\ Photo Credit: Afghan refugee families waiting to return home at UNHCR’s voluntary repatriation centre in Chamkani, Peshawar in the Islamic Republic of Pakistan. Upon reaching Afghanistan they will receive a cash grant to help them reintegrate, along with medical care, mine risk awareness, and advice on accessing education, civil documentation, and available land. ©UNHCR /S. Rich Table of Contents Acknowledgements ....................................................................................................................................... 4 Key Findings................................................................................................................................................... 5 1. Introduction .......................................................................................................................................... 7 Research Objectives .............................................................................................................................. 8 2. Data sources and approach ................................................................................................................... 9 3. The return decisions and pre-return profile of registered Afghan refugees ....................................... 10 Individual and household characteristics ............................................................................................ 10 Living conditions in Pakistan ................................................................................................................ 11 Educational attainment and attendance ............................................................................................. 12 Labor market attachment of Afghan refugees in Pakistan .................................................................. 13 The decision to return: Regression analysis ........................................................................................ 14 4. The settlement decisions of Afghan returnees.................................................................................... 15 Basic characteristics of post-2013 returnee households ..................................................................... 15 Where do returnees settle? ................................................................................................................ 16 5. Livelihoods after return ....................................................................................................................... 18 6. Conclusion ........................................................................................................................................... 25 Sources ........................................................................................................................................................ 26 Appendix 1: Data sources ............................................................................................................................ 27 Appendix 2: Propensity Score Matching...................................................................................................... 28 Appendix 3: Additional data tables.............................................................................................................. 30 3 Acknowledgements This report is the result of a collaboration between the United Nations High Commissioner for Refugees (UNHCR) and the Poverty and Equity Global Practice of the World Bank Group (WBG). The core team consisted of Nandini Krishnan (WBG), Jeffrey Savage (UNHCR), Christina Wieser (WBG), and Thea Yde- Jensen (WBG). This effort is part of a joint data sharing and analysis collaboration between the World Bank, the UN High Commission for Refugees and the International Organization for Migration in Afghanistan. The WB phone survey was financed by a grant from the UK Department for International Development (DFID) Trust-Fund on forced displacement to generate evidence on displacement; and their support is gratefully acknowledged. Data collection for the World Bank Phone Survey was implemented by Orange Door Research in collaboration with Viamo. We are grateful to Michael Kleinman, Nicholai Lidow, and Kunwar Prashant Singh along with the rest of the survey team for an excellent collaboration. We are grateful to our WBG peer reviewers – Tara Vishwanath and Dhiraj Sharma – for their insightful comments and suggestions on this paper, Hisham Esper who diligently provided his knowledge on the determinants of return, and UNHCR who shared their in-depth knowledge of the data at various stages of this project. Finally, we would like to thank Aurvasi Patel, Deputy Representative for UNHCR Afghanistan, and Shubham Chaudhuri, WBG Country Director for Afghanistan, for their support and guidance on this project. 4 Key Findings Afghan refugees who returned to Afghanistan after 2014 tend to be worse off than those who remained in Pakistan and tended to be those who have retained ties to their native country. More specifically, Afghan refugees who returned between 2014 and 2017 were less wealthy, lived in refugee villages in Pakistan or temporary housing, had previously (5+ years earlier) considered repatriating, and visited Afghanistan regularly. However, there is some evidence that registered Afghan refugees with at least some formal education were more likely to return. Once in Afghanistan, most refugees returned to their province of origin, valuing proximity to family and friends even though these provinces tended to have lower employment rates and higher poverty rates. Given the importance of networks in finding jobs, the benefits of proximity to family may have outweighed the costs due to adverse local economic conditions. Indeed, Afghans living in their province of origin were more likely to be employed albeit less likely to have the same type of job as they did prior to returning. For those returnees who settled somewhere other than their province of origin, considerations of safety and access to economic opportunity were the main drivers. These provinces of return also tended to be more urbanized. Returnees generally experience a deterioration in the employment opportunities, wages, and job stability after returning to Afghanistan. Although most families have at least one person working for 5 pay, they have low job stability and low wages. Most returnees in the survey work as daily wage laborers in non-agricultural sectors. With large family sizes, and low reported incomes, it is likely that many returnees are living close to or below the poverty line. However, access to education improves post return for both boys and girls and the gender-gap in school attendance is reduced, driven by an increase in the number of households where all girls attend school. These findings document the economic difficulties faced by Afghan returnees, coming back after several decades to an economy that is producing few job opportunities. Social ties play an important role in finding work, somewhat compensating for return to provinces of origin that tend to be currently poorer, less urban, and with higher joblessness. Kabul and Nangarhar alone account for a third of all returnees; and returnees who do not settle in their province of origin move to relatively urban areas in search of safety, services and jobs. This tendency is likely putting additional pressure on urban and peri-urban areas, which are already hosting internally displaced populations. In addition to the availability of work, access to health and education is another critical need. While school attendance is higher post-return, educational attainment remains low among returnees. These challenges apply widely to the hosting population as well as to internally displaced persons. This report did not address returnee-specific needs related to access to land, shelter and documentation, which are likely to be important given the time away from Afghanistan. Secondary movements are also beyond the scope of this work: given the repeated exposure to risk and the reliance on migration as a coping strategy, such displacements are likely and risk further eroding precarious livelihoods. 6 1. Introduction Afghan citizens form one of the largest protracted refugee populations in the world; many Afghans have lived in exile for decades. A vast majority of registered Afghan refugees currently living in neighboring Pakistan and Iran were born and raised in these two countries of asylum - about 2.4 million Afghan citizens including some 1.4 million registered refugees, around 880,000 Afghan Citizen Card (ACC) holders and an estimated 300,000-500,000 undocumented Afghan nationals, most of whom are under the age of 301. Despite the ongoing conflict, roughly 2.4 million displaced Afghans have returned to Afghanistan since 2014 including over half a million registered refugees who returned under UNHCR’s facilitated voluntary return program and received immediate humanitarian assistance. Most returnees came from Pakistan and the number of returnees increased enormously in 2016; subsequently, voluntary returns to Afghanistan were significantly lower, including some 15,700 refugees that voluntarily returned in 2018. At the height of the wave of returns to Afghanistan in 2016, little was known about the settlement patterns and needs of the more than 2 million Afghan returnees who had come back to Afghanistan since 2014. The wave of mass return witnessed during this timeframe does not appear be explained solely by voluntary return patterns2. Rising political tensions towards Afghan refugees in Pakistan, especially following the 2014 Peshawar attacks, have likely increased pressure to repatriate the Afghan refugee populations in north- west Pakistan in particular. As a response to large influxes of returnees starting in early 2016, UNHCR also doubled its repatriation assistance amount in June 2016 to ensure the dignified reintegration of returnees, which could also have created a secondary incentive for Afghan refugees to return from Pakistan. Registration efforts during this period were led by UNHCR, IOM, and other agencies, who managed registration databases for diverse groups. Meanwhile, existing data sources had not been fully exploited to characterize diverse groups of Afghan refugees returning to Afghanistan from Pakistan. It was in light of this knowledge gap that the World Bank (WB) and the United Nations High Commissioner for Refugees (UNHCR) formalized a joint data sharing and analysis agreement to (i) better understand where the Afghan returning refugee, and internally displaced populations currently reside; (ii) understand how population movements could be tracked in preparation for the next wave of returns and displacement; and (iii) inform better-targeted humanitarian responses and development assistance for returned Afghans through a community based approach3. In a context of ongoing conflict, low economic growth and job creation relative to population growth, a major policy concern is how to absorb and reintegrate such significant number of returnees. The difficult security environment in Afghanistan reduces the humanitarian space and poses severe constraints on the feasibility of field data collection and implies that any analytical work must rely on existing data sources as well as pragmatic innovative approaches to new data collection. Given the paucity of knowledge on repatriation, a collaboration of the WB, UNHCR, and IOM, financed by a Trust Fund provided by United Kingdom’s Department for International Development (DFID) on forced displacement, complements UNHCR’s and IOM’s existing monitoring and data collection efforts, by collecting new data through 1 Ministry of States and Frontier Regions, the Office of the Chief Commissioner for Afghan Refugees and UNHCR (2012). 2 Forthcoming World Bank and UNHCR report. 3 A similar data sharing and analysis agreement is now in place with the International Organization for Migration (IOM) and the WB. 7 household level phone interviews. The DFID funding allowed expanding the sample of returnees interviewed in an attempt to generate a more representative picture of recent returnees. This mobile phone survey of returnees was implemented in two stages. In the first, random-digit-dialing was complemented by integrated voice recognition software for the identification of a broad sample of recent returnees with mobile phones. In a second stage, the sample thus identified was administered a follow-up personal phone interview on location decisions, employment and livelihoods, and access to basic social services4. Research Objectives Repatriation or the return of refugees to their country of origin has been rarely studied, and data on their socio-economic outcomes is sparsely available. In such a context, the World Bank and UNHCR teams attempted to make good use of the existing data sources and complemented it with new data collection methods to better understand the patterns and characteristics of recent Afghan refugee returns. More specifically, the team attempted to analytically connect insights between different data sources to explore (albeit imperfectly) questions of selection among Afghans who remained in Pakistan and those documented returnees who returned to Afghanistan. This report’s objective is to primarily describe and analyze the living conditions of the large Afghan refugee population who had returned from Pakistan in 2014 or after, with a particular emphasis on documented returnees5. This work contributes to our understanding of the return process, the influence of household characteristics and prevailing economic and security considerations and living conditions of Afghans who returned to Afghanistan. It also provides insights into the design of an appropriate response for future Afghan returns and of potential returns of other refugee populations to their country of origin. This note includes three types of findings: (i) factors affecting the decision of return; (ii) the decision of where to settle in Afghanistan; and (iii) an assessment of returnee livelihoods upon return to Afghanistan. First, the decision of return is analyzed using the Afghan Population Profiling, Verification and Response (PPVR) dataset, a large-scale survey of Afghan refugees living in Pakistan in 2011. We compare characteristics of documented Afghan refugees who lived in Pakistan in 2011 and returned to Afghanistan between 2014 and 2017 with a large sample of (then) documented Afghans in Pakistan in 20116. Second, the analysis of returnees’ choice of destination is analyzed using the World Bank Phone Survey (WBPS) of post-2013 returnees, collected by the World Bank Group in 2018. Finally, the assessment of livelihoods of post-2013 returnees relies on a combination of the World Bank Phone Survey (WBPS) and the PPVR dataset by combining the two data sources using propensity score matching methods7 and comparing the average change in key characteristics in the PPVR (2011) and the WBPS (2018). 4 A separate note outlines the detailed methodology of the data collection effort which can be accessed at: http://documents.worldbank.org/curated/en/298881533562809348/Afghanistan-World-Bank-Phone-Survey-of- Afghan-Returnees-Methodology-and-Representativeness. 5 With the implementation of a data sharing agreement with IOM, undocumented returnees will be covered in future analytical work. 6 Some of the unregistered/undocumented Afghans in the 2011 survey were subsequently issued documentation. 7 Propensity score matching is a statistical matching technique that attempts to estimate the effect of an intervention. 8 2. Data sources and approach This note sheds light on three aspects of Afghan refugees’ return to Afghanistan: (i) the decision to return; (ii) the choice of destination; and (iii) livelihoods upon return. We rely on three data sources to do so; the Afghan Population Profiling, Verification and Response survey (PPVR), the Voluntary Repatriation Form survey (VRF) and a recently collected World Bank Phone Survey (WBPS)8. First, we examine the characteristics of Afghan refugees who returned to Afghanistan from 2014 to 2017 and how they differ in terms of socio-economic and living conditions from Afghan refugees who did not return. We examine whether socio-economic characteristics and personal experiences influence the decision to return. The analysis of Afghan refugees’ decision to return is approximated by comparing Afghan refugees in Pakistan who returned to Afghanistan between 2014 and 2017 with Afghan refugees who remained in Pakistan. The analysis relies on the PPVR dataset and is restricted to Afghan refugee returnees who are registered by UNHCR due to lack of data of undocumented returns on an individual or household level. In total, we have information in the PPVR survey on 507,000 registered Afghans who lived in Pakistan in 2011, of which 125,000 returned to Afghanistan between 2014 and 2017. Information on the 125,000 returnees is available in the VRF. Second, we look at the choice of destination of Afghans who returned to Afghanistan since 2014. We examine post-2013 returnees’ decision of preferred destination using the WBPS and include documented as well as undocumented returnees in the analysis. Finally, we examine the living conditions of post-2013 returnees. First, a descriptive analysis of documented and undocumented post-2013 returnees’ labor market outcomes is carried out using the WBPS. To assess the change in living conditions of post-2013 returnees after their return to Afghanistan, registered returnees in the PPVR are matched with documented post-2013 returnees in the WBPS applying propensity score matching methods. More specifically, we assess the changes in key indicators by matching documented returnees in the WBPS to the nearest ten registered refugees in the PPVR (who return to Afghanistan between 2014 and 2017) with the highest likelihood of sharing similar characteristics with the WBPS sample based on age9, gender, educational level10, job type (prior to returning to Afghanistan), and province of origin11. Propensity score matching allows approximating the change in key characteristics prior to and post return as well as matching bread winners in the WBPS with individuals in the PPVR dataset based on characteristics that likely determine the bread winners of a household. For more information on the propensity score matching, see Appendix 1. Table 1 summarizes the data sources along with the registration status of the target population for each of the three analyses. 8 All three data sources are described in detail in Annex 1. The WBPS interviewed 3,575 post-2013 returnees 9 Due to few large values of age in the PPVR, age is winsorized at the 99.99 percentile. This means that all values of ‘age’ above the 99.99 percentile are set to the value of the 99.99 percentile. 10 5.6 percent of individuals older than 5 years in the PPVR who do not know their level of education. This category is replaced with missing. 11 We differentiate between ‘Nangarhar’, ‘Kabul’ and ‘other’. 9 Table 1: Analyses, data sources and comparison groups (1) Decision to return (2) Decision to settle (3) Livelihood upon return Data sources(s) PPVR and VRF WBPS WBPS and PPVR Target groups Registered refugees All returnees All returnees for descriptive analysis and documented returnees for comparative analysis 3. The return decisions and pre-return profile of registered Afghan refugees In this section, we examine the characteristics of documented Afghan refugees12 who returned to Afghanistan between 2014 and 2017 and how they differ from Afghan refugees who did not return. We also examine whether socio-economic characteristics and personal experiences influence the decision to return. The findings show that recent documented returnees resembled the average Afghan in Pakistan on a range of characteristics, with a few notable exceptions. Most returnees were either born in Pakistan or arrived during the Soviet invasion in the late 1970s and lived in temporary housing arrangements in refugee villages in Pakistan. Both documented returnees and registered Afghan refugees in general had no or low levels of education; and while women generally did not work, most men were employed as daily laborers in low skilled and non-agricultural jobs in Pakistan. We generally observe large gender disparities in educational attainment as well as school attendance while in Pakistan. In particular, while many Afghan boys attended school, this was not the case for Afghan girls in Pakistan. Despite these similarities, recent documented returnees also differ from the average Afghan refugee in Pakistan in three ways. First, fewer recent returnees were working age men and more were women. Second, Afghan refugees were more likely to return if they were less well-off economically (approximated by variables indicating whether a household owns a car and income stability), lived in refugee villages, and did not own their dwelling in Pakistan. Finally, Afghan refugees were more likely to return if they had previously considered returning and had kept ties with Afghanistan through regular visits. Individual and household characteristics Of the 507,000 registered Afghan refugees in Pakistan surveyed in the PPVR in 2011, 125,000 returned to Afghanistan between 2014 and 2017. There were as many men as women among Afghans living in Pakistan in 2011 but more women (54 percent) than men returned to Afghanistan between 2014 and 2017 (Table 2). 12 As we only have access to UNHCR data, we can undertake this analysis only for registered returnees which are under UNHCR’s mandate. However, the World Bank Group recently signed a data sharing agreement with IOM which will allow for similar analysis on undocumented Afghan undocumented returnees in the future which will include analysis on whether those who returned are different from those who remained among the population of undocumented Afghan returns. 10 Table 2: Gender distribution of documented Afghan refugees in Pakistan vs. Afghan refugees in 2011 2014-17 documented Afghan refugee returns Afghan refugees in Pakistan Female 54% 51% Male 46% 49% Total 100% 100% Source: Staff calculations using VRF and PPVR for documented Afghan returnees and PPVR for Afghan refugees in Pakistan. Note: Only documented refugees are included. Afghan refugees in Pakistan were generally young; 54 percent were younger than 18 years old and 44 percent were between 18 and 64 years old. Compared with the overall Afghan registered refugee population, females are over-represented, while young and working age males were under-represented among recent returnees (Figure 1). Figure 1: Age distribution by gender of Afghan refugees in Pakistan in 2011 Afghans in Pakistan 2015-17 Afghan returnees 65+ 18-64 Female 12-17 5-11 0-4 65+ 18-64 Male 12-17 5-11 0-4 0% 5% 10% 15% 20% 25% 30% Source: Staff calculations using VRF and PPVR for documented Afghan returnees and PPVR for Afghan refugees in Pakistan. Note: Only documented refugees are included. Ages reported in Figure 1 correspond to the age of Afghan refugee respondents in 2011. Recent documented returnees resembled the average Afghan refugee in Pakistan in terms of household size and composition. The median household included 7 household members13 among both groups. Moreover, most households (95 percent) consisted of at least two working age adults, less than 1 percent of households did not include a working age adult and the average dependency ratio of Afghan households in Pakistan was 1.7. In summary, the overall household characteristics of the recent documented returnees from Pakistan closely resemble the characteristics of the Afghan refugees in Pakistan. Living conditions in Pakistan Afghan refugees have been living in Pakistan for decades. While 70 percent of individuals were born in Pakistan, 87 percent of Afghan households arrived in Pakistan during the Soviet invasion between 1979 and 1989. Despite having lived in Pakistan for decades, most Afghans (54 percent) lived in refugee villages (Table 13 The average Afghan household living in Pakistan in 2010/11 consisted of 7.4 individuals whereas the average household of recent returnees consisted of 7.8 individuals. 11 3), with a significantly higher proportion among those who arrived during the Soviet Invasion living in refugee villages, compared to Afghans who arrived later. Recent documented returnees were generally more likely to live in refugee villages than the average Afghan refugee in Pakistan (Table 3). Table 3: Type of living location in 2011 2014-17 documented Afghan refugee returns Afghan refugees in Pakistan Refugee villages 67% 54% Urban 25% 32% Rural 8% 14% Source: Staff calculations using VRF and PPVR for documented Afghan refugee returnees and PPVR for Afghan refugees in Pakistan. Note: Only documented refugees are included. Calculated per household. May not sum to total due to rounding. The type of areas that refugees live in affects living conditions such as dwelling type and access to sanitation. Most (72 percent) registered Afghan refugees lived in mud houses or other temporary housing arrangement in Pakistan and, this was especially true for Afghans living in refugee villages (91 percent) or rural areas (59 percent). Recent documented refugee returnees were slightly more likely (4 percentage points) to live in temporary housing than the average Afghan refugee in Pakistan (Table 4), which seems to be related to a higher likelihood of living in refugee villages. Table 4: Type of dwelling in 2011 2014-17 documented Afghan refugee returns Afghan refugees in Pakistan Apartment or pakka house 12% 15% Semi-pakka 11% 13% Mud-house, temporary housing 77% 72% Source: Staff calculations using VRF and PPVR for documented Afghan refugee returnees and PPVR for Afghan refugees in Pakistan. Note: Only documented refugees are included. Other includes hut, temporary shelter/plastic, on the street/in the open. Calculated per household. May not sum to total due to rounding. A pakka house refers to dwellings that are built with solid materials such as stone, brick, cement, concrete, or timber. Few Afghan refugees in Pakistan owned their dwelling; only 18 percent of registered Afghan refugees in Pakistan and 13 percent of recent documented returnees owned their dwelling, a difference in rate of ownership which persists across location (refugee villages, rural or urban). Recent refugee returnees were more likely to rent their dwelling compared to the average Afghan refugee in Pakistan when controlling for living area (see Appendix 2, Table A.3). There are no systematic differences between recent documented returnees and the average registered Afghan refugee in Pakistan in terms of access to water and sanitation facilities. Educational attainment and attendance On average, the educational attainment of registered Afghan refugees in Pakistan was very low; recent documented returnees had similarly low rates of educational attainment. Large gender disparities in educational attainment were prevalent; less than 16 percent of Afghan working age women among this group had any formal education, while this number was twice as high (36 percent) among working age men 12 (Table 5). We also observe large disparities across age groups, with higher rates of educational attainment for young Afghans (aged 18-29) compared to their older counterparts. Recent documented Afghan refugee returnees generally had the same levels of educational attainment compared to the registered Afghan refugee population in Pakistan. Only 41% of Afghan returnee males had any formal education, and among women, this rate was only 22%. (Table 5). Just 4 percent of returning Afghan men and 1 percent of Afghan women had a higher education than secondary school. Table 5: Level of competed education by gender (12+ years) in 2011 2014-17 documented Afghan Afghan refugees in Pakistan refugee returns Males None 59% 60% Primary school 16% 15% Middle or sec. school 12% 11% Tech./college/uni. 4% 4% Religious school 5% 6% Other/Don’t know 4% 4% Total 100% 100% Females None 78% 78% Primary school 4% 4% Middle or sec. school 2% 3% Tech./college/uni. 1% 1% Religious school 8% 8% Other/Don’t know 7% 6% Total 100% 100% Source: Staff calculations using VRF and PPVR for registered Afghan refugee returnees and PPVR for registered Afghan refugees in Pakistan. Note: Only registered refugees are included. Not surprisingly, literacy rates and formal school attendance go hand in hand. Nearly all Afghan refugees in Pakistan with some level of formal schooling were literate, however, the literacy rate of those attending religious schools was only 42 percent, leading to large gender and age disparities in literacy rates and slightly lower literacy rates compared to educational attainment. School attendance of registered Afghan refugee children in Pakistan reflects the gender disparities in educational attainment; about 67 percent of Afghan boys between the age of six and fifteen attended school, but only 35 percent of girls did. We do not observe noticeable differences in school attendance of recent documented returnees and the general Afghan refugee population in Pakistan in 2011. Labor market attachment of Afghan refugees in Pakistan At first glance, registered Afghan refugee men have a strong attachment to the labor market compared to women; 85 percent of registered Afghan refugee men in Pakistan and recent registered Afghan refugees were employed (Table ). Labor market attachment for women, on the other hand, was weak; only 13 percent of registered Afghan refugee women were employed with an even lower share for recently returned refugee women at 10 percent. Despite high employment rates, refugees are employed in vulnerable professions, with the majority of registered Afghan refugee men employed in jobs with relatively low job security (with similar results for 13 recent documented returnees as well as all registered Afghans refugees in Pakistan) (Table 6). Most recently returned men worked as day laborers (45 percent) or own-account workers (22 percent) while in Pakistan in 2011 whereas only 16 percent had salaried jobs and less than 0.5 percent were employers. Table 6: Type of employment of working age (18-64 years) men in 2011 Men, working age (18-64 year) 2014-17 documented Afghan refugee Afghan refugees in returns Pakistan Day laborer 46% 45% Salaried worker 16% 16% Own-account (self-employed, independent) 21% 22% Employer 0% 0% Unpaid family worker 1% 2% Not employed 16% 15% Total 100% 100% Source: Staff calculations using VRF and PPVR for registered Afghan returnees and PPVR for registered Afghan refugees in Pakistan. Note: Only registered refugees are included and individuals in school excluded. Corresponding to the predominant job types of day labor or self-employment, most Afghan refugees in Pakistan held un- or low-skilled jobs. About 32 percent of Afghan males held elementary jobs such as garbage collectors, building construction laborers, or unskilled agricultural laborers and 23 percent worked in sales or services such as stall and market personnel, shop keepers, shop assistants, etc. (see Appendix 2, Table A.4). There were no noticeable differences in the type of occupation held by recent documented returnees and the Afghan refugee population in Pakistan. The decision to return: Regression analysis In this section, we estimate the probability of registered Afghan refugees of returning to Afghanistan in a probit model14 (Appendix 2, Table A.5). This allows for the correlation of household characteristics, as well as an assessment of the relative importance of individual characteristics on the likelihood of returning to Afghanistan. We further extend this analysis by examining whether socio-economic characteristics or personal experiences also influence the decision to return. The regression results reiterate many of the findings from our descriptive analysis; females, smaller households, and households with many children were more likely to return to Afghanistan. While the effects of education on the probability of return are less clear-cut, Afghans with some level of formal education are more likely to return than Afghans with no education. The estimation results further show that living in refugee villages or living in households with one or more disabled household members increases the likelihood of return. Wealth and income stability of the household are approximated by binary variables indicating whether a household owns a car and whether there were variations in the number of daily meals throughout the past month. The regression results 14 A probit model refers to a type of regression model in which the dependent variable can take only two values, in our case a dummy variable for whether Afghan refugees returned (1) or not (0). 14 indicate that household wealth and income stability are negatively correlated with the likelihood of return to Afghanistan. Overall, these findings suggest that economically vulnerable households in poor living conditions were more likely to return to Afghanistan. The coefficients also show that living in refugee villages (positive) and owning a house or an apartment (negative) in Pakistan are the largest predictors of return. Interestingly, Afghans who had previously considered returning or had retained ties to Afghanistan were also more likely to return which may indicate that households, who had more information about potential livelihood opportunities and had a network in Afghanistan to help re-settle, were more likely to return. Moreover, Afghans who kept ties with Afghanistan through regular visits of household members were more likely to return. Surprisingly, those who own property in Afghanistan appear to have a lower likelihood of return, which may be explained by households with property in Afghanistan being better off than refugees who do not own property in Afghanistan. The findings in this section show that recent documented returnees come from slightly worse living conditions than the population of registered Afghan refugees in Pakistan. In general, recent returnees were more likely to have lived in refugee villages and therefore more likely to live in temporary housing arrangements and less likely owning their dwelling. But those with some formal education were more likely to return than those with none, although educational attainment and school attendance was generally low among all registered Afghan refugees. Although this was especially true for registered refugee women and for girls, 75 percent of registered refugee males had completed primary school or less. Personal characteristics, living conditions and previous experiences all influenced returnees’ decision to return. Individuals and households that were somewhat worse off in Pakistan were more likely to return. Moreover, Afghans who had previously considered returning or had retained ties to Afghanistan were also more likely to return which may indicate that household who had researched their livelihood opportunities and had a network in Afghanistan to help re-settle, were more likely to return. 4. The settlement decisions of Afghan returnees In this section, we look at returnees after they have made the decision to return and crossed the border into Afghanistan. The section presents basic characteristics of post-2013 returnees (registered Afghan refugees and undocumented Afghan nationals returning to Afghanistan in 2014 or thereafter) and examines their re-settlement patterns. The analyses are based on weighted household level data collected via the World Bank Group in a phone survey. Basic characteristics of post-2013 returnee households Returnee households tend to be large and most households return to Afghanistan with children. The median (average) returnee household consists of 9 (9.6) household members15 and more than 50 percent of household members are between 0 and 16 years old (Table 7)16. 15 The national median (average) household size in Afghanistan is 7 (7.7) members based on the Afghanistan Living Conditions Survey of 2016-17. Household sizes from different data sources differ. For example, the VRF shows an average household size of 7.8 while the WBPS shows a household size of 9.6. 16 The median (average) household includes 5 (5.0) children. 15 Table 7: Share of children Post-2013 returnees Share of children in household 0% 3% ]0%-25%] 7% [25%-50%] 38% [50%-75%] 46% [75%-100%] 6% Total 100% Source: Staff calculations using the WBPS. Note: May not sum to total due to rounding. The majority (42 percent) of post-2013 returnees arrived in Afghanistan in 2016 (Figure 2) which coincided with (i) a deterioration of the protection environment for Afghan refugees in several areas of Pakistan 17 and (ii) a doubling of UNHCR’s repatriation cash grant – from USD 200 to USD 400 per person – which likely had a significant impact on the number of Afghan refugees who decided to return that year. Additional factors affecting return in 2016 were the introduction of border management at Torkham, and proactive advocacy by the Government of Afghanistan to encourage refugees to return. Four out of five returnees returning in 2014 or later did not leave anything behind in their host country and the most common belongings left behind were land/property/house (6 percent), household assets (5 percent), and cash/bank account (5 percent) (see Appendix 2, Table A.6). That returnees did not leave much behind is consistent with the findings in section Error! Reference source not found. which indicated that less well-off refugees were more likely to return. Figure 2: Year of return 42% 32% 12% 7% 7% 2014 2015 2016 2017 2018 Source: Staff calculations using the WBPS. Where do returnees settle? Many returnees interviewed in the WBPS repatriated to the eastern and central parts of Afghanistan, in particular, Nangarhar and Kabul provinces (Figure 3). These areas tend to be where post-2013 returnees originally came from, with 79 percent of all returnees settling in their province of origin and 60 percent settling in their district of origin (Figure 4). 17 Hosting communities and local authorities in Pakistan were less accepting of Afghans which resulted in a loss of self-reliance opportunities for Afghan returnees likely triggering a large-scale return. 16 Figure 3: Province of current settlement Figure 4: Living in place of origin 70% 79% 60% 19% 11% NANGARHAR KABUL OTHER CURRENT PROVINCE CURRENT DISTRICT PROVINCES IS PROVINCE OF IS DISTRICT OF ORIGIN ORIGIN Source: Staff calculations using the WBPS. Source: Staff calculations using the WBPS. Proximity to family and friends was the primary reason for returnees’ decision to settle in their province of origin (Table 8), followed by shelter, safety, and employment opportunities. Among those returnees who did not return to their province of origin, security concerns (35 percent) and lack of employment opportunities (25 percent) were the primary reasons for settling elsewhere. Table 8: Reasons for settlement For returnees settling in province of For returnees settling in provinces origin other than their province of origin Relatives/family/friends live here 53% 20% Have shelter/own house 20% 5% Safety/security 10% 35% Employment opportunities 6% 25% Have land 6% 3% Access to services (health, 4% 11% education, water, electricity) Other 1% 1% Total 100% 100% Source: Staff calculations using the WBPS. To shed more light on returnees’ decision to settle in or outside their province of origin, correlations between geographic and demographic characteristics and the probability of settling in the province of origin are examined in a probit model (Appendix 2, Table A.7). The regression results confirm responses of returnees presented in Table 8. Returnees who settled in their province of origin were more likely to be from provinces with relatively high poverty rates (based on results on poverty rates derived from the Afghanistan Living Conditions Survey 2016-17) and low employment rates. We interpret these findings as a willingness to compromise economic opportunities to be close to family and friends, which is also confirmed by the importance of networks in finding employment opportunities (more details in section 5). Regression results also show that returnees who did not settle in their province of origin were more likely to live in an urbanized province and in districts under government control, presumably seeking better/safer opportunities. Regression results further indicate that households which settle in their place of origin upon return have bread winners who are more likely to be employed but have a different type of job than they did in their 17 host country. This may reflect the usefulness of social networks when searching for jobs as well as the necessity to adjust to the employment situation upon return. The types of job held by Afghan returnees prior to and after their return are discussed in more detail in Section Error! Reference source not found.. Undocumented returnees, smaller households, and households with a smaller share of children were also more likely to settle in their province of origin. This was also true for Afghans returning in 2017 and 2018 (though the sample size is rather small in those years). Although we saw in section Error! Reference source not found. that refugees in Pakistan who returned between 2014 and 2017 were more likely to have previously considered returning to Afghanistan, these findings suggest that returnees who came to Afghanistan as a result of increased push factors were less likely to have family and other ties in Afghanistan, resulting in a smaller likelihood of returning to their province of origin. We observe that returnees who do not live in their province of origin tend to live in safer districts in more urban and less poor provinces with higher than average employment rates. However, the challenges faced by returnees living in their province of origin and those who do not are similar: about 68 percent of returnees indicate that the cost of living is too high or that they have difficulties finding jobs in the area they live in. However, despite experiencing challenges, mobility of returnees is low; 91 percent of returnee households indicate that they do not have plans to move and 90 percent live in the province that they first settled in upon returning to Afghanistan (Table 9). Of the returnees who plan to move, 6 percent do so in hopes of finding jobs or a lower cost of living. Table 9: Mobility of returnees No plans to move Plans to move Total Do not live in first place of settlement 9% 1%* 10% Live in first place of settlement 82% 8% 90% Total 91% 9% 100%/100% Source: Staff calculations using the WBPS. *Less than 30 observations (unweighted) In summary, most returnees decided to remain in their area of origin to live in proximity to family and friends. Returnees who do not return to their province of origin typically seek safer areas with better access to services and employment opportunities. 5. Livelihoods after return Afghanistan struggles with conflict, job vulnerability, and high rates of poverty. The peak of returns in 2016 coincided with increasing vulnerability of the Afghan population, with a sharp increase in poverty rates to 55 percent. At the same time, the challenging security situation over the past decades has led to large-scale population displacements. Internal displacement, displacement to other countries, and large-scale return within an unfavorable economic and security context pose risks to welfare, for the displaced and for the population at large. In this section, we examine the labor market attachment of returnee households (via the primary bread winner) to document the livelihood of returnees. To assess changes in the living situation of Afghans upon their return to Afghanistan, this section also compares key characteristics related to 18 livelihood and access to basic services such as education and health care for documented returnees, both prior and post return, using nearest neighbor propensity score matching18. In 9 out of 10 (87 percent) households, the primary bread winner of returnee households was employed and at least one male member worked for pay in 91 percent of returnee households (Table 10). However, only one man19 worked for pay in more than half (58 percent) of returnee households. With a typical monthly salary of 5,000 to 10,000 Afs per month (Figure 5) and a median household size of 9, we can assume that many returnee households live below the poverty line, as do many among the hosting population20. Table 10: Number of working adult men in Figure 5: Income of bread winner households 45% TOTAL 0 9% 23% 1 58% 14% 13% 2 20% 6% 3 9% 4+ 4% TOTAL 100% 1,500 - 3,001 - 5,001 - MORE DO NOT 3,000 5,000 10,000 THAN HAVE A AFS AFS AFS 10,000 JOB AFS Source: Staff calculations using the WBPS. Source: Staff calculations using the WBPS. Note: Due to few answers, less the 1,500 AFS monthly is excluded. Most post-2013 returnees live in households in which the bread winner has very low educational attainment and only one out of four returnees have completed more than 6th grade (Table 11). The low educational attainment is reflected in the types of jobs returnees (bread winners) hold; most returnees work as daily wage laborers or are self-employed in non-agricultural jobs ( Figure 6). Upon return, there is also a small increase in self-employment in agriculture and public sector employment, likely representing an easing of constraints in accessing land and government employment. About 57 percent of households have employed bread winners who hold the same type of job as they did prior to return and this is especially true for daily wage laborers in the non-agricultural sector (see Appendix, Table A.8) In contrast, bread winners who used to be self-employed in agriculture or held salaried 18 Nearest neighbor propensity score matching refers to a matching estimator in which we match individuals in control and treatment groups based on the “nearest neighbor”. The nearest neighbor is defined as an individual from the comparison group that is closest in terms of propensity score. 19 Employment rates among women in Afghanistan are low and the survey does not estimate the employment of women (unless the main bread winner of the household is a woman). 20 The poverty line is estimated at 2,056 Afs per person per month, equivalent to USD 30 per capita per month using current exchange rates, or roughly USD 1 per person per day 19 positions in the public sector prior to returning to Afghanistan, are most likely to change their job upon return. Table 11: Highest completed education (bread winner) Total No education 48% Primary school 26% Middle/secondary school 12% Higher/tertiary education 10% Religious school 4% Total 100% Note: Staff calculations using the WBPS. Note: Only respondents who are also the main bread winners are included in the table. Figure 6: Job type prior to and after return to Afghanistan (bread winner) Job type prior to return Job type after return 59% 43% 23% 18% 13% 9% 9% 9% 6% 4% 2% 1% 1% 0% 0% 0% EMPLOYEES PUBLIC SECTOR SALARIED, DID NOT WORK AGRICULTURE AGRICULTURE AGRICULTURE AGRICULTURE LABOR, NON- WITH 1 OR EMPLOYER DAILY WAGE DAILY WAGE PRIVATE EMPLOYED, EMPLOYED, SECTOR SALARIED, MORE LABOR, NON- SELF SELF Source: Staff calculations using the WBPS. Note: Missing information on employment are excluded. Wage laborers and self-employed persons in Afghanistan typically have high job vulnerability, and returnees are no exception. About half of the employed bread winners are paid daily for their work, 21 percent receive irregular payments, and only 20 percent are paid monthly. Daily wage laborers and self- employed workers in agriculture also tend to be paid poorly. Breaking down earnings by the four most common job types (Figure 7) shows that, independent of job type, approximately half of the bread winners earn between 5,000 and 10,000 Afs (approximately USD 70-130) per month. However, only 6 percent of those self-employed in agriculture and 3 percent of self-employed in non-agriculture earn more than 20 10,000 Afs (approximately USD 130) per month while 32 percent and 22 percent of salaried private sector employees and non-agricultural daily wage laborers, respectively, earn more than 10,000 Afs per month. Figure 7: Typical monthly income by common job type Salaried, private sector Self employed, agriculture Self employed, non-agriculture Daily wage labor, non-agriculture 67% 54% 51% 47% 47% 32% 31% 23% 22% 17% 6% 3% 1,500 - 5,000 AFS 5,001 - 10,000 AFS MORE THAN 10,000 AFS Source: Staff calculations using the WBPS. Note: 1 percent of respondents indicate that they earn less than 1,500 AFS a month. Due to the low frequency, these individuals are excluded from the chart. Most returnees21 rely on networks to find jobs and less than 2 percent use formal channels such as agencies or advertisements. More than 50 percent of respondents found their job through friends/neighbors or family, and 43 percent found their job on their own (Figure 8). Only 12 percent of the returnees with employment found their job within 1 month while 44 percent of returnees searched for more than 3 months before finding a job. Another 44 percent of bread winners who did not have a job were unemployed for more than 6 months (Table 12). Figure 8: How the respondent found his/her job Table 12: Duration of unemployment Total Less than one month 12% 1-6 months 44% More than 6 months 44% Total 100% 21 Respondents (not bread winners) were asked how they found their job. About 68 percent of respondents were employed. 21 44% 43% 11% 2% FAMILY HBOURS OWN/N OTHER O HELP ON MY FRIEND S/NEIG Source: Staff calculations using the WBPS. Source: Staff calculations using the WBPS. Afghan returnees typically came from poor living conditions in Pakistan (Section Error! Reference source not found.). To assess changes in the living situation of Afghans upon their return to Afghanistan, we compare key characteristics such as school attendance rates and various income related indicators of registered returnees both prior22 and post return. We use nearest neighbor propensity score matching methods to identify registered returnees in the PPVR who are likely to resemble the registered returnees in the WBPS prior to their return (in 2011). Access to basic services of good quality such as health care and schools can play an important role in reducing vulnerabilities of the poor. While in Pakistan, the vast majority (97 percent) of Afghan returnee households23 lived within one-hour travel time from the nearest health facility and more than half (52%) lived within 15 minutes to the nearest health facility. Likewise, 97 percent of Afghan returnee households lived within one-hour walking distance from a primary school in Pakistan. The proximity of schools and health care may be explained by the fact that most Afghans in Pakistan lived in refugee villages (Section Error! Reference source not found.) where large groups of people live in small areas and hence facilities such as schools and health care will typically be found within or close to the refugee villages. This is somewhat supported when comparing the type of health facilities Afghan returnees lived closest to prior to and after returning Afghanistan. Prior to their return, most (45%) Afghan returnees lived closest to Afghan Refugee Health Centers and private practitioners and to a lesser extent to public health facilities which tend to be found in urban areas. In contrast, after returning to Afghanistan most returnees lived in proximity to public health clinics and hospitals after which is consistent with many Afghans settling in urban areas such as Nangarhar and Kabul. School attendance is determined by a range of factors including accessibility of schools, quality of schools, wealth of households and social norms. School attendance rates among returnee children in Afghanistan display large gender disparities. While 61 percent of households with boys sent all boys to school, this was the case for only 37 percent of households with girls (Figure 9). However, school attendance rates were also lower and displayed larger gender disparities prior to returning to Afghanistan. The comparison of school attendance rates before and after return to Afghanistan suggests that more households send all boys and all girls to school, and fewer send no boys and no girls to school, after returning to Afghanistan. 22 This section compares the 911 registered returnees in the WBPS who live in their province of origin with 911 similar individuals in the PPVR. 23 Registered Afghan households in Pakistan who returned between 2014 and 2017. 22 Figure 9: School attendance rates of boys and girls prior and post return 61% 57% 52% 53% 37% 28% 27% 24% 19% 20% 11% 12% PRIOR POST PRIOR POST PRIOR POST PRIOR POST PRIOR POST PRIOR POST NO GIRLS SOME GIRLS ALL GIRLS NO BOYS SOME BOYS ALL BOYS ATTENDING ATTENDING Source: Staff calculations using registered returnees in the WBPS and in the PPVR. Vertical bars indicate 95 percent confidence intervals. Note: Nearest (10) neighbor propensity score matching. Please see appendix for more information on the matching. The change in living conditions is less encouraging when comparing labor market characteristics of returnees prior and post return to Afghanistan. The unemployment rate among returnees was significantly higher after their return to Afghanistan than in their host country. The average unemployment rate of bread winners after returning to Afghanistan was 10 percent compared to only 3 percent prior to returning to Afghanistan (Figure 10). The estimated employment rate among bread winners prior to returning to Afghanistan is likely an upper estimate as the propensity score matching method may match bread winners in the WBPS with household members who are not bread winners in the PPVR dataset. The results also show that there is a slight shift from being employed or self-employed towards daily labor work after returning to Afghanistan. Thus, as previously shown, indicates that Afghan returnees experience higher job vulnerability and lower salaries after their return to Afghanistan. Figure 10: Job type prior and post return 56% 54% 40% 35% 10% 3% PRIOR POST PRIOR POST PRIOR POST NO WORK DAY LABORER EMPLOYEE/EMPLOYER/SELF- EMPLOYED Source: Staff calculations using documented returnees in the WBPS and in the PPVR. Vertical bars indicate 95 percent confidence intervals. Note: Nearest (10) neighbor propensity score matching. Please see annex 2 for more information on the matching. Higher job vulnerability is also suggested by a lower number of days worked per week (Table 13). Afghan returnees worked on average 6 days per week (8.9 hours per day) in Pakistan whereas they only worked an average of 4.5 days per week (9 hours per day) after returning to Afghanistan. 23 Table 13: Average work days and hours Prior return Post return Average Median Average Median Weekly days of work 5.9 6 4.5 5 Daily hours of work 8.9 8 9.0 8 Source: Staff calculations using documented returnees in the WBPS and in the PPVR. Note: Nearest (10) neighbor propensity score matching. In the PPVR (prior return), weekly days of work are calculated by dividing monthly days of work with 4.33. Please see annex 2 for more information on the matching. Comparing income levels prior and after returning to Afghanistan shows a general decrease in average income levels after returning to Afghanistan (Figure 11). This decrease is driven by an increase in the share of Afghans earning between 1,500-3,000 Afs along with a decrease in the share of Afghans earning more than 10,000 Afs after returning to Afghanistan. The share of Afghans earning between 3,000 and 10,000 Afs did not change significantly prior to and after their return (which may also be a result of matching bread winners in the WBPS with individual level data in the PPVR dataset). Figure 11: Monthly income levels (Afs.) per employed person prior and post return 49% 44% 34% 31% 16% 8% 8% 5% PRIOR POST PRIOR POST PRIOR POST PRIOR POST 1500-3000 3001-5000 5001-10000 MORE THAN 10000 Source: Staff calculations using registered returnees in the WBPS and in the PPVR. Note: Nearest (10) neighbor propensity score matching. Due to the low frequency of respondents reporting an average income less than 1,500 Afs these individuals are not shown in the chart. Hence percentages do not sum to 100. An exchange rate of 0.6 is applied. Please see annex 2 for more information on the matching. The findings of this section show that recent returnees are living under difficult circumstances. Post-2013 returnee households are generally large and though most bread winners are employed, many have very little means to support their families. Post-2013 returnees have weak labor market attachment; they typically work as daily wage laborers, have high job vulnerability and low earnings. Moreover, the findings also suggest that school attendance has improved after returning to Afghanistan, whereas employment and earnings have decreased. When considering results from the Afghanistan Living Conditions Survey of 24 2013-14, which allows for the disaggregation of returnees and those Afghans who have never been displaced (host populations), we observe that returnees were more urbanized than hosts, and their higher urbanization was associated with relatively better outcomes on a range of socio-economic measures such as a higher exposure to formal education or greater access to infrastructure services compared to hosts. Yet, male returnees had lower employment-to-population ratios, lower labor force participation rates, and slightly higher unemployment rates. Returnees also suffered higher indebtedness, on average, and were less likely to own their own homes. 24 There has been a clear deterioration in living conditions and employment in Afghanistan across the distribution since the security transition in 2014. This suggests that the recent arrivals examined in this report may have faced worse conditions than previous waves of return. In addition, as shown in this report, recent returnees, are worse off in terms of employment and earnings compared to those remaining in Pakistan. Yet, returnees came back to Afghanistan in unprecedented numbers between 2014 and 2017 which is likely a result of non- economic reasons, such as the environment towards Afghan refugees in Pakistan and incentives for repatriation. 6. Conclusion This report sheds light on the return decisions and the living conditions of Afghan returnees, an issue that is more pertinent than ever with the increase in the number of refugees in protracted situations. Along with integration and third-country settlement, repatriation is considered one of the long-term solutions for forced displacement. This study expands our understanding of what drives the return decision and socioeconomic outcomes upon return of those who have lived in refuge for long, an issue where evidence is scarce. Afghanistan confronts one of the world’s most protracted and complex population displacement challenges. Millions of Afghans have returned from neighboring countries since 2002, and growing numbers of Afghan people are internally displaced. Amidst deepening insecurity and economic fragility, refugee returns strain public services in Afghanistan and intensify competition for scarce economic opportunities. This affects not only displaced people, but all Afghans. Effective management of the displacement challenge will be critical for Afghanistan’s economic and political future and a robust policy response requires understanding of the specific socio-economic conditions and needs of refugee returns. Yet, reliable socio-economic data on displacement in Afghanistan have been difficult to obtain. A recent collaboration of the World Bank and UNHCR financed by a Trust Fund provided by United Kingdom’s Department for International Development (DFID) on forced displacement, complements UNHCR’s existing monitoring and data collection efforts, aimed at collecting real -time data through household level phone interviews. In this context, the World Bank and UNHCR teams attempted to make good use of the existing data sources and complemented it with new data collection methods to better understand the patterns and characteristics of recent Afghan refugee returns. More specifically, the team aimed at connecting insights 24 For an in-depth analysis on how returnees fare compared to hosts among whom they live in Afghanistan which provides insights into questions around reintegration or broader support policies, please refer to: http://documents.worldbank.org/curated/en/294921533557045480/Afghanistan-s-Displaced-People-A-Socio- Economic-Profile-2013-2014 25 between different data sources to explore (albeit imperfectly) questions of selection among Afghans who remained in Pakistan and those documented returnees who returned to Afghanistan. The report details that Afghan refugees who returned to Afghanistan between 2014 and 2017 tend to be worse off in monetary terms than refugees who stayed in Pakistan. More specifically, Afghan refugees who returned between 2014 and 2017 were less wealthy, lived in refugee villages in Pakistan or temporary housing, had previously considered repatriating, and had retained ties to Afghanistan. However, there is some evidence that returnees were more likely to have some (albeit low) formal education) than the average registered refugee in Pakistan. Once returned to Afghanistan, our findings on the settlement decision and post-return economic outcomes provide evidence of the primacy of physical security and proximity to social networks in the decision to return, even at the cost of economic outcomes. Most refugees returned to their province of origin in proximity to family and friends or, if returned elsewhere, did so for safety and economic reasons. Moreover, Afghans living in their province of origin were more likely to be employed and less likely to have the same type of job as they did prior to returning. The survey’s most salient finding was of widespread socio -economic hardship affecting returnees. Afghan returnee households are large and although most families have at least one person working for pay, they have low job stability and low wages. Most returnees work as daily wage laborers in non-agriculture, and returnees generally experience a decrease in the employment opportunities, wages, and job stability after returning to Afghanistan. At the same time, school attendance for boys and girls is higher post-return, and the gender gap in attendance is lower, driven by a higher likelihood that households send girls to school. In general, over time, analysis of data on socio-economic outcomes of returnees may be useful in identifying opportunities for reintegration into Afghanistan’s socio-economic landscape: that is, to better understand why some groups performed unusually well, despite adversity, and whose success may hold lessons for others. Analysis like the one discussed in this report can yield their full benefit when they can be compared in series with findings from subsequent iterations of the survey, giving a sense of how socio- economic outcomes are evolving. Sources Roggio, B. and Gutowski, A. (2018): “Taliban control of Afghan districts remains unchanged despite increased US military pressure”, FDDs Long War Journal, May 1st, 2018. Ministry of States and Frontier Regions, the Office of the Chief Commissioner for Afghan Refugees and UNHCR (2012). Population Profiling, Verification and Response Survey of Afghans in Pakistan 2011. PPVR/HHS/MAR 2012, available at: https://unhcrpk.org/wp-content/uploads/2018/05/Population- Profiling-Verification-and-Response-Survey-of-Afghans-in-Pakistan.pdf World Bank (2018a). Afghanistan World Bank Phone Survey of Afghan Returnees: Methodology and Representativeness. Washington, D.C.: World Bank Group. Available at: http://documents.worldbank.org/curated/en/298881533562809348/Afghanistan-World-Bank-Phone- Survey-of-Afghan-Returnees-Methodology-and-Representativeness 26 World Bank (2018b). Afghanistan’s Displaced People: A Socio-Economic Profile, 2013-2014. Washington, D.C.: World Bank Group. Available at: http://documents.worldbank.org/curated/en/294921533557045480/Afghanistan-s-Displaced-People-A- Socio-Economic-Profile-2013-2014 27 Appendix 1: Data sources This note relies on three data sources to shed light on the movements and living conditions of post-2013 returnees; the Afghan Population Profiling, Verification and Response survey (PPVR), the Voluntary Repatriation Form survey (VRF) and a recently collected World Bank Phone Survey (WBPS). The ‘Afghan Population Profiling, Verification and Response Survey’ and ‘Voluntary Repatriation Form’ The PPVR dataset is used to approximate the characteristics and living conditions of Afghan refugees in Pakistan. The PPVR is a large-scale household survey conducted by the Government of Pakistan and the United Nations High Commissioner for Refugees (UNHCR) from November 2010 to December 2011. It includes detailed information on individual and household characteristics of 130,000 Afghan households who lived in Pakistan at the time of the survey, and given its large sample size, may be considered representative of that population at the time. In total, 975,000 individuals were interviewed, 507,000 of which were registered refugees for whom the proof of registration (PoR) number was recorded. The subset of registered Afghan refugees living in Pakistan in 2011 and returning to Afghanistan between 2014 and 201725 was identified using the Voluntary Repatriation Form survey (VRF) which was collected by UNHCR at the time of return. It includes the PoR number of registered returnees who pass through one of the Voluntary Repatriation Centers in Pakistan to apply for voluntary repatriation and then go to the Encashment Centers in Afghanistan to receive assistance. At the time of writing this report, the only data available on socio-economic conditions of returning Afghan refugees (in the form of the VRF database) was from UNHCR and thus the section comparing returned refugees and Afghans remaining in Pakistan, is based on documented returnees. The group of registered Afghan refugees living in Pakistan in 2011 and returning to Afghanistan between 2014 and 2017 will be referred to as recent documented returnees. World Bank Phone Survey A recent phone survey (WBPS) collected by the World Bank Group is used to assess the characteristics and living conditions of post-2013 Afghan returnees. The WBPS sheds light on post-2013 returnees and their (i) movements and choice of location, (ii) labor market attachment, and (iii) access to services. In total, 3,575 returnees in Afghanistan26 who returned in 2014 or later were randomly selected27 and interviewed between January 2018 and April 2018. The WBPS is a household level survey and one household member answers questions on behalf of the household. The labor market information is collected for the primary bread winner of the household, which in three-quarters of the households is the respondent him/herself. In cases where the survey respondent 25 The period of 2014 to 2017 was selected as a reference for the VRF analysis to coincide with the years of return of the bulk of respondents in the WBPS. 26 Equally distributed between documented and undocumented. 27 For details on methodology and sampling, please refer to the accompanying methodology note of the WBPS, accessible at: http://documents.worldbank.org/curated/en/298881533562809348/Afghanistan-World-Bank-Phone- Survey-of-Afghan-Returnees-Methodology-and-Representativeness. 28 is not the primary bread winner, labor market information is collected via proxy-response28. Given the nature of the survey, only households which own a mobile phone were able to participate in the survey. The survey findings are therefore biased towards households owning a mobile phone and living in areas with network coverage and may likely be better off than the average Afghan returnee. We thus consider the survey findings an upper estimate of post-2013 returnees’ living conditions. This note presents weighted results of the phone survey adjusting for regional sampling imbalances29. The participants of the WBPS will be referred to as post-2013 returnees. Appendix 2: Propensity Score Matching We apply propensity score matching methods to create a comparable sample from the 2011 PPVR data on Afghans in Pakistan, to approximate the change in living conditions after returning to Afghanistan. Propensity scores were only used to eliminate Afghan households in the pre-return data from 2011 who were not significantly similar to those covered in WBPS. Propensity scores are the predicted probability of being in the WBPS (see Table A.1) and the larger the propensity score, the larger the predicted probability that the individual would be in the WBPS. We use the propensity scores to identify individuals in the PPVR (who made a registered return to Afghanistan from 2014 to 2017) who resemble registered post-2013 returnees in the WBPS30. The propensity scores were estimated based on six personal characteristics; age (at the time of the respective surveys), gender, highest completed education (scale variable), language (binary variables), job type prior to return (binary variables), and province of origin31 (Nangarhar, Kabul or other) using STATA’s user written program psmatch2 (by Edwin Leuven, University of Oslo and Barbara Sianesi, Institute for Fiscal Studies, London, UK). A total of 911 registered post-2013 returnees (main bread winners) in the WBPS were each successfully matched with 10 registered returnees in the PPVR. To avoid matching individuals with propensity scores far away from each other, we only consider individuals under common support – individuals in the PPVR with a propensity score higher (or lower) than the highest (or lowest) propensity score of individuals in the WBPS were not included in the matching. Table A.2 reports descriptive information for the 911 matches and shows that the average values of the characteristics used for matching are very close for the matched individuals in the WBPS and in the PPVR. 28 Personal characteristics such as educational background and gender of the main bread winner of the household are only available when the respondent is the main bread winner. 29 Survey weights are based on preliminary data on the number of returnees from the IOM’s Displacement Tracking Matrix and the ALCS 2016-17. For details on survey weights, please refer to the accompanying methodology note of the WBPS. 30 Only individuals older than 13 years at the time of the respective surveys are included in the matching 31 The WBPS only includes information on province of origin if the respondents live in their province of origin. 29 Figure A.1 Propensity of being in WBPS Source: Staff calculations using VRF and PPVR for registered Afghan returnees in Pakistan and the WBPS for registered returnees in Afghanistan. Table A.2: Results of propensity score matching Source: Staff calculations using VRF and PPVR for registered Afghan returnees in Pakistan and the WBPS for registered returnees in Afghanistan. 30 Appendix 3: Additional data tables Table A.3: Living location and tenancy arrangement Refugee villages 2014-17 Afghan returnees All Afghans in Pakistan Owning 14% 16% Renting 23% 20% Relative’s house 1% 1% Given freely 62% 63% Other 0% 0% Total 100% 100% Urban 2014-17 Afghan returnees All Afghans in Pakistan Owning 6% 16% Renting 85% 75% Relative’s house 1% 1% Given freely 8% 7% Other 1% 0% Total 100% 100% Rural 2014-17 Afghan returnees All Afghans in Pakistan Owning 20% 30% Renting 60% 53% Relative’s house 1% 1% Given freely 19% 16% Other 1% 0% Total 100% 100% Source: Staff calculations using VRF and PPVR for registered Afghan returnees and PPVR for registered Afghan refugees in Pakistan. Note: Calculated per household. May not sum to total due to rounding. Table A.4: Occupation of working age (18-64 years) men Men, working age (18-64 years) 2014-17 Afghan All Afghans in Pakistan returnees Elementary Occupations 32% 35% Services and sales workers 23% 23% Plant and Machine Operators and Assemblers 12% 11% Craft and related trade workers 12% 10% Skilled agriculture, forestry and fishery workers 3% 3% Professionals 3% 3% Managers 0% 0% Tech. and ass. Professionals 0% 0% Clerical support workers 0% 0% 31 Armed Forces Occupations 0% 0% Not employed 16% 15% Total 100% 100% Source: Staff calculations using VRF and PPVR for registered Afghan returnees and PPVR for registered Afghan refugees in Pakistan. Note: Does not include individuals who are in school Table A.5: Regression results of the likelihood of returning to Afghanistan (Probit model) (1) (2) (3) (4) VARIABLES (2011 characteristics) Probability of Probability of Probability of Probability of returning returning returning returning between 2014 between 2014 between 2014 between 2014 and 2017 and 2017 and 2017 and 2017 =1 if household intended to return to Afghanistan when interviewed 0.0130** 0.0134** 0.00924 in 2011 (0.00564) (0.00567) (0.00563) =1 if anyone in the household visits Afghanistan regularly? 0.162*** 0.163*** 0.151*** 0.176*** (0.00449) (0.00461) (0.00465) (0.00462) =1 if household owns property in Afghanistan? -0.0853*** -0.0884*** -0.0824*** (0.00524) (0.00525) (0.00522) Gender (Male) -0.0992*** -0.0998*** -0.106*** (0.00419) (0.00420) (0.00417) Aged 12-17 years -0.110*** -0.115*** -0.119*** (0.00614) (0.00616) (0.00612) Aged 18-59 years -0.0913*** -0.0942*** -0.0972*** (0.00533) (0.00537) (0.00534) Aged 60+ years -0.214*** -0.222*** -0.228*** (0.0109) (0.0114) (0.0113) Own dwelling in Pakistan -0.246*** -0.248*** -0.262*** (0.00594) (0.00608) (0.00600) Household size -0.00790*** -0.00775*** -0.00884*** -0.00949*** (0.000395) (0.000398) (0.000408) (0.000408) Share of children (0-16 years) in household 0.190*** 0.102*** 0.130*** 0.144*** (0.0109) (0.0117) (0.0120) (0.0119) =1 if the number of meals the same for the whole month -0.0814*** -0.0844*** (0.00849) (0.00841) =1 if individual reports some form of disability 0.0438*** 0.0633*** (0.00696) (0.00692) =1 if lived in camp 0.525*** 0.502*** 0.499*** (0.00690) (0.00695) (0.00732) =1 if lived in rural area 0.213*** 0.183*** 0.166*** (0.00741) (0.00747) (0.00760) =1 if any member of household ever experienced serious physical or 0.0457*** 0.0962*** mental harm in Pakistan (0.0141) (0.0140) =1 if dwelling type was semi-pakka 0.0120 0.126*** (0.00857) (0.00828) =1 if dwelling type was mud house, hut, tent, temporary -0.0459*** 0.202*** shelter/plastic, on the street/in the open, other (0.00722) (0.00616) 32 Household owns a car -0.140*** -0.123*** -0.138*** -0.115*** (0.0128) (0.0129) (0.0129) (0.0129) Highest completed level of schooling is primary school 0.0710*** 0.138*** (0.00608) (0.00599) Highest completed level of schooling is middle or secondary school 0.0735*** 0.151*** (0.00635) (0.00625) Highest completed level of schooling is technical/college/university 0.0633*** 0.147*** (0.00842) (0.00834) Highest completed level of schooling is religious school 0.00749 -0.0162** (0.00714) (0.00708) Highest completed level of schooling is other/Don't know 0.0510*** 0.0590*** (0.00803) (0.00794) Constant -1.145*** -0.897*** -0.823*** -0.714*** (0.00891) (0.0110) (0.0150) (0.0135) Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Source: Staff calculations using VRF and PPVR for registered Afghan returnees and PPVR for registered Afghan refugees in Pakistan. Excluded baseline categories are: ‘age interval 5-11 years’, ‘living in rural area’, ‘living in a pakka residence’ and ‘having no education”. Note: Only registered refugees are included and refugees younger than 6 years are excluded. No weights applied. Table A.6 Belongings left behind in country of refuge Post-2013 returnees Nothing 78% Land/property/house 6% Household assets 5% Cash/bank account 5% Business/shop/enterprise 3% Other 3% Total 100% Source: Staff calculations using WBPS. Table A.7: Regression results of likelihood of returning to the province of origin (Probit model) (1) (2) (3) VARIABLES = 1 if current province = 1 if current province = 1 if current province is the province of origin is the province of origin is the province of origin Employment ratio in province of -1.623*** -1.273*** -1.279*** return (0.0113) (0.0117) (0.0133) Poverty rate in province of return 0.708*** 0.858*** 0.727*** (0.00696) (0.00717) (0.00793) Urbanization rate in province of -0.602*** -0.506*** -0.534*** return (0.00150) (0.00168) (0.00188) District controlled by government -0.460*** -0.427*** (0.00371) (0.00420) Household size -0.00569*** (0.000328) 33 Share of children (0-16y) in hh -0.350*** (0.0106) =1 if documented return (vrf no) -0.186*** (0.00348) Year of return to Afghanistan = 3, 0.169*** 2015 (0.00750) Year of return to Afghanistan = 4, 0.149*** 2016 (0.00641) Year of return to Afghanistan = 5, 0.260*** 2017 (0.00665) Year of return to Afghanistan = 6, 0.592*** 2018 (0.0107) =1 if a child (<14y) within hh has 0.125*** contributed to hh income (0.00471) =1 if anyone in hh skipped meal last 0.114*** week due to lack of food (0.00366) =1 if bread winner of hh has a job 0.0564*** (0.00583) =1 if same main source of income -0.149*** as before returning to Afghanistan (0.00352) Constant 2.426*** 2.134*** 2.406*** (0.0128) (0.0132) (0.0181) Observations 1,119,070 1,119,070 938,578 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Source: Staff calculations using the WBPS. Note: 229 households (6.4 percent of the sample) that are planning to move are not included in the regression. 2 observations did not answer the question about current province of living. ‘Urbanization rate’, ‘poverty rate’ and ‘employment rate’ are based on results using ALCS 2016-17. ‘Under government control’ is based on Roggio et al. (2018). Table A.8 Share of Afghans by current job type who held the same type of job prior to returning to Afghanistan Total Salaried, public sector 19% Salaried, private sector 51% Daily wage labor, agriculture 13% Self-employed, non-agriculture 54% Daily wage labor, non-agriculture 79% Did not work 8% Total 57% Source: Staff calculations using WBPS. Note: Due to low number of respondents, the categories ‘self-employed agriculture’ and ‘employer with 1 or more categories are excluded 34