for survival (Downes, 2007). In such a situation, governments can deter civilian populations from supporting opponents by inflicting selective violence as penalization, in which targets are chosen based on individualized suspicion (Kalyvas, 2006). Even indiscriminate violence, if committed against a small population in a relatively confined space, can help the government defeat insurgents quickly because the government interdicts their source of support (Downes, 2007). Barbarism, which involves attacking the sanctuary and social assistance of the opponent, destroys their capacity to fight (Arreguín-Toft, 2001). Indiscriminate violence not only eliminates support for the enemy but also disrupts insurgents’ activities. Lyall (2009) argues that indiscriminate violence by the government creates logistical problems for insurgencies and undermines their organization's military effectiveness by facilitating collective action of locals against insurgents. Lyall (2009) finds that indiscriminate violence reduces insurgent attacks in Chechnya.

However, whether targeting civilians is an effective strategy remains controversial. Pape (1996) argues that punishing the enemy civilian population is not successful in inducing concessions. Collateral damage lessens information sharing about insurgents with government forces (Condra and Shapiro, 2012).

External intelligence assistance may influence the recipient government's violence against civilians through such two mechanisms of violence—by altering the violence strategy and supplementing information shortages.

External support and civil wars

Research maintains that military interventions alter the balance of capabilities available to belligerents (Balch-Lindsay and Enterline, 2000; Regan, 2002). This altered relative strength affects reservation points and concessions at the negotiation table. Military interventions ultimately extend the expected duration of negotiations (Regan, 2002) and achieve a better timeline for reaching a negotiated settlement (Balch-Lindsay et al., 2008). The change in the balance of capabilities also affects belligerents’ strategy of civilian victimization. Belligerents victimize civilians owing to a shift in the balance of capabilities to compensate for the decline in relative power and to deter rising threats (Wood et al., 2012). When the balance of interventions forces a shift in favor of rebels, the group tends to reduce the use of anti-civilian violence; however, if the balance shifts against them, rebels increase violence against civilians (Wood et al., 2012). Military support is not the only form of intervention that alters belligerents’ military capacities. If rebels receive highly fungible external support, they are less likely to reach conflict termination as such support does not always boost fighting power immediately, creating high levels of uncertainty that impact any bargained settlement (Sawyer et al., 2017).

External support alters rebel groups’ capabilities and also influences rebel governance. Domestic rebels are obliged to provide desired goods to the communities they represent, but rebels with external bases and patrons are less sensitive to the demands of local communities (Salehyan, 2010). This disconnect with local politics could incentivize rebels to target civilians. Rebel groups with external patron support commit high levels of indiscriminate violence (Weinstein, 2007). Reliance on foreign state support may also allow groups to afford legitimacy costs better and engage in terrorism (Fortna et al., 2018). Wood (2013) demonstrates the conditional effects of foreign sponsorship on insurgent violence. When alternative source flows substitute for popular mobilization support and grassroots mobilization strategies, militarily more capable insurgent groups inflict greater violence on civilians (Wood, 2013).

The influence of third parties is not limited to physical changes among those receiving support; third parties can also exert their preferences over conflict outcomes, which could become instruments in the strategies of conflicting parties in civil wars. For instance, if an intervention opposes the government and attempts to modify the pre-conflict status quo of the state, punishment becomes an influential strategy; thus, interventions on behalf of the opposition coincide with more violent civil wars (Lemke and Regan, 2004). Moreover, when foreign funding for rebel groups is recognized in principal–agent terms, some external principals have stronger concerns regarding atrocities in conflict zones, resulting in less violence against civilians by rebels receiving support from the concerned principals (Salehyan et al., 2014).

The same logic applies to external intelligence assistance during civil wars. External supporters providing intelligence assistance have strategy preferences. In the extant literature, strategy analysis of counterinsurgency intelligence demonstrates that such practices follow the winning-hearts-and-minds strategy. For instance, in the Malayan case, the war against communist insurgency was fought via civilian campaigns designed to win civilians’ hearts and minds (Smith, 2001). General Sir Gerald Templer successfully reduced the Malayan Communist Party's support base and de-legitimized the party by encouraging the Malayan Chinese support and sponsoring elections (Smith, 2001). In the case of Uganda, the USA started providing military aid to the Ugandan government in its fight against the Lord's Resistance Army (LRA). Consequently, the US African Command was established, which provided training, financial support, airlifts, communications, and intelligence assistance to 3000 African Union regional task force troops (Demmers and Gould, 2018). On 24 March 2010, President Obama signed an Act that formalized the US government's commitment to providing “political, economic, military, and intelligence assistance for viable multilateral efforts to protect civilians from the LRA” (Forest, 2014: 3). This assistance was demonstrative of the winning-hearts-and-minds approach. For instance, the Obama administration contributed toward building telecommunications infrastructure and supplying cellphones and radio capacity to encourage communities to transmit information to the outside world (Forest, 2014). In the case of Iraq, the Information Operation Task Force's (IOTF) mission, rooted in the winning-hearts-and-minds strategy, aimed to derail extremist politics (Segell, 2009). In the case of Afghanistan, to foster local cooperation, the US intelligence acknowledged that addressing issues of local concern was an important intelligence task (Charters, 2012). The strategy of winning hearts and minds is considered a softer approach vs. traditional military operations, and is enacted via different forms of information operations, psychological operations, and Special Forces activities (Egnell, 2010). By adopting the winning-hearts-and-minds strategy, external intelligence assistance can shift the dynamics of popular support. For instance, Segell (2009) shows that individuals who watched more than one IOTF television commercial were more likely to reject Al-Qaeda's worldview. Such intelligence assistance differs from military assistance because while it provides additional strategy preferences, the principal function of military intervention in conflicts is to provide additional military capacity (Sullivan and Karreth, 2015).

Regarding the second mechanism of how external intelligence assistance influences civil wars, studies show that external intelligence assistance degrades and disrupts opponents’ activities. For instance, Canadian military intelligence assistance in Afghanistan led to the capture of suspected or known insurgents, including high-value targets (Charters, 2012). In Vietnam's case, the US intelligence assistance led to the capture or death of over 80,000 suspected members of the Viet Cong Infrastructure, comprising enemy political leaders and guerrilla forces (Walsh, 2008). However, there are cases where such disruption fails. For instance, US intelligence led to the destruction of several LRA strongholds in Uganda but failed to capture Joseph Kony, who regrouped and launched a series of reprisal attacks against villages in the northeastern Democratic Republic of Congo (Forest, 2014). Hence, in most cases, external intelligence assistance helps to identify suspects.

Thus, civil wars involving external intelligence assistance may be influenced by both the winning-hearts-and-minds strategy and suspect identification, which could ultimately influence the recipient government's violence against civilians.

A theory of external intelligence assistance and violence against civilians

I argue that external intelligence assistance reduces the recipient government's violence against civilians through two mechanisms: (1) enhancing the intelligence capacity of the government that solves identification problems; and (2) encouraging the recipient government to adopt the winning-hearts-and-minds strategy. I further argue that this strategy itself is not sufficient for restraint toward civilians as the counterinsurgent government still faces identification problems. External intelligence assistance supplements the information required to avoid civilian targeting as part of this strategy. This provides a secure environment for people, resulting in building trustworthy relationships, facilitating information flow and enhancing the intelligence capability of the recipient government. From the supplier's perspective, the enforcement of intelligence capability of the recipient is important for the supporter because intelligence assistance is a two-sided process, and information passed to the supporter needs to be accurate (Byman, 2006; Munton and Fredj, 2013). From the recipient's perspective, the enhanced intelligence capability results in fewer identification problems, enabling the counterinsurgent government to implement the winning-hearts-and-minds strategy. Through such an endogenous relationship, external intelligence assistance reduces violence by the recipient government against civilians (Figure 1). I present the detailed logic below.

Figure 1.

External intelligence assistance and violence against civilians.

First, identification problems suggest that belligerent groups inflict indiscriminate violence against civilians owing to the lack of information necessary to target their adversaries selectively (Kalyvas, 2006). It is not rare for rebel groups to hide in villages inhabited by civilian populations. In 2017, pursuing separatist militants in southern Kashmir, the Indian military forces cordoned off villages and conducted house-to-house searches, causing civilian casualties during the ensuing clashes.³ In Venezuela, the action against a town around La Victoria was aimed at the Tenth Front, a faction of the former Revolutionary Armed Forces of Columbia dissents. Government soldiers entered civilian homes, and interrogated and killed civilians accused of being guerrillas.⁴ Even if rebel groups do not live in the village, they participate in activities alongside civilians, which makes it difficult for the government to identify these groups. In March 2021, members of the Al-Qaeda-affiliated jihadist group in Mali, the Katiba Serma, participated in a wedding in a village, where an airstrike resulted in civilian casualties.⁵ Information shortages are a common issue causing identification problems, and consequently, civilian casualties.

External actors influence the government's ability to obtain necessary information. Intelligence assistance from external supporters, in which they assist the government in acquiring information about their targets, is a dynamic whereby external actors supply information. For instance, the USA is expanding its intelligence-sharing with Saudi Arabia, which involves the vetting of targeting information prepared by Saudis in its air campaign against Houthi militias in Yemen.⁶ In the case of the Malayan Insurgency, intelligence gathering campaigns were established with assistance from British experts, such as “Operation Letter Box”, whereby Chinese villagers were encouraged to divulge information about guerrillas through anonymous questionnaires (Smith, 2001). Such forms of intelligence assistance supplement governments’ information on targets, allowing them to specifically target insurgents separate from civilians. Indeed, past intelligence assistance succeeded in identifying suspects (Charters, 2012; Walsh, 2008).

Next, I argue that external supporters prefer the winning-hearts-and-minds strategy, and the target government adopts such a strategy owing to three mechanisms: (1) a thirst for reliable information; (2) a focus on long-term objectives; and (3) democratic oversight of intelligence agencies. Indeed, anecdotal evidence shows that intelligence practices tend to follow the winning-hearts-and-minds strategy (Charters, 2012; Segell, 2009; Smith, 2001).

First, intelligence assistance is not a one-sided action. To receive external support, the external state must be prepared to support the actor, and the actor must be willing to accept the support (Salehyan et al., 2011). Once support is granted, the recipient and supplier cooperate. Intelligence assistance collects, assesses and analyzes information. Information may be obtained through the receipt either independently or dependently. From the supporter's perspective, the information must be accurate (Byman, 2006) because it can challenge the supporter state's views or undermine the supporter's preferred policies (Munton and Fredj, 2013), ultimately posing a security threat. To enhance information accuracy, the supporter prefers the winning-hearts-and-minds strategy as the strategy enhances the recipient government's intelligence capability.⁷

Intelligence is the principal source of information during a conflict, and the population provides information. For instance, to identify and capture members of the Boko Haram in Nigeria, the Civilian Joint Task Force provided information on the Boko Haram-occupied locations and the surrounding environment (Agbiboa, 2019). Thus, civilians can provide crucial intelligence to governments related to the locations and coercive capabilities of insurgents and the surrounding sheltered environments.

While the local people are a critical source of information and intelligence, they refuse to talk about insurgents unless they feel secure (Galula, 1964). In Afghanistan, the local population shared information only after the constantly patrolling troops demonstrated their commitment, convincing the people that they could be trusted to protect them (Charters, 2012; Windsor et al., 2008). A secure environment enables counterinsurgency operations to build personal relationships with the population, producing reliable intelligence (Duyvesteyn, 2011). If the government selectively targets people who provide information, this signals the consequences of betrayal and deters defection (Downes, 2007), discouraging the civilian population from passing on information to the government. Indeed, Condra and Shapiro (2012) and Shaver and Shapiro (2021) find that civilian targeting minimizes information sharing about insurgents with government forces. As such, the winning-hearts-and-minds strategy enhances the intelligence capability, facilitating local intelligence flow. This enables the supporter and the recipient government to obtain more accurate local information.

Second, I argue that intelligence assistance is operationalized consistently with long-term political objectives rather than short-term military ones, preferring the winning-hearts-and-minds strategy. Intelligence supports operations with the ultimate goal of ensuring state security (Bruneau, 2008). In fact, intelligence assesses long-term trends of interest to government officials (Johnson and Wirtz, 2004). The ultimate goal of security leads to long-term perspectives. However, not all leaders have the same time horizon in assessing security. Those who discount the future tend to invest in short-term securities (Loewenstein and Prelec, 1992), while leaders feeling politically secure tend to adopt a long-term view (Li, 2009). Providers of external intelligence assistance could adopt a long-term horizon in evaluating security as the battlefield is outside of the country, and the ongoing conflict of the target state is less likely to lead to an imminent threat of removal from the office of the leader assisting. Considering long-term security, external supporters providing intelligence would consider long-term political control of the target state through the winning-hearts-and-minds strategy, where the state obtains legitimacy and civilian support. While coercion and sanction are detrimental mechanisms of social control, legitimacy motivates compliance and achieves social control (Hurd, 1999). Using coercive power only fails to govern a society long term (Tyler, 2001). Without legitimacy, the state will probably fail in the future (Lake, 2016). By enhancing the legitimacy and social control of the target state, intelligence assistance ensures the target state's security.

Counterinsurgency is a long-term strategy, and from this perspective, once the locations of rebel groups are identified and even if it is easy to expel them from the area, dispersing them does not end the insurgency. Preventing groups from reforming is also necessary. A victory of counterinsurgency constitutes permanent isolation of insurgents from the population, where the isolation is not enforced upon but maintained by the population (Galula, 1964). As insurgencies are highly dependent on the local civilian population for food, shelter, supplies, recruits, and information (Downes, 2007), the return of guerrilla units or rebel groups requires cooperation from the population. In a long-term vision, establishing and maintaining a trustworthy relationship with the local population helps to garner political support and establish the regime's legitimacy. Thus, the success of counterinsurgency depends on establishing a perception of legitimacy (Fitzsimmons, 2008).

For short-term military objectives, violence against civilians can effectively reduce insurgent attacks (Lyall, 2009). As insurgencies depend on civilians, indiscriminate violence and destroying the sources of support can be effective in defeating insurgencies (Downes, 2007). However, a temporal reduction in insurgent attacks does not indicate that their will is silenced. The success of insurgents arises not from a military victory but from the progressive attrition of their opponents’ political capability to wage war (Mack, 1975: 177). Shattering their will to fight is more important than destroying their army (Griffith, 1963; Mack, 1975). In the long-term, enhancing the government's legitimacy with civilian political support is more effective in destroying the will of the current and potential opponents rather than temporarily destroying insurgents’ existing capabilities.

Acknowledging and sharing such a long-term objective with the recipient government incentivizes restraint toward civilians because violence against civilians diminishes the government's legitimacy. Violence perpetrated against civilians creates grievances (Hashim, 2006) and a sense of insecurity among citizens (Kalyvas, 2006; Mason and Krane, 1989). Simultaneously, this induces support for rebels and allows them to provide state functions through security, constituting a source of legitimacy for rebel governance (Terpstra and Frerks, 2017; Worrall, 2017) and decreases regime legitimacy. Arreguín-Toft (2001) argues that if the desired objective is long-term political control, barbarism invariably backfires. Considering such long-term consequences, counterinsurgency intelligence encourages governments to avoid targeting civilians to prevent them from supporting the governments’ opponents.

Third, external intelligence assistance encourages the recipient government to adopt the winning-hearts-and-minds strategy, as democratic control curbs intelligence to a certain extent. External intelligence assistance faces constraints from domestic and external audiences. While democratic control is direct in democracy through intelligence review, in autocracy, the mechanism is indirect because naming and shaming by foreign audiences turn covert abuses into common knowledge and lead to domestic demands for compliance with international human rights (Hendrix and Wong, 2013), making the recipient of intelligence assistance minimize violence against civilians.

Organizations can supervise counterinsurgency directives and priorities for interagency coordination (Matei and Bruneau, 2011). Since the 1970s, the supervision of security intelligence agencies has spread to many states (Gill, 2007). Through this supervision, the effectiveness of external intelligence assistance can be constantly reviewed by evaluating the associated costs of intelligence assistance. For instance, the official National Commission of Inquiry, the Agranat Commission, was established to investigate why Israel's intelligence failed in 1973. This review could also occur during ongoing intelligence assistance. As policymakers have strong political incentives to disregard unsettling estimates, unsatisfied policymakers can even create ad hoc bodies to review intelligence (Rovner and Long, 2005).

In an intelligence review, civilian casualties are considered a cost of the assistance. In a review report conducted by the US Military Operations, minimization of civilian casualties is referred to as a factor that maintains the support of partner governments.⁸ Operations that cause high civilian casualties not only indicate gaps in intelligence and procedures but also damage the leadership credibility of governments providing intelligence assistance (Federici and Dalton, 2017). Thus, if a state receives intelligence assistance but does not adopt the winning-hearts-and-minds strategy, the support could be terminated after a review.⁹

Autocracy is also responsive to a review of intelligence. However, in this case, the mechanism of democratic control can be indirect. Intelligence services in an autocracy can be reviewed in other countries. For instance, when the Syrian government carried out chemical weapon attacks in 2017, the USA accused Russia, although there was no consensus to corroborate Russian involvement.¹⁰ In autocratic regimes, damage to international reputation can invoke citizens’ disapproval of leaders through both formal and informal channels when the government's behavior and effectiveness are observable (Li and Chen, 2021). Although the effectiveness of intelligence sharing may not be directly observable to the domestic audience, when domestic sources of information on human rights abuses are lacking and covert, the regime is more sensitive to international naming and shaming because it turns covert abuses into common knowledge, improving human rights outcomes (Hendrix and Wong, 2013). Even without citizens’ reaction, the intelligence service incurs international reputational costs. Thus, intelligence assistance is expected to push back toward the winning-hearts-and-minds strategy when it deviates from such a strategy.

Governments avoid civilian casualties through the winning-hearts-and-minds strategy. However, the strategy itself is insufficient to prevent civilians from being targeted. Even counterinsurgent governments using such strategies face identification problems, especially if they lack intelligence capability. First, a lack of intelligence potential generates a need for external intelligence assistance. Once external intelligence assistance is granted, with supplemented information and a counterinsurgent strategy, the recipient government commits less violence against civilians. This results in the provision of a secure environment and the building of a trustworthy relationship with the population, thus enhancing the recipient government's intelligence capability. As a result, the identification problem is endogenously mitigated, further reducing civilian casualties.

These discussions lead to the following hypothesis:

Hypothesis: Governments receiving external intelligence assistance perpetrate less violence against civilians.

Research design and methodology

To test the hypothesis, I conducted an empirical analysis using country-year as the unit of analysis because the hypothesis investigates violence perpetrated by governments. I used data from the Uppsala Data Conflict Program (UCDP), One-sided Violence Dataset version 20.1 (Eck and Hultman, 2007; Pettersson and Öberg, 2020). In this dataset, one-sided violence is defined as any use of armed force by the government or by a formally organized group against civilians that results in at least 25 deaths. The dependent variable was considered the best estimate of the number of civilian deaths caused by government violence in a given year. While constructing the sample, I aggregated the expanded non-state actors (NSA) dataset (Cunningham et al., 2009) at the country-year level. Thus, the sample covered all country-years where at least one active non-state actor identified in the NSA dataset existed in a given country-year. The dependent variable's minimum, maximum, and mean values were 0, 5801,¹¹ and 94.598, respectively. As the distribution of the dependent variable was positively skewed, I discuss the appropriate model in a following section. The total number of observations was 483,¹² including 57 countries between 1990 and 2008.

Main explanatory variables

To test the relationship between external intelligence assistance and recipient government's violence against civilians, I used intelligence assistance for G_t−₁ as the main explanatory variable. The variable takes a value of 1 if a government received any form of intelligence from an external supporter, such as maps relaying enemy positions, information on troop capability, and/or data on the whereabouts of rebel leaders (Högbladh et al., 2011) in a given year; otherwise, it takes the value of 0. The information was obtained from the UCDP External Support Dataset (Högbladh et al., 2011). Owing to reverse causality issues, I used the one-year lag of this variable. Additionally, there may be a lag in the improvement of the intelligence capability of the recipient government's intelligence assistance. However, as a one-year lag will not be able to address endogeneity concerns regarding governments attracting intelligence assistance by adopting the winning-hearts-and-minds strategy, after explaining the control variables in the next subsection, I explain how I addressed this problem other than lagging the main explanatory variable.

Table B (Online Appendix) shows that among the sample countries, 13 countries received external intelligence assistance. Owing to the missing values of the control variables for Afghanistan in 2003 and 2004, these were not included in the analysis. Thus, 31 observations out of 483 were coded as 1 for the one-year lag of the independent variable. The theoretical logic presented earlier indicates how identification problems, intelligence assistance, and the winning-hearts-and-minds strategy of intelligence assistance apply not only to government intelligence assistance but also to non-governmental intelligence assistance. Thus, I included observations for Colombia 1994–1999 and Myanmar 2005, where the intelligence assistance provider was not the government. However, the logic regarding democratic control of intelligence assistance may only apply to the state; thus, in the Online Appendix, I present the results on intelligence assistance by removing Colombia 1994–1999 and Myanmar 2005. To check if the US intelligence assistance drove the result, I estimated a model using the same only in the Online Appendix.

Control variables

I also included several control variables. External actors affect the balance of capability (Licklider, 1993; Regan, 2002), influencing a conflict actor employing violence tactics toward civilians (Wood et al., 2012). According to the principal–agent theory, external support may incentivize killing innocents to demonstrate commitment toward sponsors (Salehyan et al., 2014). As external support other than intelligence assistance may influence the balance of power, I controlled for external support for G_t−₁, indicating if the government received any external support in a given year. The measure was taken from the UCDP External Support Dataset (Högbladh et al., 2011). In my sample, external intelligence assistance accompanied other types of support except for only one observation, India in 2008. The other observations that received intelligence assistance received other types of support also. About 214 observations received external support but not intelligence assistance. As a robustness check, I re-estimated the models without this general external support variable (Online Appendix Table H).

Weak rebel groups may be more likely to remain incognito in the civilian population for their survival, rendering identification difficult and increasing demand for both external intelligence assistance and state violence against civilians. Thus, I controlled for parity/stronger rebel_t−₁, which takes the value of 1 if the rebel group's strength relative to government forces is at parity or stronger, and otherwise, 0. When more than one rebel group exists in the country, the maximum value is used for analysis; hence, the value 1 is assigned if there is at least one parity or stronger rebel group in the country during a given year. The information used was obtained from the NSA dataset (Cunningham et al., 2009).

Territorial control by rebel groups influences the government's strategy to target civilians. When a rebel group controls a territory, a government may increase the demand for external intelligence assistance owing to difficulty in collecting intelligence from the local population where the rebel governs. On the one hand, territorial conquest encourages the government to eliminate enemy combatants, thereby increasing civilian targeting (Downes, 2006); however, on the other hand, territorial control by rebel groups may lead to less civilian targeting as it could change the form of civil war from insurgency to conventional warfare, resulting in fewer identification problems. The expected effect of territorial control is mixed, either positive or negative. Thus, territorial control_t−₁ was included as a binary indicator, measuring if at least one rebel group controlled a territory. The measure was taken from the NSA dataset (Cunningham et al., 2009).

I also controlled for conflict characteristics: intensity_t−₁, number of rebels_t−₁ and rebel violence_t−₁. First, conflict severity is positively associated with violence against civilians (Downes, 2006; Hultman, 2012; Salehyan et al., 2014; Wood et al., 2012). Intensity_t−₁ takes the value of 1 if the number of battle-related deaths exceeds 1000 in a country during a given year, and 0 otherwise. The information used was taken from the UCDP Battle-Related Deaths Dataset (Pettersson and Öberg, 2020). Second, Wood et al. (2012) showed that the presence of multiple rebel groups increased the level of violence enacted against civilians by governments. When there are multiple rebel groups, these groups may form alliances (Christa, 2012), complicating the conflict structure and increasing the need for intelligence assistance. Thus, I controlled for the log-transformed¹³ number of rebels_t−₁, which counts the active number of rebel groups during a given year in a country. Finally, the opponents’ level of civilian targeting may influence armed actors inflicting violence against civilians either negatively (Hultman, 2012; Wood et al., 2012) or positively (Salehyan et al., 2014; Wood, 2013). To account for such dependencies of violence, I controlled for log-transformed rebel violence_t−₁. This counts the aggregate number of instances of violence committed by rebel groups in a country during a given year.¹⁴ I did not control for types of conflict (irregular/conventional) as many insurgents engage in both (Hazelton, 2017).

To control for state characteristics, I included lnGDP p.c._t−₁, ln Population_t−₁ and Democracy_t−₁. First, weak states face a higher risk of killing civilians through a genocide (Anderton and Carter, 2015). A government may receive intelligence assistance as a substitute for the weak state capacity. Thus, I controlled for log-transformed GDP per capita from the Varieties of Democracy Dataset (Inklaar et al., 2018). Second, larger populations may enable more violence against civilians (Salehyan et al., 2014) or may lead to less violence (Hultman, 2012). I controlled for log-transformed population from the World Bank Development Indicators. Third, the regime type influences the depth of intelligence cooperation because the absence of democratic norms, such as contingent consent and bounded uncertainty, constrains cooperation (Bock, 2015). On the one hand, a democratic government is less likely to commit domestic democide owing to its institutional constraints (Rummel, 1995). On the other, from rebels’ perspective, a government's democratic accountability incentivizes rebels to target civilians (Hultman, 2012). Therefore, I measured Democracy_t−₁ using the Polity IV score. This variable takes the value of 1 if the Polity 2 score at time t − 1 is greater than 5, and 0 otherwise. For all the control variables mentioned earlier, I used a one-year lag to avoid the problem of reverse causality.

While those control variables are likely to address problems of omitted variable bias, a potential underlying endogeneity problem has not been addressed. First, in the theoretical section, I have argued that external intelligence assistance tends to induce the recipient government to adopt the winning-hearts-and-minds strategy. However, one may suspect that governments pursuing the winning-hearts-and-minds strategy are more likely to attract intelligence assistance in the first place, reversing the causal mechanism. As external support is not one-sided but depends on both demand- and supply-side perspectives (Salehyan et al., 2011), the recipient government with long-term objectives may opt for external intelligence assistance owing to the matched demand- and supply-side objectives. Responsible governments that refrain from targeting civilians are most likely to receive intelligence assistance. To account for the endogeneity problem, I controlled for the logged and one-year lagged cumulative best estimate of civilian targeting by governments. Second, as external support is provided systematically to a recipient government (Salehyan et al., 2011; Sawyer et al., 2017), non-random assignment must be considered Thus, I used a matching technique. Matching permits pairing cases of treatment (external intelligence assistance) and control (no external intelligence assistance) where values of other covariates are similar. I excluded certain control variables during matching because including all covariates significantly reduced the number of observations. I chose covariates to be included based on Table C (Online Appendix), which shows where intelligence assistance goes. As coarsened exact matching (CEM) reduces imbalance, CEM was performed using the MatchIt package (Ho et al., 2011) in R. Continuous variables were matched with the class using the Sturges formula. After the matching, 40 observations remained.¹⁵ Figure B-1 (Online Appendix) shows the standardized mean difference for all samples and matched samples. The bias reduced after matching.¹⁶ However, inferences with CEM may be inefficient if too many treated units are discarded and, in such a case, widening the degree of maximum imbalance can be used as a remedy (Iacus et al., 2012). Thus, I also used 1/4 intervals for continuous variables in the matching. After this matching, 138 observations remained,¹⁷ although the quality of the matches was reduced (Figure B-2).

I checked the variance inflation factor score; no variables exceeded the score of 2.13 and the mean score of the variance inflation factor was 1.53, implying no multicollinearity problems.

Results and discussion

The dependent variable was a count variable of the number of civilian deaths caused by the government, which was overdispersed. The mean was 94.598 and its variance was 128407.4, yielding a positive skew. A binomial distribution could yield a more general estimator (King, 1989). However, as 332 out of 483 observations were coded as 0 for the dependent variable, I used the zero-inflated negative binomial (ZINB) model as the main model. Table D of the Online Appendix contains the results from the negative binomial regression. To address temporal dependence and serial correlation, I included a lagged and log-transformed dependent variable. Tables 1 and 2 present the results from the ZINB models. Table 1 reports a parsimonious model.¹⁸ I added the remaining control variables in the models in Table 2. The table reports no violence against civilians (Inflate) and violence against civilians (Count) stages separately. The inflate stage predicts the absence of violence against civilians. In Table 2, the same control variables were used for both stages, except for two variables—intensity_t−₁ variable, which was included only in the count stage, and the log of time since the last civil war, which was included only during the inflate stage. Table 3 presents the results from the negative binomial model with the matched sample.¹⁹

Table 1.

One-sided violence by the government (ZINB model) without control variables.

	Count component		Inflate component
Intelligence assistance for G_t−1	−0.705***	(0.253)	0.869*	(0.505)
ln Government violence _t−1	0.181***	(0.043)	−0.630***	(0.064)
ln time since the last civil war			3.617*	(2.031)
Constant	4.953***	(0.218)	1.777***	(0.184)
Observations	483
Wald χ²	23.68
Prob > χ²	0.000
Alpha	1.151	(0.173)
Log-likelihood	−1194.215
AIC	2404.431

*p < 0.10, **p < 0.05, ***p < 0.01.

Note: Standard errors are clustered on country.

Table 2.

One-sided violence by the government (ZINB model) with control variables.

	Count component		Inflate component
Intelligence assistance for G_t−1	−0.495**	(0.213)	0.589	(0.613)
External support for G_t−1	0.221*	(0.119)	0.360	(0.270)
ln cumulative gov violence_t−1	−0.125**	(0.054)	−0.042	(0.066)
Parity/stronger rebel_t−1	−0.848*	(0.462)	1.042**	(0.498)
Territorial control_t−1	0.240	(0.198)	−0.688*	(0.391)
Intensity_t−1	0.497***	(0.129)
ln Number of rebels_t−1	−0.772***	(0.252)	−0.328	(0.527)
ln Rebel violence_t−1	−0.035	(0.042)	0.066	(0.057)
ln GDP p.c._t−1	−0.106	(0.090)	0.194	(0.186)
ln Population_t−1	−0.141***	(0.052)	0.123	(0.189)
Democracy_t−1	−0.585**	(0.254)	0.791**	(0.402)
ln Government violence_t−1	0.208***	(0.034)	−0.544***	(0.058)
ln time since the last civil war			3.913*	(2.141)
Constant	9.169***	(1.406)	−1.769	(3.102)
Observations	483
Wald χ²	187.60
Prob > χ²	0.000
Alpha	0.881	(0.105)
Log-likelihood	−1158.746
AIC	2371.492

*p < 0.10, **p < 0.05, ***p < 0.01.

Note: Standard errors are clustered on country.

Table 3.

One-sided violence by the government (negative binomial regression) (after matching).

	CEM with Sturges		CEM with 1/4 intervals
Intelligence assistance for G_t−1	−1.880**	(0.842)	−1.359*	(0.813)
ln Government violence_t−1	0.525***	(0.108)	0.737***	(0.083)
Constant	3.207***	(0.592)	1.782***	(0.330)
Observations	40		138
Wald χ²	24.77		80.41
Prob > χ²	0.0000		0.0000
Alpha	9.918	(4.678)	11.201	(5.626)
Log-likelihood	−130.229		−374.635
AIC	268.458		757.271

*p < 0.10, **p < 0.05, ***p < 0.01.

Note: Standard errors are clustered on country.

The results strongly support the theory. The parsimonious model in Table 1 shows that the effect of external intelligence assistance is negative and statistically significant in the count portion of the model. This result did not change when the remaining control variables were added to the models in Table 2. To examine the substantive effect of external intelligence assistance for governments on one-sided government violence, Figure 2 presents the marginal effects results from Tables 1 and 2. Figure 2 indicates a change in the predicted number of instances of one-sided violence enacted by the government, with a change from 0 to 1 for the external intelligence assistance variables. In the Count model in Table 1, when a government received external intelligence assistance, the predicted number of incidents of one-sided violence enacted by the government decreased by 35.992 on average, with a 95% confidence interval of [ − 56.870, − 15.115], while holding values of other variables at their mean values for continuous variables and at their median values for dichotomous variables. The substantive negative effect was similar when the control variables were added to the model in Table 2. The external intelligence assistance reduced recipient governments’ violence against civilians by 22.971 on average with a 95% confidence interval [−42.424, −3.518].²⁰ On the one hand, considering that the median value of violence against civilians, which removed cases without violence, is 142, the reduction in 22.9 can be considered as a sufficient effect. On the other, considering that the standard deviation of violence against civilians, which removed cases without violence, is 590.983, the reduction only constitutes 3.875% of such a value. However, as standard deviation is affected by the mean value, which is influenced by the outlier, and if the histogram of the violence (Online Appendix Figure A) is considered, the reduction of 22.9 can be meaningful.²¹

Figure 2.

Marginal effects of intelligence assistance for a government on one-sided violence by the government.

In terms of robustness check, the main results were robust to different model specifications where I used a negative binomial regression in the Online Appendix. In Table 3, I estimated the model with the matched sample. Owing to the small number of observations remaining after the matching, Table 3 presents a parsimonious negative regression model without the control variables except for the logged lag of the dependent variable. The results indicate that the impact of external intelligence assistance on violence against civilians was negative and statistically significant. Figure 3 shows that the marginal effect in the case of CEM with Sturges is −53.970 with a 95% confidence interval of [−119.443, 11.504]. The marginal effect in the case of CEM with 1/4 intervals is −15.121 with a 95% confidence interval of [−29.296, −0.946]. The results from the matching analysis indicate that the impact of external intelligence assistance was not driven by any selection bias. However, regarding the effect of intelligence assistance on no violence against civilians by the government, it is not robust as observed in the inflate component of Tables 1 and 2. This may be attributable to the fact that intelligence assistance avoids mass violence against civilians by the recipient government through the winning-hearts-and-minds strategy, but the low intelligence capability, which compelled the recipient government to initially demand intelligence assistance, could result in targeting some civilians.

Figure 3.

Marginal effects of intelligence assistance for the government on one-sided violence by the government (after matching).

I also discuss the other substantive endogeneity issue that the matching technique may not have covered. As mentioned in the definition of intelligence, from the supplier's perspective, intelligence assistance is provided to fulfill the ultimate goal of state security. Thus, intelligence assistance is more likely to be provided to countries with high security threats, and insurgents that are difficult to locate. From the recipient's perspective, governments may demand external intelligence assistance when their own intelligence capability is low. The two reasons for providing intelligence assistance include severe identification problems and low intelligence capability of the recipient government, which may cause higher levels of violence against civilians. Although these two factors were not considered during the matching process, the potential direction of the omitted variable bias can be discussed. These two factors could generate a positive omitted variable bias in the relationship between intelligence assistance and violence against civilians. However, the results suggest a negative relationship between intelligence assistance and violence against civilians. Thus, considering the presence of a negative relationship notwithstanding the expected positive omitted variable bias, this finding further supports the hypothesis.

Next, I checked the extent to which the results were driven by US intelligence assistance. Table E (Online Appendix) presents the results. The US intelligence assistance was negative and statistically significant. When I included both the non-US intelligence assistance and US intelligence assistance, there was still a statistically significant negative relationship between the non-US intelligence assistance and the violence perpetrated by the government against civilians. These findings support the hypothesis.

As regards the control variables, first, the cumulative violence by the government had a negative and statistically significant effect in the count component of Table 2. The attributable reason could be that as cumulative civilian victimization increases, criticism by international communities increases. This could lead to material and political costs for the government, thereby preventing further violence against civilians (DeMeritt, 2012; Krain, 2012). Second, considering that external support could influence the balance of capability (Licklider, 1993; Regan, 2002; Sullivan and Karreth, 2015; Wood et al., 2012), general external support increases civilian killings by the government, as shown in the count component in Table 2. Although a case of civilian killings by the government, this finding aligns with the logic that external support decreases legitimacy costs (Fortna et al., 2018) or encourages agents’ commitment to the patron (Salehyan et al., 2014), thereby increasing violence against civilians. Importantly, the contrast between intelligence assistance and general support shows that while latter increases the risk of civilian killings by the government, the former diminishes such a risk. This finding shows that different objectives and aims of external support exist depending on the type of support. The winning-hearts-and-minds strategy in intelligence assistance is considered a softer approach than conventional military operations (Egnell, 2010). It may be the case that when external military support is provided, the focus becomes short-term military objectives rather than long-term political objectives, prioritizing the destruction of the existing insurgent capability and leading to indiscriminate violence.

Some variables related to the attributes of rebel groups had a significant influence on violence committed against civilians by governments. When a government faces a stronger rebel group, it is less likely to commit violence against its civilians (Inflate component, Table 2). A possible explanation is that weaker rebel groups have a greater incentive to target civilians to overcome collective action problems (Kalyvas, 2006; Wood, 2010). Given this strategy of rebel groups, the government may face a lower risk of losing popular support owing to the violence as civilians become less certain about who to hold accountable. Second, Table 2 shows that when a rebel group controls a territory, government is more like to commit violence against its civilians, although the relationship is only significant at the 90% interval level. The effect of territorial control is not observed in the count component. Perhaps, this may be because although territorial conquest encourages the government to eliminate enemy combatants (Downes, 2006), territorial control by a rebel group could develop an irregular war that is similar to a conventional war, wherein the former applies to a case where the civilian population is the “sea” in which guerrilla “fishes” swim (Valentino et al., 2004), leading to fewer identification problems. Compared with the case of an international war where a territorial conquest leads to the victimization of civilians owing to evicting them from the conquered territory (Downes, 2006), in a civil war, the government uses a strategy to maintain civilians in the territory.

As regards indicators of conflict characteristics, first, the effect of the intensity is positive and significant, as depicted by the count component in Table 2. The finding aligns with previous studies (Downes, 2006; Hultman, 2012; Salehyan et al., 2014; Wood et al., 2012). Second, contrary to a previous study (Wood et al., 2012), as the number of rebel groups increases, the government commits less violence against civilians. This may be because when there are multiple rebel groups, it becomes infeasible for the government to eliminate these groups through indiscriminate violence. Third, rebel violence against civilians is not statistically significant.

Finally, regarding country characteristics, first, a democratic country is less likely to commit violence against civilians, and the level of violence against civilians is lower. This supports the finding by Rummel (1995). Second, in alignment with Hultman (2012), a larger population results in less violence against civilians. The finding supports the logic presented by Downes (2007) that indiscriminate violence is effective when it is directed against a small population. In a country with a large population, indiscriminate violence would not be able to interdict the insurgent's sources of support; thus, the prospects of successfully eliminating insurgents through indiscriminate violence decreases. Finally, GDP p.c. was not statistically significant. However, the directions of coefficients in both the components in Table 2 align with a previous finding that weak state capacity results in more violence against civilians or mass killings (Anderton and Carter, 2015; Hultman, 2012).

Conclusion

Despite the efforts of external parties to provide intelligence to minimize violence by isolating the revolutionary elite from their followers (Joes, 2004), very little is known about whether such assistance minimizes the violence perpetrated by belligerents. This study aimed to improve our understanding of external intelligence assistance and the recipient government's violence against civilians through the logic of identification and the counterinsurgent strategy of intelligence assistance. I argued that intelligence assistance influences both identification problems and the recipient government's counterinsurgency strategy, usually emphasizing the winning-hearts-and-minds strategy. Accordingly, supporters encourage the recipient government to avoid civilian targeting, facilitating the establishment of reliable intelligence from the local population. The study revealed that when a government receives external intelligence assistance, government violence against civilians decreases. The results were found robust to different model specifications.

This study offers important insights for policymakers who must decide whether to offer intelligence assistance to another country to meet the security threats posed by rebel groups. I argue that external intelligence assistance provided to a government helps to avoid civilian casualties during a civil war. However, interpretation of the net conflict management effect of intelligence assistance must be done cautiously. When one-sided violence by rebel groups was used as the dependent variable using the same sample and variables in the Online Appendix, the result indicated no significant effect of intelligence assistance for the government on country-level violence against civilians by rebel groups. This implies the possibility that intelligence assistance could readily fail in terms of violence against civilians by groups rebelling against the government. The rebel groups may adopt a counterstrategy to undermine the external party's willingness to provide support and destroy leadership credibility. Therefore, in the short term, the external supporter is faced with the choice of whether or not to promote international cooperation that could undermine the reputation of the supporter. Moreover, although the finding showed that external intelligence assistance reduces the recipient government's violence against civilians in the short term, the effect of the winning-hearts-and-minds strategy itself is controversial. In the long-term, such a strategy sometimes fails or violates human rights (Hazelton, 2017). For instance, the British approach of the winning-hearts-and minds-strategy in Malaya involved high levels of force, resulting in human rights abuses (Dixon, 2009).

Future research could investigate the long-term consequences of intelligence assistance. For instance, in Uganda, intelligence assistance led to the establishment of the telecommunications infrastructure, which contributed to the country's development. As another long-term consequence of external intelligence assistance, future research could investigate how such assistance influences conflict outcomes or popular support for the recipient government. Another avenue for future research on violence against civilians is the long-term consequences of such aggressions. While existing research focuses on the short-term consequences, such as the decrease in insurgent attacks, investigating how violence against civilians during civil wars influences post-conflict elections would be relevant to the relationship between long-term political objectives and violence against civilians. Such long-term consequences of violence against civilians, such as political support or even the country's political stability, could ultimately influence the practice of providing intelligence assistance to mitigate potential security threats.

Supplemental Material

sj-docx-1-cmp-10.1177_07388942221129047 - Supplemental material for External intelligence assistance and the recipient government’s violence against civilians

Supplemental material, sj-docx-1-cmp-10.1177_07388942221129047 for External intelligence assistance and the recipient government’s violence against civilians by Wakako Maekawa in Conflict Management and Peace Science

Supplemental Material

sj-zip-2-cmp-10.1177_07388942221129047 - Supplemental material for External intelligence assistance and the recipient government’s violence against civilians

Supplemental material, sj-zip-2-cmp-10.1177_07388942221129047 for External intelligence assistance and the recipient government’s violence against civilians by Wakako Maekawa in Conflict Management and Peace Science

Footnotes

Acknowledgements

I gratefully thank the editor and anonymous reviewers for excellent comments and suggestions.

Funding

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

ORCID iD

Wakako Maekawa

Supplemental material

Supplemental material for this article is available online.

Notes

References

Agbiboa

(2019) Eyes on the street: Civilian joint task force and the surveillance of Boko Haram in northeastern Nigeria. Intelligence and National Security 33(7): 1022–1039.

Anderton

Carter

(2015) A new look at weak state conditions and genocide risk. Peace Economics, Peace Science and Public Policy 21(1): 1–36.

Arreguín-Toft

(2001) How the weak win wars: A theory of asymmetric conflict. International Security 26(1): 93–128.

Balch-Lindsay

Enterline

(2000) Killing time: The world politics of civil war duration, 1820–1992. International Studies Quarterly 44(4): 615–642.

Balch-Lindsay

Enterline

Joyce

(2008) Third-party intervention and the civil war process. Journal of Peace Research 45(3): 345–363.

Bock

(2015) Anglo-Soviet intelligence cooperation, 1941–45: Normative insights from the dyadic democratic peace literature. Intelligence and National Security 30(6): 890–912.

Bruneau

(2008) Democracy and effectiveness: Adapting intelligence for the fight against terrorism. International Journal of Intelligence and CounterIntelligence 21(3): 448–460.

Byman

(2006) Friends like these: Counterinsurgency and the war on terrorism. International Security 31(2): 79–115.

Charters

(2012) Canadian military intelligence in Afghanistan. International Journal of Intelligence and CounterIntelligence 25(3): 470–507.

10.

Christa

(2012) Alliance Formation in Civil Wars. Cambridge: Cambridge University Press.

11.

Clayton

Thomson

(2016) Civilianizing civil conflict: Civilian defense militias and the logic of violence in intra-state conflict. International Studies Quarterly 60(3): 499–510.

12.

Condra

Shapiro

(2012) Who takes the blame? The strategic effects of collateral damage. American Journal of Political Science 56(1): 167–187.

13.

Cunningham

Gleditsch

Salehyan

(2009) It takes two: A dyadic analysis of civil war duration and outcome. Journal of Conflict Resolution 53(4): 570–597.

14.

DeMeritt

JHR

(2012) International organizations and government killing: Does naming and shaming save lives? International Interactions 38(5): 597–621.

15.

Demmers

Gould

(2018) An assemblage approach to liquid warfare: AFRICOM and the ‘hunt’ for Joseph Kony. Security Dialogue 49(5): 364–381.

16.

Dixon

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

17.

Downes

(2006) Desperate times, desperate measures: The causes of civilian victimization in war. International Security 30(4): 152–195.

18.

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.

19.

Duyvesteyn

(2011) Hearts and minds, cultural awareness and good intelligence: The blueprint for successful counter-insurgency? Intelligence and National Security 26(4): 445–459.

20.

Eck

Hultman

(2007) One-sided violence against civilians in war: Insights from new fatality data. Journal of Peace Research 44(2): 233–246.

21.

Egnell

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

22.

Federici

Dalton

(2017) Operational security, accountability, and civilian casualties. Center for Strategic & International Studies, https://www.csis.org/analysis/operational-security-accountability-and-civilian-casualties.

23.

Fitzsimmons

(2008) Hard hearts and open minds? Governance identity and intellectual foundations of counterinsurgency strategy. Journal of Strategic Studies 31(3): 337–365.

24.

Forest

(2014) U.S. Military Deployments to Africa: Lessons from the hunt for Joseph Kony and the Lord’s Resistance Army. JSOU Report 14–4. MacDill AFB, FL: Joint Special Operations University.

25.

Fortna

(2015) Do terrorists win? Rebels’ use of terrorism and civil war outcomes. International Organization 69(3): 519–556.

26.

Fortna

Lolito

Rubin

MAR

(2018) Don’t bite the hand that feeds: Rebel funding sources and the use of terrorism in civil wars. International Studies Quarterly 62(4): 782–794.

27.

Galula

(1964) Counterinsurgency Warfare: Theory and Practice. Westport, CT: Praeger.

28.

George

(2013) Intelligence and Strategy. In: Baylis

Wirtz

Gray

(eds) Strategy in the Contemporary World. 4th ed. Oxford: Oxford University Press, pp. 151–169.

29.

Gill

(2007) Evaluating intelligence oversight committees: The UK intelligence and security committee and the “war on terror”. Intelligence and National Security 22(1): 14–37.

30.

Gleditsch

Wallensteen

Eriksson

, et al. (2002) Armed conflict 1946–2001: A new dataset. Journal of Peace Research 39 (5): 615–637.

31.

Griffith

(1963) Sun Tzu: The Art of War. New Yok: Oxford University Press.

32.

Hashim

(2006) Insurgency and Counter Insurgency in Iraq. Ithaca, NY: Cornell University Press.

33.

Hazelton

(2017) The “hearts and minds” fallacy: Violence, coercion, and success in counterinsurgency warfare. International Security 42(1): 80–113.

34.

Hendrix

Wong

(2013) When is the pen truly mighty? Regime type and the efficacy of naming and shaming in curbing human rights abuses. British Journal of Political Science 43(3): 651–672.

35.

Imai

King

, et al. (2011) Matchit: Nonparametric preprocessing for parametric causal inference. Journal of Statistical Software 42(8): 1–28.

36.

Högbladh

Pettersson

Themnér

(2011) External support in armed conflict, 1975–2009: Presenting new data. In: 52nd Annual International Studies Association Convention, Montreal, Canada, March 2011, pp. 16–19.

37.

Hultman

(2012) Attacks on civilians in civil war: Targeting the achilles heel of democratic governments. International Interactions 38(2): 164–181.

38.

Hurd

(1999) Legitimacy and authority in international politics. International Organization 53(2): 379–408.

39.

Iacus

King

Porro

(2012) Causal inference without balance checking: Coarsened exact matching. Political Analysis 20(1): 1–24.

40.

Inklaar

de Jong

Bolt

, et al. (2018) Rebasing “Maddison”: New income comparisons and the shape of long-run economic development. GGDC Research Memorandum GD-174. Groningen Growth and Development Center, University of Groningen.

41.

Joes

(2004) Resisting Rebellion: The History and Politics of Counterinsurgency. Louisvill, KY: University Press of Kentucky.

42.

Johnson

Wirtz

(2004) Strategic intelligence: Windows into a secret world. California: Roxbury.

43.

Kalyvas

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

44.

King

(1989) Event count models for international relations: Generalizations and applications. International Studies Quarterly 33(2): 123–147.

45.

Krain

(2012) J’accuse! Does naming and shaming perpetrators reduce the severity of genocides or politicides? International Studies Quarterly 56(3): 574–589.

46.

Lake

(2016) The Statebuilder’s Dilemma: On the Limits of Foreign Intervention. New York: Cornell University Press.

47.

Lemke

Regan

(2004) Intervention as influence. In: Diehl

(eds) The Scourge of War: New Extensions on an Old Problem. Ann Arbor, MI: University of Michigan Press, pp. 145–168.

48.

(2009) Democracy, autocracy, and expropriation of foreign direct investment. Comparative Political Studies 42(8): 1098–1127.

49.

Chen

(2021) Public opinion, international reputation, and audience costs in an authoritarian regime. Conflict Management and Peace Science 38(5): 543–560.

50.

Licklider

(1993) Stopping the Killing: How Civil Wars End. New York: New York University Press.

51.

Loewenstein

Prelec

(1992) Anomalies in intertemporal choice: Evidence and an interpretation. The Quarterly Journal of Economics 107(2): 573–597.

52.

Lyall

(2009) Does indiscriminate violence incite insurgent attacks? Evidence from Chechnya. Journal of Conflict Resolution 53(3): 331–362.

53.

Mack

(1975) Why big nations lose small wars: The politics of asymmetric conflict. World Politics: A Quarterly Journal of International Relations 27(2): 175–200.

54.

Mason

Krane

(1989) The political economy of death squads: Toward a theory of the impact of state-sanctioned terror. International Studies Quarterly 33(2): 175–198.

55.

Matei

Bruneau

(2011) Intelligence reform in new democracies: Factors supporting or arresting progress. Democratization 18(3): 602–630.

56.

Munton

Fredj

(2013) Sharing secrets: A game theoretic analysis of international intelligence cooperation. International Journal of Intelligence and CounterIntelligence 26(4): 666–692.

57.

Pape

(1996) Bombing to Win: Air Power and Coercion in War. Ithaca, NY: Cornell University Press.

58.

Pettersson

Öberg

(2020) Organized violence, 1989–2019. Journal of Peace Research 57(4): 597–613.

59.

Regan

(2002) Third-party interventions and the duration of intrastate conflicts. Journal of Conflict Resolution 46(1): 55–73.

60.

Rovner

Long

(2005) The perils of shallow theory: Intelligence reform and the 9/11 commission. International Journal of Intelligence and CounterIntelligence 18(4): 609–637.

61.

Rummel

(1995) Democracy, power, genocide, and mass murder. Journal of Conflict Resolution 39(1): 3–26.

62.

Salehyan

(2010) The delegation of war to rebel organizations. Journal of Conflict Resolution 54(3): 493–515.

63.

Salehyan

Gleditsch

Cunningham

(2011) Explaining external support for insurgent groups. International Organization 65(4): 709–744.

64.

Salehyan

Siroky

Wood

(2014) External rebel sponsorship and civilian abuse: A principal–agent analysis of wartime atrocities. International Organization 68(3): 633–661.

65.

Sawyer

Cunningham

Reed

(2017) The role of external support in civil war termination. Journal of Conflict Resolution 61(6): 1174–1202.

66.

Segell

(2009) Creating intelligence: Information operations in Iraq. International Journal of Intelligence and CounterIntelligence 22(1): 89–109.

67.

Shaver

Shapiro

(2021) The effect on civilian casualties on wartime informing: Evidence from the Iraq war. Journal of Conflict Resolution 65(7–8): 1337–1377.

68.

Sims

(2006) Foreign intelligence liaison: Devils, deals, and details. International Journal of Intelligence and CounterIntelligence 19(2): 195–217.

69.

Siroky

Dzutsati

(2015) The empire strikes back: Ethnicity, terrain, and indiscriminate violence in counterinsurgencies. Social Science Quarterly 96(3): 807–829.

70.

Smith

(2001) General Templer and counter-insurgency in Malaya: Hearts and minds, intelligence, and propaganda. Intelligence and National Security 16(3): 60–78.

71.

Sullivan

Karreth

(2015) The conditional impact of military intervention on internal armed conflict outcomes. Conflict Management and Peace Science 32(3): 269–288.

72.

Terpstra

Frerks

(2017) Rebel governance and legitimacy: Understanding the impact of rebel legitimation on civilian compliance with the LTTE rule. Civil Wars 19(3): 279–307.

73.

Tyler

(2001) A psychological perspective on the legitimacy of institutions and authorities. In: Jost

Major

(eds) The Psychology of Legitimacy: Emerging Perspectives on Ideology, Justice, and Intergroup Relations. New York: Cambridge University Press, pp. 416–435.

74.

Valentino

Huth

Balch-Lindsay

(2004) “Draining the sea”: Mass killing and guerrilla warfare. International Organization 58(2): 375–407.

75.

Walsh

(2008) Intelligence sharing for counter-insurgency. Defense & Security Analysis 24(3): 281–301.

76.

Walter

(2019) Explaining the number of rebel groups in civil wars. International Interactions 45(1): 1–27.

77.

Weinstein

(2007) Inside Rebellion: The Politics of Insurgent Violence. Cambridge: Cambridge University Press.

78.

Windsor

Charters

Wilson

(2008) Kandahar Tour: The Turning Point in Canada’s Afghan Mission. Toronto: Jhon Wiley.

79.

Winward

(2021) Intelligence capacity and mass violence: Evidence from Indonesia. Comparative Political Studies 54(3–4): 553–584.

80.

Wood

(2010) Rebel capability and strategic violence against civilians. Journal of Peace Research 47(5): 601–614.

81.

Wood

(2013) Opportunities to kill or incentives for restraint? Rebel capabilities, the origins of support and civilian victimization in civil war. Conflict Management and Peace Science 31(5): 1–20.

82.

Wood

Kathman

Gent

(2012) Armed intervention and civilian victimization in intrastate conflicts. Journal of Peace Research 49(5): 647–660.

83.

Worrall

(2017) (Re-)emergent orders: Understanding the negotiation(s) of rebel governance. Small Wars & Insurgencies 28(4–5): 709–733.

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.

0.03 MB

0.00 MB

0.42 MB