Abstract
This article proposes a descriptive typology of civilian resettlement patterns in civil wars. The patterns vary in two dimensions: whether or not displaced civilians cluster together or resettle independently, and if they remain within their home country or not. The combination of the factors leads to four resettlement patterns: expulsion, segregation, integration, and dispersion. Expulsion and segregation occur when the displaced cluster, either within the home state (segregation) or beyond it (expulsion). Integration and dispersion occur when the displaced do not cluster but seek to blend in with other communities, either abroad (dispersion) or within core cities and towns in the state (integration). After introducing the typology and illustrating it with examples, the article engages in theory-building to explain variation in resettlement patterns. It argues that resettlement forms are based on the type of displacement that civilians experience, and the perpetrator of the violence. The displacement type influences individuals’ best strategy for achieving relative safety. Within and across wars, groups that experience political cleansing are likely to cluster together for safety. The best destination options for the displaced to resettle depend on the perpetrator, which lead to clustering either within a state if the actor is non-state, or outside the state if the actor is the state or an ally. The argument is illustrated with examples. Finally, the article considers the implications of resettlement patterns for violence, conflict, and state-building.
Introduction
On the eve of the US invasion of Iraq in 2003, humanitarian organizations anticipated thousands of refugees and constructed camps along the Syrian border to receive them (Margesson, Bruno & Sharp, 2009: 7). But after the war started, something strange happened: hardly anyone came. In fact, families who had been forcibly relocated under Saddam Hussein’s regime began seeking ways to return to their original communities. Though there were thousands of refugees and internally displaced people (IDPs), the next two years passed with a surprisingly low level of displacement. Then Samarra’s shrine was bombed in February 2006, and everything changed. Sectarian violence was vicious and displacement tripled within months. But the displaced did not leave the country, as the United Nations High Commissioner for Refugees (UNHCR) initially anticipated. Instead, nearly three-quarters remained within Iraq. How can we understand these changes over time? Why did the displaced remain in Iraq rather than move to a refugee camp?
The existing literature gives us a partial answer: as violence increases, so does displacement (e.g. Davenport, Moore & Poe, 2003; Zolberg, Suhrke & Aguayo, 1989). The second question, why the majority of the displaced remained within Iraq, is a puzzle. The predominant approach to resettlement draws on the ‘push-pull’ model from migration studies, 1 and reaches the logical conclusion that individuals and households will move to places with better conditions than where they reside. Several macro factors are associated with the scale of resettlement in general, such as violence and economic opportunity (Moore & Shellman, 2006). But it is hard to argue that Iraq between 2006 and 2008 was better off in these terms than Syria, or Jordan.
I argue that one reason that the Iraq case is puzzling stems from a key assumption of the push-pull model: that individuals independently decide whether and where to go. The safety of each individual is taken to be the result of the characteristics of the possible locations that the displaced could move to, not an outcome that depends on the decisions of others who are also displaced. 2 In contrast, I argue that this is an important aspect of variation in resettlement: that the displaced anticipate whether their safety depends in part on others who are also displaced, or not.
When the displaced think their safety will be improved if they resettle with others who have also been displaced, I expect that they will cluster together in new locations. I introduce a descriptive typology in this article that combines this dimension, whether or not the displaced cluster, with the more common distinction of whether or not the displaced are refugees or internally displaced. The interaction of how and where the displaced resettle creates four ideal-type patterns: expulsion, segregation, dispersion, and integration. Expulsion and segregation occur when the displaced resettle in clusters, beyond and within the home state, respectively. Dispersion and integration occur when the displaced resettle independently, again outside and within a state, respectively. This typology offers a basis for comparison across and within wars – one that moves beyond analyzing the scale of displacement to considering the nature of resettlement as well. These resettlement patterns can also be consequential for outcomes such as ongoing violence, conflict, and state-building.
The typology also provides the tools to frame the questions about the Iraq case differently. Not only did a majority of the displaced remain within Iraq, but thousands remained in Baghdad, forming sectarian enclaves (Lischer, 2008). What can account for this segregation of the internally displaced? The typology helps us characterize a key aspect of the displacement, and opens up new lines of inquiry such as: what explains when and where we observe each resettlement pattern within and across wars? I argue that the form of displacement civilians experience shapes whether or not they will try to resettle with others, and the actor that perpetrates the violence explains if they will resettle within their home state or abroad.
The theory builds on recent work that characterizes forms of displacement based on how armed groups target civilians and how civilians respond. There are three types of displacement: (1) individual escape, when people react to selective targeting by an armed group; (2) mass evasion, when civilians avoid indiscriminate violence; and (3) political cleansing, when armed groups expel civilians through collective targeting, based on a shared trait (Steele, 2017). I argue that those who experience cleansing are the most likely to cluster together in their new location, because they face an ongoing security risk. Because the displaced are targeted based on a shared trait, such as ethnicity, sect, or political identity, that is difficult to shed, they are vulnerable to further violence. If a household targeted for their identity resettles independently, they will stand out and face potential harm again. From a household’s perspective, this risk is in part mitigated by resettling with others similarly targeted (Steele, 2018). In contrast, those who experience selective targeting can try to seek anonymity in cities or new communities, appearing as a ‘normal’ migrant. Civilians who face indiscriminate violence can evade it by relocating, sometimes even for a brief period. Unlike for those collectively targeted, the violence the indiscriminately targeted experience is related to their location; as a result, changing their location can reduce the threat they face, whether or not they resettle with others.
Whether the displaced are likely to cluster within the home state or abroad depends on the perpetrator of the violence. If a state agent or ally targeted the group, then the displaced are less likely to find safety in areas that the state controls: they will remain at risk in their new location within the state. As a result, they are likely to seek safety beyond the reach of the state, across an international border. If targeted by a rebel group, then the targeted group is more likely to remain within their home state’s borders, closer to state protection.
The descriptive and theoretical framework helps us grasp what happened in Iraq, and why. As non-state militias threatened civilians based on their sect, displacement increased and resettlement took the form of segregation.
This article draws on a variety of cases across regions to illustrate the types and the argument. It also demonstrates the typology’s utility by describing the research agenda that it permits, and the implications that it reveals for violence, conflict, and state-building. The article continues in the next section by reviewing the literature on civilian displacement and resettlement. The third section introduces the typology of resettlement patterns. The fourth develops an explanation of when and where we are likely to observe each resettlement pattern, based on the form of displacement, and the perpetrator of the violence. In section six, I discuss the political implications of the resettlement types. Section seven concludes.
Destinations of the displaced in the literature
The ‘push-pull’ model of migration has had an enduring influence on the displacement literature (Adhikari, 2012; Ibáñez, 2008; Kunz, 1973; Petersen, 1958). 3 It reasons that the displaced are analogous to migrants, and that they compare the conditions in their home community with those in other regions when deciding whether and where to go. While understanding individuals’ and households’ decisionmaking is inherently important, it is a limited approach for understanding the broader patterns of displacement and resettlement. In part, this is because the model treats individuals’ decisions as independent of one another, even though we know that networks matter, both for migrants and the displaced (e.g. Hein, 1993; Lubkemann, 2008). Further, the primary aggregate implication of individuals’ decisions is the scale of displacement and resettlement. Moore & Shellman (2006), for instance, consider the state-level factors that influence the displaced to cross an international border or remain within their home state. They compare the ratio of refugees to IDPs based on ‘push’ factors, such as the type of violence (genocidal or not), and ‘pull’ factors in the potential destinations, such as stability and regime type. However, the number of people who are internally displaced or refugees is only one way to describe resettlement. It also differs qualitatively.
Lischer (2005), for instance, identifies types of refugee populations depending on their political and security situation: the persecuted, the state-in-exile, and the situational. She argues that states-in-exile are the most likely to initiate conflict, while the persecuted are most likely to suffer violence. Importantly, she links the causes of displacement to the potential consequences the refugees face in terms of violence. In more recent work, Lischer (2007) conceptually disaggregates displaced populations by seven types of wars or conflicts, and introduces a typology of ‘displacement crises’ that relate to four aspects of resettlement: location, demographics, political legitimacy, and humanitarian status. While both are descriptively rich, Lischer (2007) does not theorize the connections between them.
A common thread through the literature is the tendency to theorize, sometimes implicitly, from the civilians’ perspective. On average, more dire circumstances are likely to create displacement among a wider segment of the population. This is not illogical, but it is incomplete: civil wars involve strategic actors who are likely to have preferences over what civilians do. The push-pull model conceives of the violence itself as exogenous to the civilians or the community. However, recent work connects the characteristics of those exposed to violence, and the likelihood of the violence and the forms it takes (Balcells, 2017; Balcells & Steele, 2016). I argue that incorporating armed groups explicitly, to consider why they perpetrate violence that will cause some to flee, reveals important factors that shape civilians’ resettlement decisions.
Armed groups have been the focus of two related literatures on displacement and resettlement. Work on ethnic cleansing and resettlement as a counterinsurgency strategy consider the incentives and motives for armed groups to expel civilians from their communities. In general, civilians who are viewed as disloyal to the government, or uniquely different from the citizenry that the state aims to create, are vulnerable to such expulsion campaigns (Bulutgil, 2016; Mann, 2005). Whereas ethnic cleansing is an attempt to remove a population from a state’s territory, counterinsurgency resettlement is perceived as a strategy to reduce the contact that civilians have with rebel groups (Zhukov, 2014). Lichtenheld (2018) shows that what he terms ‘forced relocation’ occurs more often than commonly believed, including among states with relatively few resources. Even though ethnic cleansing and forced relocation are distinct processes, both literatures tend to study the question as a centralized strategy, especially in terms of why a government decides to pursue cleansing or resettlement (for examples of subnational variation, see Ron, 2003; Bulutgil, 2015; Hägerdal, 2019). A question that is more often left either implicit or unexplored, particularly in the case of ethnic cleansing, is where the expelled are likely to resettle when it is not determined by armed groups. In the next section, I propose a new way to characterize resettlement patterns.
A typology of resettlement patterns
One important aspect of refugee and IDP movements is the scale of the movements, or the number of individuals and households displaced. However, I argue that scale is not the most essential aspect of variation. Instead, I characterize patterns of resettlement during civil wars based on two dimensions: the extent to which the displaced cluster together in their new location, and whether or not they remain within their home state. The intersection of these dimensions yields a descriptive typology of resettlement patterns. 4 By clustering, I mean whether or not the displaced resettle with other displaced people. The most obvious example of clusters are camps for refugees and internally displaced persons (IDPs). Even without official camps, the displaced often create informal enclaves within their original state, or in a new one. Clustering together provides several advantages: it allows the displaced to tap existing social networks for mutual support in new settings, and as I explain below, it can also provide a relative degree of safety. Even with these possible advantages, not all displaced seek out others when they resettle. Some strike out independently into new cities or rural communities. The displaced may resettle close to other IDPs or refugees, but it would be incidental rather than integral to their decision to relocate there.
The second dimension of variation for resettlement patterns is a common characteristic considered when discussing resettlement: whether the displaced remain within their home state or not (Lischer, 2007; Moore & Shellman, 2006; Weiner, 1996). However, this dimension alone is too limited: refugees and IDPs do not all relocate to the same areas abroad or within their home country. Combined with whether the displaced resettle in clusters or not, this captures important variation in civilian-led resettlement.
Table I outlines the four ideal-types of resettlement patterns. The four categories – expulsion, dispersion, segregation, and integration – are named from the perspective of the home country. 5 Following Arjona (2014: 1376), the typology is useful for three reasons. First, it identifies important conceptual variation that minimizes variation within type but maximizes it across types. Second, it is parsimonious. 6 Third, it captures variation that is important to understand, and which may be important for other outcomes, as I elaborate on below. 7 I turn to describing each type, and I offer examples from a range of cases that illustrate the utility of the typology.
Resettlement patterns in war
Expulsion
When the displaced cluster together and relocate across a border of their home state, expulsion has occurred.
8
Refugee camps are good indicators of clustering, though clustering can occur outside of officially recognized camps. Often, the clustering of refugees precedes the formation of a camp, as Crisp & Jacobsen (1998: 28) note: [In] many mass influx situations, refugees and their leaders organize themselves into camp-like settlements before UNHCR or any other humanitarian organization has arrived on the scene and established an assistance program […] the refugees themselves […] congregate in large groups, forming large-scale settlements which eventually become institutionalized.
The Syrian civil war has produced the expulsion of millions who have resettled in camps in neighboring Jordan, Turkey, and Lebanon. The two Liberian civil wars spanning 1989–96 and 2000–03 were characterized by expulsion as well. Over half of the country’s population was uprooted by 1994, and 700,000 relocated to camps and communities in neighboring states (Nmoma, 1997). Camps were established in Sierra Leone, Ghana, and Nigeria, while in Ivory Coast and Guinea, refugees largely resettled with local populations (UNHCR, 1999: 112).
Dispersion
Dispersion is characterized by the displaced relocating independently of other displaced individuals and families in new countries. ‘Independently’ does not imply that the displaced cannot resettle in an area where other refugees live, but that such proximity would be a coincidence or convenience rather than a requirement for the displaced. They may seek assistance and shelter in a city or small community.
Following the intensification and spread of the violence in Nepal in 2001, Nepalis began to cross the border with India to a greater extent. However, no camps were established, and the refugees largely drew on their existing social networks and migratory patterns to cope with the displacement (Martinez, 2002: 13). This is an example of dispersion. Some Mozambican refugees also tapped personal networks to resettle throughout South Africa (Lubkemann, 2008). Malkki (1995) studies two different communities of Burundian refugees in Tanazania, one in a camp, and another living in a city. Onoma (2013) studies dispersed refugees from Sierra Leone and Liberia in Guinea.
I refer again to the UNHCR estimates, which suggest that nearly half of all refugees in sub-Saharan Africa between 2000 and 2010 are ‘self-settled’, the closest indicator of dispersed refugees (Fisk, 2014: 263). Through 2002, approximately two-thirds of the refugees in Asia were estimated to be dispersed (Schmidt, 2003).
The next two types of resettlement patterns occur within the state of origin, and refer to internally displaced people (IDPs).
Segregation
Segregation occurs when the displaced cluster together in cities and regions within the origin state, forming enclaves. An example of a clustering pattern within the origin state is found in Sri Lanka, where Tamils comprised the majority of the displaced population, but tended to remain in the Tamil regions of the north and east. Liberia also experienced segregation. The majority of those displaced – an estimated 1.3 million – lived in camps controlled by ECOMOG, the African regional economic union that intervened in the war (Nmoma, 1997). Around Monrovia, 20 formal IDP camps were established, which produced a de facto separation between the displaced and non-displaced.
During Colombia’s civil war in the mid-20th century known as La Violencia, liberal and conservative partisans from mixed communities separated, leaving many municipalities dominated by one party’s supporters (Karl, 2011). In Iraq, segregation of Sunnis and Shi’a across regions and within the city of Baghdad began in 2006 (Weidmann & Salehyan, 2013).
Extending these examples, we could further subdivide segregation into types based on which actor has authority over the IDPs: the state, a non-state armed group, a third party such as the UNHCR or other international actor, or areas where no centrally organized actor has authority. This further disaggregation may be important for the consequences of segregation.
Integration
Integration occurs when the displaced remain within their home country and resettle independently, such as IDPs relocating to urban areas within their home state.
During the first period of the Nepal civil war, between 1996 and 2001, integration was the predominant form of resettlement. Displaced individuals and families during this early period tended to arrive in district capitals and larger cities like Nepalganj and Kathmandu, areas under government control (Martinez, 2002: 12). Another example of integration comes from the Aceh Civil War (1999–2005). While some Achenese sought refuge in Malaysia, the majority remained within the district. The majority of those who did not return to Aceh resettled in the neighboring district of North Sumatra (IDMC, 2001: 88).
The Democratic Republic of the Congo (DRC) has one of the largest displaced populations in the world, and the majority of the displaced are scattered throughout the DRC rather than across borders. According to the Internal Displacement Monitoring Center (IDMC), ‘A vast majority of IDPs in the DRC live outside camps, with relatives, members of the same ethnic groups and church communities often providing support’ (IDMC, 2017). This example shows that ‘integration’ does not necessarily refer to integration into the state, but into communities. As with segregation, remaining within the borders of the origin state does not imply that the state has authority over the displaced.
These examples from a range of different wars indicate that variation in resettlement patterns exists across civil wars, but also across groups and over time within one war. Further, the typology is consistent with the way that the UNHCR classifies and estimates refugees (in camps, or self-settled), but has the virtue of incorporating an analogous distinction for IDPs as well. In this way, researchers can consider IDPs and refugees together without artificially limiting the comparison. But what explains the resettlement patterns? In the next section, I engage in theory-building to account for the resettlement patterns.
Explaining resettlement patterns: Forms of displacement and perpetrators
Where and when do expulsion, segregation, dispersion, and integration emerge? I argue that resettlement patterns are explained by two factors: (1) the form of displacement that civilians experience and (2) the perpetrator of the violence. The interaction of these factors shapes civilians’ decisions and leads to the resettlement patterns in the aggregate. In other words, the theory outlines a logic of ‘civilian-led resettlement’ (Lichtenheld, 2018), meaning that the argument applies to civilians’ decisions in the absence of a coercive actor imposing a destination on them. This simplification indicates a baseline set of expectations. The next section considers how additional actors, such as host state governments and international humanitarian organizations, change the expectations.
The theory applies to civil wars. 9 Within civil wars, I focus on the interaction between civilians and state and non-state armed groups. I assume that civilians prioritize their safety, and that as a result, on average, safety concerns trump material ones. Civilians’ decisions about their safety are informed by the behavior of armed groups and in some cases by other civilians. I assume that armed groups are strategic about at least some proportion of the violence they perpetrate, and that this strategic violence includes a mix of selective, collective, and indiscriminate targeting (described below). Finally, I assume that armed groups are mobile and can pursue the displaced in new locations if it is in the armed group’s interests, but cannot extend across international borders. 10
I argue that civilians’ choices about resettlement are related to the form of displacement they experience, and which actor perpetrated the violence that led to the displacement. I discuss each dimension in turn.
Forms of displacement and clustering
Drawing on previous work, I connect the form of displacement that civilians experience to the likelihood that they will cluster together with others who are displaced. Previously, I have argued that the type of violence civilians face from armed groups influences whether and how they are displaced during wars (Steele, 2017).
Forms of displacement depend on how armed groups target civilians. Three types of targeting characterize observed violence against civilians: selective, indiscriminate, and collective (Gutiérrez Sanín & Wood, 2017; Steele, 2009), and a specific form of displacement accompanies each: individual escape, mass evasion, and political cleansing, respectively (Steele, 2017). 11 While individual escape and mass evasion are unintentional outcomes from the perspective of armed groups, political cleansing is purposeful. Each type has different implications for the safest resettlement strategy.
Individual escape occurs when armed groups target individuals selectively for their behavior or suspected behavior, such as collaborating with a rival. In general, selective targeting falls into two categories: political and private. The former entails individuals and households targeted for failing to collaborate with an armed group, or for defecting to a rival. Private selective targeting is based on a denunciation for motives unrelated to supporting one armed group or another (Kalyvas, 2006). Selective targeting is associated with a form of displacement that I call individual escape. If individuals are alerted to the danger they face when selectively targeted, they can try to avoid it by leaving their communities. 12
Once they leave, civilians who faced selective targeting are most likely to avoid future violence by resettling in a place where they will be anonymous and difficult to find (Steele, 2009). As a result, they do not need to resettle with others, but can try to blend in independently.
Mass evasion stems from indiscriminate violence, which is not related to any trait or behavior, so anyone is potentially a victim. This type of targeting is the one most often associated with the idea of ‘collateral damage’. During bombing campaigns, civilians can move out of the area to avoid violence.
In cases of mass evasion, civilians may relocate nearby until the fighting or violence has ended, unless the violence destroys local infrastructure or shelter. People who experience mass evasion can improve their safety by moving to an area with a lower amount of indiscriminate violence. Victims of indiscriminate violence will not improve their safety by clustering with other displaced people, so they can resettle independently.
Political cleansing is the expulsion of a group of civilians based on a shared characteristic (Steele, 2017). 13 This means that anyone sharing the trait is a potential victim. (Ethnic cleansing is a form of political cleansing.)
When civilians experience political cleansing, they are not necessarily safe even when they leave their communities. Armed groups may continue to target them for some distance. Their ‘profile’ is usually discernible in distant communities, and even in non-ethnic settings. The timing and location of the cleansing can shape perceptions of displaced people’s loyalties or ‘guilt’, because armed groups politically cleanse territories to gain or consolidate territorial control within wars (Steele, 2018).
Because it is difficult to shed their identity or perceived guilt, victims of cleansing cannot escape the risk of violence only by moving. Instead, the safety of those collectively targeted is in part dependent on others who are similarly targeted. By resettling together, households will reduce the risk that they will suffer direct violence compared to the alternative of trying to resettle alone and risking identification. This incentive to cluster together to reduce the household’s risk has a perverse consequence: the group itself may be in greater danger because it is more visible and potentially threatening to armed actors (Steele, 2018).
Explaining patterns of resettlement
The next section explains whether the displaced are likely to remain in their country of origin as IDPs, or to cross a border and become refugees.
Perpetrators and destinations
The second factor that accounts for resettlement patterns is the perpetrator of the violence. The displaced not only take into consideration what type of violence they faced, but also which actor perpetrated the violence and their ability to inflict it in new locations. A safe destination will be one where the perpetrator does not have effective capacity or interest to pursue.
If civilians experience political cleansing by a state or state ally, they are likely to try to cross an international border for relative safety and to resettle together. If the state or state-allied group targets violence indiscriminately, the civilians who evade the violence are also likely to try to cross an international border. The reason is that the perpetrating state or allies should be more constrained in their ability to target violence there.
If civilians are targeted for political cleansing by a non-state armed group, they are likely to move closer to the state and remain within its borders. In contrast to when the state is the victimizer, the displaced may believe that the state will be able to offer some degree of protection. The same applies to civilians who experience indiscriminate or selective targeting by non-state actors: they are likely to resettle within the state’s borders. The argument is summarized in Table II, and I illustrate the argument with examples below.
Expulsion
When state armed forces or their allies perpetrate political cleansing, expulsion is the likely resettlement pattern. The resettlement of the Rohingya in Bangladesh, who have been ethnically cleansed by the Myanmar military, falls in this category. The majority of those targeted have sought refuge abroad. This expectation is also consistent with the finding that genocidal violence by the state is most likely to lead to refugee flows rather than internal displacement (Moore & Shellman, 2006); genocidal violence is a lethal form of collective targeting.
Against the expectations of the argument, Burundian rebels in the 1990s apparently sought to drive Hutus across the border and into camps in order to recruit them into their ranks (Lischer, 2007: 147). However, on average, I expect expulsion to be more associated with political cleansing by state or state allied forces.
Dispersion
I argue that dispersion, the other form of refugee resettlement, stems primarily from individual escape and mass evasion of state-perpetrated violence. Refugees will seek protection relatively independently, but far from the offending state in a new country. Of the ‘self-settled’ refugees estimated by the UNHCR, I hypothesize that the majority are fleeing state-perpetrated selective or indiscriminate targeting. One illustration is from the second period of the war in Nepal, when a number of those fleeing state violence crossed the border with India, relying on personal networks and existing migration routes (Martinez, 2002).
Segregation
As with refugees, IDPs can cluster or resettle independently. Political cleansing by non-state actors leads to segregation. In these cases, the displaced cluster together in regions within their home state, forming enclaves.
Worth (2013) describes a process of segregation in Syria, as Alawites were targeted by rebel groups in some areas and relocated to Damascus even though violence was still intense in the city. This pattern contrasts starkly with Syrian refugees who have fled political cleansing and indiscriminate bombing by the Assad regime.
Muslim Bosniaks targeted by Bosnian-Serb armed groups tended to move to particular enclaves within Bosnia. Compared to the resettlement in Rwanda, the pattern in Bosnia was segregation, even though there were many similarities in the ethnic profiling of targets.
Integration
When a non-state actor uses selective or indiscriminate violence, civilians are more likely to integrate by remaining within the country but to move to a new region, community or city for relative safety. In the early phase of the Nepalese Civil War, when civilians were targeted by Maoists, they relocated to cities, in contrast to the dispersion of the later phase of the war (Martinez, 2002).
Before turning to testable implications of the theory, I first consider how the expectations shift by relaxing core assumptions.
Additional factors: Assistance and policy
The theory assumes that external constraints on resettlement choices are inconsequential, but rather civilians choose where to go and whether or not to cluster. I also assumed that civilians prioritize security over material concerns. This section relaxes these assumptions to consider how additional factors can also shape aggregate resettlement patterns. In particular, I focus on two: the offer of assistance by state or international actors, and host state policies.
The displaced seek security, but they also need assistance: they have lost most of their assets, typically must care for children, and have suffered trauma (Moya, 2018). Assistance can be provided by the origin state itself, the receiving state, social and kin networks, strangers in new communities, or the international community. What influence does this have on resettlement? The international humanitarian regime has increased its foreign aid to refugees and IDPs substantially over the last four decades (Fearon, 2008). While this assistance was initially restricted to refugee camps, the UNHCR has increasingly offered aid to IDPs as well, as in the case of Liberia. In some cases, then, the presence of refugee camps or IDP camps could alter what the destination of the displaced would otherwise have been.
One example is Darfur. The vast majority of the displaced in Darfur – 90% by some estimates – moved into camps within Sudan, despite the political cleansing they suffered at the hands of state-allied militias. So why did the Darfuri remain in the country, rather than cross the border? One reason could be that the humanitarian community was able to provide assistance within Darfur, rather than across the border.
At the same time, it is not clear how much weight the displaced give to assistance. Again, in some cases, the placement of camps is guided by the choices of the displaced themselves (Crisp & Jacobsen, 1998). Returning to the case of Iraq, the presence of UNHCR camps in Jordan did not significantly affect the predominant form of resettlement that emerged in 2006, when IDPs remained within Baghdad or their home region.
A second additional factor that could affect resettlement is the policy of potential receiving states. Some states implement policies that redirect or block where refugees would resettle otherwise. Host states sometimes prefer to keep refugees in camps rather than allow them to resettle independently. In Kenya, in the 1990s refugees lived on the coast and in Mombasa, but eventually the government decided to force refugees into camps (Crisp, 2000). Motivations for state policy vary (Fisk, 2014: fn 63), but security is often a key justification, as was the case in Kenya (Jacobsen, 2002: 587). Another motive is often an attempt to avoid dispersion and long-term residence in the host state (Crisp, 2000: 617), a demand which may come from local communities (Jacobsen, 2002: 591).
If on average states that face large refugee populations favor refugee camps, then we should observe more expulsion compared to what we would expect from the baseline argument. State policies can also effectively group several clusters together, as in Kenya: Crisp (2000: 623, 629) points out that the largest camps in Kenya included inhabitants from ten countries and 20 ethnic groups.
In other words, as international assistance increases and state policies favor camps, expulsion should become relatively more frequent. At the same time, if assistance is offered more and more within the host state, segregation and integration are more likely (depending on the provider of the assistance). Incorporating this set of factors is a promising avenue to refine the expectations of the theory, especially over time.
Observable implications
The conceptual and theoretical frameworks advanced here can be tested empirically, forming the basis of a research agenda to advance our understanding of civilian resettlement. As a step in this direction, I distill some testable implications of the argument. One advantage of this framework is that even though it is based on micro-level logic, it leads to implications at the aggregate level, which should allow for testing across and within wars.
A first challenge is to create a measure of resettlement patterns. This would involve identifying clusters of displaced people, both within home states and abroad. One guide could be the UNHCR distinction between camp and self-settled refugees, and it could be extended to IDPs as well. In the absence of reliable estimates of camp sizes, one indicator could be the number of official camps in proportion to the estimated IDP population (which is collected by the Internal Displacement Monitoring Center on behalf of UNHCR). Careful case studies can validate such cross-country data: it is likely that counting formal camps will overlook informal enclaves and underestimate expulsion and segregation as a result, and that integration and dispersion will suffer from undercounting as well.
An implication of the theory is that wars that feature higher levels of political cleansing should produce higher levels of segregation and expulsion, relative to wars that have more indiscriminate and selective violence. For instance, if the war is an ethnic civil war, in which collective targeting based on ethnic (or sectarian) identity is common, then a high proportion of the population may be ensnared by political cleansing. As a result, segregation and expulsion should be relatively more frequent than in wars with no clear identity-based cleavage for collective targeting. Cross-cutting cleavages should make integration and dispersion more common. Process-tracing methods could then identify perpetrators of the cleansing and whether or not the expected destinations match the argument.
We could also draw on existing scholarship to reason about the likely relative frequency of resettlement patterns depending on warfare type. The forms of targeting and displacement are likely to transcend warfare type (Balcells & Steele, 2016). However, different types of warfare may generate different frequencies of targeting type and forms of displacement. Lichtenheld (2018) finds that cleansing is more common in conventional civil wars than irregular wars or symmetric non-conventional wars (SNCs) (Kalyvas & Balcells, 2010). Armed groups in SNCs, by contrast, may not have enough resources to hold territory so may resort more frequently to indiscriminate targeting. As a result, we can test the extent to which conventional civil wars are associated with segregation and expulsion, compared to irregular civil wars, which are more likely to produce a wider range of patterns. SNCs should be most associated with dispersion and integration.
A final implication can leverage changes over time within one war. When the predominant cleavage of a war shifts, we should expect a shift in the patterns of resettlement as well. An example is what occurred in Iraq in 2006: once the cleavage of the war coalesced around sectarian identities, the pattern of resettlement shifted to segregation – non-state armed groups targeted civilians, who sought to resettle with one another, remaining in Iraq. While testing these implications is beyond the scope of this article, it is a potentially fruitful direction for future research.
The potential implications of resettlement patterns for violence, conflict, and state-building
The resettlement patterns typology not only introduces a new dependent variable, but also offers a new way to think about the importance of wartime displacement for violence, conflict, and state-building through resettlement patterns. This exercise indicates that the typology could serve as a guide for case selection and scope conditions when theorizing the likely forms of conflict and violence associated with refugees and IDPs.
Violence and conflict
Resettlement patterns may be linked to the likelihood of violence associated with refugees and IDPs in both the receiving and origin states. A rich literature addresses violence against refugees and IDPs (e.g. Bohnet, Cottier & Hug, 2018; Fisk, 2018; Lischer, 2005, 2008; Savun & Gineste, 2019; Terry, 2002; Weidmann & Salehyan, 2013). One of the earliest frameworks to link the characteristics of the displaced with the risk for violence was developed by Lischer (2005). While Lischer (2005) studied vulnerability among camps, Fisk (2018) finds that large clusters of self-settled refugees are more likely to be attacked by armed groups than those living in camps across sub-Saharan Africa. Such clusters might also reflect expulsion even though they were not living in formal camps.
Onoma (2013) finds that among dispersed refugees, those who resettle in ‘closed’ communities, in which they have to seek permission from local elites, are more protected from the influence of state leaders who occasionally try to mobilize citizens to target refugees for violence. Counter-intuitively, refugees seeking safety in open communities that do not require submission to local authorities are more vulnerable when violence does break out. Onoma’s careful work reveals that while dispersed refugees might not suffer violence at the hands of the original perpetrators, they face different risks depending on the communities they join.
Among IDPs, segregation and integration have been linked to different patterns of violence and conflict as well. More recently, the influence of segregation on violence has been studied in Northern Ireland and Iraq (Balcells, Daniels & Escribà-Folch, 2016; Lischer, 2007; Weidmann & Salehyan, 2013). The redistribution of large numbers of a subset of the population is likely to have implications for the territorial reach of competing armed groups. Clustering itself can also endanger refugees or IDPs (Steele, 2018). However, there may also be a level at which violence associated with segregation reaches a plateau, as Balcells, Daniels & Escribà-Folch (2016) found with Northern Ireland, and Weidmann & Salehyan (2013) in Iraq.
In terms of the likelihood of conflict at the local level, the form of resettlement might also play an important role in shaping the type of conflict we should expect. For example, processes of segregation effectively ‘sort’ groups into different regions of the country, which could be accompanied by the occupation of abandoned property on both sides. Following La Violencia in Colombia, part of the reconciliation measures involved formalizing these swaps (Karl, 2017). This could be a specific challenge in some postwar settings that experience segregation (compared to land reform, for instance). In contrast, IDPs that tend to integrate are more likely to create dislocations in local labor markets and potentially to generate resentment among local populations, though to what extent this is likely to occur and which sectors it will affect should vary depending on the setting (Calderón & Ibáñez, 2015). A similar logic may apply, again, to dispersed refugees: their presence may stimulate local economies, but hurt some sectors in the process (Jacobsen, 2002; Maystadt & Verwimp, 2009).
State-building
The implications of resettlement patterns on state-building are relevant for both origin and host states. From the home country’s perspective, history shows that expulsion was often part and parcel of state formation processes. Population transfers were the ‘internationally legitimate means to overcome the discrepancy’ between diverse populations and the goal of homogeneous nations (Haddad, 2008: 120). Zolberg (1983: 28) observes that categories of people become targeted most often during the shift from empires to nation-states: ‘[m]inorities had existed before; but they had now been turned into political misfits’. 15 Arendt argues that in addition to ‘birthright’, ‘loyalties’ figured prominently in the calculation of who could remain and who was expelled by state authorities during the interwar period (Arendt, 1948: 278). Though expulsion is no longer a legitimate means to state-build, it still occurs within the context of civil wars.
Expulsion and dispersion also have implications for the state-building and nation-building potential of host countries. Jacobsen (2002: 578, 589) points out that refugees have led to the formation of new bureaucracies in several receiving states, as well as the deployment of the military. At the same time, the opportunities for state-building depend on settlement type: where there are camps, international assistance will be available to the host country, whereas if the population is dispersed, refugees will be in a better position to contribute to the host country’s economy (Jacobsen, 2002: 593).
Long-term assimilation in the host state may also relate to expulsion and dispersion. Malkki (1995) shows that dispersed refugees were much less likely to share an ethnic, nationalist identity with their home state. This is one way that resettlement connects to the literature on nation-building: Mylonas (2012) finds that ‘non-core groups’ are likely to be assimilated, accommodated, or expelled depending on the foreign policy goals of the host government (revisionist or not), and the relationship it has with the external patron of the non-core group. Refugee resettlement patterns could influence the relationship the host state has with the sending state, and what policies the host state adopts towards the refugees over the long term.
By 2016, refugees accounted for only one-third of the overall population of the displaced, while two-thirds remained within their origin state. Whether those IDPs segregate or integrate is likely to be consequential for state-building. In cases of segregation, it may be important which actor is in control of the territory where the displaced resettle. If the displaced resettle on the state’s periphery, segregation can lead to the creation of a radicalized periphery. In Colombia following the displacement of Liberals during La Violencia, several armed groups emerged alongside the new colonization of these regions. The incipient FARC was one: it helped organize and regulate the new communities. Over time, these areas became strongholds of the insurgency (Steele, 2017). Peripheral areas of resettlement can be more difficult to govern in the future because of armed or unarmed resistance to government encroachment. This is one example of how resettlement upends possibilities for post-conflict reconstruction – an area that is still understudied (Salehyan, 2007: 137). At the other end of the spectrum, Toft (2005) argues that segregated groups are more likely to engage in separatist conflict.
Segregation between territories controlled by the state and insurgents can reinforce polarization through physical segregation. In the case of Iraq, segregation even led to calls for partition of the country (Biden & Gelb, 2006). Over the long term, the distribution of the population can form the basis of long-term cleavages that shape political order, for instance through political party formation.
Integration is a possible mechanism for contemporary state-building, especially in states where cities become an attractive destination. In some cases, integration itself might drive urbanization as more and more displaced try to make a place for themselves in cities. While the strain of demands on urban growth can be difficult for a developing country to meet, it can also spur administrative innovation. Further, the ‘human geography’ can also allow a state to monitor and respond to its citizens more effectively (Kocher, 2002).
Conclusion
This article has presented a new characterization of civilian resettlement in civil wars. Patterns of resettlement diverge depending on whether or not the displaced cluster together, and whether or not they cross international borders. The combination yields four ideal-types: expulsion, dispersion, segregation, and integration.
The article also proposed a theory to account for variation in patterns within and across wars that links civilians’ calculations and armed groups’ behavior during civil wars. Resettlement patterns, I argue, result from the form of displacement civilians experience, and if the actor responsible for the displacement is a state or its ally, or a rebel group. Though the theory is not tested here, the article lays the groundwork to test the implications of the argument. The first step will be to validate the typology descriptively, then assess its explanatory power.
Finally, the article also points to the ways that resettlement patterns can influence ongoing violence, conflict, and state-building. Contemporary wars have led to the displacement of more than 60 million people – more than at any other time in history. This article indicates how wartime migrations, in turn, influence states, through different patterns of resettlement.
Footnotes
Acknowledgments
I am grateful to the special issue editors Alex Braithwaite, Idean Salehyan, and Burcu Savun, as well as Graeme Davies, Isabelle Duysteyven, Kerstin Fisk, Dominika Koter, Theresa Leimpek, Adam Lichtenheld, Juan Masullo, Daniel McDowell, Quinn Mulroy, Tom Ogorzalek, Jacob Shapiro, and Seiki Tanaka for helpful feedback, and the participants at the Households in Conflict Network Annual Workshop in Barcelona, 2011, the Kobe-Essex IR workshop in Brussels, 2014, the Political Economy and Transnational Governance Seminar at the University of Amsterdam, 2015, the Dutch Political Science Association workshop on Political Violence, Leiden, 2017, and the ISA workshop on Refugees, Forced Migration and Conflict, 2018. Nicole Martinez-Moore, Nina Pouls, Benoît Siberdt, and Luis Villafañe Jorge provided outstanding research assistance. All errors are my own.
Funding
Funding for this research was provided by the Stimuleering Europees Onderzoek of the Netherlands Organization for Scientific Research (NWO) via the Political Economy and Transnational Governance group at the University of Amsterdam, and by the US Army Research Office through the Minerva Initiative under grant number W911NF1810089.
