Jump to ContentJump to Main Navigation
States in DisguiseCauses of State Support for Rebel Groups$

Belgin San-Akca

Print publication date: 2016

Print ISBN-13: 9780190250881

Published to Oxford Scholarship Online: October 2016

DOI: 10.1093/acprof:oso/9780190250881.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2017. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: null; date: 24 February 2017

(p.159) Appendix 2 Dangerous Companions

(p.159) Appendix 2 Dangerous Companions

Cooperation Between States And Nonstate Armed Groups (NAGs)

States in Disguise

Belgin San-Akca

Oxford University Press

A Triadic Level Time-Series Dataset on Support for NAGs by States


v. April 2015

Belgin San-Akca


Department of International Relations

Koç University

Istanbul, Turkey

This research is funded by Marie Curie International Reintegration Grant (Proposal Ref. No. FP7-268486 and Grant ID No. REA.P3(2010)D/3202).


San-Akca, Belgin. “Dangerous Companions: Cooperation between States and Nonstate Armed Groups (NAGs),” ver. April/2015. nonstatearmedgroups.ku.edu.tr.

San-Akca, Belgin. States in Disguise: Causes of External State Support for Rebel Groups. New York: Oxford University Press. 2016.


San-Akca, Belgin. “Supporting Non-state Armed Groups: A Resort to Illegality?” Journal of Strategic Studies 32 (4) (August 2009): 589–613.

Research Team

  • Coders

  • Gizem Türkarslan (Research Assistant)

  • Barış Arı

  • Efe Çoban

  • Aytaç Denk

  • Melissa Meek

  • Ilayda Bilge Önder

  • Burcu Sağıroğlu

  • Paulina Schenk

  • Burcu Yiğiter

  • Data Portal and Website

  • Burak Demir (Research Assistant)

  • Ferit Demircioğlu

  • Özge Nur Kamcı

  • Suat Alper Orhan

  • Halit Sezgin

  • Irena Atnaguzina

  • Elif Burcu Gundoğdu

A2.1 Introduction

The Dangerous Companions Project (DCP) aims to realize two objectives with respect to the interactions between states and nonstate armed groups (NAGs). Building on a novel conceptualization of state and armed rebel group relations, it (1) collects data on state-rebel relations and (2) builds a continuously maintained data portal, which is accessible by the public and includes information on individual profiles of NAGs and their state supporters including the sources used for coding each case. The DCP operates on the understanding that NAGs are not mere agents of states, simply serving to realize the objectives of states, which support them. Obviously, when trying to provide support to an armed opposition group, a state engages in a decision-making process since it is a risky experiment. Nonetheless, states have historically used these alternative actors in international politics to pursue certain foreign policy goals, regardless of whether it proved to be a successful strategy. Similarly, the leaders and members of NAGs go through a decision-making process in deciding whether to receive external support from other states and in deciding which states are likely to provide the most effective grounds for extracting human and material recourses. In other words, the current conceptualization of state-NAGs relations goes beyond a simple treatment of these actors as a part (p.161) of proxy wars between two major states. When it comes to the motives of actors—states and NAGs—the theoretical framework is further developed in States in Disguise so as to capture the varieties of state-NAGs relations.

Each of these decision-making processes are referred to as States’ Selection Model and Rebels’ Selection Model, respectively, with the understanding that states might select the NAGs to which to provide support, but it does not tell us the entire story about the ways that NAGs select and extract resources from external states. For examples, Hamas received support from several countries, including Jordan, Syria, and Iraq, states that, since 1993, provided, on occasion, safe haven to its leaders and/or members. On the other hand, Hamas counted on supporters in the United States and the United Kingdom, who raised funds for the organization and transferred them to Hamas, despite the lack of observable evidence related to state sponsorhip from these countries. These are two distinct processes referenced by the framework developed in the current project. In the former, states choose intentionally to support Hamas. In the latter, states do not create channels with the goal of supporting the organization; rather, are exploited due to the individual freedom and liberties intrinsically found in democracies. Though in either case Hamas was able to acquire resources, they are not necessarily the same, thus they should be treated and coded separately.

NAGs refer to any armed opposition group that uses violent means to pursue certain political objectives. It is an overarching concept used to refer to ethnic and religious insurgents, revolutionary movements, and terrorists. Insurgency, terrorism, and revolution are various forms of violence that NAGs resort to in realizing their objectives. Therefore, it is not useful for the purposes of the current project to refer to these groups by politically loaded terms, such as revolutionaries or terrorists. It is commonly accepted now that NAGs acquire resources through various channels, such as support from elements in the diaspora, fund-raising by charity organizations, smuggling of drugs and weapons, and engaging in illicit trade, among others. The purpose of the current project is to examine general patterns concerning the nature of states in which NAGs most frequently conduct such activities to acquire these resources. Therefore, the current project stands in direct challenge to the post-9/11 assumption that the major source abetting terrorism and armed rebels, in general, is ungoverned territories or weak states. Designated borders between states divide the vast majority of the world’s territory. So it must be the case that more than 90 percent of the time, armed groups are operating within the borders of states, which claim a monopoly over the legitimate use of violent means. What kind of domestic and international environment attracts armed groups to certain states when it comes to acquiring resources within their territories?

The State-NAGs Cooperation dataset (NAGs dataset) is an attempt to operationalize external state support for NAGs that are engaged in violent conflict against one or more governments within or outside the state(s) in which they live. The groups and the states they target are borrowed from the UCDP/PRIO Armed Conflict database (ver. 4-2014a) (Gleditsch et al. 2002; Pettersson and Wallensteen 2014). In total, information is available on 455 NAGs in existence during the post-1945 period. The first version of the dataset (ver.4/2015) covers the period between 1945 and 2010. Each case in the dataset is a triad that involves a NAG, a target (the country subject to violence by the NAG), and a supporter (the country that has provided one or more types of support, such as training camps, safe havens, arms and equipment, funds, and troops to the NAG). A detailed description of each variable (p.162) is given in the table below. Next, case selection, each variable, and corresponding coding rule have been explained in detail. A triad is listed for the entire period of a NAG’s activity if a state ends up supporting it for at least one year during the period it is active.

The other contribution of this project is the public data portal site with information on each rebel group and states from which they acquire resources. Given that it is a challenging task to find and code information about state support, whether it is an intentional act on the part of a state or a case of rebels selecting a country from which to extract resources, the best one could do is to transparently share the sources used to collect and code such information. The DCP data portal presents profiles of each NAG, listing thousands of sources used to gather the required information. In addition, it provides data visualization by using maps and profiles for each group.

A2.1.1 How Is the NAGs Dataset Placed Next to Other Existing Datasets on the Issue of External State Support?

There are several other existing datasets on nonstate armed groups and third-party interventions. Cunningham, Gleditsch and Salehyan’s Non-state Actor dataset (NSA) (2009) is a dyadic dataset with information on each NSA’s military strength and capacity, leadership characteristic, popular support, and political linkages as well as external sponsorship. However, external state support is not broken into diverse types. We only know if explicit or implicit support is given by external states. Similarly, UCDP’s External Support Data is another dyadic dataset, which also codes external state supporters that give support to a NAG in a given year from 1975 to 2010 (Hogbladh et al. 2011). These data are limited in their temporal domain, not going beyond 1975 and not distinguishing between state selection and rebel selection processes as described in detail later. Finally Regan and Aydin’s data on external interventions into civil conflicts look particularly at military, economic, and diplomatic third-party interventions (2006). This dataset takes intrastate conflicts as one single unit and does not distinguish between rebel groups when multiple groups are involved in a given intrastate conflict (Regan et al. 2009). The NAGs dataset lists each rebel group and contains information on its ideational characteristics, detailed objectives, and whether support emerges as a result of state or rebel selection processes.

A2.1.2 What Is New about the DCP and NAGs Dataset?

The NAGs dataset differentiates between state and NAGs selection cases. Though it might be misleading to refer to interactions emerging at the end of both processes as state support or cooperation, to some extent it is justified since states turn out to be de facto supporters of rebels at the end of the rebel selection process. We do not have a way of understanding the intentions of each state about whether its politicians really want to support or curb support for a given NAG, if they found themselves selected by them. As a result, the healthiest way to go about coding such information relies on observed behavior. In the case of NAGs selection, some countries turn into facilitators of violent operations of NAGs despite the absence of state sponsorship in these countries. Therefore, it is not wrong to refer to both cases as state support, which emerges either from a state’s intended or unintended acts. In addition, the NAGs dataset starts coding a NAG as soon as it declares a name regardless of whether, at the onset, it conducts violent operations, while the ACD takes the start of a NAG as the year in which at least one battle-related death occurs for (p.163) the first time. This is despite the fact that sometimes NAGs might have existed long before they resort to violence. Another contribution of this project is the detailed coding of group objectives and ideational characteristics.

Byman (2005b) took the initiative in classifying several paths by which armed opposition movements, specifically terrorists, acquire resources from states. He refers to passive support of terrorists by states under three conditions: “(1) the regime in question itself does not provide assistance but knowingly allows other actors in the country to aid a terrorist group, (2) the regime has the capacity to stop this assistance or has chosen not to develop this capacity, and (3) often passive support is given by political parties, wealthy merchants, or other actors in society that have no formal affiliation with the government” (Byman 2005a, 118). Except for the last criterion, the first two are very difficult to determine in each case even if given extensive time and resources to do so. It is very difficult to determine whether “a government chooses not to develop capacity” to curb support of rebels within its territories. And it is very difficult to know whether a regime or government “knowingly allows other actors in the country to aid” rebel groups.

Though the way to conceptualize passive support of terrorism is to be commended, the ambitious data collection and coding project specified under DCP requires developing a standardized set of criteria that will make coding a large number of cases possible. Such criteria can be developed if we rely on what is observed rather than what is intended, since the intentions do not always lead to observable outcomes. Rather than treating capacity as a selection criterion when coding cases, it is better to treat it as a variable. This way it is possible to detect whether NAGs or armed rebels select states with weak or low capacity to organize their activities. Each variable is defined and explained in the following sections.

A2.2 List of Variables

Variable Short Name

Variable Long Name & Measurement


Triad identifier—Unique Triad Id Number for a given NAG, its target, and supporter


Dyad identifier—UCDP/PRIO Dyad Code


Conflict identifier—UCDP PRIO Conflict ID


Year of observation

Yr_Active (not coded)

1—the year the group was formed, 2—the year in which at least one BRD (battle-related death) is observed (Startdate1 in UCDP/PRIO), 3—the year in which at least 25 BRDs are observed (Startdate2 in UCDP/PRIO), 4—formation date and one BRD year are the same, 5—formation date, one BRD, and 25 BRD year are the same year, 6—one BRD and 25 BRD year are the same.


Foundation year


Year in which at least one BRD is observed in UCDP/PRIO


Year in which at least 25 BRDs are observed in UCDP/PRIO


Target country name—abbreviation


Numeric Gleditsch and Ward ID of target country


Numeric COW ID of target country


Full name of the NAG


Numeric code of the nonstate armed group


Numeric UCDP/PRIO Actor code of the NAG




Name of territory


Identity of NAG (Numeric): 1: NOID, 2: Ethno-nationalist, 3: religious, 4: leftist, 5: other


Ethnic identity of the NAG (Name)


Religion the nonstate actor belongs to (Name)


Whether a NAG is a leftist revolutionary group (Binary)


Whether a NAG is a right-wing group (Binary)—fascist or conservative or other (specify)


Whether a NAG has democratic aspirations (Binary)


Whether a NAG aspires to establish an authoritarian regime (Binary)


Whether a NAG is supporting a dictatorial regime (Binary)


Whether a NAG is supporting a military regime (Binary)


Whether a NAG aspires to establish a theocratic regime (Binary)

NAGObj 1-6

  • Objective of the NAG (Numerical): 1: toppling an existing leadership, 2: change of regime type (transition from autocracy to democracy or the reverse regime change), 3: demands for autonomy, 4: secession/territorial demand, 5: demands for policy change, 6: Other—specify

  • Each category is coded as separate binary variables. A group may have more than a single objective.


Political party dummy—whether the group has a party wing


Political party name


Country providing support to a NAG—abbreviation


Gleditsch and Ward country code of the supporter


COW country code of the supporter

  • StateSup

  • (State Selection)

Binary variable of state selection cases of support

S_Precision 1-4

Support precision 1: supporter state clear intention, 2: reliable sources document support, 3: support is highly suspected by reliable source, 4: target state accuses supporter state without documentation


Safe haven to members


Safe haven to leadership




Training camp




Weapons and logistics aid


Financial aid


Transport of the military equipment, military advice




Any other kind of intentional support

  • De facto Support

  • (NAGs Selection)

Binary variable of de facto Support (NAGs Selection Cases)

DS_Precision 1-4

De facto support precision 1: supporter state clear intention, 2: reliable sources document support, 3: support is highly suspected by reliable source, 4: target state accuses supporter state without documentation


Safe haven to members


Safe haven to leadership




Training camp




Weapons and logistics aid


Financial aid


Transport of military equipment, military advice


Any other kind of de facto support


Domestic support dummy


1—not much confident, 2—somehow confident, 3—confident


Support termination dummy


Target country pressure


International community pressure


Regime change in the target country


Regime change in the supporting country


Leadership change in the supporting country


Group ceased activity


Group signed a cease-fire


Group turned into a political movement


Other termination—explain (p.164) (p.165)

(p.166) A2.3 Cases

A2.3.1 Identifying the Targets

To identify the states that have been targets of nonstate violence in the post–World War II period, I rely on the UCDP/PRIO Armed Conflict dataset ver. 4-2014a, 1946–2014 (Gleditsch et al. 2002; Themnér and Wallensteen 2014).

A2.3.2 Identifying the Groups

The groups included in the dataset have to meet the criterion of “25-battle related deaths” according to the UCDP/PRIO Armed Conflict dataset (ACD). I use the same groups as in UCDP/PRIO Armed Conflict dataset ver.4-2014a, 1946–2014 (Gleditsch et al. 2002; Pettersson and Wallensteen 2014). The ACD identifies an opposition organization as following: “Any non-governmental group of people having announced a name for their group and using armed force to influence the outcome of the stated incompatibility.” Two types of incompatibility are identified by the ACD: (1) “incompatibility concerning government: type of political system, the replacement of the central government, or the change of its composition,” (2) “incompatibility concerning territory:Incompatibility concerning the status of a territory, e.g. the change of the state in control of a certain territory (interstate conflict), secession or autonomy (internal conflict)” (Gleditsch et al. 2002; Pettersson and Wallensteen 2015). As described below, the NAGs dataset has a further detailed coding of group objectives.

A2.3.3 Identifying the Period

The temporal domain in the dataset is identified as the “opportunity period,” that is, the period during which a state has the opportunity to support it. This basically stems from the fact that a group has to be active in order for a state to have an opportunity to support a group. Therefore, the year variable specifies the years during which a group is active. In some cases, a group ceases activity for a while and then reinstates violence. As long as the group and the government it targets do not sign an aggrement and the dispute is not entirely resolved, these brief cease-fires are treated as activity years. In other words, once a group is caught to declare a name, whether or not it resorts to violence initially, the opportunity to support the group emerges. Concerning the information about conflict termination, this dataset draws upon the UCDP Conflict Termination dataset (ver. 201-1) (Kreutz 2010). The UCDP Conflict Termination dataset identifies seven types of termination: Peace agreement, cease-fire agreement with conflict regulation, cease-fire agreement, victory, low activity, other, and joining alliance. As long as the conflict does not end with the first four types of outcome, it is not considered that a NAG terminated.

UCDP/PRIO Armed Conflict Dataset codes two different start years for a group. The first year (Startdate1) is coded when there is at least one battle-related death in a conflict. And a second start date (startdate2) is coded when there are at least twenty-five battle-related deaths in a conflict. The second start date is considered as the onset of a conflict within a given government-opposition group conflict. And there might be multiple conflicts within a given government-opposition organization conflict. The (p.167) current data code the first year when a group is formed as the beginning of the activity period regardless whether violence is observed. This way it is also possible to observe whether receiving external support prompts groups to resort to violence. Of course, this is attainable when the formation year is before the first year in which a minimum of one battle-related death is observed. See description of Yr_Active variable below for a more detailed coding.

A2.3.4 Definition of Variables:

  • TriadID: Each row in the dataset represents a triad year. A triad consists of a target, a supporter, and a NAG that resorts to violent means against a country (target) to achieve its political objectives. This number is constituted in the following way: target COW ID*1,000,000 + NAGcode*1,000 + potential supporter COWID. For example, in calculating the triad ID number for Algeria (target), Armed Islamic Group (NAG), and France (supporter);

    • COW ID for Algeria: 615

    • NAGcode for the Armed Islamic Group: 2

    • COWID for France: 220

    • Triadid = [(615 * 1000000) + (2 * 1000) + 220] = 615002220

  • DyadID: A unique identifier generated by UCD/PRIO Armed Conflict Dataset Dyadic Codebook to identify each dyad of a rebel group and its target state (Gleditsch et al. 2002; Pettersson and Wallensteen 2015).

  • ConflictID: Conflict identifier from the UCDP/PRIO dataset.

  • Year: The year of observation. Each triad is listed for the period during which a NAG is active, beginning from its formation year regardless whether there is violence in that year.

  • Yr_Active: 1: the year groups was formed, 2: the year in which at least one BRD (battle-related death) is observed (Startdate1 in UCDP/PRIO), 3: the year in which at least 25 BRDs are observed (Startdate2 in UCDP/PRIO), 4: formation date and one BRD are the same year, 5: formation date, one BRD, and 25 BRDs are the same year, 6: One BRD and 25 BRDs are the same year.

When coding the activity periods of some NAGs, one issue concerns the groups that capture control of the government for a brief period during their insurgency. For example, the Armed Forces Revolutionary Council (AFRC) staged a coup against the government of Sierra Leone in 1997 and controlled the capital of Freetown until forces of the Economic Community of Western African States Monitoring Group (ECOMOK) drove them out in 1998. In other words, they did not secure acknowledgment by the international community. In such cases, the rebel group is not considered as representing the government. Rather, its activity is considered as continuing in these periods even though the group controlled the government. Another case is the Khmer Rouge in Cambodia. Although the Khmer Rouge began as a rebel group, it won control of the Cambodian government between 1975 and 1979. It allied with the North Vietnamese government and no regional and/or international efforts were launched to remove them from (p.168) power. In such cases, the rebel group is not considered as continuing its activity in the period it is in charge of the government.

  • Foundyr: The UCDP/PRIO dataset starts coding a group after at least one BRD death is observed. However, some groups existed long before a battle-related death occurred. Therefore, this variable is coded separately.

  • Startdate1: Adopted directly from UCDP/PRIO dataset.

  • Startdate2: Adopted from UCDP/PRIO dataset.

  • Target: The country facing a threat from a NAG.

  • TarNum_GW: Numeric Gleditsch and Ward ID of target country.

  • TarNum_COW: COW country code for target.

  • NAG_name: Full name of the group.

  • NAGcode_1: The numeric code of the NAG (ranges from 1 to 455). In the end, a list of groups, their codes in the dataset, and the period of activity are listed. As long as the group did not sign an agreement with the government, it is not considered terminated.

  • NAGcode_2: Numeric UCDP/PRIO code of the NAG.

  • Incomp: Incompatibility as coded by UCDP/PRIO ACD.

  • Terr: Name of territory under dispute (from UCDP/PRIO).

  • NAGID: Identity of the NAG (Numeric):

    • 1: NOID

    • 2: Ethno-nationalist

    • 3: Religious

    • 4: Leftist

    • 5: Other (specify)

These categories are not mutually exclusive. In many cases, it is possible to associate a NAG with multiple identities. Groups such as the Palestinian Islamic Jihad and Hamas can be identified both as an ethnic and as a religious-oriented group. The ideational identity of these groups has been recorded for both variables of ethnic and religious identity. If a group does not associate itself with any identity and/or ideology, such as the Cocoyes in the Democratic Republic of Congo, then it is coded as not having any ideational identity. Although the group has aspirations to change the leadership, it does not carry out propaganda for a specific ethnic or religious group and/or political ideology. Rather it aspires to be inclusive by bringing together multiple ethnic groups in the southern Congo.

Determining the ethnic, religious, or ideological aspirations of each group requires extensive analysis of the components of the ideational spectrum with which they identify. For example, the Moro National Liberation Front (in contrast to the Moro Islamic Liberation Front) does not aspire to found an Islamic state in the southern Philippines. Rather it seeks autonomy for areas populated by Moro Muslims. This group is coded as an ethno-nationalist group. Although Islam is a part of their identity, it is not the main driving force. In cases in which there are multiple identities, the one that is overwhelmingly emphasized is coded in addition to recording each component of a group’s identity under the corresponding variables below. To attract international attention and support from major powers such as the ex-Soviet Union, (p.169) China, and the United States, some NAGs have claimed to follow a “communist” or a “democratic” ideology despite the fact that their actions clearly do not conform to either of these. For example, the National Liberation Front in Algeria has been given different ideological labels over the course of its history, including calling it an anti-colonial, nationalist, and communist movement. In such cases, their discourse has been disregarded and their actions have been taken as a basis for coding.

NAGeth: Ethnic identity of the group. Ethnic identity is not only coded for ethno-nationalist movements, but for all movements. The Ethnic Power Relations dataset (EPR) was used to code the ethnic identity of each group (Cederman et al. 2010). The EPR dataset is hosted by the GROWup portal (http://growup.ethz.ch/), which matches each UCDP/PRIO Armed Conflict data rebel group with corresponding ethnic groups in each target state (Girardin et al. 2015). However, at the time of the coding of the NAG dataset, ethnic identities of groups were not yet available. Therefore, each NAG’s ethnic identity was coded according to EPR identities by the Dangerous Companions Project team. In cases in which a group’s ethnic identity was not clear, the identity of the group’s leader was coded instead. For multiethnic NAGs whose composition included members with more than three different ethnic backgrounds, only the top three ethnicities found in their country of origin have been used in matching them with supporter major groups.

NAGrel: Religious identity of the group. It does not necessarily mean that the group seeks to establish a religious regime or identifies itself openly with a religious affiliation. For example, the Kurdistan Workers’ Party (PKK) never cites religion as part of its identity, yet it is coded in accordance with the religious identity to which group members overwhelmingly belong. The religion categories are coded according to the indicators of the World Religion Project (Maoz and Henderson 2013). If the specific religion with which the group identifies itself is readily found (e.g. Sunni, Shia, Catholic, Orthodox, etc.), such labels are used in coding. Otherwise, this variable is coded as broad religious identities (e.g. Muslim, Christian). The categories coded for religion can be found in the list below. Similar to the NAG ethnic identity variable, for multireligious NAGs, whose composition included members with more than three different religious backgrounds, only the top three religious identities found in their country of origin have been included.

  • NAGleft: Dummy variable for whether a NAG is a leftist revolutionary group or not.

  • NAGright: If the NAG has fascist or conservative or other aspirations that we may relate to right-wing views.

  • NAGdemoc: If a group claims that it has democratic aspirations, this variable is coded as “1,” and “0” otherwise. There is a problem that any group may argue that it will bring democracy. Usually any ethnic group that has aspirations for secession or partial autonomy has demands about reforms advancing individual political rights and liberties. Indeed, such groups usually claim to represent ethnic minorities, such as the Basque people who live in Spain (ETA) or the Kurdish people who live in Turkey (PKK). They claim to seek further rights for the minorities they claim to represent. When we code this variable, we do not take into consideration such demands. Both PKK and ETA do not seek primarily to bring democracy to the country in which they live. Rather, they prioratize other (p.170) aspirations for their own ethnic communities. An example of a group that aims to bring democracy to a country is the All Burma Students’ Democratic Front (ABSDF). The primary motivation of ABSDF has been to overthrow the military regime in Burma and establish a democratic regime.

  • NAGauth: If a group is fighting for a form of autocratic regime other than theocracy, dictatorship, and military government, this variable is coded “1,” “0” otherwise.

  • NAGdict: If a group is fighting for a dictatorial regime, this variable is coded as “1,” otherwise “0.”

  • NAGmil: If a group is fighting for a military regime, this variable is coded as “1,” “0” otherwise. Most NAGs that carry out military coups fall under this category.

  • NAGtheo: If a group is fighting for a form of theocracy, this variable is coded “1,” “0” otherwise. Most fundamentalist Islamist groups fall under this category.

  • NAGobj: Goals pursued by a NAG might be various and can change over time. The UCDP/PRIO dataset codes the demands for government change and autonomy as main forms of incompatibility over government and territory, respectively. The objective of the group is recoded in a more detailed manner in the NAGs Dataset. A demand for a change of leadership is different from a demand for a change of regime. By the same token, the ACD take the stated incompatibility in the beginning of the conflict as if it continues until the end. It is known that various groups can change their objectives throughout the conflict and this in itself is a very important variable that is captured in the NAGs Dataset.

    • 1: toppling an existing leadership

    • 2: change of regime type (transition from autocracy to democracy or the reverse regime change)

    • 3: demands for autonomy

    • 4: secession/territorial demand

    • 5: demands for policy change

    • 6: Other—specify (p.171)

  • PolParDummy: Binary variable indicating whether there is a political party affiliated with the group in a given year. Affiliation is described as whether the party shares similar aspirations as the group and there is evidence that the party leaders communicate with militants.

  • PartyName: Name of the political party affiliated with a NAG.

  • Supporter: The state that supports the NAG in a given year.

  • SupNum_GW:Numeric Gleditsch and Ward ID of supporter country.

  • SupNum_COW: The COW country code of the supporter.

Table A2.1 World Religion Project Religion Categories Coded for NAGs

Variable Label



Christianity—Protestants—No. of Adherents


Christianity—Roman Catholics—No. of Adherents


Christianity—Eastern Orthodox—No. of Adherents


Christianity—Anglican—No. of Adherents


Christianity—Others—No. of Adherents


Christianity—Total No. of Adherents


Judaism—Orthodox—No. of Adherents


Judaism—Conservatives—No. of Adherents


Judaism—Reform—No. of Adherents


Judaism—Others—No. of Adherents


Judaism—Total No. of Adherents


Islam—Sunni—No. of Adherents


Islam—Shi’a—No. of Adherents


Islam—Ibadhi—No. of Adherents


Islam—Nation of Islam—No. of Adherents


Islam—Alawite—No. of Adherents


Islam—Ahmadiyya—No. of Adherents


Islam—Other—No. of Adherents


Islam—Total No. of Adherents


Buddhism—Mahayana—No. of Adherents


Buddhism—Theravada—No. of Adherents


Buddhism—Other—No. of Adherents


Buddhism—Total No. of Adherents


Zoroastrian—Total No. of Adherents


Hindu—Total No. of Adherents


Sikh—Total No. of Adherents


Shinto—Total No. of Adherents


Baha’i—Total No. of Adherents


Taoism—Total No. of Adherents


Confucianism—Total No. of Adherents


Jain—Total No. of Adherents


Syncretic Religions- Total No. of Adherents


Animist Religions—Total No. of Adherents


Non-Religious—Total No. of Adherents


Other Religions—Total No. of Adherents

(p.172) 2.3.5 Intentional versus De Facto Support

Support is an action that implies an intentional act on the part of an external actor. The post-9/11 debate about terrorism has focused on weak states and how they have become safe havens for various terrorist organizations. Yet the fact that weak states may turn into safe havens for terrorists or other NAGs does not qualify for “support of nonstate violence.” State capacity should not be used as a coding criterion when deciding whether or not a form of support is provided. It should be treated as an independent variable in explaining the ability of a government to control its borders in determining whether they become as safe havens or sources of other forms of support for NAGs. As previously argued, the best way to code whether a NAG is able to acquire resources from other states is to focus on the observable outcomes rather than intentions, since the latter can be hard to determine. Therefore, the current coding protocol treats cases in which evidence is clear that states create channels to abet certain groups as intentional support emerging through state selection process. Furthermore, multiple reliable sources are used to confirm information for each case of support. On the other hand, when a rebel group is able to operate within the borders of a country without a clear evidence of sponsorship of that country’s state, it is treated as an incident of NAGs selection, or de facto support. In that case, two criteria have been used to code state support of NAGs:

  1. (1) Whether there was an observable indication that a given NAG was operating within the borders of a country; i.e. leaders finding safe havens, a source for raising funds, weapon smuggling, etc.

  2. (2) Whether the government or leadership in a given country knowingly creates channels to support a given NAG. For example, Egypt knowingly let the fedayeen operate within its borders until the Suez Crisis, after which the Egyptian government expelled them from the country.

  3. (3) In the absence of confirmable information that the government or leadership in a given country provides support to a given NAG or creates channels to facilitate its activities, it is assumed to be de facto support, i.e. NAGs selecting the countries in which to acquire resources to sustain their operations against their targets.

The following set of sources are used in confirming intentional state support and de facto support incidents:

  1. (1) News wires and press releases from credible sources, such as Agence France-Presse (AFP), the United Press International (UPI), Xinhua News Agency, Reuters, Aljazeera, CNN, BBC Monitoring, etc.

  2. (2) Major newspapers, such as the New York Times, the Washington Post, the Guardian, the Financial Times, the Globe and Mail, etc.

  3. (3) Scholarly research articles, books, book chapters, and research notes published in academic and peer-reviewed journals.

In coding state selection cases, the emphasis was on whether a government directly provides assistance to facilitate violent conduct of a NAG. In other words, when making a decision about coding a case of support, some evidence was required with respect to the government or a (p.173) political actor or organization formally affiliated with the government providing support. The Revolutionary United Front (RUF) targeted Sierra Leone between 1991 and 2001. It received intentional, direct support from Liberia, Burkina Faso, and Libya in the form of safe haven for members and leaders, funds, arms, logistics, and troops. In coding state support for RUF, the following is an exemplary statement adopted from a news source:

Nine years ago, the state's collapse, the poverty of its people and the eternal tussle for Sierra Leone's diamonds led to war. A cashiered army corporal named Foday Sankoh joined his vague notions of revolution with money and guns from Libya and Liberian warlord--now president--Charles Taylor to form the Revolutionary United Front. The RUF seized diamond fields, smuggled gemstones and became one of Africa's most thuggish militias. (Rupert 2000)

In 1991, while still fighting in Liberia, Taylor helped launch the civil war in Sierra Leone by providing troops, training and supplies to Foday Sankoh, leader of the Revolutionary United Front. Richie was assigned to Sankoh's forces for their first incursion into Sierra Leone and has been fighting here ever since. (Douglas 2001)

Another task when coding support is determining the time and duration of support. In some cases it is easy to find out from the sources used for collecting data on a particular NAG. Yet it can sometimes prove challenging to code the time and duration of support. When an external support is mentioned in the sources but the period of support is not clear, the release date of the sources is used as an approximate date of support. The All Tripura Tiger Force (ATTF) was an ethno-nationalist group fighting the Indian state between 1992 and 2010 in seeking an independent state for the Tripuri people. The sources, dated mostly in 2002, 2008, and 2010, pointed out that ATTF received support from Bangladesh and Pakistan. Thus the support is coded as continuing from 2002 to 2010 given that we confirm this information with some scholarly case studies.

Furthermore, for each type of support coded, a precision level is determined. For the ATTF and Indian conflict, the above-stated sources mentioned the accusations or allegations of the Indian government. Therefore, when coding, the lowest precision level was assigned to this particular group. Another example is the Nicaraguan Democratic Force (FDN), which fought against the Sandinista regime between 1981 and 1990. Between 1981 and 1984, support for the FDN was authorized by the U.S. Congress (Cody 1984; Woodward et al. 1984). Once Congress stopped overt support channels from the U.S. government, the administration of President Reagan approved covert efforts to support the FDN, which resulted in the notorious Iran-Contra affair. It is a clear case of state support with a very high precision level. Indeed, congressional reports indicate clearly that support was given during the specified period.

  • Precision (S_Precision & DS_Precision): To specify how confident the coder is that there is evidence of active support, the variable receives the following rating:

    • 1: The supporter stated its intention and/or type of support, and/or the support was officially documented by that state or another.

    • 2: A journalist on the field, a scholar, or a media outlet records the support and provides convincing evidence and there are other sources that confirm this information. (p.174)

    • 3: Support is highly suspected by a reliable source (such as a journalist, scholar, or media outlet) but cannot be confirmed by other sources.

    • 4: One state accuses another state of supporting a group, but it cannot provide official documentation beyond allegations.

  • SUPPORT TYPES: States selection cases are denoted by “S” and NAGs selection cases are denoted by “DS” (De facto support).

For each rebel group, a table of direct citations, including stories and news from reliable sources, has been created by using the Lexis-Nexis academic web program, Keesing’s Archives, and published secondary sources, including political science journals, journals focusing on particular regions of the world, books, and book chapters. Each coder received training and was given a sample NAG with which to code. After inter-coder reliability is confirmed at the end of the sample group coding, they were assigned groups on a weekly basis. Regular meetings were held with the coders to respond to questions and concerns. In Lexis-Nexis Academic, a keyword search was done for each group for all available dates. Each coder was given a questionnaire, which is available on the DCP website, with directions and guidance about how to conduct research in online databases and sources to find and collect the required data. To determine the supporters and the type of support provided, the following keywords have been searched in the Lexis-Nexis categories “Major U.S. and World Publications,” “News Wire Services,” and “TV and Radio Broadcast Transcripts” with each group’s name: support, assistance, sponsor, safe haven, sanctuary, training camps, camps arms, weapons, funds, troops.

A2.3.6 Coding Rules for State Support

After a preliminary analysis and coding of 20 percent of all the rebel groups and their supporters, the following rules were applied with respect to some ambiguous forms of support referenced in the sources used for data coding:

  1. 1. If a state provides health services to a group’s members or leaders, it qualifies as providing safe haven for members or leaders.

  2. 2. In some instances, states become hosts to negotiations and meetings between a group’s leaders and the target government. This does not qualify as a form of support.

  3. 3. Some states host the headquarters of rebel groups. These headquarters organize propaganda and fund-raising activities of a group and provide communication with the militants at home. This is coded as a form of support under the name of “headquarters or opening offices.”

  4. 4. Some states host TV channels and radio stations operated by rebels that are used to disseminate information about the group. This does not qualify as a form of support for the purposes of this project but can be the subject matter of another research project.

  5. 5. A state may provide one or more of the specified support types.

  6. 6. Some specific cases proved to be particularly complex. One such case was Palestinian militant groups finding safe havens in Lebanon. Lebanon was under the occupation of Syria and Israel between 1979 and 2005 and between 1982 and 2000, respectively. Prior (p.175) to the civil war and Syrian occupation, Lebanon served as a host territory for several Palestinian militant groups as well as Palestinian refugees, primarily in the southern part of the country. It is not clear the degree to which the Lebanese government might have resisted the pressures of strong Arab states, such as Egypt and Syria; yet, under the Cairo agreement, the country was designated a safe haven for several groups. Initially, it is coded as a state support case. Yet later after foreign occupation, no support was coded for Lebanon and for several groups, such as the Palestinian Islamic Jihad, PFLP, and Fatah. Furthermore, Hezbollah emerged in southern Lebanon in response to the Israeli occupation. So Hezbollah did not choose to reside in southern Lebanon. In other words, it could not have emerged in Jordan or Egypt. Yet if Hezbollah sought support from external states, such as Iran and Syria, then it is coded as a clear case of state support by Iran and Syria.

  7. 7. In some cases, NAGs establish a presence in a foreign country with the assistance of another rebel group targeting that foreign country. The Maoist Communist Center of India (MCC) operated safe havens and training camps in Nepal. The Nepalese government was not involved in this assistance; Maoist insurgents fighting against Nepal helped the MCC to establish facilities inside Nepal. This is a choice on the part of the MCC, thus it is coded as de facto support from Nepal to the MCC against India.

    SafeMem: Providing safe havens to members. A certain number of militants are present within the territories of a state or they establish some bases. Safe havens are defined as “geographical spaces where non-state armed groups members are able to establish organizational and operational base that allows them to engage in financing activities, developing a communications network for command and control, achieving access to weapons and developing logistics network to enable travel, the movement of money and weapons” (Kittner 2007, p. 308). Geographical spaces in which militants acquire operational space for training are coded separately as “training camps.” This does not annul the fact that training camps are also operational spaces. Mere refugee camps do not qualify as safe havens. There needs to be some proof that militants are found in these camps and operate from these places.

Providing safe havens to members of a rebel group is different from providing training camps or access to existing camps. More often than not, the neighbors of a state that experiences civil war or ethnic conflict end up accepting refugees within their own borders. Opening the borders to refugees does not qualify for providing safe havens to an armed group that is fighting its target government unless the group is engaging in violent cross-border attacks. The members of the Karen National Union, which has been fighting Myanmar’s government for more than five decades, frequently escape into neighboring Thailand. They occasionally organize armed attacks back into Myanmar. The following statement illustrates the type of evidence used to determine whether a state provides safe havens to a group:

Thai television reported that Burma was preparing to attack Karen refugee camps inside Thailand. Mortars reportedly were fired at one camp across the border in the Teakaplaw region, forcing thousands of refugees to flee. The fighting comes two weeks after a Karen splinter group supported by government launched cross-border raids against three (p.176) camps of refugees loyal to the Karen National Union inside Thailand. Two camps were burned to the ground and 8,000 refugees fled into the Thai jungle. (“Burmese Army Launches… .” 1997)

This statement indicates that the KNU had safe havens in Thailand. Whether the support is provided intentionally by the state is discussed more generally in the beginning of this section in explaining intentional versus de facto support. In addition, multiple sources were used to determine whether the KNU members were engaging in cross-border attacks into Myanmar.

  • SafeLead: Providing sanctuary to leadership. Providing safe havens to leaders of a group is different from providing safe havens to its members. Group leaders live in other states due to reasons such as being expelled from their target countries or no longer feeling safe in the target countries. Of the total years that leaders of rebels spent in external safe havens, 35 percent were in democratic states, in contrast to the 65 percent spent in autocratic states. Despite that fact, democratic states might be preferred by leaders due to the individual freedoms and liberties enjoyed there that make their arrest difficult. After the assassination of Indian president Rajiv Gandhi, the Liberation Tigers of Tamil Eelam (LTTE) lost its support base and funding from India. As a result, it set up offices in western European countries, such as Switzerland, France, and the United Kingdom, as well as in the United States and Canada. Evidence is clear that the Sri Lankan government put pressure on these countries to stop the fund-raising activities of the group and to return the group’s leaders back to Sri Lanka. The United States banned the group and its fund-raising activities in 1997 in passing an anti-terrorism law and declaring LTTE a terrorist organization (“Tamil Tigers, from a Rag-Tag Band….” 1997). The United Kingdom and Canada did not ban fund-raising activities of the group until 2001 (Jayamaha 2000).

  • Headquarters/Open Offices: The group has a physical office that does not oversee the violent activities of the organization or is used to spread propaganda and raise funds, not necessarily directed toward violence. Usually if a supporter country provides headquarters for a NAG or allows it to open offices within its territories, there is a high probability that the country provides a safe haven to its leadership.

  • TrainCamp: Providing training camps. Providing rebels with training camps requires extra effort on the part of the supporters than providing safe havens. Training camps are expected to be furnished with military equipment to help the members of a group in organizing and implementing violent attacks against their targets. During the Syrian occupation of Lebanon from 1976 to 2005, various Palestinian groups were trained in Lebanese territories with the assistance of Syria (“Qom Meeting of Fundamentalist Groups….” 1996). For instance, Palestinian Islamic Jihad members trained in the camps in Lebanon. Although the headquarters of the group had been in Damascus since its foundation, the training camps were not in Syrian territories. In coding the support of PIJ by Syria, providing training camps is not coded among the support types but providing safe havens to leaders is coded among the types of support. (p.177)

  • Training: In addition to training camps, some states provide training not necessarily within their own borders. This refers to the temporary assignment of a state’s security forces to train the militants.

  • WeaponLog: Providing weapons and logistics aid. This variable is coded if there is clear evidence that the arms originated from the supporting country. The evidence for whether a state provides arms to rebels is not easily attainable. Mere allegations by the target states are not enough to prove that a state provides arms to a rebel group. In the following narration directly cited from the source, it is clear that the giving of arms by the Libyan government to the Irish Republic Army (IRA) was not a mere allegation by the United Kingdom:

Histories of the IRA have identified Mr Murphy as an IRA weapons smuggler who helped to procure supplies by travelling to Libya using false passports. In the 1980s, Libya supplied the Provisional IRA with more than 100 tonnes of weaponry. (Sharrock 2007)

  • FinAid: Fund-raising is different from receiving money from the supporter state’s government. While in some cases, such as Iran and Hezbollah, governments provide funds to a rebel group, in many others the groups themselves manage to raise funds within the borders of another state, such as the IRA raising funds in the United States. When this is the case, the support is assumed to be de facto, i.e. rebels select certain states as supporters without necessarily any intentional effort on the part of the supporter. It is possible to argue that the United States had the capacity to control the IRA’s activities, in which case the support of the group would have been intentional. However, making this judgment requires a more extensive analysis of each case in the dataset, which is not an attainable goal within the time frame of the current project. The specific type of support the IRA obtained in the United States is called “passive support” by Byman (2005a). It is coded as de facto support in the NAGs Dataset.

  • Transport: Providing transport of the military equipment and military advice. If a state serves as a transport point for a rebel group, it is coded separately from providing arms and military supplies. Cambodia has for years become a de facto transport point for arms smuggling for many NAGs in Asia. (Bonner 1998). Zaire (now the Democratic Republic of Congo) was the major transport point for the weapons sent by the United States to the National Union for the Total Independence of Angola (UNITA), which was fighting the communist regime in Angola (Lewis 1987).

  • Troop: In some cases, states allow their troops to fight on the side of the rebels against their targets. When civil wars or ethnic conflicts cross the borders of other states, there is a risk that the latter will act to protect its borders. This variable is not coded for de facto support since it is impossible for a state to send its troops to help a NAG and do this without the sponsorship of the state in question. This leads to the accusation of providing troop support to rebels. Myanmar accused the Thai army multiple times of providing the KNU with troops during the cross-border operations of Myanmar’s government into the Karen National Union camps in Thailand. The following illustrates the type of statements and narratives used to code troop support:


Angola, allied to Sassou Nguesso's Cobra militia, staged a weekend attack along the border between its oil-producing Cabinda enclave and southwestern Congo, sending some 1,000 troops into Congo, according to diplomats. (“Angolan Tanks and Troops Enter….” 1997)

Following the 1979 establishment of the Islamic Republic and as a response to Israel’s invasion of Lebanon in 1982, Iran organized, equipped, and trained Hezbollah. Tehran deployed 1,500 personnel from its IslamicRevolutionaryGuardCorp (IRGC)—a semi-autonomous vanguard of Iran’s military used to foment regional disorder and support terrorist organizations—to Lebanon. (Wilner 2012, pp. 19–20)

Only a total of 6 percent of the binary support years involves states that provide troops to rebels. This is normal if we consider that troop support is a very risky strategy, since it means directly engaging with the target of a rebel group. The primary purpose of supporting a group is to avoid direct confrontation with the adversary, besides trying to undermine the power of an adversary.

  • Other: Any other kind of support not listed above.

  • DomSup: Whether there is a support basis from within the target or the supporter country.

  • DomSup_P: The confidence by which we can claim domestic support from a NAG’s target or supporter. The domestic support refers to support from among the people rather than the political leadership. 1: not very confident, 2: somewhat confident, 3: confident.

  • SupTerm: Why did the support end?

    1. 1. Pressures from the target of a given NAG

    2. 2. Pressures from the international community in general: other states (other than the target)

    3. 3. Regime change in the target country

    4. 4. Regime change in the supporting country

    5. 5. Leadership change in supporting country

    6. 6. Group ceased activity

    7. 7. Group signed a cease-fire

    8. 8. Group turned into a political movement

    9. 9. Other: describe

Support termination is coded as missing when there is no external state support for a NAG in a given year.


___. 1997. “Angolan Tanks and Troops Enter Congo’s Oil Town Pointe Noir.” Deutsche Presse-Agentur. (October 15). Accessed on August 17, 2007.

___. 1997. “Burmese Army Launches Offensive against Karen Rebels.” International News section. (February 13). Accessed on August 21, 2008.

___. 1997. “Tamil Tigers, from a Rag-Tag Band to a Fighting Force.” Associated Press Worldstream. (October 15). Accessed on August 20, 2008.

(p.179) ___ “UCDP/PRIO Armed Conflict Dataset Codebook, ver.4. 2014a.” Available at http://new.prio.no/CSCW-Datasets/Data-on-Armed-Conflict/. Accessed on June 10, 2015.

Bonner, Raymond. 1998. “Tamil Guerrillas in Sri Lanka: Deadly and Armed to the Teeth.” New York Times. (March 7). Available at www.nytimes.com. Accessed on August 15, 2008.

Byman, Daniel. 2005a. “Passive Sponsors of Terrorism.” Survival 47(4): 117–144.

Byman, Daniel L. 2005b. Deadly Connections: States That Sponsor Terrorism. Cambridge, UK: Cambridge University Press.

Cederman, Lars-Erik, Brian Min, and Andreas Wimmer. 2010. “Ethnic Power Relations Dataset.” http://hdl.handle.net/1902.1/11796. Harvard Dataverse, Version1. Accessed on June 20, 2015.

Cody, Edward. 1984. “Contras Say Supplies Sufficient Until June”. The Washington Post. (April 24).

Cunningham, David E., Kristian Skrede Gleditsch, and Idean Salehyan. 2009. “It Takes Two: A Dyadic Analysis of Civil War Duration and Outcome.” Journal of Conflict Resolution 53(4): 570–597.

Douglas, Farah. 2001. “‘Commander Poison’ Sheds Name, and Hits Taste for War.” The Washington Post. (June 4).

Girardin, Luc, Philipp Hunziker, Lars-Erik Cederman, Nils-Christian Bormann, and Manuel Vogt. 2015. “GROWup: Geographical Research On War, Unified Platform.” ETH Zurich. Available at http://growup.ethz.ch/.

Gleditsch, Nils Peter, Peter Wallensteen, Mikael Eriksson, Margareta Sollenberg, and Havard Strand. 2002. “Armed Conflict, 1946–2001: A New Dataset.” Journal of Peace Research 39(5): 615–637.

Högbladh, Stina, Therése Pettersson, and Lotta Themnér. 2011. “External Support in Armed Conflict, 1975–2009: Presenting New Data.” Paper presented at the 52nd Annual International Studies Association Convention, Montreal, Canada, March 16–19, 2011.

Jayamaha, Dilshika. 2000. “AP Photos COL101-102.” Associated Press Worldstream. (December 11). Accessed on August 20, 2008.

Kittner, Cristina C. Braffman. 2007. “The Role of Terrorist Safe Havens in Islamist Terrorism.” Terrorism and Political Violence 19(3): 307–329.

Kreutz, Joakim. 2010. “How and When Armed Conflicts End: Introducing the UCDP Conflict Termination Dataset.” Journal of Peace Research 47(2): 243–250.

Lewis, Neil A. 1987. “Washington Talk: Working Profile; Chester Croker: Inside, Making Policy on Africa.” (June 9). Accessed on August 28, 2007.

Pettersson, Therése, and Peter Wallensteen. 2015. “Armed Conflicts, 1946–2014.” Journal of Peace Research 52(4): 536–550.

Regan, Patrick M., and Aysegul Aydin. 2006. “Diplomacy and Other Forms of Intervention in Civil War.” Journal of Conflict Resolution 50 (5): 736–756.

Regan, P., R. Frank, and A. Aydin. 2009. “Diplomatic Interventions and Civil War: A New Dataset” Journal of Peace Research 46(1): 135–146.

Rupert, James. 2000. “Africa Journal.” The Wahington Post. Available https://www.washingtonpost.com/archive/politics/2000/07/10/africa-journal/fbed9a6f-f0ee-4e81-87ed-abf81b82eb7b/. Accessed on April 20, 2015.

(p.180) Sharrock, David. 2007. “IRA Baron Charged: On Tax.” (November 10). Weekend Australian. Accessed on July 24, 2008.

Themne´r, Lotta, and Peter Wallensteen. 2014. “Armed Conflict, 1946-2013.” Journal of Peace Research 51:541–54.

Wilner, Alex S. 2012. “Apocalypse Soon? Deterring Nuclear Iran and Its Terrorist Proxies.” Comparative Strategy 31: 18–40.

Woodward, Bob & Shapiro, Margaret & Feinman, Barbara. 1984. “CIA Sought 3rd-Country Contra Aid”. The Washington Post. (May 19).

Zeev Maoz, and Errol A. Henderson. 2013. “The World Religion Dataset, 1945–2010: Logic, Estimates, and Trends.” International Interactions 39: 265–291.