Policy Research Working Paper 10054 The Effects of Internally Displaced Peoples on Consumption and Inequality in Mali Jeremy Foltz Sakina Shibuya Social Sustainability and Inclusion Global Practice May 2022 Policy Research Working Paper 10054 Abstract A series of civil conflicts in Mali has generated more than hosting communes relative to non-IDP host communes. 346,000 internally displaced people (UNHCR, 2020). This study also finds some partial evidence of increasing This study estimates the effect of conflict-generated inter- consumption at the household level although inequality nal displacement on consumption, poverty, and inequality and poverty at the commune level remain the same. The in host communities. Using comprehensive nationwide evidence suggests a fairly successful hosting and aid process household survey data this study finds that wealth at the in Mali for IDPs in terms of mitigating economic disrup- commune and household level is non-decreasing in IDP tion for host communities. 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 jdfoltz@wisc.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 The Effects of Internally Displaced Peoples on Consumption and Inequality in Mali Jeremy Foltz∗ Sakina Shibuya∗ Keywords: Internal displacement, Conflicts, Inequality, Poverty JEL Codes: D74, I30, O15 ∗ Respectively, Professor and Ph.D. candidate, Department of Agricultural and Applied Economics, University of Wisconsin, Madison (Jdfoltz@wisc.edu, Sshibuya2@wisc.edu). We thank Massa Coulibaly, Nouhoum Traor´ e, and Aly Sanoh for help with accessing the data used in this work. We also thank anony- mous reviewers for their helpful comments. 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. 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. 1 Introduction In the last decade, Mali has become the epicenter of one of the world’s worst civil conflicts (ACLED, 2020), generating more than 346,000 internally displaced peoples (IDPs) (UN- HCR, 2020). Real and perceived inequalities between and within communities have helped fuel this conflict (Pezard & Shurkin, 2015). The plight of IDPs in Mali is severe, which has been recognized and they are actively aided by national and international organizations. Hoogeveen, Rossi, and Sansone (2019), which studies the dynamics of intentions to return after the early part of Mali’s conflict subsided, shows that some but not all IDPs intend to return home. IDPs fleeing this conflict and locating in other Malian towns potentially affect social cohesion by changing economic conditions including wealth, poverty, and inequality among the population in host communities. Policymakers dealing with an internal displace- ment crisis are often challenged to mitigate shocks caused by a sudden inflow of people on host populations, while providing adequate support for incoming displaced peoples. If IDPs foster increased inequality or negatively affect local wealth, then they may affect social co- hesion in ways that exacerbate the conflict. This work tries to answer how does the arrival of IDPs in host communities in Mali affect wealth, consumption, poverty, and inequality? We analyze the effects of IDP presence and numbers on consumption, poverty, and in- equality of host communities in Mali. We use comprehensive nationwide household survey data on over 36,000 households from 2013 - 2019 combined with data on IDP populations to estimate econometric models of the effects of IDP presence and numbers. Our modeling strategy uses multiple econometric methods, difference-in-difference, instrumental variable, and propensity score matching to provide the most robust results to concerns about the endogenous location choices of IDPs. A now burgeoning literature analyzes the economic impact of displaced peoples on host communities, see Verme and Schuettler (2020) for a systematic review. While the literature review finds a low probability of a decline in well-being for households in host communities, there is little evidence on what hosting displaced peoples does to inequality. A positive 1 effect of IDPs on inequality is potentially important, given that inequality between groups can lead to further conflicts (Østby, 2008). In addition, while this literature has studied many instances of refugee location there are relatively few studies on IDPs especially in low-income country settings such as Mali, Hoogeveen et al. (2019) and Sedova, Ludolph, and Talevi (2021) being some recent exceptions. While aid to refugees and IDPs has generally been shown to produce positive externalities on host community incomes and consumption (see e.g. Taylor et al. (2016)), it can also change inequality, especially between ethnic groups. This effect on inequality is not yet well studied in part because of the need for larger and longer-term data sets, which we partially solve in this study. How might the influx of IDPs affect host commune consumption and inequality over time and space in Mali? We hypothesize that an increase in IDPs increases average host community incomes and therefore consumption, but that this increase in income is not evenly distributed about the population. We further hypothesize that the heterogeneity in consumption benefits from IDPs might favor certain types of households (e.g., farmers) and disfavor others (livestock herders, landless). The results in this work show that, in the Malian context, IDP host communities and the households in them fare at least as well as other Malian communities and households that do not host IDPs. The results show positive effects of IDP hosting on household consumption, under some modeling assumptions, and zero effects on poverty and inequality at the community level. The results are suggestive of the benefits in IDP hosting communities being fairly evenly distributed across all residents, and not different for households with farming as their major occupation. These results provide important insights into under- researched questions about how the internal displacement caused by the Malian and broader Sahelian conflict has affected host communities. Understanding that with proper support IDP populations do not exacerbate inequality in host communities is an important piece for policymakers in deciding how to work with and create social cohesion in host communities. Our work complements other studies in this project that examine the effects of refugee 2 and IDP influx on host communities. While a number of the other studies (Aksoy & Ginn, 2021; Betts, Stierna, Omata, & Sterck, 2021; Pham et al., 2021; Zhou & Grossman, 2021) investigate the effect of hosting displaced people on host residents’ perceptions toward the displaced, we focus on the effects of IDP influx on communities’ economic outcomes: non- agricultural earnings, consumption, inequality, and poverty. Our work is most closely related to the work by Sedova et al. (2021) who also investigate the IDP effect on wealth and inequality in host communities in Nigeria. While our findings generally show a null effect of IDP presence on inequality, that work finds, in the context of Nigeria, that an IDP influx is associated with a decrease in the perception of wealth and equality. These contrasting effects suggest the importance of local context, institutions, and interventions for understanding how IDP presence affects wealth and inequality. 2 Context: Conflict and Internal Displacement in Mali Since its independence in 1960, Mali has had a series of rebellions in the north, primarily led by northerners, primarily from parts of the Tuareg ethnic group, unhappy with the government rule led by mostly southern Malians. The first three of these rebellions happened in 1963–64; 1990 – 96; 2006 – ’09; and were ended by peace accords and agreements, which were not effective at resolving the underlying problems and reducing the probability of future conflict (Nomikos, 2019; Pezard & Shurkin, 2015). The current civil conflict started in 2012 with a fourth northerner rebellion,1 which took on a new character with the insertion of international violent extremists who both had a stronger ability to fight and control territory and had a wider appeal beyond the ethnically determined communities through their ability to weaponize local disputes. A military coup d’etat in 2012, in part due to the government’s ineffective campaign against the northern rebellion, also exacerbated the conflict. This wider 1 While these rebellions take the form of northerners who oppose the central government and would like to secede from the country, they are by no means uniformly supported by all northerners. They are as much power struggles between the different groups and clans within the north of Mali as much as they are a conflict between north and south. 3 support for rebellion, rekindling of old rivalries, and dissatisfaction of the population with the status quo helped lead to the expansion of the conflict from northern to central Mali. In 2013, after northern rebels and their international violent extremist allies had con- quered much of northern Mali a French-led military intervention, Operation Barkhane, pushed them out of the area. Violent events spiked in 2013, especially in northern Mali as the French fought with violent extremists and their local allies. That violence subsided somewhat after 2013 due to the deployment of French troops and a UN peacekeeping mission MINUSMA (ACLED, 2020; Nomikos, 2019). The early wave of fighting in 2012/13 produced a wave of displacement, but much of it was short-lived as the French intervention pushed the violent extremist connected groups out of the major population centers of the north, Gao and Timbuktu. Hoogeveen et al. (2019) show great heterogeneity in those who decided to return to their hometowns from this initial wave of displacement. While the French-led intervention tamped down the immediate conflict, the violence started to spread to other countries (Niger and Burkina Faso) and other parts of Mali. We show in figure 1 the location of violent events in Mali as measured by ACLED for the period of this study. What is obvious is the spread of violence from being mostly in the north in 2014 to expanding into more parts of the country by 2017 and increasing in intensity (as seen by larger/darker circles on figure 1). Starting in 2014 the violence that starts to pick up in central Mali becomes in part driven by long-simmering issues between ethnic groups where the spread of violence is pushed by discontent with the status quo, deficiencies in national security, and justice institutions. The violent extremist groups provide the promise of personal security and rudimentary justice for rural residents who do not see the government as able to provide that. In opposition to the violent extremist groups and their local allies, many villages developed their own self- defense or communal militias as their own response to the inadequacy of government justice and security institutions. These self-defense groups along with the violent extremists have raised levels of inter-communal violence and have been a major source of fatalities, especially 4 in central Mali (Hoogeveen et al., 2019). For example, in the Mopti region, there have been increasingly violent conflicts between livestock herders and sedentary agriculturalist groups. A number of observers have claimed that the violent extremist movements have forged alliances with Peulh pastoralist groups eroux-B´ (Assanvo, Dakono, Th´ ıga, 2019; Benjaminsen & Ba, 2019; Diallo Aly, enoni, & Ma¨ 2017) and some have suggested that Peulh livestock sales partially finance these groups (Daniel, 2020). The central Mali conflict builds on long-standing conflicts between livestock herders and agriculturalists that in the past had been settled through customary means (Turner, 2004). Increases in the population, in the number of livestock, in the returns to agriculture, and the variability of the climate (Benjaminsen & Ba, 2019; Nomikos, 2019; Raleigh, 2010) as well as the ready availability of small and large arms in the area has meant that old disputes have led to increasing violence. By 2017, central Mali becomes the major locus of violence in Mali (ACLED, 2020). At the same time in the last few years, some conflict has spread into southern Mali including the northern parts of the Segou and Koulikoro regions, and the eastern Sikasso region, particularly areas close to the Burkina Faso border. While the conflict in Mali may have started out, and is often seen from outside, as a bilateral conflict between northerners and the Malian government, it ensnares Malians of all ethnicities and backgrounds. The conflict, especially within central Mali, is much more complicated in its make-up and often pits neighbors against each other. Such a conflict produces a wide range of displaced persons both within the country and abroad. As of the beginning of 2021, the conflict in Mali that started in 2012 has generated over 346,000 internally displaced peoples as well as well as more than 165,000 refugees, most of whom are in the countries immediately neighboring Mali: Mauritania, Burkina Faso, and Niger (UNHCR, 2021). The displacement of peoples in Mali has grown from 62,000 IDPs in 2015, dropping to 37,000 in 2016, before starting to rise exponentially from 2018 onwards, to its current level of more than a quarter of a million people. Figure 1 shows the evolution 5 of the location of IDPs in Mali from 2014 to 2019, the period of our study. The IDP situation in Mali appears to be fairly typical of IDPs throughout the world. The IDPs come from across all parts of the conflict zone and some areas near the conflict. From discussions with in-country workers, it appears that IDPs come from all different ethnic groups, rather than being concentrated in a single ethnic group. In many cases, IDPs from multiple ethnic groups will move to the same commune, sometimes because the fighting targeted multiple ethnic groups and other times because they are non-adherents to the conflict by their co-ethnics. IDP’s in Mali can be individuals or families: that are fleeing the fighting, that has had their crops or livestock or access to land taken away, that are ethnic minorities in their particular village, that are minorities in their region, or that have beliefs and ways of living different from the local governing polity, which may be led by violent extremists, local self- defense groups or the local government. Where violent extremists predominate they often chase out government functionaries such as school teachers. Such government functionaries may make some IDPs in Mali wealthier and better connected than might be typical elsewhere. The available evidence suggests that IDPs in Mali are well received within their host communities. A number of communities in the north (e.g. Sofara and Konna) that host IDPs have held successful financing appeals to the diasporas from their community in the capital, Bamako, and abroad to help house, feed, and clothe IDPs. We are unaware of any incidents in which IDPs themselves have become a source of conflict, but that always remains a possibility. In some central Malian areas, people have concentrated in villages of the same ethnicity for safety, especially where there are ethnically based self-defense groups, while in other areas IDPs of all ethnicities have moved to single locations, usually larger towns with more diverse populations. The international community has come together to provide large amounts of aid to Mali to alleviate this conflict-induced crisis and the plight of IDPs in the country. The overall budget for the humanitarian situation in Mali was $354 million in 2016 (UNOCHA-Mali, 2016) and 6 $262 million in 2018 (UNOCHA-Mali, 2018). This humanitarian operation has typically been funded by 40 - 50 different international humanitarian organizations and implemented by 150+ humanitarian organizations on the ground in Mali (UNOCHA-Mali, 2017). It has provided protection services, housing, nutrition, health, cash transfers, education, and other resources to IDPs as well as others within the conflict zone. At least a portion of the support provided to IDPs appears to be in the form of direct cash grants to recipients, which is likely to create positive economic spillover effects on local communities in the form of increased local purchases of goods and services. 7 Figure 1: The Number of Conflicts and IDP HHs per Commune Source: ACLED Notes: The large area with IDPs between 3000 8 to 5000 in 2016 is the commune of Tombouctou. Northern Mali is largely uninhabitable due to its acrid climate; thus, IDP are likely concentrated in the south of the commune. Figure 2: Ethnic Diversity in 2006 Source: The Demographic and Health Surveys, 2006 Notes: The geographical unit is commune. Light gray indicates NA. Ethnic diversity is shown as an inverted Herfindahl-Hirshmann index, where a larger value indicates a greater diversity. The large area in the north of the country is the commune of Tombouctou. Northern Mali is largely uninhabitable due to its acrid climate; thus, its population is concentrated in the south of the commune. 9 Figure 3: Inequality in Mali Notes: The total household consumption-based Gini coefficients are commune mean values over 2014-2019, and calculated using the EMOP data. Light gray indicates NA. A greater value indi- cates higher inequality. The large area in the north of the country is the commune of Tombouctou. Northern Mali is largely uninhabitable due to its acrid climate; thus, its population is concentrated in the south of the commune. 3 Theoretical Motivation 3.1 Literature Review The literature investigating the effects of forced displacement has grown rapidly in recent years with many of the recent studies focused on understanding the effects of a large inflow of Syrian refugees crossing international borders on the economies of host countries. Much of the attention has been given to middle-income host counties which have the smaller fiscal capacity to accommodate large inflows of refugees compared to high-income host countries, but receive less international humanitarian aid relative to low-income host countries such as West African countries (Verme & Schuettler, 2020). Most existing studies have focused their efforts on the effects of displaced peoples, particularly refugees, on labor and real estate market outcomes. For example, Depetris-Chauvin and Santos (2018) analyzes how IDPs in 10 Colombia affect rental markets and crime and Rozo and Sviatschi (2021) shows how Syrian refugees spike rental housing market prices in Jordan. Few studies have focused on internally displaced people and their effects on economic welfare in a low-income country context.2 We have identified several studies relevant to the effects of refugees on income, consumption, inequality, and poverty on host communities in a low-income country setting, but did not find many studies that investigate the effects of forced internal displacement on inequality and poverty. While refugees and IDPs both fall under the category of displaced peoples, there are differences that might mean they have different effects on the local economy. First, IDPs are more likely to be ethnically similar to their hosts than refugees would be. Second, IDPs are still within their own country, they may receive more care from the central government 3 because they are citizens of that country. Third, the host population may be more willing to accept IDPs because they are citizens of the same country and potentially of similar ethnic groups. Studies on the effects of forced displacement on household economic outcomes in the low-income country context tend to report positive effects of the influx of people on wealth outcomes for host communities. Most studies find a positive correlation between an inflow of refugees and host community household consumption and wealth. These studies typically attribute this positive correlation between displaced peoples and host community wealth to increased economic opportunities due to the inflow of new people including international aid workers and sometimes the creation of new infrastructure such as roads. An example of this work, Taylor et al. (2016), conducts a Monte Carlo simulation to estimate the effects of cash transfer to Congolese refugees in Rwanda on the local economies 2 Braun, Kramer, Kvasnicka, and Meier (2021) is the main economic paper we could find focused on economic welfare and community hosting of IDPs but its analysis of the effects of the post-WWII internal displacement of Germans from Eastern Europe to Western Germany is in an economic context that differs greatly from our context. 3 Note that this effect really depends on the nature of the conflict and the relationship of the displaced persons to the government. In some cases, such displacement during the Rwandan conflict, displaced peoples may be better accepted in a neighboring country than their own. 11 within the 10 km radius from three refugee camps. Their results indicate that an additional cash transfer recipient can increase the annual income of a local household by USD 205 to 253. Their study provides clear short-term evidence of significant positive economic externalities of refugees on host communities. The effects they measure of forced displacement combine both population effects and potential spill-over effects of the cash transfer program in place in those particular refugee camps. Loschmann, Bilgili, and Siegel (2019) study the same context, and document a shift away from agricultural production to wage employment in the host communities, which may or may not be beneficial in the long-term for the farm population. Other recent work investigates the effects of refugee presence on host community house- hold consumption, which may provide a more consistent way of depicting the economic experience of local households in host communities. Maystadt and Duranton (2019) investi- gate the effects of the inflow of refugees from Burundi and Rwanda due to the civil wars in the 90s, and estimates the effects on household consumption and poverty. They find that a higher refugee index is associated with higher per adult consumption, and a lower poverty rate. It is not clear, however, how the results from refugees should be expected to translate to IDPs. In addition, some of their results come from decades of residence by refugees and major infrastructural investments by aid agencies and governments, which may also not be relevant for the rapidly changing IDP situation in the Sahel. While there seems to be a general finding that the presence of displaced peoples (primarily refugees) has a positive effect on host community wealth, this effect is complicated by a potential increasing inflation on prices on food and housing, which may undermine some measures of wealth and consumption. For example, Alix-Garcia and Saah (2009) study the effects of the inflow of Brunudian and Rwandan refugees from 1993 to 1994 on western Tanzanian host communities. While they find positive effects on household wealth in rural areas, they find negative effects in urban areas. Combined with their presentation of evidence of increases in the prices of agricultural commodities consumed largely by local people, they 12 argue that the refugee inflow might have benefited local farmers while it might have had adversely affected urban Tanzanians due to price effects. Similarly, Alix-Garcia, Bartlett, and Saah (2012) suggests a framework for analyzing the effects of forced displacement on host communities, and emphasizes the importance of price dynamics in order to obtain a full picture of how forced displacement affects households in host communities. 3.2 Hypotheses The existing literature generally suggests that, if the effects of hosting IDPs follow the same pattern as those of refugee-hosting communities, one should expect to see positive relationships between an inflow of IDPs and the economic well-being of households in host communities. There are two channels through with an introduction of IDPs can affect host communities’ economies, as suggested by Verme and Schuettler (2020). First, a sudden inflow of IDPs can directly affect host communities by increasing their population and labor force levels. This mechanical effect can lead to a second effect with fiscal implications, when, in the absence of outside or national level interventions, host communities may have to internalize fiscal burdens to accommodate IDPs by providing direct assistance and expanding public services. In the Malian context, a large-scale humanitarian assistance program came from various international organizations, along with the help from the diaspora and the central government in Bamako. All of this suggests that there would likely be some positive economic shock along with the introduction of IDPs. This study asks: How does the influx of IDPs affect host commune consumption, poverty, and inequality over time and space in Mali? Based on the literature, we hypothesize that an increase in IDPs increases average host community incomes and consumption, but that this increase is not evenly distributed about the population. We further hypothesize that the heterogeneity in benefits from IDPs will favor certain types of households (male-headed households, farmers) and disfavor others. Finally, we hypothesize that these inequalities in consumption may widen as IDP numbers increase in a community. 13 4 Research Design In order to test our hypotheses about the potential effects of IDPs on household and village level consumption and inequality we turn to an empirical estimation using data from 2011 - 2019, spanning the beginning, early, and middle periods of the current conflict in Mali. Our data provides fine-grained observations on households across more than three-quarters of the communes in Mali. In terms of methods, we take a broad-brush view of the appropriate methods, presenting different ways to measure the outcomes of interest-based on different assumptions. The point of this multi-method technique is to provide the most robust under- standing of the outcomes across different issues of data quality and data generating processes. We first present the data, then the empirical strategy, the details of which are more fully outlined in the online appendix. 4.1 Data This work merges three main data sources on household consumption and IDPs to capture their effects on household and village consumption, poverty, and inequality. These are: a comprehensive household income and consumption data set (EMOP) collected by the same unit responsible for Mali’s Living Standard Measurement Survey (LSMS) data collection, data from the International Organization for Migration (IOM) on IDP movements, and data from the Demographic and Health Surveys (DHS). In addition, we access Malian census data and use conflict data from the Armed Conflict Location and Event Data Project (ACLED). ete Modulaire et Permanente aupr` We use eight years (2011, 2013-2019) of Mali’s Enquˆ es enages (EMOP), a nationally representative household survey on employment, in- des M´ come, consumption, with indicators on demographics and well-being (Institut National de La Statistique, n.d.). This data set offers detailed documentation on household non-agricultural earnings and expenditures (consumption) which enable us to measure inequality and poverty. The EMOP data has 38,625 household observations in 607 of Mali’s 703 communes, with 14 coverage in 9 of Mali’s 10 regions. The 10th region, Kidal, is not included due to insecurity in the zone. The sample size is large enough to estimate commune-level inequality and 90% of the communes in the data set are sampled more than twice, giving us a panel at the commune level over time. Within a commune, the choice of household is random, such that at the household level the data represents a repeated-cross section. Furthermore, the EMOP data are largely representative at the commune level.4 Combined with the sampling strategy at the commune level, these data provide us with consistent data to estimate effects at the commune level including both urban and rural communes.5 The commune-level IDP data come from the Displacement Tracking Matrix (DTM) pro- vided by the IOM (IOM, n.d.). These data derive from approximately monthly surveys on the number of IDPs in host communities from 2014 to the present. With this data set, we are able to calculate the total yearly number of IDPs and households in each host community identified by the IOM, and also construct an indicator variable of IDP presence. While the data collection involves multiple layers of data validation, this data set will mostly reflect the number of IDPs recognized by regional authorities, NGO representatives, civil society organizations, who are potentially beneficiaries of humanitarian assistance, as the IOM relies on such local actors as major informants. Therefore, IDP households who temporarily live with relatives to escape violence without seeking assistance are unlikely to be counted in the data.6 The ethnic diversity measure we use as an instrument in some specifications is calculated 4 The average number of sampled households in the EMOP data is 15 households. There are, however, two communes out of the 370 communes surveyed in 2019 with only one household interviewed in each village. These villages are Didieni in the Djidieni arrondissement in the Koulikoro region, and Niantaga in the M’pessoba arrondissement in Sikasso region. 5 We have tested aggregating the data to the arrondissement level and find similar effects as reported here, though they have less precision because our measures are necessarily less precise at a higher level of aggregation. The EMOP data we use show 46% of the households as urban, which almost exactly matches national statistics on urbanization rates, suggesting that our data are broadly representative of both urban and rural areas. 6 This implies that our measure of IDP presence and population should be considered as a measure of formally recognized IDPs who are likely receiving services. To the extent that there might be large numbers of IDPs in communes who are not receiving services, our data do not measure them. The available evidence that we can glean from the EMOP survey suggests that the number of people who fit such a category is very small. 15 e, Traor´ with the self-reported ethnicity data from the 2006 DHS (Samak´ el´ e, Ba, Demb´ e, & Diop, 2006) as an inverted Herfindahl-Hirschman Index (HHI). The HHI, first developed to study levels of market competition, is commonly used in the literature to measure ethnic fractionalization indices. It is simply an inverse of the sum of the squares of the share of each ethnic group per population in each commune.7 Figure 2 shows the distribution of the index across the country for communes where we have DHS survey data. In the figure a greater index value and darker color indicates a greater level of ethnic diversity. Additionally, we gather the 2009 commune population from the Malian census reported on the Wikipedia pages (Wikipedia, n.d.).8 Lastly, all maps are made with the administrative geographical boundaries made available by the Database of Global Administrative Areas (University of California Berkely, n.d.). For outcomes measures, we use per-capita household non-agricultural income and con- sumption from EMOP and then calculate commune-level poverty measures and the Gini coefficient as a measure of inequality from the consumption data. One important caveat on the EMOP income data is that it captures non-agricultural household income rather than total household income. The EMOP questionnaires are structured in a way that is more suitable to collect data on income with stable flows. We believe that this likely ignores most agricultural income. In fact, about a quarter of observations on household income is reported to be zero, while there is no report on zero consumption. Therefore, we interpret the EMOP income data to be reflective of non-agricultural household income, rather than total household income. 7 Please refer to Equation 5 in the online appendix for details on the creation of the index. 8 The Wikipedia page offer both circle- and commune-level population counts. We checked its internal consistency by comparing each circle-level value to a sum of commune-level values for each circle. After we verified internal consistency, we took the sum of all the commune population values to obtain the country population value, and compared it to Mali’s population count available on the World Bank’s DataBank. The data aggregated up accurately and we concluded that the data posted on the Wikipedia page is the valid, true data. 16 4.2 Descriptive statistics We now turn to descriptive statistics of IDPs and wealth levels in IDP and non-IDP com- munes. The number of IDP host communes fluctuates over time in our sample time period as shown in Figure 4a, which shows the share of IDP host communes to the total of 703 communes in Mali. We see a large spike in 2016, which corresponds to the increase in the number of IDP households shown in Figure 1. The IDP host communes in our analysis sample are also diverse in terms of the number of IDP households they host (Figure 4b). Figure 4: Distribution of IDP Host Communes (b) Distribution of IDP Communes across IDP (a) Share of IDP Host Communes in Mali Population Levels Notes: There are total 703 communes in Mali. Figures 5 and 6 present yearly trends for our outcome variables: non-agricultural income, consumption, poverty, and inequality. On average across all communes, IDP communes have fewer poor people than non-IDP communes. In addition, inequality based on both household and per capita consumption is higher in IDP communes than in non-IDP communes. House- holds in the IDP and non-IDP communes seem to earn a similar level of non-agricultural income on average, except the non-IDP trend is a lot more volatile. Lastly, at least on the raw descriptive level, households in the IDP communes generally consume less on average 17 than those in the non-IDP communes. Figure 5: Trends of the Outcome Variables (Commune Level) (a) Consumption-based Gini (b) Per capita Consumption-based Gini (c) Poverty Headcount Ratio (d) Poverty Gap 18 Figure 6: Trends of the Outcome Variables (HH Level) (a) Non-agricultural HH Income (b) Total HH Consumption Tables 1 and 2 in the online appendix describe other household and commune character- istics, which we use as control variables in our econometric analysis. In our analysis sample, 8% of the households have female household heads, 2% are Christian, 39% have literate household heads, and the mean age of households heads is 49 years old. Among households, they have an average of 6 members, 27% are polygamous, 47% have farming as a primary occupation, while only 0.4% are livestock herders. 46% of the sample households live in urban areas. Lastly, 78% of the sample households have total household consumption under the national poverty line (INS, 2017). In terms of commune-level characteristics, the sample communes are, on average, fairly equal in terms of Gini coefficients for average household and per capita consumption: 0.22 and 0.2, respectively. Meanwhile, the average commune-level consumption-based poverty headcount ratio and gap index are 0.86 and 0.46. Our sample communes are broadly equally poor across households. The average commune population in 2009 was 25,259 with a stan- dard deviation of 28,018, which says that communes vary greatly in population size. About 16% of the sample communes host IDPs. These host communes on average accommodate 19 170 IDP households and about 885 IDPs. However, as shown in Figure 4b the number of IDPs hosted varies greatly between host communes with a standard deviation of 1,765 for IDP households, and 8,438 individuals. 4.3 Empirical Strategy and Identification Estimating the effects of IDP presence on host communes’ economic outcomes is made dif- ficult by the possibility that communes with and without IDPs may differ in fundamental and potentially unobservable ways.9 This complicates our estimation because we want to attribute the observed differences in consumption, inequality, and poverty to the difference in IDP presence after controlling for observable characteristics. If communes with and with- out IDPs are different in ways that are unobservable in our data and uncorrelated with control variables, for example, the level of hospitality and tolerance toward outsiders, that complicates the estimation and potentially the validity of our inference from the results. In particular, we are concerned about being able to control for: unobservable differences in the initial conditions in IDP and non-IDP villages, IDPs choosing wealthier or more equal villages (endogenous selection), and other endogenous processes or reverse causality between IDP presents and household or village level wealth outcomes. In order to be able to link the observed difference in the economic outcomes to IDP presence, we use three types of econometric tools: difference-in-difference (DID), instrumen- tal variable (IV), and propensity score matching (PSM). Each of these common estimation techniques helps address a different potential type of bias in our estimate and each one comes with different sets of assumptions that might be more or less valid given the empirical setting in Mali and the data we can access. Taken together these estimates provide a broad picture of the relationship between IDPs and wealth in Mali. For policymakers, they provide estimates on all the key outcome variables of interest, although in some cases the estimates are weak or based on strong assumptions. The aim of this paper is to present a broad view 9 For a technical discussion of our identification strategy, please refer to the online appendix. 20 of the relationship between these factors.10 The first econometric approach we use is the Difference in Differences (DID) estimation approach, which compares how households or communes with and without IDPs move from initial conditions before IDPs arrive to after their arrival. Like the matching methods, this method assumes that once one has controlled for observable differences, communes or households would have followed similar paths, typically called “parallel trends”, had IDPs not arrived in their communes. This approach is particularly helpful in controlling for the initial differences between IDP and non-IDP communes before a sequence of conflicts had occurred in 2013 which have triggered the mass internal displacement we observe in Mali today. This method is only valid where we have strong evidence of parallel trends, and where we do not, such as for consumption levels, we do not report results. Additionally, given IDPs arrive in various years in host communes, we implement our DID estimation using the sequential treatment approach suggested by Callaway and Sant’Anna (2020), which accounts for staggered treatments as is the case here. In this analysis, we estimate the effect of IDP for groups of communes that have received IDPs for the first times in 2014 and 2016. In addition to estimating the IDP effect on each treatment group, we also present the overall effect. The second approach is the instrumental variable (IV) approach. In this method, to control for potential endogeneity of the placement of IDPs in communes, we estimate a first stage that uses an instrument, correlated with IDP location but uncorrelated with our main outcomes of interest, to “clean” the IDP location variable of the endogeneity of location choice. We use ethnic diversity from the previous decade as the instrumental variable to help identify the variation in IDP presence.11 This IV technique would be valid under the assumption that more diverse places would be more attractive and hospitable to IDPs, 10 In the results and policy implications section, we base our analysis on the results we believe are most robust across models and assumptions. 11 We also tested using conflict incidence within geographic bands of the communes in our data set as an instrument. These tests showed that conflict incidence did not pass the first stage tests and so are not presented here. 21 but that diversity itself a decade ago would not change current economic activity. This methodology is particularly useful in mitigating potential bias due to the possibility that the economic differences between IDP and non-IDP communes actually attract IDPs, rather than the differences being caused by IDP presence. This method is only valid where we have a strong first-stage relationship between ethnic diversity and IDP presence and numbers, and we only report it for those cases. The third approach is Propensity Score Matching (PSM) estimation, which seeks to match communes or households between IDP hosting communes and non-hosting communes. This method allows us to construct counterfactuals of communes and households that are very similar in characteristics to the actual IDP communes. Rather than using the full analysis sample, we construct two groups of IDP and non-IDP communes whose only difference is in the presence of IDPs, and compare them. Given the panel nature of the commune- level data, we implement an approach suggested by Imai, Kim, and Wang (2020). The PSM technique works well for the commune level estimates where we have good successes matching observable household characteristics between IDP and non-IDP communes (Figure 14), but does not work well at the household level since the same households are not sampled every round of the survey. In this exercise, we identify commune observations similar in characteristics to the actual IDP communes in the year of the event and the year before. 5 Results We now present the results from our regression analysis in three groups. First, we show the estimated results of IDP presence on economic outcomes in host communes using the binary indicator of IDP presence as the main explanatory variable. Second, we present the effects of IDPs at the intensive margin with the IDP population in communes as the main explanatory variable. Third, we investigate the possible existence of heterogeneous effects of IDP presence at the household level for different types of households. We use all three 22 different estimation approaches for estimates of IDP presence, only IV for IDP population, and only DID for the heterogeneity estimates. The conventional propensity score matching does not achieve the covariates balance necessary in the household level analysis, so we do not include it. 5.1 Effects of IDP presence on host communes We first discuss the estimation results from our analysis in which we represent the existence of IDP households with a simple binary variable. This section presents estimates from the DID, IV, and the PSM estimation approaches. First, we present the DID results. We conduct the DID analysis on inequality, and poverty at the commune level, and on non-agricultural income, and total household consumption at the household level. Figure 12 and 13 in the online appendix show the parallel trend assumption holds in each of the timing groups. The DID estimates show mixed effects on non-agricultural income and total household expenditures (consumption). We observe a statistically meaningful positive correlation be- tween IDP presence and per capita consumption at the household level analysis (Figure 7). This indicates that households residing in the IDP communes have on average higher expen- ditures. We observe, however, no statistically significant relationship between IDP presence and either household consumption or non-agricultural income. In both cases, the estimates produce precisely estimated zeros. Figure 8 presents the DID estimation result of the effects of hosting IDPs on inequality and consumption.12 The graphs show that at the commune level the existence of IDP households has no statistically meaningful effect on any of the measures of inequality and poverty. 12 See Table 4 in the online appendix for the regression coefficients and model statistics. 23 Figure 7: DID Estimates of The IDP Effects (Household Level) (a) Non-ag. Income (b) Consumption (c) Consumption per capita Notes: Estimates plotted above are from the DID estimation at the household levels based on Call- away and Sant’Anna (2020). The bars represent the 95% confidence intervals. The corresponding table 3 is presented in the appendix. Robust cluster SE at the household level. 24 Figure 8: DID Estimates of The IDP Effects (Commune Level) (a) Consumption Gini (b) Per Capital Consumption Gini (c) Poverty Share (d) Poverty Gap Notes: Estimates plotted above are from the DID estimation at the communes levels based on Callaway and Sant’Anna (2020). The bars represent the 95% confidence intervals. The corre- sponding table 4 is presented in the appendix. Robust cluster SE at the commune level. Next, we turn to our findings from the IV analysis, which allows us to estimate effects for more outcome variables than is possible with DID, and also better account for potential reverse causality. To estimate the effect of IDP presence using ethnic diversity in 2006 as an instrument, we must first verify that the instrumental variable has a sufficiently strong 25 correlation with IDP presence.13 The results shown in the online appendix of strong first stages leads us to conclude that 2006 ethnic diversity can be used as an instrument to the IDP presence indicator for the commune- and household-level estimation under the assumption that lagged ethnic diversity does not have a direct effect on wealth outcomes post-2014. Figure 9a presents the results from the IV estimation of the household-level regressions of IDP presence on non-agricultural household income, and total household consumption. The results show a statistically significant positive relationship on household consumption. More specifically, they suggest that when a commune hosts IDP households, total household expenditures on average increases by 25.4%. Such a level of consumption change would be consistent with the scale of effects found for refugees by Taylor et al. (2016). There is, however, no statistically significant evidence in the IV regressions that IDP presence has an effect on non-agricultural household income. Figure 9b presents the results from the IV estimation of commune-level regressions of IDP presence on inequality and poverty. The results show no statistically significant relationships between IDP presence, and host community inequality, and poverty. Note that the IDP coefficients on all the inequality and poverty measures are pretty close to 0 or negative with substantially larger imprecision. These results suggest that the true effect of IDP presence on poverty and inequality may be close to 0, consistent with our finding from the DID analysis. 13 Columns 1 and 2 of Table 7 in the online appendix show the results from the OLS estimation of re- gressions of IDP presence on ethnic diversity. The column 1 estimate at the commune level indicates that ethnic diversity is not meaningfully correlated with the number of IDP households. However, The column 2 shows that it has a strong statistical association with IDP presence with a F-statistic of 61. Both of the endogenous variables are strongly correlated with 2006 ethnic diversity at the household level as Columns 3 and 4 show. We have also considered the number of conflicts and fatalities within 50km outside of each commune’s boundary as an instrumental variable to IDP presence. As Table 8 in the online appendix shows no meaningful statistical relation between nearby conflicts and IDP presence. Thus, we only use the ethnic diversity index as an instrument in this section. 26 Figure 9: IV Estimates of The Effects of IDP Presence (a) Household Level (b) Commune Level Notes: The IV estimates shown here have robust cluster standard errors at the commune level. Fixed effects included are region and year. The independent variable is a binary variable which is 1 if a commune has IDPs. The inner bar shows the 90% confidence interval, and the outer bar shows the 95% confidence interval. All models are instrumented with the ethnic diversity index which is an inverted Herfendalh-Hirshman index with greater values indicating greater diversity. The dependent variables are log total household consumption and income for the household-level analysis, and household and per capita consumption-based Gini; consumption-based poverty head count ratio; and consumption-based poverty gap index. PR stands for poverty headcount ratio, and PG for poverty gap index for the commune-level analysis. Covariates included are: if HHH female; if Christian; if HHH literate; HHH age; and HH size. A corresponding tables 9 and 10 are presented in the appendix. Third, we now bring our attention to the Propensity Score Matching estimation. A proper estimation of the effects of IDP presence is possible only when we can establish a counterfactual to the group of IDP host communes from the non-IDP units in the year contemporaneous to the event year, as well as the year before the event. To verify this, we check if our matching is successful by making sure that these groups of communes have similar characteristics. Figure 14 in the online online appendix plots the difference in standard deviation between the IDP and non-IDP groups on commune characteristics: the share of female household heads; the share of Christian household heads; the share of literate household heads, the average age of households; the share of household heads whose primary occupation is a farmer; the share of households living in urban areas; the ethnic diversity 27 index; and commune populations. The graph shows that the post-matching difference in most of the characteristics is close to zero in this time period, except for the share of female household heads. This result indicates that the counterfactual constructed through the matching procedure is reasonably similar to the IDP group, and can be used to conduct a proper comparison on the outcome variables. Figure 10 shows the results of the PSM estimation of the effects of IDP presence on consumption-based Gini coefficients, per capita consumption-based Gini coefficients, poverty ratio, and poverty gap. The result indicates that there are no statistically significant rela- tionships between hosting IDP households and changes in these outcome variables. Figure 10: PSM Estimates of The IDP Effects (Commune Level) Notes: Estimates plotted above are from the PSM estimation at commune level. The bars represent the 95% confidence inter- vals. The corresponding table 13 is presented in the appendix. Standard errors are bootstrapped. 5.2 Effects of the Number of IDP Households Now, we investigate whether IDP presence can have effects on household-level economic outcomes at an extensive margin by using the number of IDP households as the treatment 28 variable, rather than the binary indicator used above. For this estimation, we use an IV approach to test the effects of increasing numbers of IDPs on host community outcomes.14 We again start by checking the strength of our instrumental variable, ethnic diversity in 2006, by estimating its effect on the number of IDP households.15 We conclude that 2006 ethnic diversity can be used as an instrument for IDP populations at the household level, but not at the commune level. Figure 11 provides the results from the IV estimation of regressions of the household economic outcome on the IDP population with ethnic diversity in 2006 as the instrumental variable. The results demonstrate a positive and significant effect on total household con- sumption, while we observe a positive but insignificant result on non-agricultural income. The consumption result implies that an inflow of 1,000 new IDP households into a commu- nity would lead to a substantial 10% increase in total household consumption. The scale of the point estimate for non-agricultural income is even higher, but since it is insignificant we cannot conclude that the true value is different from zero.16 14 The estimation with DID and PSM approaches requires the explanatory variable to be a binary variable. 15 Columns 1 and 3 of Table 7 in the online appendix indicates that the correlation between ethnic diversity and IDP population is negative but insignificant at the commune level, while it is positive and significant at the 1% level at the household level with an F-statistic of 365. 16 Additionally, we have considered the number of conflicts within 50km outside of each commune’s bound- ary as an instrumental variable to the number of IDP households. As Table 8 in the online appendix again shows the statistical relation between nearby conflicts and the number of IDPs. Thus, we use these conflicts variables to instrument the number of IDPs in the commune-level estimation. The estimated results (Tables 11 and 12) shows generally null effects on inequality and poverty. While Columns 1 and 3 of Table 11 indicates statistical significance at the 10% level, the magnitudes of the effects are nearly zero. This further adds to our evidence that the influx of IDP may not necessarily impact the inequality and poverty of host communities. 29 Figure 11: IV Estimates of The Effects of IDP Population on Income and Consumption (HH Level) Notes: The IV estimates shown here have robust cluster standard errors at the commune level. Fixed effects included are region and year. The independent variable is the number of IDPs. The inner bar shows the 90% confidence interval, and the outer bar shows the 95% confidence interval. All models are instrumented with the ethnic diversity index which is an inverted Herfendalh- Hirshman index with greater values indicating greater diversity.. The dependent variables are log total household consumption and income. Covariates included are: if HHH female; if Christian; if HHH literate; HHH age; and HH size. A corresponding table 9 is presented in the appendix. 5.3 Heterogeneity We next explore the possible existence of heterogeneous effects of IDPs on the economic outcomes of different types of households within their host community using the DID esti- mation models. We test two types of heterogeneity, by gender of the household head and by principle occupation of the household head, in particular, whether they are farmers.17 Table 5 in the online appendix shows the results of the heterogeneity check by the gender 17 For this exercise, we run regression on each of the following sub-sample groups: only female house- hold heads, only male household head, only primary farming households, and only non-primary farming households, and compare the difference in the magnitudes of the coefficients with the Clogg test (Clogg, 1995). 30 of the household heads. We find that, the female coefficients are statistically significant on consumption and consumption per capita, the male coefficients are only significant for consumption per capita. When comparing these female and male coefficients, we learn that the gender difference is positive statistically significant for consumption (Column 2, Panel C). In other words, a female-headed household tends to consume more as a household than its male counterpart in an IDP host community. However, the statistical significance of this difference disappears in per capita consumption, potentially indicating that accounting for household size there may not be any difference between female- and male-headed household in IDP host communities. In our other heterogeneity analysis by household head job type, IDP presence has no statistically different affects on farming and non-farming households in either consumption or non-farm income as shows in Table 6 in the online appendix. Despite a literature that suggests there might be heterogeneity in effects of IDP presence between farmers and non-farm households, we do not find any meaningful differences across groups by job type in our sample.18 6 Policy and Program Implications In this work, we have estimated the effects of IDP presence on consumption, inequality, and poverty in Mali, a country with a burgeoning IDP presence due to the ongoing conflict in the Sahel. In contrast to a literature that has sometimes found negative effects, we find consistent evidence that IDP presence in Malian communes is either beneficial or at the very least not detrimental to economic outcomes for the local population across multiple different estimation techniques and modeling assumptions. Our results suggest that overall household consumption goes up on average and that poverty and inequality measures are stable. We do not find much of a differential effect of the scale of IDP populations, which is suggestive of the effects being driven by the UN or NGO presence in helping IDPs rather than the scale of the operation. 18 We were unable to test between agriculturalist and herder households due to small sample sizes. 31 The results presented here suggest a number of implications for the current IDP crisis in Mali. It appears overall that the UN, NGOs, and the Malian government working with IDPs in Mali have succeeded in providing resources to IDPs and IDP hosting villages in a way that modestly enhances average village level consumption as well as not exacerbating inequality or poverty within the hosting communes. The results generally do not suggest that IDPs would have a detrimental effect on social cohesion due to the economic effects from their presence. But one should be cautious since we are observing the effects of IDP presence when there is a large humanitarian operation to help them, and acknowledge that results might differ where humanitarian operations are of a smaller scale or unable to reach IDP populations due to the conflict. There are some limitations that are worth highlighting before we turn to policy and program implications. First, our measure of IDP presence comes from the UN and NGOs that work with IDPs and may miss IDPs who are not known of or registered with the UN or NGOs in the area. Our EMOP data do suggest that there are small numbers of people who might fit this category, but not of a scale likely to significantly bias our results. Nonetheless, the effects reported here should be considered effects of officially registered IDPs. Second, this work only measures the effects of IDP presence on non-IDP households, such that our measures of poverty inequality are only within the host community estimate and do not measure inequality between host community and IDP wealth. It is along this dimension that some of the potential effects of inequality on social cohesion might be most felt. Third, the results presented here are specific to the Malian cultural context, a country that has a long-standing cultural history of accepting and housing outsiders. Called “Diatiguiya” in the local language, Bambara, which translates as the keeper of hospitality, this feature of Malian society means that all members of a community are expected to provide hospitality to strangers. Other places without a tradition of Diatiguiya may not have the same results. Fourth, due to the inability to measure prices at a fine enough scale, the current analysis does not account for potential localized inflation in communes with IDPs that might bias 32 our consumption and wealth findings. The results we present here should be taken as one piece of information, rather than thorough evidence on the economic effects of IDPs on host communities, because of the data limitations described above. Given these limitations, we recommend that policy-makers and program-implementers allocate more resources toward collecting comprehensive data on IDPs, as well as on economic indicators, in Mali and other countries with IDPs. Doing so can 1) further our collective understanding of the relationships between IDPs and their host communities; and 2) provide knowledge to inform policy making for both IPDs and their host communities. We specifically recommend improving on the types of data collected. First, finding ways of quantifying and obtaining data on those IDPs who are not served by aid organizations as they, for instance, temporarily live with relatives or in slums in urban areas would provide a more comprehensive view of what happens to host communities’ economies when IDPs arrive in a large numbers. Second, following up on the pioneering work by Hoogeveen et al. (2019), tracking individual IDP households over time, for example every month for a year or over multiple years, to generate data on how they cope with displacement, and interact with their host communities. Such data can shed light both on the results presented here and on general social cohesion in the host communities. Third, providing support to the Malian statistical authority to collect more frequent price data at a granular geographical level will help researchers understand the macroeconomic effects of IDPs. An effort to collect these types of data ultimately will help policy-makers and program- implementers design better projects that support both IDPs and host communities. 33 References ACLED. (2020). Ten Conflicts to Worry about in 2020. (January). Retrieved from www.crisisgroup.org/connect%0Ahttps://foreignpolicy.com/2018/01/02/ 10-conflicts-to-watch-in-2018/ Aksoy, C. G., & Ginn, T. (2021). Host Country Policies and Attitudes toward Refugees: A Global Evaluation. Unpublished Working Paper. Commissioned as part of the “Prevent- ing Social Conflict and Promoting Social Cohesion in Forced Displacement Contexts” Series. Washington, DC: World Bank Group.. Alix-Garcia, J., Bartlett, A., & Saah, D. (2012). Displaced Populations, Humanitarian Assistance and Hosts: A Framework for Analyzing Impacts on Semi-urban Households. World Development , 40 (2), 373–386. Retrieved from http://dx.doi.org/10.1016/ j.worlddev.2011.06.002 doi: 10.1016/j.worlddev.2011.06.002 Alix-Garcia, J., & Saah, D. (2009). The effect of refugee inflows on host communities: Evi- dence from Tanzania. World Bank Economic Review , 24 (1), 148–170. Retrieved from https://academic.oup.com/wber/article/24/1/148/1734945?login=true doi: 10.1093/wber/lhp014 Assanvo, W., Dakono, B., Th´ enoni, L.-A., & Ma¨ eroux-B´ ıga, I. (2019). Violent extremism, organised crime and local conflicts in Liptako-Gourma (Vol. 2019; Tech. Rep. No. 26). Retrieved from https://issafrica.org/research/west-africa-report/violent -extremism-organised-crime-and-local-conflicts-in-liptako-gourma Benjaminsen, T. A., & Ba, B. (2019). Why do pastoralists in Mali join jihadist groups? A political ecological explanation. Journal of Peasant Studies , 46 (1), 1–20. Retrieved from https://doi.org/10.1080/03066150.2018.1474457 doi: 10.1080/03066150 .2018.1474457 Betts, A., Stierna, M. F., Omata, N., & Sterck, O. (2021). Social Cohesion and Refugee-Host Interactions : Evidence from East Africa. 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.(June). Braun, S. T., Kramer, A., Kvasnicka, M., & Meier, P. (2021). Local labor markets and the persistence of population shocks: Evidence from West Germany, 1939-1970. Journal of Economic Geography , 21 (2), 231–260. doi: 10.1093/jeg/lbaa013 Callaway, B., & Sant’Anna, P. H. (2020). Difference-in-Differences with multiple time periods. Journal of Econometrics (xxxx), 1–31. Retrieved from https://doi.org/ 10.1016/j.jeconom.2020.12.001 doi: 10.1016/j.jeconom.2020.12.001 Clogg, C. C. (1995). Statistical Methods for Comparing Regression Coefficients Be- 34 tween Models Author ( s ): Clifford C . Clogg , Eva Petkova and Adaman- tios Haritou Published by : The University of Chicago Press Stable URL : http://www.jstor.com/stable/2782277 Statistical Meth. , 100 (5), 1261–1293. es s’intensifient. Daniel, S. (2020, oct). Dans le centre du Mali, les combats entre groupes arm´ Retrieved from https://www.rfi.fr/fr/afrique/20200410-mali-centre-pays -les-combats-entre-groupes-arm{\unhbox\voidb@x\bgroup\let\unhbox\ voidb@x\setbox\@tempboxa\hbox{e\global\mathchardef\accent@spacefactor\ spacefactor}\let\begingroup\def{}\endgroup\relax\let\ignorespaces\ relax\accent19e\egroup\spacefactor\accent@spacefactor}s-sintensifient Depetris-Chauvin, E., & Santos, R. J. (2018). Unexpected guests: The impact of internal displacement inflows on rental prices in Colombian host cities. Journal of Develop- ment Economics , 134 (April), 289–309. Retrieved from https://doi.org/10.1016/ j.jdeveco.2018.05.006 doi: 10.1016/j.jdeveco.2018.05.006 Diallo Aly, O. (2017). Ethnic Clashes, Jihad, and Insecurity in Central Mali. Peace Review: A Journal of Social Justice , 29 , 299–306. Retrieved from https://doi.org/10.1080/ 10402659.2017.1344529 Hoogeveen, J. G., Rossi, M., & Sansone, D. (2019). Leaving, Staying or Coming Back? Migration Decisions During the Northern Mali Conflict. Journal of Development Stud- ies , 55 (10). Retrieved from http://www-wds.worldbank.org/external/default/ WDSContentServer/WDSP/IB/2013/12/18/000442464 20131218151209/Rendered/ PDF/793790JRN0Natu00Box0379850B00OUO090.pdf Imai, K., Kim, S. I., & Wang, E. (2020). Matching methods for causal inference. Retrieved from https://imai.fas.harvard.edu/research/files/tscs.pdf INS. (2017). EMOP 2017 (Tech. Rep.). Institut National de La Statistique. Retrieved from http://www.instat-mali.org/contenu/eq/ranuel17 eq.pdf Institut National de La Statistique. (n.d.). Enquˆ ete Modulaire et Permanente aupr` es des M´enages (EMOP). IOM. (n.d.). Displacement Tracking Matrix. International Organization for Migration. Retrieved from https://displacement.iom.int/mali ¨ & Siegel, M. (2019). Considering the benefits of hosting refugees: Loschmann, C., Bilgili, O., evidence of refugee camps influencing local labour market activity and economic wel- fare in Rwanda. IZA Journal of Development and Migration , 9 (1). Retrieved from https://izajodm.springeropen.com/articles/10.1186/s40176-018-0138-2 doi: 10.1186/s40176-018-0138-2 Maystadt, J. F., & Duranton, G. (2019). The development push of refugees: Evidence from Tanzania. Journal of Economic Geography , 19 (2), 299–334. doi: 10.1093/jeg/lby020 35 Nomikos, W. G. (2019). Mali Country Report. Risks from the EU’s South- ern Border (Vol. 5; Tech. Rep. No. 769886). Retrieved from https:// www.cidob.org/en/publications/publication series/project papers/ eu listco/mali country report risks from the eu s southern border Østby, G. (2008). Polarization, horizontal inequalities and violent civil conflict. Journal of Peace Research , 45 (2), 143–162. doi: 10.1177/0022343307087169 Pezard, S., & Shurkin, M. (2015). Achieving Peace in Northern Mali: Past Agreements, Local Conflicts, and the Prospects for a Durable Settlement. doi: 10.7249/rr892 Pham, P., O’Mealia, T., Wei, C., Bindu, K. K., Makoond, A., & Vinck, P. (2021). Host- ing New Neighbors: Perspective of Host Communities on Displacement and Social Cohesion. Unpublished Working Paper. Commissioned as part of the “Preventing So- cial Conflict and Promoting Social Cohesion in Forced Displacement Contexts” Series. Washington, DC: World Bank Group.. Raleigh, C. (2010). Political Marginalization, Climate Change, and Conflict in African Sahel States. International Studies Review , 12 (1), 69–86. Retrieved from https://www .jstor.org/stable/40730710?seq=1#metadata info tab contents doi: 10.1111/ j.1468-2486.2009.00913.x Rozo, S. V., & Sviatschi, M. (2021). Is a refugee crisis a housing crisis? Only if housing supply is unresponsive. Journal of Development Economics , 148 . doi: 10.1016/j.jdeveco.2020 .102563 Samak´ e, S., Traor´e, S. M., Ba, S., Demb´el´ ´ & Diop, M. (2006). Enquˆ e, E., ete D´emographique et de Sant´ e du Mali 2006 (Tech. Rep.). Calverton, Maryland, USA: Cellule de Planification et de Statistique du Minist` ere de la Sant´ e - CPS/MS/Mali, Direc- tion Nationale de la Statistique et de l’Informatique du Minist` ´ ere de l’Economie, de l’Industrie et du Commerce - DNSI/MEIC/Mali and Macro International. Retrieved from http://dhsprogram.com/pubs/pdf/FR199/FR199.pdf Sedova, B., Ludolph, L., & Talevi, M. (2021). Inequality and Security in the Aftermath of Internal Population Displacement Shocks : Evidence from Nigeria. Unpublished Working Paper. Commissioned as part of the “Preventing Social Conflict and Promot- ing Social Cohesion in Forced Displacement Contexts” Series. Washington, DC: World Bank Group., 1–53. Taylor, J. E., Filipski, M. J., Alloush, M., Gupta, A., Valdes, R. I. R., & Gonzalez-Estrada, E. (2016). Economic impact of refugees. Proceedings of the National Academy of Sciences of the United States of America , 113 (27), 7449–7453. doi: 10.1073/pnas.1604566113 Turner, M. D. (2004). Political ecology and the moral dimensions of ”resource conflicts”: The case of farmer-herder conflicts in the Sahel. Political Geography , 23 (7 SPEC.ISS.), 36 863–889. Retrieved from https://www.sciencedirect.com/science/article/pii/ S0962629804000654?casa token=cbRIVO0lxFoAAAAA:CRZd3AfCq3N2xXip8kmTvy0Cn bzFLHrYkfphridaPilCslfTLo2Gk3OivcLyE88qfhc9xOu doi: 10.1016/ j.polgeo.2004.05.009 UNHCR. (2020). External Operational Update UNHCR Sahel Crisis Response. (June), 1–12. UNHCR. (2021). Refugee statistics. Retrieved from https://www.unhcr.org/refugee -statistics/ University of California Berkely. (n.d.). Global Administrative Areas. Retrieved 2020-06-22, from https://gadm.org/download country v3.html UNOCHA-Mali. (2016). Plan de R´ eponse Humanitaire-2016 (Tech. Rep.). Bamako. Re- trieved from www.humanitarianresponse.info/en/operations/mali UNOCHA-Mali. (2017). Plan de R´ eponse Humanitaire-2017 (Tech. Rep.). Bamako. Re- trieved from www.humanitarianresponse.info/en/operations/mali UNOCHA-Mali. (2018). Plan de R´ eponse Humanitaire-2018 (Tech. Rep.). Bamako. Verme, P., & Schuettler, K. (2020). The Impact of Forced Displacement on Host Commu- nities. A Review of the Empirical Literature in Economics. Wikipedia. (n.d.). Communes of Mali. Retrieved 2021-02-01, from https://en.wikipedia .org/wiki/Communes of Mali Zhou, Y., & Grossman, G. (2021). When Refugee Exposure Improves Local Development and Public Goods Provision : Evidence from Uganda Can Greater Exposure to Refugee Settlements Improve Public Service Delivery and Development. Unpublished Working Paper. Commissioned as part of the “Preventing Social Conflict and Promoting So- cial Cohesion in Forced Displacement Contexts” Series. Washington, DC: World Bank Group., 1–35. 37