Vulnerability and Protection of Refugees in Turkey Findings from the Rollout of the Largest Humanitarian Cash Assistance Program in the World Vulnerability and Protection of Refugees in Turkey Findings from the Rollout of the Largest Humanitarian Cash Assistance Program in the World P. Facundo Cuevas O. Kaan Inan Aysha Twose Çiğdem Çelik With Data Collection by © 2019 The World Bank and the World Food Programme 1818 H Street NW, Washington DC 20433, USA. Telephone: 202-473-1000; Internet: www.worldbank.org. Some rights reserved This work is a product of the staff of The World Bank and the World Food Programme. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of the Executive Directors of The World Bank or the governments they represent , or those of the World Food Programme. The World Bank and the World Food Programme do not guarantee the accuracy of the data included in this work. The bound- aries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank or the World Food Programme concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this work is subject to copyright. Because The World Bank and the World Food Programme encourage dissemination of their knowl- edge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Attribution—Please cite the work as follows: “Cuevas, P. Facundo, O. Kaan Inan, Aysha Twose, and Çiğdem Çelik. 2019. Vulnerability and Protection of Refugees in Turkey: Findings from the Rollout of the Largest Humanitar- ian Cash Assistance Program in the World. © World Bank and World Food Programme.” All queries on rights and licenses, including subsidiary rights, should be ad- dressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org. Table of Contents Acknowledgments vii Acronyms ix Executive Summary 1 1. Introduction 5 2. Background: The Emergency Social Safety Net for Refugees in Turkey 7 3. Conceptual Approach: The Multiple Dimensions of Vulnerability 9 4. Empirical Approach to Assessing Vulnerability and ESSN Protection Performance 11 4.1 Data 11 4.2 Methodology 13 5. What vulnerabilities affect the population of ESSN refugees? 19 Poverty and resources 19 Access to key services: Education 23 Food security 23 Capacity to cope 23 Household composition 26 Skills and Livelihood 27 Debt burden 29 6. How well does ESSN target and protect vulnerable refugees in Turkey? 31 Comparative vulnerability between refugees eligible and ineligible to ESSN 31 Coverage and targeting 33 Adequacy of transfers 35 Benchmarking performance to international experience 36 Benchmarking performance to alternative untargeted design 39 7. Policy Discussion and Way Forward 41 References 43 List of Tables Table 1: Distribution of ESSN applicants by eligibility and region 13 Table 2: Per capita monthly expenditure of ESSN refugees, average and median 20 Table 3: Poverty and inequality among ESSN refugee population 20 Table 4: Use of consumption coping strategies in ESSN households 24 Table 5: Demographics of ESSN refugee households 26 Table 6: Median debt of ESSN households, total and as a proportion of expenditure 29 Table 7: Comparative vulnerability between ESSN eligible and ineligible across dimensions 34 Table 8: Performance indicators of simulated untargeted transfer 40 List of Figures Figure 1: Dimensions of Vulnerability 10 Figure 2: Map of subnational strata of the PAB survey 12 Figure 3: Extreme poverty rate and number of poor among ESSN refugees 21 Figure 4: Poverty rate and number of poor among ESSN refugees 21 Figure 5: Expenditure shares among ESSN households 22 Figure 6: Percentage of ESSN households with all school-age children out of school 23 Figure 7: Percentage of ESSN households with unacceptable food consumption 24 Figure 8: Consumption-based Coping Strategies Index among ESSN households 24 Figure 9: Most severe livelihood coping strategies used by ESSN households 25 Figure 10: Types of livelihood coping strategies used by ESSN households 26 Figure 11: Share of ESSN households with at least one member with Turkish language abilities 27 Figure 12: Main income sources of ESSN households 28 Figure 13: Primary income source of ESSN households, by region 29 Figure 14: Poverty rate among eligible and ineligible ESSN refugee population 32 Figure 15: Extreme poverty rate among eligible and ineligible ESSN refugee population 32 Figure 16: At any value of poverty line, ESSN eligible refugees are poorer 33 Figure 17: Coverage of ESSN, by poverty status (individual level) 34 Figure 18: ESSN beneficiary incidence and inclusion error 34 Figure 19: ESSN eligibility and poverty status across applicant refugee population 35 Figure 20: Adequacy of ESSN transfer size (as percentage of per capita expenditure) 36 Figure 21: Poverty and coverage rates of unconditional cash transfer programs across the world 37 Figure 22: Poverty and beneficiary incidence of unconditional cash transfer programs across the world 38 Figure 23: Poverty and adequacy of unconditional cash transfer programs across the world 39 Acknowledgments This report was authored by P. Facundo Cuevas (Senior Economist, World Bank), O. Kaan Inan (Consultant, World Bank), Aysha Twose (Head of VAM/M&E, World Food Programme), and Çiğdem Çelik (M&E Officer, World Food Programme), with support from Nesrin Semen (M&E Officer, World Food Programme) in the early stages of the draft. The authors would like to specially thank the Turkish Red Crescent (TRC) team led by Servet Avci (M&E Team Leader, TRC) for collecting the data that made this study possible. The report benefitted from valuable feedback and guidance from World Bank and WFP management, peer reviewers and colleagues. These include, on the World Bank side, Carolina Sanchez-Paramo (Senior Director, Poverty and Equity Global Practice), Carlos Silva (Acting Practice Manager, Poverty and Equity Global Practice), Ximena del Carpio (Turkey Program Leader, Social Inclusion), Ruth Hill (Lead Economist, Pov- erty and Equity Global Practice), Ugo Gentilini (Sr. Social Protection Specialist, Social Protection and Jobs Global Practice), and Phillippe Leite (Sr. Social Protection Econ- omist, Social Protection and Jobs Global Practice); and on the WFP side, Jonathan Campbell (Deputy Country Director, WFP Turkey); Nils Grede (Country Director, WFP Turkey); Susanna Sandstrom (Economist, WFP); Edgar Luce (M&E Officer, WFP Re- gional Bureau, Cairo); Mohamed Salem (M&E Officer, WFP Regional Bureau, Cairo); Enas Ali (Regional Data Analyst, WFP Regional Bureau, Cairo). The Ministry of Family, Labor and Social Services provided useful comments to earlier draft of the report. In addition, the work profited from presentations of findings at various stages of preparation conducted at Turkey’s Ministry of Family, Labor and So- cial Services; ECHO; WFP; Syria Task Force; and Cash-Based Interventions Technical Working Group. All their contributions are gratefully acknowledged. The team would like to thank ECHO and Sweden for financially supporting this com- prehensive assessment. Finally, the WFP asked the World Bank to lead the targeting and performance analysis (in section 6) to refrain from assessing its own program and for the report to preserve an objective perspective. | vii Acronyms AIDA Asylum Information Database ASPIRE Atlas of Social Protection Indicators of Resilience and Equity CTP Cash-Transfer Programming CVME Comprehensive Vulnerability Monitoring Exercise DGMM Directorate General of Migration Management ECHO European Civil Protection and Humanitarian Aid Operations ESSN Emergency Social Safety Net FRiT European Union’s Facility for Refugees in Turkey FCS Food Consumption Score HBS Household Budget Survey ISAS Integrated Social Assistance System LCS Livelihood Coping Strategies LCSI Livelihoods Coping Strategies Index LIC Low-Income Country LMIC Lower-Middle-Income Country MoFLSS Ministry of Family, Labor and Social Services M&E Monitoring and Evaluation PAB Pre-Assistance Baseline PDM Post-Distribution Monitoring PDMM Provincial Directorate of Migration Management PPP Purchasing Power Parity rCSI reduced (Food) Coping Strategies Index SASF Social Assistance and Solidarity Foundation SSC Social Service Center TRC Turkish Red Crescent UMIC Upper-Middle-Income Country UNHCR United Nations High Commissioner for Refugees VAM Vulnerability Assessment and Mapping WFP World Food Programme | ix x | V ulnerabilit y and P rotection of R efugees in T urke y Executive Summary T urkey hosts more refugees than any oth- er country in the world, with near 4 mil- lion refugees living in the country by December beneficiaries as of December 2018, ESSN is the largest humanitarian assistance program in the world. The program is funded by the Euro- 2018.1 The overwhelming majority, 3.6 million, pean Union member states, and implemented are from Syria. The remainder originate mainly nationwide in partnership with the Ministry of from Afghanistan and Iraq. Turkey’s response Family, Labor and Social Services (MoFLSS), the to the refugee crisis has drawn international World Food Programme (WFP), and the Turkish praise, highlighted for providing a non-camp Red Crescent (TRC). solution to the refugee influx, and granting ac- cess to public services. Notwithstanding this, This report provides local and international pol- the nature of the crisis calls for humanitarian icy actors, stakeholders, and those interested assistance to help refugees cope with it. in the developments of one of the most signifi- cant humanitarian crises of our time with In November 2016, the Emergency Social Safe- ty Net (ESSN) program was introduced with the a. A comprehensive view of the vulnerability objective of supporting the most vulnerable ref- situation of refugees eligible for ESSN and ugees meet their basic needs through monthly b. An assessment of how well ESSN targets, cash transfers. The targeting criteria were de- supports, and protects the most vulnera- veloped by identifying household demograph- ble refugees. ic characteristics as best proxy to being poor, unable to afford basic needs. With 1.5 million The timing of the analysis takes place after pro- gram eligibility is determined but before trans- fers are disbursed. Future work will focus on 1 The report uses the term refugee regardless of country of the impact of transfers. origin, although Syrians are under temporary protection sta- tus, and non-Syrians under international protection law. For Readers in Turkey and abroad might find this more details: Asylum Information Database (AIDA), Intro- report valuable for at least two reasons. It fills a duction to the Asylum Context in Turkey. knowledge gap in terms of assessing refugees’ | 1 vulnerabilities using population-representative implications. First, a proportion of refugees data. Existing studies have used purposive and suffer from food insecurity and face constraints non-representative samples that constrain the in children’s access to education, which bear generalizability of their findings. This report enduring human capital consequences. Sec- uses nationwide representative data to char- ond, there is an intensive use of ‘costly’ coping acterize the living conditions of refugees that strategies, which cripple ESSN households’ applied to ESSN. It is a significant step forward longer-term livelihood capacity. in achieving a more accurate and unbiased pic- ture of refugees’ vulnerabilities on the ground There is an ample degree of heterogeneity and understanding population-level trends rel- across regions in terms of incidence of vulner- ative to existing studies. ability dimensions, and there is no simple re- gional pattern, which speaks to the complexity Second, the report documents the rollout re- of refugee vulnerability. sults of a unique humanitarian cash assistance program, the largest of its kind, at a time when • Refugees in Istanbul and Aegean regions there is a call in the international community are less vulnerable to poverty, lack of for more cash-based programming of human- skills, and precarious labor income sourc- itarian aid. While cash-transfer programming es than the rest but show the highest vul- (CTP) is regarded as an important area of in- nerability in access to education. novation in humanitarian assistance, with • Refugees in Anatolia show the least vul- great potential to meet assistance needs more nerability to food insecurity, perhaps giv- efficiently and more effectively, it accounted en the proximity to agricultural activities, for only 10 percent of humanitarian aid in the but they suffer the worst access to skilled world in 2016. labor opportunities. • Refugees in the Southeast show better Findings from Vulnerability Assessment access to education than the rest but suf- of Refugees Targeted by ESSN fer the highest vulnerabilities in food se- curity and language skills. The report assesses vulnerability of refugees • The single unifying pattern is, perhaps, eligible for ESSN before they start receiving that refugees in all regions resort to inten- transfers, using a multidimensional approach sive use of detrimental coping strategies that includes the following areas: poverty, that cripple the productive capacity of the household resources, access to key services, household, its resilience, and its ability to food security, capacity to cope, skills and live- face adversity in the future. lihood sources, and debt burden. Findings from ESSN Rollout on Targeting The vulnerabilities of the refugee population and Protection of Refugees targeted by ESSN are multiple and complex: there is a pervasive prevalence of poverty, af- The analysis here focuses on issues of target- fecting 76 percent of ESSN refugees, com- ing, coverage, errors of exclusion and inclusion, pounded by other types of vulnerabilities with and adequacy of the transfer size relative to inter-temporal and inter-generational poverty pre-assistance refugee budgets. 2 | V ulnerabilit y and P rotection of R efugees in T urke y The report finds the following: Policy Discussion and Way Forward a. The ESSN targeting criteria was effec- The report identifies four emerging policy lessons. tive in identifying a relatively poorer and more vulnerable population, compared The first policy lesson that emerges is ‘yes we to the population that applied but do not can’. In a constrained resource, data, and time fit the program’s eligibility criteria. This context, ESSN achieves good performance holds both at the national level and with- marks. The complementarities in design and in every region. implementation capacity of the WFP-TRC-Mo- b. Relative to international experience in un- FLSS partnership have been among the key conditional cash transfers, ESSN achieves drivers of this performance, which made it high coverage of the poor, as well as high possible to receive, process, and validate thou- (ex ante) adequacy and protection of the sands of applications across Turkey. poor, though it allows higher inclusion error. Among ESSN applicants, 6 percent The second policy lesson is that within the de- are extreme poor and ineligible, 17 per- bate of whether assistance should be target- cent are poor (but not extreme poor) and ed or untargeted, the ESSN’s targeting already ineligible, and 14 percent are nonpoor looks more ‘universal’ than other cash transfer and eligible. programs, since ESSN achieves relatively high c. In the context of humanitarian assis- coverage, at the expense of higher inclusion er- tance to refugee populations, prior- ror, but providing meaningful assistance to sup- itizing coverage and inclusion at the port beneficiaries’ basic needs. An untargeted expense of accuracy is probably the design would have to make large sacrifices on preferred balance, since the nonpoor the adequacy of the benefit level. population that represents the inclusion error is relatively vulnerable. A third of While ESSN decision makers gave priority to nonpoor beneficiaries live with a budget the advantages of the targeted design given ES- that is within 10 percent of the poverty SN’s programmatic objective and budget con- line. straints, a third policy implication is that still the d. If ESSN had gone with an untargeted de- ineligible population is in need of support since sign, such that every applicant is eligible poverty is high among them. A promising area to receive some assistance, it could have for policy action could be to improve access to addressed the exclusion error borne by economic opportunities. In a sense, the eligi- the chosen targeted design. But, given bility criteria capture lack of capacity to work, resource constraints, this would require which implies that ineligible refugees have a large reduction in the size of the trans- relatively better prospects for income genera- fer provided to each beneficiary, that is tion with the right support, such as Turkish lan- a significant decrease in adequacy. This guage skills. would compromise the ability of each household to meet their basic needs, As a fourth policy implication, efforts must focus which is the main programmatic objec- on decreasing ESSN’s exclusion error among tive of ESSN. the extreme poor, with complementary actions E x ecutive S ummar y | 3 to increase their inclusion in the program. The Turkish Social Assistance and Solidarity Foun- dation (SASF) Allowance, started in November 2018, presents a promising step forward. It is recommended that this allowance finds ways to incorporate knowledge of community leaders, Muhtars, and local organizations, who could give refugees referrals to the SASFs to foster equitable access. Moving forward, on the analytical front, WFP and the World Bank will jointly evaluate the impacts of the ESSN assistance on beneficia- ry lives. Additionally, on the operational front, as the conflict underlying the refugee influx becomes more protracted, ESSN should tran- sition from a humanitarian-type to a develop- ment-type response to promote a sustainable exit from poverty and vulnerability. In that re- gard, ESSN stakeholders are starting to focus on designing strategies to transition beneficia- ries toward better income opportunities. Future developments in ESSN should be docu- mented and made publicly available, to ensure the ESSN experience can be productively used for policy discussions and programming re- sponses in other humanitarian challenges and contexts. This report is a building block toward that larger, longer-term objective. // 4 | V ulnerabilit y and P rotection of R efugees in T urke y 1. Introduction Turkey is host to more refugees than any other remain to increase coverage of these services, country in the world. Close to 4 million refugees particularly with regard to work permits and were living in Turkey by December 2018.2 The schooling. overwhelming majority, 3.6 million, are from Syria.3 The remainder originate mainly from Af- In November 2016, the Emergency Social Safe- ghanistan and Iraq.4 Turkey’s response to the ty Net (ESSN) program was introduced to help refugee crisis has drawn international acclaim, the most vulnerable refugees meet their basic highlighted for providing a non-camp solution needs through monthly cash transfers. ESSN to the refugee influx, and granting access to was set up with financial support from the Eu- public services.5 A reduced number of refu- ropean Union’s Facility for Refugees in Turkey gees, about 5 percent of the total, are hosted (FRiT) through its humanitarian arm, the Direc- in camps near the Syrian border.6 Refugees in torate-General for European Civil Protection Turkey can obtain identity cards, with which and Humanitarian Aid Operations (ECHO). The they can access health care in government fa- program is implemented nationwide in part- cilities and enroll in schools. Working-age refu- nership with the Ministry of Family, Labor and gees can also potentially obtain work permits. Social Services (MoFLSS), the World Food Pro- Notwithstanding this, other significant barriers gramme (WFP), and the Turkish Red Crescent (TRC). With 1.5 million beneficiaries as of De- cember 2018, ESSN is the largest humanitarian 2 The report uses the term refugee regardless of country of cash transfer program in the world. origin, although Syrians are under temporary protection sta- tus, and non-Syrians under international protection law. For This report has two objectives. First, to provide more details: Asylum Information Database (AIDA), Intro- an assessment of the vulnerability situation of duction to the Asylum Context in Turkey. refugees eligible for ESSN. Second, to evaluate 3 There are 3,585,738 Syrians under temporary protection how well the ESSN targets, supports, and pro- in Turkey. Source: DGMM (2018). tects the most vulnerable refugees. To do so, 4 There are 164,351 Afghans, 142,576 Iraqis, 37,732 Ira- the report uses data from program applicants nians, 5,518 Somalis, 11,515 Other countries. Source: UN- after their eligibility into the program has been HCR (2018). determined but before the distribution of the 5 See, for example, World Bank (2015), New York Times cash assistance. This allows to examine issues (2017). of targeting, errors of exclusion and inclusion, 6 The refugee camp population is 177,376, while the off- and adequacy of the transfer size relative to camp population is 3,408,362 (DGMM 2018). pre-assistance budgets of refugees. | 5 This report makes two contributions to the aid in the world in 2016, up by 2.5 percent from growing literature on refugee crises, forced 2015.9 displacement, and policy responses. First, it contributes to fill in a gap in terms of vulnera- The study is intended for the wide audience of bility assessments of refugees in Turkey. There stakeholders, both in Turkey and abroad, con- are many studies and reports on the subject, cerned about refugees’ vulnerabilities and pol- but they are based on purposive samples rath- icy and program responses to refugee crises. er than representative data, which prevents The findings may be of interest, among others, knowing how generalizable or biased findings to those interested in monitoring and evalu- are. This report uses nationwide representa- ation, those focused on program design and tive data from a probability sample of refugees operational issues, and those keen on know- that applied to ESSN. While it has its limitations, ing more about the policy responses to one of that is, the study is not representative of the the most significant humanitarian crises of our non-applicant refugee population in Turkey, it time. is a significant step forward in achieving a more accurate and unbiased picture of refugees’ vul- The report is organized in the following sections. nerabilities on the ground and understanding Section 2 describes the main design features population-level trends.7 of the ESSN program. Section 3 presents the report’s conceptual approach to the multiple Second, the report documents the rollout re- dimensions of vulnerability, giving a consistent sults of a large-scale and unique humanitarian framework to the assessment. Section 4 de- assistance program, the largest of its kind, at scribes the analytical approach and data sourc- a time when there is a call in the international es used in the study. Sections 5 and 6, the main community for more cash-based programming sections of the report, present the findings of of humanitarian aid.8 While cash-transfer pro- the analysis: a vulnerability assessment of refu- gramming (CTP) is regarded as an important gees targeted by ESSN, and a review of different area of innovation in humanitarian assistance, aspects of ESSN performance at pre-assistance with great potential to meet assistance needs distribution time. Finally, section 7 discusses more efficiently and more effectively, it ac- policy implications and the way forward. counted for only 10 percent of humanitarian 7 As discussed in the Data section, data from a complemen- tary survey called Comprehensive Vulnerability Monitoring Exercise (CVME) shows that non-applicants and ineligible applicants have similar poverty levels, and non-applicants are not poorer than eligible applicants. Since the CVME sam- pling is not representative beyond the survey subjects, this is indicative evidence (CVME 2017). 8 See High-Level Panel on Humanitarian Cash Transfers (2015); Bailey and Harvey (2017); Grand Bargain Signato- ries (2016). 9 Cash Learning Partnership (2018). 6 | V ulnerabilit y and P rotection of R efugees in T urke y 2. Background: The Emergency Social Safety Net for Refugees in Turkey ESSN was launched nationwide in November or equal to 1.5 (essentially, at least three de- 2016 to support the most vulnerable refugees pendents for every two able-bodied adults); (2) in Turkey.10 The main partners of the program families with four or more children; (3) single are ECHO, MoFLSS, WFP and TRC. It provides females; (4) elderly headed households; (5) unconditional cash transfers of TL 120 per single parent households; and (6) households person per month to promote access to basic with one member at least 40 percent dis- needs through easing households’ constrained abled.13 budgets.11 Comprehensive outreach efforts have been Eligibility of a household to benefit from ESSN undertaken to reach the intended population is determined based on six demographic crite- and key stakeholders. This included the use of ria, selected as proxy measures of household multiple media (workshops, printed materials, poverty and vulnerability to meet basic needs.12 social media) in four languages to provide ac- These are (1) a dependency ratio greater than curate information on the program and support ESSN reach the target population and relevant actors—for example, Turkish Social Assistance 10 More precisely, the program covers foreigners residing in and Solidarity Foundations (SASFs), Social Ser- Turkey under International Protection, including Temporary vice Centers (SSCs), Nüfus, Provincial Director- Protection for Syrians. For brevity, these foreigners will be ate of Migration Management (PDMM) offices, referred to as refugees, although not all of them have the Provincial Governors, Kaymakams (heads of legal status of refugees within Turkey. district), and Muhtars (community leaders). 11 In addition to the regular monthly transfers, beneficiary households also receive flat quarterly top-ups designed to All ESSN applications are digitized and consol- support smaller households in meeting their basic needs. idated into the Integrated Social Assistance Household size 1–4: TL 250; Household size 5–8: TL 150; Household size 9+: TL 50. For more details on the ESSN pro- gram, please refer to its website: http://kizilaykart-suy.org/. the model, refer to the WFP Vulnerability Assessment and 12 Using quantitative data from a 2015 household survey, Mapping (VAM) report (WFP 2016). a model was developed to understand the relationship be- 13 Before June 2017, the dependency ratio threshold was tween household characteristics and household welfare above 1.5 (not including those with ratio equal to 1.5) and (measured using per capita expenditure). These results the disability threshold was two members (not one). The were used, in combination with other secondary data, to targeting criteria was revised to increase the coverage of develop the demographic targeting criteria. For details on vulnerable refugee households. | 7 System (ISAS), or Bütünleşik Sosyal Yardım Bil- gi Sistemi, an e-government system developed by Turkey and introduced in 2009 that elec- tronically integrates and facilitates all steps re- lated to the management of social assistance. The government of Turkey adapted ISAS to handle ESSN application processes, and assess eligibility into the program, including verifica- tion of required documents. As a result, ESSN applicants and beneficiaries are integrated into the same registry with Turkish applicants and beneficiaries of MoFLSS social assistance pro- grams. This integration makes Turkey an inter- esting and leading example of ‘adaptive social protection’ to effectively respond to a surge in social assistance needs caused by the human- itarian emergency situation created by the ref- ugee crisis. Monitoring and evaluation is an essential com- ponent of the program. This includes a Pre-As- sistance Baseline (PAB) survey, which is used as the main source of data for this report, and Post-Distribution Monitoring (PDM) surveys collected with quarterly frequency. The surveys are collected by a TRC call center, with techni- cal support from the WFP and the World Bank. Given the ESSN characteristics, findings of this report will be useful to a wider audience focused on the links between humanitarian assistance and national social protection systems, ‘adap- tive social protection’, and forced displacement contexts.14 Understanding the ESSN rollout and performance to date helps inform these devel- opments moving forward. 14 A discussion is already undergoing in Lebanon: http://odi- hpn.org/blog/cash-debate-lebanon/. 8 | V ulnerabilit y and P rotection of R efugees in T urke y 3. Conceptual Approach: The Multiple Dimensions of Vulnerability The analysis of the report is conducted within Income is an important part of the story, but a ‘multidimensional vulnerability’ conceptual there are dimensions of being vulnerable that framework. Vulnerability refers to a situation go well beyond income. Income is necessary, where the household has limited ability to meet but not sufficient, to escape vulnerability. For basic needs and a constrained capacity to cope example, for refugee populations, access to le- with risks that can negatively affect the achieve- gal status and capacity to cope with trauma are ment of those needs. In the face of ongoing two dimensions that cannot be captured by just risks, this constrained household capacity can focusing on income. bear short- and long-term consequences on its members’ well-being. In being vulnerable, While there are considerations to make for there is an element that pertains to a current each specific context, the following seven di- situation (for example, meeting basic needs at mensions provide a relatively comprehensive a point in time) and an element that relates to account useful for a study of household vulner- the susceptibility or fragility of that situation to ability: change and worsen in the future (for example, high risk of not meeting basic needs tomorrow). 1. Poverty and resources: Does the house- As such, vulnerability can manifest itself along hold have enough resources to meet ba- multiple dimensions. For example, a household sic needs? may be considered to be in a vulnerable condi- 2. Access to key services: Do school-age tion if… children have access to education? Do households have access to housing, • … it is affected by food insecurity health services, legal status? • … it struggles to access basic services 3. Food security: Can a household procure • … it has a large number of dependents a diet with the necessary caloric and nu- per able working-age member tritional content for all its members? • … it lacks adequate coping strategies to 4. Capacity to cope: Can the household re- face adverse shocks sort to safe coping strategies to face adver- • … it burdens itself with debt to meet to- sity? Or does the household use strategies day’s basic needs that may be damaging for future produc- tivity (including sale of assets, reductions These vulnerabilities are of diverse nature: in human capital and risky behaviors)? some are economic, some demographic, some 5. Household composition: Do households are more static, and some are inter-temporal. have a structure that puts them in a more | 9 fragile situation (such as a large number complementary. For example, a household that of dependents, single parent, female afford basic needs could still be vulnerable to head)? food insecurity if the composition of the diet 6. Skills and livelihood: Do household is not adequate. Each dimension captures a adults have the skills required by the con- standalone aspect of being vulnerable. Figure text? Can they access income sources for 1 provides a graphic summary of the approach. self-reliance? 7. Debt burden: Do household incur in debt The first part of the report assesses the vulnera- that can become unsustainable? bility of refugees in the ESSN program using each of these dimensions, first at the national level and Naturally, these dimensions are inextricably then across regions of the country. In addition, in intertwined. They are not linked by a unidirec- the targeting section, this vulnerability frame- tional causal relation and they are not mutually work will be used to compare refugees eligible to exclusive either but somewhat overlapping and receive ESSN with those ineligible for ESSN. Figure 1: Dimensions of Vulnerability Poverty & Basic Resources Services Debt Food Burden Security Household Vulnerability Skills & Coping Livelihood Capacity Household Composition 10 | V ulnerabilit y and P rotection of R efugees in T urke y 4. Empirical Approach to Assessing Vulnerability and ESSN Protection Performance 4.1 Data because it means that findings in those surveys cannot be generalized beyond those cases that The main data source for the analysis is the PAB answered the survey and have limited value survey, which was collected by TRC as part of for decision making or program design. In this the monitoring and evaluation of the program, sense, the PAB is unique—by using the registry and is representative of refugee households of ESSN applicants as the sampling framework that applied to ESSN between the start of the to randomly select respondents, PAB findings program and May 2017.15 In population terms, are generalizable to the 1.6 million people who the survey is representative of 1.6 million peo- applied to the program. ple.16 At that time, there were 3.2 million refu- gees in Turkey, so the PAB is representative of A second important strength of the PAB is that half of the refugee population. the survey sample was also designed to be rep- resentative of both the eligible population and The PAB survey is innovative in being represen- the ineligible population that applied to ESSN, tative of a refugee population in the Turkey con- which allows to conduct comparative analysis text. There are many surveys collected around between these two groups of interest. the country but due to a lack of sampling frame- work, their samples are not representative of A third advantage of the PAB survey is that it a population.17 This is an important limitation is not only representative of the population of applicants at the national level but also at the subnational level. In other words, survey re- 15 This section summarizes the main features of the data used spondents were selected at random within five in the analysis. For further technical detail see Cuevas et al regional strata. Therefore, the data can be bro- (2019). ken down for analysis across the following five 16 Since May 2017, about 500,000 additional people have regions: Southeast, Anatolia/Thrace, Istanbul, applied. 17 This happens, for example, in surveys that choose their respondents in specific locations such as a market, a com- calculate sampling weights to make the observations in the munity center, a neighborhood. Without a comprehensive sample representative of the population where they come understanding of all the households that make up the popu- from. For a regularly updated list of assessments and sur- lation of interest (‘the universe’), these surveys do not have veys see: http://data.unhcr.org/syrianrefugees/documents. a way to assess what population are the cases included in php?page=2&view=grid&Language%5B%5D=1&Country the study representative of. In technical terms, they cannot %5B%5D=224&Type%5B%5D=4. | 11 Figure 2: Map of subnational strata of the PAB survey Kırklareli Bartın Kastamonu Sinop Istanbul Zonguldak Ardahan Edirne Tekirdağ Karabük Samsun Artvin Kocaeli Düzce Rize Ordu Trabzon Yalova Sakarya Çankırı Amasya Giresun Bolu Çorum Gümüşhane Tokat Kars Bayburt Bursa Bilecik Erzurum Çanakkale Ankara Iğdır Balıkesir Eskişehir Kırıkkale Yozgat Erzincan Ağrı Sivas Kütahya Kırşehir Tunceli Bingöl Muş Manisa Afyon Nevşehir Uşak Kayseri Malatya Elazığ Bitlis Aksaray Van Izmir Batman Diyarbakır Isparta Konya Niğde K. Maraş Siirt Aydın Denizli Adıyaman Burdur Şırnak Hakkâri Ozmaniye Mardin Muğla Karaman Şanlıurfa Antalya Adana Gaziantep Mersin Kilis Hatay Strata Aegean Anatolia/Thrace Istanbul Mediterranean South-East Mediterranean, and Aegean. Figure 2 illustrates plicants are similar to those of ineligible appli- the composition of each region. cants. It also shows that non-applicants have entered Turkey more recently than applicants The survey was collected by the TRC call center and have not applied to ESSN because their and was designed by the WFP with inputs from registration with the authorities for identity the World Bank. The TRC call center team was cards is still in process.18 recruited and trained to conduct interviews in three languages: Arabic, Turkish, and English, Second, the phone-based modality poses an to minimize the number of surveys that could important limitation on the amount of infor- not be completed due to a language barrier. mation that can be collected from survey re- spondents. A face-to-face survey can collect Although the PAB data brings important value information about every household member added, it faces two limitations that are useful to and can ask about living standards with a fine highlight to motivate the data-collection agen- level of detail through hundreds of questions.19 da going forward. First, the PAB survey does not A phone-based survey is constrained to a sub- sample the non-applicant population. There- fore, the vulnerability and targeting assessment is based on those that applied. If, for example, 18 CVME Survey (WFP 2017): Sample of non-applicants all non-applicants were poor, then the targeting and ineligible applicants was obtained using neighborhood conclusions of the report would be too ‘optimis- snowball sampling, so the results are not representative of tic’. However, there is indication that that is not the refugee population beyond the sample. the case. A survey of applicants and non-appli- 19 For example, Turkey’s household expenditure survey asks cants shows that the poverty levels of non-ap- respondents about expenses on more than 200 items. 12 | V ulnerabilit y and P rotection of R efugees in T urke y Table 1: Distribution of ESSN applicants by eligibility and region Households Individuals Number Share (%) Number Share (%) Total 268,602 100 1,627,874 100 Eligible 141,330 53 956,653 59 Ineligible 127,272 47 671,221 41 Region Istanbul 37,855 14 236,018 15 Aegean 29,377 11 175,718 11 Anatolia/Thrace 59,259 22 363,679 22 Mediterranean 24,454 9 150,895 9 Southeast 117,657 44 701,564 43 Source: PAB data. set of questions and to basic information about PAB survey through different variables and in- household living conditions. dicators, as follows:20 The PAB sample size is 8,690 households, 1. Poverty and resources: A household is which were interviewed after the ESSN pro- considered poor if its per capita expen- gram had determined their eligibility but diture is below the poverty line. Expen- before households knew their eligibility sta- diture is measured by monthly expens- tus and therefore before receiving any ESSN es on food, housing, utilities (electricity, transfers. The majority of those who applied heating, gas), water, phone/internet ser- were considered eligible. The eligibility rate vice, health, hygiene, education, trans- reached 53 percent of households, which, portation, and other non-food items (like taking into account their members, meant clothes, tobacco, and so on) reported by that nearly 60 percent of all individuals who households in the PAB survey and adjust- applied were assessed as eligible. Among re- ed by price differences between regions.21 gions, the provinces in the Southeast provided more than 40 percent of the ESSN applica- tions. Table 1 summarizes the distribution of 20 This section provides a summary of the methodology. For applicants and their status. further technical detail see Cuevas et al (2019). 21 Following Deaton and Saidi (2002), expenditure on cel- 4.2 Methodology ebrations and debt repayment are excluded, while expen- diture on health is included due to their high elasticity. Ex- The multiple dimensions of vulnerability dis- penditures are adjusted using spatial price indexes from cussed in section 3 can be measured with the Turkey’s National Statistics Office (TUIK) to account for 4 . E mpirical A pproach to A ssessing V ulnerabilit y and E S S N P rotection P erformance | 13 The main poverty line used in the analy- as 4 out of 5 days per week. Access to sis is equal to US$5.5 per capita per day, health services (mental or physical), ad- in 2011 purchasing power parity (PPP), equate housing (for example, sanitation, equivalent to TL 284 per capita per month housing materials, number of rooms), in 2016. This is an internationally compa- and legal services are not captured in the rable line calculated by the World Bank to survey.24 monitor poverty in upper-middle-income 3. Food security: To assess households’ contexts like Turkey.22 A complementary access to an adequate diet, the analysis line is used to capture extreme poverty, uses the Food Consumption Score (FCS). which takes the value of US$3.20 per The FCS is constructed with questions on capita per day (2011 PPP), equivalent consumption frequency of 9 different food to TL 165 per capita per month in 2016 groups during the past 7 days. Therefore, terms.23 The analysis looks also at those it considers not only the frequency but that are closely above the poverty line, also the diversity of the diet.25 It is a stan- since while they are making ends meet dard WFP indicator used globally to mea- on the day of survey, they are vulnerable sure food security and allows to assess if to fall below the poverty line if there is a households achieve acceptable or unac- small decrease in income. ceptable food consumption.26 2. Access to key services: The PAB allows 4. Capacity to cope: This is measured by to construct three indicators to measure two indicators: the reduced (Food) Cop- access to education. First, the share of school-age children that are regularly at- tending school. Second, whether none of 24 Access to identification and legal status is not collected the school-age children in the household by the survey since registration with the DGMM (temporary are regularly attending school. Third, protection for Syrians, international protection for non-Syr- whether all the school-age children in ians) is a prerequisite to apply for ESSN. In addition, ESSN the household are regularly attending officers helped refugees get their papers in order with the school. Regular attendance was defined DGMM when needed (for example when the person regis- tered in one province but moved to another). 25 For more details on the FCS, refer to WFP (2008). purchasing power and cost of living differences across re- 26 FCS classifies household diets into three groups: accept- gions (the value of TL 1 in Istanbul is not the same as TL 1 able, borderline, or poor. In this analysis, all households with in Sanliurfa). poor and borderline FCS are grouped into unacceptable. 22 For technical details, see Jolliffe and Prydz (2016). Borderline food consumption equates to daily consumption 23 This is an internationally comparable line calculated by of staples and vegetables, with frequent (4 days/week) con- the World Bank to measure poverty in lower-middle-income sumption of oil and pulses. As sugar and oil are consumed countries. For the Turkey context, it can be taken as a proxy as part of the daily diet in the context of Middle East popu- for extreme poverty, whereby households cannot afford food lations where refugees come from, thresholds are set to in- needs. Many countries calculate and use national poverty clude oil and sugar within the poor and borderline classifica- and extreme poverty lines based on cost of basic needs and tion. Therefore, a borderline or worse diet is very restricted, cost of food needs, but that is not the case in Turkey. If those usually including no or extremely low consumption of meat, lines were available, they would be used in the analysis. dairy, pulses, or fruit. 14 | V ulnerabilit y and P rotection of R efugees in T urke y ing Strategies Index (rCSI) and the Liveli- ty through the household’s dependency hoods Coping Strategies Index (LCSI). ratio and female headship. Households a. The rCSI is constructed with questions with higher dependency or headed by a about reductions in quantity or quality woman are in a more vulnerable situation of food intake to cope with adversity. to afford adverse shocks than households It includes five specific consumption with lower dependency or headed by a coping strategies, each given a stan- man. The dependency ratio is calculat- dard severity weight and aggregated ed as the number of children and elder- into an index.27 ly members (ages 0–17 and 60 or more, b. The LCSI assesses the stress and se- respectively), by the number of work- verity of coping mechanisms used by ing-age members in the household (ages households and their implications for 18–59). While the ESSN targeting criteria longer-term productive capacity. It is related to the dependency ratio accounts derived from a series of 14 questions for disability, the PAB survey does not col- regarding households’ experience with lect information on disability status. livelihood stress and asset depletion 6. Skills and livelihood: In a high literacy during the 30 days before the survey. context like Turkey, language skills are a Coping strategies are classified into three key competency to escape vulnerability categories: stress, crisis, and emergency, and access economic opportunities. The based on the severity of the impact of analysis uses two indicators to capture the strategy on household resilience and this: first, whether at least one house- the ability to cope with future livelihood hold member can read/write Arabic and shocks. The questions used for the PAB’s second, whether at least one household LCS module were validated and weight- member can read/speak Turkish. In addi- ed based on focus group discussions tion, access to livelihood sources is cap- conducted with the affected population, tured in the PAB by asking respondents to ensure that they are appropriate and about their three main sources of income: representative for the current context. skilled labor, unskilled labor, borrowing, When the respondent reports that the gifts, remittances, and others. household has not used a given coping 7. Debt burden: While having access to strategy, he or she reports whether it is credit to weather shocks can be consid- because the household has exhausted ered positive, it can put households in a the use of such strategy (for example, very vulnerable situation if the amount of selling assets) or simply did not need to debt becomes unsustainable. To capture engage in that behavior. this, the analysis uses the ratio between 5. Household composition: The data at the household’s stock of debt and the hand allow to capture the household household’s monthly expenditure. composition dimension of vulnerabili- To tackle the second question of interest, that is, how well does ESSN target and support its 27 For more details on the rCSI, refer to the Coping Strategies beneficiaries, the analysis will use the following Index: Field Methods Manual (Maxwell and Caldwell 2008). approach. 4 . E mpirical A pproach to A ssessing V ulnerabilit y and E S S N P rotection P erformance | 15 • Comparative vulnerability assessment If the poverty rate is 100 percent, then of ESSN eligible households to ESSN by definition the targeting inclusion error ineligible households: Using the multi- will be zero regardless of any targeting dimensional vulnerability approach and strategy. the indicators described above, eligible • Adequacy of the transfer size: It is of little households are compared to ineligible use to have a great targeting system if the households to assess whether the pro- transfers do not provide meaningful sup- gram was able to target the more vulner- port to those eligible. Benefit adequacy able population. Ideally, the comparison represents the ratio between the monthly would also be done with households that value of transfers that eligible households did not apply to ESSN, but data for that do will receive and their pre-transfer month- not exist. ly budget. In this way, adequacy portrays • Coverage rate and targeting exclusion the capacity of the program to adequately error: An important indicator to evaluate support expenses of beneficiaries.29 the performance of any assistance pro- • Benchmarking: To complement the as- gram is its coverage of the target popu- sessment of the coverage, targeting, and lation. In this case, that is measured by adequacy rates of ESSN, the analysis share of eligible population among the makes use of a benchmarking approach. poor. Conversely, the complement of the Comparing ESSN performance indicators coverage rate represents the targeting to other situations brings perspective exclusion error. That is, the share among into the analysis. The goodness of ESSN’s the poor that the targeting criteria of the support to vulnerable refugees cannot be program failed to identify as ‘in need of measured by how far it is from a utopia of ESSN support’.28 zero targeting errors but by using realistic • Targeting inclusion error: This concerns alternatives as benchmarks. The analysis those cases that were not part of the tar- makes use of two different yardsticks. get population of the program, that is the First, ESSN is benchmarked with respect nonpoor, but were assessed eligible by to the international evidence on perfor- the program’s targeting criteria. The in- mance of cash transfer programs around clusion error is calculated as the share of the world. The analysis uses data from nonpoor population among the eligible. the Atlas of Social Protection Indicators of This error is a function of the poverty rate. Resilience and Equity (ASPIRE), the larg- est and most comprehensive database on worldwide safety net performance in- 28 Instead of poverty, other indicators of vulnerability could dicators. Second, the existing ESSN pro- be used to assess exclusion error. However, the poverty indi- tection performance is benchmarked to cator is at the core of ESSN’s targeting strategy—the target- ing criteria were defined as the best proxies to being under the poverty line that can be measurable and verifiable in this 29 As a reminder, the PAB survey was collected before any context. In addition, the nature of the poverty measure, with ESSN transfers were distributed, so the adequacy calcula- a clear threshold to classify households, makes the analysis tions are made using the amounts that eligible households clearer and the findings simpler to communicate. will start receiving. 16 | V ulnerabilit y and P rotection of R efugees in T urke y an alternative design of universal, untar- geted transfers, as a simulated counter- factual. This is a scenario where, keeping the total transfers budget constant, all those who apply receive a cash transfer, but lower in size than the one under the existing targeted design. This is a design question that most programs struggle with: giving a bit of support to all regard- less of their vulnerability versus trying to give a bit more to those who need it more. • To assess the performance of an untar- geted, universal transfer and deal with the challenge of estimating how many people would apply if the transfer was known to be untargeted, the analysis proceeds with a method of bounds, keeping total trans- fer budget constant. In the lower-bound scenario, the number of applicants to the untargeted transfer equals the num- ber of applicants to the existing targeted design. That is 1.6 million people accord- ing to the PAB data, and the value of the transfer would be TL 70 per person per month.30 In the upper-bound scenario, the number of applicants to the untar- geted transfer equals the number of ref- ugees registered in the country. That is 4 million people as of December 2018, and the value of the transfer would be TL 39 per person per month.31 Performance in- dicators are calculated under each simu- lated counterfactual to bound what would be observed if an untargeted transfer was implemented. The reality would sit some- where in between. 30 Calculated as (TL 120 × 956,653 ESSN eligible people in the PAB) / 1,627,874 applicant people in the PAB. 31 Calculated as (TL 120 × 1.3 million ESSN eligible in admin data) / 4 million refugees in admin data. 4 . E mpirical A pproach to A ssessing V ulnerabilit y and E S S N P rotection P erformance | 17 18 | V ulnerabilit y and P rotection of R efugees in T urke y 5. What vulnerabilities affect the population of ESSN refugees? Based on the multidimensional vulnerability to education. Anatolia shows the least framework presented in section 3, the main vulnerability to food insecurity, given the findings of the first research question of the re- proximity to agricultural activities, but port are the following: suffers the worst access to labor income. The Southeast benefits from the best ac- a. The vulnerabilities of the refugee popu- cess to education but shows the highest lation targeted by ESSN are multiple and vulnerabilities in food security and lan- complex: there is a pervasive incidence guage skills. The single unifying pattern of poverty, affecting 76 percent of ESSN is, perhaps, that given refugees’ limited refugees, compounded by three other resources and opportunities and disad- types of vulnerabilities with wide-ranging vantages of each region, all regions show inter-temporal and inter-generational im- a vulnerable side in some dimension. In plications. First, a proportion of refugees addition, all regions resort to intensive are food insecure and have limited ac- use of detrimental coping strategies that cess to education for their children, which cripple the productive capacity of the bears detrimental human capital implica- household, its resilience, and its ability to tions. Second, there is an intensive use of face adversity in the future. ‘costly’ coping strategies, which cripple ESSN households’ longer-term livelihood It is important to remember that the report capacity. And finally, there are important assesses vulnerability of refugees targeted by constraints with regard to skills and live- ESSN after they are assessed as eligible and lihood sources to better participate in the before any transfer is disbursed to them. This labor market, which limits the short-term section summarizes the findings in each of the ability to emerge out of poverty. seven dimensions of vulnerability described by b. There is an ample degree of heteroge- the conceptual approach. neity across regions in their exposure to each vulnerability dimension, and there is Poverty and resources no simple regional pattern, which speaks to the complexity of refugee vulnerabili- The average refugee eligible for ESSN lives on ty. The Istanbul and Aegean regions are a median budget of TL 219 per month. Expen- less vulnerable to poverty, lack of skills, diture levels are highest in Istanbul, reflecting and precarious labor income sources but better economic opportunities, with a me- show the highest vulnerability in access dian of TL 242 per month. Meanwhile, ESSN | 19 Table 2: Per capita monthly expenditure Table 3: Poverty and inequality among of ESSN refugees, average and median ESSN refugee population Per capita monthly Poverty Headcount Poverty expenditure (TL) indicators ratio gap Mean Median Median Extreme poverty 23.8% 5.8% Eligible 233 219 Poverty 76.0% 25.0% Eligible by Istanbul 257 242 Region Aegean 243 227 Anatolia/ 219 202 Top10/ Thrace   Gini Bottom10 Mediterra- 226 209 nean Inequality 0.228 5.0 Southeast 233 216 indicators Source: PAB data, authors’ calculations. Note: Extreme poverty line = TL165 per person per month. Poverty line = TL284 per person per month. Top10/Bot- tom10 = Average Expenditure of richest 10% / Average expenditure of poorest 10%. Source: PAB data, authors’ refugees in Anatolia and Mediterranean regions calculations. show the lowest budget levels, with medians of 202 and 209, respectively.32 native Turkish population. Based on expenditure data from the Household Budget Survey (HBS), The overwhelming majority of the refugee pop- about 10 percent of Turks live in poverty, and 2 ulation targeted by ESSN is poor. Poverty—liv- percent are affected by extreme poverty.33 The ing with a budget below TL 284 per person per comparison, however, is only indicative. Although month—affects 76 percent of ESSN refugees. the same poverty line is used for both refugees Meanwhile extreme poverty—having a budget and Turks, the survey used to measure poverty of lower than TL 165 per person per month—af- Turkish locals is quite different from the PAB and fects 24 percent of the ESSN population. Basi- that poses methodological comparability limita- cally 3 out of 4 ESSN refugees are poor and 1 tions. PAB consumption aggregate is built upon out of 4 is extreme poor. In addition, 12 percent 10 questions, while HBS consumption aggregate live with budgets that are above, but within 20 uses 300 consumption items. Findings from pre- percent of, the poverty line and are therefore vious research imply that refugee poverty rates vulnerable to fall into poverty in the face of a would be lower if they were measured with a moderate income decline. That is, 88 percent long and detailed questionnaire similar to HBS.34 of ESSN refugees are poor or ‘nearly’ poor. The poverty rates of the ESSN refugees living in 33 Source: World Bank calculations using Turkey’s House- Turkey are significantly higher than those of the hold Budget Survey (HBS) 2016; see Poverty and Equity Brief (World Bank 2018a) for more details on poverty mon- itoring in Turkey. 32 As explained in the Methodology section, spatial deflators 34 The comparison is useful to illustrate the gaps, but it are used to adjust for cost of living and purchasing power should considered carefully. The PAB consumption aggre- differences between regions. gate is built upon 10 questions, while HBS consumption ag- 20 | V ulnerabilit y and P rotection of R efugees in T urke y Figure 3: Extreme poverty rate and number of poor among ESSN refugees Extreme Poverty Rate Extreme Poor Population Anatolia/Thrace 32% Anatolia/Thrace Mediterranean 28% Mediterranean Southeast 23% Southeast National 24% National Aegean 18% Aegean Istanbul 14% Istanbul 0% 20% 40% 60% 80% 0 100,000 200,000 300,000 Source: Extreme poverty line = TL165 per person per month. PAB data, authors’ calculations. Figure 4: Poverty rate and number of poor among ESSN refugees Poverty Rate Poor Population Anatolia/Thrace 80% Anatolia/Thrace Mediterranean 78% Mediterranean Southeast 76% Southeast National 76% National Aegean 73% Aegean Istanbul 69% Istanbul 0% 20% 40% 60% 80% 100% 150,000 350,000 550,000 750,000 Source: Poverty line = TL284 per person per month. PAB data, authors’ calculations. Poverty rates vary significantly across regions. least twice as high in the Anatolia and Medi- The share of ESSN refugees living in extreme terranean regions than in Istanbul. In addition, poverty, with TL 165 or less per month, is at the share living in poverty reaches 80 percent in Anatolia but 69 percent in Istanbul. Regional poverty disparities are somewhat reflective of gregate uses more than 300 questions. Previous research the spatial variation in regional economic de- shows (for example, Beegle et al. 2012) that detailed survey velopment. The Istanbul and Aegean regions modules lead to higher consumption measures. This implies are home to some of the most developed in- that the PAB poverty rates would be lower if they were mea- dustries in the country, with a higher number of sured with a long and detailed questionnaire similar to HBS. relatively better job opportunities. 5 . W h at vulne r abilities affect t h e p o pulati o n o f E S S N r efu g ees ? | 21 Figure 5: Expenditure shares among ESSN households Other 4.9% Food 37.3% Social 0.3% Transport 3.8% Communications 2.2% Debt 2.3% Water 3.4% Education 2.2% Hygine 5.4% Health 3.6% Rent 25.9% Utilities 8.8% Source: PAB data, authors’ calculations. For program design and budget allocations dedicated to food. In addition, about a quarter across regions, it is important to look at the (26 percent) is spent on rent. When food, rent, number of poor in addition to the headcount and utilities are accounted for, ESSN house- poverty rate. Two regions with the same rate holds are left with just 28 percent of their total of poverty can present different challenges monthly budget for all other needs. This pres- with regard to policy solutions if they differ in ents a precarious situation; any unexpected the size of the poor population. In fact, among cost, such as a medical bill or a funeral, could ESSN refugees the rankings in poverty rates leave a household unable to pay rent or meet and poor population are different. Overall, their food needs. there are around 227 thousand extreme poor and 727 thousand poor individuals among the Rent and food expenditure shares vary signifi- ESSN refugees. The largest poor populations cantly across regions. In Istanbul, rent accounts are in Southeastern Turkey, which houses 41 for 32 percent of total expenditure versus only percent of ESSN refugees in poverty. Istan- 25 percent in the Southeast. Accordingly, food bul, despite having a lower headcount rate accounts for about 40 percent of expenditure in using the TL 284 per month line, has more the Southeast but only 32 percent in Istanbul. poor refugees than the Mediterranean and the This reflects the fact that rent costs are signifi- Aegean. cantly higher in Istanbul, and households are forced to adjust their budgets accordingly. In An additional measure of vulnerability of ex- Istanbul, median rent paid for a household with penditure looks at expenditure composition. five members was TL 550 in rent per month The data show that 37 percent of expenditure is versus TL 350 in the Southeast. 22 | V ulnerabilit y and P rotection of R efugees in T urke y Access to key services: Education Regional differences are related to a higher proportion of children working in the Aegean, With regard to access to education, in the aver- Anatolia, and Istanbul regions, as the liveli- age ESSN refugee household just over half (54 hoods coping analysis will show, and the pres- percent) of school-age children are regularly ence of Temporary Education Centers with les- attending school. Moreover, in many house- sons taught in Arabic in the Southeast, which holds, none of the school-age children attend- facilitate school attendance for children who do ed school. In more than a quarter of all ESSN not speak Turkish. households, none of the school-age children are in school. In contrast, 31 percent of ESSN Food security households have all their children regularly at- tending school. The proportion of eligible households with unacceptable food consumption reaches 24 There is certain degree of variation in access to percent. This is driven by constraints in con- education across regions. Households in the Ae- sumption of fruits, pulses, vegetables, animal gean, Anatolia, and Istanbul regions show the protein (eggs, meat, fish), and dairy, which are highest incidence of vulnerability with regard consumed only 1, 2, 2.8, 2.9, and 3.6 days a to children’s schooling. These regions have the week, respectively. Only 6 percent of the ESSN highest proportion of households with no chil- refugee population eats fruits at least 4 times dren attending school, at 28–29 percent. Access per week. to schooling is significantly better in the South- east, where only 24 percent of ESSN households The Southeast and Mediterranean regions have none of their children attending school. present higher levels of food insecurity, with 27 and 25 percent of the eligible population hav- ing an unacceptable (that is, poor or borderline) Figure 6: Percentage of ESSN households with all school-age children out of school food consumption score, respectively. In com- parison, for example, 33 percent of the eligible population live in households that eat vegeta- bles at least 4 days per week in Anatolia ver- Total Eligible 26.5% sus only 21 percent in the Southeast. Similarly, while 41 percent of the population consume South-east 24.4% animal protein (eggs, fish, meat) at least 4 days a week in Anatolia, the corresponding share is Mediterranean 27.1% 27 percent in the Southeast. Anatolia/Thrace 28.4% Capacity to cope Aegean 28.9% Overall, 90 percent of eligible households en- Istanbul 27.6% gaged in some form of consumption-based % of eligible households coping to face shortages in income and means of support. The most commonly employed Source: PAB data, authors’ calculations. strategy was relying on less expensive or less 5 . W h at vulne r abilities affect t h e p o pulati o n o f E S S N r efu g ees ? | 23 Figure 7: Percentage of ESSN households Table 4: Use of consumption coping with unacceptable food consumption strategies in ESSN households Mean days per All Strategy week that the Total Eligible 23.8% strategy is used Sex of Houshold head Less preferred, less expen- 4.4 Male 22.3% sive foods Female 26.0% Borrow food from friends or 0.6 neighbors Region Reduced number of meals South-east 26.8% 2.1 per week Mediterranean 25.3% Reduce portion size at 2.0 meals Anatolia/Thrace 22.0% Reduce adult consumption 2.0 so children can eat Aegean 19.2% Source: PAB data, authors’ calculations. Istanbul 20.3% % of eligible population Figure 8: Consumption-based Coping Source: PAB data, authors’ calculations. Strategies Index among ESSN households All Total Eligible 15.7 preferred foods. On average, ESSN households Region resort to this strategy 4.4 days per week. It is South-east 16.8 also relatively frequent for eligible households to reduce the number of meals and portion Mediterranean 13.5 sizes and decrease adult intake in favor of chil- dren. These coping strategies are used 2 days Anatolia/Thrace 15.7 per week, on average. It is not so common to Aegean 16.5 borrow food from friends or neighbors (0.6 days per week). Istanbul 13.4 The consumption-based coping strategies Source: PAB data, authors’ calculations. index reaches a relatively high level for the ESSN targeted households. The index also reveals that there are three regions that are The incidence of ‘costly’ livelihood coping strat- relatively more vulnerable to this dimension: egies (LCS), which could cripple ESSN house- Southeast, Anatolia, and Aegean all suffer holds’ longer-term livelihood capacity, is also from significantly higher consumption-coping relatively high. Overall, 89 percent of house- vulnerability than the Istanbul and Mediterra- holds applied stress strategies (at least once), nean regions. 63 percent of households applied crisis strat- 24 | V ulnerabilit y and P rotection of R efugees in T urke y egies (at least once), and 39 percent applied emergency strategies (at least once). Figure 9: Most severe livelihood coping strategies used by ESSN households Households are then classified by the most se- Region vere type of coping strategy they adopted. In 18.2% 36.5% 40.2% order of severity, emergency is worse than cri- sis, and crisis is worse than stress. Therefore, Istanbul if a household used both crisis and emergency 24.4% 33.4% 35.8% LCS, it is classified as emergency. Accordingly, households are put into one of four mutually Aegean exclusive groups: did not use any LCS, the most 19.5% 37.5% 37.1% severe LCS used was stress, the most severe LCS used was crisis, the most severe LCS used Anatolia/Thrace was emergency LCS. In turn, only 3.7 percent 24.2% 36.8% 36.8% of ESSN households did not engage in any livelihood coping strategies. Moreover, most Mediterranean households resorted to emergency coping as 23.7% 34.5% 40.2% their most severe strategy, such as sending children to work or relocating the household Southeast to an entirely different location.35 Crisis cop- All ing strategies are almost as frequently used as emergency ones. Both types involve costly con- 22.0% 35.6% 38.7% sequences that are harder to reverse and likely Total Eligible to reduce future productivity and resilience to shocks. Overall, 39, 36, and 22 percent of the None Stress Crisis Emergency ESSN households used emergency, crisis, and Source: PAB data, authors’ calculations. stress coping strategies, respectively, as their most severe LCS. Looking within each of the LCS types allows to understand what individual coping actions are driving the use of each type. Within emergency nisms are reducing investments in health and LCS, the most commonly used coping actions education (43 and 37 percent, respectively). are child labor and household relocation (18 Within stress LCS, borrowing money and buy- and 21 percent, respectively). Within crisis ing food on credit are typically the most resort- LCS, the most frequently used coping mecha- ed ones.36 35 The reference period of the LCS questions is the past 30 36 These are unconditional averages. For example, the per- days. Therefore, this is only capturing the refugees who re- centage using child labor as coping represents the average cently moved into Turkey, not the ones who have been in among all ESSN households. It is not conditional on having living in Turkey for more than a month. used at least one emergency strategy. 5 . W h at vulne r abilities affect t h e p o pulati o n o f E S S N r efu g ees ? | 25 Figure 10: Types of livelihood coping strategies used by ESSN households Emergency Sold household assets 26.9% Spent savings 16.9% Bought food on credit 64.6% Borrowed money 67.3% Gathered unusual types of food 14.2% Crisis Withdrew children from school 15.3% Sold productive assets 3.8% Reduced health expenditure 42.6% Reduced education expenditure 37.0% Stress Entire household moved elsewhere 18.4% Sent children to work 21.1% Begging 1.0% Member of the household returned to Syria 5.1% % of eligible households Source: PAB data, authors’ calculations. Household composition Table 5: Demographics of ESSN refugee households The ESSN targeting conditions are based on vulnerability criteria related to household de- Demographics mographic structure. Therefore, targeted ref- Percentage with male household head 60% ugee households show a high incidence of Percentage with female household head 40% female headship as well as high dependency Number of household members 6.8 ratios. In ESSN households, for every work- Share of members age 0–5 20.3% ing-age adult there almost two dependents to take care of, making refugee households vul- Share of members age 6–17 38.1% nerable to adverse changes in livelihood. In Share of members age 18–59 37.8% addition, 40 percent of ESSN households are Share of members age 60+ 3.8% headed by women. Dependency ratio 1.9 Source: PAB data, authors’ calculations. The regional breakdown shows moderate vari- ations across regions in household composition indicators. In terms of dependency ratios, the 2.00, respectively. Regarding female head- largest variation is observed between South- ship, the highest share of female-headed ESSN east and Anatolia, displaying ratios of 1.86 and households is seen in Istanbul with 43 percent, 26 | V ulnerabilit y and P rotection of R efugees in T urke y while the Anatolia region exhibits the lowest In any case, the more important language skills share, with 38 percent. to decrease vulnerability and enhance the prospects of integration (accessing public ser- The average refugee household in ESSN has vices, DGMM registration and Nüfus [Popula- 6.8 members and has a relatively young age tion Department] registration) are Turkish. Only structure, with 20 percent being children under 46 percent of ESSN households have a member 5 years, 38 percent children in the 6–17 age who can speak Turkish, and 24 percent have a group, another 38 percent in 18–59 group, and member who can read Turkish. 4 percent in the 60-or-more years-old group. There is significant variation in Turkish lan- Skills and Livelihood guage skills across regions. The Southeast has the lowest proportion of ESSN households with About 90 percent of ESSN households have at Turkish language abilities—only 41 percent can least one member who can read or write Ara- speak Turkish and 21 percent can read Turkish. bic. This does not necessarily indicate that the This phenomenon is related to the presence remaining 10 percent are illiterate, however. of many Arabic speaking local populations in While some may be illiterate, many are likely Southeastern provinces, allowing Syrians to coming from non-Arabic speaking countries, communicate without the need to know the such as Afghanistan and Iran. Unfortunately, Turkish language. Furthermore, Southeastern the PAB survey does not allow to distinguish Turkey has the largest numbers of Syrian and between the two hypotheses. Iraqi refugees, allowing for wider Syrian/Iraqi communities to develop, which reduces the need to learn Turkish for daily interactions. In addition, the Temporary Education Centers, Figure 11: Share of ESSN households with at least one member with Turkish which teach in Arabic, are mostly in the South- language abilities east—therefore children do not enroll in local schools and do not learn Turkish.37 All 23.7% Total Eligible 46.3% In terms of income sources, about 90 percent Region of ESSN refugees obtain their main source of South-east 21.4% livelihood from labor. For the vast majority, 64 40.6% percent, unskilled labor is the main income 23.8% Mediterranean 51.5% source. Skilled labor is the main livelihood Anatolia/Thrace 26.1% source for 25 percent of eligible refugees. Very 49.4% few refugees report living from assistance, beg- 27.0% Aegean 53.0% ging, or gifts from friends. 24.4% Istanbul 50.1% % of Eligible Households 37 It is expected that the Temporary Education Centers will Read Turkish Speak Turkish be closed in phases, and children enrolled there will be Source: PAB data, authors’ calculations. transferred to local schools with Turkish curricula. 5 . W h at vulne r abilities affect t h e p o pulati o n o f E S S N r efu g ees ? | 27 ployers of Syrians under temporary protection Figure 12: Main income sources of ESSN have the option to apply for work permits since households 2016, very few refugees have obtained work per- Remittances Unskilled labor mits; only 21,000 were issued in 2017.39 Second, 1.5% 63.8% to reach the more vulnerable households, ESSN All other eligibility rules deem ineligible those households 2.2% with any member enrolled in social security (a re- sult of formal employment). With regard to the second source of income, 61 percent of households reported borrowing money or credit. When asked about a third in- come source, 61 percent stated that they do Credit- not have any access to cash besides these two borrowing sources. Around 17 percent of households stat- 4.6% ed their second or third main source of income Gifts from family Skilled labor to be assistance from other organizations or the 2.9% 24.9% Turkish government. Source: PAB data, authors’ calculations. Results across the five geographical areas This partly reflects the protracted nature of the show significant differences. In Istanbul over conflict. As most refugees have been residing in 50 percent of households reported skilled labor Turkey for an extended period, they have been as their main source of income, with 43 percent able to find basic livelihood opportunities and relying on unskilled labor. In the Aegean region, generate their own income. The issue is that the 60 percent of households report unskilled la- labor income they are able to generate is still bor as their main income source, followed by too low and does not provide enough resources skilled labor with 35 percent of households. to meet basic needs and escape from poverty. These proportions look very different in the oth- As shown earlier, 64 percent of targeted refu- er regions. In Anatolia, the Mediterranean, and gees live with a budget below the poverty line. Southeast, the clear majority of households en- Therefore, vulnerability in the livelihood sourc- gage in unskilled labor, and only 18–20 percent es dimension for refugees targeted by the pro- gain most of their income through skilled work. gram is relatively high.38 Almost all refugees enter Turkey through the In addition, even among the 25 percent of house- Southeast. Population movements toward Istan- holds that get most of their income through skilled bul and the Aegean are driven by the search for labor, employment conditions are vulnerable and better work opportunities. The higher incidence largely informal, for two reasons. First, while em- of skilled labor as a main source of income among 38 As a reminder, all of this is analyzed before eligible refu- 39 Source: Ministry of Family, Labor and Social Services, gees start receiving ESSN transfers. Work permit statistics: http://bit.ly/2j04WVH. 28 | V ulnerabilit y and P rotection of R efugees in T urke y Figure 13: Primary income source of ESSN households, by region South-east 69.5% 19.1% 6.0% Mediterranean 72.0% 20.2% 2.5% Anatolia/Thrace 64.6% 17.5% 4.7% Aegean 59.4% 35.1% 1.9% Istanbul 43.2% 50.6% 3.6% 0.0% 20.0% 40.0% 60.0% 80.0% 100.0% % of Eligible Households Unskilled labor Skilled labor Credit-borrowing Gifts from family Remittances All other Source: PAB data, authors’ calculations. Table 6: Median debt of ESSN households, total and as a proportion of expenditure Total debt (TL, Total debt (TL, re- As proportion of nominal) gionally adjusted) expenditure (%) Istanbul 1,000 876 61.4 Aegean 1,000 935 58.9 Strata Anatolia/Thrace 900 924 60.7 Mediterranean 1,000 986 65.2 Southeast 1,000 1,039 70.5 Female 1,000 935 59.2 Household head Male 1,000 1,039 69.4 Total Eligible 1,000 1,020 64.8 Source: PAB data, authors’ calculations. households in these regions is consistent with indebted ESSN households, the median amount the drivers of refugee population movements. of accumulated debt is TL 1,000, representing a risky burden relative to households’ monthly bud- Debt burden gets. For the median ESSN household, debt rep- resents 65 percent of the monthly budget. Put in Among ESSN households, only 20 percent have perspective, a household is in debt for an amount not incurred any accumulated debt. This varies per that equals the value of their monthly food and region with a low of 15 percent in the Mediterra- rent expenditures combined (which represents nean, and a high of 27 percent in Anatolia. Among 63.2 percent of expenditure on average). 5 . W h at vulne r abilities affect t h e p o pulati o n o f E S S N r efu g ees ? | 29 30 | V ulnerabilit y and P rotection of R efugees in T urke y 6. How well does ESSN target and protect vulnerable refugees in Turkey? 40 Comparative vulnerability between percent, the incidence of poverty among the in- refugees eligible and ineligible to ESSN40 eligible is a cause for concern. The first important result in terms of assess- A similar trend can be observed in extreme ing how well ESSN targets the most vulnerable poverty rates. At the national level, 24 percent refugees is that poverty incidence among the of eligible individuals are in extreme poverty, ESSN eligible refugees is significantly higher while the rate is 14 percent for the ineligible. than among ineligible refugees. 41 Extreme poverty is 71 percent higher for ESSN refugees. In addition, a large gap between the Nationwide statistics show that 76 percent of two groups can be observed in each of the five eligible refugees live with a budget below the regions. In three of the five regions, the inci- poverty line of TL 284 per month, while 55 per- dence of extreme poverty among the ESSN ref- cent of ineligible do so. The incidence of pover- ugees is double or higher than extreme poverty ty among ESSN refugees is higher than among among the ineligible. ineligible refugees by a proportion of 38 per- cent.42 Moreover, the difference remains sub- It is important to make sure that this finding is stantial across all regions of the country. In pro- not the result of choosing two arbitrary pover- portional terms, ESSN refugees are poorer by a ty lines. Therefore, a complementary analysis magnitude that ranges from 33 to 51 percent. is conducted comparing eligible and ineligible The ESSN criteria were effective in identifying along the entire distribution of per capita ex- a relatively poorer population, at the national penditure. It goes to show that at any possible level, and within every region. That said, at 55 poverty line, the finding that the incidence of poverty is significantly higher among ESSN ref- ugees is unequivocal. On average the eligible 40 The WFP asked the World Bank to lead the targeting and population is substantially poorer than the in- performance analysis for the report to preserve an objective eligible applicant population, and this is robust perspective since WFP also leads ESSN’s implementation. to whatever threshold is chosen to determine The results of section 6 were therefore led by the World poverty status. Bank alone. 41 This section presents the main findings of the analysis. For It is also useful to study proximity to the pover- additional and detailed results see Cuevas et al (2019). ty line as a measure of vulnerability. It is inter- 42 Calculated as (76/55 – 1) × 100. esting to consider the size of the groups on both | 31 Figure 14: Poverty rate among eligible and ineligible ESSN refugee population 90% 80% 76% 78% 76% 80% 73% 69% 70% 55% 55% 57% 60% 51% 51% 53% 50% 40% 30% 20% 10% 0% National Istanbul Aegean Anatolia/ Mediterranean South-East Thrace Ineligible Eligible Note: Poverty line = TL 284 per person per month. Source: PAB data, authors’ calculations. Figure 15: Extreme poverty rate among eligible and ineligible ESSN refugee population 100% 80% 60% 40% 32% 28% 24% 23% 14% 14% 18% 14% 17% 20% 9% 9% 12% 0% National Istanbul Aegean Anatolia/ Mediterranean South-East Thrace Ineligible Eligible Note: Extreme poverty line = TL 165 per person per month. Source: PAB data, authors’ calculations. sides of the poverty line: (a) those vulnerable refugees have an expenditure level just 10 per- to fall into poverty, with per capita expenditure cent below the poverty line, close to exiting above but close to the poverty line and (b) those poverty. prone to exit poverty, with resources below but close to the poverty threshold. To complete the comparative assessment of vulnerability between eligible and ineligible, Visually, this can be done by inspecting the ex- the analysis turns to the other six dimensions penditure cumulative distributions. This reveals of vulnerability. The findings reveal that the that 7 percent ESSN refugees live with budgets population targeted by ESSN is more vulner- just 10 percent above the poverty line of TL 284 able across most, though not all, dimensions. per month. On the flip side, 10 percent of ESSN In particular, the most noticeable differences 32 | V ulnerabilit y and P rotection of R efugees in T urke y Figure 16: At any value of poverty line, ESSN eligible refugees are poorer Cumulative Distribution of Expenditure per Capita by Eligible and Ineligible Status 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 100 200 300 400 500 600 700 800 900 Eligible Ineligible 165 TL Pov Line 284 TL Poverty Line Source: PAB data, authors’ calculations. are found in the use of coping strategies. ESSN target population is poor households, and the refugees have a consumption coping index that targeting criteria were developed by identifying is 32 percent higher than ineligible refugees. the demographic characteristics as best proxy They also make use of crisis livelihood coping to the population that could not meet basic strategies and emergency livelihood coping needs. The ESSN coverage rate, the percent- strategies 20 and 15 percent more than ineligi- age of the poor population assessed eligible, ble, respectively.43 reaches 66 percent. Ideally, the coverage of the poor would be 100 percent. The gap between Coverage and targeting the coverage rate and 100 is typically called ex- clusion error, which in this case amounts to 34 The first standard metric used to evaluate the percent of the poor. Among the extreme poor, performance of an assistance program is cov- the coverage rate is 71 percent, and the exclu- erage of the target population. For ESSN, the sion rate is therefore 29 percent. The other side of the coin of assessing a pro- 43 There are also large differences in household composi- gram’s targeting performance is the inclusion tion, such as dependency ratios, but that difference arises error. This is given by the share of nonpoor refu- by construction as a result of the targeting criteria, making it gees among all eligible refugee population. For redundant to report. the ESSN program, the inclusion error stands at 6 . H o w well d o es E S S N ta r g et and p r o tect vulne r able r efu g ees in T u r k e y ? | 33 Table 7: Comparative vulnerability Figure 17: Coverage of ESSN, by poverty between ESSN eligible and ineligible status (individual level) across dimensions Proportion in which vulnerability is high- 80% 71% er among eligible relative to ineligible, by 66% dimension (%) 70% 59% Share of households with no school-age 60% 14 children attending school 50% 43% Percentage of population with Unac- 9 40% ceptable Food Consumption Average consumption-based Coping 30% 32 Strategies Index 20% Percentage of population that used 20 10% crisis livelihood coping strategies Percentage of population that used 0% 15 Extreme Moderate Non-Poor Overall emergency livelihood coping strategies Poor Poor Percentage of population living in households where no member can 10 Source: PAB data, authors’ calculations. speak Turkish Percentage of population living in households where no member can read 3 Figure 18: ESSN beneficiary incidence Turkish and inclusion error Percentage of population living with Beneficiary Incidence 6 unskilled labor as main income source Median debt as share of total expendi- Non-Poor Extreme Poor 1 ture 24.0% 23.8% Note: For each indicator, calculated as (value for eligible / value for ineligible) – 1. Source: PAB data, authors’ calcu- lations. 24 percent. Given the data and time constraints under which the system was designed, the ratio of only 1 nonpoor eligible person out of every 4 eligible refugees seems in principle like a rel- atively good performance. Holding the budget constant, if the program had been allocated at Poor random within the applicant population, the in- (Not extreme) 52.2% clusion error would have been 33 percent. The targeting strategy was able to reduce it to 24 Source: PAB data, authors’ calculations. percent. markers of performance are calculated among The exclusion rate stands at 34 percent of the different populations (different denominators), poor, and the inclusion error reaches 24 per- makes it difficult to compare its magnitudes. To cent of the eligible. The fact that these typical get around this, it is useful to look at coverage 34 | V ulnerabilit y and P rotection of R efugees in T urke y Figure 19: ESSN eligibility and poverty status across applicant refugee population Extreme Poor Poor but not Extreme Poor Non-Poor 14% 6% 31% 17% 14% 19% Eligible Ineligible Eligible Poor Ineligible Poor Eligible Ineligible Extreme Poor Extreme Poor (Not extreme) (Not extreme) Non-Poor Non-Poor 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Source: PAB data, authors’ calculations. rates by poverty groups in a mutually exclusive receiving as a share of their total pre-transfer format, as a percentage of the overall refugee expenditure.45 Adequacy figures portray the ca- population. This approach yields that 6 percent pacity of the program to adequately support ba- are extreme poor and ineligible, 17 percent are sic needs. In other words, even if targeting was poor (but not extreme poor) and ineligible, and perfect, a program would have low adequacy if 14 percent are nonpoor and eligible. Two pol- it made cash transfers that were too small to icy implications can be taken from this. First, make a difference in people’s lives. Helping ref- it means that even if the perfect targeting tool ugees meet basic needs is a core programmatic was available, a budget neutral reallocation objective of ESSN. would not be able to cover all the poor. Second, complementary targeting efforts, perhaps us- In the case of ESSN, the adequacy ratio is 96 ing community leaders’ knowledge, should be percent among the extreme poor and 63 per- deployed to reach the 6 percent of the appli- cent among all poor. The size of the ESSN cash cant population that does not meet any ESSN transfers is sufficient to significantly raise ben- eligibility criteria but is in dire need.44 eficiaries’ budgets, measured as a proportion of pre-transfer expenditure levels.46 Adequacy of transfers Benefit adequacy represents the total value 45 The adequacy indicator is sometimes also called gener- of transfers that eligible households will start osity, alluding to how generous a program is with the size of its transfers. 46 One limitation of these magnitudes is that the PAB data 44 In November 2018, the SASF ‘discretionary allowance’ likely underestimate household expenditures due to the was rolled out nationwide. This permits the local SASF of- phone-based limited-length expenditure module. Alterna- fices to identify and include extremely poor households who tively, these can be interpreted as an upper bound of the are ineligible by the demographic criteria. estimate of the adequacy ratios. 6 . H o w well d o es E S S N ta r g et and p r o tect vulne r able r efu g ees in T u r k e y ? | 35 The first benchmark is how well programs cov- Figure 20: Adequacy of ESSN transfer size er the poorest 20 percent of the population (the (as percentage of per capita expenditure) bottom quintile). The ASPIRE database shows 100 96.1 there is ample variation across countries, with the median coverage rate at about 16 percent % of average per capita expenditure 90 and the average at 24 percent of the bottom 80 quintile. A conservative estimate of ESSN cov- 70 62.9 erage of the poorest 20 percent of the popu- 60 lation is 36 percent.48 This places the coverage 51.4 performance of ESSN above the median among 50 international experience. 40 32.5 30 Cross-country comparisons can be difficult 20 to interpret given that poverty rates vary from 10 country to country. The coverage of a cash transfer program will need to be larger or 0 Extreme Overall Non-Poor All smaller depending on how high the incidence Poor Poor Eligible of poverty is. For example, a coverage rate of 50 Note: Calculated as the ratio of TL 120 to average per capita percent among the poorest 20 percent cannot expenditure per month of each group. Source: PAB data, authors’ be interpreted the same way in a country with a calculations. poverty rate of 10 percent as in a country with a poverty rate of 60 percent. For the former, that Benchmarking performance to coverage rate would mean a good performance, international experience but for the latter the opposite would be true. To value ESSN’s targeting and adequacy perfor- Therefore, the analysis brings context-specif- mance, it is helpful to contextualize it in terms ic poverty rates into the mix. ESSN, compared of global experience. The World Bank’s ASPIRE to countries with poverty rates similar to those database (Atlas of Social Protection Indicators of Resilience and Equity) provides informa- tion on the performance of a variety of social 48 At the time of the PAB survey, there were 3.2 million refu- assistance programs across the world.47 The gees in Turkey. Half of them applied to ESSN. The coverage database includes performance indicators of of the bottom quintile in the applicant population in the PAB unconditional cash transfer schemes, offering is 71.5 percent. If the income distribution of non-applicants a valuable opportunity to benchmark the ESSN is similar to the applicants, the coverage of the bottom quin- program to a comprehensive set of compara- tile in the overall population is calculated as 71.5 / 2 = 35.75 tors around the world. percent. It turns out this is a conservative estimate (a low- er bound) since (a) a survey shows indicative evidence (not representative) that poverty among non-applicants is lower than among applicants (CVME Survey, WFP 2017) and (b) it 47 See Ivaschenko et al. (2018) and ASPIRE (World Bank assumes none of the 1.6 million that had not applied at the 2018b). time of the PAB would apply later on. 36 | V ulnerabilit y and P rotection of R efugees in T urke y Figure 21: Poverty and coverage rates of unconditional cash transfer programs across the world ESSN has high coverage rate among programs for populations affected by high poverty rates 100 90 Coverage Rate of Bottom Quintile 80 70 60 50 40 ESSN 30 20 10 0 0 10 20 30 40 50 60 70 80 Context-specific Poverty Rate Note: Context-specific Poverty Rates use the following lines: For low-income countries (LICs) US$1.90, for lower-middle-income countries (LMICs) US$3.20, for upper-middle-income countries (UMICs) US$5.50 (all in per capita per day PPP 2011 terms). Coverage is measured as share of the poorest 20 percent of the population included in the program. Source: PAB and ASPIRE data, authors’ calculations. of the refugee population, shows a relatively Compared to programs that serve populations high coverage. ESSN’s coverage of the poorest affected by poverty rates in the vicinity of 60 20 percent stands at 36 percent, higher than percent, ESSN ranks low in reducing inclusion most programs serving populations with pov- error. This is driven, in part, by ESSN’s high cov- erty rates in the neighbourhood of 60 percent erage rate. In a ‘perfect world’, if ESSN covered reaches. In turn, ESSN’s targeting does a rela- all poor applicants (67 percent), and none of tively good job in reducing exclusion error. the nonpoor, the share of eligible that come from the poorest 20 percent would be equal In terms of beneficiary incidence of the bot- 30 percent (20/67x100), which is moderately tom quintile, the share of eligible that come higher than the 24 percent attained by ESSN. from the poorest 20 percent of the population, ESSN shows a modest performance compared In sum, in the trade-off between being more to other countries. A conservative estimate of inclusive versus being more accurate, the ESSN beneficiary incidence in ESSN sits at 24 per- cent, which implies a larger inclusion error.49 cants shows they have lower poverty and higher expendi- tures than beneficiaries and applicants (CVME Survey, WFP 49 It is a conservative estimate because the PAB only in- 2017). This is indicative, since the CVME survey is not repre- cludes the applicant population, and a survey of non-appli- sentative and that is why it was not included in the analysis. 6 . H o w well d o es E S S N ta r g et and p r o tect vulne r able r efu g ees in T u r k e y ? | 37 Figure 22: Poverty and beneficiary incidence of unconditional cash transfer programs across the world ESSN has low Beneficiary Incidence among programs for populations affected by high poverty rates 80 Beneficiary Incidence of Bottom Quintile 70 60 50 40 30 ESSN 20 10 0 0 10 20 30 40 50 60 70 80 Context-specific Poverty Rate Note: Context-specific Poverty Rates use the following lines: For LICs US$1.90, for LMICs US$3.20, for UMICs US$5.50 (all in per capita per day PPP2011 terms). Beneficiary Incidence is measured as share of beneficiaries that belong to the poorest 20 percent of the population. Source: PAB and ASPIRE data, authors’ calculations. targeting performance is balanced toward in- The final benchmark is in regard to adequacy clusion, while the majority of unconditional of support, the size of the transfer as a propor- cash transfer programs in the global context tion of pre-transfer expenditures. The ESSN are tilted toward more accuracy. ESSN’s perfor- adequacy ratio is significantly above all other mance can be characterized by a ‘lower exclu- unconditional cash transfers. This comparison sion error-higher inclusion error’ balance with- is indicative; since the report analyzes ESSN in the set of possible combinations delimited by performance before transfers start, the ESSN international experience. adequacy ratio is calculated ex ante. In the AS- PIRE data, adequacy is computed after trans- In the context of humanitarian assistance to fers start, so it reflects the ex post situation. refugee populations, prioritizing coverage and inclusion at the expense of accuracy is per- The larger the poverty gap—which captures how haps the preferred balance, since after all, the far below the poverty line are the budgets of the nonpoor population that contributes to the in- poor—the higher the transfer and the adequacy clusion error are relatively vulnerable. About need to be. This is confirmed by international 30 percent of nonpoor beneficiaries live with practice across cash transfer programs. The a budget that is within 10 percent of the pov- theoretical, that is, pre-transfer, performance erty line. of ESSN in this regard is positive. 38 | V ulnerabilit y and P rotection of R efugees in T urke y Figure 23: Poverty and adequacy of unconditional cash transfer programs across the world ESSN has high adequacy ratio among programs for populations with high poverty gaps 120 Adequacy Ratio among Bottom Quintile Beneficiaries 100 ESSN 80 60 40 20 0 0 5 10 15 20 25 30 35 Context-specific Poverty Gap Note: Context-specific poverty gaps use the following lines: For LICs US$1.90, for LMICs US$3.20, for UMICs US$5.50 (all in per capita per day PPP 2011 terms). Adequacy is measured as transfer value as a share of pre-transfer welfare, among the poorest 20 percent of the population. Source: PAB and ASPIRE data, authors’ calculations. Benchmarking performance to cants would likely be different than the exist- alternative untargeted design ing one. First of all, more people would apply if they knew everyone was eligible. Second, the If instead of the current targeted design, ESSN composition of the applicant pool would like- had gone for an untargeted, universal trans- ly change in terms of vulnerability. In the first fer to all refugees, while keeping the budget months of operation, it may be that the popu- constant, how would it perform? This counter- lation that applies to the program under either factual, hypothetical question does not have a targeted and untargeted scenarios is similar. straightforward answer. Because of the policy But as time goes by and it becomes known that relevance of the untargeted versus targeted everyone applying can get a cash transfer, the question and despite the methodological dif- marginal applicants under the untargeted de- ficulties, this section makes an informed at- sign would be less vulnerable than under the tempt to contribute to a better understanding targeted scenario. of the issue. To simulate the performance of an untargeted, If the program distributed transfers in an un- universal transfer design, the analysis makes targeted, universal form, the pool of appli- use of bounds, as explained in the methodology 6 . H o w well d o es E S S N ta r g et and p r o tect vulne r able r efu g ees in T u r k e y ? | 39 Table 8: Performance indicators of simulated untargeted transfer (%) Coverage of Exclusion Inclusion Adequacy the poor error error among the poor Current targeted design 66 34 24 63 Simulated untargeted design - 100 0 33 37 lower bound Simulated untargeted design - 100 0 >33 20 upper bound Source: PAB data, authors’ calculations. section. In the lower-bound scenario, the un- expenditure levels. In contrast, the current tar- targeted design would achieve full coverage of geted transfer would increase the resources the poor, exclusion error would be eliminated, of the poor by 63 percent of their pre-transfer and inclusion error would be maximized and expenditures. The adequacy of the untargeted equal to 33 percent, that is, the share of the ap- transfer is much lower than the adequacy of the plicant population in the PAB that is nonpoor. targeted transfer. The untargeted transfer’s adequacy among the poor would be 37 percent. Under the upper-bound scenario, exclusion er- ror would also be fully addressed, and inclusion error would be at least 33 percent, depending on the share of the population that is not poor among the refugee population at large. With a monthly per capita transfer of TL 39, the ad- equacy of the untargeted transfer among the poor would just be 20 percent. For a given budget, the untargeted design pres- ents clear contrasts with respect to the existing targeted design. Its most important advantage in terms of protection is that it would address the exclusion error borne by any targeted de- sign. Currently, the ESSN targeting criteria miss 34 percent of the poor. But the price paid by the untargeted transfer is maximizing inclusion er- ror and, more importantly, significantly reduc- ing the support given to the poor. An untargeted transfer would increase the resources of the poor by 20 to 37 percent of their pre-transfer 40 | V ulnerabilit y and P rotection of R efugees in T urke y 7. Policy Discussion and Way Forward ESSN is the largest humanitarian assistance strained by data and time, ESSN achieves good program in support of refugees and forcibly dis- performance marks. Among the key drivers of placed populations across the world. The pro- this have been the complementarities in design gram confronted the challenge of supporting and implementation capacity of the WFP-TRC- the most vulnerable refugees residing in Turkey MoFLSS partnership, which made it possible to with a given budget. This report analyzes the receive, process, and validate thousands of ap- rollout of the program, its coverage, and target- plications across the country. 50 ing, as well as the vulnerabilities of its eligible population, before transfers get distributed. The second policy lesson is that in the univer- What can be learned from its choices and ex- sal versus targeted debate, the report shows perience? that ESSN’s targeting already looks more ‘uni- versal’ than other cash transfer programs while ESSN uses demographic-based targeting crite- at the same time avoiding the low-adequacy ria chosen as proxy for being poor and unable to drawback that untargeted transfers suffer from. meet basic needs. The current design of ESSN Benchmarking with international experience is able to identify and cover 66 percent of the and a simulated untargeted transfer shows that poor, and 76 percent of its eligible population ESSN achieves relatively high coverage, at the lives within a budget below the poverty line. expense of higher inclusion error, but provides Relative to pre-transfer household expendi- meaningful support to its poor beneficiaries. tures, ESSN is designed to raise the budgets of An untargeted transfer would compromise the the poor by 63 percent. However, the current ability of each household to meet their basic design misses a third of the poor, and a quarter needs, which is the main programmatic objec- of its beneficiaries are nonpoor. Overall, ESSN tive of ESSN. does relatively well when compared to uncon- ditional cash transfers programs around the While ESSN decision makers gave priority to world. the advantages of the targeted design over the untargeted one given ESSN’s programmatic ob- The first policy lesson that emerges is ‘yes we jective and budget constraints, a third policy can’. ESSN shows a relatively positive coverage implication of the analysis is that some form of and targeting performance within the range of results observed in the global arena, espe- cially considering that comparator programs 50 See Maunder et al. (2018) for a thorough description of have been in place for longer. In a context con- the ESSN process and how it builds on government systems. | 41 support is needed for the ineligible population, program in the world. Ongoing data collection since poverty is still high among them. A prom- and analysis will be fundamental for improve- ising space for policy action could be found ments in ESSN. Moving forward, on the analyti- in improving access to economic opportuni- cal front, the WFP and the World Bank will con- ties. After all, the eligibility criteria are closely duct an evaluation of the impacts of the ESSN related to the lack of capacity to work, which assistance on beneficiary lives. implies that the ineligible have relatively better prospects for employability and income gen- Additionally, on the operational front, as the eration with the right support, such as Turkish conflict underlying the refugee influx becomes language skills. more protracted, ESSN should transition from a humanitarian-type to a development-type As a fourth policy implication, the report sug- response to promote a sustainable exit from gests that something needs to be done to de- poverty and vulnerability. In that regard, ESSN crease exclusion error. If the budget constraint stakeholders are starting to focus on designing cannot be relaxed, complementary actions strategies to transition beneficiaries toward need to be put in place to increase inclusion of income opportunities going forward. The FRiT the poor. The SASF Allowance, implemented as Office of the Presidency of Turkey and MoFLSS of November 2018, presents a promising way to have recently outlined a strategy to guide these reduce the exclusion error. It is recommended efforts.51 that this allowance finds a way to incorporate knowledge of community leaders, Muhtars, and Future developments in ESSN should be docu- local organizations, who could make referrals mented and made publicly available, to ensure into the SASFs for consideration within the al- the ESSN experience can be productively used lowance to foster equitable access. for policy discussions and programming re- sponses in other humanitarian challenges and An alternative design, in pursuit of balance of contexts. This report is a building block toward the two protection objectives of coverage and that larger, longer-term objective. adequacy, that is, to reach the poor and mean- ingfully support them, could be to have a com- // promise solution between the targeted and untargeted cases. That is, to offer cash trans- fer support to anyone who applies and allocate top-ups to provide additional cash assistance to those in more vulnerable conditions. This would incorporate the advantages of both ap- proaches, although it would still require con- sidering a moderate increase in the available program budget. The evidence presented here intends to inform the humanitarian community of key lessons 51 Refer to FRiT Office of the Presidency of Turkey and from the largest humanitarian cash transfer MoFLSS (2019) for details. 42 | V ulnerabilit y and P rotection of R efugees in T urke y References AIDA (Asylum Information Database) (2018). Deaton, Angus, and Zaidi, Salman (2002). Introduction to the Asylum Context in Tur- “Guidelines for Constructing Consump- key. Retrieved October 2018. https://www. tion Aggregates for Welfare Analysis.” asylumineurope.org/reports/country/tur- LSMS Working Paper No. 135, World Bank. key/introduction-asylum-context-turkey. https://openknowledge.worldbank.org/ handle/10986/14101. Bailey, S., Harvey, P. (2017). Time for Change: Harnessing the Potential of Humanitar- Development Initiatives (2017). Global Human- ian Cash Transfers. Overseas Develop- itarian Assistance Report. Development ment Institute. www.odi.org/publications Initiatives (DI). http://devinit.org/post/ /10764-time-change-harnessing-poten- global-humanitarian-assistance-2017/. tial-humanitariancash-transfers. DGMM (Directorate General of Migration Man- Beegle, Kathleen; De Weerdt, Joachim; Fried- agement) (2018). Temporary Protection man, Jed; Gibson, John (2012). “Methods Statistics. Ministry of Interior, Directorate of Household Consumption Measurement General of Migration Management, De- Through Surveys: Experimental Results cember 20 2018, www.goc.gov.tr/icerik6/ from Tanzania.” Journal of Development temporary-protection_915_1024_4748_ Economics 98 (1): 3–18. icerik. Cash Learning Partnership (2018). The State FRiT Office of the Presidency of Turkey and Min- of the World’s Cash Report: Cash Transfer istry of Family, Labor and Social Services Programming in Humanitarian Aid. CaLP. (2019). Exit Strategy from the ESSN Pro- http://www.cashlearning.org/downloads/ gram. Ankara, Turkey. https://ailevecalis- calp-sowc-report-web.pdf ma.gov.tr/media/3725/essn-exit-strate- gy-1.pdf Cuevas, P. Facundo; Inan, O. Kaan; Twose, Aysha; Çelik, Çiğdem (2019). “Vulnerabil- Gentilini, Ugo (2016). The Other Side of the ity and Protection of Refugees in Turkey: Coin: The Comparative Evidence of Cash Ex-Ante Analysis of the Largest Human- and In-Kind Transfers in Humanitarian Sit- itarian Cash Assistance Program in the uations. World Bank Studies. Washington, World.” Policy Research Working Paper, DC: World Bank. https://openknowledge. World Bank, Washington, DC. worldbank.org/handle/10986/24593. | 43 Grand Bargain Signatories (2016). The Grand Bar- sites/14/2018/09/01.-UNHCR-Turkey- gain - A Shared Commitment to Better Serve Fact-Sheet-August-2018.pdf. People in Need. Available at: http://www. agendaforhumanity.org/initiatives/3861. World Bank (2015). Turkey’s Response to the Syrian Refugee Crisis and the Road Ahead. High-Level Panel on Humanitarian Cash Trans- Policy Report No. 102184. Washing- fers (2015). Doing Cash Differently: How ton, DC: World Bank Group. Available at: Cash Transfers Transform Humanitarian https://openknowledge.worldbank.org/ Aid. Report of the High-Level Panel on Hu- handle/10986/23548. manitarian Cash Transfers. https://www. odi.org/sites/odi.org.uk/files/odi-assets/ ——— (2018a). Turkey Poverty and Equity Brief. publications-opinion-files/9828.pdf. Poverty and Equity Global Practice. The World Bank. http://databank.worldbank. Jolliffe, Dean; Prydz, Espen Beer (2016). “Es- org/data/download/poverty/33EF03BB- timating International Poverty Lines from 9722-4AE2-ABC7-A A2972D68AFE/ Comparable National Thresholds.” Policy Global_POVEQ_TUR.pdf. Research Working Paper No. 7606, World Bank, Washington, DC. ——— (2018b). ASPIRE: The Atlas of Social Pro- tection Indicators of Resilience and Equity. Maunder, N.; K. Seyfert; M. Aran; G. Baykal; M. (accessed July 2018) http://datatopics. Marzi; and G. Smith (2018). Evaluation of worldbank.org/aspire/. the DG ECHO Funded Emergency Social Safety Net (ESSN) in Turkey. Oxford Policy WFP (World Food Programme) (2008). Food Management. Consumption Analysis: Calculation and Use of the Food Consumption Score in Maxwell, Daniel, and Caldwell, Richard (2008). Food Security Analysis. Technical Guid- Coping Strategies Index: Field Methods ance Sheet. Rome. WFP Vulnerability As- Manual. 2nd edition. CARE and the World sessment and Mapping Unit. Food Programme Vulnerability Assess- ment and Mapping Unit. ——— (2016). Basic Needs Programming in Tur- key: Establishing Targeting Criteria and a New York Times (2017). “How to Treat Refu- Minimum Expenditure Basket. WFP Vul- gees with Dignity: A Lesson from Turkey.” nerability Assessment and Mapping re- Op-Ed by Rula Jebreal, New York Times, port. Available at: https://documents.wfp. September 27, 2017. Available at: https:// org/stellent/groups/public/documents/ www.nytimes.com/2017/09/27/opinion/ ena/wfp284420.pdf. turkey-syrian-refugees.html. ——— (2017). Comprehensive Vulnerability Mon- UNHCR (United Nations High Commission- itoring Exercise. CVME Round 1. https:// er for Refugees) (2018). Turkey Fact- www.wfp.org/content/turkey-comprehen- sheet: August 2018. Available at: http:// sive-vulnerability-monitoring-exercise. www.unhcr.org/tr/wp-content/uploads/ 44 | V ulnerabilit y and P rotection of R efugees in T urke y Vulnerability and Protection of Refugees in Turkey: Findings from the Roll- out of the Largest Humanitarian Cash Assistance Program in the World assesses the targeting performance and benefit level design of the Emergency Social Safety Net (ESSN) program for refugees in Turkey. It also provides a comprehensive look at the vulnerability of ESSN eligible households using a multidimensional lens, drawing from novel represen- tative data. The ESSN provides monthly cash transfers to help the most vulnerable refugees meet their basic needs, and complement Turkey’s response to the crisis. With near 4 million refugees, Turkey hosts more refugees than any other country in the world. The program is funded by the European Union member states, and implemented nationwide in partnership with the Ministry of Family, Labor and Social Services, the World Food Pro- gramme, and the Turkish Red Crescent. The study finds that the vulnerabilities of the ESSN refugee population are multiple and complex. Refugees in the ESSN program suffer from a shortage of resources today, but also resort to coping strategies that cripple their resource-generating capacity tomorrow. The ESSN targeting criteria are relatively effective in selecting the most vulnerable refugees, but exclude a share of the poor. This issue is starting to get addressed by decentralized allowances targeted with community-level information. The ESSN cash transfer value, is found to be adequate to support basic needs. An untargeted design would have minimized exclusion errors, but would reach everybody with smaller transfers, insufficient to meet basic needs. Future analysis will focus on the impact of the transfers on house- hold welfare.