Policy Research Working Paper 10044 Livelihood Impacts of Refugees on Host Communities Evidence from Ethiopia Solomon Zena Walelign Soazic Elise Wang Sonne Ganesh Seshan Social Sustainability and Inclusion Global Practice May 2022 Policy Research Working Paper 10044 Abstract Most refugee hosting communities are characterized by show that refugee inflow brings substantial benefits to host high levels of poverty with precarious livelihood conditions, communities by creating significant jobs, in which people low access to public services, and underdeveloped infra- engage as secondary occupations, and triggers an increasing structure. While the unexpected inflow of refugees might demand for livestock products. Specifically, while no effect bring both constraints and opportunities for improving was found on diversification of activities such as a primary and maintaining local livelihoods in these communities, occupation and crop product sales, a 1 percent increase the understanding of these effects remains limited. Using a in refugee inflow leads to a 2.7 percent rise in diversifica- household level micro data set from a 2018 baseline survey tion of livelihood activities as a secondary occupation and of the Ethiopia Development Response to Displacement a 15.9 percent increase in the value of livestock product Impacts Project, this paper assesses the impact of refugee sales. These effects tend to be heterogeneous across refu- inflow on the livelihood strategies of host communities gee hosting regions and the gender of the household head: with respect to diversification and agricultural commer- negative effects were mainly observed in Gambella region, cialization. The endogeneity of refugee inflow is addressed which hosts the largest refugee population in the country, by exploiting differences in factors that influence refugee and male-headed households were more likely to benefit arrival in the host communities. Specifically, the analysis from the refugee presence for the whole sample. The paper uses potential refugee inflow as an instrument, which is identifies households’ increased engagement in different the product of population density and intensity of con- livelihood activities and access to markets as a potential flicts (number of fatalities per event) in the closest region mechanism for the observed effects. The findings add to of the origin country to the refugee camp weighted by the the growing literature on the socioeconomic impacts of distance of the refugee camp to the closest region. The refugee inflow on host communities by showing an overall paper also constructs an aggregate index to proxy house- positive effect on the livelihoods and welfare of receiving holds’ livelihood diversification strategies. The findings 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 swalelign@worldbank.org/solezena@googlemail.com, swangsonne@worldbank.org, and gseshan@worldbank.org. 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 Livelihood Impacts of Refugees on Host Communities: Evidence from Ethiopia Solomon Zena Walelign1, Soazic Elise Wang Sonne, Ganesh Seshan Keywords: Livelihood, activities, commercialization, refugees, host communities, diversification, Ethiopia JEL Codes: J21, J43, E23, F61, F22 1 Corresponding Author 1. Introduction We are in the midst of protracted refugee crises. According to the latest UNHCR trends report, at the end of 2020, 76 percent of refugees globally (15.7 million) were in a protracted situation (UNHCR 2021). 2 Most refugees reside in low-income countries, and more than eight of every 10 refugees (86 percent) live in countries within territories affected by acute food insecurity and malnutrition (UNHCR 2021a). Refugee receiving host communities also tend to be poor, experience precarious livelihood conditions and face many socio-economic challenges, such as low economic status, poor access to public services, and infrastructural development. For these communities, refugees might bring both challenges and benefits. On the one hand, refugees increase competition for natural resources (e.g., wood for energy, construction, land), public services and infrastructure (e.g., education, health, water supply), and economic opportunities (e.g., traditional livelihoods, labor employment). Refugee inflow may also affect the local market by mainly depressing wages and raising product prices (Vemuru et al. 2020). On the other hand, refugees might also bring benefits to the local communities by: (i) providing skilled and unskilled labor, potentially leading to the establishment of new firms and also improving the performance of existing firms; (ii) creating extra demand for both agricultural and non-agricultural products in the local economy, leading to further intensification and commercialization of livelihood activities; and (iii) attracting humanitarian assistance and, increasingly, development aid (Alix-Garcia et al. 2018; Maystadt and Verwimp 2014). The presence of refugees may also attract infrastructural development projects to host communities. While these changes affect people’s livelihoods – both negatively and positively – in host communities and while the overall effect would depend on the dominance of one effect over the other, our understanding of the effect of refugee inflow on people’s livelihoods in host communities is limited. Hence, this paper contributes to bridging this gap by assessing the impact of refugee inflow on the livelihood strategies of host communities. The paper focuses on two main livelihood strategies at a household level: i) the diversification of livelihood activities and ii) the degree of agricultural commercialization. We measure livelihood diversification using two main variables: the degree of diversification of activities as a primary occupation and the degree of diversification of activities as a secondary occupation. 3 The degree of agricultural commercialization is also measured using two variables: the value from the sale of crop and livestock products. We measure refugee inflow (presence) as the number of refugees (population) in the nearest refugee camp to the household location weighted by the household's inverted distance to the camp. The impact of refugee inflow on household livelihood strategies can be causal if there are no confounding factors that affect livelihoods in host communities when refugee inflow changes. This is unlikely as refugee flow and the location of refugee camps are not random (see e.g., Baez 2011). Refugee camps are often situated close to international borders, among others, to allow for easy repatriation of the refugees when stability is restored in their countries of origin. In addition, refugees often seek shelter in the nearest refugee camp once they arrive in the host country, which is arguably true in most hosting countries as refugees often travel on foot for 2 According to UNHCR, a protracted refugee situation is a situation in which at least 25,000 refugees from the same nationality have been in exile for at least five years in a given host country. 3 Diversification of activities is calculated using the inverse Simpson diversity index. In constructing the index, we considered both agricultural and non-agricultural livelihood activities. 2 hours (Ruiz and Vargas-Silva 2018). Hence, to identify the causal impact of refugee inflow on livelihood diversification and commercialization, we employ a two-stage least squares (2SLS) econometric specification strategy using potential refugee inflow as an instrument. Potential refugee inflow is constructed as the product of population density and intensity of conflicts (number of fatalities per event) in the closest region of the origin country to the refugee camp weighted by the inverted distance of the refugee camp to the closest region 4 (i.e., the shortest distance to the border between the refugee camp and the bordering country of origin). Similar (weighted) instruments have been used in the literature (e.g., Baez 2011; Fallah et al. 2019) and proved to be an appropriate instrument to study the socio-economic impact of refugees on host communities. Livelihood diversification and agricultural commercialization are the two main common strategies that people in low-income countries adopt to improve or maintain their livelihood and welfare. Given the prevailing under-developed insurance market in the event of shocks, households tend to pursue several income generating activities. However, potential barriers such as low asset endowment hinder households' successful livelihood diversification (Ellis 2000; Martin and Lorenzen 2016; Loison 2015). As most households in low-income countries do not have access to cash-based income earning activities (e.g., salary employment), they often sell their products (e.g., crop, livestock) in the market to make a living and meet cash requirements (school fees, buying fuel, etc.). However, due to limited access to markets and to information on market prices, the degree of agricultural commercialization is low (Newshan et al. 2018). An unexpected and sudden inflow of refugees might bring both benefits and constraints to the local economy, which might then create or shrink opportunities for livelihood diversification and agricultural commercialization. Hence, it is important to understand how livelihood strategies of people in host communities change in the presence of refugees. The current paper attempts to address whether refugee inflow expands or diminishes opportunities for livelihood diversification and agricultural commercialization. Our findings show that refugees bring substantial benefits to host communities through creating more jobs that people engage in as a secondary occupation, and raising demand for livestock products. Refugee inflow has a positive impact on diversification of livelihood activities as a secondary occupation and on commercialization of livestock products. Specifically, a 1 percent increase of the refugees’ presence 5 leads to a 2.7 and 15.9 percent increase in the diversification of livelihood activities as a secondary occupation and value of livestock product sale, respectively. It should be noted that this analysis is taking place during a period where refugees in Ethiopia were prohibited by law from seeking work outside designated camps. This has changed after 2019 because of the revised Ethiopian Refugee Law. These effects tend to be heterogeneous across regions and to a limited extent, vary depending on the gender of the household head. The negative effects tend to be concentrated in Gambella, a region that hosts most of the refugee population in Ethiopia and where the refugee population is as large as the population of the region. Overall, compared to women-headed households, households with a male head seem to benefit through increased diversification of activities as a secondary 4 Region refers to the administration level 1 from the Database of Global Administrative Areas (GADM). The nearest region to the refugee camp is identified as the one that has the shortest straight distance to the refugee camp among all neighboring regions in the major refugee source countries. 5 As explained above, refugee presence is the number of refugees (population) in the nearest refugee camp to the household location weighted by the household’s inverted distance to the camp. 3 occupation and livestock product sale. We identify households' engagement in different individual livelihood activities and access to market as potential mechanisms for the observed effects. 6 This paper adds to the emerging literature on the microeconomic consequences of refugee presence in host communities. Existing studies have examined the impact of refugee presence or shock on several host community outcomes, including consumption and wealth (Alix-Garcia et al. 2018; Becker and Ferrara 2019; Ayenew 2021, Kadigo et al. 2022), child health (Baez 2011; Tatah et al. 2016; Wang Sone and Verme 2019), labor market participation (Becker and Ferrara 2019; Fallah et al. 2019; Ruiz and Vargas-Silva 2015), education (Bilgili et al. 2019), and gender roles (Ruiz and Vargas-Silva 2018). While the majority of these studies found a positive effect of refugee presence on local communities, there are some notable exceptions, such as Baez (2011) who documented a negative impact on child health and educational outcomes in Tanzania and Ayenew (2021) who reported a negative effect on consumption. The effects tend to be heterogeneous by observable characteristics of household in the host communities – rural residents, low skilled individuals, the poor and women tend to bear the negative consequences while relatively high skilled individuals tend to benefit (see e.g., Ruiz and Vargas-Silva 2015; 2019; Ayenew 2021). Although prior studies cover several socio- economic outcomes, to our knowledge, none have investigated the effect of refugee presence on household’s livelihood strategy choices. Hence, the current paper contributes to this literature by investigating the impact of refugee presence on people’s livelihood strategies in refugee host communities. The paper is also related to the literature on the determinants of an improved livelihood (welfare) in low- and middle-income countries. The existing literature has (i) investigated the nature and drivers of poverty and poverty dynamics (see e.g., Adato et al. 2006; Naschold 2012) and (ii) emphasized understanding on the nature and dynamics of livelihood of people using household income portfolio from different activities, asset endowment of different types and contexts (see Angelsen et al. 2014; Jiao et al. 2017). While the literature acknowledges the importance of exogeneous shocks (e.g., refugee inflow, drought, land appropriation) in influencing people’s welfare (livelihood) as well as welfare (livelihood) dynamics (see e.g., Wunder et al. 2014; Giesbert and Schindler 2012), the current literature tend to focus on investigating the effect of natural (e.g., drought) and policy shocks (e.g., land appropriation) rather than a shock arising from a potential large and sudden influx of people. To the best of our knowledge, there is a paucity of empirical evidence on the nexus between refugee inflow and livelihood strategies. Fallah et al. (2019) examined the impact of Syrian refugees on employment outcomes of host communities in Jordan and documented a positive effect even though the hosts changed the type of work they engaged in, and the refugees competed with less educated individuals. Tumen (2016) and Bağır (2018) examined the impact 6 The conflict in Northern Ethiopia has caused substantial internal displacement and hindered effective delivery of humanitarian assistance to the affected populations. About 2.1 million, 250,000, and 112,000 people in Tigray, Amhara, and Afar region, respectively, have been internally displaced. As a result of the conflict, two refugee camps in Tigray (i.e., Hitsats and Shimelba) were destroyed in January 2021, and thousands of refugees that sheltered in the camps fled. UNHCR, together with the Ethiopian Agency for Refugees and Returnees Affairs (ARRA), has been working to locate the refugees. The two institutions have also been facilitating the relocation of thousands of Eritrean refugees in three camps (i.e., Mai Aini and Adi Harush in Tigray region and Berhale in Afar region) (UNHCR 2021b). These substantial internal displacements and changes in the refugee situation could affect our findings in Afar and Tigray region. 4 of Syrian refugees on labor market outcomes in Turkey and reported that refugees negatively impacted the outcomes of low-skilled and less-experienced workers and workers participating in the informal labor market. Taking the case of Rwandan and Burundian refugees in Tanzania, Ruiz and Vargas-Silva (2015) investigated the effect of refugee inflow on the likelihood of adult’s outside employment (i.e., wage and salary employment activities) and found a negative impact. Using the same data set as in Ruiz and Vargas-Silva (2015), Ruiz and Vargas-Silva (2018) made intrahousehold comparison of time allocated to farming, firewood, and water, outside employment and schooling activities and found that the refugee inflow decreases the likelihood that women engage in outside employment although the results tend to vary depending on the baseline literacy and numerical skills (prior to the arrival of refugees). Maystadt and Verwimp (2014) investigated occupation-differentiated welfare impact of refugee presence on host communities in Tanzania. They documented a positive impact on the welfare of host communities, which is heterogeneous across host communities’ occupation – agricultural wage workers are worse off while self-employed agriculturalists are better off. Kadigo et al. (2022) also reported occupation differentiated welfare impacts of refugee presence in Uganda. Using the case of Congolese refugees in three Rwandan camps, Loschmann et al. 2019 found that living closer to refugee camps increases wage employment, asset ownership and women’s engagement in business activity with no effect on subjective wellbeing. In a meta-analysis of 59 studies on the economic impact of forced displacement covering 19 major episodes of forced displacement crises, Verme and Schuettler (2021) confirmed that people in the informal market, and low skilled, young, and female workers are the ones most affected, in terms of losing employment or wages. They also found that the negative effect of refugee presence on host communities verifies in the short-term and tends to dissipate with time. There is also an extensive literature on the effect of migration, particularly voluntary, on labor market outcomes in host communities (see e.g., Borjas 1999; Card 2005; Dustmann et al. 2005). However, this literature tends to focus on high-income countries and has mixed conclusions on the labor market outcome of migration (Longhi et al. 2005; Dustmann et al. 2015). This paper departs from existing studies in two ways. First, most of the studies only consider a subset of occupation or livelihood activities (mainly employee-based) and do not provide a full picture of the livelihood impacts of refugee presence on host communities. We consider an exhaustive set of livelihood activities in which households (individuals) in the impacted communities may engage. 7 Second, prior studies tend to focus on the different livelihood activities separately (i.e., whether the individual adult members or the household engage in each of the livelihood activities). 8 Therefore, they are unable to infer whether households are diversifying or specializing their livelihoods or are engaging more on the commercialization of activities. 9 The current paper goes beyond the allocation of labor to individual (specific) 7 As the data we used does not have a good welfare indicator (e.g., income, consumption, and assets), we could not explore the welfare impact of refugee inflow. 8 We examined households’ engagement in individual livelihood activities as a mechanism for household livelihood strategies. 9 Generally, households tend to diversify their livelihood when facing negative shocks (e.g., conflicts, droughts) to minimize risk (Ellis 2000a, b). In the case of refugee inflow, households may either diversify or specialize as refugee inflow could be both a negative shock (through increase competition for resources, services, and employment) and a positive shock (through creating opportunities, such as high demand agricultural products, provision of cheap labor). 5 occupations and looks at household strategies in terms of the degree of occupation diversification and commercialization. 2. Refugees in Ethiopia Ethiopia has a long history of welcoming refugees fleeing from conflicts, political repression, forced military service, drought, and conflict-induced food insecurity in neighboring countries (Carver 2020; Abebe 2018). In the late 1980s, the country established Hartisheik and Itang refugee camps, among the largest in the world, to host refugees from Somalia and Sudan, respectively. Ethiopia is the third largest refugee-hosting country in Africa and refugees are primarily from South Sudan, Somalia, and Eritrea (Vemuru et al. 2020; Nigusie and Carver 2019). 10 Most of the refugees are in Gambella, Benishangul-Gumuz, Somali, Afar and Tigray regions (see Figure 1). Most of the refugee camps are located close to the international border and in the least-developed areas of the host regions (Vemuru et al. 2020). At the time of data collection (in 2018), the majority of refugees were living in 26 refugee camps, depended on humanitarian assistance and had limited access to services and job opportunities. A significant number of refugees (mainly from Eritrea and Somalia) also live in Addis Ababa with their own arrangements and networks (Nigusie and Carver 2019). 10 At the end of January 2022, Ethiopia was the second largest refugee-hosting country in Africa (830,305 refugees in total) with 99 percent of them originating from four countries, namely, South Sudan, Somalia, Eritrea, and Sudan. 6 Figure 1: Map of regions in Ethiopia, location of refugee camps, and refugee source countries. Source: Database of the Global Administrative Areas (GADM) (https://gadm.org/data.html, accessed on November 20th, 2020) In Ethiopia, the Refugees and Returnees Services (RRS, former Agency for Refugees and Returnees Affairs (ARRA)) is responsible for managing refugee camps and its oversight by making sure that the commitment of the federal government is met (Nigusie and Carver 2019). Except for Eritrean refugees, most of whom are eligible for out of camp policy, arriving refugees, at the time the data was collected, were allocated to one of the 26 refugee camps spanning the five refugee hosting regions. Refugees living outside of camps represent about 10 percent of the refugees in Ethiopia (Abebe et al. 2018). The allocation tends to be based on shared identity between the refugee and the host communities and the distance of the refugee camps from the border of the source country. The South Sudanese refugees are hosted in the refugee settlements in Gambella, except the few who were relocated to the refugee camps in Benishangul-Gumuz region. Most of these refugees arrived during the civil conflict in South Sudan in 2013 (Nigusie and Carver 2019). Refugees from Somalia are allocated to camps in the Somali region. Refugees from Eritrea are diverse in ethnic composition and are allocated in camps across two regions: refugees from the Afar ethnic group are allocated to camps in Afar region while refugees from other ethnic group (largely Tigre) are placed in camps in Tigray region (Nigusie and Carver 2019). 7 Refugees typically have limited access to education and employment and face restricted mobility outside the camps (Abebe 2018). However, the Ethiopian government has recently undertaken several measures, which are indications of a policy shift from the longstanding encampment policy to a mix of encampment, out of camp and local integration approaches. Ethiopia was one of the 17 refugee-hosting countries that signed the UN Declaration for Refugees and Migrants during the UN’s Leaders’ Summit on Refugee in New York in September 2016. At the event, Ethiopia, the co-host, announced its Job Compact – an industrialization effort to create over 100,000 jobs for Ethiopians and refugees. Ethiopia also signed up for the Comprehensive Refugee Response Framework (CRRF). The declaration and the framework advocate for shared responsibility to secure refugee self-reliance and local integration of refugee communities (Nigusie and Carver 2019). In 2019, Ethiopia also signed on to the Global compact on Refugees. Furthermore, the Ethiopian government passed a new proclamation, which repeals the 2004 refugee law, in February 2019. Unlike its predecessor, the 2019 proclamation grants refugees the right to work along with rights extended to other foreign nationals. In Ethiopia, the refugee communities have been documented to affect the local host in several ways (Vemuru et al. 2020; Watol and Assefa 2018). Economically, the arrival of the refugee communities and the accompanying relief activities create a market for local agricultural products and services, increase commercial activity and trade, and provide a unique set of skills and knowledge to the local economy. Socially, refugees and host communities interact with each other at markets, religious ceremonies, wedding and funeral services, sport places, and public infrastructure while using services (e.g., health, education). Intermarriages, which create social connections between host and refugee communities, are also common. On the other hand, the arrival of refugees also raised the price of basic goods and services, created competition for employment opportunities, and increased pressure on the local environment. Refugee presence can also breed local insecurity, including petty thefts and violent robbery, where economic precarity and ethnic differences (between refugee and host communities) were prevalent. In host communities, where adolescents are the largest demographic group, refugee presence could also increase Gender-Based Violence, leading to a high prevalence of sexual violence, unintended pregnancies, and sexually transmitted diseases, such as HIV (Gebrehiwet et al. 2020). In terms of public services, such as water, education, and health, refugee presence has been associated with increased access to those services (Vemuru et al. 2020). These suggest that refugee presence can affect the livelihood of hosts in Ethiopia both negatively and positively. The overall net effect depends on the dominance of the positive effects over the negative effects, and vice versa. 3. Conceptual framework This study is conceptually grounded on the canonical sustainable livelihood framework (Scoones 2015; Ellis 2000), and the household livelihood strategy framework (e.g., Nielsen et al. 2013; Walelign and Jiao 2017), which is an improved version of the former. Figure 2 presents the analytical framework of the current paper. The framework, adapted from Nielsen et al. (2013), and Walelign and Jiao (2017), is suitable to study the impact of refugee inflow 8 on livelihood strategies in host communities. Under this framework, the unit of analysis can be both a household and an individual household member. 11 The main elements used in livelihood analyses are assets, activities, strategies, outcomes, and contexts. Assets (natural, social, physical, human, and financial capital) play critical roles in determining households’ or individuals’ livelihood; they directly influence choices of livelihood activities, strategies, and outcomes. Livelihood activities are occupations households or individuals engage in. A livelihood strategy, the portfolio of livelihood activities that a household or an individual adopts, is dynamic and adaptable to availability of resources (e.g., labor, physical, financial assets) and to the changing contexts (Davis et al. 2010; Nielsen et al. 2013). Diversification and commercialization of activities are two major rural livelihood strategies. The contexts, encompassing human and natural forces (e.g., policies, shocks), enable or constrain the use of assets to influence livelihoods positively or negatively. Hence, contexts can create both opportunities to improve or maintain livelihoods and precipitate conditions to depress livelihoods. Contexts are often exogeneous and hardly under the full control of the households or individuals. Refugee hosting communities face refugee inflow as an additional contextual factor on top of existing ones. Refugee inflows increase the demand for goods and services (demand effect) and increase competition for labor and agricultural products (supply effect) (Baez 2011; Fallah et al. 2019; Alix-Garcia et al. 2018), which in turn affect assets, livelihood activities, livelihood strategies, and livelihood outcomes in the host communities through influencing existing contextual factors (e.g., population density, policies (mainly related to refugee)). Hence, refugees can either create opportunities or bring constraints for livelihood improvement in host communities. This means that refugees can expand or inhibit livelihood strategies: livelihood diversification and commercialization. If refugees compete with the hosts through supplying labor and agricultural products in the local market, refugee inflow inhibits household livelihood diversification and agricultural commercialization as this process depresses labor and product prices. If refugees significantly increase demand for local goods, services, and labor, refugee inflow promotes livelihood diversification and commercialization as this process raises product and factor prices. The overall effect depends on the dominance of one effect over the other, which differs across contexts depending on - e.g., the degree of substitutability between refugee and host community’s labor, the degree and medium of support (whether cash or kind) to the refugee community, or the livelihood (strategy) component under consideration. However, the effect is more likely to be heterogeneous across socio-economic groups in host communities as refugee inflows affect people differently depending on their contextual factors, livelihood activities, and outcomes. Unlike other contextual factors, refugee presence in host communities is also triggered by negative shocks (e.g., conflicts, natural disasters) in the source communities. Furthermore, refugee inflow in host communities depends on the population size exposed to as well as the intensity of the negative shocks in the source communities, and distance between the exposed and the host communities (Figure 2). This implies that refugee inflow in a particular community is not random, but rather depends on the contextual factors in source countries and their regions. We use this notion to construct a composite variable using distance from origin to the host community (measured in kilometers), intensity of conflicts (measured in number of deaths/fatalities per conflict), and population density in the origin region (potential refugee 11 Following the Poverty Environment Network (PEN) approach, a household is defined as a group of people (family and non-family members) who live under the same roof and share resources and income. Individual household members are those who have lived in the household for at least six months during the year of data generation (PEN, 2007). In the current paper, the unit of analysis is the household. 9 inflow). We then use the composite variable as an instrument to identify the causal impact of refugee presence on livelihood diversification and commercialization in host communities. Figure 2: Household livelihood strategy framework under refugee inflow Source: Adapted from Nielsen et al. (2013) and Walelign and Jiao (2017) 4. Data sources The major data source of the current study is the World Bank’s Development Response to Displacement Impacts Project (DRDIP) 12 baseline survey from Ethiopia. The Ethiopia DRDIP survey was administered between September 2017 and August 2018. The survey covers 113 Kebeles (wards) in 16 Woredas (districts) from the top five refugee-hosting regions in Ethiopia. The selection of the sample households follows stratified random sampling with proportion to size (the number of households) using Woredas as a geographic stratum. The sample originally comprised a total of 3,390 households, who were selected using systematic random sampling within each Woreda. We used data from 3,375 households, as 15 of them were excluded due to missing location information (GPS). The sample households are located within varying distance from the nearest refugee camp (approx. 67 to 76,665 meters) (see Figure 3). 12 DRDIP aims to improve access to basic social services, expand economic opportunities, and enhance environmental management for communities hosting refugees through providing funding for community driven projects. 10 Figure 3: Location of refugee camps in Ethiopia and the Ethiopia Development Response to Displacement Impacts Project (DRDIP) sample households Source: Authors’ compilation using the database of the Global Administrative Areas (GADM) (https://gadm.org/data.html, accessed November 20, 2020) and the Humanitarian Data Exchange (HDX) database for the refugee location (https://data.humdata.org/dataset/ethiopia- refugee-camp-locations, accessed November 20, 2020). From the Ethiopia DRDIP data set, we derive two measures of livelihood diversification and two measures of agricultural commercialization (all at household level). The measures of diversification include: (i) the degree of labor diversification in different productive livelihood activities (e.g., farming, wage employment) as a primary activity (occupation), and (ii) the degree of labor diversification in different livelihood activities as a secondary activity (occupation). 13 These two outcomes were constructed using the inverse Simpson diversity 1 index as ∑ 2 , where is the share of the number of adult labor engages in ℎ livelihood activity to total active adult household labor and ranges from 1 to the number of livelihood activities that a household engages in (Valdivia et al. 1996). The measures of commercialization, which were adjusted for adult equivalent units, include the value of several 13 Primary and secondary occupations are quite distinct as different livelihood activities differ in income generation and time allocation. This approach has been used in the literature to understand livelihood strategies (see e.g., Pender et al. 2014; Jacobs and Makaudze 2012). 11 crop products (e.g., wheat, potatoes) and the value of five livestock products (i.e., milk, egg, butter, hides, and honey) sold in the market. 14 Several other data sources were utilized. First, the Ethiopian refugee camps location data set from the Humanitarian Data Exchange (HDX) 15 and the total number of refugees by camps from the United Nations High Commissioner for Refugees (UNHCR), Addis Ababa office. We use data from 26 official UNCHR refugee camps in Ethiopia that were operational in 2018 (see Figure 1; 3). Second, we use administrative data sets for Ethiopia and refugee source countries from the database of Global Administrative Areas (GADM). 16 We also use the conflict data set from the Armed Conflict Location and Event Data Project (ACLED) 17 and the population data from the Gridded Population of the World (GPW) data set. 18 On the basis of these data sets and the location of sample households from Ethiopia DRDIP data set, we generated the following variables: i) distance of sample households to the nearest refugee camp, the nearest region (administration level 1 in GADM) to the refugee camps, ii) distance of the refugee camps to the nearest border of the refugee source country, iii) the intensity of conflicts, and iv) population density in refugee host countries by region. Appendix B presents the list of variables along with description and summary statistics. 5. Empirical strategy To estimate the impact of refugee inflow 19 on host community’s livelihood strategy choice, we use the following basic econometric model: = + + + + (1) Where, indexes a household, is an outcome variable of interest (livelihood diversification or commercialization of agriculture), is the measure of refugee inflow, i.e., the refugee population (average of 2017 and 2018) in the nearest refugee camp weighted by the inverse of distance of the household to the refugee camps, is a set of household controls, is kebele fixed effects, and is the error term. Several variables, from the DRDIP data set, were used as controls in our model. These include total livestock values, size of land owned, distance to all weather roads, membership in 14 We focus on livestock products as we do not have data on the number of live animals sold. 15 https://data.humdata.org/dataset/ethiopia-refugee-camp-locations. 16 https://gadm.org/data.html. 17 https://acleddata.com/curated-data-files/. 18 https://sedac.ciesin.columbia.edu/data/set/gpw-v4-population-count-rev11. 19 One could argue that our estimate commingled with refugee camp and international aid. We believe that refugee inflow and refugee camps are the same if most of the refugees live in the camps as the camps and refugee attributes are the same and could be used interchangeably as it is in the literature. We, however, prefer refugee inflow to refugee camp as our main explanatory variable is constructed as the number of refugees weighted by inverted distance. Drawing experiences from Ethiopia and other refugee host communities, we believe that international aid and other benefits coming to our study communities are because of refugee flow (establishment of a refugee camp) and hence should be looked at as a mechanism through which refugee camps (refugee inflow) affects local communities’ livelihood. Unfortunately, we could not get data on refugee related international aid in the communities to undertake a causal mediation analysis. If there is any cause of concern on its omission, the Kebele fixed effects absorb the bias that could arise as refugee induced international aid is less likely to vary within Kebeles. If there are any other international aid programs that are not related with refugee inflow/the establishment of refugee camps, the aids are more likely to be placed at a Kebele level and hence less likely to vary within kebeles and the Kebele-fixed effects absorb most of the variation to mitigate the resulting bias. 12 community organizations, amount of loan for productive purposes, household head’s gender, marital status, and education. Following Angelsen et al. (2014), all assets shared by household members in production and consumption (e.g., land holding) as well as all production and income values were divided by adult equivalent units (aeu) of a household, to adjust for household size and allow inter-household comparisons. The coefficient β, in equation (1), can be interpreted as causal if refugee inflow is exogeneous, i.e., among others, there are no excluded variables in the model that affect the outcome variable of interest as the distance to and (or) the number refugees in the nearest refugee camp change(s). This is highly unlikely to hold as the location of refugee camps as well as refugees’ choice of camps or allocation of refugees to the camps, is not random for two reasons. First, refugee camps are often located along international borders for logistical purposes and easy repatriation of the refugees when the situation in the source countries become stable. Second, refugees travel long distance on foot and would like to stay in the closest refugee camps to their origin country (Baez 2011; Ruiz and Vargas-Silva 2018). Consequently, refugee inflow is endogenous and the estimated β from equation 1 is likely to be biased. We instrument refugee inflow with an Instrumental Variable, , which is constructed as: = (2) Where, is the population density (in 2010) of the nearest region in the nearest source country to the refugee camp, is the intensity of conflict (between 2001 and 2018) in the nearest region in the nearest source country to the refugee camps, and is the distance of the camp to the nearest border (of the nearest region or source country). 20 Subscripts , , and stand for region in source country, refugee camp, and border to the nearest source country, respectively. In principle, the constructed potential refugee inflow is a valid instrument as (i) most refugee camps in Ethiopia are located close to the border, 21 and people seeking refugees are allocated to the nearest refugee camp given the capacity of the camp and (ii) the magnitude of refugee flow depends on the population density and intensity of conflicts in the source country (Baez 2011; Ruiz and Vargas-Silva 2018; Fallah et al. 2019). In addition, potential refugee inflow, , is less likely to affect the outcome variables of interest directly except through a household’s exposure to actual refugee inflow, . This is confirmed by F-statistics (>10) in most of the first stage regression (see below and Appendix C; E). Previous studies (e.g., Fallah et al. 2019; Baez 2011; Alix-Garcia, J., 2018) employed similar instruments to estimate the causal impact of refugees on host communities. However, we anticipate two potential threats to our identification strategy. First, the intensity of conflict in the nearest region of source countries could affect economic activities and livelihoods in refugee host communities if there is significant cross-border trade between host communities and refugee source countries. But this is less likely because: (i) most of the host 20 Following the literature of the impact of refugees (see e.g., Baez et al. 2019), we prefer to use the composite index, rather than its components, as an instrument because the composite index gives us variation at the refugees’ camp level. The two components (population density and intensity of conflicts) mainly vary at refugee host regions and does not provide sufficient variation to explain refugee exposure if used individually. 21 The mean and median distances to the border are 53,921 and 48,924 meters, respectively, with the minimum of 3,439 meters, and maximum of 138,489 meters. 13 communities in our sample are located close to, but not just at the border; 22 and (ii) most cross- border trade between host communities and refugee source countries occurs between major towns and most of our study kebeles are far away from these towns. Second, the reported number of conflicts in the neighboring regions could be associated with communication infrastructure in the place of origin. However, most neighboring regions in source countries are underdeveloped with very similar levels of communications infrastructure and accessibility, so we do not anticipate any significant systematic differences in the number of casualties reported. We, however, control for any potential effect of these two threats by including the nearest region of the neighboring countries fixed effect in our regression models so that a significant part of the effect of conflict can be attributed to refugee inflow. We use two-stage least squares (2SLS) 23 to estimate the causal impact of refugee inflow on livelihood diversification, and commercialization. The 2SLS is implemented by estimating the following two equations as a first and second stage regression, respectively: = 1 + 1 + 1 + 1 + 1 (3) � + 2 + 2 + 2 = 2 + 2 (4) Where, � is the fitted refugee inflow from equation (3) and all the notations are similar to the ones in equation 1 and 2. The estimate 2 is the adjusted effect of refugee presence on the outcome variable of interest and can be interpreted as a causal effect assuming our instrument ( ) is valid. The regression results and summary statistics of first-stage regressions, with and without controls, are presented in Appendix C. Both first-stage regressions have a significant statistics ( (1,112) = 18.283 and = 19.838 , < 0.01 with and without controls, respectively) and Kleibergen-Paap rk Wald statistic ( 2 (1) = 19.02 and = 21.86 , < 0.01), suggesting the instrument is a good predictor of refugee presence in the first stage regression. Kleibergen-Paap Wald rk F statistic is 18.81 and 20.94 with and without control, respectively, both the critical value of 16.38 (for a 10% maximal IV size), suggesting our model is well identified. We start our main estimation for overall effects using the full sample, without the covariates and then include covariates to check the robustness of our results. The set of controls include whether the household is a member of a community and social organization, the amount of loan for productive purpose (in Birr; aeu adjusted), total land owned by the household (in hectares; aeu adjusted), value of livestock owned (in Birr; aeu adjusted), distance to all- weather road (in Kilometers), level of education, gender, and marital status of the household head. In all models, we include kebele fixed effects to control for unobserved heterogeneity at the kebele level, and cluster standard errors at the village level. 22 The mean and median distances to the border are 44,966 and 39,888 meters, respectively, with the minimum of 37 meters, and maximum of 107,406 meters. 23 2SLS is an estimation technique used to estimate the causal impact of an explanatory variable of interest (i.e., refugee inflow) on outcome variable of interest (i.e., livelihood diversification and commercialization) in the presence of endogeneity threats. It is implemented with two regressions: the first regression estimates the value of the endogenous variable using an additional variable (i.e., potential refugee exposure) and the second regression uses the predicted values to estimate the impact of the endogenous variable. 14 6. Results and discussion Overall effects Table 1 shows results from our 2SLS estimation using, as a dependent variable, the two measures of livelihood diversification – diversification of activities as a primary and secondary occupation (Panel A), and the two measures of agricultural commercialization – sales of crop and livestock products (Panel B). The table also presents OLS estimates (with controls) for comparison. Both dependent variables and refugee inflow are Inverse Hyperbolic Sine (IHS) transformed, so the coefficients can be interpreted as elasticities. 24 The OLS estimate results are lower in magnitude than the 2SLS, indicating that the OLS estimate would underestimate the effect of refugee presence. 25 Here, we will discuss the estimates from the 2SLS. The 2SLS results show statistically significant and intuitively signed coefficients for diversification of activities as a secondary occupation and sale of livestock products, both with and without controls. The estimates with and without controls are similar, showing the results are robust to the exclusion of controls. Regarding livelihood diversification, we find that a 1% increase in refugee inflow leads to more than 1% (2.5 and 2.7% without and with controls, respectively) increase in diversification of activities as a secondary occupation. We find no effect on diversification of activities as a primary occupation across model specifications. Putting the results together suggests that refugee presence significantly increases availability of livelihood activities and households in host communities that pursue these activities as a secondary occupation (because of the positive elasticity effect on secondary occupation) while retaining their primary activity (because of the robust null effect on primary occupation). This can be attributed to the nature of primary and secondary activities in the study areas, where many individuals engage in activities that require skills or livelihood assets as a primary occupation rather than a secondary occupation. For instance, about 7.6% of the individuals who have a primary occupation engage in salaried employment (comparable to 1.5% of individuals who have salary employment as a secondary occupation), which requires knowledge of the local working culture and some level of education (see Appendix D). Hence, people tend to keep their primary activity and either take on or diversify in a secondary activity. While less skilled and/or asset-intensive (secondary) occupations face fierce competition from immigrants or refugees as observed by Morales (2017) in Colombia and Ruiz and Vargas-Silva (2018) in Tanzania, such competition in host communities in Ethiopia is absent as refugees in Ethiopia, except for some Eritrean refugees, experience restricted movement out of camps and are not permitted to work according to the Ethiopian refugee law prior to 2019. This precludes the argument that households engage in more activities as a coping strategy to maintain their existing living standards. 24 The elasticities are calculated from the regression coefficients following Bellemare and Wichman (2020) to account for distortions in actual elasticities due to IHS transformation. If a percentage change in refugee presence leads to a more than a percentage change in our outcome variable of interest (i.e., the coefficient is greater than one), the observed effect is elastic. In contrast, if a percentage change in refugee presence leads to a less than a percentage change in outcome variables of interest (i.e., the coefficient is less than one), the observed effect is inelastic). 25 This could be for several reasons, but the major one is our main explanatory variable of interest – refugee inflow – is not randomized and hence, the estimates from OLS are less consistent and more biased. 15 Turning to agricultural commercialization, a 1% increase in refugee inflow leads to a very high increase (16.0% and 15.9% without and with controls) in livestock sales while we do not see a statistically significant effect on sale of crop products. This could be attributed to two facts related to the refugee situation in Ethiopia. First, most refugees are typically provided with cereals to meet their food requirements. This leads to the: (i) creation of a market and increase demand for livestock or its products relative to crop products, and (ii) selling of the cereals that the refugees received as assistance, increasing the supply of crop products, which could depress prices and make commercialization of crop products less attractive (Vemuru et al., 2020). These differential changes in markets and prices of livestock and crop products may lead to structural changes in livelihood: household and individuals allocate their labor primarily to livestock production than crop production. We will turn to this in the mechanisms section (see below). Second, most of the refugees are pastoralists, who demand more livestock products (e.g., milk) than crop products due to their dietary traditions. The demand effect is larger, also observed around Kakuma refugee camp in Kenya by Alix-Garcia et al. (2018), as dairy products are rarely included in the refugee food assistance bundle. Table 1: OLS and 2SLS estimation of the impact of refugee inflow on livelihood diversification and agricultural commercialization at household level (1) (2) (3) (4) (5) (6) OLS 2SLS 2SLS OLS 2SLS 2SLS Panel A: Livelihood diversification IHS(Primary activity IHS(Secondary activity diversification) diversification IHS (refugee presence) 0.014 -0.392 -0.338 0.047 2.508** 2.733** (0.029) (0.285) (0.297) (0.110) (1.143) (1.274) Panel B: Agricultural commercialization IHS(Crop product sale) IHS(Livestock product sale) IHS (refugee presence) 0.264 1.305 -1.336 0.435 16.036** 15.845** (0.262) (1.983) (2.927) (0.309) (7.239) (7.610) N 3375 3375 3375 3375 3375 3375 Kebele fixed effects Yes Yes Yes Yes Yes Yes Nearest region fixed Yes Yes Yes Yes Yes Yes effects Controls Yes No Yes Yes No Yes ***significant at 1%, **significant at 5%. *significant at 10%, village level clustered standard errors in parenthesis; IHS stands for inverse hyperbolic sine transformation. Region and gender differentiated effects Table 2 presents the 2SLS estimation by region. The results show that the overall findings tend to be heterogeneous across regions. Positive effects of refugee presence on diversification of livelihood activities as a primary occupation were documented in Tigray region while a negative effect was documented in Afar and Somali. All these effects were inelastic. Regarding diversification of activities as a secondary occupation, a positive effect was observed in Tigray, Afar and Somali regions while a negative effect was observed in Gambella region. All these effects were inelastic, except for the Somali region. The refugee presence had an inelastic negative effect on sale of crop products in Tigray and Gambella regions. In contrast, the refugee presence had an inelastic positive effect in Tigray region and an elastic positive effect in the Somali and Benishangul-Gumuz regions. We observed a marginally significant positive effect of the refugee presence on diversification of activities as a primary occupation (very inelastic) and sale of crop products (very elastic) in Benishangul-Gumuz region. The results revealed the 16 negative effect on most livelihood diversification and commercialization outcomes tends to occur in Gambella region while the more positive effects tend to occur in Tigray (except sale of crop products), Afar (except diversification of activities as a secondary occupation), and Somali (except diversification of activities as a secondary occupation). These differential impacts across regions could be attributed to the large number of refugees that Gambella hosts: The region hosts about half of the refugee population in Ethiopia which is almost as large as the population of the region. Table 2: 2SLS estimation of the impact of refugee inflow on livelihood diversification and agricultural commercialization by region. (1) (2) (3) (4) (5) Tigray Afar Somali Benishangul- Gambella Gumuz Panel A: IHS(Primary activity diversification) IHS (refugee presence) 0.220*** - -0.283** 0.516* -0.058 0.104*** (0.016) (0.027) (0.112) (0.275) (0.114) Panel B: IHS(Secondary activity diversification) IHS (refugee presence) 0.333*** 0.439*** 2.456*** 1.584 - 0.807*** (0.035) (0.045) (0.618) (1.071) (0.278) Panel C: IHS(Crop product sale) IHS (refugee presence) -0.774*** -0.156 0.045 9.537* - 0.430*** (0.127) (0.138) (0.922) (4.986) (0.139) Panel D: IHS(Livestock product sell) IHS (refugee presence) 0.537*** -0.091 14.533** 4.610** -0.812 * (0.098) (0.096) (3.774) (2.080) (1.423) N 833 390 990 536 626 Kebele fixed effects Yes Yes Yes Yes Yes Nearest region fixed Yes Yes Yes Yes Yes effects Controls Yes Yes Yes Yes Yes ***significant at 1%, **significant at 5%. *significant at 10%, village level clustered standard errors in parenthesis; IHS stands for inverse hyperbolic sine transformation; the summary statistics of first-stage regressions for each region are presented in Appendix E. The first-stage regressions have a significant F statistic greater than 10 for all regions subsample regressions, except Benishangul-Gumuz and Gambella regions. Hence, we use limited information maximum likelihood (LIML) estimator, instead of 2SLS estimator, for Benishangul-Gumuz and Gambella regions sample, which minimizes the bias in 2SLS due to the weak instrument by weighted linear combination of OLS and 2SLS estimates. The results in Table 3 present a gender-differentiated impact of the refugee presence on livelihood diversification and agricultural commercialization. The refugee presence had a positive effect on the diversification of activities as a secondary occupation for male-headed households and sale of livestock product: a 1% increase in refugee inflow led to a 2.7% and 15.9% increase in the diversification of activities as a secondary occupation and sale of livestock products, respectively. We further find the refugee presence had no effect in any of the diversification and commercialization measures for female-headed households, suggesting that male-headed households are more likely to take livelihood opportunities created by refugee inflow than female-headed households. This could be partly attributed to the Ethiopian culture that encourages men to seek employment outside the home compared to women. In other 17 words, women tend to engage in housework. Such gender differentiated effects were also documented by Ruiz and Vargas-Silva (2020) in Tanzania. Table 3: 2SLS estimation of the impact of refugee inflow on livelihood diversification and agricultural commercialization by gender of household head. Female Male (1) (2) (3) (4) (5) (6) (7) (8) IHS(Primary IHS(Secondar IHS(Cro IHS(Livestoc IHS(Primary IHS(Secondar IHS(Cro IHS(Livestoc activity y activity p product k product activity y activity p product k product diversification diversification sale) sale) diversification diversification sale) sale) ) ) ) ) IHS (refugee -0.005 5.001 24.112 21.516 -0.338 2.733** -1.336 15.845** presence) (0.305) (3.669) (25.492) (14.285) (0.297) (1.274) (2.927) (7.610) N 550 550 550 550 2825 2825 2825 2825 Kebele fixed Yes Yes Yes Yes Yes Yes Yes Yes effects Nearest Yes Yes Yes Yes Yes Yes Yes Yes region fixed effects Controls Yes Yes Yes Yes Yes Yes Yes Yes ***significant at 1%, **significant at 5%. *significant at 10%, village level clustered standard errors in parenthesis; IHS stands for inverse hyperbolic sine transformation; Mechanisms At the household level, based on the overall sample, our findings show that refugee presence has a positive effect on diversification of activities as secondary occupation (livelihood diversification), and on sale of livestock products (livelihood commercialization). We also find a null effect of refugee inflow on diversification of activities as a primary occupation and on sales of crop products. While several mechanisms could support these findings, based on data availability and the conceptual framework; we study two major mechanisms: household member’s engagement in livelihood activities,26 and access to markets. The results are presented in Tables 4 and 5. Refugee presence has both a negative and positive effect on the number of adult household members that engaged in the different activities as a primary occupation (Table 4). Specifically, we find a positive effect on the number of household members engaged in salaried employment and livestock production (with an elasticity coefficient of 6.0 and 7.0, respectively) but a negative effect on crop production (with an elasticity coefficient of -2.7). The very negative and positive effects could result in no effect on the diversification of activities as a primary occupation. The very elastic positive effect on livestock production and the very elastic negative effect on crop production suggests household’s reallocation of adult labor from crop to livestock production. Such change of occupations in response to refugee inflow is also documented elsewhere, e.g., in Jordan (Fallah et al. 2019). 27 Regarding secondary occupations, refugee inflow had a positive effect on housework and crop production (with an elasticity coefficient of 6.1 and 7.2, respectively), resulting in the observed aggregated positive effect of refugee inflow on diversification of activities as a secondary occupation. The elastic effect of refugee presence on crop production provides further evidence on the adoption of crop production as a secondary activity rather than a primary activity for the reasons mentioned above. The increase in the number of household members in livestock production as a primary 26 One could argue that these variables are more of an outcome than a mechanism. However, as we showed in our conceptual framework individual household members engagement in different livelihood activities influences livelihood diversification (i.e., livelihood strategies). 27 Fallah et al. (2019) reported that the hosts in Jordan change the type of work they do (mainly characterized by a shift from private to public sector) in response to refugee inflow. 18 occupation and crop production as a secondary occupation contrast with the main finding of Ruiz and Vargas-Silva (2015) wherein refugee inflow does not have an effect on traditional employment activities. Table 4: 2SLS estimation of the impact of refugee inflow on the number of household members engage in each livelihood activities. (1) (2) (3) (4) (5) (6) IHS IHS IHS IHS (Crop IHS IHS (Salaried (Housework (Commerce or production (Livestock (Other employment ) other business) ) production activity ) ) ) Panel A: as a primary activity IHS (refugee 6.024*** 0.236 1.451 -2.708*** 7.033** -2.038 inflow) (2.208) (0.960) (1.216) (0.775) (3.046) (3.498) Panel B: as a secondary activity IHS (refugee 24.770 6.096*** 0.650 7.145** 3.897 1.497 inflow) (24.742) (2.146) (0.559) (3.290) (2.376) (1.955) N 3375 3375 3375 3375 3375 3375 Kebele fixed Yes Yes Yes Yes Yes Yes effects Nearest region Yes Yes Yes Yes Yes Yes fixed effects Controls Yes Yes Yes Yes Yes Yes Notes: ***significant at 1%, **significant at 5%, *significant at 10%; village level clustered standard errors in parenthesis; other activity includes causal labor, mining, religious leader, and other activities than listed above; IHS stands for inverse hyperbolic sine transformation; the summary statistics of first-stage regressions for each model are as presented in Appendix C as we used the same sample and set of covariates. We also find evidence that refugee presence improved households’ access to market, measured by the distance to markets in kilometers and minutes. Specifically, a 1% increase in refugee presence led to a decrease in distance to markets in minutes and kilometers by 2.6% and 3%, respectively. This can potentially be attributed to the creation of new markets (mainly driven by demand) due to refugee inflow, as observed by Vemuru et al. (2020), which reduced distance to markets and improved access to markets. This led to commercialization of activities through sale of livestock products as our main analysis revealed. Table 5: 2SLS estimation of the impact of refugee inflow on market access measured in distance to market in minutes. (1) (2) IHS (distance to market, in IHS (distance to market, in kilometers) minutes) IHS (refugee inflow) -2.644*** -3.092*** (0.924) (0.989) Kebele fixed effects 3375 3375 Nearest Region fixed Yes Yes effects Controls Yes Yes Kebele fixed effects Yes Yes Notes: ***significant at 1%, **significant at 5%, *significant at 10%; village level clustered standard errors in parenthesis; IHS stands for inverse hyperbolic sine transformation. 7. Summary and conclusion This paper aims to assess the impact of refugee inflow on livelihood strategies in host communities. The study considers two major livelihood strategies: livelihood diversification 19 and commercialization. Livelihood diversification is measured with diversification in livelihood activities as a primary occupation and diversification in livelihood activities as a secondary occupation. Agricultural commercialization is measured with value of crop product and livestock products sold in the market. Refugee inflow is measured as the number of refugees in the nearest refugee camp to a household weighted by the distance between the household and the camps. To address the endogeneity concern of refugee inflow arising from non-random location of refugee camps and refugees’ choice of a refugee camp (Baez 2011; Ruiz and Vargas-Silva 2018; Fallah et al. 2019), we employed two-stage least squares (2SLS) using potential refugee inflow as an instrument. Potential refugee inflow is the product of population density and intensity of conflicts (number of fatalities per event) in the closest region of the source country to the refugee camp weighted by the distance between refugee camp and the nearest border with the source country. The summary statistics from the first stage regression suggest that the instrument is a good predictor of refugee presence. Does inflow to refugee-hosting communities create or diminish opportunities for livelihood diversification and agricultural commercialization? Our findings provide evidence that refugee inflow creates opportunities for livelihood diversification and agricultural commercialization, in line with the existing literature documenting that refugees can have a positive effect on host communities, particularly on labor employment (see e.g., Loschmann et al., 2019). However, the finding tends to depend on the livelihood strategy element under consideration as well as the regional and socio-economic context. Regarding the livelihood strategy at play, we find positive impacts of refugee inflow on diversification of activities as a secondary occupation and commercialization of livestock products, while we find no effect on diversification of activities as a primary occupation, and commercialization of crop products. Our findings are robust to the exclusion of control variables. The positive effect on diversification of activities as a secondary activity can be attributed to the fact that refugee presence created jobs in different activities, and the host communities take the created jobs as a secondary occupation. The positive effect on the commercialization of livestock products is related to the fact that refugees are often provided with crop products (cereals) and tend to demand livestock products due to their customary diets. We identify access to market and household’s increased engagement in different livelihood activities as a potential mechanism for the observed effects. Regarding the regional and socio-economic context, our findings show heterogeneity across regions and gender of the household head and household members. Specifically, we find a negative impact of the refugee presence on refugee host communities in Gambella region that host most of the refugee population (almost the same size as the regional population) in Ethiopia. However, we mostly find positive effects in the Tigray, Somali, and Afar regions, and marginal positive effects in Benishangul-Gumuz region. These regions host a relatively smaller proportion of the refugee population in Ethiopia. The observed regional difference can be attributed to competition between host and refugee communities for work as documented in other places (Morales 2017; Ruiz and Vargas-Silva 2018). Our findings also show that male- headed households seem to benefit through increased diversification of activities as a secondary occupation and commercialization of livestock products, which could be attributed to the fact that female household members are usually responsible for housework in Ethiopian culture. Hence, our findings speak to the literature documenting that refugee communities have heterogeneous effects across various groups (see e.g., Maystadt and Verwimp 2014). 20 The findings of the study will have implications for the 2019 Ethiopian proclamation which repeals the 2004 refugee law to grant all refugees in Ethiopia the right to work and free movement, among others. The proclamation could lead to intense competition between refugee and local communities for work and other services and consequently depress the observed positive effects on host communities in most refugee hosting regions. The effect of the policy could be larger in major refugee hosting communities in Ethiopia, particularly in Gambella region, where the number of refugees is as large as the population of the region. This could have a substantial negative consequence on host communities’ livelihood, which may lead to hostilities between host and refugee communities, e.g., in Gambella. Interventions that create more jobs and maintain or build public infrastructure in host communities, such as the World Bank’s Development Response to Displacement Impacts Project (DRDIP), may favor a smooth transition to the new policy environment. This study has three limitations. First, in the absence of data on net income (gross income minus cost) for all income sources, we could not measure livelihood diversification in terms of net income from each activity, which is the common approach in the literature. However, income is the return of labor allocated to the different activities, which make us to believe that measuring livelihood diversification based on labor is a good proxy for income-based livelihood diversification. Second, as we did not have the price of livestock products, or the value of total livestock production, we measure agricultural commercialization in terms of gross sale of livestock products. We used a similar measure for crop commercialization for the sake of consistency. As these measures do not consider the scale of production, we could not confirm that the observed effect on livestock sale is due to actual market participation. However, total sale is still a good measure for commercialization, whatever is produced. Finally, we were not able to assess whether the observed effects lead to better livelihood outcomes as our data does not contain information on consumption, full spectrum of assets and income sources. While this needs to be explored further in the presence of refugee communities, previous studies have shown that livelihood commercialization and diversification are associated with better livelihood and welfare outcomes (see e.g., Ochieng et al. 2019; Loison 2015). Acknowledgments We are grateful to Matthew Stephens, Sandra Viviana Rozo Villarraga, and Kibrom Tafere at the World Bank for providing comments on earlier versions of the paper. We thank the UNHCR office in Addis Ababa for providing the refugee population data in Ethiopia and to Berhanu Ayalew at Food and Agricultural Organization for providing information on local contexts. 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 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. 21 References Abebe, T.T., 2018. Promises and challenges of Ethiopia’s refugee policy reform. Institute for Security Studies, Policy Brief 118. Adato, M., Carter, P.M.R., May, J., 2006. Exploring poverty traps and social exclusion in South Africa using qualitative and quantitative data. The Journal of Development Studies 42, 226–247. Alix-Garcia, J., 2018. Do refugee camps help or hurt hosts? The case of Kakuma, Kenya. Journal of Development Economics 130, 66–83. Angelsen, A., Jagger, P., Babigumira, R., Belcher, B., Hogarth, N.J., Bauch, S., Börner, J., Smith-Hall, C., Wunder, S., 2014. Environmental Income and Rural Livelihoods: A Global-Comparative Analysis. World Development 64, S12–S28. https://doi.org/10.1016/j.worlddev.2014.03.006 Ayenew, Ashenafi Belayneh. 2021. Welfare Impact of Hosting Refugees in Ethiopia. Policy Research Working Paper #9613. World Bank, Washington, DC. Baez, J.E., 2011. Civil wars beyond their borders: The human capital and health consequences of hosting refugees. Journal of Development Economics 96, 381–408. Becker, S.O., Ferrara, A., 2019. Consequences of forced migration: A survey of recent findings. Labour Economics 59, 1–16. Bellemare, M.F., Wichman, C.J., 2020. Elasticities and the inverse hyperbolic sine transformation. Oxford Bulletin of Economics and Statistics 82, 50-61. Bilgili, Ö., Loschmann, C., Fransen, S., Siegel, M., 2019. Is the Education of Local Children Influenced by Living near a Refugee Camp? Evidence from Host Communities in Rwanda. International Migration 57, 291–309. https://doi.org/10.1111/imig.12541 Carver, F., 2020. Refugee and host communities in Ethiopia: 2018-2019 integrated national survey. Overseas Development Institute, London. Davis, B., Winters, P., Carletto, G., Covarrubias, K., Quiñones, E.J., Zezza, A., Stamoulis, K., Azzarri, C., Digiuseppe, S., 2010. A cross‐country comparison of rural income generating activities. World Development 38, 48–63. Ellis, Frank, 2000. The Determinants of Rural Livelihood Diversification in Developing Countries. Journal of Agricultural Economics 51, 289–302. Ellis, F., 2000. Rural Livelihoods and Diversity in Developing Countries. Oxford University Press. Fallah, B., Krafft, C., Wahba, J., 2019. The impact of refugees on employment and wages in Jordan. Journal of Development Economics 139, 203–216. Gebrehiwet, K., 2020. The social health impact of Eritrean refugees on the host communities: the case of May-ayni refugee camp, Northern Ethiopia. BMC Research Notes 13, 182. Giesbert, L., Schindler, K., 2012. Assets, Shocks, and Poverty Traps in Rural Mozambique. World Development 40, 1594–1609. Jacobs, P. and Makaudze, E., 2012. Understanding rural livelihoods in the West Coast district, South Africa. Development Southern Africa 29, 574–587. Jiao, X., Pouliot, M., Walelign, S.Z., 2017. Livelihood Strategies and Dynamics in Rural Cambodia. World Development 97, 226–278. Kadigo, M. M., Diallo N. O., Maystadt J-F. 2022. How to Cope with a Refugee Shock? Evidence from Uganda. Policy Research Working Paper #9950. World Bank, Washington, DC. 22 Loison, S.A., 2015. Rural Livelihood Diversification in Sub-Saharan Africa: A Literature Review. The Journal of Development Studies 51, 1125–1138. Loschmann, C., 2019. Considering the benefits of hosting refugees: evidence of refugee camps influencing local labour market activity and economic welfare in Rwanda. IZA Journal of Development and Migration 9, https://doi.org/10.1186/s40176-018-0138-2. Martin, S.M., Lorenzen, K., 2016. Livelihood Diversification in Rural Laos. World Development 83, 231–243. Maystadt, J.-F., Verwimp, P., 2020. Winners and Losers among a Refugee-Hosting Population 62, 769–809. Naschold, F., 2012. “The Poor Stay Poor”: Household Asset Poverty Traps in Rural Semi-Arid India. World Development 40, 2033–2043. Newsham, A., Kohnstamm, S., Naess, L.O., Atela, J., 2018. Agricultural commecialization pathways: climate change and agriculture. Agriculture Policy Research in Africa Working paper #09. Nielsen, Ø.J., Rayamajhi, S., Uberhuaga, P., Meilby, H., Smith-Hall, C., 2013. Quantifying rural livelihood strategies in developing countries using an activity choice approach. Agricultural Economics 44, 57–71. Ochieng J., Knerr B., Owuor G., Ouma E. 2019. Food crops commercialization and household livelihoods: Evidence from rural regions in Central Africa. Agribusiness 36, 318-338. PEN (2007). The PEN technical guidelines. Bogor, CIFOR. Available at: https://www2.cifor.org/pen/the-pen-technical-guidelines/. (accessed on March 05, 2021). Pender, J., Jagger, P., Nkonya, E. and Sserunkuuma, D., 2004. Development pathways and land management in Uganda. World Development 32, 767–792. Ruiz, I., Vargas-Silva, C., 2020. The impacts of refugee repatriation on receiving communities 0, 1–26. Ruiz, I., Vargas-Silva, C., 2018. The impact of hosting refugees on the intra‐household allocation of tasks; A gender perspective.pdf. Review of Development Economics 22, 1461–1490. Ruiz, I., Vargas-Silva, C., 2015. The Labor Market Impacts of Forced Migration. American Economic Review: Papers & Proceedings 105, 581–586. Scoones, I., 2015. Sustainable livelihoods and rural development. Practical Action Publishing, Rugby. Tatah, L., Delbiso, T.D., Rodriguez-Llanes, J.M., Cuesta, J.G., Guha-Sapir, D., 2016. Impact of Refugees on Local Health Systems: A Difference-in-Differences Analysis in Cameroon. PLOS ONE 11(12), e0168820. UNHCR, 2021a. Global Trends; forced displacement in 2020. Copenhagen, Denmark: Statistics and Demographics Section UNHCR Global Data Service. https://www.unhcr.org/60b638e37/unhcr-global-trends-2020; accessed December 26, 2021). UNHCR, 2021b. Tigray situation update. https://reliefweb.int/sites/reliefweb.int/files/resources/UNHCR%20Ethiopia%20Tigra y%20Update%20%2313.pdf; accessed December 26, 2021) Valdivia, C., Elizabeth, D., & Christian, J. 1996. Diversification, a risk management strategy in an Andean agropastoral community. American Journal of Agricultural Economics, 78, 1329–1334. 23 Vemuru, V., Sarkar, A., Woodhouse, A.F., 2020. Impact of Refugees on Hosting-Communities in Ethiopia. A Social Analysis. The World Bank, Washington, DC. Verme, P., Schuettler, K., 2021. The impact of forced displacement on host communities: A review of the empirical literature in economics. Journal of Development Economics 150: 102606. Walelign, S.Z., Jiao, X., 2017. Dynamics of rural livelihoods and environmental reliance: Empirical evidence from Nepal. Forest Policy and Economics 83, 199–209. https://doi.org/10.1016/j.forpol.2017.04.008 Wang Sonne, S. E., Verme, P. 2019. Intergenerational Impact of Population Shocks on Children’s Health Evidence from the 1993–2001 Refugee Crisis in Tanzania. Policy Research Working Paper #9075. World Bank, Washington, DC. Watol, B.S., Assefa, D.T., 2018. The Socio-Economic Impact of Refugees on the Neighboring Countries: The Case of Sherkole Refugee Camp, Western Ethiopia. Global Journal of HUMAN-SOCIAL SCIENCE: E Economics 18, 36–48. Wunder, S., Börner, J., Shively, G., Wyman, M., 2014. Safety Nets, Gap Filling and Forests: A Global-Comparative Perspective. World Development 64, S29–S42. 24 Appendices Appendix A: Data collection We used a semi-structured survey questionnaire to generate the required data. The questionnaire has four modules. The first module collected information about the demographic characteristics (e.g., education, age, occupations) of individual household members. The second module collected information on sustainable environmental management practices, which include irrigation, soil, and water conservation. The third module collected information regarding household’s livelihood attributes, including livestock and land-based assets, crop production, use and sale; livestock production and sale; income from non-farm activities, access to inputs, technologies, extension services, and credit; and membership and participation in community organizations. The fourth module collected information on presence of and access to social and economic services and infrastructures, including heath, education, market, water, energy, and roads. All the modules, except the second, were collected at the household level: the second module was collected at the individual household member level. The questionnaire was translated to the local languages by professional editors, and field tested prior to full-scale implementation. Both the field-testing and the actual data collection were carried out using Computer-Assisted Personal Interviewing (CAPI) mode of data collection. Participants in data collection (supervisors, enumerators, and survey managers) were trained both in class and field on how to administer the questionnaire using the CAPI tablet interface. 25 Appendix B: description and summary statistics of the variables used in the analysis Variable Description Mean Panel A: Variables used for instrument and explanatory variable of interest construction Refugee camp distance Continuous: distance of refugee camps to the nearest border of the border of 53921.02 the nearest region of a source country in meters (32061.04) [3438.80, 138488.60] Population density Continuous: population density in the nearest region in the source country 94.354 (181.063) [6.464, 698.271] Number of conflicts Continuous: number of battles and violence against civilians in the nearest 116.461 region of source countries between 2001 and 2018 (171.478) [7, 674] Number of conflict Continuous: number of deaths battles and violence against civilians in the nearest 366.946 fatalities region of source countries between 2001 and 2019 (474.277) [8, 1939] Household distance Continuous: household distance to the nearest refugee camp in meters 13504.64 (12775.02) [65.630, 76632.78] Refugee population Continuous: the refugee population living in the nearest refugee camps in 2017 23263.61 and 2018, on average (19598.36) [3671, 83658] Panel B: Instrument Potential refugee inflow Continuous: the product of population density and intensity of conflicts (number 0.028 of fatalities per event) in the closest region of the origin country to (0.070) the refugee camp weighted by the distance of the refugee camp to [9.060X10-5, 0.333] the closest region Panel C: Variables used for explanatory variables of interest Refugee inflow Continuous: the number of refugees (population) in the nearest refugee camp 5.592 to the household location weighted by household’s inverted (14.329) distance to the camp [0.198, 195.364] Panel D: Explanatory variables of interest Refugee inflow Continuous: the refugee population weighted by the inverse of distance to the 1.562 nearest refugee camp (1.078) [0.197, 5.968] Panel E: Outcome variables of interest Diversification in Continuous: diversification of livelihood activities as a secondary occupation 1.596 primary occupation (.577) [0, 4] Diversification in Continuous: diversification of livelihood activities as a primary occupation 0.928 secondary occupation (0.805) [0, 4] Livestock Continuous: aeu adjusted value of livestock product sale 665.606 commercialization (2079.695) [0, 54346.2] Crop commercialization Continuous: aeu adjusted value of crop product sale 237.497 (1210.527) [0, 42314.29] Panel F: Control variables Group membership Continuous: the number of community associations and/or organizations that 0.322 the household is a member of (0.467) [0, 1] Productive load Continuous: aeu adjusted loan for productive purposes 115.869 (1022.659) [0, 33333.33] Land owned Continuous: aeu adjusted land owned by the household in hectares 0.344 (0.514) [0, 14.4] Livestock value Continuous: aeu adjusted total value of livestock owned by the household 30401.01 (45351.4) [0, 563000] Distance to all weather Continuous: distance of the household to the nearest all the weather road in 60.777 road minutes (149.761) [0, 990] Head education Continuous: the education level of the household head 2.909 (4.576) [0, 17] Head gender Dummy: gender of the head of the household 0.163 (0.370) [0, 1] Head marital status Dummy: marital status of the head of the household 0.808 (0.394) [0, 1] Values in parenthesis are standard deviations of the mean; values in square brackets are the minimum and maximum values. 26 Appendix C: first-stage regression coefficient and summary statistics for the household level analysis (1) (2) Without With controls controls IHS(Refugee presence) -0.569*** -0.487** (0.178) (0.199) IHS(number of groups a household is a member) -0.041* (0.023) IHS(loan for productive purpose in Birr, aeu adjusted) 0.003 (0.004) IHS(total land owned by a household in hectares, aeu adjusted) -0.083*** (0.024) IHS(total livestock owned by a household in Birr, aeu adjusted) 0.000 (0.002) IHS(distance to all-weather road in minutes) -0.019* (0.011) IHS(household head education in number of years attended) 0.006 (0.005) Gender of the household head (1=female) -0.083*** (0.029) Marital status of the household head (1=married) 0.186*** (0.030) N 3375 3375 Kebele fixed effects Yes Yes F (1, 112) 23.20*** 17.85*** Kleibergen-Paap rk Wald statistic, X2(1) 21.86*** 19.02*** Kleibergen-Paap Wald rk F statistic 20.94 18.18 27 Appendix D: proportion of individuals engage in different occupations as a primary and secondary activity. Livelihood activities Primary Secondary Singifican # of individual Percentage # of individual Percentage ce (1) (2) (3) (4) difference (Z-test) (2) vs (4) Salaried 559 7.57 75 1.45 10.71*** employment Housework 2549 34.51 2155 41.8 -5.14*** Own business 563 7.62 189 3.67 9.94*** Crop production 2746 37.18 904 17.53 10.96*** Livestock 910 12.32 1792 34.76 -12.39*** production Other 59 0.8 41 0.8 0.79 Overall 7386 100 5156 100 - 28 Appendix E: First-stage regression summary statistics for regional and gendered subsample regressions for the household level analysis Region Robust F (1,1208) Panel A: Region Tigray 749.21*** Afar 490.86*** Somalia 26.02*** Benishangul-Gumuz 4.68** Gambella 6.61** Panel B: Gender Female 58.35*** Male 26.94*** ***significant at 1%;**significant at 5% *significant at 10% 29