Policy Research Working Paper 10102 What it Takes to Return UN Peacekeeping and the Safe Return of Displaced People Vincenzo Bove Jessica Di Salvatore Leandro Elia Social Sustainability and Inclusion Global Practice June 2022 Policy Research Working Paper 10102 Abstract Can the international community enable conditions for 2018. To mitigate concerns about non-random subnational voluntary, safe and sustainable return of displaced people? assignment of peacekeepers, the paper exploits variations As conflict is key in the decision to leave and to return, in the presence of previous infrastructures and information this paper investigates whether the deployment of UN on the total supply of troops to African countries from peacekeeping operations can reduce the insecurities driv- each troop-contributing country. The paper finds that UN ing displacement and delaying return. It explores the case peacekeeping affects both the magnitude and the quality of of South Sudan, which hosts the second largest UN peace return. Displaced people are more likely to return home if operation in the world. It combines information on peace- peacekeepers are deployed in their county of destination. At keepers’ subnational deployment with data on individuals’ the same time, the local presence of peacekeepers mitigates intention to move and host communities’ perceptions of host communities’ negative perception of IDPs; they also returnees and internally displaced people (IDPs) using two enable the delivery of support to communities that seem surveys, one carried out between 2015 and 2017 and one in to improve attitudes toward returnees and IDPs. This paper is a product of the Social Sustainability and Inclusion Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at V.Bove@warwick.ac.uk. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team What it takes to Return: UN Peacekeeping and the safe return of displaced people∗ Vincenzo Bove† Jessica Di Salvatore‡ Leandro Elia§ University of Warwick University of Warwick Marche Polytechnic University Keywords: Internally displaced people; Returnees; UN Peacekeeping; Conflict; Percep- tions of returnees; Social cohesion. JEL classification: D74; H56; I31; O15. ∗ Social Sustainability and Inclusion Global Practice; Preventing social conflict and promoting social cohesion in forced displacement contexts; Susan Wong and Audrey Sacks. † Corresponding Author. Address: Department of Politics and International Studies, University of Warwick, UK; Email: v.bove@warwick.ac.uk. This paper was commissioned by the World Bank Social Sustainability and Inclusion Global Practice as part of the activity “Preventing Social Conflict and Promoting Social Cohesion in Forced Displacement Contexts.” The activity is task managed by Audrey Sacks and Susan Wong with assistance from Stephen Winkler. The program is funded by UK aid from the United Kingdom’s Foreign, Commonwealth and Development Office (FCDO), it is managed by the World Bank Group (WBG) and was established in partnership with the United Nations High Commissioner for Refugees (UNHCR). The scope of the program is to expand the global knowledge on forced displacement by funding quality research and disseminating results for the use of practitioners and policy makers. This work does not necessarily reflect the views of FCDO, the WBG or UNHCR. The authors thank the Folke Bernadotte Academy for supporting financially the data collection. ‡ Address: Department of Politics and International Studies, University of Warwick, UK; Email: jessica.di- salvatore@warwick.ac.uk § Address: Department of Economics and Social Sciences, Marche Polytechnic University, Italy; Email: l.elia@univpm.it 2 1 Introduction In recent years, natural disasters, environmental degradation and conflict have forced mil- lions of people to flee their homes. By one estimate, 9.9 million people were newly displaced due to conflict and violence worldwide in 2020, whereas the total number of people living in conflict-driven internal displacement reached a staggering 48 million.1 This displacement is oftentimes protracted as conflict may last for decades and conditions for return may take time to materialize. As a consequence, the rate of returns does not keep up with the rate of forced displacement (UNHCR, 2018), and this mismatch has lasting consequences on economic development and political stability of war-torn countries. Against this backdrop, International Organizations such as the United Nations High Commissioner for Refugees (UN- HCR), the International Organization for Migration (IOM) and the World Bank Group (WBG) as well as numerous national and international non-governmental organizations have long played a crucial role in providing humanitarian assistance to refugees and internally displaced people (IDPs), in supporting local public services under pressure, and in searching for practical solu- tions to forced migration problems. As conflict is one of the key factor in the decision to leave and to return (Czaika and Kis-Katos, 2009; Adhikari, 2013), one would expect the provision of a stable and secure environment to help address forced displacement in conflict-ridden regions. We explore whether the deployment of UN peacekeeping operations - by reducing insecurity and mitigating the very violence that drives displacement and delays return in the first place - can enable the safe return of refugees and the resettlement of IDPs.2 Despite some tragic failures a chronic lack of resources, UN peacekeeping can be an effective tool to stop atrocities, even when deployed in the most difficult and trying contexts (e.g., Hultman, Kathman and Shannon, 2013; Di Salvatore and Ruggeri, 2017; Ruggeri, Dorussen and Gizelis, 2017; Bove and Ruggeri, 2019; Bove, Ruffa and Ruggeri, 2020). Peace operations in recent years saw a dramatic expansion in the scope of their mandate, including protecting refugees and IDPs from violence and supporting their return. In addition to providing security, often peace missions undertake or encourage violence-reduction interventions and engagement and resilience-building activities in local communities; they also support local governments in reintegration efforts (van der Lijn, 2019). Perhaps more importantly, the local presence of peacekeepers also provide a safe environment for aid delivery. In this respect, as the arrival of IDPs and returnees brings economic and societal pressure on host communities, we might expect UN peacekeeping not only to encourage the returns, but also to directly or indirectly mitigate the perceived costs of new arrivals for communities. We investigate the complexity of return dynamics from two angles. First, we explore whether and how the security umbrella provided by UN missions affects the magnitude of returns, that is individuals’ decision to return to their home communities. Second, we study how peacekeeping shapes the quality of return, that is attitudes toward displacement and return from the perspective of host communities. In particular, we assess whether peacekeeping can mitigate negative perceptions of people on the move during displacement (i.e. towards IDPs) and after displacement (i.e. towards returnees).3 1 See IDMC database: https://www.internal-displacement.org/database/displacement-data. 2 We define UN peacekeeping operations as military interventions into potential, ongoing, or recently ended violence conflicts by the United Nations. So as not to conflate UN peacekeeping missions with activities carried out by NGOs or other International Organizations, we refer to all activities associated with the UN peacekeeping mission as ‘peacekeeping’. We refer interchangeably to ‘peacekeepers’ or ‘Blue Helmets’ as armed personnel in UN missions. 3 Note that while our paper explores how peacekeeping shapes the prevalence and quality of return, return per se is only one of the potential durable, development-oriented solutions for forcibly displaced persons and 3 We address these issues by analyzing the “hard case” of South Sudan, a country which is currently experiencing one of the direst humanitarian crises involving about 4.3 millions of currently displaced people.4 South Sudan is also hosting the UN Peacekeeping Mission in South Sudan (UNMISS), the second largest UN mission after DRC, with more than 16,000 uniformed personnel operating with a mandate to protect civilians and promote safe, informed, voluntary and dignified return of displaced households. The large number of displaced people and the size of the UN deployment make South Sudan a suitable test-bed to examine our hypotheses on the impact of peacekeeping on the safe return of displaced people. At the same time, as we elaborate, lingering instability also makes South Sudan a hard test to study return dynamics. In the context of ongoing violence, like in South Sudan, we expect peacekeepers to restore and maintain a minimum level of security needed to shape individuals’ intention to return and support communities’ capacity to receive them. Delivering a relatively secure environment in one of the most politically unstable regions is thus a necessary condition for returns. This security umbrella has a direct impact on local populations, but also indirect benefits as it encourages NGOs and development agencies to direct aid and assistance to the host countries. In fact, often peacekeeping missions start at the same time as development assistance programs. Finally, improved security conditions translate into a stimulus to local economic activities, particularly when the presence of the mission improves subjective perceptions of safety among households (Caruso et al., 2017; Bove, Di Salvatore and Elia, 2021). In sum, these improvements may facilitate returns as peacekeepers remove insecurity triggers and restore normalcy. Displaced households are also concerned about the social conditions of their returns and communities’ willingness to accept them, or in other words what we call the quality of return. We expect the same violence-mitigating effect of peacekeepers, which reduces concerns over safety and encourages economic recovery, to improve also the quality of return. Most impor- tantly for communities, in this regard is how missions’ security umbrella provides a conducive environment for aid delivery in areas where humanitarian workers are often denied access. In a context of ongoing conflict and large-scale displacement such as South Sudan, communities may perceive that their limited resources will have to be shared not only with returnees but also with IDPs that may be resettling. As such, alleviating concerns over limited resources should help mitigate tensions between host and displaced/return communities in aid-receiving areas. If anything, activities carried out by development and humanitarian actors should mitigate negative perceptions of and relations with returnees and IDPs in host communities rather than exacerbating tensions. To test these expectations, we use information on UNMISS peacekeepers’ deployment loca- tions, as documented in the UN Secretary General reports on the mission. We then combine this information with two key household-level data sources. First, we rely on the South Sudan High Frequency Survey (HFS), a representative survey carried out in four waves (2015-2017) in South Sudan by the World Bank across counties with and without UNMISS contingents. The survey includes information on returns and perceptions of IDPs’ impact on economic and security conditions. Second, we leverage community-level survey conducted by the IOM in the two neighboring towns of Bentiu and Rubkona. Both towns have benefitted from the presence of a UN contingent, allowing us to explore the extent to which peacekeepers enable the work of other humanitarian actors and NGOs that further shape relations with returnees and IDPs. Being so close to each other, the survey allows to compare reasonably similar contexts affected by displacement and conflict, and how communities have responded to the arrival of both returnees and IDPs. Unfortunately neither survey sample IDPs and returnees specifically their hosts. 4 https://www.unrefugees.org/emergencies/south-sudan/ 4 as both are based on a representative sample of South Sudanese population (HFS) and the two surveyed towns (CPS). However, while our data does not provide information from the perspective of displaced and returnees, communities’ attitudes toward them likely shape the quality of return and, to some extent, may still affect displacement decisions. To mitigate concerns on the non-random subnational assignment of peacekeepers we use three complementary strategies. First, we test for parallel trend in violent events between exposed and unexposed counties before the UN deployment to exclude the possibility that non- random placement of troops may shape the peacekeepers-returnee relationship. Second, we use an instrumental variable approach that leverages plausibly exogenous variations in the presence of previous infrastructures built in early 2000s to host a previous mission (UNMIS). Third, we exploit information on troop-contributing countries (TCCs) in UNMISS to report instrumental variable (IV) estimates, where peacekeeping size is instrumented using a ‘shift-share’ instrument commonly adopted in the migration literature (Altonji and Card, 1989). Two basic results emerge. First, we find that the local presence of UN peacekeeping troops affect the magnitude of returns as individuals are more likely to move and return to their homes if peacekeepers are deployed in their county of destination. Second, UNMISS presence improves the quality of return as host communities are less likely to believe that IDPs have a negative impact on job opportunities and on the overall security in their neighborhood if peacekeepers are deployed in their county. In addition, the community-level survey reveals that additional support provided to communities under UNMISS’ security umbrella may further mitigate sources of tension. This is particularly true for returnees and IDPs that have settled in the host community; no intervention scrutinized, however, seems to affect perceptions of IDPs living in protection of civilians (PoC) sites. Our paper complements several other papers that are commissioned as part of this project. By investigating whether UN peacekeeping increases the return of IDPs, we complement the recent studies of Parry and Aymerich (2021), Kaplan (2021) and Walk, Garimella and Chris- tia (2021) on what factors shape resettlement decisions. While the authors focus on peace agreements, information dissemination and “irrational” belief systems, our study focuses on the provision of security and relief as a key factor shaping returns and resettlement dynamics. Furthermore, by studying how peacekeeping can mitigate negative perceptions of and relations with returnees and IDPs in host communities, our findings speak to the studies of Ruiz and Vargas-Silva (2021), Murard (2021), Albarosa and Elsner (2021) and Groeger, Le´ on-Ciliotta and Stillman (2021) on the impact of returnees on social cohesion, hostility and discrimination in receiving communities. En route, as important tools available to address negative attitudes towards returnees are the provision of cash transfers (as discussed by Ag¨ uero and Fasola, 2021) and vocational training programs (as discussed by Ferguson et al., 2021), we explore whether non-economic factors can also be effective at reducing discrimination and promoting more pos- itive immigration attitudes in refugee-hosting setting. More generally, while we know that peacekeeping saves lives and supports peace (see Di Salvatore and Ruggeri, 2017, for a recent review), we have far more limited knowledge about its implications for population movements and the welfare of those benefiting from the security that Blue Helmets should bring about.5 We contribute to the above studies by providing some evidence that UN peacekeepers may sup- port safe and voluntary return by improving perceptions of returnees and IDPs in displacement 5 A notable exception is Sundberg (2020) who provides the first quantitative evidence of how peacekeeping affects forced displacement. He shows that whereas UN troop size is not associated with forced displacement in South Sudan, there is some evidence that UN peacekeepers acts as a “pull factor” for IDPs. Whereas he uses data on events of mass displacement, we focus on individual attitudes towards migration. At the same time, we rely on a different research design. 5 settings, and they may reduce risks of tensions between resettled and host communities. The remainder of paper proceeds as follows. Section 2 provides background information on the historical context in South Sudan, the conflict and displacement dynamics as well as the UN response to the crisis. Section 3 discusses the channels through which the deployment of peacekeepers can affect IDPs’ intentions to return and their movements as well as individuals’ perception of returnees. We then move to the description of the data and the empirical strategy in Section 4. Here we also address concerns over endogeneity, particularly selection bias. Results are reported in Section 5, and Section 6 concludes. 2 Context In this section, we highlight some key features of both the South Sudanese case and the UN response to the conflict. We also provide an overview of the displacement context in the country. To understand the current instabilities, we need to look all the way back to the Sudan’s colonial past. Between 1899 to 1956 the UK and Egypt ruled present-day Sudan and South Sudan through a dual colonial government known as the Anglo-Egyptian Condominium. To facilitate the control of the region, the British authorities segregated the animist or Christian Africans, living predominately in the south, from the Muslim Arabs living overwhelmingly in the north. In fact, the British administration sought to preserve social hierarchy throughout Sudan between the rich and high status Northern Muslims and Egyptian nationalists of the modernized North and the black Africans, the sudani (see e.g., Sharkey, 2008). For example, British authorities placed northern riverine peoples in positions of power and authority, specifically the Shaigiyya, Jailiyyin, and Dongola groups. This social hierarchy contributed to creating distrust, fear, and conflict between among Sudanese peoples. A North-Side divide was also enhanced by the decision to allow the spread of Christianity through the southern region, while maintaining Islam in the northern regions to preserve their independence and identity and also for fear of galvanizing religious resentments. Subsequently, in recognition of the ethnic and cultural differences, in 1930 the British colo- nial authorities introduced the so-called “Southern Policy”, through which the three southern provinces were administered as a different entity from the North. In particular, the policy excluded the Arabs from the administration, trade, and settlement in the southern part of the country, and created a protectorate in the south. Although scholars are divided on the legacy of the “Southern Policy”, many argue that it contributed to the rift between North and South Sudan and the pervasive tensions between the Arabs and the African ethnic groups (see, e.g., Mayo, 1994). Since its independence in 1956, Sudan became the theatre of two civil wars. The First Sudanese Civil War, (1955-1972), was a conflict between the northern part of Sudan and the southern Sudan region that demanded representation and more regional autonomy. The main belligerents in the war were the central government of Sudan and the Southern Sudan Liberation Movement (SSLM). The 1972 Addis Ababa Agreement failed to address lingering tensions and grievances, which resulted in the Second Sudanese Civil War, (1983-2005). The war saw the rebels from South Sudan forming the Southern Peoples Liberation Army (SPLA), led by John Garang, to fight the central government in Khartoum. In July 2002, the Gov- ernment of Sudan and the SPLM reached an agreement but talks continued into the following years until 2005, when the Government of Sudan and the SPLA signed the Comprehensive Peace Agreement. South Sudan gained independence from Sudan in July 2011, six years after the 2005 Comprehensive Peace Agreement ended 22 years of guerrilla warfare that had killed at least 1.5 million people. 6 The country today counts more than sixty cultural and linguistic groups. A decentralization of power was prescribed in the Interim Constitution of 2005 with the aim to creating shared sovereignty and allows many differentiated groups to participate in decision-making. In Figure A.1 (Appendix), we report summary statistics of selected variables to show the large within country-variation in the characteristics of South Sudanese people. 2.1 UNMISS and the South Sudanese Civil War Following the birth of the Republic of South Sudan, the newest country in the world, the UN Security Council established in July 2011 through resolution 1996 the United Nations Mission in the Republic of South Sudan (UNMISS) for an initial period of four years. Up to 7,000 military personnel and 900 civilian police personnel were initially authorized by the council to be rapidly deployed in the country. The mission had the mandate to consolidate peace and security and create security conditions conducive to timely and unimpeded humanitarian assistance. According to the UNHCR data (UNHCR, accessed 2021), in 2011 South Sudan was already hosting more than half a million internally displaced people, though numbers steadily declined in the following two years. Violence erupted in December 2013, following a political struggle between President Salva Kiir and opposition leader and former Vice President Riek Machar over the leadership of the Sudan People’s Liberation Movement (SPLM). The political tensions between Kiir and Machar led to the creation of the Sudan People’s Liberation Movement in Opposition (SPLM-IO), a political party and rebel group that split from SPLM. The conflict reopened unresolved political and ethnic grievances. As President Kiir accused his former deputy Machar of attempting a coup, violence quickly diffused and spread to several states, including Central Equatoria, Jonglei, Lakes, Unity and Upper Nile. After the civil war erupted in 2013, Uganda’s forces also crossed into South Sudan to assist with evacuations of their citizens, but they also remained to help President Salva Kiir secure the capital Juba.6 As the situation deteriorated, violent clashes between armed groups, inter-communal vio- lence, widespread insecurity and lack of services drove million people from their home. Within just the first four weeks of the 2013 crisis, almost 500,000 persons were displaced within South Sudan and around 74,300 people crossed into neighboring countries (UNHCR, accessed 2021). By the end of 2020, the Internal Displacement Monitoring Centre (IDMC) reported that more than 1 million and 400 thousands had been internally displaced by armed conflict.7 The con- flict was characterized by a marked ethnic dimension, as mainly Dinka members of government forces were loyal to President Kiir, whereas Nuer army defectors and their allied militias were loyal to ex-Vice-President Machar. As a result, numerous Dinka, Nuer and Shilluk civilians have been targeted and displaced on the basis of their ethnicity and (assumed) political af- filiations.8 For example, government forces and the Padang Dinka military and political elite pushed the Shilluk population from the east bank of the White Nile to maintain control of the area. And several large towns have been depopulated of their traditional ethnic communities, whereas some are being repopulated by members of the dominant Dinka ethnic group (see e.g., Sullivan, 2018). Against the backdrop of widespread violence and humanitarian crisis, and faced with UN- MISS limited capacity to cope with the deterioration of security, in December 2013 the Security 6 https://www.bbc.co.uk/news/world-africa-25759650; https://www.reuters.com/article/us-sou thsudan-security-idUSKCN0ZU13P 7 Figures available online at https://www.internal-displacement.org/countries/south-sudan 8 https://www.amnesty.org/en/latest/news/2014/05/south-sudan-civilians-killed-and-raped-v iolence-spirals-and-famine-looms/ 7 50 14000 12000 40 Peacekeepers Countries 10000 30 8000 20 6000 2011 2012 2013 2014 2015 2016 2017 2018 # of countries # of peacekeepers Figure 1: Total number of troops and total contributing countries in UNMISS, 2011-2018. Dashed vertical line indicates conflict outbreak in 2013 . Council by its resolution 2132 increased the overall troop and police strength of the mission through the rapid deployment of about 6,000 security forces, in addition to 7,600 peacekeepers already in the country. This raised the troop level to 12,500 personnel and the police component to 1,323 personnel. Figure 1 shows the total number of UN peacekeepers deployed in South Sudan, as part of UNMISS and the number of countries contributing soldiers to the mission. Between 2011 and 2015, the total number of UN troops more than doubled, which improved the capacity of the mission to contain violence. Alongside with an increase in the total number of “boots on the ground” - i.e., soldiers from national armies - also the number of donor countries have gradually increased from 20 in 2011 to 50 in 2015. In addition to increasing the size of the mission, in May 2014, the Security Council, through resolution 2155, also re-prioritized the mandate of UNMISS towards the protection of civilians, human rights monitoring and the support for the delivery of humanitarian assistance for the implementation of the Cessation of Hostilities Agreement. Figure 2 shows the tasks that were mandated before the eruption of violence in December 2013 and the subsequent changes to the mandate.9 Since the outbreak of conflict, armed groups have targeted civilians along ethnic lines and destroyed and looted villages, quickly precipitating the country into one of the direst ongoing humanitarian crisis (UNOCHA, 2020). The displacement context that characterizes the ongoing humanitarian crisis in South Sudan is complex, as both violence and food insecurity have contributed to rising levels of displacement. As of April 2021, more than 7 millions South Sudanese are severely affected by food insecurity. Among the 1.4 million that continued to be displaced in the country, some are currently hosted in IDPs camps protected by UNMISS’ protection-of-civilians umbrella. Hundreds of thousands of civilians sought refuge at UN bases in South Sudan. These sites came to be referred to as Protection of Civilians (PoC) sites, guarded by UNMISS military operating under a robust PoC mandate. The establishment of the 9 Based on the Peacekeeping Mandate Dataset (Di Salvatore et al., 2020). Only few tasks remain relevant to the mission post-December 2013; most importantly for our argument and analysis, it is noteworthy that the revised mandate become more concerned with humanitarian relief (particularly the protection of humanitarian actors) and the protection and support for refugees and IDPs. 8 Figure 2: UNMISS mandate before and after December 2013. protection sites has been central to the UNMISS protection strategy. It should be noted, though, that we do not aim to assess whether PoC sites have shaped displacement in South Sudan; rather, we are interested in whether the peacekeeping mission and its security and economic effects (which do not hinge on PoC sites as scope condition) have shaped displacement and returns. We return to the PoC sites in the conclusions, when reflecting on policy implications and UNMISS’ decision to hand over some protection camps to the government. Consistent with UNMISS’ mandate to assist humanitarian relief, several local and interna- tional humanitarian organizations work in South Sudan, including the International Committee of the Red Cross, the Norwegian Refugee Council, MSF, Midair, Nile Hope and Universal In- tervention Development Organization.10 Many of these organizations also support IDPs and returnees. As the conflict broke out in late 2013, and dozen humanitarians lost their lives in the line of duty,many international humanitarian actors, ranging from UN agencies to NGOs, sought to live and work inside the PoC sites to be close to the war-affected populations they serve, and to protect their own staff from violent attacks (Sutton, 2018). In 2013, the UN Humanitarian Country Team in South Sudan approved new guidelines for the constructive co- ordination between humanitarian actors and UNMISS, and to allow them to operate effectively within the same environment.The latter is an important point for this paper as it highlights that peacekeepers can improve security and allow the work of humanitarian actors, which in turn improves the lives of displaced people and enable their return. In Figure 3 we report a map of South Sudan in 2013 and 2017, where for each county we show whether Blue Helmets were deployed (black dots) and the number of returnees, using data from the IOM. The figure is helpful to illustrate the spatial distribution of returnee populations in relation to UN peacekeeping deployment, though the visual inspection does not reveal a clear pattern. For example, focusing on the most violent counties of the Jonglei State (Center-East) 10 The number of South Sudanese NGOs registered as members of the South Sudan NGO Forum was 263 in 2019 (Moro et al., 2020) 9 Figure 3: Returns and PKO presence in 2013 and 2017 based on IOM data. and Upper Nile (North-East), locations with UNMISS presence seem to have recorded more returns in 2017 (darker blue shades), though data for several counties is not available. Other locations in the West/South-West states (Western Bahr el Ghazal or Western Equatoria) also appear to have more returns recorded in 2017 particularly in areas that hosted UNMISS at some point. The maps also illustrate that UNMISS deployment did not significantly change over time, and was scattered throughout South Sudan’s territory. In relation to this, the High-Frequency Survey carried out by the World Bank (discussed in section 4.1) provides important fine-grained information on the experience of displacement and in particular on what would be needed to enable returns. Figure 4 shows the four most common items that respondents indicate as important to help their return. The left panel shows the frequencies for respondents that had been displaced during the 2013 conflict, while the right panel focuses on the population that returned home. The single most important factor mentioned by respondents is security, followed by other conditions related to access to services and food availability. Notably, economic opportunities are mentioned by returnees but are never mentioned by displaced people. This evidence has significant policy implications when we consider which are, among these, the factors that we can reasonably expect peacekeepers to contribute to, hence supporting sustainable and voluntary returns. 3 Theoretical Motivation In this section, we overview the main channels through which the deployment of peace- keepers can affect the living conditions of individuals. Ultimately, we are interested in how these channels connect the deployment of peacekeepers to decisions to return and the quality of (re)settlement. The first channel is direct and goes through an improvements in the security environment. 10 Main Needs to Help Returns Internally Displaced by 2013 Conflict Moved to Return Home .5 .5 .4 .4 .3 .3 .2 .2 .1 .1 0 0 Security Education/Health Food Employment Security Education/Health Food Employment Figure 4: Respondents reporting on what they need to help them returning home. Conflict is a major driver of displacement, as populations flee from death threats, indiscriminate violence, kidnapping, and forced recruitment, the so-called “journey to safety” (Lozano-Gracia et al., 2010; Balcells and Steele, 2016). Forced displacement is oftentimes observed in the wake of rising violence, particularly when households are directly targeted by armed groups or when they perceive a high risk of victimization in the near future. In other words, “people flee after being directly victimized by armed groups (reactive displacement) or to avoid being the victim of an attack (preventive displacement)” (Lozano-Gracia et al., 2010, p.160). Displacement, however, is also a strategy that belligerents may use to identify loyalties (Lichtenheld, 2020). When resettling elsewhere, civilians may reveal their preferences by selecting specific locations which may ultimately make them more susceptible to targeting (Steele, 2018). In light of this, we could expect peacekeepers to provide an additional layer of security in the eye of forcibly displaced people. Research shows that peace missions reduce violence and the risk of conflict re-occurrence by acting as security guarantors and mitigating the commitment problem among former warring parties (Walter, 1997, 2002; Fortna, 2008). This pacifying effect also exists at the local level and when civil wars are still active. Anecdotal evidence shows that there have been instances of UNMISS’ inability to fulfill its mandate to protect civilians. For example, when the July 2016 violence erupted in Juba, peacekeepers abandoned one of the protection sites (Burke, 2016). There were, however, also instances of peacekeepers opening the gates of the bases to let civilians in and monitoring their safety, in accordance to the mandate (CIVIC, 2016). Against these contrasting anecdotes, research shows that, on average, peacekeepers do improve security. More importantly for our micro-level focus on households’ displacement and return, peacekeepers do not only reduce conflict-related insecurities; they can also improve household’s perceptions of personal safety (Bove, Di Salvatore and Elia, 2021). This is particularly im- portant for displacement context where perceptions and fears of victimization are fundamental motivations for leaving, and in light of unavoidable reporting bias plaguing media-based conflict data (see Weidmann, 2016). On aggregate, this data may provide a general picture of trends of violence, which reduces concerns over measurement error but at the local level it may misrepre- sent security conditions. Indeed, peacekeepers may improve perceptions of safety among local populations by signaling presence and deterring armed actors via highly visible activities such as regular community policing and proactive patrolling. This is the reason why peacekeepers 11 can improve safety even if they are not necessarily heavily armed, as in the case of UN police officers (Hultman, Kathman and Shannon, 2013; Di Salvatore, 2019). Notably, the opposite dynamic is also plausible: peacekeepers may reduce civilian victimization by engaging with non-state armed actors, and such armed engagements may possibly heightening perceptions of insecurity among civilians. This view is consistent with a policy-based understanding of what peacekeepers can do in displacement contexts. Returnees, refugees and IDPs are often part of a broader group of “vul- nerable people” that peace operations are mandated to protect. Indeed, the key goals set for peace operations on matters of displacement concern facilitating humanitarian assistance for displaced populations and creating conditions for their voluntary, safe, dignified and sustain- able return (UNSC, 2006). While numerous UN peace operations since the 1990s have been mandated to support IDPs, refugees and returnees, the lack of an explicit authorization in the Security Council resolution has not prevented missions from supporting returns; in fact, “even when return is not written into a mission mandate, the civil affairs section of a UN operation is often involved in activities such as negotiating that returnees can move back into their oc- cupied houses” (van der Lijn, 2019, 12). Relatedly, the security umbrella UN missions provide is also qualitatively different from the security that local non-state actors and the government itself may provide for two main reasons. First, UN troops are locally deployed but missions act within a broader national intervention that involves national stakeholders. This means that UN missions, differently from local-level security providers, can ensure governments’ commitment to peace processes. Second, and relatedly, peacekeepers’ presence signals a relatively secure environment for international and national humanitarian actors. Their third-party presence actively provides a protection umbrella that enables humanitarian action and aid delivery that would otherwise require continued negotiations with local actors. For the same reason, UN pro- tection umbrella is not exclusionary; local armed actors and governments, on the other hand, may provide security and other public goods to their own constituencies, especially when a civil war is ongoing. In sum, the security provided by UN peacekeepers is diffused and unbiased, thus qualitatively different from security that warlords and national governments may provide in ongoing civil wars. Yet, the influence of conflict on outmigration is also indirect. Conflict leads to the disrup- tion of economic activities and the destruction of productive human capital, critical infrastruc- tures and transportation facilities, with the least-developed societies enduring the highest costs (Gates et al., 2012). As such, by causing economic deterioration, conflict also shapes other leading factors that in turn induce migration (Abel et al., 2019). Improvements in security resulting from a peace mission can help revitalize the economy of the host country and set the stage for reconstruction in two main ways. For one, by improving individuals’ safety, peacekeep- ing encourage them to engage in economic activities and enhance households’ living conditions. When security conditions improve, households have more incentives to resume the economic habits that violence altered. For example, households may have more access to markets or they feel safe enough to participate in the labor market and return to work when peacekeepers are deployed close-by (Bove, Di Salvatore and Elia, 2021).11 This is a necessary condition for a functioning economy as lack of safety depresses economies and reduces incentives for economic exchanges because of the unpredictability and uncertainty of future returns. Secondly, the eco- nomic benefit of hosting peacekeepers is not only linked to their boost to local economies, but also to their catalytic role for development support from the international community, includ- 11 This is not unique to the case of South Sudan. In Liberia, for example, Aning and Edu-Afful (2013) illustrate how local economies were reinvigorated by both the security linked to UNMIL presence and infrastructural improvements, particularly in areas hosting military bases. 12 ing international and national organizations. The presence of peace operations may encourage donors to support national governments in their management of migration flows. By doing so, UN missions indirectly alleviate the fiscal and financial burden for governments. Ultimately, peace missions may have both a micro-level “economic revival” effect that may explain the di- rection of subnational flows, but also a macro-level “economic relief” effect that could improve living conditions for displaced populations and host communities. Finally, as stressed above, the security umbrella that peacekeepers provide allow the working of humanitarian and devel- opment actors. It should be noted that the employment opportunities that missions generate, especially in the case of UNMISS, are often limited. However, if the humanitarian community is more likely to work in areas where peacekeepers provide security, we would expect them to also provide much more numerous employment opportunities. If peacekeepers not only address security concerns that hinder returns but also have positive externalities on local economic con- ditions, then we should expect peacekeeping to also affects individuals’ migration decisions.12 Overall, this discussion leads to the formulation of our first hypothesis: Hypothesis 1: Magnitude of Resettlement : (a) UN peacekeeping increases the flow of returnees in locations where they are deployed Having discussed the effect of peacekeeping on the size of resettlement, we also expect the operation to improve the quality of resettlement as such. Refugee population and IDPs can intensify existing problems associated with economic scarcity, particularly in weak or failing states (Homer-Dixon, 2010). Low-capacity states have limited or no ability to provide services to incoming displaced people. As people migrate to other areas to escape violence, they often clash with native populations over access to already scarce resources or the distribution of essential goods. The presence of displaced people may generate grievances and tensions with natives, especially when they see refugees and IDPs as a threat or feel that they are depriving ohmelt, Bove and the local population of what is rightfully theirs (Braithwaite et al., 2019; B¨ Gleditsch, 2019; Savun and Gineste, 2019) The presence of peacekeepers can help address the challenges associated with refugees and responding to their needs by creating a “wider peace dividend”. The arrival of both IDPs and returnees puts pressures on communities that already struggle to provide basic goods and generate local population’s grievance. In this context, the effective provision of humanitar- ian aid mitigates anger and anti-refugee hostility because of positive spillover effects on host communities (Lehmann and Masterson, 2020). In fact, aid mitigates the perceived economic costs of refugees by increasing refugees’ capacity to appease host communities through sharing aid or supporting demand for local goods and services. Furthermore, aid can increase positive contact with locals (Lehmann and Masterson, 2020). More context-sensitive reports have high- lighted the potential risks of humanitarian and development actors to have an unpredictable effect on communities in South Sudan. Access to aid and services have become sensitive is- sues if, for example, they have been perceived as unequal (e.g. reinforcing land claims of one particular group) or as incentivizing displaced people to remain in the area (Deng, 2020). In the last update in December 2020, the Conflict Sensitivity Resource Facility has reported that most communities have perceived a positive impact of aid from development and humanitar- ian agencies, though the reports raises important concerns over negative perceptions possibly linked to the amount of aid provided in Malakal and Wau counties (CSRF, 2021). 12 Note that the primary aim of our analysis is to capture the overall effect of peacekeeping on the safe return of internally displaced people. Plausibly, not all locations or individuals are affected by the same mechanisms, and, as a matter of fact, the same individual might be pushed by more than one driver. Unfortunately the data does not allow us to isolate the main mechanisms. 13 We posit that peacekeeping missions can reduce the pressure the arrivals may have on com- munities and local institutions by enabling the provision of public goods and services, including relief and aid provided by international and national agencies. For one, peacekeepers oftentimes support local authorities in the provision of public services. When conflict dissipated in Mundri, South Sudan, people started returning home. The community also hosted a peacekeeping base. However, one of the immediate effect of this sudden returns was water shortages, which pro- duced significant disruptions to the community’s schools and local economy. In response, the mission started working on a quick impact project to restore normal provisions (UN, 2002b). In addition to supporting local authorities, the security umbrella provided by the UN mission enables the provision of humanitarian aid including food and medicines. For example, Leer County Commissioner Wal Yach argued that while general security improved in the county after peacekeepers deployed, “with the presence of UNMISS, the presence humanitarians at Leer also increased” (UNMISS, 2015). In South Sudan, active conflict has often prevented their distribution in many areas of the country, and armed groups have looted aid compounds and health centers across the country. In this situation, the presence of peacekeepers is necessary for the safe and effective distribution of aid. In fact, UNMISS is tasked to “ensure adequate security for the protection sites in its bases and worked with humanitarian partners to provide sufficient assistance to displaced persons”.13 For example, in May 2017 peacekeepers were urgently deployed to Aburoc in the Upper Nile region to “provide humanitarian groups with the confidence they need to resume the provision of urgent assistance to tens of thousands of people”. One of the most urgent humanitarian need were the provision of safe drinking water to avoid the risk of an outbreak of diarrhea or cholera; in this context, peacekeepers secured the roads to provide safe passage for the delivery and collection of water (UNMISS, 2017). At the same time, UN missions often engage with local communities and foster reconciliation (Smidt, 2020), thus further reducing the tension that returnees may bring about. Peacekeepers also work with local authorities and community leaders to ensure support for resettlements. Consultations activities to ensure peaceful coexistence in expectations of more returns had been carried out in the DRC (UNMISS, 2020),and resolution mechanisms had been put in place by UNMSIL in Sierra Leone to address tension over returnees’ houses and lands (UN, 2002a). This mitigates individuals’ concerns about the impact of resettlement within their community. It is important to highlight that mitigating host communities’ concerns over displaced and returnees arrival says little about how the latter perceive the quality of the resettlement/return. Even if communities welcome them (back) and material support is provided, we should not assume IDPs and refugees will positively assess their well-being and welfare. Hence, we highlight that our conceptualization of the quality of return and resettlement entails how welcoming host communities are toward these people, regardless of how the latter experience it. This limitation is uniquely due to data availability.14 With this caveat in mind, our argument leads to the following hypotheses: Hypothesis 2: Quality of resettlement : (a) Members of host communities are less likely to perceive a negative impact of IDPs and returnees on security and economic opportunities in locations where peacekeepers are 13 https://unmiss.unmissions.org/background 14 The UNHCR has conducted an intention survey among South Sudanese refugees in neighboring countries. Similarly to the data we use in this study, this survey asks about intention and motivation to return, with no special attention to how refugees assess their current resettlement conditions or their concerns about how communities will receive their return. 14 deployed (b) In locations under PKO protection, members of host communities are less likely to per- ceive IDPs and returnees as detrimental to social cohesion where support is provided to communities 4 Research Design In this section we describe the data sources, the key variables, and present the empiri- cal approach taken in this paper to investigate and identify the impact of peacekeeping on migration. 4.1 Data Our analysis is based on the combination of fine-grained data on peacekeepers’ subnational deployment (presence and size) with two surveys: (i) three waves of the nationally representative World Bank High Frequency Survey (HFS) and (ii) the Community Perceptions Survey (CPS) carried out by the IOM in Bentiu and Rubkona communities. First, we use georeferenced data on the peacekeeping mission in South Sudan based on the RADPKO (Hunnicutt and Nomikos, 2020). The data estimates the presence and size of blue helmets based on the UN Secretary General reports on peacekeeping missions and the deployment maps included in these documents. We link these deployment information to South Sudan counties and use presence and logged size of contingents as our main treatment variables.15 Second, respondents in the High Frequency Survey carried out by the World Bank from 2015 to 2017 are also geolocated in South Sudanese counties within the 7 (out of 10) former states surveyed. Jonglei, Unity and Upper Nile states are not included in the survey due to lack of security conditions that made fieldwork unfeasible. It should be noted, though, that the HFS may well survey individuals that left from those states as long as they moved to one of the states that was surveyed. However, individuals that left or returned to their homes without moving outside the three states are not part of the survey. This is unfortunate considering the political relevance of states such as Unity in the post-2013 crisis in the country. Thus, we need to interpret our results as representative of the population surveyed in HFS and excluding the three most unstable regions of the country. In total, the HFS includes respondents from 46 counties in 7 states and is based on the 2008 census.16 We focus on household heads as respondents - in lieu of all household members - as some key questions we use in our analysis were only asked to them. Second, the CPS in Bentiu and Rubkona are included in the analysis of host communities’ attitudes toward (former) displaced to assess whether the provision of 15 An important concern about UNMISS provision of security is its geographical extent. In our analysis we claim that security umbrella extends on the whole county where UN personnel was deployed. We cannot precisely gauge the radius of UNMISS’ security umbrella but mission’s materials and documents provide important insights. UNMISS adopted a ‘hub and spoke’ model that made security provision more mobile and effective in remote areas. Temporary bases have been also used to respond swiftly to evolving security threats and protect civilians with minimum delay. At the same time, and partly because of the need to protect PoC sites, UNMISS’ resources for long-range patrols has been limited and the risk-averse posture of the mission has limited security spillovers across counties (see for example CIVIC (2016) and Rolandsen (2015)). However, short-range patrols have been regularly conducted by UNMISS. If anything, our results may be underestimating the actual security effect it had on counties hosting UN bases. 16 It should be noted that after 2015, the government re-organized the administrative division in South Sudan, which now includes 10 states and 180 counties. 15 food aid, non-food and shelter aid and NGOs’ work improves perceptions of displaced people in the eye of host communities. The CPS surveyed 546 respondents, 266 located in Bentiu town and 280 in Rubkona, selected as a representative sample based on the 2008 census, like the HFS survey. It is interesting to highlight that the two towns were selected as they have both experienced displacement and returns to similar extent, with a protection site being created between the two towns. Hence the goal of the survey was exactly to “better understand the potential for tension between the host community residing in the towns and returnees and IDPs [...]. Bentiu and Rubkona towns, as well as the PoC that was created in 2013, have borne the brunt of displacement during the ongoing crisis in Unity State”. The survey was also conducted to assess whether, as stakeholders in Juba expected, the presence of the UN and the provision of services and facilities to IDPs and returnees could lead to tensions within host communities in the area (Kang, 2019). 4.2 Outcome Variables 4.2.1 Motivation to Return (HFS; H1a) To estimate the impact of peacekeeping on population movements, we focus on reported motivations to return. More specifically, we code a dummy for respondents reporting that they moved to the county where they currently are “to be home”. To assess whether peacekeepers attract respondents not only if they want to be home but also for other reasons, we code dummy variables for respondents reporting economic opportunities, access to services and family reasons as alternative pull factors. Note that these motivations are mutually exclusive answers of the same question. 4.2.2 Perceptions of IDPs (HFS) and Returnees (CPS) (H2a, H2b) The HFS survey includes two key questions that allow us to measure attitudes toward IDPs’ impact on host communities. First, respondents are asked whether they agree that IDPs have had a negative impact on job opportunities. Second, they are also asked whether IDPs have made host communities less secure. We create two dummy variables that reflect negative assessments of IDPs’ impact on the economic and security environment respectively. Table A.1 in the Appendix provides summary statistics of all variables used in the analysis. In addition, we use the CPS survey to analyze perceptions of host communities toward returnees and IDPs that live either inside or outside the PoC site. From this survey, we code three variables of interest. First, we create an ordinal variable measuring how friendly is the relationship of the community with returnees according to the respondent. In particular, the binary variables equals one for respondents reporting “friendly” relations with returnees, zero for individuals reporting neutral relation and -1 for “not get along” responses. It is interesting to note that the same question is asked regarding IDPs living in the PoC site and outside of it; notably, while relations with IDPs range from “not getting along at all” to “very friendly”, relations with returnees are never assessed that negative.ly More commonly, these are assessed as “neutral”, hence suggesting that overall returnees are likely more welcome than IDPs. For this reason, the dependent variable measuring relations with returnees is binary (neutral or positive) rather than ordinal. As main independent variables, we code whether respondents indicate that three types of support are provided: food, shelter and NGOs/civil society support. 16 4.3 Empirical strategy and identification The baseline model specification is of the following form: yit = δpkc + βk xit + αk zct + fs + ft + εict The variable of interest is pkc , which is an indicator for the presence or size of the peacekeep- ing mission. As we use the first three waves of the HFS in our analysis, we measure UNMISS presence and size as a cross-section. When measuring presence,the variable takes value one if UN military forces operate in a county c between 2015 and 2016; it takes the logged number of troops in county c in the same time window when we focus on the deployment size. Both variables equal zero when no UN presence is coded in the data. As such, with δ we evaluate the effect of the presence (1) vs the absence (0) of peacekeepers. Models using size instead of presence are discussed in the results section, and fully reported in the Appendix, Section A.5. The vector xit contains an array of individual/household characteristics that we retrieve from the HFS and from the CPS. When using HFS, we control for whether the household moved after the 2013 conflict, their experience of violence, perceptions of safety, assessment of living conditions, employment status, distance from markets, gender (1 for women), education (3 dummies for primary, secondary and tertiary education), age groups (2 dummies for 15-64 and 64+), married status, rural households, ethnicity (dummy for Dinkas) and frequency of food availability. When using the CPS, we control for whether respondents are migrants themselves, their gender (1 for women) and age, whether they are the household head, whether they expect increasing arrivals of IDPs and returnees and the total size of the household. We also include other control variables related to availability of water, security, markets and livelihoods within the towns. Details on the coding of variables from the HFS and the CPS are in Appendix A.6. To remove county-level heterogeneity that might affect simultaneously the likelihood of UN- MISS deployment and the level of variables yit in a county, we add a vector zct of administrative- level characteristics. For the HFS which is at the county-level, we include pre-deployment deter- minants of peacekeeping using the share of land devoted to agriculture and pasture, population size, distance to the capital Juba, night lights intensity, all taken from the PRIO data (Tollefsen et al., 2016). All variables from PRIO are measured in 2010 - one year before deployment - and interacted with survey waves dummies to allow them to differ by wave. Additional unobserved state-level heterogeneity is controlled for with the inclusion of state fixed effects fs . The vector ft is a set of wave dummies that capture the effect of macro-shocks affecting all units in a given wave. For the CPS, we include respondents’ reported availability of markets, security, economic opportunities, infrastructures and local government’s support for IDPs and returnees. As the CPS is a one-off survey, we do not include unit and time fixed effects. Finally, εit is the disturbance term. We report robust standard errors clustered at the household level throughout the analysis using the HFS in order to control for arbitrary group- wise heteroskedasticity. Standard errors are clustered at the respondent level for the analysis based on the CPS, as it only involves respondents in two distinct towns. We estimate linear model specifications (OLS) for both analyses of the HFS and the CPS. 4.4 Threats to identification A fair concern is that peacekeepers’ deployment locations are not randomly chosen and depend on local conditions. However, as Mvukiyehe and Samii (2018) also point out, the information available to peacekeepers for making local basing decisions is often limited to coarse information from maps of conflict events, road infrastructure and terrain. As such, if we 17 400 PK deployment No Deployment 300 UNMIS Violent events (ACLED) end date --> 200 UNMISS start date -------------> 100 0 1997 1999 2001 2003 2005 2007 2009 2011 2013 2015 2017 Year Figure 5: Violent events in counties with/without prospective UNMISS deployment, 1997-2017 managed to control for these factors, we should reasonably be able to identify locations with similar odds of hosting a base. We build on previous literature on peacekeeping - in particular by looking at which characteristics are relevant determinants of peacekeeping deployment (see e.g., Ruggeri, Dorussen and Gizelis, 2018) - and then add these characteristics to the specification. In particular, for the HFS we use the share of land devoted to agriculture and pasture, population size, distance to capital Juba and night lights intensity. To further mitigate concerns about selection bias, we employ a number of complementary strategies for our analysis of the HFS survey. First, we check for the presence of parallel trends in violent events between exposed and unexposed counties before the UN mission deployment. Figure 5 shows the evolution of violent events across the two groups between 1997 and 2017 using conflict data from ACLED (Raleigh et al., 2010). Counties are considered exposed if they ever receive the treatment at some point (i.e. UNMISS deployment). A visual inspection of the time series reveals that the two groups exhibit similar violence trends, which assuages worries about potential correlation between persistent trends in violence and UNMISS deployment. Second, we explore whether deployment is driven by levels of violence using leads and lags of conflict events at the county level respectively for 5 years before and 4 years after UNMISS arrival (see Table A.2 in the Appendix). The long time-series allows us to exploit the time variation in PKO deployment to perform this check. Statistically insignificant coefficients on the lags would mitigate concerns on potential endogeneity of UNMISS deployment. All coefficient on lags are statistically insignificant at conventional levels and we do not detect any correlation between past violence and deployment. If anything, the presence of peacekeepers may be associated with more violence two and four years later. As such, the deployment of troops does not seem fully driven by previous violence in a county, rather by logistical considerations. Third, we use an instrumental variable approach that leverages plausibly exogenous varia- tions in the presence of previous infrastructures built in early 2000s to host a previous mission (UNMIS). The location of these infrastructures is reasonably independent from current con- ditions as UNMIS was concerned with a different set of conflict and actors and was deployed before the independence of South Sudan. The credibility of the identification depends on the fact that bases are not moved after they are set up initially. More importantly, UNMIS de- ployment locations likely followed a different logic from UNMISS ideal deployment locations as the two mission had a strikingly different mandates (as a consequence of dealing with different 18 conditions on the ground). The outbreak of civil war in 2013 and the consequent reconfigura- tion of UNMISS, made the difference between the two missions even more remarkable. As of 2021, the size of UNMISS is almost twice the size of UNMIS at its maximum deployment in the entire Sudanese territory. Furthermore, some of the UNMIS bases that UNMISS took over as County Support Bases had been put up to support the 2011 referendum, hence were not established in reaction to evolving conflict dynamics (da Costa and de Coning, 2013). UNMIS had originally planned to set up 70 referendum support bases, but only 24 were ready when the referendum took place in 2011 (UN News, 2010; UNMIS Press Conference, 2011). Among the completed bases, only some were taken over by UNMISS, and they did require some infras- tructural work to be used as military compounds (UNGA, 2012), exactly because they had not been originally designed as military bases. While we do not argue that were are able to rule out all possible violations of the exclusion restriction, the use of the instrument is supported by the specificity of UNMISS’ set up, its mandate, the reconfiguration post-2013 violence and the re-use of UNMIS infrastructure that was not originally intended for conflict reduction. More specifically, we use UNMIS infrastructures to instrument future presence on UNMISS in 2015- 2017. To instrument the size of UNMISS deployments, we propose a third strategy building on the migration literature and exploiting information on troop-contributing countries (TCCs) to construct a “shift-share” instrument as proposed by Altonji and Card (1989) (see Section 5.3). 5 Results In this section we present our main findings. For the sake of conciseness, in the following tables we only report the results for the main independent variables. The tables with all parameters are available in the Appendix, section A.4. 5.1 Does Peacekeeping Pull Returns? The first set of specifications we present in Tables 1 and 2 aims at explaining whether peacekeepers can shape returns. More specifically, these tables focus on respondents from the HFS that have moved from a county different from the one where they are currently located, and each column shows (i) a baseline model with state and wave fixed effects, (2) results with individual-level control variable, (iii) results with the inclusion of county-level variables and finally (iv) results with interacted state-wave fixed effects. We first show the results from OLS models in Table 1 and then address issues of self-selection in Table 2; in the latter, we present the estimates from the IV models that exploit plausibly exogenous variations in the presence of previous infrastructures. Quite consistently, all columns in Table 1 show that peacekeepers’ presence is associated with a higher likelihood that respondents moved there in order to return home. The coefficient is always positive and significant statistically, though the magnitude of the effect is not particularly large. On average, respondents are approximately 8 percentage points more likely to have returned home if peacekeepers are deployed in the county. If for purely illustrative purposes one interprets the OLS estimates as causal, then, this suggests that peacekeepers may have pulled individuals that were attempting to go back home. Reassuringly, the results are similar when we move to the IV models in Table 2 both in terms of size and statistical significance. As we can see the F-test statistics have values well above conventional levels that characterize weak instruments. Yet, the inclusion of interacted state- wave fixed effects returns marginally not statistically significant results for PKO presence. This is expected since the inclusion of state-wave fixed effects increases first-stage regression fitting, reduces the PKO dummy’s sample variation, in turn inflating variance of the IV parameters. 19 Table 1: PKO Presence and Probability to Return Home. OLS Estimates. Baseline Individual County-level State-Time FEs Controls Controls PKO presence 0.078∗∗ 0.084∗∗ 0.093∗∗ 0.078∗∗ (0.032) (0.035) (0.037) (0.037) Observations 1027 917 917 917 ∗p < 0.10, ∗ ∗ p < 0.05, ∗ ∗ ∗p < 0.01. Standard errors clustered by household (in parentheses). Regressions include wave and state FEs. Controls not shown. As mentioned in Section Data, our dataset does not include Unity state, which was a pivotal state for the escalation of conflict in 2013. The lack of information on households in Unity can pose one issue, that is, that results presented so far reflect decisions to return solely for the population surveyed in the HFS. This is true as long as households in Unity were less/more prone to return as compared with households in other counties. We cannot a priori exclude this possibility. As Bentiu hosted one of the largest PoC centre that supplied health, food and education to their residents as well as people from close towns, it is possible that refugees in search of security and services may have returned anyway home, and particularly in towns close to PoC site encouraged by the accessibility to services provided by the PoC site. Unfortunately, our data do not allow us to further explore the extent to which the exclusion of Unity can severely affect our estimates. That said, our results remain valid and representative of households included in the HFS. Table 2: PKO Presence and Probability to Return Home. 2SLS Estimates. Baseline Individual County-level State-Time FEs Controls Controls PKO presence 0.079∗ 0.102∗∗ 0.080∗ 0.070 (0.044) (0.049) (0.044) (0.044) First-stage regression UN previous infrastructure 0.700∗∗∗ 0.684∗∗∗ 0.878∗∗∗ 0.868∗∗∗ (0.0227) (0.0250) (0.0221) (0.0222) F-test of excl. instr. 951.31 748.14 1579.59 1533.23 Observations 1027 917 917 917 ∗p < 0.10, ∗ ∗ p < 0.05, ∗ ∗ ∗p < 0.01. Standard errors clustered by household (in parentheses). Regressions include wave and state FEs. Controls not shown. PKO presence is instrumented by the location of previous mission’s (UNMIS) infrastructure. To dig deeper into individuals’ motivation to move and how these may be affected by peacekeepers, we assess whether UNMISS presence may attract people on the move for reasons different than returning home. In Table 3 we leverage the alternative motivations that indi- viduals provide to explain why they moved to the location where they are being interviewed, namely economic opportunities, better access to services, family reasons and a residual cate- gory for ‘other’ motivations. Interestingly, peacekeepers’ presence does not explain population movements motivated by these alternative motivations. To clarify, notice that this finding does not imply that UN presence does not attract returnees because they do not provide economic relief. Rather, it shows that among the many reasons that pull individuals to specific locations, peacekeeping supports returns but does not affect decisions to move for, e.g., economic needs. 20 Table 3: PKO Presence and Reason to Move Here. 2SLS Estimates. Econ. Opp. Access Serv. Family Others PKO presence -0.012 -0.043 -0.014 0.031 (0.034) (0.027) (0.036) (0.040) First-stage regression UN previous infrastructure 0.890∗∗∗ (0.0203) F-test of excl. instr. 1913.18 Observations 1026 1026 1026 1026 ∗p < 0.10, ∗ ∗ p < 0.05, ∗ ∗ ∗p < 0.01. Standard errors clustered by household (in parentheses). Regressions include wave and state FEs. Controls not shown. PKO presence is instrumented by the location of previous mission’s (UNMIS) infrastructure. 5.2 Does Peacekeeping Mitigate concerns over Arrivals? In this section, we consider how individuals assess the impact of people-on-the-move on host communities in two moments: first, during displacement and, second, after displacement (i.e. once displaced people return). We expect peacekeeping operations to mitigate the possible tensions between host communities and returnees or displaced people that we have identified in section 3. Peacekeepers can do so by providing a larger peace dividend that locations with no deployment do not benefit from. In Table 4, we begin assessing whether respondents believe IDPs have a negative impact on job opportunities and on overall security in their neighborhood. The first key result is that peacekeeping per se does not affect how people assess the impact of IDPs on the labor market. However, to better grasp the dynamics at play, we investigate what happens in locations that host peacekeepers and have IDPs settling in by means of simple interaction terms. We find that IDPs’ presence in locations that do not host UNMISS basis tends to somewhat exacerbate individual’s negative assessments of IDPs’ economic impact, but peacekeepers attenuate these negative perceptions. Interestingly, these dynamics seem to also play out when respondents assess the negative security impact of IDPs presence, hence suggesting the mitigating effect of peacekeepers can also reduce the negative impact of displacement on security conditions.17 The corresponding IV specifications are shown in Table 5. The indicator for PKO presence and its interaction with the IDPs dummy variable are instrumented by previous UN infrastructures and its interaction with the IDPs dummy variable. As we can see, both instruments highly predict UNMISS presence and its interaction with IDPs indicator and the F-test statistics are always above the benchmark value of 10. Table 5 confirm the result for economic security, suggesting that individuals who are under the Blue Helmets’ umbrella and live close to IDPs are 17 to 21 percentage points less likely to report that IDPs have a negative impact on local labor market as compared to individuals living in areas without IDPs. However, the results for neighborhood security are not statistically significant, weakening the evidence of a positive impact of peacekeepers on locals’ security concerns regarding displacement. Estimates of Table 4 and 5 may be affected by endogenous settlement, that is, IDPs may relocate in areas where people of the same ethnicity live and that this may affect host communities’ perceptions, leading to more positive attitude towards IDPs. If so, the effect we associate to UNMISS is (partly) due to more affinity between host communities and IDPs. In terms of our specifications, this bias would result in less negative 17 Notably, this result is consistent with research showing how peacekeepers can improve subjective perceptions of safety among locals living in proximity of UN basis (Bove, Di Salvatore and Elia, 2021). 21 assessment of IDPs’ impact on the labor market and on security. We may expect that displaced people will move to PoC sites if they believe their ethnicity may result in hostile attitudes were they to settle within ethnically different communities; conversely, individuals from the same ethnic group of a given community would be less concerned about inter-ethnic tension and will resettle in informal settlements within such communities. If ethnic affinity is biasing our results, then this bias should be particularly evident when we remove PoC-hosting counties from our samples. More specifically, if the size of the coefficient of the interaction term between PKO and IDPs’ presence is larger in absolute value, this would entail that the bias due to ethnic affinity is substantive. Table 4: PKO Presence and Perceptions of IDPs. OLS estimates. Neg. Effect on Jobs Less Safe Neighborhood Baseline Interaction State-Time Baseline Interaction State-Time Model FEs Model FEs PKO presence 0.012 0.048 0.043 0.002 0.017 0.016 (0.029) (0.033) (0.033) (0.017) (0.018) (0.018) PKO presence × IDPs -0.160∗∗ -0.132∗ -0.102∗ -0.109∗∗ (0.071) (0.074) (0.053) (0.053) IDPs 0.098∗ 0.074 0.037 0.039 (0.057) (0.058) (0.032) (0.032) Observations 1852 1771 1771 5083 4837 4837 ∗p < 0.10, ∗ ∗ p < 0.05, ∗ ∗ ∗p < 0.01. Standard errors clustered by household (in parentheses). Regressions include wave and state FEs. Controls not shown. To this aim, we re-estimate specifications of Table 4 and 5 by excluding counties where PoC sites are installed. This allows us to consider only counties where the issue of endogenous IDPs settlement is likely more pervasive and check whether and to what extent the size of the interaction term PKO×IDPs varies. Among the 6 PoC sites present in South Sudan as of 2017, only two counties included in the HFS hosted one: Wau and Juba counties. Results of these additional specifications, reported in the Appendix A.3, show that when PoC sites are excluded from the sample the coefficient of the interaction term is, in fact, smaller (rather than larger) or virtually the same, indicating that the supposed bias seem not to be a sever concern for our estimates. Table 5: PKO Presence and Perceptions of IDPs. 2SLS estimates. Neg. Effect on Jobs Less Safe Neighborhood Baseline Interaction Model State-Time FEs Baseline Interaction Model State-Time FEs PKO presence 0.035 0.078∗∗ 0.081∗∗ -0.002 0.010 0.008 (0.034) (0.039) (0.039) (0.020) (0.021) (0.022) PKO presence×IDPs -0.175∗ -0.206∗∗ -0.064 -0.071 (0.096) (0.098) (0.085) (0.086) IDPs 0.107∗ 0.108∗ 0.023 0.025 (0.063) (0.064) (0.040) (0.040) First-stage regressions first-stage first-stage first-stage first-stage (i) (ii) (i) (ii) (i) (ii) (i) (ii) ∗∗∗ ∗∗∗ ∗∗ ∗∗∗ ∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ UN previous infr. 0.905 0.919 0.020 0.919 0.016 0.905 0.913 0.023 0.910 0.023∗∗∗ (0.015) (0.016) (0.007) (0.016) (0.006) (0.010) (0.010) (0.003) (0.010) (0.003) UN pr. infr.×IDPs -0.091∗∗ 0.732∗∗∗ -0.106∗∗ 0.765∗∗∗ -0.106∗∗∗ 0.606∗∗∗ -0.106∗∗∗ 0.606∗∗∗ (0.032) (0.034) (0.035) (0.040) (0.026) (0.039) (0.025) (0.039) F-test of excl. instr.(i) 3496.56 1744.83 1707.83 7532.89 4006.59 4046.33 F-test of excl. instr.(ii) 363.26 398.75 311.87 304.57 Observations 1852 1771 1771 5083 4837 4837 ∗p < 0.10, ∗ ∗ p < 0.05, ∗ ∗ ∗p < 0.01. Standard errors clustered by household (in parentheses). Regressions include wave and state FEs. Controls not shown. PKO presence and its interaction with IDPs presence are instrumented by the location of previous mission’s (UNMIS) infrastructure and its interaction with IDPs presence. 22 23 Table 4 and 5 focus on how communities perceive the arrival and settlement of displaced people, which may suggest how receptive they are to arrival of groups that may affect their access to services, their security and their economic opportunities. However, it is difficult to infer how returnees would be received. Similarly to IDPs, communities may face challenges in enabling the resettlement and reintegration of returnees, and may require the support of the government or international organizations. We expect peacekeepers’ presence to improve the quality of resettlement of returnees as well. Unfortunately, the HFS does not ask about returnees. We rely on the Community Perceptions Surveys conducted by the IOM in Bentiu and Rubkona in 2018 to shed more light on whether peacekeepers can enable other forms of support that improve conditions of returns and displacement. Table 6: Community Perceptions and Forms of Support. DV in column (i) is binary; DV in columns (ii) and (iii) is ordinal. Relations Returnees Relations IDPs Relations IDPs PoC Food Aid provided Negative 0.000 0.001 (0.002) (0.003) Neutral 0.003 0.004 (0.032) (0.012) Positive 0.072∗ -0.003 -0.005 (0.037) (0.034) (0.015) NGOs support Negative -0.005 0.002 (0.004) (0.002) Neutral -0.087∗∗ 0.007 (0.035) (0.007) Positive 0.007 0.092∗∗ -0.008 (0.026) (0.038) (0.009) Non-Food and Shelter provided Negative -0.004 0.001 (0.003) (0.003) Neutral -0.070∗ 0.003 (0.036) (0.010) Positive -0.007 0.074∗∗ -0.003 (0.040) (0.038) (0.013) Observations 538 538 533 ∗p < 0.10, ∗ ∗ p < 0.05, ∗ ∗ ∗p < 0.01. Marginal effects shown. Standard errors are robust to heteroskedasticity. Controls not shown. The results of the analysis are reported in Table 6, which shows that different interventions seem to affect perceptions of returnees and IDPs differently. First, relations with returnees are reported to be better for respondents that also report receiving food aid, though other interventions do not seem to affect perceptions at all. Relationships with IDPs living in the community are more likely to be assessed as positive in presence of NGOs and non-food support (e.g. shelter), and food does not appear as a concern associated with IDPs’ presence. It is surprising though that none of the above affects respondents’ assessment of relations with IDPs living in the PoC camps. The lack of significant results does not imply that relations with PoC 24 residents are necessarily more hostile; in fact, it could be that their impact on communities is mitigated by the fact that they get most of the support they need from the site itself. But it is interesting that communities may have different perceptions of displaced depending on whether they reside within the community or UN PoC sites. This suggests the need for more research on how communities perceive IDPs in PoC sites, and the extent to which such sites creates additional challenges for IDPs’ resettlement.18 5.3 Deployment Size In the main analysis we use the presence of peacekeepers, rather than the deployment size, for two reasons. First, our expectations are based on how the presence of peacekeepers per se, by affecting violence and perceptions of safety, can produce a safer environment that encourages return. Especially for returnees, one could expect that the decision to return home is more linked to whether Blue Helmets are deployed or not, rather than how many operate on the ground (which may be a difficult piece of information to get precisely in any case) The second reason concerns the fact that subnational data on peacekeeping relies on estimated numbers. Unfortunately, we do not have precise data from the UN on the number of peacekeepers sent to different locations in host countries. Hence, we prefer to use a dummy variable to reduce issues with measurement errors.19 For robustness, however, it is important to check whether our results remain overall consis- tent when we use deployment size instead of a deployment dummy. Even if some measurement errors exist, they should not completely offset our estimates and provide a much different pic- ture. The tables in Section A.5 (Appendix) replicate the analysis of the paper by replacing the peacekeeping dummy with the log number of peacekeepers deployed in a county. We calculate this as the average number of troops in each month defining a survey wave. To address the endogeneity problem of the location choice, we exploit information on troop- contributing countries (TCCs) in UNMISS to report instrumental variable (IV) estimates, where peacekeeping size is instrumented using a ‘shift-share’ instrument (Altonji and Card, 1989). This is constructed by interacting the initial distribution of peacekeepers across South Sudanese counties and the supply of troops from each TCC to all African missions.20 In this way, we rely on variation stemming from the interaction of the time-varying national growth rates in the supply of troops from each TCC and TCC contributions in 2014.This information is combined with information on the nationality of contingents from different TCCs in South Sudanese counties based on the GeoPKO datatset (Cil et al., 2020). These strategies will ensure identification along both the extensive margin of whether peacekeepers are present and the intensive margin of how many peacekeepers are deployed. We find only partial support for the results in the main analyses that uses a dummy vari- able to capture peacekeepers’ presence. For example, we find that the higher the number of peacekeepers in a county, the higher is the probability that an individual moved there in or- 18 As noted above, our analysis focuses on the mission overall, rather than the specific strategies it used to protect civilians. More importantly, as PoC sites presence does not equate UNMISS presence, this is neither a scope condition of our theory nor an implication of our empirical results. Hence, our findings do not speak to the question of whether PoC sites work or not. We return to this point later. 19 The two currently available data sources on UN peacekeeping are Cil et al. (2020) and Hunnicutt and Nomikos (2020). Both data projects use deployment maps provided by the UN to estimate the number of peacekeepers in a certain location. While both estimates are made on reasonable assumptions, this could make the precision of the inference more problematic. 20 We have used an alternative version of the shift-share instrument which considers the annual supply of troops from TCC to all world missions. Results (not shown) are similar. 25 der to return home. Perhaps not surprisingly, given issues of measurement errors, the results are only statistically significant using the IV strategy (see Table A11). In line with results in Table 3, we similarly find that peacekeeping presence does not affect decisions to move linked to economic or family concerns, with the exception of the residual ‘other’ category (see Table A12). Finally, and perhaps more importantly, we find evidence that an increase in the size of the deployment leads to improvements in local populations’ perceptions of IDPs particularly with respect to perceptions of IDPs impact on job opportunities (see Table A13). However, results become statistically insignificant in the IV specifications (see Table A14) primarily due to weak instruments. As such, the very presence of the Blue Helmets, rather than the specific size of the local contingent, appears to affect the quality of the return. 6 Policy and Program Implications We investigate whether UN peacekeeping operations, by improving perceptions of security, encourage the safe return of vulnerable populations in the location of deployment. In addition to the magnitude of return, we also expect the presence of peacekeepers to provide a conducive environment for aid delivery in areas where humanitarian workers are often denied access. This should mitigate tensions and perceptions of insecurity that these vulnerable populations may bring to host communities, thus improving the quality of return. We use South Sudan as a case study and show that the local presence of peacekeepers positively and significantly affects population flows and individuals’ intention to move. At the same time they also shape the quality of return by improving host communities’ perceptions of displaced people. Our analyses depart from existing research examining the security effect of peace operations based on conflict data, and our findings shed new light on the positive spillovers of UN deployment and how peacekeeping can contribute to creating the conditions for voluntary, safe and sustainable return. With its relatively long history of internal conflicts, and the presence of a robust UN mission, South Sudan provides a laboratory for understanding how peacekeeping can help establish a safe environment that mitigate insecurities driving displacement and delaying return. A main advantage of our study is the use of the World Bank high-frequency household-level data, which allow comparison of several units to mitigate the problem of unit heterogeneity. Yet, our research design and data present some caveats and limitations, and we hope that some important avenues for further research might emerge from these limitations. First, the HFS survey had to exclude respondents located in Upper Nile, Unity, and Jonglei, which needs to be considered when interepreting the results. These states also hosted UNMISS bases, and are the opposition stronghold where a large-scale displacement took place. As we note, there is therefore a risk that our finding on the effect of peacekeeping are biased although it is not straightforward to predict in which direction. Second, we lack specific survey questions on refugees, as opposed to IDPs, and future research will try to isolate the impact of peacekeeping on different population groups. Relatedly, we were only able to discuss the quality of return based on an indirect measure, i.e. reception from communities. While these can shape displaced people and returnees experiences, we also need to design surveys that focus on their assessment of the conditions of return and resettlement. Third, the geographical areas of return are limited to only two towns, Bentiu and Rubkona. In the future, the availability of high-frequency individual level dataset in several war-torn villages, whose collection requires logistical and financial costs, will allow more rigorous investigations across different locations. As with any case study, our estimates of the effect of peacekeeping in South Sudan may not necessarily hold in all other countries hosting peacekeepers, we note that the crisis in South Su- 26 dan is not unique and similarly high humanitarian needs are present in Somalia, Sudan (Darfur) and DRC, not to mention the most severe crises in Yemen and Syria (UNOCHA, 2020). While South Sudan may exhibit some typical features with regard to the conflict dynamics compared to some other African countries, it does represent a particularly challenging case when we con- sider the universe of cases where UN peacekeepers are deployed. The ongoing violence makes it even more difficult to detect a meaningful effect of the mission on the magnitude and quality of displacement. Hence, South Sudan is a hard test in terms of the severity of its humanitarian crisis, the number of people in need of help, and levels of violence it is experiencing while the UN attempts to achieve its protection mandate. In relation to the latter, the UN response to the crisis and the mandate of UNMISS is substantially in line with the last generation of peace missions that operate with a robust mandate, a focus on civilians’ protection and peacebuilding goals. This further highlight that it is not that UNMISS’ mandate is particularly ambitious; rather, conditions on the ground make its implementation more challenging. Our findings speak to the wider debate on how to address population displacement and hu- manitarian crises. This paper assesses in particular how to create conditions that enable returns and reduce social tensions in host communities through multilateral military interventions. By establishing the necessary security conditions conducive to the provision of humanitarian assis- tance and return of refugees and IDPs, UN peace operations may help facilitate the repatriation of refugees and return of IDPs. Supporting the resettlement and reintegration of the displaced - wherever they want to be - is one of the key objectives of the aid community. Our paper focuses on whether peace- keeping affects the prevalence and quality of return, where the return is one potential durable, development-oriented solution for forcibly displaced persons and their hosts. The quantitative estimates reported in this research provide evidence that UN peacekeeping can affect both the magnitude and the quality of return. This finding by no means suggests that peacekeeping alone is the answer to displacement. In fact, peacekeepers are not deployed to contain displacement; they are deployed as conflict-mitigation tools. Hence, our analysis suggests that when peace- keepers work to suppress violence, they also create more favorable conditions for displaced and returnees. Yet, UN missions should be an instrument that complements other actions under- taken by the international community, such as those of the UNHCR and development agencies like the World Bank, to address the issue of returns and resettlement. In fact, in the last few years, particularly in the wake of the European refugee crisis, the inter- national community has been trying to forge collective solutions to the displacement challenge. Recent examples are the 2018 Global Compact on Refugees, to strengthen global architecture for international cooperation, or the 2019 Global Refugee Forum, which generated a range of commitments to accelerate the realization of the Compact’s promise. The World Bank Group’s International Development Association (IDA) has also supported development opportunities for refugee and host communities through the Window for Host Communities and Refugees, whereas regional initiatives like the Intergovernmental Authority on Development (IGAD), have generated initiatives for the displacement situation South Sudan. And whereas our empirical expectation on the benefits of peacekeeping is itself based on a burgeoning literature that finds conflict reduction on a monthly basis, there is still limited academic evaluations of the lasting effects of interventions after missions close and peacekeepers exit. Our study contributes to our knowledge of what happens when peacekeepers are deployed, but we still know little about the lasting legacies of those operations, after mission closure.21 As the collective attention of the international community and the type and amount of resources available can shift away from 21 The 2020 forum on International Peacekeeping sets the stage for this ongoing research (see e.g., introduction article by Gledhill, 2020). 27 host states after missions close, we need to better understand whether internationally-backed peace operations leave legacies of sustainable peace. Related to this issue is the UNMISS decision to hand over some protection sites to the government in 2020, while the mission re-orients its focus in the South Sudanese hinterland (Harragin, 2020). The handover of PoC camps from UNAMID (Darfur) to the Sudanese gov- ernment has already illustrated the risks of such decisions, and the security challenges resulting from alleged improper uses of the camps and the presence of Sudanese soldiers inside the camps (Forti, 2019). In the case of South Sudan, reports indicate that people have often decided to remain the camps, for either security reasons or access to services not available otherwise (Har- ragin, 2020). Our analysis does not specifically hinge on protection sites as shaping factors for the magnitude and quality of return. These sites are not a long-term solution and rushing with- drawals - before stabilization has occurred and trust for the government is restored - presents numerous risks for both residents and humanitarian actors.22 The Revitalized Agreement signed by the government of South Sudan in 2018 did include a commitment to develop a national framework for returns and reintegration (launched in 2019), but as of 2021 humanitarian needs remain rampant (UNOCHA, 2021). Finally, as peace cannot be sustainable without development, humanitarian and develop- ment actors are pivotal in ensuring that long-term development needs are met and a relapse in conflict is prevented. Mitigating the devastating effects of conflicts is one of the core mis- sions of the UK DfID, the World Bank and the UNHCR. We believe that this objective can be accomplished when complementarities between donors are leveraged. Yet, while the work of humanitarian agencies to address displacement gives a temporary relief to the population and the state, a prolonged service delivery role can reduce the capacity of the recipient governments and undermines their legitimacy in the eyes of people through the creation of a parallel service delivery mechanism. Bypassing national and local institutions could undermine sustainable long term capacity, and weaken national or local ownership. Effective interventions should also take into due account the role of host states and direct efforts to strengthen core state capacity. References Abel, Guy J, Michael Brottrager, Jesus Crespo Cuaresma and Raya Muttarak. 2019. “Climate, conflict and forced migration.” Global Environmental Change 54:239–249. Adhikari, Prakash. 2013. “Conflict-induced displacement, understanding the causes of flight.” American Journal of Political Science 57(1):82–89. uero, Jorge M and Eniola Fasola. 2021. “Distributional Policies and Social Cohesion in Ag¨ a High-Unemployment Setting.” Unpublished Working paper. Commissioned as part of the “Preventing Social Conflict and Promoting Social Cohesion in Forced Displacement Con- texts” Series. Washington, DC: World Bank Group. Albarosa, Emanuele and Benjamin Elsner. 2021. “Forced Migration, Social Cohesion and Con- flict: The 2015 Refugee Inflow in Germany.” Unpublished Working paper. Commissioned as part of the “Preventing Social Conflict and Promoting Social Cohesion in Forced Displace- ment Contexts” Series. Washington, DC: World Bank Group. Altonji, Joseph G and David Card. 1989. The effects of immigration on the labor market outcomes of natives. Technical report National Bureau of Economic Research. 22 See Harragin (2020) for a discussion on possible actions for UNMISS. 28 Aning, Kwesi and Fiifi Edu-Afful. 2013. “Unintended impacts and the gendered consequences of peacekeeping economies in Liberia.” International Peacekeeping 20(1):17–32. Balcells, Laia and Abbey Steele. 2016. “Warfare, political identities, and displacement in Spain and Colombia.” Political Geography 51:15–29. ohmelt, Tobias, Vincenzo Bove and Kristian Skrede Gleditsch. 2019. “Blame the victims? B¨ Refugees, state capacity, and non-state actor violence.” Journal of Peace Research 56(1):73– 87. Bove, Vincenzo and Andrea Ruggeri. 2019. “Peacekeeping effectiveness and blue helmets? distance from locals.” Journal of Conflict Resolution 63(7):1630–1655. Bove, Vincenzo, Chiara Ruffa and Andrea Ruggeri. 2020. Composing Peace: Mission Compo- sition in UN Peacekeeping. Oxford University Press. Bove, Vincenzo, Jessica Di Salvatore and Leandro Elia. 2021. “UN peacekeeping and house- holds’ well-being in civil wars.” American Journal of Political Science . Braithwaite, Alex, Idean Salehyan, Burcu Savun et al. 2019. “Refugees, forced migration, and conflict: Introduction to the special issue.” Journal of Peace Research 56(1):5–11. Burke, Jason. 2016. “UN peacekeepers refused to help as aid workers were raped in South Sudan.” The Guardian . October 6. Caruso, Raul, Prabin Khadka, Ilaria Petrarca and Roberto Ricciuti. 2017. “The economic im- pact of peacekeeping. Evidence from South Sudan.” Defence and peace economics 28(2):250– 270. ee Nilsson. 2020. “Mapping blue helmets: Cil, Deniz, Hanne Fjelde, Lisa Hultman and Desir´ introducing the geocoded peacekeeping operations (Geo-PKO) dataset.” Journal of Peace Research 57(2):360–370. CIVIC. 2016. “Under Fire: The July 2016 Violence In Juba and UN Response.”. CSRF. 2021. “Conflict Sensitivity Resource Facility: Context Update.” Conflict Sensitivity Resource Facility October-December 2020:1–5. Czaika, Mathias and Krisztina Kis-Katos. 2009. “Civil conflict and displacement: Village-level determinants of forced migration in Aceh.” Journal of peace research 46(3):399–418. da Costa, Diana Felix and Cedric de Coning. 2013. UNMISS County Support Bases: Peacekeeping-Peacebuilding Nexus at Work? Technical report. Deng, David. 2020. “Conflict Sensitivity Analysis: Considerations for the Humanitarian Re- sponse in Mangalla.” Conflict Sensitivity Resource Facility October 2020:1–7. Di Salvatore, Jessica. 2019. “Peacekeepers against Criminal Violence - Unintended Effects of Peacekeeping Operations?” American Journal of Political Science 63(4):840–858. Di Salvatore, Jessica and Andrea Ruggeri. 2017. Effectiveness of peacekeeping operations. In Oxford Research Encyclopedia of Politics. 29 Di Salvatore, Jessica, Magnus Lundgren, Kseniya Oksamytna and Hannah M Smidt. 2020. “Introducing the peacekeeping mandates (PEMA) dataset.” Working Paper SSRN . Ferguson, Neil T. N, Rebecca J. Wolfe, Laila Amine, Eric Ramadi and Lina Shahin. 2021. “Building Stability Between Host and Refugee Communities: Evidence from a TVET Pro- gram in Jordan and Lebanon.” Unpublished Working paper. Commissioned as part of the “Preventing Social Conflict and Promoting Social Cohesion in Forced Displacement Con- texts” Series. Washington, DC: World Bank Group. Forti, Daniel. 2019. “Navigating Crisis and Opportunity: The Peacekeeping Transition in Darfur.” IPI Working Paper December 2019:1–40. Fortna, Virginia Page. 2008. Does peacekeeping work? Princeton University Press. avard Hegre, H˚ Gates, Scott, H˚ ard and H˚ avard Mokleiv Nyg˚ avard Strand. 2012. “Development consequences of armed conflict.” World Development 40(9):1713–1722. Gledhill, John. 2020. “The pieces kept after peace is kept: Assessing the (post-exit) legacies of peace operations.” International Peacekeeping 27(1):1–11. Groeger, Andre, Gianmarco Le´ on-Ciliotta and Steven Stillman. 2021. “Immigration, Labor Markets and Discrimination: Evidence from the Venezuelan Exodus in Per´u.” Unpublished Working paper. Commissioned as part of the “Preventing Social Conflict and Promoting Social Cohesion in Forced Displacement Contexts” Series. Washington, DC: World Bank Group. Harragin, Simon. 2020. “Back on their feet: The role of PoCs in South Sudan and the potential for returning ‘home’.” Conflict Sensitivity Resource Facility October 2020:1–22. Homer-Dixon, Thomas F. 2010. Environment, scarcity, and violence. Princeton University Press. Hultman, Lisa, Jacob Kathman and Megan Shannon. 2013. “United Nations peacekeeping and civilian protection in civil war.” American Journal of Political Science 57(4):875–891. Hunnicutt, Patrick and William G Nomikos. 2020. “Nationality, Gender, and Deployments at the Local Level: Introducing the RADPKO Dataset.” International Peacekeeping pp. 1–28. Kang, Tarnjeet K. 2019. Urbanization as a result of displacement - A case study of Bentiu, South Sudan. Technical report Internal Displacement Monitoring Centre. Kaplan, Oliver. 2021. “Superstitions and Civilian Displacement: Evidence from the Colombian Conflict.” Unpublished Working paper. Commissioned as part of the “Preventing Social Con- flict and Promoting Social Cohesion in Forced Displacement Contexts” Series. Washington, DC: World Bank Group. Lehmann, M Christian and Daniel TR Masterson. 2020. “Does aid reduce anti-refugee violence? evidence from Syrian refugees in lebanon.” American Political Science Review 114(4):1335– 1342. Lichtenheld, Adam G. 2020. “Explaining population displacement strategies in civil wars: a cross-national analysis.” International Organization 74(2):253–294. 30 Lozano-Gracia, Nancy, Gianfranco Piras, Ana Maria Ib´ an˜ez and Geoffrey JD Hewings. 2010. “The journey to safety: conflict-driven migration flows in Colombia.” International Regional Science Review 33(2):157–180. Mayo, David Nailo N. 1994. “The British southern policy in Sudan: an inquiry into the closed district ordinances (1914-1946).” Northeast African Studies 1(2/3):165–185. Moro, Leben, Naomi Pendle, Alice Robinson and Lydia Tanner. 2020. “Localising humanitar- ian aid during armed conflict Learning from the histories and creativity of South Sudanese NGOs.” London School of Economics . Murard, Elie. 2021. “Mass refugee inflow and social cohesion in the long-run Evidence from the Greek population resettlement.” Unpublished Working paper. Commissioned as part of the “Preventing Social Conflict and Promoting Social Cohesion in Forced Displacement Con- texts” Series. Washington, DC: World Bank Group. Mvukiyehe, Eric and Cyrus Samii. 2018. “Peacekeeping and development in fragile states: micro-level evidence from Liberia.” Policy Research Working Paper . Parry, Jacqueline and Olga Aymerich. 2021. “IDPs with perceived ISIL affiliation: Can local peace agreements facilitate their return?” Unpublished Working paper. Commissioned as part of the “Preventing Social Conflict and Promoting Social Cohesion in Forced Displacement Contexts” Series. Washington, DC: World Bank Group. avard Hegre and Joakim Karlsen. 2010. “Introducing Raleigh, Clionadh, Andrew Linke, H˚ ACLED: an armed conflict location and event dataset: special data feature.” Journal of peace research 47(5):651–660. Rolandsen, Øystein H. 2015. “Small and far between: peacekeeping economies in South Sudan.” Journal of Intervention and Statebuilding 9(3):353–371. Ruggeri, Andrea, Han Dorussen and Theodora-Ismene Gizelis. 2017. “Winning the peace locally: UN peacekeeping and local conflict.” International organization 71(1):163–185. Ruggeri, Andrea, Han Dorussen and Theodora-Ismene Gizelis. 2018. “On the frontline every day? Subnational deployment of United Nations peacekeepers.” British Journal of Political Science 48(4):1005–1025. Ruiz, Isabel and Carlos Vargas-Silva. 2021. “Refugee Return and Social Cohesion.” Unpublished Working paper. Commissioned as part of the “Preventing Social Conflict and Promoting Social Cohesion in Forced Displacement Contexts” Series. Washington, DC: World Bank Group. Savun, Burcu and Christian Gineste. 2019. “From protection to persecution: Threat environ- ment and refugee scapegoating.” Journal of Peace Research 56(1):88–102. Sharkey, Heather J. 2008. “Arab identity and ideology in Sudan: The politics of language, ethnicity, and race.” African Affairs 107(426):21–43. Smidt, Hannah M. 2020. “United Nations peacekeeping locally: Enabling conflict resolution, reducing communal violence.” Journal of Conflict Resolution 64(2-3):344–372. 31 Steele, Abbey. 2018. “IDP resettlement and collective targeting during civil wars: Evidence from Colombia.” Journal of Peace Research 55(6):810–824. Sullivan, Daniel P. 2018. Displaced Nation: The Dangerous Implications of Rushed Returns in South Sudan. Technical report Refugees International. Sundberg, Ralph. 2020. “UN Peacekeeping and Forced Displacement in South Sudan.” Inter- national Peacekeeping 27(2):210–237. Sutton, Rebecca. 2018. “Safeguarding “distinction” inside the wire: Humanitarian-peacekeeper interactions in South Sudan’s Protection of Civilian sites.”. URL: https://iow.eui.eu/2018/06/28/safeguarding-distinction-inside-wire-humanitarian- peacekeeper-interactions-south-sudans-protection-civilian-sites/ Tollefsen, Andreas Forø, Karim Bahgat, Jonas Nordkvelle and Halvard Buhaug. 2016. “PRIO- GRID v. 2.0 Codebook.” Journal of Peace Research 49(2):363–374. UN. 2002a. Handbook on United Nations Multidimensional Peacekeeping Operations. Technical report Peacekeeping Best Practices Unit, Department of Peacekeeping Operations. UN. 2002b. “UNMISS set to mitigate water shortages in Mundri caused by returnees attracted by peace.”. URL: https://peacekeeping.un.org/en/unmiss-set-to-mitigate-water-shortages-mundri- caused-returnees-attracted-peace UN News. 2010. UN to help deliver voter registration material for referendum on southern Sudan. Technical report. UNGA. 2012. A/67/610. Technical report United Nations General Assembly. UNHCR. 2018. “Forced Displacement in 2018.” United Nations High Commissioner for Refugees: Geneva, Switzerland . UNHCR. accessed 2021. UNHCR Statistical Online Population Database. Technical report UNHCR. UNMIS Press Conference. 2011. Near-verbatim transcript of the Press Conference by the UNMIS Regional Coordinator, David Gressly. Technical report. UNMISS. 2015. “Presence of UNMISS in Leer improves security, humanitarian assistance.”. URL: https://unmiss.unmissions.org/presence-unmiss-leer-improves-security- humanitarian-assistance UNMISS. 2017. “UNMISS DEPLOYS PEACEKEEPERS TO ABUROC TO ENABLE DELIVERY OF HUMANITARIAN AID.”. URL: https://unmiss.unmissions.org/unmiss-deploys-peacekeepers-aburoc-enable-delivery- humanitarian-aid UNMISS. 2020. “Pinga: Reassured by the presence of peacekeepers and the army, families have returned home.”. URL: https://peacekeeping.un.org/en/pinga-reassured-presence-of-peacekeepers-and-army- families-have-returned-home 32 UNOCHA. 2020. Global Humanitarian Overview 2020. Technical report UN Office for the Coordination of Humanitarian Affairs. URL: ht tp s: // ww w. un oc ha .o rg /s it es /u no ch a/ fi le s/ GH O-2020_ v9. 1. pd f UNOCHA. 2021. “South Sudan Humanitarian Needs Overview 2021.” January 2021. UNSC. 2006. Resolution 1674 (2006). Technical report United Nations Security Council. ır. 2019. “Multilateral Peace Operations and the Challenges of Irregular Mi- van der Lijn, Ja¨ gration and Human Trafficking.” SIPRI Background Paper . Walk, Erin, Kiran Garimella and Fotini Christia. 2021. “Displacement and Return in the Inter- net Era: How Social Media Captures Migration Decisions in Northern Syria.” Unpublished Working paper. Commissioned as part of the “Preventing Social Conflict and Promoting Social Cohesion in Forced Displacement Contexts” Series. Washington, DC: World Bank Group. Walter, Barbara F. 1997. “The critical barrier to civil war settlement.” International organi- zation 51(3):335–364. Walter, Barbara F. 2002. Committing to peace: The successful settlement of civil wars. Prince- ton University Press. Weidmann, Nils B. 2016. “A closer look at reporting bias in conflict event data.” American Journal of Political Science 60(1):206–218.