Policy Research Working Paper 10019 Social Cohesion, Economic Security, and Forced Displacement in the Long-Run Evidence from Rural Colombia Juan F. Tellez Laia Balcells Social Sustainability and Inclusion Global Practice April 2022 Policy Research Working Paper 10019 Abstract Millions of people around the world are internally displaced draws on a rich set of covariates and outcomes to provide and yet—compared to other forms of wartime victimiza- plausible estimates on the long-run effects of internal dis- tion—scholars know relatively little about the long-run placement, finding that a decade or more after displacement, consequences of displacement for victims. This gap in the victims experience substantial negative welfare effects yet literature is problematic since displacement is distinct from exhibit higher levels of social cohesion than their coun- other forms of victimization and because IDPs face unique terparts. Finally, combining a prediction framework with challenges in post-conflict transitions. This study contrib- interviews with key stakeholders and displacement victims, utes to the literature on the effects of displacement in three the study explores variation in outcomes among victims, ways. First, the study brings to bear a unique sample of particularly why some return home and seek reparations. households in Colombia that is largely homogeneous along The results reveal a wide assortment of consequences from key confounders–mostly poor, rural, and conflict-afflicted— displacement and should help inform policy-making bear- yet varies in their exposure to displacement. Next, the study ing on support for internally displaced people. 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 laia.balcells@georgetown.edu. 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 Social cohesion, economic security, and forced displacement in the long-run: Evidence from rural Colombia* Juan F. Tellez† Laia Balcells‡ JEL codes: O15, R23, D74, H56 Keywords: IDPs, social cohesion, civil war, displacement * 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. This work is part of the program “Building the Evidence on Protracted Forced Displacement: A Multi-Stakeholder Partnership”. 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. We would like to thank Ana Maria Ibañez for making data on the Fondo de Tierras available and one anonymous reviewer for their helpful comments. † Assistant Professor of Political Science, University of California, Davis. Contact: jftellez@ucdavis.edu ‡ Provost’s Distinguished Associate Professor of Government, Georgetown University. Contact: laia.balcells@georgetown.edu 1 INTRODUCTION The United Nations High Commissioner for Refugees estimates that more than 45 million peo- ple around the world are internally displaced, comprising roughly half a percent of the global population 1 (UNHCR, 2020). For many, displacement is a “life long sentence”, as IDPs can re- main in a state of protracted crisis for years and even decades (OCHA, 2017). The magnitude and duration of displacement raises questions about how the experience of displacement af- fects victims, particularly in the long-run. Surprisingly, while there is now a large literature on the consequences of wartime victimiza- tion for a broad range of social and political outcomes (Bauer et al., 2016), academic research has relatively less to say about displacement, especially in the long-term. This is in spite of the fact that displacement is a unique form of victimization: in being forced to relocate, victims are separated from their communities and social networks; they face violence and social exclu- sion in new host communities (Rozo and Vargas, 2021); and they are often separated from prop- erty, assets, and savings that they struggle to recover (Thapa and Hauff, 2005; Ibáñez and Moya, 2010). These characteristics also mean that displacement effects are especially likely to persist and evolve over time. Understanding how displacement affects victims – both in the short and long-term – is thus important not just for the academic value but also to inform the design of policy aimed at addressing one of the most pressing global humanitarian challenges. In this paper we bring new evidence to bear on the effects of internal displacement through an original survey of 1,500 rural households spread across fifty municipalities in Colombia. Colom- bia is a country where almost one in ten citizens have been internally displaced by armed con- flict (Internal Displacement Monitoring Centre, 2021) and where the success of a recent, fragile peace agreement may in part rest on the extent to which the plight of IDPs can be redressed.2 We combine the unique nature of the sample – households that are largely homogeneous along key confounders, such as poverty, rurality, and conflict exposure, yet vary substantially in their incidence of displacement – with a rich set of covariates and causal inference approaches to pro- vide plausible estimates of the effects of displacement on a varied set of victim outcomes3 . Then, 1 Internally displaced people (IDPs) remain within the boundaries of their country of origin. By contrast, refugees by definition cross international borders. 2 We would like to note that Colombia has also had significant number of refugees since 2015, particularly from Venezuela. In this paper, we focus only on Colombian IDPs. 3 The sample was not designed to be representative of all IDPs in Colombia, a limitation we discuss more fully in the appendix and research design section. That said, we would argue there is still great value in the sample, given that data collection on IDPs is relatively rare, particularly in the remote areas we cover. We would also argue, along the lines discussed in Aronow and Samii (2016), that a nationally-representative sample analyzed in a regression framework would not avoid the generalizability concerns associated with our sample. 2 we use a predictive framework, a set of modules on IDP life after displacement, and qualitative interviews with key stakeholders4 and victims of displacement to provide rare insight into two key post-displacement questions: 1) who returns home after they are displaced, and why?; 2) why do some victims seek out reparations for lost property from the state but not others? Our results suggest displacement leaves victims transformed in varied ways, even years af- ter the displacing event. Even within the largely poor and conflict-affected population we study, IDPs are substantially worse off in economic terms than their counterparts (Ibáñez and Moya, 2010). They are also, surprisingly, much more mobilized and invested in collective action, per- haps attesting to the need for self-advocacy in the face of state failures in post-conflict policy- making (Sanín, Huertas and Hernández, 2019). IDPs also appear more invested in the prospect of peace-building, even as they expect insecurity to remain constant (Tellez, 2019b). On the other hand, displacement does not seem to generate changes in attitudes towards the state or land security, which are generally low across the board. Finally, our analysis of who returns home and who seeks restitution highlights continuing insecurity (Prem et al., 2020) and low state capacity vis-a-vis post-conflict reparations (Sanín, Huertas and Hernández, 2019) as ma- jor obstacles to the resettlement of IDPs. 2 CONTEXT: THE INTERNAL DISPLACEMENT CRISIS IN COLOMBIA Colombia has one of the largest populations of internally displaced people in the world – approx- imately 5 million in a country of 50 million people (Internal Displacement Monitoring Centre, 2021). The displacement crisis also spans the country’s territory: civilians have been displaced in every department of Colombia. At the same time, IDPs are disproportionately concentrated in ethnic minorities (CODHES, 2016, p. 826) and households headed by women (CODHES, 2015). Despite an historic peace agreement in 2016 (Tellez, 2019a) that brought an end to large-scale confrontations between the Colombian government and the country’s largest rebel group, the FARC, forced displacement continues to plague the country, with an estimated 106,000 new cases in 2020 alone (OCHA, 2020). At the center of the displacement crisis in Colombia is a protracted, multi-party armed con- flict that has lasted several decades, involving varied insurgent groups, the Colombian state, and numerous paramilitary organizations. The dynamics of displacement in Colombia has been a function of civilians fleeing in order to avoid being caught in the cross-fire of competing groups, but also the result of groups deliberately ‘cleansing’ territory of civilians in areas controlled (or 4 The stakeholders we interviewed include representatives of NGOs and government agencies working in land restitution in Colombia, as well as victims of displacement. 3 suspected to be controlled) by another group (Steele, 2011; Balcells and Steele, 2016).5 Once dis- placed, IDPs often resettle in the outskirts of big cities such as Cali, Medellín or Bogotá, where they face physical and economic insecurity (Ibáñez and Moya, 2010; Steele, 2017). To be properly understood, the current displacement crisis must be set against a broader context of weak tenure security, widespread informality, and a long history of violent land expro- priation in the country. The Colombian countryside has been characterized by cycles of forced expulsion and resettlement of peasants, which have fueled land conflicts long before the con- temporary civil war (LeGrand, 1986). Land ownership is highly concentrated and interventions to increase peasant access to land have often faltered or been derailed (Albertus and Kaplan, 2013). As a result of these and other factors, rural informality is rampant, with an estimated 48% of rural parcels lacking a registered title.6 These long-running land issues interact with the displacement crisis in complex ways. For instance, there is evidence that elites and other actors have leveraged tenure insecurity to ex- propriate peasant land during the conflict (Vargas Reina, 2021). Influxes of IDPs have also gener- ated land conflicts among marginalized peasant and indigenous communities competing over resources and territory (de Benito, 2015). And finally, with respect to IDP resettlement and resti- tution, the fact that many rural landholders have no formal ownership of their land means that achieving the legal return of property lost during conflict is rife with judicial and administra- tive barriers (CODHES, 2015). The result is that displacement both results from and feeds into Colombia’s land historical land problems. Given the scale of the crisis, a key element of post-conflict policy has been the question of how to resettle IDPs. While prior legislation and elements of the peace process provide for the restitution, or legal return, of land and property to conflict victims, progress has been slow and impeded by numerous factors7 . Figure 1 visualizes the scale of the restitution problem, high- lighting how only a fraction of the estimated number of restitution-eligible victims have begun the restitution process, and among those, an even smaller percent have finished the process. The substantial gap between what the restitution need the state anticipated (based on the inci- dence of displacement) and the actual number of restitution claims received (Garay et al., 2016, p. 18), raises the question of why few victims pursue restitution – a question we shed light on in the study.8 5 The process of political decentralization and democratization since the 1990s has also played a role in the po- litical targeting of civilians (Steele, 2017). 6 Figure retrieved from USAID’s Land-Links information portal. https://www.land-links.org/ country-profile/colombia/#1528464011915-6f6e82e5-9a53 7 For a comprehensive review of barriers to restitution in Colombia, see CODHES (2016) 8 Measuring the amount of restitution need is not straightforward and there are varying, competing estimates of total need. The estimates presented here come from De Justicia. 4 Status of case: No request Request filed Request completed (Total: 360,000) (225,000) (124,000) (10,400) 1,000 households Figure 1: Progress on land restitution in Colombia as of 2019. Source: De Justicia. 5 While there is important research on Colombian IDPs by economists and political scientists (e.g., (Ibáñez and Moya, 2010; Ibáñez and Vélez, 2008; Steele et al., 2021)) there is still much that is unknown about the effects of displacement on IDPs, and particularly its effects on social and political attitudes and behaviors. We know even less about what comes after displacement: why some IDPs are ultimately able to return home and seek out restitution, while others are unable or unwilling to do so. In what follows we shed some light on these questions. 3 THEORETICAL MOTIVATION 3.1 The effects of victimization on social cohesion Displacement is a form of wartime victimization in which civilians are forced to flee their home, either because they are fleeing the cross-fire of armed actors or because they are deliberately expelled by combatants (Steele, 2009). Research on the effects of displacement can be broadly separated into research that focuses on: (1) how displacement affects the victims of displace- ment; and (2) how displacement affects the communities which IDPs either arrive to or are ex- pelled from. Our focus is on how displacement affects the displaced and, as a result, our analyt- ical starting point is a broader literature on the general effects of victimization. This larger literature has found some robust links between victimization and a range of so- cial and political behaviors, including outcomes related to social cohesion. Early work by Bel- lows and Miguel (2009) and Blattman (2009) document a positive effect of exposure to violence on post-conflict civic and political participation (measured by activities such as attending com- munity meetings, joining social and political groups, voting, etc.) in Sierra Leone and Uganda, respectively. Likewise, more recent research in other parts of the globe has documented that those exposed to wartime violence generally are more likely to participate socially, join more social groups, and assume increased leadership roles within their communities (Bauer et al., 2016). In behavioral studies, people who have had greater exposure to conflict have been found to exhibit more pro-social behaviors, measured through altruistic donations, public good con- tributions, and investment in trust-based transactions, among others (Gilligan, Pasquale and Samii, 2014). In contrast to these studies are a sizable set of findings linking victimization to ‘anti-social‘, conflictual, or vengeful attitudes and behaviors (Vinck and Pham, 2009). In a set of behavioral experiments in rural Colombia, Vélez et al. (2016) find that victimization affects subjective inse- curity and has a negative effect on social cohesion, cooperation, and trust, although it does have a positive effect on altruism. Likewise, other work has found that pro-social behavior is not nec- 6 essarily correlated with increased political participation (Steele et al., 2021). Finally, some work finds that victimization increases anger and punitiveness towards out-groups, undermining so- cial cohesion in communities where victimization rates are high (Beber, Roessler and Scacco, 2014; Balcells, 2017). The mechanisms underlying these disparate findings vary, but they generally center on psy- chological theories that link traumatic experiences to behavior. Studies showing that victimiza- tion decreases social cohesion and other pro-social behaviors often point to the development in victims of post-traumatic stress disorder (PTSD), which has been linked to antisocial attitudes and behaviors in both conflict settings (Vinck and Pham, 2009) and the general population (Landwehr and Ojeda, 2021).9 Other psychological mechanisms emphasize the social dimension of vic- timization, where feelings of shame, fear, or ostracism cause victims to retreat from public life (Wood, 2008; Nordås and Cohen, 2021). By contrast, studies that find victimization increases social cohesion and other pro-social behaviors often point to “post-traumatic growth” as a po- tential mechanism, where individuals respond to victimization with an increased sense of re- silience and a desire to engage more actively in their communities (Bauer et al., 2016; Bateson, 2012). Schon (2020) argues, for instance, that witnessing violence delays the decision to flee among Syrians because post-traumatic growth leads people to stay and try to help their home communities. Precisely why studies on victimization have appeared to produce contradictory results with competing psychological mechanisms is perhaps one of the key standing questions in the lit- erature. One possibility is that in testing for the effects of ‘victimization’, broadly conceived, researchers are masking substantial heterogeneity among different types of victimization ex- periences. For instance, there is mixed evidence that suggests victimization effects can change over time (Balcells, 2012; Rohner, Thoenig and Zilibotti, 2013; Villamil, 2020). The context in which victims find themselves can also moderate the impact of victimization: scholars have doc- umented how wars can create “hierarchies of victimhood” that make some victims more likely to mobilize and participate than others (Berry, 2017) and that patterns of rebel governance can impede or mobilize victim activism (e.g. Arjona, 2016),10 The type of victimization itself should also matter, as some forms of violence impact not just the individual but also the community 9 However, not everyone who experiences trauma suffers from PTSD, and there is great diversity in the way PTSD can emerge (Galatzer-Levy and Bryant, 2013). There is also a debate in the psychology literature on whether PTSD is a universally experienced condition or whether it is relevant only to specific cultural groups. 10 Justino and Stojetz (2019) find that individual exposure to governance in times of war increases participation in long-term local planning and provision of public goods, and stimulates interest in and commitment to politics, governance, and collective action, even if only locally. Steele et al. (2021) argue, for their part, that exposure to rebel mechanisms of competitive state-building increase informal political engagement (e.g., participating in strikes and protests, joining civil society organizations, etc.) once a conflict has ended. 7 (Wood, 2008), with downstream implications for social cohesion and mobilization. There are thus potentially different effects and mechanisms at play depending on the type of victimization, the timing of the victimization, and the context in which victims find them- selves. With this in mind, we turn now to a discussion of how displacement constitutes a distinct form of victimization. 3.2 Displacement as a distinct form of victimization In other strands of the conflict literature – for instance, on the causes of violence – it is today well- established that displacement is a type of violence that should be studied separately (Ibáñez and Vélez, 2008; Steele, 2017). Yet when it comes to research on the effects of displacement on victim outcomes, displacement tends to be lumped in with other forms of victimization (Fearon and Shaver, 2021). For example, Lupu and Peisakhin (2017) focus on the deaths of relatives among the survivors of a large forced displacement campaign, but don’t consider how the displacement itself might have affected victims. Similarly, Hazlett (2020) studies the determinants of atti- tudes towards peace among refugees but focuses on the effects of their exposure to violence and not on the experience of displacement itself.11 The key way in which displacement differs from other forms of victimization is that, in dis- placement, victims are forced to relocate, often far from their homes. This forced relocation has a number of distinct implications for victim outcomes. First, forced relocation means that dis- placed households are often separated from substantial proportions of their wealth, in the form of lost property, assets, and the costs associated with moving and resettling (Ibáñez and Vélez, 2008). Resettlement also means moving to areas where there is often a mismatch between vic- tims’ human capital and local labor market demand – seen most strikingly, for instance, with rural, agrarian households that are forced to relocate to urban environments (Calderón and Ibañez, 2009). The implication should be reduced economic welfare of displaced households even in the longer-term, as these effects are not one-off shocks but likely to persist over time. Second, forced relocation also means that long-standing networks of cooperation, reciprocity, and exchange within communities are often severed (Dorff, 2017; Torres and Casey, 2017). While this should ultimately reduce social cohesion and the capacity for collective action among vic- tims of displacement, there are also some potential counter-trends in how forced displacement affects social cohesion. For one, refugees often cluster in particular places: a function of geogra- 11 An exception is a recent study by Steele et al. (2021), who survey victims in Colombia –including IDPs. However, they mainly focus on the political effect of IDPs registering as victims. They find that IDPs who register with the state are more likely to engage in politics through formal channels (for example, voting), in line with Voytas (2021)’s findings of Pinochet’s victims in Chile. 8 phy, information networks, and other factors (Steele, 2009; Schon, 2018). Such clustering may provide victims with the proximity and common interest necessary to spur social cohesion and collective action. Relatedly, the welfare losses associated with displacement – while undermin- ing the economic security of victims – may also present victims with the motive and shared in- terest to pursue collective action as they try to recover land, property, or win reparations from the state. Agitation for wartime reparations may be particularly salient in post-conflict settings, where victims must often navigate a complex ‘state bureaucracy of victimhood’ in order to ac- cess services (Steele et al., 2021; Cronin-Furman and Krystalli, 2021). ’ Finally, displacement is a particularly long-lasting form of victimization. Psychological re- search on the mental well-being of refugees shows that IDPs can display significant distress symptoms for as long as their status as displaced people lasts (Vinck and Pham, 2009). This means that displacement effects should be more likely to persist over time than other forms of wartime victimization. The duration of displacement may also moderate the overall effects of displacement – for instance, it is possible that in in the short run the network-destroying dimen- sion of displacement would reduce overall social cohesion, while in the long-run it is possible IDPs could become more engaged and cohesive in their demands of the state for reparations. In sum, while the effects of displacement on household economic welfare are likely to be unambiguously negative, the overall effect of displacement on social cohesion and other pro- social attitudes could potentially run in either direction depending on the balance of effects on victim’s social networks and capacity and motivation for collective action. Displacement effects should also be particularly persistent, in that many of the dynamics generated by displacement are likely to persist and change over time. In addition to exploring overall displacement effects, we also leverage our data collection effort to explore heterogeneity among IDPs in their post-displacement experiences. We focus on two important yet understudied dimensions of life post-displacement: first, why some dis- placed people ultimately return to their origin communities while others do not; and second, why some IDPs pursue formal reparations from the state while others are unwilling or unable to do so. Research on return is still scarce, and it is more focused on refugees than on IDPs. There is not much evidence on the preferences for return among IDPs, and even less for desire for return in the longer-term, as is the case in our study.12 Research on reparations is also scarce, particu- larly on the determinants of who advocates for (and ultimately is able to win) reparations.13 12 Some recent contributions to this topic are: Vinck and Pham (2009), Camarena and Hägerdal (2020), Schwartz (2019), Ghosn et al. (2011), and Fakhoury (2021). 13 Some recent exceptions look at the political effects of receiving reparations (Voytas, 2021; Steele et al., 2021). 9 4 RESEARCH DESIGN 4.1 Data Data for the project comes from a survey of rural households in Colombia fielded between March and June of 2017. The survey was commissioned as part of the performance evaluation14 of US- AID’s Land and Rural Development Program (LRDP), a program designed to improve institu- tional capacity in relation to land formalization, land restitution, and broader rural develop- ment outcomes. The survey was fielded at the halfway point of the LRDP’s lifespan, part of a broader effort to measure the program’s strengths and weaknesses up to that point in time. The survey captured almost 1,500 responses across fifty rural and conflict-affected munici- palities, 25 of which were selected because of their involvement in or relevance to the LRDP and an additional 25 which were selected through matching on observables to serve as comparable municipalities that did not experience any LRDP programming (Figure A.1). Survey responses were collected on tablets and in-person, in or near the respondent’s home. More detailed infor- mation on data collection, sampling, implementation, and ethical considerations of sampling in violent contexts are available in the Appendix. This survey data has a number of strengths for the study of displacement. First, the sam- pling frame explicitly over-sampled areas and communities that have experienced a lot of dis- placement and as a result, produced a sample where approximately half of respondents have been displaced (sampling characteristics can be seen in Figure A.2). Second, and relatedly, the survey contains long modules both on displacement experiences as well as on a variety of post- displacement outcomes that can be used to explore the effects of displacement. Finally, the sample is otherwise relatively homogeneous, capturing rural areas with a relatively high de- gree of exposure to conflict. This means that – following statistical adjustment to control for potential confounds in who is displaced – there is good coverage for comparing respondents who have experienced displacement against those who have not experienced displacement. An important limitation, however, is that the sample is not representative of the universe of IDPs in Colombia.15 In terms of which outcomes to focus on with respect to displacement effects, we consider a broad range, which we group into six categories. Given prior work on the welfare-diminishing 14 Generally, performance evaluations differ from impact evaluations in that the latter tend to have built-in ran- domized treatment and control groups for subsequent estimation of effects, while the former tend not to or are often more descriptive in their evaluation goals. 15 While the distribution of respondent characteristics in the household survey largely resemble comparable re- spondent characteristics from the recent (nationally representative) rural waves of LAPOP in Colombia (see Figure A.3), we cannot strictly generalize our findings outside of these areas. 10 consequences of displacement (e.g., Ibáñez and Moya (2010)), we look at survey items captur- ing different dimensions of economic well-being (welfare). Following a large literature on the social and political consequences of victimization (Bauer et al., 2016), we also look at items mea- suring social cohesion and community cooperation (cohesion), interest or involvement in pol- itics (politics), and trust in state institutions (trust). Finally, we consider two other sets of outcomes that are particular to displacement and the Colombian context: the first are attitudes bearing on land reform and tenure security (land), which may be affected by displacement and are important elements of politics in countries with rural economies (Albertus, 2020). The sec- ond are attitudes towards the peace process between the Colombian government and the FARC (peace), which prior work suggests may be influenced by victimization experiences (e.g., Tellez (2019b); Hazlett (2020)). 4.2 Estimating displacement effects Estimating the effect of displacement – as with other victimization experiences – is difficult. Some civilians are more likely to experience displacement than others, either as a result of back- ground characteristics that expose them to conflict at higher rates (e.g., poverty, insecurity, weak property rights) or characteristics that single them out as targets during war (e.g., perceived loy- alties, location in relation to battle-lines, lootable resources). The decision to flee itself – i.e., not just who is exposed to conflict, but who “chooses” to flee rather than stay remain in the face of exposure (Schon, 2019) – may also be endogenous to outcomes of interest. And, unlike other ‘treatments’ of interest in the social sciences, field experiments and randomized control trials do not present a solution to these identification problems. Our approach to addressing these identification problems relies on inverse probability weight- ing (IPW) (Imbens, 2000; Cole and Hernán, 2008). In short, IPW resembles matching methods where researchers attempt to balance pre-treatment characteristics among treated and control groups to create ‘apples-to-apples’ comparison groups that vary only in treatment status. IPW has the added advantage of not discarding observations that are left unmatched, instead re- weighting the sample towards observations that are similarly likely to have received treatment. The weights create a synthetic sample in which subjects who have received a treatment (the displaced) and subjects who have not received a treatment (the non-displaced) are balanced on pre-treatment characteristics (i.e., where the distribution of pre-treatment covariates is in- dependent of treatment assignment) (Austin, 2011). In our application, weights are generated by the covariate balancing propensity score (CBPS) methodology, which models treatment as- signment while optimizing covariate balance (Imai and Ratkovic, 2014), though our results are 11 robust to other approaches. An important part of any approach to estimating treatment effects with observational data is deciding what covariates to adjust for. Table A.1 lists and provides justification for all covari- ates included in our IPW procedure. We include covariates (both at the individual and municipal- level) that are likely affect both the likelihood of being displaced and the class of outcomes we are interested in, while avoiding adjustment for post-treatment variables (e.g., completed years of education, which are likely to be affected by displacement) (Montgomery, Nyhan and Torres, 2018).16 A first step in evaluating our estimation strategy is examining covariate balance post-weighting. Figure 2 shows remarkable improvements in balance before (‘unadjusted‘) and after weighting (‘adjusted‘), and also showcases the problem with naively comparing outcomes of IDPs and non- IDPs. Prior to adjustment, IDPs and non-IDPs are fundamentally different in many ways that make comparison difficult (e.g., IDPs tend to have larger families). Post-weighting, many of these baseline differences shrink to zero. This can be seen, for instance, in the LRDP’s sampling design itself, which actively sought out places with displacement: respondents in LRDP areas are substantially more likely to have been displaced than those in other areas. Post-weighting, this difference effectively disappears. Table A.2 provides further summary statistics and also shows that we retain a relatively large post-weighting effective sample size (ESS). 16 Importantly, our municipal-level controls are based on the municipality of origin for respondents in our sample (i.e., for the displaced this is the municipality they were displaced from; for the non-displaced, it is where they currently live), out of concern that characteristics associated with the current location of the displaced are post- treatment. 12 Propensity score (distance) White/non-white Sex (male) Population (logged) Other victimization experiences Number of people displaced (avg) Municipal poverty index Municipal homicides (avg) LRDP programming area literacy Household size FARC presence (prop. of years) Distance to capital AUC presence (prop. of years) Altitude (logged) Age (years) % rural population 0.0 0.5 Standardized differences between treatment and control Statistic: Unadjusted Adjusted Figure 2: Love plot of covariate balance before and after weighting. One potential confound that is specific to displacement but unaddressed by the set of co- variates discussed above is that exposure to displacement may be a function of an individual’s (real or perceived) political loyalties (Steele, 2011; Balcells and Steele, 2016), which are in turn correlated with outcomes of interest. In the Colombian context, this might mean that civilians who are (perceived to be) loyal to the left-wing guerrillas are more likely to be targeted by right- wing groups, or the military. Using an endorsement experiment designed to indirectly elicit truthful attitudes towards the FARC (Bullock, Imai and Shapiro, 2011), we find no evidence that the displaced in our sample17 are any more likely to have favorable views of the FARC than their counterparts (Table A.6; details on design available in the Appendix). While the experiment im- perfectly addresses this concern, it does present some evidence against the possibility that the displacement effects we estimate are confounded by ideology.18 17 Of course, we are not claiming that (perceived) loyalties have no bearing on the incidence of displacement in Colombia writ large. We are instead suggesting that, among the people we interviewed, loyalties do not seem a major potential confounding difference in attributes between IDPs and non-IDPs. 18 In particular, the endorsement experiment measures attitudes in the present and not attitudes at the time of displacement, which may have taken place years in the past. 13 5 RESULTS 5.1 Main Results: Outcomes of Displacement We compile the main results of the project into Figure 3, which plots effect estimates (points) and 95% confidence intervals (lines) for each of the six types of outcomes. Overall, a few distinct patterns emerge. First, displacement seems to have substantial negative welfare effects: IDPs report lower levels of income today, are more likely to say their income is not enough to subsist on, and expe- rience food insecurity more frequently.19 They are also more likely to be at risk of eviction, and to expect to have a harder time transferring their property, patterns that reflect the economic damages of displacement. We also test for whether displacement causes shifts in general risk- seeking behavior, but find no evidence of differences between IDPs and non-IDPs.20 Our evidence also suggests displacement may increase levels of social cohesion and mobi- lization: IDPs are more likely to say they’ve recently helped solve a local problem, support col- lective action (“Would you get together with other rural families to demand the state improve your sit- uation? ”), are more willing to pay taxes if they were used to help the needy, and are more likely to say they’ve attended a variety of local meetings21 . These findings echo prior work on the mo- bilizing effects of victimization (Blattman, 2009; Bauer et al., 2016), though to our knowledge these patterns have not been documented with respect to displacement. Displacement also appears to influence attitudes towards peace-building. People who have been displaced are more likely to express support for the peace process (“to what extent do you support the implementation of the peace accords with the FARC?”) and to be optimistic about the ability of ex-FARC members to reintegrate into society (“How optimistic or pessimistic do you feel about demobilized FARC members successfully reintegrating into society?”) than those who weren’t displaced, even as they are less likely to say they feel safer now. This latter pattern bearing on expectations of future safety is more mixed: they are no more likely to fear the formation of new groups, nor to expect less conflict after the end of the peace process.22 19 These patterns are in line with prior work showing that IDPs have extreme poverty rates over three times higher than the national average (CODHES, 2016). 20 We picked three items from a “Risk-Acceptance Index” used in the American National Elections Survey (ANES) (Kam, 2012). The items we selected were very weakly correlated, and thus we did not combine them into an index. 21 The “women’s meetings” item was only administered to respondents who identified as women. 22 We delve more deeply into the effects of displacement on peace attitudes through a conjoint experiment that randomly varies the attributes of hypothetical peace agreements (Hainmueller, Hopkins and Yamamoto, 2014). Overall, we find no evidence that the displaced and non-displaced react differently to the peace agreement at- tributes, using the preference heterogeneity test in Leeper, Hobolt and Tilley (2020) (Table A.4). Attribute-level results available in Table A.5. 14 Cohesion Land Peace Would pay more tax Property safe from outsiders Supports peace Property safe from govt Helped solve local problem Less conflict future Property is respected Favors collective action Land rights protected Knows land victim law Attends women's meetings Land rights clear Feel safer now Attends political party meetings Land help available Fear new groups Attends farmer meetings Land conflicts unlikely Attends community meetings Belief farc can demobilize Knows of lru 0.0 0.1 0.2 0.3 -0.2 -0.1 0.0 0.1 0.2 -0.2 0.0 0.2 0.4 Politics Trust Welfare Trust in state Risk aversion 2 Supports land invasions Trust in police Risk aversion 1 Supports expropriation Trust in personeria Monthly income Trust in mayor Invested in property Interest in politics Trust in incoder Income not enough Did not vote Trust in courts Has enough food Belief in titling Eviction risk Belief restitution is fair Belief in state land reform Ease of property transfer -0.1 0.0 0.1 0.2 0.3 -0.2 0.0 0.2 -0.8 -0.4 0.0 Coefficient estimate on displacement Figure 3: Estimated effect of displacement across outcomes. In terms of political attitudes, differences are more muted. The displaced are no more likely than the non-displaced to claim they will vote in the next presidential elections, or to say they are generally interested in politics. They are, however, more likely than the non-displaced to en- dorse land expropriation as a policy tool to improve access for the landless and (unsurprisingly) to believe that the restitution process is generally fair and equitable. Where there is consistently little to no evidence of displacement effects is in attitudes to- wards land tenure and trust in the state. Across the board, displaced people are no more likely than the non-displaced to feel their land tenure is insecure, that their property is safe, that land conflicts are likely to emerge, even as they are more aware of the government’s land restitution unit (LRU). In terms of trust, the displaced are similarly no more or less likely to express trust in varied state institutions, though they do appear more optimistic about the ability of the state to implement land reform as part of the peace process. Jointly, these patterns perhaps attests to the widespread nature of tenure insecurity in the rural Global South (Deininger, 2003; Broe- gaard, 2005). The results are robust to a few other, common modeling choices. With IPW, one decision- point is what algorithm or process to use in constructing the propensity scores underlying the 15 weights. Figure A.4 shows estimates are largely consistent across five different methodologies for constructing propensity scores. A different decision-point concerns how to identify ‘extreme’ weights and whether or not to remove them prior to analysis. Figure A.5 shows results are largely similar even while truncating extreme weights (above the 99th percentile). In sum, the picture that emerges of the effect of displacement on civilians centers on house- hold welfare and social cohesion. IDPs are clearly worse off in economic terms than their peers, even among the very poor and rural population that we sampled, and many years after the fact: the median respondent in our sample was displaced sixteen years before their responses were collected. This finding mirrors prior research on the negative welfare consequences of displace- ment (Ibáñez and Moya, 2010). The displaced also appear much more mobilized and invested in community than their coun- terparts, in part reflecting prior work on the potentially ‘positive’ effects of displacement on so- cial cooperation, though at the same time clashing with work that suggests these pro-social ef- fects are a function of social ties (which have often been destroyed for IDPs) (Dorff, 2017). One possibility, as discussed in Steele et al. (2021) is that the displaced have something at stake – recovering lost land or property, or returning to origin communities – that makes them more likely to mobilize and experience social cohesion than the non-displaced. We explore these two outcomes in the following section. 5.2 What happens after displacement? Beyond estimating the effects of displacement, our data also provide us a rare view into the post- displacement experiences of IDPs. Here, we focus on two key questions of post-displacement life: first, which victims return home after displacement? And second, which victims pursue legal restitution of their land? 5.2.1 IDP RETURN Figure 4 presents a few descriptive patterns related to IDP return. More than half of IDPs in our sample do eventually return to their place of origin, and among those, the most common motivations for return are economic and social. Notably, almost no one in our sample listed state assistance as a factor facilitating return, which speaks to persistent low state capacity in redressing wartime victims (Voytas, 2021). Return for many is also a long process: the median victim who returned home did so after five years, but many spend much longer. 16 Were you able to return home? Why did you decide to return? economic necessity Yes family/friends security improved No destination worse state helped return 0% 20% 40% 60% 0% 10% 20% 30% 40% 50% Percent of respondents Percent of respondents (N = 730) (N = 399) Length of time until return Why were you unable (in years) to return? 50 security concerns 40 30 property destroyed 20 can't recover land 10 destination better 0 0 5 10 15 20 25 30 35 40 45 50 55 0% 10% 20% 30% 40% 50% Percent of respondents (N = 327) Figure 4: Summary of descriptive characteristics concerning ability of displaced to return home. Among those who never return, security concerns and an inability to recover property (ei- ther because it was destroyed, or because of legal issues) are the dominant obstacles.23 Inse- curity in the countryside is driven by various factors, including the violent expansion of armed groups into areas previously controlled by the FARC (Prem et al., 2020). Such insecurity can limit the ability of victims to return, even for those who are already relatively near. Team interviews with a Yupka indigenous community in the department of Cesar, for instance, highlighted how paramilitary groups continue to occupy Yupka land and forbid access to state officials working to map out the boundaries of the community’s territory.24 The return of IDPs to areas where property rights are characterized by informality can also generate social conflict between different groups laying claim to the same land. In Colombia, land conflicts between indigenous groups and peasants have surged over questions of who owns what (de Benito, 2015). These patterns were echoed in our interviews with members of the Yupka community, who argued that returning IDPs were creating conflict because “they [the 23 These findings echo reports by the Consultancy on Human Rights and Displacement (CODHES), who in a 2014 study found that security concerns are among the most important barriers to IDP return (CODHES, 2015). 24 Research team focus group discussions in Valledupar, Cesar. March 27th, 2017. 17 peasants] are being given back land that was part of our ancestral territory”.25 As discussed in the next section, informality can also exacerbate conflicts between returning IDPs and locals laying claim to the same parcel of land (CODHES, 2016, p. 453). 5.2.2 SEEKING LAND RESTITUTION These dynamics demonstrate that the return of IDPs is often entangled with the question of whether IDPs are able to recover lost property through the restitution process. As seen in Fig- ure 5, only a quarter of IDPs in our sample formally seek out restitution of property from the state. While the majority among these have some professional entity advocating their case, many claim they are serving as their own advocates. The process is also lengthy: among those who achieve restitution, the average respondent spent 2.25 years in the process, but others at different stages of restitution spend much longer. Have you sought restitution? Who is advocating your case? Public defender Yes No representation Private lawyer Other No Family NGO 0% 20% 40% 60% 80% 0% 10% 20% 30% 40% 50% Percent of respondents Percent of respondents (N = 726) (N = 166) Time in restitution process 30 20 10 0 Under review Request accepted Verdict reached, Land restituted no restitution In years (N = 111) Still waiting for restitution? No Yes Figure 5: Summary of descriptive characteristics concerning restitution. 25 Research team focus group discussions in Valledupar, Cesar. March 27th, 2017. 18 The restitution process itself is opaque, costly, and difficult for victims to navigate. Author interviews with state officials and advocacy organizations point to problems with institutional coordination, low state capacity, and the difficulty of adjudicating competing claims as key ob- stacles in the restitution process. Officials noted how courts and legal services are concentrated in urban areas whereas restitution cases are concentrated in rural areas, limiting access for vic- tims and creating coordination problems among entities26 . Even when a restitution case is adju- dicated by the courts, agencies often lack the capacity to implement and enforce court rulings.27 . Restitution cases are also often complicated by the presence of secondary occupants who live on the land abandoned by the IDP 28 . Indeed, of the 144 IDPs we interviewed who had a pending restitution case, about one quarter of them faced an opposing claimant. As (CODHES, 2016) notes, competing claims introduce additional normative and legal complexity in balancing the needs of IDPs against those of secondary occupants, who may have purchased the land in good faith and depend on it for survival. 5.2.3 PREDICTIVE ANALYSIS Given these obstacles, a natural follow-up question is what factors explain variation in who re- turns and who seeks out restitution. Given the nature of this question and the data – a relatively small sample of displaced respondents (N = 733), many potential explanatory variables, and few ex ante theoretical expectations – we adopt a prediction framework (Shmueli, 2010; Hill Jr and Jones, 2014), and ask which variables are most predictive of these outcomes. To retain inter- pretability, we use LASSO regression, a common variable selection methodology (James et al., 2013). LASSO uses a penalization parameter to shrink coefficient estimates to zero based on how the resulting model performs in cross-validation. Our process is as follows. First, we select a relatively large set of explanatory variables that are reasonably relevant to each outcome. Next, we use 10-fold cross-validation to tune the pe- nalization parameter, using the explanatory variables in the training data. We then evaluate the predictive out-of-sample performance of the resulting model on the test set. We report the main results of the LASSO model in Figure 6. In terms of who returns, men and those who report greater comfort with risk-seeking were more likely to have returned home. Conversely, those who were deliberately displaced by force (as opposed to fleeing general inse- 26 Author interview with Director of Access to Justice Program (March 9th, 2017) and interview with Director of Defensoría Pública and head of Group for the Judicial Representation of Victims (March 7th, 2017). 27 Author interview with Secretary of Government of Santander de Quilichao, Cauca. April 3rd, 2017. 28 Author interviews with Director of Defensoria Publica and head of Group for the Judicial Representation of Victims. Bogota, Colombia, March 7th, 2017. 19 curity) were less likely to return, as were respondents who identified as white/mestizo (as op- posed to other racial categories, such as Afro-Colombian or indigenous). That non-white IDPs are more likely to return is in many ways surprising: multiple interviews with state officials high- light a dearth of specialists in ethnic issues and a “need for greater dialogue” with minority com- munities affected by the conflict.29 On the other hand, if victims are returning largely without the help of the state it may be that these communities have a higher capacity for collective ac- tion that facilitates return (Arjona, 2016). With respect to restitution, men also tend to be more likely than women to formally seek out restitution of lost property. This gender gap is also reflected in an earlier national survey by CODHES (2015), who find that although women-headed households constitute 45% of the dis- placed population, they comprise only 30% of restitution cases (pg. 130). In addition, respon- dents who have attended local meetings for victims of the conflict also appear more likely to seek out restitution, while the probability that a victim seeks out restitution decreases in how rural their origin municipality is. Interestingly, despite the legalistic nature of restitution pro- cesses, income and education seem to have little predictive power in who seeks restitution. 29 Author interview with Director of Access to Justice Program (March 9th, 2017) and interview with Director of Defensoria Publica and head of Group for the Judicial Representation of Victims (March 7th, 2017). 20 Was respondent able to return home? Sex (male) Risk-seeking Years of education Years since displacement Monthly income Age (years) Household size News consumption White/non-white Displaced by force -0.2 0.0 0.2 Variable importance (LASSO) Did respondent seek restitution? Attends victims' meetings Sex (male) Age (years) Years since displacement Years of education White/non-white Risk-seeking News consumption Monthly income Knows Victim's Law Household size Displaced by force % rural population -0.2 0.0 0.2 0.4 Variable importance (LASSO) Figure 6: Penalized coefficient estimates from LASSO regression. Variables with missing bars had estimates pulled to zero. 21 How well do these models predict who returns and who seeks out restitution? We report receiver operating (ROC) curves for out-of-sample (test set) predictions in Figure A.6. The area under curve metric is .68 for the decision to return and .7 for restitution-seeking, which repre- sents an improvement of 36% and 40% over guessing at random, respectively. In overall terms, these improvements are modest, which suggests other, unknown factors may be more predic- tive of these outcomes. 6 POLICY AND PROGRAM IMPLICATIONS Displacement is a deeply traumatizing experience that affects millions around the world. The prospect of resettling, restituting, and rebuilding communities destroyed by displacement also represents a daunting challenge for societies in conflict and global humanitarian efforts. Un- derstanding how displacement shapes victim’s trajectories is thus crucial for crafting informed policy-making and advocacy. In this project, we advance our understanding of the effects of dis- placement using original data from rural households with a high incidence of displacement in Colombia. The results have numerous implications for both the academic literature and policy- makers. First, the results provide strong evidence that there can be substantial heterogeneity of ex- perience within the broad umbrella term of “wartime victimization”. Even after adjusting for other forms of wartime victimization, the experience of displacement appears to both under- mine and mobilize victims in distinct ways that may set them apart from other conflict victims. Disaggregating wartime victimization may thus be one way of working towards resolving what are at times contradictory findings in the literature on the effects of victimization. With respect to policy, this potentially means designing interventions that take into account the dispropor- tionate economic burden faced by IDPs (Arias et al., 2014), or that leverages IDPs apparently higher capacity for collective action in ‘bottom-up’ peacebuilding efforts (Autesserre, 2021). Second, the study underscores that there is substantial variation among IDPs in what comes after displacement. Explaining and addressing this variation is likely crucial for the long-term well-being of IDPs. With respect to return, ongoing security concerns originating from post- conflict instability (Prem et al., 2020) and difficulty recovering property appear as major impedi- ments. Intervention design should thus take into account the fraught security situation present in many post-conflict settings. In terms of restitution-seeking and the recovery of property, a key insight is that few victims formally pursue restitution, and that many in our sample claim to lack access to representation in what is an often labyrinthine legal process (Sanín, Huertas and Hernández, 2019). A key area for policymakers to intervene is thus in facilitating the resti- 22 tution process for victims who want restitution, and to explore how the differential uptake of restitution may exacerbate inequities in post-conflict societies. Interventions to improve the restitution process and formalize land ownership should also, in turn, help to reduce local con- flicts among IDPs and secondary occupants and ensure the rights of each in the process. Finally, it is important to note that the results presented here are from a particular time (the post-accord peace process) and a particular place (rural Colombia). Displacement effects can vary significantly across countries in ways that are difficult to synthesize. More theoretical and empirical work is needed on the effects of displacement, especially work that leverages re- search design strategies to identify displacement effects. One clear avenue for future research is work that identifies the determinants of IDP return, particularly research that distinguishes between the desire and ability of IDPs to return (Arias, Ibáñez and Querubin, 2014; Camarena and Hägerdal, 2020). Another avenue is to explore how IDPs navigate post-conflict programs and policies related to victim’s reparations and restitution (Voytas, 2021; Steele et al., 2021). Even if such programs are key to rebuilding after war, their impact will be limited if IDPs face significant barriers to uptake. 23 REFERENCES Albertus, Michael. 2020. “Land reform and civil conflict: Theory and evidence from Peru.” American Journal of Political Science 64(2):256–274. Albertus, Michael and Oliver Kaplan. 2013. “Land Reform as a Counterinsurgency Policy Evidence from Colombia.” Journal of Conflict Resolution 57(2):198–231. Arias, María Alejandra, Adriana Camacho, Ana María Ibáñez, Danuel Mejía and Catherine Rodríguez. 2014. Costos económicos y sociales del conflicto en Colombia:¿ Cómo construir un posconflicto sostenible? Ediciones Uniandes-Universidad de los Andes. Arias, María Alejandra, Ana María Ibáñez and Pablo Querubin. 2014. “The desire to return during civil war: Evidence for internally displaced populations in Colombia.” Peace Economics, Peace Science and Public Policy 20(1):209–233. Arjona, Ana. 2016. Rebelocracy. Cambridge University Press. Aronow, Peter M. and Cyrus Samii. 2016. “Does regression produce representative estimates of causal effects?” American Journal of Political Science 60(1):250–267. Austin, Peter C. 2011. “An introduction to propensity score methods for reducing the effects of confounding in observational studies.” Multivariate behavioral research 46(3):399–424. Autesserre, Séverine. 2021. The Frontlines of Peace: An Insider’s Guide to Changing the World. Oxford University Press, USA. Balcells, Laia. 2012. “The Consequences of Victimization on Political Identities: Evidence from Spain.” Politics & Society 40(3):311–347. Balcells, Laia. 2017. Rivalry and Revenge: The Politics of Violence during Civil War. New York: Cambridge University Press. Balcells, Laia and Abbey Steele. 2016. “Warfare, Political Identities, and Displacement in Spain and Colombia.” Political Geography 51:15–29. Bateson, Regina. 2012. “Crime Victimization and Political Participation.” American Political Science Review . Bauer, Michal, Christopher Blattman, Julie Chytilová, Joseph Henrich, Edward Miguel and Tamar Mitts. 2016. “Can war foster cooperation?” Journal of Economic Perspectives 30(3):249–74. Beber, Bernd, Philip Roessler and Alexandra Scacco. 2014. “Intergroup Violence and Political Attitudes: Evidence from a Dividing Sudan.” Journal of Politics 76(3):649–65. 24 Bellows, John and Edward Miguel. 2009. “War and Local Collective Action in Sierra Leone.” Journal of Public Economics 93(11):1144–1157. Berry, Marie E. 2017. “Barriers to Women’s Progress After Atrocity: Evidence from Rwanda and Bosnia-Herzegovina.” Gender & Society 31(6):830–853. Blattman, Christopher. 2009. “From Violence to Voting: War and Political Participation in Uganda.” American Political Science Review 103(2):231–247. Broegaard, Rikke J. 2005. “Land tenure insecurity and inequality in Nicaragua.” Development and Change 36(5):845–864. Bullock, Will, Kosuke Imai and Jacob N. Shapiro. 2011. “Statistical analysis of endorsement experiments: Measuring support for militant groups in Pakistan.” Political Analysis 19(4):363–384. Calderón, Valentina and Ana María Ibañez. 2009. “Labor market effects of migration-related supply shocks: Evidence from internally displaced populations in Colombia.”. Working paper. Camarena, Kara Ross and Nils Hägerdal. 2020. “When Do Displaced Persons Return? Postwar Migration among Christians in Mount Lebanon.” American Journal of Political Science 64(2):223–239. CODHES. 2015. “La intencíon de retorno de la población desplazada.”. CODHES. 2016. El reto ante la tragedia humanitaria del desplazamiento forzado: Análisis sobre el estado de cosas inconstitucional. Technical report Consultoría para los Derechos Humanos y el Desplazamiento. Cole, Stephen R. and Miguel A. Hernán. 2008. “Constructing inverse probability weights for marginal structural models.” American journal of epidemiology 168(6):656–664. Cronin-Furman, Kate and Roxani Krystalli. 2021. “The things they carry: Victims’ documentation of forced disappearance in Colombia and Sri Lanka.” European Journal of International Relations 27(1):79–101. de Benito, Cynthia. 2015. “Campesinos frente a indígenas, el otro conflicto colombiano.”. URL: https://www.elespectador.com/colombia/mas-regiones/ campesinos-frente-a-indigenas-el-otro-conflicto-colombiano-article-552931/ Deininger, Klaus W. 2003. Land policies for growth and poverty reduction. World Bank Publications. Dorff, Cassy. 2017. “Violence, Kinship Networks, and Political Resilience: Evidence from Mexico.” Journal of Peace Research 54(4):558–573. 25 Fakhoury, Tamirace. 2021. “Refugee return and fragmented governance in the host state: displaced Syrians in the face of Lebanon’s divided politics.” Third World Quarterly 42(1):162–180. Fearon, James D and Andrew Shaver. 2021. “Civil War Violence and Refugee Outflows.” ESOC Working Paper Series . Galatzer-Levy, Isaac R. and Richard A. Bryant. 2013. “636,120 Ways to Have Posttraumatic Stress Disorder.” Perspectives on Psychological Science 8(6):651–662. Garay, Luis Jorge, Fernando Barberi, Clara Ramirez, Fernando Vargas, Ivonne Moreno, Monica Jimenez, Juan Carlos Betancur, Diana Mendez, Roberto Lucio and Felipe Cifuentes. 2016. Lineamientos para la construccion de un plan estratégico interinstitucional de restitución de tierras. Technical report The United States Agency for International Development. Ghosn, Faten, Tiffany S. Chu, Miranda Simon, Alex Braithwaite, Michael Frith and Joanna Jandali. 2011. “The Journey Home: Violence, Anchoring, and Refugee Decisions to Return.” American Political Science Review pp. 1–17. Gilligan, Michael J, Benjamin J Pasquale and Cyrus Samii. 2014. “Civil War and Social Cohesion: Lab-in-the-Field Evidence from Nepal.” American Journal of Political Science 58(3):604–619. Hainmueller, Jens, Daniel Hopkins and Teppei Yamamoto. 2014. “Causal inference in conjoint analysis: Understanding multi-dimensional choices via stated preference experiments.” Political Analysis 22(1):1–30. Hazlett, Chad. 2020. “Angry or Weary? How Violence Impacts Attitudes toward Peace among Darfurian Refugees.” Journal of Conflict Resolution 64(5):844–870. Hill Jr, Daniel W. and Zachary M. Jones. 2014. “An empirical evaluation of explanations for state repression.” American Political Science Review pp. 661–687. Ibáñez, Ana María and Andrés Moya. 2010. Do conflicts create poverty traps? Asset losses and recovery for displaced households in Colombia. In The economics of crime: Lessons for and from Latin America. University of Chicago Press pp. 137–172. Ibáñez, Ana María and Carlos Eduardo Vélez. 2008. “Civil conflict and forced migration: The micro determinants and welfare losses of displacement in Colombia.” World Development 36(4):659–676. Imai, Kosuke and Marc Ratkovic. 2014. “Covariate balancing propensity score.” Journal of the Royal Statistical Society: Series B: Statistical Methodology pp. 243–263. Imbens, Guido W. 2000. “The role of the propensity score in estimating dose-response functions.” Biometrika 87(3):706–710. 26 Internal Displacement Monitoring Centre. 2021. Data retrieved from Global Internal Displacement Database: https://www.internal-displacement.org/database/displacement-data. James, Gareth, Daniela Witten, Trevor Hastie and Robert Tibshirani. 2013. An introduction to statistical learning. Vol. 112 Springer. Justino, Patricia and Wolfgang Stojetz. 2019. “Civic legacies of wartime governance.” WIDER Working Paper . Kam, Cindy D. 2012. “Risk attitudes and political participation.” American Journal of Political Science 56(4):817–836. Landwehr, Claudia and Christopher Ojeda. 2021. “Democracy and Depression: A Cross-National Study of Depressive Symptoms and Nonparticipation.” American Political Science Review 115(1):323–330. Leeper, Thomas J, Sara B Hobolt and James Tilley. 2020. “Measuring subgroup preferences in conjoint experiments.” Political Analysis 28(2):207–221. LeGrand, Catherine. 1986. Frontier expansion and peasant protest in Colombia, 1850-1936. University of New Mexico Press. Lupu, Noam and Leonid Peisakhin. 2017. “The legacy of political violence across generations.” American Journal of Political Science 61(4):836–851. Publisher: Wiley Online Library. Montgomery, Jacob M., Brendan Nyhan and Michelle Torres. 2018. “How conditioning on posttreatment variables can ruin your experiment and what to do about it.” American Journal of Political Science 62(3):760–775. Nordås, Ragnhild and Dara Kay Cohen. 2021. “Conflict-Related Sexual Violence.” 24(1):193–211. OCHA. 2017. Breaking the impasse: Reducing protracted internal displacement as a collective outcome. Technical report United Nations Office for the Coordination of Humanitarian Affairs. OCHA. 2020. “Results of Humanitarian Need Overview.”. https://kutt.it/TIhSOu. Prem, Mounu, Andrés Rivera, Dario Romero and Juan F. Vargas. 2020. “Selective civilian targeting: The unintended consequences of partial peace.” Available at SSRN 3203065 . Rohner, Dominic, Mathias Thoenig and Fabrizio Zilibotti. 2013. “Seeds of distrust: conflict in Uganda.” Journal of Economic Growth 18(3):217–252. Rozo, Sandra V and Juan F. Vargas. 2021. “Brothers or invaders? How crisis-driven migrants shape voting behavior.” Journal of Development Economics 150:102636. 27 Sanín, Francisco Gutiérrez, Rocío del Pilar Peña Huertas and María Mónica Parada Hernández. 2019. La tierra prometida: Balance de la política de restitución de tierras en Colombia. Vol. 4 Editorial Universidad del Rosario. Schon, Justin. 2018. “Using ERGMs to Disaggregate Displacement Cascades.” Journal of Social Structure 19(1):1–40. Schon, Justin. 2019. “Motivation and opportunity for conflict-induced migration: An analysis of Syrian migration timing.” Journal of Peace Research 56(1):12–27. Schon, Justin. 2020. Surviving the War in Syria: Survival Strategies in a Time of Conflict. Cambridge University Press. Schwartz, Stephanie. 2019. “Home, Again: Refugee Return and Post-Conflict Violence in Burundi.” International Security 44(2):110–145. Shmueli, Galit. 2010. “To explain or to predict?” Statistical science 25(3):289–310. Steele, Abbey. 2009. “Avoiding Displacement and Choosing Destinations in Civil Wars.” Journal of Peace Research 46(3):419–429. Steele, Abbey. 2011. “Electing Displacement: Political Cleansing in Apartadó, Colombia.” Journal of Conflict Resolution 55(3):423–445. Steele, Abbey. 2017. Democracy and Displacement in Colombia’s Civil War. Ithaca: Cornell University Press. Steele, Abbey, Marianne Dahl, Havard Mokleiv Nygard and Michael Weintraub. 2021. “Wartime victims and forms of political engagement during post-conflict transitions.” Working Paper. University of Amsterdam. Tellez, Juan F. 2019a. “Peace agreement design and public support for peace: Evidence from Colombia.” Journal of Peace Research 56(6):827–844. Tellez, Juan Fernando. 2019b. “Worlds Apart: Conflict Exposure and Preferences for Peace.” Journal of Conflict Resolution 63:1053–1076. Thapa, Suraj Bahadur and Edvard Hauff. 2005. “Psychological distress among displaced persons during an armed conflict in Nepal.” Social psychiatry and psychiatric epidemiology 40(8):672–679. Torres, Jacqueline M and Joan A. Casey. 2017. “The centrality of social ties to climate migration and mental health.” BMC Public Health 17(1):1–10. UNHCR. 2020. “Figures at a glance.” https://www.unhcr.org/en-us/figures-at-a-glance.html. Accessed 01-09-2020. 28 Vargas Reina, Jenniffer. 2021. “Coalitions for land grabbing in wartime: State, paramilitaries and elites in Colombia.” The Journal of Peasant Studies pp. 1–21. Villamil, Francisco. 2020. “Mobilizing memories: The social conditions of the long-term impact of victimization.” Journal of Peace Research 3:399–416. Vinck, Patrick and Phuong Pham. 2009. “PeaceBuilding and Displacement in Northern Uganda: A Cross-sectional Study of Intentions to Move and Attitudes towards Former Combatants.” Refugee Survey Quarterly 28(1):59–77. Voytas, Elsa. 2021. “What Can Reparations Do? the Political Consequences of Compensation.”. OSF Preprints. At doi:10.31219/osf.io/akz26. Vélez, María Alejandra, Carlos Andres Trujillo, Lina Moros and Clemente Forero. 2016. “Prosocial Behavior and Subjective Insecurity in Violent Contexts: Field Experiments.” PLOS ONE 11(7):1–23. Wood, Elisabeth Jean. 2008. “The Social Processes of Civil War: The Wartime Transformation of Social Networks.” Annual Review of Political Science 11:539–561. 29 Appendix 1 SURVEY DATA: SOURCE OF DATA, SAMPLING FRAME Data used in the rural household analysis come from a survey collected in March and June of 2017 as part of USAID’s performance evaluation of the Land and Rural Development Programme (LRDP), which had been operating in the following departments/regions: Cauca, Tolima, Meta, Cesar, and Montes de Maria. The LRDP is a multi-faceted development program that include ef- forts to address challenges in rural areas bearing on land restitution for displaced victims of the war, the formalization of land tenure, and broader rural development. Program interventions varied, but largely focused on improving bureaucratic capacity and cooperation among govern- ment agencies that interact with these policy areas. More details of the program are available in Section 2 of [citation removed to ensure blind review]. The survey data used in this paper come from the quantitative portion of the performance evaluation of the LRDP, which aimed to assess the effectiveness of the LRDP at the halfway point of its lifespan. As part of this evaluation, approximately 1,548 rural households were selected across 50 municipalities that included locations where the LRDP had been operating and other, comparable municipalities. The decision-making process underlying where the LRDP chose to operate involved a number of factors, but largely centered on need related to restitution and formalization, conflict-exposure, low development, and rurality. In terms of selecting munici- palities for evaluation, a subset of the municipalities where the LRDP had been active were ran- domly selected, and then using data on municipal characteristics, a set of comparable munici- palities were identified through matching. The result is a set of municipalities characterized by high rurality, high exposure to conflict, and a high incidence of poverty. More details available in Annex 1 of [citation removed to ensure blind review]. Moving from the municipality to the household, rural parts of the municipality were em- phasized for data collection, in particular veredas (or ‘villages’) within municipalities. Village se- lection was largely random, though in some municipalities where the LRDP was active specific 1 villages were chosen because of their relevance to the LRDP program. Roughly 15 households in each village were surveyed. Enumerators performed random-walks through the village until the specified number of surveys were completed in each village. In terms of sample diversity, the sample frame involved quotas on gender and age, though not for ethnicity. That said, many respondents do belong to minority ethnic groups in Colombia, especially Afro-Colombian (10% of sample) and indigenous (13% of sample). More details available in Annex 1 of [citation re- moved to ensure blind review]. Once in the field, there were a few unexpected constraints. In a small number of cases, cost constraints prevented traveling to very remote or poorly connected villages. There were also a few instances where it was deemed too dangerous to survey a particular village. In these in- stances, a new village was chosen at random that had a high chance of having similar charac- teristics. These constraints are not out of the ordinary for survey work in rural Colombia. More details on page 239 of [citation removed to ensure blind review]. The resulting data captures a random sample (though I note a few constraints, above) of households in rural and highly conflict-afflicted regions of the country. Given the sampling frame emphasis on rurality and conflict-exposure, the resulting sample is not intended to be nationally representative of all households, nor of rural households. 2 Figure A.1: Location of surveys, including LRDP status. 3 Figures A.2 and A.3 present distributions of sample characteristics in the survey and from the last two years of the rural portion of data from Colombia collected by LAPOP, respectively. In contrast to the survey, the item used to measure displacement in LAPOP asks respondents whether a family member was displaced during the conflict, so direct comparison is not perfect. Overall, there is a fair amount of overlap in these distributions, though the sample (by design) captures a much larger proportion of people who were displaced. 4 Household size Respondent age Years of education 0 5 10 15 20 25 50 75 100 0 5 10 15 20 Displaced? Marital status Widowed NA Separated Permanent partner Yes Never married Married No Engaged Divorced 0% 10% 20% 30% 40% 50% 0% 10% 20% 30% 40% Race Sex NA White Male Other Mulatto Mestizo Female Indigenous Black 0% 10% 20% 30% 0% 10% 20% 30% 40% 50% Percent of sample Figure A.2: Sample respondent characteristics from survey. 5 Family member displaced Gender Race White NA Woman Other Yes Mestizo Indigenous Man No Black 0% 20% 40% 60% 0% 10% 20% 30% 40% 50% 0% 10% 20% 30% 40% 50% Percent of sample Age Number of kids Years of education 25 50 75 0 5 10 15 0 5 10 15 20 Figure A.3: Respondent characteristics from rural module of LAPOP Colombia (2018 and 2019, combined). 6 1.1 Implementation, data quality, and ethical considerations Participants took the survey in or near their home. An informed consent form was read to par- ticipants outlining the purpose of the study, steps taken to protect the confidentiality of their responses, and potential risks faced by participating. Participants were assured that their pro- tection was voluntary. They were also informed that their responses would be shared through public publication in a way that protected their identity. Participants received no benefit or ex- pectation of participation and were instructed that they could end the interview at any time. Enumerators administered the survey using tablets, and the interview lasted approximately 45 minutes to an hour. Participants who agreed to participate gave consent orally, and consent was recorded in the tablet. The survey included modules on topics in restitution, land tenure, land formalization policy, and other questions bearing on rural development. The full instrument is available in Appendix 3 of [citation removed to ensure blind review]. In terms of data quality, enumerators received four days of training on best practices in sur- vey collection, including sampling methodology, and electronic data collection using platforms and Survey CTO (the survey platform selected for electronic data collection). On the field, enu- merators received spot-checks by supervisors, phone verification by the survey team of answers with a random subset of participants, and data quality checks by the performance evaluation team (e.g., plausible length of time answering questions; geo-location while collecting survey). More details on page 102 of [citation removed to ensure blind review]. Data collection received IRB approval from XXX University in March 2017. I have no reason to expect the survey adversely affected participants, though I recognize the difficulty people can face in discussing painful memories. Given that the survey instrument touches on sensitive sub- jects, the survey procedure involved ensuring that others were not present during the interview who might overhear, to the best extent possible. Enumerators also received training on ethics in survey research and research with human subjects, which explicitly included training on not ‘pushing’ participants for answers on questions deemed sensitive. Participants were not paid or compelled to participate in the survey. The survey also did not differentially benefit or harm any particular group. Finally, given the setting, some participants can be considered vulnerable or marginalized. As a result, great care was taken to remove identifying information from the data and ensure respondent anonymity both in data storage and analysis. 2 ADDITONAL ANALYSIS 7 Table A.1: Motivation for including variable in covariate balancing procedure. For simplicity, out- come variables are discussed as ‘outcomes’, which includes beliefs, preferences, sense of secu- rity, etc. Covariate Reasoning White/non-white Ethnic minorities more likely to be targeted for displace- ment (?); identity affects outcomes Sex (male) Men more likely to die, women often disproportionately affected (?); men and women often systematically differ in outcomes Age (years) Age shapes decision to flee; age shapes outcomes Household size Family size shapes decision to flee; family status affects outcomes % rural population Conflict clustered in rural areas; rurality shapes outcomes Population (logged) Size of town shapes decision to flee; size shapes out- comes Distance to capital Remote areas more likely to be targeted; remoteness shapes outcomes Municipal poverty index Poverty shapes decision to flee; poverty shapes outcomes Altitude (logged) Conflict clustered in mountainous terrain; rugged terrain shapes outcomes Municipal homicides (avg) Relative danger shapes decision to flee; danger/risk shapes outcomes AUC presence (prop. of years) Presence of AUC increases displacement; right-wing group shapes outcomes via targeting, unobserved loyal- ties FARC presence (prop. of years) Presence of FARC increases displacement; left-wing groups shape outcomes via targeting, unobserved loyal- ties ELN presence (prop. of years) Presence of ELN increases displacement; left-wing groups shape outcomes via targeting, unobserved loyalties Number of people displaced Places with high displacement might be fundamentally (avg) different Other victimization experiences Victims are more likely to be displaced; victimization shapes outcomes LRDP programming area LRDP shapes sample, influencing both displacement and outcomes literacy Human capital shapes decision to flee and outcomes, lit- eracy is arguably pre-treatment 8 Covariate Balancing Propensity Score Entropy balancing Propensity score (glm) Weighting approach: Energy balancing Propensity score (gbm) attends community meetings attends farmer meetings attends political party meetings attends women's meetings belief FARC can demobilize belief in land reform belief in restitution 0.2 0.20 0.4 0.3 0.3 0.2 0.15 0.3 0.1 0.1 0.2 0.2 0.10 0.2 0.1 0.1 0.1 0.0 0.0 0.05 0.1 0.0 0.0 0.0 -0.1 0.0 0.00 -0.1 belief in titling belief restitution is fair ease of property transfer eviction risk favors collective action fear new groups feel safer now 0.04 0.0 0.05 0.0 0.3 0.04 0.09 0.2 0.02 -0.1 -0.1 0.2 0.03 0.06 0.1 0.00 0.02 -0.2 -0.2 -0.02 0.1 0.01 0.03 0.0 -0.3 -0.3 0.00 -0.4 0.0 0.00 -0.1 has enough food helped solve local problem income not enough interest in politics invested in property knows land victim law knows of LRU Coefficient estimate on displacement status 0.0 0.20 0.25 0.08 0.20 0.20 0.15 -0.2 0.10 0.15 0.04 0.15 0.15 0.10 0.10 0.10 0.00 0.10 -0.4 0.05 0.05 0.05 0.05 -0.04 0.05 -0.6 0.00 0.00 0.00 0.00 0.00 land conflicts unlikely land help available land rights clear land rights protected less conflict future monthly income property is respected 0.1 0.2 0.2 0.05 0.0 0.15 0.1 0.2 0.00 0.10 0.1 -0.3 0.0 0.0 0.1 -0.05 0.05 0.0 -0.10 -0.6 0.00 -0.1 -0.1 0.0 -0.9 -0.05 -0.15 -0.1 -0.2 -0.10 property safe from govt property safe from outsiders risk aversion 1 risk aversion 2 supports expropriation supports land invasions supports peace 0.1 0.1 0.2 0.3 0.2 0.05 0.4 0.0 0.0 0.00 0.2 0.1 0.0 -0.05 0.1 0.0 0.2 -0.1 -0.1 -0.2 -0.10 -0.1 -0.2 -0.2 -0.15 0.0 0.0 -0.4 trust in courts trust in INCODER trust in MARD trust in mayor trust in personeria trust in police trust in state 0.0 0.1 0.0 0.1 0.1 0.1 0.0 0.2 -0.1 0.1 -0.1 0.0 0.0 0.0 -0.1 -0.2 -0.2 0.0 -0.1 -0.1 -0.2 -0.3 -0.3 -0.1 -0.3 -0.1 -0.4 -0.2 -0.2 -0.4 -0.2 -0.4 -0.2 voted for Santos would pay more tax 0.15 0.10 0.10 0.05 0.05 0.00 0.00 Figure A.4: Displacement effect estimates across weighting algorithms. 9 Table A.2: Summary statistics of the qualities of the distributions of weights. Balance Status Coef. of variation MAD Entropy ESS (control) ESS (treated) Unweighted Treated 0.5 0.33 0.82 783 725 Weighted Control 0.46 0.29 0.72 648.77 580.48 Overall 0.48 0.31 0.77 NA NA 10 Cohesion Land Peace Would pay more tax Property safe from outsiders Supports peace Property safe from govt Helped solve local problem Less conflict future Property is respected Favors collective action Land rights protected Knows land victim law Attends women's meetings Land rights clear Feel safer now Attends political party meetings Land help available Fear new groups Attends farmer meetings Land conflicts unlikely Attends community meetings Belief farc can demobilize Knows of lru 0.0 0.1 0.2 0.3 -0.2 -0.1 0.0 0.1 0.2 -0.2 0.0 0.2 0.4 Politics Trust Welfare Trust in state Risk aversion 2 Supports land invasions Trust in police Risk aversion 1 Supports expropriation Trust in personeria Monthly income Trust in mayor Invested in property Interest in politics Trust in incoder Income not enough Did not vote Trust in courts Has enough food Belief in titling Eviction risk Belief restitution is fair Belief in state land reform Ease of property transfer 0.0 0.1 0.2 0.3 -0.4 -0.2 0.0 0.2 -0.8 -0.4 0.0 Coefficient estimate on displacement Figure A.5: Displacement effect estimates, truncating weights above the 99th percentile. 11 3 CONJOINT RESULTS 12 Table A.3: Conjoint experiment attributes and levels. Provision Type Attribute Levels FARC members don’t go to jail, Transitional Justice Retributive Justice Only human right violators go to jail, All FARC members go to jail Demobilized fighters cannot compete in elections, Political FARC Elections Demobilized fighters can compete in elections, Demobilized fighters can compete in elections and have 5 guaranteed seats in Congress No transfer, Land Land Distribution Small transfer of land to landless peasants, Large transfer of land to landless peasants Aerial fumigation, Drug Policy Drug Policy Reform Manual eradication, Cultivation substitution program 13 Table A.4: Test of heterogeneity (f-test) for conjoint experiment. Null model = attributes only; alternative model = attributes interacted with displacement status. Resid. Df Resid. Dev Df Deviance F Pr(>F) 8, 567 2, 086.793 8, 558 2, 083.007 9 3.786 1.728 0.077 14 Table A.5: . Conjoint results. Model interacts displacement status and profile attributes. Errors clustered at respondent-level. Model 1 Intercept 0.424*** (0.021) displaced? 0.043 (0.031) drugs (medium) -0.007 (0.018) drugs (high) -0.049*** (0.018) elections (medium) -0.065*** (0.018) elections (high) -0.095*** (0.018) land (medium) X displaced -0.025 (0.026) land (high) X displaced -0.064** (0.027) justice (medium) X displaced -0.044 (0.027) justice (high) X displaced 0.032 (0.028) drugs (medium) X displaced 0.003 (0.026) drugs (high) X displaced 0.008 (0.027) elections (medium) X displaced -0.029 (0.026) elections (high) X displaced -0.016 (0.026) N 8576 * p < 0.1, ** p < 0.05, *** p < 0.01 15 4 ENDORSEMENT EXPERIMENT Three endorsement experiments were administered to measure respondent support for the FARC. These experiments were administered out of concern that respondent fear, social desir- ability bias, and non-random refusal to participate would bias measures of rebel support. Policy questions were selected to be: a) plausible or existing policy questions in domestic Colombian politics; b) relevant to the rural context, to maximize probability that respondents have opin- ions or intuitions about them; c) policies that the FARC either has or would plausibly endorse. The mechanics of the endorsement experiment is as follows: a randomly-selected half of all respondents are asked to express their opinion toward a policy, which they are told has been endorsed by the FARC (treatment group). The other respondents are asked for their opinion, without the FARC endorsement (control group). Higher levels of support under the treatment condition is interpreted as evidence of support for the endorsing actor. The three policy questions are the following: • A recent proposal [TREATMENT: by the FARC] calls for shifting away from importing food- stuffs from foreign countries and instead producing food domestically, so that the major- ity of food consumed in the country is made by Colombians. How do you feel about this proposal? • A recent proposal [TREATMENT: by the FARC] calls for redistributing land from large land- holders, in order to give landless peasants greater access to land. The proposal would mean that anyone who owns very large tracts of land might be subject to having some of it taken away. How do you feel about this proposal? • A recent proposal [TREATMENT: by the FARC] calls for the legalization of coca cultivation in Colombia. This proposal would allow coca cultivators to sell coca legally, and has been argued will end the drug trade and the need for counter-narcotics efforts on the part of the state. How do you feel about this proposal? To analyze whether underlying support for the FARC varies substantially across displace- ment status, I simply interact the treatment indicator with an indicator of whether the respon- dent was displaced during the conflict. 16 Table A.6: Endorsement experiment results. Models interact FARC endorsement treatment with displacement status. Dependent variable: Agr. policy Coca policy Land policy (1) (2) (3) Endorse treatment X displaced 0.042 0.017 0.175 (0.098) (0.112) (0.117) Observations 1,500 1,456 1,484 ∗ Note: p<0.1; ∗∗ p<0.05; ∗∗∗ p<0.01 17 5 LASSO Outcome: respondent able to return 1.00 sensitivity 0.75 0.50 0.25 Out-of-sample AUC: 0.68 0.00 0.00 0.25 0.50 0.75 1.00 1 - specificity Outcome: respondent sought restitution 1.00 sensitivity 0.75 0.50 0.25 Out-of-sample AUC: 0.7 0.00 0.00 0.25 0.50 0.75 1.00 1 - specificity Figure A.6: Receiver operating characteristic curve for LASSO models, out-of-sample data. 18