Camp settlement and communal conflict in sub-Saharan Africa

Abstract

Are areas that host encamped refugees more likely to experience communal conflict, and under what conditions? Building on insights from the refugee studies literature suggesting that settling refugees in camps can intensify intercommunal tension in host communities, this article investigates the effect of refugee encampment on the occurrence of communal conflict at the subnational level in sub-Saharan Africa. It first tests for a general relationship between the overall presence and population intensity of encamped refugees and communal conflict before assessing whether this relationship is moderated by local-level characteristics, including interethnic linkages and political and economic marginalization within the host region. The basic findings show that communal conflict occurs more frequently in regions where refugees are camp-settled. Tests for interactive effects indicate that refugee camps have a significant marginal effect on conflict only if they are located in areas with politically marginalized host groups. Origin country/host region ethnic ties are shown to exert significant moderating effects. Moreover, results from an extended set of analyses show that the form of refugee settlement matters, as the presence and population intensity of self-settled refugees are related to decreases in the occurrence of communal conflict.

Keywords

communal conflict inequality political exclusion refugees refugee settlement sub-Saharan Africa

Introduction

Researchers interested in the relationship between population movements and conflict have, in recent years, focused their attention on how refugee populations might affect the onset and dynamics of civil war, or armed conflict between formally organized rebel groups and the state (Fisk, 2014a; Lischer, 2005; Salehyan & Gleditsch, 2006; Shaver & Zhou, 2015). Yet several of the theoretical explanations for the anticipated civil war/refugees association also link logically to other, non-state forms of violence that have yet to be investigated. The central research question in this article is: are areas that host encamped refugees more likely to experience communal conflict, and under what conditions? Past studies argue that refugee populations can increase the likelihood of civil war when they ‘compete with locals over scarce resources such as employment, housing, land, and water, constituting an economic “threat”’ which ‘may lead to a setting that invites violence against migrants as well as more general dissatisfaction with political and economic conditions’ (Salehyan & Gleditsch, 2006: 344). While mobilizing against the central government is one way to respond to these grievances (and communal conflict can transform into civil war (Brosché & Elfversson, 2012)), the cost and risk thresholds for engaging in other, non-state forms of conflict are comparatively lower (Fjelde & von Uexkull, 2012; Hendrix & Salehyan, 2012). This is the case for communal conflict in particular, which involves ‘local encounters between identity-based groups including ethnic, regional, religious or livelihood communities’ and tends to ‘emerge over territorial disputes, local power disparities, resource access and historical disagreements’ (Raleigh, 2014: 93). As Brosché (2014: 14) notes, shared communal identity may not only form along the lines of common ethnic or religious identity, but also may correspond to shared identity as ‘original inhabitants of an area’ in contrast to ‘more recent settlers’ which are cast as ‘the other’.

Along these lines, forced migration and refugee studies scholars have long debated how refugees’ patterns of settlement, via their effects on local host economies and resources, influence host community relations and security. In particular, there is a longstanding and unresolved debate regarding the effects of refugee encampment.¹ Some argue that by concentrating aid and service provision to refugees (which the host population resents), refugee encampment exacerbates tensions in the host community (Duncan, 2005; Jacobsen, 1997). This could, in turn, increase the potential for conflict. Others, in contrast, maintain that refugee camps directly benefit local hosts by boosting employment opportunities and the local economy more generally, which could reduce the chances of conflict. While camp settlement may have security implications for the host community – refugees as well as local host populations – its effects have not yet been examined beyond a small subset of case studies.

This article investigates the relationship between settling refugees in camps and communal conflict – conflict between non-state groups formed along communal identity lines – in 39 countries in sub-Saharan Africa. It first examines whether there is a general relationship between refugee encampment and the occurrence of communal conflict in an administrative unit (e.g. a region, state or province). It then moves forward to explore whether the effect of refugee encampment on communal conflict is moderated by other potentially important considerations identified in the refugee studies literature, namely, the relative political and economic status of local hosts and the presence of ethnic ties between the refugee-sending country and refugee hosts. The findings show that the presence and population intensity of camp settlements increase the counts of communal conflict that occur in a host region. I also find evidence that communal conflict increases when larger intensities of camp-settled refugees are located in a region where a host group is politically marginalized, and where there are ethnic ties between the refugees’ origin country and the host population, than otherwise. Additionally, a secondary set of analyses find that the incidence of communal conflict decreases where refugees are self-settled² among local populations. Taken together, these results indicate that while host governments attempt to justify refugee encampment policies on security grounds, congregating refugees in camps can reduce the physical security of refugees and hosts alike.

Given that the vast majority of refugees today are in protracted situations and repatriation levels are at historic lows,³ it is increasingly important to understand the security impacts of encampment policies. Whether camp settlement is correlated with communal conflicts centered on ‘traditionally based contests for local level resources and micro political dominance’ (Raleigh, 2014: 93) is an empirical question that has important implications for refugee policy. The following section of the article briefly surveys the existing political science literature on refugees and host-country conflict and introduces arguments and hypotheses pertaining to communal conflict in particular. The third and fourth sections discuss methods and present the findings, respectively. The concluding section discusses implications for domestic and international refugee policies.

Refugees and conflict dynamics in the host state

Previous studies that investigate the security implications of hosting refugees theorize how refugee influxes can directly and/or indirectly destabilize host countries. For instance, refugee populations are thought to be directly destabilizing when they include ‘refugee warriors’ (Zolberg, Suhrke & Aguayo, 1989) who create and operate bases in the host country, using camps in particular for sanctuary, humanitarian aid, and recruitment purposes (e.g. Stedman & Tanner, 2003; Lischer, 2005; Terry, 2003). Scholars also theorize that refugee populations can be directly destabilizing when they extend rebel networks from the origin country into the host state, providing the know-how and arms necessary for domestic opposition groups to successfully mobilize and challenge the government (e.g. Salehyan & Gleditsch, 2006). Prior work has found that hosting larger refugee populations from geographically contiguous countries is not only related to the duration of civil war in the sending state (Salehyan, 2009), but also heightens the likelihood of civil war onset in the host (Salehyan & Gleditsch, 2006). More recently, research has begun to examine how variation in refugees’ location and form of settlement can condition the relationship between refugee populations and the subnational occurrence of civil conflict (e.g. Fisk, 2014a; Shaver & Zhou, 2015). The studies in this special issue further examine important case variation related to refugees and conflict – including the presence of ethnic ties (Rüegger, 2019) and the capacity of the host government to attenuate the effects of hosting refugees (Böhmelt, Bove & Gleditsch, 2019). The focus of the present analysis is the relationship between refugee settlement type – specifically refugee encampment – and the occurrence of intercommunal violence in host regions. In the following section, I integrate insights from the distinct literatures on communal conflict and forced migration and formulate hypotheses concerning the impacts of refugee encampment on communal conflict.

Refugee encampment, settlement location, and communal conflict

As mentioned in the introduction, a primary debate in the field of refugee studies concerns how refugee encampment impacts host communities (Black, 1998; e.g. Crisp & Jacobsen, 1998; Harrell-Bond, 2000; Smith, 2004). This literature yields several insights with regard to how hosting refugee camps and collective centers could contribute to communal tension and conflict – including intergroup conflict among the refugees, between refugees and the host population, and among local groups – in host communities. One such consideration is that refugee encampments may spatially concentrate individuals who fled (and may be on different sides of) the same conflict, as well as refugees of different nationalities. Intergroup conflict among refugees is considered a greater risk ‘when members from political, ethnic, and/or religious groups that have a history of enmity and conflict are hosted in the same camps’ (Nathan, 2000 quoted in Mogire, 2011: 64). This is largely due to political differences, given that ‘refugees are highly politicized and often support competing and hostile political and rebel factions’ (Mogire, 2011: 65). Conflict between refugees from different countries, though less common, also can result (Mogire, 2011: 68). In Kakuma camp in Kenya, for example, there have been bouts of violence involving Dinka and Nuer refugees from South Sudan, as well as violence between South Sudanese refugees and refugees from the Great Lakes region.⁴ Militarization and other forms of manipulation by armed groups seeking potential recruits and other ‘refugee resources’ may be more likely where there are greater concentrations of refugees, which can increase the risk of communal conflict if the host population blames the refugees for the armed groups’ infiltration (Mogire, 2011).

In addition, with very few exceptions in the African context,⁵ encamped refugees are segregated spatially and face restrictions on their ability to move and find employment beyond organized settlements (Hunter, 2009; Kibreab, 2003). In Tanzania, for example, refugees are not permitted to move more than 4 km outside the radius of their camp. Chiasson (2015) notes that, ‘[i]n order to leave their DA (designated area) […] a refugee must obtain a temporary movement permit from an official’. A refugee who does not comply with the permit limit (generally 14 days) may be convicted and imprisoned for up to six months, fined 50,000 shillings, or both.

These types of restrictions effectively induce refugees’ economic dependency upon the aid and assistance provided by international and nongovernmental organizations and/or the host government (Hyndman, 1997; Kibreab, 1989). Interrefugee fighting can result from food shortages and other acute forms of resource scarcity (due to heightened competition over limited resources and associated grievances) within the camps. At the same time, a consequence of the dedicated provision of food supplies, clean water, education, medicine, and other forms of assistance to encamped refugees is that it can create resentment among local hosts based on the perception that the refugees’ livelihoods are being privileged over their own. In other words, refugees living in organized settlements may be ‘seen as a privileged group in terms of services and welfare provisions’ (Agblorti, 2011; also Aukot, 2003; Betts, 2009; Loescher & Milner, 2005: 32). Along these lines, Duncan’s study of host communities in Indonesia found that ‘[m]any locals were angered by the sight of government trucks filled with foodstuffs going into the camps’ (2005: 35). The hosts believed that the displaced populations among them were ‘lazy and simply sat around waiting for handouts […] With all of these supposedly unemployed and idle people, the camps came to be seen as anarchic places’ (Duncan, 2005: 34–35; also see Lawrie & Van Damme, 2003). Given that the overall amount and frequency of aid provision should be greater in areas where camp-settled refugees reside, the view of refugees as privileged and the potential for resentment and conflict are likely to be more prevalent in these locations. In the rare instances (e.g. Uganda) in which settlement-based refugees are allocated plots of land to farm, the size and sustainability of the plots and the degree to which they enable refugees to supplement food aid provisions decrease as the number of refugees increases. The spatial segregation of and restrictions on camp-based refugees can also prevent local hosts from taking advantage of and benefiting from the refugee presence in ways that could simultaneously decrease the refugees’ economic dependency – for instance by hiring laborers and through the exchange of goods and services in camps and surrounding markets (Whitaker, 2002: 352).

Third, as Kibreab notes, ‘[t]he aim of warehousing refugees in camps is to prevent them from developing relations with the host populations so that they maintain their separate national identities and way of life until return […] becomes possible’ (2012: 101). Encamped refugees are relatively less able to integrate/familiarize themselves with the host community when they are contained in spatially segregated and controlled sites. Hosting larger numbers of non-integrated, camp settlement-based refugees may therefore exacerbate ‘us’ and ‘them’ identification, tensions that form along communal lines, and subsequently the likelihood of conflict. Hovil (2014), for instance, argues that in general, ‘[c]amps play into the narrative that refugees are outsiders, foreigners or a security threat, demanding close scrutiny until such time as they can return home’.

Another characteristic of refugee encampment that can exacerbate host community tension and lead to conflict is the host government’s appropriation of land for camp/settlement construction, particularly if the hosts rely on access to the land for their own livelihoods.⁶ For example, Sebba (2006: 1) finds that ‘land conflicts between refugees and nationals are a result of government policy of settling refugees in gazetted areas in Uganda’. Both wealthy and poorer African hosts can be negatively impacted by land appropriation: ‘the wealthier hosts may lose more from loss of grazing, and the poor may lose more from loss of firewood and thatching grass’ (Chambers, 1986: 254).⁷ Thus, the land appropriation and pressure associated with encampment may have negative livelihood impacts on hosts, heightening communal tension and increasing the likelihood of subsequent communal conflict between hosts and refugees, as well as among local hosts.

Fifth, previous research demonstrates that increased resource scarcity and competition significantly increase the likelihood of communal conflict (Fjelde & von Uexkull, 2012; Hendrix & Salehyan, 2012; Kahl, 2006). The scale of environmental degradation and resource scarcity associated with hosting refugee populations can be linked to encampment. Both tend to be more severe where populations are more densely concentrated (in camps and centers) rather than dispersed throughout a host region (Black, 1994). An analysis of environmental degradation surrounding the Dadaab refugee complex in Kenya, for instance, blames mobility restrictions in a low-productivity location – a by-product of the refugees’ encampment (Nordic Agency for Development and Ecology, 2010). Camp settlement also can exacerbate environmental problems because it necessitates clearing land and its resources for camp construction and concentrates and restricts the movement of people, who then more heavily depend on the resources (e.g. fuelwood, water) that exist in the near-to-immediate vicinity (Jacobsen, 1997: 24). Indeed, camp-based refugees who are not given adequate provisions often attempt to secure them outside of immediate camp confines. Martin (2005: 332) notes that ‘[i]t is rare for refugees to be provided with construction materials or fuel for cooking, and these resources will often by necessity be collected from local environments’. His study of Bonga Camp in Ethiopia found that, due to greater resource scarcity, ‘[h]ost communities […] strongly expressed the view that life was considerably easier before the arrival of the camp’ (Martin, 2005: 336).

To reiterate, there are theoretical reasons to anticipate that factors associated with the camp settlement of refugee populations can lead to communal conflict – including among refugees, among local hosts, and between hosts and refugees. The preceding discussion leads to the following baseline expectation:

Hypothesis 1: Regions hosting camp-settled refugee populations experience higher rates of communal violence compared to other locations in the host state.

In addition to investigating how refugee encampment relates to levels of communal violence, it is important to explore whether context-specific conditions shape the impact of camp settlement on communal conflict. Insights from refugee studies suggest that camp settlement might have a greater impact on intercommunal tension when the camps/centers are situated among marginalized hosts (Chambers, 1986). The relative political status of host groups might therefore exert moderating effects such that the effect of refugee encampment on communal violence is amplified when the host population is marginalized politically. As Fjelde & Østby (2014: 743) observe, ‘[i]n many states in sub-Saharan Africa […] political elites in control of the government often seek legitimacy by favoring co-ethnics in the distribution of state patronage and provision of collective goods’; having more political power thus translates into greater access to resources and economic privileges. In effect, politically marginalized host populations – whom Chambers (1986: 246) calls ‘hidden losers’ – may be denied employment opportunities associated with the camps, which tend to be awarded to better educated and/or better politically positioned individuals. According to local Turkana hosts in Kenya, for instance, ‘all NGOs are headed by non-Turkana who practice nepotism, tribalism, favoritism, and sideline them because they are “primitive” and “unqualified”’ (Aukot, 2003: 77). Aukot found that only a small percentage of workers at Kakuma camp came from the host community; the vast majority were flown in from the capital. Compared to local host groups with political power, and thus privileged access to resources and economic privileges, host groups that are marginalized politically also may have a stronger sense of grievance related to the allocation of aid, infrastructure, services, and other privileges rendered to refugees. Therefore, communal conflict may result as ‘part of a strategy by those living in areas of minimal government intervention to regulate access to critical livelihood components such as water and land and to acquire wealth’ (Raleigh, 2010: 79). I therefore posit the following moderating effect:

Hypothesis 2: Regions hosting camp-settled refugee populations experience higher rates of communal violence when they are home to a politically marginalized group than otherwise.

In related terms, the economic characteristics of the host population may moderate the effect of camp settlement on communal conflict. Specifically, hosts in regions characterized by higher levels of economic inequality may have greater perceived grievances and a more acute sense of relative deprivation in relation to hosting camp-settled refugees compared to hosts in regions where inequality is lower. Competition over land is, for instance, likely to be a more pronounced problem for more economically marginalized hosts, who tend to be less mobile compared to those who are better off, and so ‘are liable to be trapped where they are’ (Chambers, 1986: 260). Inflated prices for food and other staple items like kerosene have been shown to more profoundly impact more economically marginalized hosts (Whitaker, 2002: 347). Whitaker (2002) also shows how wealthier and better-connected hosts were able to take advantage of the refugee presence in Tanzania, while poorer hosts who were not well positioned suffered disproportionately (also see Maystadt & Verwimp, 2014). Given that, as Whitaker (2002) argues, some hosts are considered ‘winners’ while poorer hosts consider themselves ‘losers’ related to the refugee presence, this real or perceived discrepancy may generate more intense intergroup (particularly interhost) comparisons. These comparisons may then take the form of collective economic grievances that facilitate group mobilization and result in intercommunal conflict between refugees and hosts, as well as conflict among local hosts that may or may not involve refugees directly. Interrefugee tension and conflict also may be more prevalent where refugees are encamped among economically marginalized host communities, given that resource/aid scarcity as well as a lack of viable opportunities for refugees may be more acutely perceived and experienced in these locations. In line with this discussion, I anticipate the following:

Hypothesis 3: Regions hosting camp-settled refugees experience higher rates of communal violence at higher levels of interpersonal economic inequality.

Finally, recent work on the flight patterns of refugees shows that they tend ‘to flee to kin groups in neighboring countries’ (Rüegger & Bohnet, 2018: 83). Where such ‘ethnic ties’ between groups in the refugee-sending country and host region exist, host groups may be more likely to perceive and cast the refugee presence as either a boon or a threat to their security, ways of life, and relative power – and respond by increasingly identifying along communal lines. If, for instance, there are disputes in the host community or among the refugees that fall along ethnic lines, an ethnically linked host or refugee population may be more likely to become involved than those who do not share such ties. The potential for conflict also may increase if hosts feel threatened because they are part of the same ethnic group that clashed with the refugees in the country of origin. If the refugees are ethnically linked to a local host group in the country of arrival, hosts from other ethnic groups to which the refugees are not tied may perceive/brand the refugees as demographic threats to the ‘ethnic balance’ (Salehyan & Gleditsch, 2006; also see Rüegger, 2019). This could heighten hostility among those who perceive their relative status to be under threat and those ethnically linked to the refugees, resulting in communal violence that may or may not involve refugees directly. Whitaker (2005), for example, demonstrates how politicians in several African countries have attempted to label (and thus discredit) their political opponents as non-citizen ‘foreigners’ by highlighting their opponents’ ethnic ties to refugees. Such xenophobic appeals may broaden social cleavages and lead to communal conflict between refugees and hosts, as well as between host groups. The presence of ethnic ties between groups in the refugee-sending country and groups in the host region might, therefore, intensify the effect of refugee encampment on communal conflict.

Hypothesis 4: Regions hosting camp-settled refugee populations experience higher rates of communal violence when there are ethnic ties between the host population and the refugees, compared to when there are no such ties.

In the following sections, I set out to determine if camp settlement independently impacts the occurrence of violent communal conflict. I then turn to assess whether the host population characteristics discussed above moderate these relationships.

Methods

The unit of analysis in this study is the first level administrative unit/year (e.g. province, region, state) within 39 countries⁸ in sub-Saharan Africa (2000–10), the region of the world where the majority of communal conflicts take place (Brosché & Elfversson, 2012; Sundberg, Eck & Kreutz, 2012). While some disaggregated studies use geographical grid squares as the units of analysis (typically 100-km² grid cells) in order to test constant (homogenously sized) cross-sections, grid cells are relatively non-intuitive and generally arbitrarily determined (Østby, Nordås & Rød, 2009: 309).⁹ The first level units come from the United Nations’ Second Level Administrative Boundary (SALB) dataset. The SALB data code intermittent boundary changes and have been confirmed by each country. Due to over-dispersion in the dependent variable, I estimate negative binomial event count models with population averaged robust standard errors clustered on subnational administrative units.¹⁰

Dependent variable

The dependent variable in each of the models below, Communal, relies on non-state conflict data from the Uppsala Conflict Data Program’s Geo-Referenced Event Dataset (UCDP GED, version 17.1 (2016); Croicu & Sundberg, 2017). UCDP defines a non-state conflict as ‘the use of armed force between two organized armed groups, neither of which is the government of a state, which results in at least 25 battle-related deaths in a year’ (Sundberg, Eck & Kreutz, 2012). Within these non-state conflicts meeting the 25 annual deaths threshold, an individual conflict event is included in GED if it results in a minimum of one fatality. Since GED does not differentiate among different forms of non-state conflict events, it was necessary to cross-reference GED with the UCDP Non-State Conflict Dataset (version 17.1; Allansson, Melander & Themnér, 2017; Sundberg, Eck & Kreutz, 2012) to separate communal conflict from the other forms contained in the dataset. Thus, my dependent variable Communal includes only the total number of deadly conflict events (count) among ‘groups that define themselves along identity lines, be it ethnic, clan, religious, national, or tribal identities’ (Sundberg, Eck & Kreutz, 2012: 353) in a subnational unit/year.¹¹ A negative binomial count model is appropriate for the data because there is skewness and overdispersion (the variance is greater than the mean) in incidents of communal violence – and the vast majority of these locations/years experience multiple communal conflict events.

Explanatory variables

I operationalize camp settlement in two ways. First, the predictor variable Refugee camp is a binary indicator (0/1) for whether or not the region hosts a planned settlement (‘camps, etc.’ in UNHCR data) in a subnational region/year.¹² Second, the predictor variable Camp-host proportion indicates the population size of camp-settled refugees as a proportion of the total, local host population size. I include a second indicator because several of the arguments outlined above suggest that the intensity (the ratio of refugees to local hosts) of the camp-based population is influential. For instance, population pressure on common natural resources is likely to be greater when larger numbers of camp-settled refugees are concentrated in more heavily populated areas than where there are smaller settlements, making resource scarcity more concentrated/acute and the potential for conflict greater in these locations. Larger numbers of spatially segregated, non-integrated refugees relative to local hosts may also be more likely to contribute to hosts’ perception that the refugee presence is a threat, increasing the likelihood of conflict. In order to assess the proposed moderating effects, I include the interaction terms Politically marginalized × Refugee camp and Politically marginalized × Camp-host proportion in order to account for whether areas hosting camp-settled refugees are home to a host group that is ‘powerless’, ‘discriminated’, or ‘irrelevant’ according to the Ethnic Power Relations dataset (= 1) or not (= 0). I also include the interaction terms Asset inequality × Refugee camp and Asset inequality × Camp-host proportion, accounting for the anticipated moderating effect of interpersonal economic inequality on the relationship between hosting camp-based refugee populations and communal conflict at the subnational level. Finally, I include the interaction terms Ethnic ties × Refugee camp and Ethnic ties × Camp-host proportion to account for the hypothesized effects of hosting camp-settled refugees where there are ethnic ties (0/1) between the local host population and the refugees’ country of origin.

The data on refugees are taken from the Geo-Refugee Dataset (Version 1.0; Fisk, 2014b).¹³ Each variable contains two categories of refugees: (1) legally recognized refugees and (2) those in a ‘refugee-like situation’, or who are determined by the UNHCR to be refugees, despite not yet having received official refugee status (UNHCR, 2010: 15). These populations are differentiated according to their accommodation type (‘camps, etc.’, ‘urban’, or ‘rural/dispersed’). ‘Camp-settled’ refugees are those whom the UNHCR determines as living in a planned refugee settlement (e.g. a camp, a collective center). The time frame for the analysis is 2000–10, as these are the years for which the subnational refugee location and demographic data are currently available.

I generated the variables representing host population political marginalization by cross-referencing two datasets, the Ethnic Power Relations dataset (EPR) and the Geo-Referenced Ethnic Power Relations dataset (Geo-EPR) (Vogt et al., 2015). The GeoEPR dataset is a dynamic, geographically referenced version of the Ethnic Power Relations (EPR) dataset that includes information on politically relevant ethnic groups based on information taken from expert surveys. The operational definition of ethnicity is based on Weber’s (1978: 427) conception: ‘any subjectively experienced sense of commonality based on the belief in common ancestry and shared culture’. The binary indicator Politically marginalized accounts for whether or not the region is home to a group that experiences one or more of the following ‘types’ of political marginalization: political powerlessness, political discrimination, and/or political irrelevance. Powerless and discriminated-against groups are, by definition, politically relevant in their country, meaning that ‘at least one political organization claims to represent it in national politics or […] its members are subjected to state-led political discrimination’ (Cederman, Wimmer & Min, 2010: 99). Politically irrelevant groups, in contrast, typically are ‘historically marginalized, demographically small, geographically disparate, and largely considered the poorest and most vulnerable in society’ (Raleigh, 2014: 95).

The data on interpersonal economic inequality come from Fjelde & Østby’s (2014) subnational inequality data, which they constructed using the household asset and geographic information available in household Demographic and Health Surveys (DHS) for 32 African countries. The authors use DHS to compile a ‘household assets index’ based on ‘information on whether the household has electricity, a radio, a television, a refrigerator, a bicycle, a motorcycle, and/or a car’ which they then use to calculate Gini coefficients for household assets for first-level administrative units. Higher Gini coefficients indicate ‘higher levels of interpersonal inequality in asset distribution’ (Fjelde & Østby, 2014: 748–749).

The ethnic ties variable captures whether the refugees are hosted in a subnational region in which any host ethnic group shares ties with an ethnic group in the refugees’ origin country (= 1) or not (= 0).¹⁴ I constructed this variable by first determining the ethnicity/ethnicities of the local host population in each refugee-hosting region using Geo-EPR (Vogt et al., 2015). Next, I determined whether the refugees in each host region originated from a country where the same ethnic group is present, using EPR. The indicator therefore combines subnational-level data on refugees’ country (or countries) of origin from UNHCR with subnational-level data on the locations and local names of ethnic groups. While recent work shows that ethnic refugee groups often flee to countries where kin groups are located (Rüegger & Bohnet, 2018), data on the ethnicity of refugee populations is not currently available at the subnational level. I am thus unable to verify that the refugees have direct ethnic ties to their local hosts. Establishing whether or not there is an ethnic linkage between the local host population and the refugees’ country of origin is nevertheless important given that the study is focused on communal conflict dynamics, which center at the local level.

Controls

The controls are based on explanations for communal conflict in previous literature. I control for Population (logged) at the subnational level using data from the Gridded Population of the World dataset (version 3), since the possibility for conflict may be greater among larger populations in general. Given that interregional variation in wealth has been linked to conflict in past studies (e.g. Buhaug et al., 2011), I include a control for the absolute level of Income per capita in each administrative region. For this, I use data from Fjelde & Østby (2014), who first calculated administrative-level per capita income data by overlaying administrative polygons with Nordhaus’s (2006) G-Econ data (version 4.0), in which average income estimates in terms of purchasing power parity are generated at five-year intervals for 1° × 1° grid cells, and calculated the area weighted sum of the cells that fall within administrative-level polygons. They then divided this aggregate estimate by population size for each administrative unit (province, region, state, etc.). I control for Capital distance, the logged distance in kilometers from the capital region to other regions, since an alternative explanation for communal violence is a lack of state capacity or territorial control in areas located farther away from the government’s base in the capital (Raleigh, 2010). Controlling for distance from the capital also helps account for the possibility that the periphery contains populations that unequally benefit from government patronage, programs, and infrastructure (and thus are economically marginalized), which by itself could increase the likelihood of conflict in peripheral locations. I include a binary control variable for whether or not the administrative unit is located within a democratic country. Democracy takes a value of 1 if the region is located in a country that has a Polity score of +5 or higher and 0 if the country’s Polity score falls below +5 in any given year (Polity 2 Index; Marshall, 2014). Since communal conflict may coincide with armed conflict, I control for Battles (logged) between government forces and insurgents using data on state-based conflict events in GED. Past work, in addition, suggests that communal violence is likely to occur alongside host government incitements and acts of violence against a refugee community (e.g. Onoma, 2013). Others maintain that host populations are likely to blame refugees for rebel group attacks on host communities, potentially leading to intergroup violence (Mogire, 2011). I therefore include a control for One-sided violence, which is the logged incident count of intentional violence against civilians perpetrated by the government or armed groups in each administrative unit. These data also are taken from GED. Finally, I include a spatial lag, Conflict adjacent, and the temporal lag, Previous communal, since violence may be more common in locations with neighboring units in conflict, and in areas where communal violence occurred previously.

Table I.

Refugee encampment and the occurrence of communal conflict

	(1)	(2)
Variables	Communal	Communal
Refugee camp	0.596*
	(0.290)
Camp–host proportion_(log)		1.012*
		(0.444)
Population_(log)	0.546**	0.559**
	(0.163)	(0.165)
Income_{(log, lag)}	0.027	0.022
	(0.173)	(0.175)
Democracy	–0.734*	–0.743*
	(0.303)	(0.298)
Capital distance_(log)	0.150^†	0.148^†
	(0.082)	(0.082)
Previous communal_{(log, lag)}	0.722*	0.705*
	(0.312)	(0.315)
Conflict adjacent	1.543**	1.521**
	(0.349)	(0.351)
Battles_{(log, lag)}	–0.280	–0.278
	(0.250)	(0.250)
Onesided_(log)	0.375**	0.365*
	(0.146)	(0.143)
Constant	–11.88**	–12.01**
	(2.889)	(2.947)
Observations	5,799	5,799
Number of administrative units	539	539

Standard errors in parentheses. **p < 0.01, *p < 0.05, ^† p < 0.1.

Findings

Table A1 in the Online appendix displays the summary statistics. Models 1 and 2 in Table I present the primary results. Model 1 shows that the coefficient for the refugee camp variable is positive and statistically significant (0.60, p < 0.5). The coefficient for Camp-host proportion also is positive and statistically significant in Model 2 (1.01, p < 0.05). I therefore find support for Hypothesis 1, which anticipates that hosting camp-settled refugees is a significant predictor of intercommunal conflict events. Incident rate ratios (IRRs) indicate that the effect of simply hosting camp-settled refugees, with all other variables held constant, is to increase the frequency of communal conflict events by about 82% (IRR = 1.82). The effect of camp-host proportion is more pronounced: a one-unit increase in the log of camp-host proportion increases the frequency of communal conflict events by a factor of 2.75 (IRR = 2.75). These estimates indicate that while both variables are substantively meaningful in explaining the pattern of communal conflict events, the intensity of the encamped refugee population has a more significant influence.¹⁵

Results for the controls generally are consistent across Models 1 and 2. Regional population size is positively and significantly related to communal conflict in both models. The finding that higher levels of regional income per capita are correlated with higher rates of communal conflict does not fall in line with previous findings; however, this variable is not statistically significant in either model. Regions within democratic countries experience significantly less communal conflict, a finding in line with the previous literature. Distance from the capital is positively correlated with communal conflict, and this effect is statistically significant in both models. Controls for spatial and temporal diffusion are positive and significant – communal conflict in a region is significantly more likely if it took place in the previous year, as well as if at least one neighboring region experienced conflict in the same year. While communal conflict is less common in regions that experience ‘state-based’ battles between rebels and the government, communal conflict is shown to be more likely to occur in regions that experience one-sided violence against civilians.¹⁶

Next, I turn to Models 3–6 in Table II, which explore the conditional hypotheses related to host population political and economic inequality. To refresh, Hypothesis 2 proposed that the effects of refugee encampment on communal conflict might be amplified where a host group is politically marginalized. Results in Models 3 and 4 in Table II show that the signs on both interaction term coefficients Politically marginalized × Refugee camp and Politically marginalized × Camp-host proportion are positive in line with expectations, but neither is statistically significant. Hypothesis

Table II.

Refugee encampment, host population characteristics, and the occurrence of communal conflict

	(3)	(4)	(5)	(6)
Variables	Communal	Communal	Communal	Communal
Refugee camp	–0.091		–0.809
	(0.812)		(1.438)
Politically marginalized	0.059	0.037
	(0.348)	(0.348)
Politically marginalized × refugee camp	0.872(0.866)
Camp–host proportion		–0.073		–0.963
		(1.223)		(2.094)
Politically marginalized × camp-host proportion		1.367(1.295)
Asset inequality			1.607	1.436
			(1.041)	(1.066)
Asset inequality × refugee camp			2.416
			(2.277)
Asset inequality × camp-host proportion				3.415(3.230)
Population_(log)	0.536**	0.549**	0.693**	0.720**
	(0.167)	(0.168)	(0.190)	(0.200)
Income_{(log, lag)}	0.024	0.021	0.776**	0.774**
	(0.172)	(0.174)	(0.226)	(0.229)
Democracy	–0.728*	–0.742*	–0.296	–0.316
	(0.298)	(0.291)	(0.345)	(0.336)
Capital distance_(log)	0.131	0.132	0.043	0.043
	(0.081)	(0.082)	(0.102)	(0.103)
Previous communal_{(log, lag)}	0.701*	0.684*	0.520	0.492
	(0.318)	(0.320)	(0.327)	(0.325)
Conflict adjacent	1.519**	1.504**	1.443**	1.416**
	(0.350)	(0.351)	(0.432)	(0.433)
Battles_{(log, lag)}	–0.284	–0.281	0.003	0.016
	(0.245)	(0.246)	(0.141)	(0.139)
Onesided_(log)	0.366*	0.355*	0.505**	0.492**
	(0.144)	(0.142)	(0.116)	(0.115)
Constant	–11.60**	–11.77**	–19.44**	–19.72**
	(3.004)	(3.034)	(4.052)	(4.196)
Observations	5,799	5,799	4,424	4,424
Number of administrative units	539	539	416	416

Standard errors in parentheses. **p < 0.01, *p < 0.05, ^† p < 0.1.

3 proposed that higher levels of host asset inequality intensify the effects of refugee encampment on communal conflict. Again, the interaction coefficients for Asset inequality × Refugee camp and Asset inequality × Camp-host proportion are positive but non-significant, as shown in Models 5 and 6.

As noted above, the coefficients on the interaction terms in Models 4 and 5 are not statistically significant, which suggests that the hypothesized moderating effects are not present (Pepinsky, 2018). While the marginal effects of the moderating variables’ values are not shown to be significantly different from one another, it is nevertheless important to go beyond looking at the signs and significance of the interaction terms to evaluate the substantive effects of the moderating variables and in particular determine whether the marginal effects for some values of the moderating variables are statistically distinguishable from zero (e.g. Brambor, Clark & Golder, 2006; Berry, Golder & Milton, 2012; Kingsley, Noordewier & Vanden Bergh, 2017).¹⁷ I therefore calculate

Figure 1.

Estimated marginal effect of encamped refugee intensity moderated by host group political marginalization

the marginal effects at different levels of the moderating variables. In the interest of space, only the marginal effects results for the encamped refugee intensity indicators are shown below. The marginal effects of the binary refugee camp indicators are qualitatively very similar and are reported in the Online appendix.

Figure 1 shows that when a politically marginalized group is not present (Marginalized = 0), the marginal effect of refugee intensity on communal conflict is negative and not statistically different from zero. When a politically marginalized group is present (Marginalized = 1), the effect is positive and significantly different from zero at a 95% confidence interval. Thus, the findings for Hypothesis 2 are mixed – the interaction term was not significant, though the marginal effects are consistent with this hypothesis insofar as that there is only an effect if the refugees are located in areas with politically marginalized groups. This provides some suggestive evidence that, in the context of this dataset, the relationship between refugee intensity and communal conflict is stronger in areas with such marginalized populations. Next, I turn to evaluate Hypothesis 3, which anticipates that intraregional asset inequality amplifies the effect of refugee intensity on communal conflict. Although Figure 2 shows that the effect of encamped refugee intensity on conflict increases at higher levels of intraregional asset inequality, there are no statistically significant marginal effects across any of the values of the moderating variable. Hypothesis 3 is therefore not supported.¹⁸

Figure 2.

Estimated marginal effect of encamped refugee intensity moderated by intraregional asset inequality

Turning to examine models that correspond to Hypothesis 4 (Models 7 and 8 in Table III), I find that the interaction between hosting a refugee camp and the presence of origin country/host population ethnic ties is positive and significant in Model 7 (1.93, p < 0.05).¹⁹ The results in Model 8 likewise indicate that the interaction between camp proportion and ethnic ties is positive and statistically significant (3.74, p < 0.01). In order to further gauge support for the hypothesis positing that ethnic ties moderate the effects of encamped refugee intensity on communal conflict, I calculate the marginal effects of the coefficients.

Figure 3 indicates that the marginal effect of refugee intensity on communal conflict increases more substantially when there are ethnic ties between the refugees’ country of origin and groups in the host region than when there are no such ties. More specifically, the effect of refugee intensity on conflict is positive and significantly different from zero at the 95% confidence interval when ethnic ties are present, but not when ethnic ties are absent. This suggests that there is some evidence of moderating effects in line with Hypothesis 4.

Table III.

Refugee encampment, ethnic ties, and the occurrence of communal conflict

	(7)	(8)
Variables	Communal	Communal
Refugee camp	0.164
	(0.398)
Ethnic ties	–1.243^†	–1.832*
	(0.642)	(0.866)
Refugee camp × ethnic ties	1.933*
	(0.782)
Camp–host proportion		0.289
		(0.612)
Camp–host proportion × ethnic ties		3.735**
		(1.432)
Population_(log)	0.535**	0.553**
	(0.163)	(0.167)
Income_{(log, lag)}	0.017	0.015
	(0.165)	(0.168)
Democracy	–0.746**	–0.755**
	(0.283)	(0.281)
Capital distance_(log)	0.150^†	0.148^†
	(0.085)	(0.087)
Previous communal_{(log, lag)}	0.693*	0.672*
	(0.310)	(0.316)
Conflict adjacent	1.521**	1.495**
	(0.348)	(0.353)
Battles_{(log, lag)}	–0.263	–0.255
	(0.257)	(0.260)
Onesided_(log)	0.363*	0.364*
	(0.150)	(0.150)
Constant	–11.60**	–11.80**
	(2.817)	(2.904)
Observations	5,799	5,799
Number of administrative units	539	539

Standard errors in parentheses. **p < 0.01, *p < 0.05, ^† p < 0.1.

Figure 3.

Estimated marginal effects of encamped refugee intensity moderated by origin/host ethnic ties, 2000–10

Extended analyses: Refugee self-settlement

A related empirical question subject to much debate in the field of refugee studies centers on the relative merits and drawbacks of refugee self-settlement for host populations. Kibreab (1989), for instance, argues that self-settlement can be detrimental to hosts because self-settled refugees enter into direct competition with locals for jobs. At the same time, self-settled refugees tend to be integrated/assimilated into the host community and local economy (Hansen, 1979), which could have a dampening effect on the likelihood of communal conflict associated with their presence. While a more comprehensive treatment and full analysis of these dynamics is beyond the scope of this study, I perform some initial tests to probe the impact of hosting self-settled refugee populations at the subnational level.²⁰ The initial results (which I report in the Online appendix due to space constraints) indicate that the general presence of a self-settled refugee population is negatively and significantly related to communal conflict (Table A3, Model 1). The proportion of self-settled refugees relative to the host population also is negatively and significantly correlated with communal conflict (Table A3, Model 2).

Discussion and conclusions

Communal conflict, despite its prevalence and deadliness, is under-researched and therefore not yet well understood. In this study, I focus on a research question that has been understudied but deserves greater attention due to implications for the security needs of refugees, host populations, and overall refugee policy: does settling refugees in camps increase communal conflict? In an initial set of tests, I find that regions hosting camps in general, as well as those hosting greater population intensities of camp-based refugees, experience significantly higher rates of communal conflict at the subnational level. I then move forward to investigate whether particular host population characteristics moderate the effect of refugee encampment on communal conflict, and I find some suggestive evidence that this relationship is stronger in areas where host groups are politically marginalized.²¹ I find evidence that origin country/host region ethnic ties have a significant moderating effect, in line with my expectations. Moreover, results from an extended set of analyses further show that the form of refugee settlement matters, as the presence and population intensity of self-settled refugees are related to decreases in the occurrence of communal conflict.

It is, by now, clear that host governments in general prefer planned refugee settlement to self-settlement (e.g. Mogire, 2011) and attempt to justify encampment policies on security grounds. International agencies likewise have long considered camps ‘a “cornerstone” of humanitarian and state responses’ to refugee crises (Peteet, 2005: 29). Yet a growing body of research demonstrates the ways in which encampment policies can be detrimental. This article’s findings provide further indication that refugee encampment often negatively impacts the physical security of host populations and refugees. While additional tests show that hosting a self-settled refugee population is negatively and significantly related to communal conflict, additional research is needed to investigate the nature (including the possible conditionality) of this relationship. The basic findings nevertheless imply that policies could be designed to reduce the likelihood that host regions experience such violence. For instance, they suggest that efforts to redress host population marginalization and bolster mutually beneficial economic and social interactions between locals and refugees could help to mitigate communal violence. Future work should examine other measures of host population economic marginalization; the non-finding for intraregional asset inequality could be tied to a lack of available data for some refugee-hosting countries. Future studies also might look more directly at how refugee self-settlement influences refugee–host dynamics, as well as investigate effects of some of the more specific theoretical mechanisms related to camps outlined above – for instance, how variation in levels of humanitarian assistance and environmental degradation impact conflict.

Footnotes

Replication data

The dataset and do file for the empirical analyses in this article, as well as the Online appendix, can be found at .

Acknowledgments

I’m grateful to all of the participants at the Refugees, Aid, and Conflict workshop at the University of Arizona, in particular Alex Braithwaite, Faten Ghosen, Will Moore, Idean Salehyan, and Burcu Savun. I also thank Melissa Ziegler-Rogers, Jennifer Merolla, Prachi Jain, Puspa Amri, and the anonymous reviewers for their helpful comments.

Funding

The research reported here was supported by Harvard Business School.

Notes

References

Agblorti

Samuel Kofi Miledzi

(2011) Humanitarian assistance to refugees in rural Ghana: Implications for refugee–host relations. Norwegian Journal of Geography 65(2): 75–82.

Allansson

Marie

Melander

Erik

Themnér

Lotta

(2017) Organized violence, 1989–2016. Journal of Peace Research 54(4): 574–587.

Aukot

Ekuru

(2003) ‘It is better to be a refugee than a Turkana in Kakuma’: Revisiting the relationship between hosts and refugees in Kenya. Refuge 21(3): 73–83.

Bakewell

Oliver

(2014) Encampment and self-settlement. In: Fiddian-Qasmiyeh

Elena

Loescher

Gil

Long

Katy

Sigona

Nando

(eds) The Oxford Handbook of Refugee and Forced Migration Studies. Oxford: Oxford University Press, 127–138.

Berry

William D

Golder

Matt

Milton

Daniel

(2012) Improving tests of theories positing interaction. Journal of Politics 74(3): 653–671.

Betts

Alexander

(2009) Development assistance and refugees, towards a North–South grand bargain? Forced Migration Policy Briefing 2. Refugee Studies Center, University of Oxford.

Black

Richard

(1994) Forced migration and environmental change: The impact of refugees on host environments. Journal of Environmental Management 42(3): 261–277.

Black

Richard

(1998) Putting refugees in camps. Forced Migration Review 2(August): 1–7.

Böhmelt

Tobias

Bove

Vincenzo

Gleditsch

Kristian Skrede

(2019) Blame the victims? Refugees, state capacity and non-state actor violence. Journal of Peace Research 56(1): 73–87.

10.

Brambor

Thomas

Clark

William Roberts

Golder

Matt

(2006) Understanding interaction models: Improving empirical analyses. Political Analysis 14(1): 63–82.

11.

Brosché

Johan

(2014) Masters of War: The Role of Elites in Sudan’s Communal Conflicts. Unpublished PhD dissertation, Uppsala University.

12.

Brosché

Johan

Elfversson

Emma

(2012) Communal conflict, civil war, and the state: Complexities, connections, and the case of Sudan. African Journal on Conflict Resolution 12(1): 33–60.

13.

Buhaug

Halvard

Gleditsch

Kristian Skrede

Holtermann

Helge

Østby

Gudrun

Tollefsen

Andreas

(2011) It’s the local economy, stupid! Geographic wealth dispersion and conflict outbreak location. Journal of Conflict Resolution 55(5): 814–840.

14.

Cederman

Lars-Erik

Wimmer

Andreas

Min

Brian

(2010) Why do ethnic groups rebel? New data and analysis. World Politics 62(1): 87–119.

15.

Chambers

Robert

(1986) Hidden losers? The impact of rural refugees and refugee programs on poorer hosts. International Migration Review 20(2): 245–263.

16.

Chiasson

Sophie

(2015) The state of freedom of movement for refugees in Tanzania: An overview. International Refugee Rights Initiative, 1 September (https://reliefweb.int/report/united-republic-tanzania/state-freedom-movement-refugees-tanzania-overview).

17.

Crisp

Jeff

Jacobsen

Karen

(1998) Refugee camps reconsidered. Forced Migration Review 3(December): 27–30.

18.

Crisp

Jeff

Long

Katie

(2016) Safe and voluntary refugee repatriation: From principle to practice. Journal on Migration and Human Security 4(3): 141–147.

19.

Croicu

Mihai

Sundberg

Ralph

(2017) UCDP GED Codebook Version 17.1. Department of Peace and Conflict Research, Uppsala University.

20.

Duncan

Christopher R

(2005) Unwelcome guests: Relations between internally displaced persons and their hosts in North Sulawesi, Indonesia. Journal of Refugee Studies 18(1): 25–46.

21.

Eck

Christine

(2014) The law of the land: Communal conflict and legal authority. Journal of Peace Research 51(4): 441–454.

22.

Fisk

Kerstin

(2014a) Refugee geography and the diffusion of armed conflict in Africa. Civil Wars 16(3): 255–275.

23.

Fisk

Kerstin

(2014b) Geo-refugee dataset, version 1. Harvard Dataverse.

24.

Fisk

Kerstin

(2018) One-sided violence in refugee-hosting areas. Journal of Conflict Resolution 62(3): 529–556.

25.

Fjelde

Hanne

Østby

Gudrun

(2014) Socioeconomic inequality and communal conflict: A disaggregated analysis of sub-Saharan Africa, 1990–2008. International Interactions 40(5): 737–762.

26.

Fjelde

Hanne

Uexkull

Nina von

(2012) Climate triggers: Rainfall anomalies, vulnerability and communal conflict in sub-Saharan Africa. Political Geography 31(7): 444–453.

27.

Hansen

Art

(1979) Once the running stops: Assimilation of Angolan refugees into Zambian border villages. Disasters 3(4): 369–374.

28.

Harrell-Bond

Barbara

(2000) Are refugee camps good for children? New Issues in Refugee Research, Working Paper no. 29 (http://www.unhcr.org/3ae6a0c64.pdf).

29.

Hendrix

Cullen S

Salehyan

Idean

(2012) Climate change, rainfall, and social conflict in Africa. Journal of Peace Research 49(1): 35–50.

30.

Hovil

Lucy

(2014) With camps limiting many refugees, the UNHCR’s policy change is welcome. Guardian 2 October (https://www.theguardian.com/global-development/poverty-matters/2014/oct/02/unhcr-policy-change-refugee-camps).

31.

Hunter

Meredith

(2009) The failure of self-reliance in refugee settlements. POLIS Journal 2(Winter): 1–46.

32.

Hyndman

Jennifer

(1997) Refugee self-management and the question of governance. Refuge 16(2): 16–22.

33.

Jacobsen

Karen

(1997) Refugees’ environmental impact: The effect of patterns of settlement. Journal of Refugee Studies 10(1): 19–36.

34.

Kahl

Colin

(2006) States, Scarcity, and Civil Strife in the Developing World. Princeton, NJ: Princeton University Press.

35.

Kibreab

Gaim

(1989) Local settlements in Africa: A misconceived option? Journal of Refugee Studies 2(4): 468–490.

36.

Kibreab

Gaim

(2003) Displacement, host governments’ policies, and constraints on the construction of sustainable livelihoods. International Social Science Journal 5(175): 57–67.

37.

Kibreab

Gaim

(2012) Invisible integration in the Greater Horn region. In: Mengisteab

Kidane

Berekteab

Redie

(eds) Regional Integration, Identity and Citizenship. Martlesham: Boydell & Brewer, 69–110.

38.

Kingsley

Allison F

Noordewier

Thomas G

Bergh

Richard G Vanden

(2017) Overstating and understating interaction results in international business research. Journal of World Business 52(2): 286–295.

39.

Lawrie

Nicolette

Damme

Wim Van

(2003) The importance of refugee–host relations: Guinea 1990–2003. Lancet 362(9383): 575.

40.

Lischer

Sarah Kenyon

(2005) Dangerous Sanctuaries. Ithaca, NY: Cornell University Press.

41.

Loescher

Gil

Milner

James

(2005) Protracted Refugee Situations: Domestic and International Security Implications. New York: Routledge.

42.

Marshall

Monty

(2014) Polity IV project: Political regime characteristics and transitions, 1800–2013 (http://www.systemicpeace.org/polity/polity4.htm).

43.

Martin

Adrian

(2005) Environmental conflict between refugee and host communities. Journal of Peace Research 42(3): 329–346.

44.

Maystadt

Jean-Francois

Verwimp

Philip

(2014) Winners and losers among a refugee-hosting population. Economic Development and Cultural Change 62(4): 769–809.

45.

Mogire

Edward

(2011) Victims as Security Threats: Refugee Impact on Host State Security in Africa. Farnham: Ashgate.

46.

Nordhaus

William D

(2006) Geography and macroeconomics: New data and new findings. Proceedings of the National Academy of Sciences 103(10): 3510–3517.

47.

Nordic Agency for Development and Ecology (NORDECO) (2010) Impacts of Dadaab Refugee Camps on Host Communities. Draft Study Report. Copenhagen.

48.

Onoma

Ato Kwamena

(2013) Anti-Refugee Violence and African Politics. Cambridge: Cambridge University Press.

49.

Østby

Gudrun

Nordås

Ragnhild

Rød

Jan Ketil

(2009) Regional inequalities and civil conflict in sub-Saharan Africa. International Studies Quarterly 53(2): 301–324.

50.

Pepinsky

Thomas B

(2018) Visual heuristics for marginal effects plots. Research & Politics 5(1): 1–9.

51.

Peteet

Julie

(2005) Landscape of Hope and Despair: Palestinian Refugee Camps. Philadelphia, PA: University of Pennsylvania Press.

52.

Raleigh

Clionadh

(2010) Political marginalization, climate change, and conflict in African Sahel states. International Studies Review 12(1): 69–86.

53.

Raleigh

Clionadh

(2014) Political hierarchies and landscapes of conflict across Africa. Political Geography 42(September): 92–103.

54.

Rüegger

Seraina

(2019) Refugees, ethnic power relations, and civil conflict in the country of asylum. Journal of Peace Research 56(1): 42–57.

55.

Rüegger

Seraina

Bohnet

Heidrun

(2018) The ethnicity of refugees: A new dataset for understanding flight patterns. Conflict Management and Peace Science 35(1): 65–88.

56.

Salehyan

Idean

(2009) Rebels Without Borders: Transnational Insurgencies in World Politics. Ithaca, NY: Cornell University Press.

57.

Salehyan

Idean

Gleditsch

Kristian Skrede

(2006) Refugees and the spread of civil war. International Organization 60(2): 335–366.

58.

Sebba

Kalyango Ronald

(2006) Land Conflicts and Their Impact on Refugee Women’s Livelihoods in Southwestern Uganda. Research Paper no. 127. Geneva: UNHCR.

59.

Shaver

Andrew

Zhou

Yang-Yang

(2015) Questioning refugee camps as sources of conflict. Meeting of the International Studies Association, 25 September (https://pdfs.semanticscholar.org/bb53/49abe427c3f21482e3a7ae6d6e92d9589ac1.pdf).

60.

Smith

Merrill

(2004) Warehousing refugees: A denial of rights, a waste of humanity. World Refugee Survey (http://refugees.org/wp-content/uploads/2015/12/Warehousing-Refugees-Campaign-Materials.pdf).

61.

Stedman

Stephen John

Tanner

Fred

(2003) Refugee Manipulation: War, Politics, and the Abuse of Human Suffering. Washington, DC: Brookings Institution.

62.

Sundberg

Ralph

Eck

Kristine

Kreutz

Joakim

(2012) Introducing the UCDP Non-State Conflict dataset. Journal of Peace Research 49(2): 351–362.

63.

Terry

Fiona

(2003) Condemned to Repeat? The Paradox of Humanitarian Action. Ithaca, NY: Cornell University Press.

64.

UNHCR (2010) UNHCR Statistical Yearbook, 2010 (http://www.unhcr.org/uk/statistics/country/4ef9cc9c9/unhcr-statistical-yearbook-2010-10th-edition.html).

65.

Vogt

Manuel

Bormann

Nils-Christian

Rüegger

Seraina

Cederman

Lars-Erik

Hunziker

Philipp

Girardin

Luc

(2015) Integrating data on ethnicity, geography, and conflict: The ethnic power relations data set family. Journal of Conflict Resolution 59(7): 1327–1342.

66.

Von Uexkull

Nina

Pettersson

Therese

(2013) What they are fighting for: Conflict issues in African non-state armed conflicts 1989–2011. Paper presented at the meeting of the European Network of Conflict Research (ENCoRe), Amsterdam, 24–26 April.

67.

Weber

Max

(1978) Economy and Society: An Outline of Interpretive Sociology. Berkeley, CA: University of California Press.

68.

Whitaker

Beth Elise

(2002) Refugees in Western Tanzania: The distribution of burdens and benefits among local hosts. Journal of Refugee Studies 15(4): 339–358.

69.

Whitaker

Beth Elise

(2005) Citizens and foreigners: Democratization and the politics of exclusion in Africa. African Studies Review 48(1): 109–126.

70.

Zolberg

Aristide

Suhrke

Astri

Aguayo

Sergio

(1989) Refugee Warriors: A Problem of Our Time. Oxford: Oxford University Press.