Issues and Actors in African Nonstate Conflicts: A New Data Set

ABSTRACT Armed nonstate conflict without the direct involvement of the state government is a common phenomenon. Violence between armed gangs, rebel groups, or communal militias is an important source of instability and has gained increasing scholarly attention. In this article, we introduce a data collection on conflict issues and key actor characteristics in armed nonstate conflicts that provides new opportunities for investigating the causes, dynamics, and consequences of this form of organized violence. The data builds on and extends the Uppsala Conflict Data Program Non-State Conflict data set by introducing additional information on what the actors in the conflict are fighting over, alongside actor characteristics. It covers Africa during the time period 1989–2011. The data set distinguishes between two main categories of issues, territory and authority, in addition to a residual category of other issues. Furthermore, we specify sub-issues within these categories, such as agricultural land/water as sub-issue for territory and religious issues for other issues. As actor characteristics, the data set notes whether warring parties received military support by external actors and whether religion and the mode of livelihood were salient in the mobilization of the armed group. The article presents coding processes, key features of the data set, and point to avenues for new research based on these data.

Armed nonstate conflict without the direct involvement of the state government is a common phenomenon. Inter-rebel clashes are important elements of complex civil wars such as in Syria or Sudan. Other countries, such as Kenya and Mexico, are largely spared by civil conflicts against the state, yet see devastating violent nonstate conflict between gangs or communal militias. The Uppsala Conflict Data Program (UCDP) registered a total of 454 such nonstate conflicts around the world in the time period 1989-2011. These violent conflicts together caused more than 100,000 fatalities. Around 70% of nonstate conflicts and the fatalities incurred were registered in Africa (Sundberg et al. 2012).
One factor commonly acknowledged to shape the causes, dynamics, and consequences of armed conflict is the issue over which the involved parties fight (Bell and Long 2016;Diehl 1992;Hensel, McLaughlin Mitchell, Sowers, and Thyne 2008). For example, conflicts over territorial issues are found to present specific challenges to resolution (Huth 1998;Toft 2014). Similarly, actor characteristics, such as assistance received from foreign patrons, are known to greatly affect conflict dynamics and risk of recurrence (Kalyvas and Balcells 2010;Karlén 2017). These findings emerge from studies and data on intrastate and interstate conflict, pitting the state against a nonstate actor or another government. However, for nonstate conflicts, fought between two organized groups neither of which is a government, systematic information about the issues at stake has been lacking so far and there are little systematic data on actor characteristics available. This has left researchers hamstrung in conducting similarly nuanced analyses of nonstate conflicts as of civil and interstate wars.
Addressing this gap, this paper introduces a novel data collection on issues and actor characteristics in nonstate conflicts. We specify two main categories of issues, territory and authority; and an additional residual category of other issues. Each of these issue clusters is further disaggregated into subcategories. These subcategories cover for example agricultural land/ water as subcategory for territory and religion as subcategory for other issues. As actor characteristics, the data cover the occurrence and form of military support by foreign or domestic governmental or nongovernmental actors. It also notes whether religion and the mode of livelihood were salient in the mobilization of the armed group. Information on the occurrence of conflicts and involved actors is taken from the UCDP Non-State Conflict Dataset v. 2.5-2016. Our data cover all African nonstate conflicts, 1989-2011. With the focus on Africa, the vast majority of all nonstate conflicts and deaths during this period are accounted for.
The data set opens up a number of new avenues for research. First, it allows researchers to focus on theoretically delimited subsets of nonstate conflicts. For example, land conflicts (for example, Eck 2014) and violent leadership struggles within armed opposition movements in civil war (for example, Lilja and Hultman 2011) are subsets of nonstate conflicts that have received specific attention in previous research. Due to the lack of more disaggregated data, researchers have often been forced to test arguments related to specific subsets of conflicts on broad categories of data. Moreover, being integrated with the UCDP Georeferenced Event Dataset (GED) data (Sundberg and Melander 2013) the data set can be used to study fine-grained spatial and temporal patterns in violence. Studying subnational patterns in violence has become an important subject of scientific inquiry (for example, Raleigh and Choi 2017). The data set is also a suitable starting point for case selection for small-and medium-N studies (for example, Ide 2015).

Rationale for taking into account conflict issues and actor characteristics
The issue fought over is commonly acknowledged to be an important feature of an armed conflict. Research comparing different issue categories in conflicts involving the state demonstrates differences in causes, duration, and dynamics of armed conflicts (Buhaug 2006;Hensel et al. 2008;Sambanis 2001). The issue over which a conflict is fought provides distinct obstacles to resolution and thereby also affects conflict duration (Toft 2014). There are also distinct patterns with regard to drivers of conflict. For example, the relationship between regime type and conflict risk differs for civil conflicts over government power compared to civil conflicts concerning a particular territory within the state (Buhaug 2006). Similarly, actor characteristics affect conflict dynamics. For instance, civil wars that see external supporters providing logistical assistance or weapons are more likely to recur (Karlén 2017).
While a growing number of data sets provide more nuanced information on intrastate conflict, for nonstate conflicts systematic information about the issues at stake has been lacking so far and there are only very limited data on actor characteristics. 1 Nonstate conflict is defined here as the use of armed force between two organized groups neither of which is the state, which results in at least 25 battle-related deaths in one calendar year (Sundberg et al. 2012). The actors in these conflicts are organized groups, such as rebel groups, militias, or gangs, as well as informally organized actors, such as supporters of political parties and communal identity groups. Addressing the gaps in existing data collection efforts, this data set introduces information on the issue over which the parties fight, as well as key actor characteristics related to the mobilization and behavior of the warring actors, namely the form and extent of military support, and whether religion and the mode of livelihood are salient in mobilization. 2 Our data on issues and actor characteristics would allow researchers to probe a range of different research questions which have hitherto received scant attention due to lack of good data. First and most directly, the study of subgroups of conflict delineated by specific issues or actors is greatly advanced by 1 To our knowledge, the only existing data on actors in nonstate conflict are Wig and Kromrey (2018), who identify the ethnic identity of communal groups involved in conflict by linking them to the Ethnic Power Relations (EPR) data (Cederman et al. 2010). For issues, the SCAD data set, which covers various forms of protest, strikes, riots, and other events of social disturbances, notes "the source of tension" for events included in their database (Salehyan et al. 2012;Salehyan and Stewart 2016). 2 The mode of livelihood, and in particular whether groups pursue settled or pastoralist livelihood strategies, is regularly a salient source of identity for communal groups involved in violent conflict (Sundberg et al. 2012).
Other salient delineations between groups follow the lines of religious or ethnic identities. As data on the ethnic identity of communal groups in Africa are available from other sources, we do not duplicate existing efforts here.
the new data. For example, conflict over land has received considerable attention in the environmental security literature (Kahl 2006;Theisen 2012). Land-related violence is also key in studies of electoral violence (Boone 2011;Klaus and Mitchell 2015) and the sons of the soil literature (Côté and Mitchell 2015). Similarly, fighting among different factions of opposition groups in civil war has received great scholarly interest (Fjelde and Nilsson 2012;Lilja and Hultman 2011;Nygård and Weintraub 2015;Warren and Troy 2014). For example, the new data allow for systematically investigating whether inter-rebel conflicts occur over leadership in the opposition movement against the government as commonly claimed (Fjelde and Nilsson 2012;Lilja and Hultman 2011). Conflicts along the lines of dominant livelihoods are another relevant field of scientific inquiry. The relationship between nomadic or seminomadic pastoralist communities and settled farmers has been emphasized as dispute-prone (Benjaminsen and Ba 2009;Turner 2004;Turner et al. 2011). There is also a distinct literature on inter-pastoralist violence that often, but not always, evolves around livestock raiding (Butler and Gates 2012;Detges 2014;Meier, Bond, and Bond 2007). Our new data allow for narrowing down a sample to the scope conditions of particular theoretical arguments and, thus, a better match between theoretical expectations and empirical inquiry. Second, there is a wide range of research questions related to issues and actor characteristics as causes of particular dynamics in conflicts. To give some examples, the conflict issue is important for understanding the targets of an armed nonstate actor (Salehyan and Stewart 2016), the incentives of a government to intervene in nonstate conflict (Elfversson 2015) and, thereby, key to understanding the scope of observed fighting (Arena and Hardt 2014). External actors, in particular governments, often bear some responsibility for the violent mobilization of a nonstate actor, for example exploiting local cleavages for weakening opposition groups (Brosché and Elfversson 2012;Kahl 1998). Our data on external support allow for more in-depth study of these relationships. One case in point is the Sudanese government providing financial support to the Southern Sudan Defense Force, which was fighting against the rebel group Sudan's People Liberation Movement/Army, at the same time as the latter was involved in a long-standing civil war against the government in the 1990s.

Scope and data collection
The new data set on conflict issues and actor characteristics includes all countries in Africa and covers the time period 1989-2011. It builds on the UCDP Non-State Conflict Dataset v. 2.5-2016, which covers conflicts between formally organized groups, such as rebel groups, and informally organized groups, such as supporters of political candidates and communal groups (Sundberg, Eck, and Kreutz 2012). 3 By focusing on Africa, the data cover around 70% of the nonstate conflicts registered for the entire world during this time period. Nonstate conflict typically takes the form of short outbursts of fighting and the 313 different nonstate conflicts covered are made up of 468 unique conflict-years (see Figure 2).
The data collection was carried out in four steps. First, the UCDP Encyclopedia (www.ucdp.uu.se) and the sources referenced there were consulted. Second, we studied the sources reported in the UCDP GED (Sundberg and Melander 2013), which includes all violent events connected to the respective actors and conflicts. In the third step, we searched for statements on disputed issues from leaders and/or other representatives of the warring groups in media reporting. This was done in the Dow Jones Factiva aggregator. Finally, books and reports on the actors and/or conflicts in question were used to find additional information after consulting area specialists. If a nonstate conflict restarts after at least 1 year of inactivity (that is not crossing the 25 fatalities threshold required for inclusion in the UCDP data), the conflict issues at stake and actor characteristics were investigated again.
In the first two steps, we drew on the work carried out by the UCDP. In terms of sources, UCDP uses at least one of the global newswires AFP, Reuters, Xinhua, or Agencia EFE, in addition to BBC monitoring, which in turn contains information from local newspapers. Other sources systematically covered by the UCDP coders are country-specific sources such as NGO reports, for instance from Human Rights Watch or the International Crisis Group, as well as research articles and books. The material for this data set was compiled and coded by full-time long-term employees of UCDP and reviewed by the project managers to make sure that it is consistent with the definitions and coding procedures.
Ideally, the conflict issues are classified based on what the parties themselves state is the incompatibility. However, as information on stated incompatibilities is not always available from public sources, we also noted the issues reported in secondary sources. 4 In order to keep up with high reliability standards of other UCDP data and to increase transparency, we noted the type of source in the data set (primary or secondary, referring to a specific outbreak of violence or a background issue). Detailed information on the data collection procedure, the sources used, and example coding can be found in the online appendix.

Issue categories
An issue is here conceived as a disputed point or question, the subject of a conflict or controversy (Diehl 1992). Importantly, the issue is not to be conflated with the driver or the cause of a tension or dispute. In other words, it is what parties ''choose to fight over, not the conditions that led to the choice of military force as the means'' (Diehl 1992, 333).
Our approach to identifying relevant issue clusters strives to reconcile two conflicting ideals: parsimony on the one hand, and reflecting the multidimensionality and nuances identified in previous research on nonstate conflict on the other. A parsimonious approach, summarizing issues in as few categories as possible, has dominated key data collections on interstate or intrastate conflicts (Diehl 1992;Gleditsch, Wallensteen, Eriksson, Sollenberg, and Strand 2002;Hensel et al. 2008). For example, for intrastate conflicts coded by the UCDP, the stated incompatibility classifies the conflict as either being over territory or government, or both. As armed conflict per se is a rare event, further disaggregation in smaller subsamples may not always be desirable.
However, previous research on nonstate conflict points to a range of issues that a mutually exclusive two or tripartite ensemble of issues categories hardly would do justice to. For example, nonstate conflict in the context of general elections often relates to both infected issues of land ownership and at the same time the question of what party candidate prevails in the electoral competition (Klaus and Mitchell 2015). In these cases, assigning one issue would be arbitrary, and it would also be useful to know whether the conflict concerned traditional leadership structures, or a formal role in government.
We therefore adopt an approach that leaves a great degree of flexibility to the user. First, the issue categories are not coded as mutually exclusive. This means that the coders code all issues mentioned in relevant sources and do not assign what the conflict primarily was about, which could be arbitrary. Second, we inductively identify sub-issue categories based on the empirical material, but also informed by extant research on nonstate conflict. This list of subissues displayed in Figure 1 could be relevant for specific research questions. For example, it enables the study of conflict over water and agricultural land and livestock specifically, which are particularly important for investigating claims on environmental drivers of conflict (Ide 2015;Kahl 2006).
As main categories, similar to the UCDP categories for intrastate conflicts, we distinguish between authority and territory conflicts. Authority is defined here literally as the power or right to give orders, make decisions, and enforce obedience (Oxford Dictionary 2013). 5 In the data set, authority issues are 5 This category is related to conflict over government known from UCDP data sets on intrastate conflict. However, as the nonstate conflicts often revolve around local power struggles, and including over traditional institutions, rather than government structures, the broader term authority is chosen.
competing claims over which of the warring parties exerts control over the other group through the state apparatus or informal power structures. Territorial issues concern control or use of the land, but not authority over the other warring party. Thus, in territorial conflicts the groups do not want to exercise power over the other group through the government or informal authority structures, but strive for control of the land, for example for own settlement or local resource use. Under a third cluster, we subsume other issues, with livestock and religious issues being the most frequent and empirically the most relevant issues. In the following, we describe the categories of issues, give examples, and provide some descriptive statistics.

Conflicts over authority
Authority is the conflict issue in 36% of the 468 conflict years coded in the data set. In the most obvious form, we see nonstate conflicts over authority Although formal authority is by far the most common form of conflict issue concerning authority, the authority category is more inclusive than this. In 8% of all conflict years fought over authority, the issue at stake was not control of the state apparatus, but leadership within a group or a community. These informal or traditional authorities play important roles in many regions of Africa. For example, in Nigeria in 1999, fighting erupted between the two Muslim groups Hausa and Ninzam over who should be appointed Emir, the traditional religious leader in the area. Also, disputes over leadership within rebel movements belong to this category. For example, in 2002, a dispute over the position as chairman of the Somali rebel group Rahanwein Resistance Army turned violent. Taking into account these differences, informal authority is coded as the second sub-cluster of the authority category.
Nigeria has experienced the highest number of conflicts over authority of all countries included in our data. Oftentimes, violence has occurred in connection with elections pitting supporters of political candidates and rival ethnic groups against each other. Nigeria is home to more than 250 different ethnic groups and ethnicity has been used as a base for political inclusion and exclusion. Following the end of military rule in 1999, the constitution emphasized the concept of "indigeneity" granting exclusive rights to those belonging to an ethnic community that first settled in an area. As indigeneity and ancestral territories became a legitimate basis on which to distribute benefits such as local political appointments and government services, numerous conflicts over authority emerged (International Crisis Group 2006). Figure 2 shows that territory is the most frequent conflict issue in the data collection. A full 66% of conflict years coded had territory as a conflict issue. In this category, we observe a great number of land-use conflicts where water or agricultural lands are the bones of contentions. In 26% of all years coded access to agricultural land or water resources were an issue at stake. One example is when in 2011 the Kenyan pastoralist groups Borana and Turkana fought each other after the Turkana accused the Borana of uprooting them from their grazing land.

Conflicts over territory
Yet, even more frequent than these forms of land-use conflicts are other territorial issues for example concerning the borders of administrative districts, such as clashes over homelands in Darfur, Sudan, or the border of districts that are dominated by different ethnic groups in Ethiopia. We registered other territorial issues in 45% of the conflict years included in the data set. We specify two subsets of territory conflicts, agricultural land/ water and other territorial issues.

Other issues
The above categories are largely in line with categories of stated incompatibilities assigned for UCDP intrastate or interstate conflicts (Gleditsch et al. 2002). However, the statements of warring parties and the scientific literature on nonstate conflict point to additional salient issues beyond these categories. Among these, livestock and religion are particularly salient and are assigned specific sub-issues.
A whole 25% of the conflict years saw fighting over livestock. Livestock raiding involving deadly and massive violence is a frequent phenomenon in the Sahel region with roots in pastoralist culture (Hagmann and Mulugeta 2008;Meier et al. 2007). In the data, Kenya is the country with the highest number of conflicts over livestock. In fact, in more than 80% of the nonstate conflict years recorded for Kenya livestock raiding is a conflict issue.
In contrast to widespread perceptions about the increasing importance of religion for armed conflict, religious issues in nonstate conflicts are rare and found in only 5% of the conflict years (see Figure 3). Where religion is the conflict issue, it is most often combined with other issues, like territory. One example is the conflict between the Somali groups Ahlu Sunna Waljamaca and Al-Shabaab, active between 2008 and 2010, which included issues of territorial control, but also religious aspects as the Sufi group Ahlu Sunna Waljamaca accused the Salafi group Al-Shabaab of destroying religious symbols and trying to ban Sufi tradition.

Actor characteristics
As actor characteristics, we note external support received by the warring parties, as well as important identity features of the groups, namely religious identity and mode of livelihood for communal groups.

Religious identity
Religious identity is coded for all actors in the data set. We note when religion is salient in the mobilization of the armed groups involved in the nonstate conflict. In 21% of the conflict years included in the data set religious identity was coded for at least one of the parties. In total, the data contain 401 different nonstate actors, of which 16% were coded as having a religious identity. A prominent example of a religious actor is the Alliance for the Re-Liberation of Somalia/Union of Islamic Courts in Somalia, which was made up of Sharia courts and had a clear Islamist agenda.

Livelihood of the actor
Competing livelihood requirements can lead to disputes and at times to violent conflicts. Rather than the source of income for the group, the livelihood should here be seen as shaping the identity of actors, the communal actor's needs and opportunities to compromise with another group. Where the actors of the nonstate conflict belong to informally organized groups ("organization level 3") in the UCDP nonstate data set, we add a variable noting the main livelihood of the group (pastoralist, farmer, or agro-pastoralist, with the latter being groups that combine agriculture with livestock keeping (Owuor, Mauta, and Eriksen 2011). 6 Most conflicts in the data set are noted for groups with a pastoralist livelihood. Thirty percent of the conflict years were coded as pastoralist groups fighting each other. Somewhat unexpectedly, fighting over livestock was less common than territory issues among these herders. In 83% of these conflict years, territory was reported as being a disputed issue, while only 10% was reported as being over authority.
The share of farmer-pastoralist violence is almost as large as for pastoralist conflict. In our data, 18% of the conflict years with livelihoods coded for both sides are fought between farmers and pastoralists. Another 11% is fought between farmers and agro-pastoralists. In contrast, farmer-farmer nonstate conflict is rare. The most common conflict issue between farmers and pastoralists was territory. Most of the farmer-pastoralist conflicts over territory took place in Ethiopia and Kenya.

External military support
External actors, in particular governmental actors, often have a great impact on the violent mobilization of nonstate actor and the conflict dynamics, for example exploiting local cleavages for instigating violence that will weaken line opposition groups (Brosché and Elfversson 2012;Kahl 1998). We provide information on external military support to the warring actors in a nonstate conflict that captures these dynamics. These data build on the definitions of the UCDP External Support Dataset (Högbladh et al. 2011). We code measurable support in the forms of weapons, military aid, training, intelligence, logistics, and provision of safe havens which is delivered by an actor external to the conflict. Actors providing external support to a warring party can be governments, either the national government or foreign governments, as well as groups residing in or outside the country in which the nonstate conflict is taking place. The party providing the support should be a 6 For other types of actors, for example organized groups or supporters of political parties, livelihood is most often not named as important for mobilization. Therefore, we refrained from coding this variable for other than communal groups. state or an organization (in the widest sense of the term) and not an individual. The data are restricted to non-warring support, meaning that the externally supporting actor is not involved with own troops, but provides other forms of support to the other side. It is provided at the level of the primary warring party so that it can be identified which nonstate group in the data got support in its struggle against another nonstate group. In 28% of conflict years included in the data, we observe some form of external support to either side of the conflict.

Illustrating prospects for research
The data could inform important research fields, such as questions relating to multiparty civil wars, electoral violence, environmental security, and sons of the soil conflicts. In order to illustrate the use of the data, we look into how issues and actors are linked. Specifically, we investigate whether conflicts over particular issues are more likely to see involvement by external actors. We hypothesize that belligerents are more likely to attract external support when the issue at stake relates to key questions of authority, such as elections to local or national decision bodies. In order to investigate this question, we identify all conflict years where authority was the only, or one of several, issues. Among the remaining conflict years, we then select all conflict years where territory was one of the conflict issues (but not authority) and lastly a third cluster of all other issues. We thereby turn the data into three mutually exclusive categories. We then compare the rate of external involvement for conflicts over authority to the other two clusters of conflict.
As Figure 4 illustrates, indeed, authority issues seem to be associated with support to one or both sides in the conflict in line with expectations. Whereas among all conflict years fought over authority almost half of the 167 conflict years see some kind of involvement by external supporters, among the 81 conflict years with other issues the share of involvement is below 20% and for territorial issues it is around 34%. We thus see that the likelihood of involvement is much higher in conflicts over authority, also compared to territorial issues. 7 A more in-depth analysis of this pattern is beyond the scope of this paper. Yet, nonstate conflicts over issues related to governance and thus core state interest seem to be particularly likely to attract interests by other nonstate and state actors, which has implications for the scope of fighting, conflict dynamics, and prospects for conflict resolution.

7
This difference in external involvement remains substantially unchanged if we aggregate the conflicts fought over several issues in a different manner in order to achieve mutually exclusive categories. In absolute terms, it is conflicts over territory that see most involvement by external actors, as there are more conflicts fought over territory compared to authority and several conflicts are fought over both.

Conclusion
An expanding field of research has increased our understanding of the causes and dynamics of nonstate conflicts. Yet, unresolved questions remain in explaining when and why these conflicts occur and how they can be resolved. Another gap is understanding links between nonstate conflicts and other forms of violence and actors. In this article, we present data on issues and critical actor characteristics facilitating the advance of research on these questions.
The new data presented in this paper reveal a number of interesting patterns. Territorial issues are the most frequent conflict issue. This is also the case for pastoralist communal groups that are often portrayed as being involved in endemic fights over livestock with other groups. In contrast, religious issues are very rare. The data also suggest that the conflict issue is linked to actor characteristics. External military support to belligerents is more frequent when the conflict is fought over an authority issue. Further leverage on the determinants and dynamics of nonstate conflict could be gained by combining these data with other available information on actors, such as ethnicity and associated information on the location, size, and political status of respective ethnic groups (Vogt, Bormann, Rüegger, Cederman, Hunziker, and Girardin 2015;Wig and Kromrey 2018). To conclude, we hope that these data will allow the academic community to  . Average external involvement for conflicts over authority, compared to conflicts over territory (but not authority) and other issues. make progress and provide guidance to policymakers for the prevention, management, and resolution of nonstate conflict.