Asylum Seekers in the European Union: Building Evidence to Inform Policy Making 2 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making © 2018 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW Washington DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and conclusions ex- pressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they rep- resent. The World Bank does not guarantee the accuracy of the data includ- ed in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judg- ment on the part of The World Bank 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 encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Any queries on rights and licenses, including subsidiary rights, should be addressed 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. Cover design, interior design: Carlos Reyes, Reyes Work Studio. Table of Contents Acknowledgements 5 Acronyms 7 Overview 9 I. Finding Evidence to Support Policy Making 13 The Policy Challenge 14 Study Objective and Main Findings 15 II. Context, Data, and Methodology 17 The Context 17 Data Sources 21 Methodology 24 III. Sociodemographic Profiles of Asylum Seekers 27 Three Broad Groups of Asylum Seekers 27 Nationals of High-Recognition Countries 30 Nationals of Low-Recognition Countries 35 The Decision to Migrate 38 Primary Movers, Secondary Movers, and Stayers 38 IV. The Migration Experience 45 A Long and Perilous Journey 45 The Financial Cost 47 The Physical and Emotional Toll 48 Intended Destinations 52 Few Planned to Return Home 52 V. Education, Skills, and Work Experience 55 Education and Country of Origin 56 Work Experience 58 Speakers of European Languages 60 Literacy Proficiency Levels Compared 61 Asylum Seekers and Other Recent Migrants in Europe 62 VI. Anxiety and Depression 67 Anxiety, Depression, and Exposure to Violence in Transit 68 VII. Conclusions and Insights into Policy 73 References 76 Appendix 79 5 Acknowledgements This study was led by the World Bank Poverty and Equity Global Practice and the Re- search and Data Groups of the Development Economics Vice-Presidency and prepared by a team comprising Mohamed Abdel Jelil, Paul Andres Corral Rodas, Anais Dahmani Scuitti, María E. Dávalos, Giorgia Demarchi, Neslihan Demirel, Quy-Toan Do, Rema Han- na, Deivy Houeix, Sara Lenehan, and Harriet Mugera, with contributions from Rima Al- Azar, Kartika Bhatia, Daniele Bolazzi, Sédi-Anne Boukaka, Gero Carletto, Setou Diarra, Rawaa Harati, Jonathan G. Kastelic, Nealia Khan, and Taies Nezam. The literacy assess- ment was developed by the OECD, particularly Marta Encinas and Francois Keslair, who also provided analytical support for the study. The work was done under the leadership of Arup Banerji, Asli Demirguç-Kunt, Shan- tayanan Devarajan, Xavier Devictor, Francisco Ferreira, Saroj Kumar Jha, Luis Felipe Lopez Calva, Ana L. Revenga, and Carolina Sanchez Paramo. It was financed by Trust Funds from the Fragility, Conflict, and Violence Cross-Cutting Solutions Area and the Umbrella Facility for Gender Equality. Data collection was managed by the Center for Development Data (C4D2), the World Bank’s Rome-based hub for innovation in household surveys and agricultural statistics, and conducted in collaboration with the Ministry of Interior in Italy and the Ministry of Migration Policy in Greece. The fieldwork was carried out by the Centro Studi di Politica Internazionale in Italy and Kapa-Research in Greece. The team is grateful for comments from peer reviewers Andrew Dabalen, Supriyo De, Roberta Gatti, Rainer Münz, Abla Safir, Kinnon Scott, and Kirsten Schuettler. The team also received valuable comments from Caroline Bahnson, Christian Bodewig, Jishnu Das, Xavier Devictor, Jo de Berry, Ruth Hill, Kristen Himelein, Nandini Krishnan, Daniel Lederman, Federica Marzo, Moritz Meyer, Ana Maria Munoz, Jacob Shapiro, Aki Stav- rous, Paolo Verme, Tara Vishwanath, and Roy van der Weide. The team also appreciates the support throughout the preparation of this report from Julia Barrera, Marta Mueller 6 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Guicciardini, Sergio Lugaresi and Nikolaos Schmidt, and the logistical support of Ar- manda Carcani and Indiana Taylor. The report benefitted from editing by Anne Grant, and cartographic inputs from Bruno Bonansea. The team thanks the staff of the Unit- ed Nations High Commissioner for Refugees (UNHCR) in Italy and Greece for their help throughout the study and the local authorities and staff in centers who made fieldwork possible. The report benefitted in Brussels from comments of the European Council, the Euro- pean Commission Directorate General for Migration and Home Affairs, the Director- ate General for European Neighborhood Policy and Enlargement Negotiations, the Directorate-General for International Cooperation and Development, the Director- ate General for Employment, Social Affairs and Inclusion and the European Political Strategy Centre; in Greece from the Ministry of Migration Policy; and in Italy from the Ministry of Interior and the Prime Minister’s Migration Sherpa. The team also received valuable comments from the International Organization for Migration team in Italy and Greece, and the UNHCR team in Greece, Italy, and Switzerland. 7 Acronyms CTRPI Commissione Territoriale per il Riconoscimento Protezione Internazionale [Regional Commission for the Granting of International Protection] CFR Council on Foreign Relations EASS Euro Asylum Seekers’ Survey 2017 EC European Commission EU European Union FGD Focus group discussion GSI Global Severity Index ICD International Statistical Classification of Diseases and Related Health Problems IDI In-depth interview ILO International Labour Oganization IOM International Organization for Migration KRI Kurdistan Region of Iraq OECD Organisation for Economic Co-operation and Development PIAAC Programme for the International Assessment of Adult Competencies PTSD Post-traumatic stress disorder UNHCR UN Refugee Agency, Office of the United Nations High Commissioner for Refugees Distinctions Asylum seekers: Individuals who are the subjects of this study because they have applied for international protection and are awaiting a decision. Migrants: All people on the move, regardless of their presence or legal status in Europe. Refugees: Individuals who have been granted asylum or another form of international protection. Primary or first movers: Those who moved from their home countries directly to Greece or Italy. Secondary movers: Those who had settled in another country foreign to them be- fore moving to the EU. Recognition rate: The ratio of asylum applicants granted protection to the total number of asylum decisions in a given year. High-recognition country: Those with recognition rates of 40 percent or more. Low-recognition country: Those with recognition rates below 40 percent. 8 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making 9 Overview The need to build evidence an assessment of skills through a computer-based literacy test comparable to the OECD adult literacy Policy needs to be informed by facts: the more that test, and an evaluation of the respondents’ symp- is known about those who may be affected, the toms of anxiety and depression. The skills assess- more realistic and achievable a policy’s goals will ment complements the self-reported educational be. However, while there have been massive move- attainment data. The screening of anxiety and de- ments of people across borders in recent years, pression has implications for how asylum seekers many entering the European Union (EU), there is can be helped to adapt to new situations. little systematic data about them available. Most of the evidence that is available is anecdotal and Who are the asylum seekers? journalistic; it deals primarily with the tragedies of migrants in transit rather than providing hard data The asylum seekers in Greece are different from on which policy-makers can base policies. those in Italy. Those in Greece mostly came as families from the Syrian Arab Republic, Iraq, and In 2015 and 2016, migrant flows into the EU surged, Afghanistan and, based on past EU recognition with Greece and Italy the main entry points. Many rates, are more likely to be granted international of the migrants applied for international protec- protection status. Those in Italy were mainly single tion in Europe, becoming asylum seekers. This young men, a large share from West Africa, with spike in EU asylum seekers, as well as the in- many facing a low probability of being granted any creasing numbers of those granted refugee sta- legal status in Europe. tus, brought a need for information on who they are—their sociodemographic characteristics; their The journey to the EU followed a few main routes: education and work experience; their experience through Niger for West Africans or Sudan for on the journey to Italy and Greece; and what it cost East Africans on their way to Libya to make the them not only financially but also physically and sea crossing to Italy; or through Turkey, passing emotionally to get there. through Iran for Afghans, to make the sea crossing to Greece. The journey was much longer for those This study took a rigorous approach to ensure in Italy, who spent about a month and a half in that it produced hard data to support policy deci- each of about three main transit points, compared sions—decisions made not only in receiving coun- to over a week in each for those in Greece. Partic- tries but also in countries of origin and transit. The ularly on the routes towards Italy, asylum seekers first step, in early 2017, was to survey adults in asy- encountered violence in transit—that happened lum centers in Greece and Italy, to learn more de- to nearly half of those in Italy, even before the per- tails about them and about their experience on the ilous crossing of the Mediterranean in inflatable journey. The study made two novel contributions: boats. 10 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Where there are data for comparison, it appears have settled in Europe for the past five years. In that many of the asylum seekers surveyed were other words, they were similar to recent migrants wealthier than those who stayed at home and did already living in Europe. not migrate—a finding that is perhaps not surpris- ing considering that for Sub-Saharan Africans the A large share of asylum seekers in Italy, about 62 average cost of the journey was about three years percent, had work experience. Among those who of income for a person living in extreme poverty. had held a job, most had worked in basic occupa- tions, like construction and agriculture, before ar- However, not all moved directly from their country riving in Italy and nearly 70 percent did some work of origin to the EU. Some had previously settled in transit. Many West Africans worked without pay in third countries, such as Libya and Iran, before during the journey. Among those in Greece, work embarking on a second migration journey. For ex- experience was minimal, partly because of gender ample, one in five asylum seekers in Italy was one gaps in employment: the population of asylum of these “secondary movers.” Of those, 43 percent seekers there is more balanced between men and were Sub-Saharan Africans who had been living in women, and about 70 percent of the women from Libya and departed when Libya itself erupted in the Syrian Arab Republic, Iraq, and Afghanistan conflict. For many Sub-Saharan Africans, the con- had no work experience. Overall, younger asylum flict likely turned Libya from a destination into a seekers (18–25) are less likely to have ever had a transit country. Among those in Greece, about one job, possibly because they have only recently en- in four were secondary movers, mostly Afghans tered the labor market. who had been living in Iran (of whom about a third had been born there) and Syrians who had been Anxiety and depression can affect the ability of living in Turkey. asylum seekers to exploit their abilities. Instanc- es of mental distress were widespread, possibly Asylum seekers in Greece and Italy had on average caused by a combination of stresses back home, low levels of education—only 32 percent of those during their journey, and in waiting for a deci- in Italy and 29 percent in Greece had completed sion in reception centers. About 70 percent of the secondary school or above. In most cases, this asylum seekers in Greece and almost 50 percent reflects the education levels in their countries of of those in Italy showed elevated levels of severe origin. mental distress. Many asylum seekers speak a European language, Insights into Policy usually English; 80 percent of those in Italy and 45 percent in Greece speak at least one. Many Sub-Sa- While a comprehensive policy agenda on migra- haran Africans, of course, come from countries tion and forced displacement is beyond the scope where English or French are official languages. The of this report, the evidence presented here—based literacy test, administered in official languages by on a large representative sample of adult asylum country of origin, found that in general asylum seekers in Italy and Greece in 2017—can help in- seekers have limited proficiency in the designated form policymaking: language. Not surprisingly, the better-educated did better in the literacy assessment. It is worth • A response to the European influx of migrants noting that the literacy profiles of asylum seekers needs to combine humanitarian with devel- do not differ from those of other migrants who opment approaches, particularly when there Overview | 11 is a context of protracted conflict in countries This report, then, contributes to knowledge of as- of origin. But the same can be said for those pects of migration and forced displacement, but not coming from conflict settings: despite the much more has yet to be learned. financial, physical, and emotional difficulties of their journey, many still make the journey. First, the response by Governments, civil society, It is important to thoroughly understand and development partners, and other stakeholders tackle the vulnerabilities that had confronted will require reinforcement of the evidence base, people in their countries of origin or where so that the response can be adapted to chang- they had initially settled. This is critical in ef- es in the situation and to support more effective forts to find sustainable resolutions of migra- interventions. The difficulties of collecting the in- tion flows. In other words, whether or not they formation needed to support policies that affect come from countries in conflict, asylum seek- transient and vulnerable populations are obvi- ers tend to come from circumstances where, ous—among them resolving methodological and from their point of view, despite the costs and ethical considerations—and they apply in many dangers of the journey to Europe, the oppor- countries worldwide that have had to deal with tunities to be attained were worth the risk. large migrant populations and forced displace- ment. Efforts to collect data and build evidence • The design of interventions to help these mi- need to be sustained, both by systematic collec- grants lead productive lives—either in Europe tion of administrative data and by survey work for the many who will be granted protection that also profiles host communities. or in their countries of origin for those who will return—can be informed by findings on Second, additional data collection and analysis education and skills. Delivery of schooling or would be valuable for (1) comparing EU host com- training and promotion of their engagement munity data with refugee, asylum seeker, or other in the labor market as part of EU integration migrant data to analyze impacts on host commu- policies, for example, need to be adapted nities and attitudes toward migrants of all types; to their initial skills and, importantly, to the (2) profiling asylum seekers and refugees in other needs of the labor market that is expected to EU countries with special attention to vulnerable absorb them. groups and on countries where earlier there were large influxes of migrants; (3) exploring how to capi- • The support that these asylum seekers need talize on current data collection efforts, particularly goes beyond traditional interventions. The administrative, to capture more timely policy-rele- anxiety and depression findings, which pro- vant information; and (4) evaluating the cost-effec- vide an important policy-relevant dimension, tiveness of policies as they are being considered make it clear that the mental health of many in and after they are adopted, such as integration pol- the groups studied is compromised. The asy- icies, transit country measures to curb the flows, lum process can offer a unique opportunity for and roll-out of voluntary return packages. early identification and intervention; certainly, those in asylum centers could benefit directly from targeted psychological support. 12 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making 13 I. Finding Evidence to Support Policy Making The last few years have seen a surge in immi- died or disappeared crossing the Mediterranean grants into the European Union (EU), mainly (UNHCR 2018). through Greece and Italy. In 2015 alone, over 1 million people came in, the majority through Women, men, and children from numerous coun- Greece and of Syrian nationality. About 1.1 mil- tries have entered the EU. Africans mostly came lion people have entered Greece since 2014, and through Italy, 75 percent of them adult men (Figure another 630,000 have crossed the sea to Italy (Fig- 1.2). In 2016, when arrivals in Italy peaked, 20 per- ure 1.1). The sudden massive inflow of migrants cent of the immigrants were from Nigeria and 11 has impressed a sense of urgency in policymak- percent from Eritrea (Figure 1.3). Arrivals in Greece ers in host countries—not only EU Member States (Figure 1.4) were dominated by Syrians (56 percent but also others like Turkey, Jordan, and Leba- in 2016), Afghans (24 percent), and Iraqis (11 per- non—and attracted attention worldwide through cent). More children entered through Greece than international press reports of sea disasters and through Italy. UNHCR data (2018) shows, however, fatalities among migrants struggling to reach the that in 2017 the composition of immigrant flows EU. Since 2014 an estimated 16,000 people have began to change and broaden.1 Figure 1.1. Sea Arrivals, Italy and Greece, 2014–17 Figure 1.2. Demographics of Sea Arrivals, 2014–17, Percent 900,000 100 800,000 90 80 700,000 70 600,000 60 50 500,000 40 30 400,000 20 300,000 10 0 200,000 Italy Greece 100,000 Men Women Children 0 Source: UNHCR (2018). 2014 2015 2016 2017 Greece Italy 1 In 2017 and 2018, immigrants from other countries of origin began to enter, such as those from Bangladesh (8 percent of the flow in the first quarter of 2018), Tunisia (6 percent) and Mo- Source: UNHCR (2018). rocco (5 percent). 14 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Figure 1.3. Nationality of Sea Arrivals, Italy, 2016, the Mediterranean in 2017 alone,2 and the protract- Percent ed Syrian crisis, countries of migrant origin, transit, and destination must all prepare for humanitarian Nigeria 20% and development responses. Other 23% However, the evidence on which to base policy ac- Eritrea tions is minimal. Early in the European migrant cri- 11% sis, administrative data collected by governments and international agencies captured only the total Somalia Guinea number of immigrants entering the EU and some 4% 7% very basic demographics. This data scarcity was, Bangladesh unavoidably, then supplemented by anecdotal 4% Côte d'Ivoire Sudan 7% evidence and media reports on their tragic expe- 5% Mali Gambia riences, the rescue efforts, and the complexity of Senegal 7% 6% 6% managing the rapid and large influx of people into host countries. Today, although increased data col- Source: UNHCR (2018). lection and analysis have expanded knowledge of the backgrounds and experiences of refugees, asy- lum seeker, and other migrants,3 there are still sig- Figure 1.4. Nationality of Sea Arrivals, Greece, nificant gaps in the information needed to support 2016, Percent formulation of effective development responses for any group entering the EU. The gaps are even Others 4% more worrisome given the variety of countries of Iran 3% origin and the contexts the immigrants come from. Pakistan 3% One-size-fits-all policies are unlikely to work. Iraq 10% The need to build a solid foundation of knowledge to fully understand migration and forced displace- ment is not limited to the EU; numerous countries worldwide are confronted by similar challenges. In 2016 the number of displaced persons reached 65.6 million people, 22.5 million of whom are ref- Afghanistan Syrian Arab 24% Republic ugees and 44.1 million displaced within their own 56% country. In 2016 about 17 percent of refugees worldwide were estimated to have reached Turkey Source: UNHCR (2018). The Policy Challenge 2 http://data2.unhcr.org/en/situations/mediterranean. 3 For example, IOM (2016) analyzes the socioeconomic back- The policy response to the migrant crisis in the EU ground of migrants and refugees in Italy; as does Buber-Ennser must continue to evolve. With flows from Africa et al. (2016), who studied the human capital, among other as- into Italy persisting, over 3,000 dead or missing in pects, of refugees in Austria. i. Finding evidence tO SuppOrt pOlicy Making | 15 and 14 percent were in the rest of Europe.4 Other For asylum seekers, migration was very expensive: regions are also hosting a large share of refugees, for Sub-Saharan Africans the average payment is such as the 30 percent in Sub-Saharan Africa and equivalent to about three years of income for a 16 percent in the Middle East and North Africa.5 person living in extreme poverty (the US$1.9/day The scarcity of data is even more of a problem for poverty line of low-income countries). Other costs developing countries. are also significant: Among the African young men reaching Italy, nearly half experienced violence Study Objective and Main Findings en route even before the perilous crossing of the Mediterranean. Yet many still expressed hope for a The objective of this study is to systematically ana- future in Europe and few were willing to return to lyze data on asylum seekers in Italy and Greece to their country of origin. inform migration and forced displacement policies being formulated by policymakers in countries of In general, immigrants into the EU have mainly left origin, transit, and destination. It draws on a large circumstances that, in their views, made the costs sample of asylum seekers to rigorously survey and dangers of the journey to Europe worthwhile. adults in asylum centers in Italy and Greece and This realization can have implications for the de- analyze the information elicited on their sociode- sign of policies to curb the flows—combating mographic characteristics and the migration expe- smugglers, breaking down human trafficking net- rience. One novel aspect of this study was to assess works, and deterrent border control policies—but asylum seeker skills using a computer-based litera- still ensure that those in need of international pro- cy test comparable to the OECD adult literacy test; tection have access to it and also avoid raising the a second was to screen for anxiety and depression. costs and dangers of the journey. Although this report does not attempt to formulate a comprehensive policy agenda on migration and The study’s innovative skills assessment validated forced displacement, the evidence it records may the self-reported educational achievements: Asy- be helpful to policymakers in a variety of ways. lum seeker literacy scores in both Italy and Greece are very similar to those of migrants in other Eu- Asylum seekers are not a monolithic group: Over ropean countries. For those with work experience, 90 percent of those reaching Greece are from the it is mostly in low-skill jobs. These insights can Syrian Arab Republic, Iraq, and Afghanistan, who inform policies both for integration and for return all have a high probability of being granted refugee of migrants to their countries of origin for reinte- status given past EU recognition rates, and most gration. Yet it is important to keep in mind that the travelled in families. In contrast, Italy has attracted study does not cover all entrants before 2017 or significant flows from West African countries, for those who did not pass through Italy and Greece to the most part single young men with probably a enter Europe. Other groups may have more skills low chance of receiving international protection. and job experience than those surveyed here. This study also illustrates that the policy agenda needs to take into account the mental health di- mension. Anxiety and depression levels are high and nearly 70 percent of the asylum seekers in 4 As reported by UNHCR and based on the 17.2 million refugees already under UNHCR mandate. http://www.unhcr.org/5943e8a34.pdf. Greece and around 40 percent of those surveyed 5 Ibid. in Italy suffered from severe mental distress. 16 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Finally, the study sheds light on the dynamics of migration and selection patterns by comparing those who moved from their home countries di- rectly to Greece or Italy (first movers) with those who first moved to another country before moving on to Europe (secondary movers). It also distin- guishes asylum seekers from the national popula- tion in their countries of origin. For example, where data for home countries allowed comparisons, the study found asylum seekers to be wealthier than those who did not migrate. Thus, cost may be a barrier to migrating for some, even though so many consider the benefits to outweigh the costs. In what follows, Chapter II sets out asylum poli- cies and processes at the time of the survey, and describes the data collected for this study. Chap- ter III describes the sociodemographic profiles of asylum seekers and offers insights into why people chose to migrate. Chapter IV details the migration experience itself, in terms of transit, cost, and ex- posure to violence. Chapter V describes the educa- tion, work experience, and skills of asylum seekers, and Chapter VI discusses the extent of their mental distress. Chapter VII concludes with general in- sights into policy gleaned from the analysis. 17 II. Context, Data and Methodology The Context of origin to tackle migration. The Agenda outlined short- and medium-term areas of action on mi- EU Policy Responses to Date gration policy centered on reducing incentives for Early EU responses to the migrant crisis were di- migration; border management for saving lives rected to immediate action to prevent further loss and securing external borders; and strengthening of life at sea and to manage and curb the rapid both asylum and legal migration policies. There inflows (EC 2015). The budget for sea search and followed detailed packages of proposals,6 such rescue operations was expanded, and the EU initiatives as setting up hotspots and relocation moved to undercut smuggling networks by target- mechanisms, and partnerships with countries out- ing their vessels and limiting their online reach to side the EU, like the action plan agreed between potential migrants (EC 2015). Part of this immedi- the EU and African heads of State at the Valletta ate response, reflected in the “European Agenda Summit on migration in November 2015.7 of Migration” put forward in mid-2015, entailed ef- forts to share the pressures of the migration flows Migration flows into Greece plunged after the EU between Member States and work with countries and Turkey agreed on a joint action plan.8 The plan consisted of EU financial support and hu- Figure 2.1. Applications for EU Asylum, 2012–17 manitarian assistance to Turkey to support Syrian refugees and host communities there and to stem 1,600,000 illegal migration. On March 18, 2016, the parties 1,400,000 executed the formal agreement.9 The number of migrants moving from Turkey to Greece plunged 1,200,000 from 124,471 in January and February 2016 to 1,000,000 26,971 in March and 3,650 in April (UNHCR 2018). 800,000 600,000 400,000 6 See for example, http://europa.eu/rapid/press-release_IP-15- 5039_en.htm and http://europa.eu/rapid/press-release_IP-15- 200,000 5596_en.htm. 7 http://www.consilium.europa.eu/media/21839/action_plan_en.pdf. 0 8 http://www.consilium.europa.eu/en/press/press-releas- 2012 2013 2014 2015 2016 2017 es/2015/11/29/eu-turkey-meeting-statement/. 9 http://www.consilium.europa.eu/en/press/press-releas- Source: EASO (2018). es/2016/03/18/eu-turkey-statement/. 18 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making From May to December 2016, on average fewer Asylum and Relocation Processes than 3,000 people reached Greece monthly. As migrant flows spiked in 2015, the number of Although flows to Greece dropped, sea arrivals applications for asylum in the EU peaked (Figure in Italy held steady, and efforts were made to in- 2.1). Asylum seekers are those who have applied crease cooperation with Libya—the main point for international protection and await a decision. of departure for most migrants who reached Ita- The 1951 Refugee Convention (Geneva Conven- ly10—and to work with countries of origin for a de- tion, amended by the 1967 New York Protocol), velopment response to migration (EU Partnership defines who is a refugee, refugee rights, and the Framework under the European Agenda for Migra- obligations of receiving States (Box 2.1).15 Be- tion, EC 2015). These efforts continue; for example, cause EU Member States are signatories of the the European Investment Bank has approved an Geneva Convention and adhere to EU rights and increase in lending of about €3.7 billion for public directives,16 their asylum processes begin with and private projects to respond to the root causes determining refugee status. The EU registered 1.4 of migration.11 million applications for international protection in 2015, 1.3 million in 2016, and 707,000 in 2017. Efforts to facilitate integration of refugees have The vast majority were submitted by first-time ap- also increased. The European Parliament in April plicants, although in 2017 there was an 8 percent 2016 passed a resolution that spoke of “accommo- increase in re-applications. The main countries of dation, literacy and language courses, inter-cultur- origin for EU asylum seekers were the Syrian Arab al dialogue, education and professional training, Republic, Afghanistan, Iraq, Kosovo, and Albania; and also effective access to democratic structures in 2016, because there was a relative decline in in society.”12 Most Member States have adopted Western Balkan applications, the main countries their own policies on integration of asylum seek- of origin were the Syrian Arab Republic, Afghani- ers and refugees,13 and the European Commission stan, Iraq, Pakistan, and Nigeria. Applications from (EC) Action Plan on the integration of third-country West Africans increased from 2016 into early 2017, nationals (adopted in June 2016) included actions particularly from nationals of Nigeria, Guinea, Côte for integrating refugees.14 d’Ivoire, and The Gambia (EASO 2018). In 2017 about 40 percent of asylum seekers were granted international or national protection (EASO 10 See, for example, http://www.consilium.europa.eu/en/press/ 2018). How long a case awaits a decision on an in- press-releases/2017/02/03/malta-declaration/. ternational protection application varies by case 11 http://www.consilium.europa.eu/en/press/press-releas- and country. An application may be rejected; es/2018/02/27/eib-council-approves-extra-3-7-billion-to-ad- dress-migration-issues/. granted giving refugee status or subsidiary protec- 12 European Parliament resolution of 12 April 2016 on the situ- tion (international protection); or granted autho- ation in the Mediterranean and the need for a holistic EU ap- proach to migration (2015/2095(INI)). 13 See the tables mapping Asylum Seekers and Refugees Inte- gration Policies across EU Member States: http://ec.europa.eu/ 15 1951 Refugee convention: http://www.unhcr.org/ social/main.jsp?catId=1274&langId=en&intPageId=4316 en-us/1951-refugee-convention.html. 14 https://ec.europa.eu/home-affairs/sites/homeaffairs/files/ 16 EU Charter of Fundamental Rights: http://fra.europa.eu/en/ what-we-do/policies/european-agenda-migration/propos- charterpedia/article/18-right-asylum, and Council Directive al-implementation-package/docs/20160607/communication_ 2004/83/EC http://eur-lex.europa.eu/LexUriServ/LexUriServ. action_plan_integration_third-country_nationals_en.pdf. do?uri=CELEX:32004L0083:EN:HTML ii. cOntext, data and MethOdOlOgy | 19 Box 2.1. Legal Framework for Refugee Status The 1951 Convention on the status of refugees consolidates previous international policies and codifies their international rights. Article 1 of the Convention endorses a single definition of the term “refugee” that emphasizes protecting persons from political or other forms of persecution. It states that a refugee is someone who “is unable or unwilling to return to their country of or- igin owing to a well-founded fear of being persecuted for reasons of race, religion, nationality, membership of a particular social group, or political opinion.” The Convention is based on such fundamental principles as nondiscrimination, nonpenalization, and non-refoulement. It also sets out basic standards for the treatment of refugees (States may grant more favorable treatment), such as access to the courts, primary education, and employment opportunities, and provides for refugee travel and other documentation. The Office of the United Nations High Commissioner for Refugees (UNHCR), founded in 1950, is charged to provide international protection to refugees and, in cooperation with governments, seek durable solutions for them. The 1967 Protocol, which amends the Convention, removed its geographic and temporal limits, thus giving it universal coverage. The Convention has since been supplemented by refugee and subsidiary protection regimes in several regions and by progressive development of international human rights law. The Protocol obliges States to comply with the 1951 Convention with respect to all persons covered by the refugee definition in Article 1, with no restrictions. The European Union Dublin Regulationa is an EU law that requires asylum-seekers to register their application in the country where they first enter the EU. It identifies the EU country respon- sible for examining an asylum application using a hierarchy of criteria, such as family unity, pos- session of residence documents or visas, irregular entry or stay, and visa-waived entry. In practice, however, the criterion most often applied is irregular entry, which means that the Member State where the asylum-seeker first entered the EU is responsible for examining the asylum application. The Dublin system assumes that, as Member State asylum laws and practices are based on the same common standards, asylum seekers will enjoy similar protection in all EU Member States. In practice, however, asylum legislation and practice are still heterogeneous, so that asylum-seekers are treated differently from one EU State to another. The European immigrant crisis has triggered calls for general reform of the Common European Asylum System, especially the Dublin rules. The Dublin Regulation also proposes a corrective allocation mechanism to streamline and supple- ment current rules that would be triggered automatically if a Member State were confronted by disproportionate numbers of asylum-seekers. a The Dublin Regulation was originally established by the Dublin Convention, signed in Dublin, Ireland, in 1990. In 2003, the Dublin Convention was replaced by the Dublin II Regulation. In 2013, the Dublin III Regulation was adopted, replacing the Dublin II Regulation. 20 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Box 2.2: Criteria for Asylum Decisions Refugee status: a person is granted refugee status (see Directive 2004/83/EC and the Geneva Convention), if he or she, “owing to a well-founded fear of being persecuted for reasons of race, reli- gion, nationality, political opinion or membership of a particular social group, is outside the country of nationality and is unable or, owing to such fear, is unwilling to avail himself or herself of the pro- tection of that country, or a stateless person, who, being outside of the country of former habitual residence for the same reasons as mentioned above, is unable or, owing to such fear, unwilling to return to it.” Further details are provided in the Directive. Subsidiary protection status: a person is granted subsidiary protection (Directive 2004/83/EC) if he or she is “a third country national or a stateless person who does not qualify as a refugee but in respect of whom substantial grounds have been shown for believing that the person concerned, if returned to his or her country of origin, or in the case of a stateless person, to his or her country of former habitual residence, would face a real risk of suffering serious harm.” Further details are provided in the Directive. Authorization to stay for humanitarian reasons: a person is granted authorization under na- tional law to stay for humanitarian reasons if he or she is not eligible for international protection but is “nonetheless protected against removal under the obligations that are imposed on all Mem- ber States by international refugee or human rights instruments or on the basis of principles flowing from such instruments.” Source: Directive 2004/83/EC http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=CELEX:32004L0083:EN:HTML and Eurostat, http://ec.europa.eu/eurostat/statistics-explained/index.php/Glossary:Asylum_decision rizing a stay for humanitarian reasons, as happens to more equitably distribute asylum applications with certain unaccompanied minors (Box 2.2). Of across EU Member States.17 In 2016 EU open cas- cases pending at the end of 2017, 61 percent had es peaked at about 1.13 million, 7 percent higher been in process for six months or more, and 40 than in 2015 and 121 percent higher than in 2014 percent of those who did receive a response that (EASO 2016). When 2017 ended, about 460,000 year were granted international protection. Types cases were still open. The relocation program, of decision varied by nationality: for example, in which transferred asylum seekers to other EU the third quarter of 2017, 77 percent of Nigerian Member States for faster case review, applied to applications were rejected (14 percent were au- those from countries whose average EU recog- thorized to stay for humanitarian reasons), as were nition was higher than 75 percent—mostly Syr- 52 percent of Afghan and 7 percent of Syrian appli- cations (EUROSTAT 2018). 17 This is in the context of the Dublin Regulation that establishes Given the high volume of asylum applications, in the Member State responsible for the examination of the asy- 2015 the EU introduced a relocation mechanism lum application. ii. cOntext, data and MethOdOlOgy | 21 ians, Eritreans, and Iraqis—and operated from Senegal, Eritrea, Mali, Côte d’Ivoire, Guinea, So- September 2015 through September 2017.18 The malia, and Sudan—countries that accounted for original target was to relocate about 63,000 peo- 72 percent of total 2016 arrivals (UNHCR 201720). ple from Greece and 35,000 people from Italy, al- Though some of those interviewed had reached though it is argued that the EU-Turkey Agreement Italy as early as 2012, 75 percent had arrived in and the low recognition rate of those coming into 2016. Italy influenced the need for relocation.19 As of February 2018, almost 12,000 cases had been re- In Greece, data were collected from 1,680 adult re- located from Italy and about 22,000 from Greece spondents (1,101 men and 579 women) from one (UNHCR 2018a). of the top 3 countries of origin (UNHCR 201721): the Syrian Arab Republic, Afghanistan, and Iraq, Data Sources which accounted for 91 percent of total 2016 ar- rivals. Though those interviewed had arrived in This study collected three sources of data from asy- Greece at various times between 2015 and 2017, lum seekers who had arrived in Italy and Greece. 95 percent had come in 2016. Data were collect- See Appendices 1 and 2 for details on methodolo- ed in the Attica, Central Greece, and Thessaly re- gies and rationales and for preliminary descriptive gions, where most asylum seekers were housed statistics. in centers run by the government or UNHCR. (The sampling frame excluded asylum seekers hosted First, a quantitative asylum-seeker survey was con- in the Greek islands.) ducted between January and May 2017 in Italy and Greece (EASS 2017: Euro Asylum Seeker Survey). It The survey instruments took into account the collected information from a random sample of complexity of the samples. The survey was ad- adults in asylum centers on (1) basic demographic ministered in 6 languages (English, French, Ara- and socioeconomic characteristics; (2) the migra- bic, Bambara, Tigrinya, and Farsi). Some modules tion experience; and (3) respondents’ plans and (transit, aspirations) were designed to support aspirations. Box 2.3 describes the asylum process comparisons between asylum seekers; others as context for the data collection and the sample (assets, dwelling characteristics), were designed population. to support comparisons with the populations at home and borrowed questions from household In Italy, data were collected from 2,444 asylum surveys in every home country covered in the seekers (2,139 men and 305 women) who were at study. Finally, some modules (anxiety/depression, least 18 and were from one of the top 9 nation- education, self-reported literacy and numeracy) alities of arrivals entering: Nigeria, The Gambia, used standardized questions that allow for inter- national comparisons. The survey was adminis- tered using electronic tablets to simplify handling of all these conditions. 18 https://ec.europa.eu/home-affairs/sites/homeaffairs/files/ what-we-do/policies/european-agenda-migration/20170906_ relocation_and_resettlement-sharing_responsibility_and_in- creasing_legal_pathways_to_europe_en.pdf. 19 https://ec.europa.eu/home-affairs/sites/homeaffairs/files/ what-we-do/policies/european-agenda-migration/20170906_ relocation_and_resettlement-sharing_responsibility_and_in- 20 http://data2.unhcr.org/en/situations/mediterranean. creasing_legal_pathways_to_europe_en.pdf. 21 http://data2.unhcr.org/en/situations/mediterranean. 22 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Box 2.3 Legal Process, Waiting Times, and Life in a Center or Camp All asylum seekers interviewed reached Italy or Greece by sea. They had already applied for inter- national protection and were being hosted in formal reception centers or camps. However, the process and its duration differed somewhat for the two countries. Describing these processes is useful for understanding both the methodological aspects of the survey (e.g., sampling strategy) and how asylum seekers were living when interviewed. Italy: The Italian Coast Guard and Border Police typically were the first to reach migrants during rescue operations at sea.a Since early 2016, rescued individuals have been brought to a hotspot in Southern Italy; exceptions are rare. Hotspots, first entry points into the EU, are where pre-identifica- tion, registration, and photo and fingerprinting operations take place; arrivals cannot leave the facil- ity. When this study was conducted, stays in hotspots varied from 48 hours to weeks (the latter more likely if individuals refused to be fingerprinted). At this stage, individuals rescued declared their in- tention to apply for international protection, though they could not yet formally lodge a claim. When hotspot procedures were completed, individuals were transferred to temporary reception centers. Which center they went to depended on quotas given to each region based on its popula- tion and economy and on the availability of slots. For first reception, this might be a Governmen- tal Centre for Accommodation of Asylum Seekers (CARA) managed by the Ministry of Interior, or a Temporary Reception Center (CAS), which is private, and was originally temporary. to respond to emergency situations); those granted protection went to centers of second reception. CARAs typically host more individuals. Some centers (typically CASs) host only particularly vulnerable entrants, such as unaccompanied minors, vulnerable women, or families. As the number of mi- grants shot up, center boundaries blurred, but during the survey those just beginning the asylum process went to a CAS or a CARA. The time spent there was longer than the terms originally set: centers at first meant to respond to emergency situations have since had to deal with all stages of the immigration and asylum application process. Asylum applications are initially lodged at border control or local police stations (some areas have large reception centers), which transmit the documentation to the Commissione Territoriale per il Riconoscimento Protezione Internazionale, which is responsible for asylum examinations and inter- views for a certain province or group of provinces. The time between lodging an asylum claim and receiving a decision may range from a few months to two-and-a-half years. A longer time is usually because a denial decision has been appealed. Throughout the process, adult asylum seekers are provided with accommodation, food, a per diem, and medical care. Accommodations vary greatly, from converted containers in areas controlled by the military to small privately-owned buildings in the center of villages or cities. Most asylum seekers receive some training in Italian and occasion- ally some vocational training or are offered recreation options. Although all asylum seekers were free to go in and out, often center staff were their main point of contact with the outside world. Ties between asylum seekers seemed weaker in Italy than in Greece. ii. cOntext, data and MethOdOlOgy | 23 Greece: As sea arrivals on the Greek islands soared in 2015 and early 2016, a large number of in- stitutions responded. UNHCR, the International Organization for Migration (IOM), and the Greek authorities set up and managed first reception units to identify and register those who reached the islands, where temporary facilities were set up. People then made their way to the mainland. In winter 2015–16, the situation was fluid and uncertain as migrants gathered in large numbers on the Greek border with the Former Yugoslav Republic of Macedonia hoping to make their way to Central and Northern Europe. Those who remained were mostly accommodated in mainland reception centers managed by the Ministry of Interior and private accommodations managed by UNHCR. Allocation was mainly by nationality. In 2016 UNHCR embarked on a major effort to register asylum applicants. Migrants could claim asylum in Greece or lodge a claim for EU relocation or family reunification, as all the survey respondents had done; 70 percent of this group of asylum seekers had reached Greek territory in the first quarter of 2016—before the EU-Turkey agreement entered into force and all refugees on the Greek islands were transferred to the mainland. At the time of the survey the vast majority of those who arrived before January 2016 had moved on to other EU Member States, although a few might still be in Greece but not in asylum seeker facilities. Most asylum seekers were housed within tents or containers in reception centers, but some were in UNHCR-managed regular buildings. Facilities were very basic. Packaged meals were provided, but rudimentary cooking facilities were often available. Residents were usually clustered by coun- try of origin. Communities were re-created, and families often formed close ties, which happened rarely in Italy. There was a high number of children of all ages in the centers and they attended schools or informal classes in the centers organized by volunteers. During the day, asylum seekers were free to go in and out. a In some cases, NGO vessels or cargo ships may have been the first to respond to help refugees to reach the Italian coast or authorities. This study is a snapshot of the asylum seekers in dom sub-sample of those who responded to the Greece and Italy at a given time; it is by no means first phase, to complement self-reported data on designed to provide a comprehensive picture of education and skills. The study team collaborat- the entire European migrant crisis. For instance, it ed closely with the Organisation for Economic does not cover those in Greece and Italy who were Co-operation and Development (OECD) to get re- outside of centers, much less the earlier wave of sults comparable to the OECD Programme for the refugees and migrants mainly resident in other EU International Assessment of Adult Competencies Member States; migrants who did not make it to (PIAAC) test. In Italy the assessment was conduct- Europe; or unaccompanied minors. ed in English or French, and in Greece in Farsi or Arabic. This phase covered 578 asylum seekers, After the quantitative phase, the study carried out 202 in Italy and 376 in Greece. (See Chapter V for a computer-based literacy assessment of a ran- details.) 24 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Finally, in both countries qualitative surveys were asylum seekers is thus based on the 2016 recogni- carried out between January and March 2017 in tion rates for their home countries.23 Over the period both semi-structured focus groups and individual of interest, 58 percent of the study sample were from interviews, supplemented by interviews with cen- high-recognition and 33 percent from low-recogni- ter staff. The objective was to explore the context of tion countries (see Chapter III, Figure 3.1). decisions to leave home; the experience of the jour- ney and displacement; life in the reception system; Variations in recognition rates tend to reflect dif- and asylum seeker plans and aspirations. The gen- ferences in the human rights situations in asylum der dimensions of these areas were given special seeker countries of origin. They also correlate with attention. To avoid interview fatigue, the qualita- conflict intensity as measured by conflict-related ca- tive sample had no overlap with that for the quan- sualties per capita and with the World Bank classifi- titative survey. This phase spoke with 257 adult cation of fragile States: There is a positive relation- asylum seekers (149 men and 108 women) hosted ship between recognition rate, conflict intensity,24 in 16 centers in the Athens and Thessaloniki areas and the Human Rights Risk Score.25 Human rights of Greece and in Lazio and Lombardy in Italy. There in all high-recognition countries are deemed at “ex- were 118 in-depth individual interviews (IDIs) and treme risk.” All except Iran are considered fragile. 26 focus group discussions (FGDs), organized by gender and language or nationality. Respondents The second distinction this report makes is be- in Italy were from Nigeria, Eritrea, Somalia, Sene- tween asylum seekers who migrated to the EU gal, Mali, Guinea, The Gambia, and Côte d’Ivoire; directly from their country of origin and those and in Greece from the Syrian Arab Republic (some who settled in a different foreign country first. of them Kurdish), Afghanistan, and Iraq (some of Primary or first movers are those who were still in them Yazidi). Participants were selected to ensure their home country two years before the survey in- diversity in age, marital status, and (in Italy) length terview date. Secondary movers are those whose of stay in Europe. Women were oversampled in Ita- last permanent residence had been different from ly, where most asylum seekers were male. their country of origin for at least two years before the survey. In Italy 21 percent of asylum seekers Methodology were secondary movers (Figure 2.3), and 43 per- cent of those were Sub-Saharan Africans who had First, the study constructed two profile groups. Ev- been living in Libya. About 25 percent of asylum ery year, Eurostat publishes statistics on the recogni- seekers in Greece were secondary movers, mostly tion rate: the ratio of approved asylum applications Afghans who had been living in Iran (of whom 34 (for refugee, subsidiary, or humanitarian status) to percent had been born in Iran) and Syrians living the total number of decisions that year. In this study in Turkey (Figure 2.4). countries of origin are ranked by their 2016 recog- nition rates (Figure 2.2): those with rates above 40 percent are considered high-recognition-rate coun- group members are as similar as possible, while the groups tries, and those with lower rates are low-recogni- themselves are as dissimilar as possible. tion-rate countries.22 The first distinction between 23 http://ec.europa.eu/eurostat/cache/metadata/en/migr_ asyapp_esms.htm 24 The number of estimated casualties estimated by the Uppsa- la Conflict Data Program divided by total population. The choice of the 40 percent cutoff is determined by “opti- 22 25 Defined and measured by Verisk Maplecroft (Reliefweb mal clustering”, whereby groups are constructed to ensure that 2016). ii. cOntext, data and MethOdOlOgy | 25 Figure 2.2. Recognition Rates, 2016, Percent of Asylum Applications Approved 100 90 80 70 60 50 40 30 20 10 0 SYR ERI SOM IRQ AFG SDN IRN LBY ETH MLI GAB SLE GIN TCD COG CMR GMB JOR CIV LBR GHA BFA EGY SEN NGA GNB BGD NER SSD LBN MAR Source: Eurostat 2016. Country codes listed in Table A1.1 Figure 2.3. Primary and Secondary Asylum Seekers Figure 2.4. Top Nationalities of Secondary Movers, in Italy and Greece, Percent Percent of All Secondary Movers 30 100 90 25 80 70 20 60 50 15 40 30 10 20 10 5 0 Italy Greece Primary Secondary 0 SYR AFG ERI SDN NGA MLI GMB SEN GIN CIV SOM Source: Data from EASS 2017. Source: Data the EASS 2017. Country codes listed in Table A1.1. Statistics for Sudan are computed from 33 observations and should thus be interpreted with caution. 26 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making 27 III. Sociodemographic Profiles of Asylum Seekers All asylum seekers are not alike: To capture the di- briefly because this vulnerable group constitutes a versity of migration experiences while highlighting large part of total flows.27 what they have in common, this section presents sociodemographic profiles of asylum seekers. The chapter concludes by analyzing patterns of There are three distinct profile groups: unaccom- those migrating: Primary and secondary movers panied minors, those from high-recognition coun- are similar to each other but, particularly in Italy, tries, and those from low-recognition countries. their demographic profile (mainly single young By combining administrative data on unaccompa- men) is very different from the population at nied minors (a vulnerable group not studied here) home. There are also patterns of education, par- and survey data from the EASS 2017,26 this section ticularly among Syrians, that suggest differenc- quantifies the share of each group in the flows into es in the education profiles of different waves of Italy and Greece, using 2016 as the reference year. migrants. Finally, using complementary data for Afghanistan, Nigeria, and Senegal, it is estimated Asylum seekers in Italy differ considerably from those that those who migrated from these countries on of the sample in Greece. Among those least likely to average come from wealthier households than a be granted legal status (low-recognition nationals) typical one at home (which may relate to the cost are West Africans, who constitute about 75 percent of the trip). of those reaching Italy; they are primarily young sin- gle men. More likely to gain legal status (high-rec- Three Broad Groups of Asylum Seekers ognition nationals) are Syrians, Afghans, and Iraqis, who constitute nearly all those reaching Greece and In 2016 unaccompanied minors accounted for generally travel with their families, and East Africans 9 percent of the total flow into Italy and Greece: reaching Italy, who are often men traveling ahead 25,846 of the 171,309 who reached Italy and 5,192 of their spouses. The third group, unaccompanied of the 165,574 who reached Greece, according to minors, although not part of the study, is discussed UNHCR data. They are considered “the most ex- posed and vulnerable victims of migration,” and We estimate unaccompanied minors from UNHCR data, then 26 estimate the shares of the categories using the weighted pro- Unaccompanied minors were deliberately left out of the study 27 portions in our sample. due to ethical issues related to consent. 28 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Figure 3.1. Share of Each Group Entering Italy and Figure 3.2. Share of Each Group Entering Greece, Greece, Percent, 2016 Percent, 2016 Unaccompanied High minors recognition 9% 58% High recognition 97% Low recognition 33% Source: Data from EASS 2017, weighted sample, and UNHCR Source: EASS 2017– weighted sample and UNHCR administra- administrative data. tive data for unaccompanied minors. Figure 3.3. Share of Each Group Entering Italy, Italy, among other countries, has passed laws 2016, Percent forbidding their return to the border and has in- troduced “volunteer guardians.”28 This group is Unaccompanied High minors recognition discussed here because of their high share in the 15% 33% inflows into Italy and Greece, the unique policy challenges they represent, and their vulnerability; the hope is to give a more comprehensive picture of the total flow. Except for unaccompanied minors, other vulnera- ble cases are inevitably part of the survey samples but are not singled out, partly for want of a precise Low working definition (see Box 3.1) and partly because recognition a survey instrument eliciting self-reported informa- 52% tion has limited ability to screen for such cases. Source: EASS 2017– weighted sample and UNHCR administra- tive data for unaccompanied minors. The asylum-seeking populations in Italy and Greece differ dramatically. In Greece (Figure 3.2), 97 percent stan—and the other 3 percent are unaccompanied of asylum seekers are from high-recognition coun- minors. The picture in Italy is more nuanced. While tries—the Syrian Arab Republic, Iraq, and Afghani- more than 52 percent of those arriving are from low-recognition countries, a substantial 33 percent are nationals of high-recognition Eritrea, Somalia, 28 http://eumigrationlawblog.eu/the-new-italian-law-on-unac- or Sudan (Figure 3.3). The rest, 15 percent, are un- companied-minors-a-model-for-the-eu/; http://eur-lex.europa.eu/legal-content/EN/ALL/?uri=CELEX- accompanied minors, a category identified from :52010DC0213. administrative data. iii. SOciOdeMOgraphic prOFileS OF aSyluM SeekerS | 29 Box 3.1. Defining Vulnerable Cases Policy makers try to respond to the needs of asylum seekers and refugees who are also vulnera- ble on dimensions other than the specific ones that inspired them to migrate. Policymakers and others have adopted policies to recognize added risks that might require special attention and procedures for asylum applications and international protection processes. However, the notion of a “vulnerable person” is ambiguous. Because there is as yet no international- ly-recognized definition, there have been calls to formulate “a working definition of migrants in vul- nerable situations” (UN 2017). The Asylum Procedures Directive, adopted in 2013 by the European Parliament and the Council, recognizes the need of special procedures for applicants due to “age, gender, sexual orientation, gender identity, disability, serious illness, mental disorders or as a conse- quence of torture, rape or other serious forms of psychological, physical or sexual violence”; it calls on Member States to strive to identify these types of applicants. It also calls for ensuring that interview- ers in the asylum process are competent to account for vulnerability as a special circumstance,a and it references the definition of vulnerability in Directive 2013/33 (on standards for reception of appli- cants) as the one to be used to prioritize applications for international protection. Directive 2013/33 describes a nonexhaustive list of vulnerable persons, “such as minors, unaccompanied minors, dis- abled people, elderly people, pregnant women, single parents with minor children, victims of human trafficking, persons with serious illnesses, persons with mental disorders and persons who have been subjected to torture, rape or other serious forms of psychological, physical or sexual violence, such as victims of female genital mutilation.”b Beyond EU legislation, other definitions of vulnerability are also quite broad. UNHCR distinguish- es between “situation vulnerability”—risks during the journey and in the host country—and “indi- vidual vulnerability”: individual characteristics or circumstances which place a person at particular risk such as that experienced by: children, particularly those who are unaccompanied or have been separated from family; older people; those with mobility, sensory, intellectual, or other disabilities; those with chronic illnesses or other medical needs; victims or survivors of trafficking who do not fall within the scope of the refugee definition; or survivors of torture or trauma en route” (UNHCR 2017). The scope and the variability of definitions of vulnerability and the difficulty of identifying some categories considered vulnerable suggest that detecting and acting on these cases is far from simple. Nevertheless, EU Member States have been striving to put measures in place to identify and offer protection to vulnerable persons (EASO 2017 Section 4.11.). a DIRECTIVE 2013/32/EU OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL of 26 June 2013 on common proce- dures for granting and withdrawing international protection (recast). http://eur-lex.europa.eu/legal-content/EN/TXT/PD- F/?uri=CELEX:32013L0032&from=en. b Article 21 of the DIRECTIVE 2013/33/EU OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL, of 26 June 2013 laying down standards for the reception of applicants for international protection (recast), .https://eur-lex.europa.eu/legal-con- tent/EN/TXT/ ?uri=celex%3A32013L0033 30 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Nationals of High-Recognition Countries “I was happy, but after the war This section discusses the profile of asylum seek- all problems started. No job and ers from high-recognition countries and how they destruction everywhere. Before the compare to the people back home. Figure 3.4 shows the nationalities of origin. Asylum seekers war…Syria was a paradise. We from the Syrian Arab Republic and Iraq, countries had everything…” suffering from civil war and sectarian tensions, Syrian man represent 33 and 9 percent of the flow; Afghans 22 percent, Eritreans 19 percent, and other East Afri- cans, mostly from Somalia and Sudan, 17 percent. Most Syrian refugees worldwide are not in the EU but in neighboring countries. At the time when Figure 3.4. Flows from High-recognition Countries most asylum seekers interviewed in our sample by Nationality, Percent, 2016 were arriving to Greece, according to UNHCR (2016), IRQ 9% Turkey hosted 2.7 million Syrian refugees (50 per- cent of the total), Lebanon 1 million (18 percent) Other East SYR 33% Africa and Jordan more than 660,000 (11 percent).29 This 17% underestimates the total number of Syrians forcibly displaced because not all are registered refugees. However, between 2011 and 2016, 889,675 Syrians requested asylum in the EU (Eurostat 2016). Germa- ny, with more than 484,920 cumulated applications, and Sweden, with 110,380, lead European coun- tries. The rest of the applications are distributed ERI 19% AFG 22% among the other European countries, with 31,780 having been processed in Greece. Source: Data from EASS 2017 – weighted sample. Twenty-seven percent of the Syrians seeking asylum in the EU were settled in Turkey or Iraq before deciding to Asylum Seekers from the Syrian Arab Republic travel to Greece. Reasons for undertaking the second leg of the journey are many. It appears that Syrians in Syrians fleeing into the EU are a recent phenome- Turkey still find it difficult to access the labor market, non associated with the Syrian Civil War. In 2011, and in 2016, of an estimated 900,000 of school age, what started as peaceful protests became a full- 400,000 were still not in school (Crisis Group 2016). The fledged civil war, involving the army, opposition study’s qualitative interviews with Syrian asylum seek- forces, militias with diverse allegiances, and foreign ers confirm that difficulties in making a living were an States. This conflict has led to the largest outflow incentive to leave Turkey, as was the hope of securing of refugees in recent history, 5.5 million, in addition better education opportunities for children. to some 6.1 million internally displaced. By 2016 there were an estimated 470,000 casualties (Hu- man Rights Watch 2016). Figure 3.5 illustrates how 29 More recent numbers show that Turkey hosts 3.59 million Syr- Syrian migrant outflows are also large in terms of ian refugees, Lebanon around 987,000 and Jordan more than worldwide forced displacement. 666,000. http://data2.unhcr.org/en/situations/syria iii. SOciOdeMOgraphic prOFileS OF aSyluM SeekerS | 31 Most Syrian asylum seekers in Greece migrated “I wanted to give my children a with family members: 66 percent of asylum seekers better education and decided to from the Syrian Arab Republic are in Greece with family members, and 14 percent of those married go to Germany… Many people already have a spouse elsewhere in Europe (Figure where I was in the camp in Turkey 3.6). Their average age is 33, and 35 percent have had already left for Germany.” a high school or university education. Younger un- married Syrian asylum seekers are more educated Syrian woman (39 percent have a high-school education or more) than older ones arriving with their families (28 per- cent with at least a high school education). Figure 3.5. Refugees and Asylum Seekers World- wide and from the Syrian Arab Republic, 1951–2016, Millions Asylum Seekers from Afghanistan Historic Data: Worldwide and Syrians Refugees began to leave Afghanistan starting in 20 the late 1970s during the Soviet-Afghan war. They 15 peaked in 1990 at 6 million (Figure 3.7), but after the war the numbers decreased substantially as re- 10 turnees made their way home. However, more epi- Syrian Civil War sodes of forced displacement accompanied waves 5 of violence during the Taliban war and the United States invasion. Political instability in recent years 0 1968 1978 1988 1998 2008 has also contributed to outflows of refugees. Worldwide Refugees Syrians Source: UNHCR 2018. “I weighed it all up and thought to myself, when a country has been Figure 3.6. Share of Syrians Reaching Centers with at war for 38 years there is no hope Families, by Location, 2016 for the future.” 80 Afghan man 70 60 In 2016, UNHCR reported that Afghan asylum 50 seekers and refugees were mostly in Pakistan 40 (1,356,607), Iran (951,149), Europe (406,000), and 30 Turkey (118,116).30 The mean length of exile for 20 Afghan refugees is estimated at 24 years (Devic- 10 tor and Do 2018). Although Europe is now hosting 0 In Center In Europe Elsewhere more than 400,000, very few have been resettled Spouse location Children location Source: Data from the EASS 2017. 30 http://popstats.unhcr.org/en/time_series. 32 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making elsewhere through the UNHCR resettlement pro- cess since the flows of Afghan refugees began. “None of us had any papers [in From the 1980s to 2016, 8,100 Afghan refugees Iran] and we could have been were resettled in Europe, compared to 15,000 Iraq- is and 20,000 Syrians resettled. deported at any time.” Afghan woman who left Iran Figure 3.7. Afghan Refugees and Asylum Seekers, 1979–2016 Of Afghan asylum seekers in the EU, 34 percent Historic Data: Afghans are secondary movers, mostly from Iran. The first 8 wave of Afghan refugees that reached Iran dates Afghan Soviet US Soviet War Exit Invasion to the 1980s and the Soviet-Afghan war. Figure 3.8 6 captures how that population peaked in the mid- 1990s when the Taliban started ruling most of the Refugees (millions) country. The number plunged after the September 4 11 attacks and the subsequent overthrow of the Taliban by a U.S.-led coalition. In the qualitative interviews, Afghans who had lived for a long time 2 in Iran as refugees or undocumented migrants ex- plained that they had come to Europe because 0 they had few rights to asset ownership and edu- 1968 1978 1988 1998 2008 cation and few employment opportunities; those who had lived in Iran as undocumented migrants Source: UNHCR 2018. feared deportation. Figure 3.8. Iran GDP, Oil Prices, and Share of Afghan Figure 3.9. Spouses and Children of Afghan Asylum Refugees in Iran, 1980–16 Seekers by Location, Percent 1 90 80 8 70 Normalized values 60 6 Share (%) 50 40 4 30 2 20 10 1980 1990 2000 2010 0 Share Afghans in Iran Oil Price In Center In Europe Elsewhere GDP per capita in Iran Spouse Location Children Location Afghan refugees in Iran as share of Afghan refugees worldwide Source: Data from EASS 2017. Source: : UNHCR 2018, World Bank World Development Notes: The figure refers to spouses and children of household Indicators, and US Energy Information Administration. heads surveyed. iii. SOciOdeMOgraphic prOFileS OF aSyluM SeekerS | 33 Since the mid-2000s, the number of Afghan refu- Figure 3.10. Eritrean Refugees and Asylum Seekers, gees worldwide has held steady at about 2.5 mil- 1980–2016 lion people (see Figure 3.7), but the number in Iran Historic Data: Eritreans has been trending down. In Figure 3.8, the green 10 The Facto Independence of Eritrea line plots the price of oil and the yellow line Iran’s GDP. Economic circumstances in Iran do not seem 8 to explain the departure of the Afghan refugees Refugees (millions) who had been living there. Correlations of the 6 share of Afghan refugees in Iran to the price of oil is (minus) -0.74 and to Iran’s GDP is -0.85. 4 Afghans reaching Greece are mostly families trav- 2 elling together. Some families (10 percent) already have a spouse or child in Europe (Figure 3.9), but 0 in 85 percent of the cases children are in Greece 1980 1990 2000 2010 with their parents. About 25 percent of the Afghan group have secondary education or higher; and Source: UNHCR 2018. the average age is 33. Figure 3.11. Location of Eritreans Spouses and Asylum Seekers from Eritrea Children, Percent In Eritrea, military conscription is mandatory for 100 both men and women until age 40, and since 1998 90 its duration has been open-ended. After 30 years 80 of fighting annexation by Ethiopia, Eritrea became 70 independent in 1993. In 1995 its president, Isaias 60 Afewerki, introduced compulsory military service; 50 Kibreab (2013), who surveyed 190 Eritreans living 40 abroad, found that the average duration of mil- 30 itary service was 5.8 years. A UN commission in- 20 terviewed 550 Eritreans residing in third countries 10 and solicited written submissions; it concluded 0 In Center In Europe Elsewhere that some of the human rights violations in the country “may constitute crimes against humanity” Spouse Location Children Location (UN 2015). Source: Data from the EASS 2017. Of the complex factors underlying a migration de- cision, the reason most Eritreans often cite is mili- who reached Italy. Other factors often mentioned tary service. Among those in Italy, some 57 percent were such human right violations as extrajudicial reported fleeing conscription and 13 percent left killings, torture, rape, and forced labor (UN 2015). because of security risks. At 35 percent, military service is also the primary reason for driving the Eritreans reaching Italy are different from citizens emigration decision reported by Eritrean women of other high-recognition countries. Adults are 34 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Figure 3.12. Location of Families of Asylum Seekers from East African Countries Other than Eritrea, “The soldiers started to come Percent to my house in 2007. The third time they came I decided to 100 leave because I understood they 80 wouldn’t leave me in peace. In 60 Eritrea, I could only be a soldier 40 for all my life, without a salary and the possibility to see my 20 family.” 0 Eritrean woman In Center In Europe Elsewhere Spouse location Children location younger, with an average age of 26. Most are un- Source: Data from the EASS 2017. married, and those who are married have left their Asylum Seekers from Iraq families behind (Figure 3.11). Some Eritreans have tried to settle elsewhere, 24 percent in Libya or Although first waves of Iraqi refugee flows date back Sudan, before moving to Europe. Among Eritrean to the Iraq-Iran war in the 1980s, refugee flows in- asylum seekers, 33 percent are women (married or creased significantly after the Second Gulf War in single) and 15 percent of them are single—a rela- 1991 due to economic sanctions and the regime’s tively high proportion of single women compared reaction to uprisings both in the south and north to Syrians (6 percent), Afghans (8 percent), and of Iraq. Since the mid-2000s and after the invasion Iraqis (7 percent). of Iraq by U.S.-led forces, as sectarian tensions and security instability escalated, more refugees made Asylum Seekers from East African Countries their way to neighboring countries (Figure 3.13). In other than Eritrea 2007, UNHCR estimated that 450,000–500,000 Iraq- is were residents of Jordan. In 2014, the self-pro- The East Africans surveyed were mostly from So- claimed Islamic State in Iraq and the Levant (ISIL, malia and a few from Sudan. Natural hazards and also known as ISIS, IS, or Daesh) took control of Mo- 25 years of armed conflict have driven out a large sul and large territories in the West, and in the direct number of Somalis. According to UNHCR, over aftermath more than 3 million were internally dis- 870,000 Somalis are registered as refugees else- placed while others decided to leave the country.31 where in the Horn of Africa and in Yemen. In 2016, 126,756 Iraqis were registered in Turkey as either asylum seekers or refugees (UNHCR 2016), The other East African asylum seekers interviewed and between 2008 and 2016, 344,885 applied for were on average aged about 26. They were more asylum in the EU (Eurostat 2016). likely to be married than those from Eritrea, but they travelled alone, leaving families back home (Figure 3.12), and 23 percent had settled some- 31 As of date, IOM, the UN Migration Agency, identified more where else, usually Libya, before continuing to It- 3.7 million people who returned to their place of origin in Iraq. aly. About 86 percent were men. http://iraqdtm.iom.int/ iii. SOciOdeMOgraphic prOFileS OF aSyluM SeekerS | 35 The Iraqis arriving to Greece mostly came as fam- “We started thinking of leaving ilies: 70 percent of those surveyed were traveling when my brother-in-law was with family members (Figure 3.14). Of the rest, spouses were either already in Europe (17 percent) killed by a car bomb that was or elsewhere, such as Iraq or Turkey. The average placed on his car in 2014” age of Iraqis seeking asylum in Greece was 33, and Iraqi woman 18 percent had a high school or university educa- tion. Most Iraqis reaching Greece were first movers: only 4 percent had settled in another country be- fore reaching Europe. Figure 3.13. Iraqi Refugees and Asylum Seekers, 1980–2016 Nationals of Low-Recognition Countries Historic Data: Iraqis 4 The large majority of asylum seekers from low-rec- Kurdish Repression ognition countries started in Sub-Saharan Africa 3 and entered Italy via Libya (Figure 3.15). They are al- Refugees (millions) Secterian violence in Iraq most exclusively from West African countries, main- 2 ly Nigeria (34 percent) followed by The Gambia (16 percent), Mali (15 percent), Senegal (13), and Guin- 1 ea, and Côte d’Ivoire (about 11 percent each). 0 These asylum seekers tend to be unmarried men, 1970 1980 1990 2000 2010 aged on average 24. More than 27 percent have secondary education or more, and about 18 per- Source: UNHCR 2018. cent were secondary movers, 10 percent from Figure 3.14. Location of Iraqi Spouses and Children, Figure 3.15. Low-recognition Group, by Nationality, Percent Italy, 2016, Percent 90 GIN 11% 80 NGA 34% 70 CIV 11% 60 Share (%) 50 40 30 20 10 SEN 13% 0 In Center In Europe Elsewhere GMB 16% Spouse location Children location MLI 15% Source: Data from the EASS 2017. Source: Data from the EASS 2017. 36 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Libya. Almost all in this group traveled alone, with with the most trafficking victims entering the EU those married leaving their families behind. (European Commission 2017). IOM (2017) esti- mates that 80 percent of Nigerian women landing Of asylum seekers reaching Italy from low-rec- in Italy may be or become sexually exploited, and ognition countries in Sub-Saharan Africa, 90–97 the number has been rising in recent years; extrap- percent are young, 18 to 35 but mostly under 26. olating from the administrative data suggests that Women constitute 24 percent of this group; 21 up to 4,880 asylum-seeking Nigerian women are at percent of all women were Nigerians and 75 per- risk—undoubtedly an underestimate because few cent were single. Young single women make up of them are likely to apply formally for asylum. 13 percent of Nigerian asylum seekers. Qualitative surveys suggest that many may have been traf- Finally, 8 percent of asylum seekers from low-rec- ficked—reportedly Nigeria is the non-EU country ognition countries had lived in Libya for at least Figure 3.16. Asylum Seeker Sociodemographic Characteristics, Italy, Percent 100 90 80 70 60 50 40 30 20 10 0 NGA GMB MLI SEN GIN CIV NGA GMB MLI SEN GIN CIV NGA GMB MLI SEN GIN CIV NGA GMB MLI SEN GIN CIV Age: 18-35 Secondary Education Single Men Source: Data from the EASS 2017. Country codes listed in Appendix 1. table 3.1. comparison of Sociodemographic characteristics, asylum Seekers from high- and low-rec- ognition countries, 2016, percent Secondary Secondary Category Country Male Single Average Age Migration Educated Syrians 27 64 29 35 33 Afghans 34 70 32 25 33 High Iraqis 4 61 27 19 33 Recognition Eritreans 24 67 57 39 26 Other East 23 86 44 40 26 Africans Low West Africans 18 90 81 27 24 Recognition Source: Data from EASS 2017. iii. SOciOdeMOgraphic prOFileS OF aSyluM SeekerS | 37 Box 3.3. Migration and the Libyan Conflict In 2011, Libyan rebels brutally executed Muammar Gaddafi, ending his 41-year rule. By 2010, some 2.5 million African migrants were settled in the country and were at risk during the crisis. Rumors of African mercenaries being hired by Gaddafi to counter the uprising, led to a wave of anti-migrants sentiment in the country (IOM 2012). Since 2011, pervasive violence has hampered the democratic transition with an estimated 1,700 armed militia groups fighting for power. By 2015 the resulting violence had led to internal displacement of 434,869 Libyans (CFR 2016). According to the Uppsala Conflict Data Program, the violence, measured by the number of casu- alties, peaked in 2016 with 1,600 deaths (Figure B3.3.1). The increase in violence was mirrored by displacement flows: In 2016, 82 percent of the internally displaced persons (IDPs) identified were displaced in 2014–16, compared to 13 percent displaced in 2011, and 5 percent in 2012–14. The number of individuals transiting through Libya to reach Europe also went up, peaking in 2016. Figure B3.3.1. Estimated Deaths in Libya and Flow of Asylum Seekers to Italy, 2014–17 8 12.1 Migrant flow into Italy (log scale) casualties in Libya (log scale) Estimated conflict-related 12 7 11.9 11.8 6 11.7 5 11.6 2014 2015 2016 2017 Casualties in Libya Flow to Italy Source: Allansson et al. (2017), and UNHCR 2018. two years before entering Italy and were thus sec- and are men (90 percent) and single (69 percent). ondary movers. The conflict in Libya (Box 3.3) not About 27 percent have secondary education; 35 only triggered migration for them, but also made it percent report having been exposed to violence; a transit rather than a destination country for oth- and 42 percent suffer from severe mental distress. er waves of migrants. Notably, the demographic profiles of these primary and secondary movers Thus, the sociodemographic profile of asylum are not very different. These are mostly from low- seekers differs by group. As summarized in Ta- recognition West African countries (82 percent) ble 3.1, asylum seekers in Greece, mostly from 38 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making high-recognition countries, are slightly older, trav- Except for Eritreans, Somalis, and Nigerians, asylum el in families, and include both men and women. seekers in Italy generally described less direct con- Those in Italy, mostly from low-recognition coun- nection to diasporas in Europe than those in Greece. tries, are primarily single young men. A fraction of However, use of social media sometimes narrowed nationals from both groups had previously settled this gap. Moussa from Côte d’Ivoire relayed how elsewhere. friends made on Facebook heavily influenced his decision to leave: “There are persons from Côte The Decision to Migrate d’Ivoire on Facebook who told me Italy is a good place. Once you are in Italy, you will have no prob- With more than 60 percent of asylum seekers ar- lem, and you will find work.” Moussa was not the riving in Greece in families, it is not surprising that only person to discuss the move with friends made asylum seekers interviewed there report consider- online; it is not always clear whether such contacts able family consultation before migrating. Family were sought out or generated by social networks. discussions, as registered in the qualitative survey, covered whether to leave and whether to do so Primary Movers, Secondary Movers, and together or sequentially. In most cases they also Stayers reached beyond the spouse or the head of house- hold to seek counsel or rally support for migration Who actually undertakes the journey and reaches in the form of emotional support, information on Europe? What types of peoples are more inclined options, or funding. Both men and women de- to migrate? To find answers, the research first com- scribed leveraging extended families and social pared primary with secondary movers among asy- networks for material support, taking out loans, lum seekers. It then compared asylum seekers with or selling land within their families or social net- the populations in their countries of origin if there works. One Afghan woman, travelling with her hus- was a relevant comparator household survey and band, said, “I sold everything, even the doormat. there were enough observations in the EASS 2017 I also borrowed €15,000 from my relations.” In Af- as for Afghanistan (accounting for 24 percent of asy- ghanistan funding within social networks has been lum seekers in Greece in 2016), Nigeria (20 percent found to intensify social obligations to repay loans of those in Italy), and Senegal (6 percent of those in on return, to the point of discouraging return or Italy). Section 3 in the Appendix discusses the meth- encouraging remigration on deportation (Schuster odological steps to allow such comparisons. Also and Majidi 2013). compared were Syrian asylum seekers in Greece and Syrian refugees in Jordan and Lebanon; this For those in Italy, the decision to migrate seems to sample was also compared with refugees in other have relied less on family and more on conversa- EU countries based on available studies. tions with peers and friends. Focus group discus- sions (FGDs) suggest that beyond concerns about The profile of first and secondary movers from security, migration may have been motivated by low-recognition countries is distinctly different desires to realize hopes for the future. Discussing from that of populations at home. West African the move with friends and social media under- asylum seekers, both primary and secondary scores the prevalence of the information-sharing movers, are mainly young, male, and single (Fig- and aspiration formation among the young that ure 3.17). The main difference is that secondary Hernandez-Carretero and Carling (2012) have movers in Italy who had been settled in Libya were linked to migration by Sub-Saharan youth. more likely than first movers to be married. iii. SOciOdeMOgraphic prOFileS OF aSyluM SeekerS | 39 Asylum seekers from Senegal and Nigeria, whether mainly men, as are nearly all the Senegalese and a first or secondary movers, are mostly young and high share of the Nigerians, and in both cases the male; 91 percent of the Senegalese and 96 percent majority are single. However, 22 percent of Nigeri- of the Nigerians are aged 18–35, while those in this an asylum seekers are women, about 66 percent of young group constitute only about 50 percent of them single, and they are very young (18–26). the population back home aged 18–65; in fact, at 18–26, 64 percent of Nigerian and 64 percent For high-recognition Afghanistan and the Syri- of Senegalese asylum seekers are even younger. an Arab Republic, differences between first and Unlike populations at home, asylum seekers are second movers are small (Figure 3.18). However, Figure 3.17. Asylum Seekers and Population in Country of Origin Compared, Percent Nigerian population (2011) Never married (%) Nigerian Asylum Seekers Senegal population (2011) Senegal Asylum Seekers West African: Secondary Movers West African : First Movers Nigerian population (2011) Nigerian Asylum Seekers Male (%) Senegal population (2011) Senegal Asylum Seekers West African: Secondary Movers West African : First Movers Nigerian population (2011) Nigerian Asylum Seekers Age 18-35 (%) Senegal population (2011) Senegal Asylum Seekers West African: Secondary Movers West African : First Movers 0 10 20 30 40 50 60 70 80 90 100 Source: EASS 2017, Nigeria 2011 Living Standard Measurement Survey; Senegal 2011 ESPS II. Figure 3.18. Syrian Asylum Seekers and Population in other countries Compared, Percent Syrians: Lebanon (2015/16) married (%) Syrians: Jordan (2015/16) Never Syrians: Secondary Movers Syrians: First Movers Syrians: Lebanon (2015/16) Male (%) Syrians: Jordan (2015/16) Syrians: Secondary Movers Syrians: First Movers Age 18-35 (%) Syrians: Lebanon (2015/16) Syrians: Jordan (2015/16) Syrians: Secondary Movers Syrians: First Movers 0 10 20 30 40 50 60 70 80 90 100 Source: EASS 2017 and 2015-16 Syrian Refugees and Host Communities Data for Jordan, KRI, and Lebanon. 40 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Figure 3.19. Afghan Asylum Seekers in Greece and Population in Afghanistan Compared, Percent married (%) Afghan population (2013-14) Never Afghans: Secondary Movers Afghans: First Movers Afghan population (2013-14) Male (%) Afghans: Secondary Movers Afghans: First Movers Age 18-35 (%) Afghan population (2013-14) Afghans: Secondary Movers Afghans: First Movers 0 10 20 30 40 50 60 70 80 90 100 Source: EASS 2017 and 2013-14 Afghanistan Living Conditions Survey. when Syrian asylum seekers are compared with portion of Senegalese asylum seekers, by contrast, Syrian refugees in Lebanon and Jordan, there is similar to Senegal’s urbanization rate. are more men among the asylum seekers. Af- ghan asylum seekers are slightly more likely than Nearly all Nigerian asylum seekers had been living the population back home to be male, young, in the Southern part of the country before migrat- and single. Nevertheless, compared to asylum ing, and 84 percent of those for whom information seekers from low-recognition countries, Afghan was available32 had been in the South-South re- and Syrian asylum seekers traveled mostly in gion, where 17 percent of Nigerian are to be found; families, and differences between the groups are a high concentration of asylum seekers are from smaller. Benin City. It may be that social networks there fa- cilitate migration, or it could relate to smuggling or First mover and secondary mover asylum seekers human trafficking networks. are quite similar, but both differ from the popula- tions at home, particularly those from low-recog- About 60 percent of Afghan asylum seekers who nition countries. This suggests that those with cer- are primary movers were from the East region of tain demographic characteristics are more likely to the country—the most highly populated region leave home, and the similarities between first and (47 percent of total) and containing the capital city. secondary movers suggests that once someone Moreover, while 69 percent of resident Afghans live leaves, other issues—related to the timing of that in rural areas, only 19 percent of asylum seekers did. decision or factors explained below—might be rel- evant to the migration path. Education Region of Origin Education levels are similar for first and second- ary movers from West African countries. Less than Except for Nigerians, within their countries asylum seekers were not highly concentrated in a region. 32 Information on region of origin has missing values. This state- About 70 percent of Nigerian asylum seekers lived ment assumes that the region distribution reflects the full sam- in urban areas, which contain only about 40 per- ple of asylum seekers and is thus not correlated with patterns cent of the Nigerian population. The urban pro- that might have been found in missing information, iii. SOciOdeMOgraphic prOFileS OF aSyluM SeekerS | 41 Figure 3.20. West African Asylum Seekers by Educational Attainment, Percent 100 90 19 14 19 14 29 23 80 42 43 70 49 58 60 50 40 30 20 10 0 West African: First Movers West African: Secondary Movers Senegal Asylum Seekers Senegal population (2011) Senegal Asylum Seekers - Youth Senegal population (2011) - Youth Nigerian Asylum Seekers Nigerian population (2011) Nigerian Asylum Seekers - Youth Nigerian population (2011) - Youth Less than secondary education (%) Secondary education or more (%) Source: EASS 2017, Nigeria 2011 Living Standard Measurement Survey; Senegal 2011 (ESPS II). Notes: Youth = those aged 18–35 years. 30 percent of asylum seekers from West African Syrians seeking asylum in the EU are more educated low-recognition countries had secondary educa- than Syrian refugees in other countries. Of Syrians tion or more, with those who moved from tempo- forcibly displaced, the least educated are in neigh- rary settlement in Libya being slightly less educated boring countries like Jordan and Lebanon and the (Figure 3.20). For Senegal the profile is not very dif- more educated have moved on to European coun- ferent, but the average Nigerian asylum seeker aged tries beyond Greece, likely in an earlier wave (Figure 18–35 is less likely to have secondary education (43 3.21). Moreover, among asylum seekers in Greece, percent) than peers at home (58 percent). This may those who had first settled in Turkey are slightly less partly be because there are more women Nigerian educated than first movers. The education of both asylum seekers, and 68 percent of them are young, first and secondary asylum seekers in Greece is on with less than secondary education. The Nigerian average more comparable to the average Syrian in asylum seeker educational gap (Figure 3.20) widens 2011. Comparing results gives a dynamic picture by 17 percentage points when the home compari- of choices to settle in Jordan, Lebanon, Turkey, son is with only the South-South region where most Greece, or Austria and Germany. asylum seekers are from.33 Afghan asylum seekers who first settled in Iran have less education than peers at home, first movers, and asylum seekers in other European countries. 33 This region of Nigeria has the highest educational attainment: The education profile of the total Afghan asy- 19 percent of the labor force has tertiary education compared lum-seeking population is relatively similar to that to less than 33 percent in some Northern regions and 70 per- of those who moved from Iran. For instance, 75 per- cent has some junior secondary education or higher. Returns to secondary education and above are also higher in the South cent of asylum seekers aged 18–65 had less than than in other regions of the country (World Bank 2016). secondary education, compared to 80 percent back 42 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Figure 3.21. Education of Syrian Asylum Seekers in Greece and Italy, Percent 100 90 80 70 60 50 40 30 20 10 0 Syrians: Jordan Syrians: Syrians: Greece, Syrians: Greece, Syrians: Austria Syrians: (2015/16) Lebanon Secondary First Movers (2016) Germany (2016) (2015/16) Movers Less than secondary education (%) Secondary education or more (%) Source: EASS 2017; for Jordan, KRI, and Lebanon: 2015-16 SRHCS; for Austria, Buber-Ennser et al. 2016 (limited to population 20-59 years); and for Germany, the 2016 IAB-BAMF-SOEP Refugee Survey. home. However, this masks other patterns: (1) Sec- Living Standards ondary movers from Iran were much less educated than primary movers: only 14 percent had second- Welfare simulations for Afghanistan, Nigeria, and ary education or above, compared to 31 percent Senegal suggest that the average asylum seeker of first movers (Figure 3.22). They were also less comes from a wealthier household than the av- educated than those at home (20 percent). (2) First erage household back home. These findings are movers who left the East region, where a large share based on techniques that draw on known welfare of asylum seekers come from, are better-educated aggregates for countries of origin to impute welfare than the average: 41 percent of working-age asylum for these countries (see Appendix 3). It is estimated seekers had secondary education or more against that the average Nigerian asylum seeker, for exam- 21 percent of those still in the region. ple, comes from a household with daily consump- tion that is about 35 percent higher than the aver- In sum, there are no systematic education patterns age in Nigeria, and even 24 percent higher than in associated with the decision to emigrate. Among the better-off South-South region (Figure 3.23). those from high-recognition countries, first mov- ers to Greece are relatively better-educated than Moreover, asylum seekers seem to be among the those in other countries or secondary movers but richest in their countries of origin: an estimated less-educated than the first wave of asylum seek- 58 percent of the Nigerians come from the top 40 ers (see Figure 3.21). Among those from low-recog- percent in the welfare distribution and 34 percent nition countries, differences between groups are from the top 20 percent (Figure 3.24). Similarly, an less stark but it appears that the young Nigerians estimated 50 percent of Senegalese asylum seek- who reached Italy have less education than the ers and 63 percent of Afghans come from the top average young person in Nigeria. (Chapter V dis- 40 percent of households in the country; 29 per- cusses the education and skills of asylum seekers cent of Senegalese and 43 percent of Afghans are in more detail.) from the top 20 percent (Figure 3.24). iii. SOciOdeMOgraphic prOFileS OF aSyluM SeekerS | 43 Figure 3.22. Education of Afghan Secondary Movers from Iran, Percent 100 90 14 20 21 26 31 80 41 70 60 50 40 30 20 10 0 Afghans: Afghan Afghans: Afghans: First Afghan Asylum Afghan Second population Austria (2016)* Movers Seekers from population Movers (2013-14) East region from East region (2013-14) Less than secondary education (%) Secondary education or more (%) Source: EASS 2017 and 2013-14 Afghanistan Living Conditions Survey; for Austria: Buber-Ennser et al. 2016, limited to ages 25+ years. The fact that the average asylum seeker comes from (1.1 years in Senegal, 1.6 years in Nigeria). The total a better-off household suggests that for some still at cost is even higher for Afghans, who travel in fam- home affordability might be a barrier to migration. ilies. In sum, not all may be able to afford the trip At about US$1,600 for those taking the Agadez/Trip- or to forego, even temporarily, the contribution of a oli route, the average financial cost of the journey household member to the family income. for Nigerians and Senegalese is equivalent to over a year of consumption for an average person at home Figure 3.24. Asylum Seekers in Top and Bottom Welfare Quintiles at Home, Percent Figure 3.23. Average Daily per Capita Consumption, Afghan population (2013-15) 20 Share in the bottom welfare quintile Share in the top welfare quintile Nigeria, Senegal, and Asylum Seekers, US$ at from country of origin (%) Afghan Asylum Seekers* 43 Purchasing Power Parity (PPP) Senegal population (2011) 20 5.00 Senegal Asylum Seekers* 29 4.50 Nigerian population (2011) 20 4.00 Nigerian Asylum Seekers* 34 3.50 3.00 2.50 Afghan population (2013-15) 20 2.00 Afghan Asylum Seekers* 15 from country of origin (%) 1.50 1.00 Senegal population (2011) 20 Senegal Asylum Seekers* 14 0.50 0.00 Nigerian population (2011) 20 Nigerian Nigerian Senegal Senegal Nigerian Asylum Seekers* 10 Asylum population Asylum population Seekers* (2011) Seekers* (2011) 0 10 20 30 40 50 Average daily per capita consumption (US$ PPP) Source: EASS 2017: Nigeria 2011 Living Standard Measure- Source: EASS 2017: Nigeria 2011 Living Standard Measurement ment Survey; Senegal 2011 (ESPS II). Survey; Senegal 2011 (ESPS II). Notes: * Data imputed. welfare aggregates for Nigeria and Notes: * Data imputed. welfare aggregates for Nigeria and Senegal Senegal are based on consumption, and for Afghanistan on are based on consumption, and for Afghanistan on assets. assets. 44 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making 45 IV. The Migration Experience This chapter tracks the long, treacherous, and • Other: The rest, primarily West Africans, took costly journeys of the asylum seekers who reached a different route, through Ouagadougou, Italy and those who reached Greece. No matter Dakar, Bamako, Gao, Tamanrasset, and Nia- what the route, the journey was dangerous and mey, apparently bypassing Agadez or Khar- expensive not only financially but also physically toum. Some of these did not report traveling and emotionally. Because violence characterized through any major city before Tripoli. the journey through Africa particularly, it deeply af- fected asylum seekers in Italy: 45 percent of them The majority of asylum seekers in Greece (77 reported experiencing violence in transit com- percent)—mostly Afghans, Iraqis, and Syrians— pared to just 5 percent in Greece. Nevertheless, the travelled to Izmir, Turkey, to get boats to Greece inflows suggest that many still considered emigra- (orange on the map). The other 23 percent appar- tion to have net benefits. ently sailed from smaller coastal cities. (Almost 70 percent of those interviewed reached Greece A Long and Perilous Journey in the first quarter of 2016, before the EU-Turkey joint action plan was operational.) Before arriv- Although asylum seekers came from about 30 ing in Izmir or another Turkish embarkation point countries worldwide, of those who traveled to Italy their routes varied, Syrians mainly entering Turkey most took one of three routes (Figure 4.1). • West Africans (most of the Nigerians, Gambi- “I knew from Satellite TV and the ans, Senegalese, Malians, Ivoirians, and Guin- news that the route was danger- eans) travelled overland to Agadez, Niger, and ous. We had heard all the stories then continued up to Tripoli, Libya, to get boats to Italy. About 36 percent of those arriv- about people drowning in the ing in Italy took this route (green on the map), Mediterranean. I personally know an entire family who drowned. • East Africans (Eritreans, Somalians, and Su- But our lives were in danger so danese) travelled to Khartoum, Sudan, then north to Tripoli to cross the sea. About 18 per- we had no choice.” cent of asylum seekers in Italy came by this Afghan man route (purple on the map). 46 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Figure 4.1. The Three Main Asylum-Seeker Routes to the EU IBRD 43701 | JUNE 2018 KAZAKHSTAN RUSSIAN FRANCE ROMANIA FEDERATION Bishkek ITALY BULGARIA GEORGIA Tbilisi Tashkent KYRGYZ REP. ARMENIA Baku UZBEKISTAN SPAIN Ankara PORTUGAL Izmir Yerevan AZERBAIJAN TURKMENISTAN TAJIKISTAN TURKEY Dushanbe Athens Ashgabat Algiers Tunis Valletta GREECE Nicosia SYRIAN Tehran MALTA CYPRUS A. R. Kabul Rabat TUNISIA Beirut Damascus Islamabad LEBANON Baghdad ISLAMIC AFGHANISTAN Tripoli REP. OF ISRAEL Amman IRAQ MOROCCO IRAN JORDAN PAKISTAN Cairo KUWAIT Kuwait New ALGERIA Delhi LIBYA ARAB REP. OF Manama BAHRAIN Western EGYPT Doha Riyadh QATAR Abu Dhabi INDIA Sahara U.A.E. Muscat SAUDI ARABIA OMAN MAURITANIA CABO NIGER SUDAN Nouakchott VERDE MALI REP. OF Agadez ERITREA Praia SENEGAL Khartoum YEMEN Asmara Sanaa Dakar CHAD BURKINA THE GAMBIA Banjul Bamako Niamey FASO DJIBOUTI Bissau Ouagadougou N’Djamena Djibouti GUINEA-BISSAU GUINEA NIGERIA Conakry BENIN Addis CÔTE Abuja Freetown TOGO SIERRA LEONE D'IVOIRE GHANA SOUTH Ababa CENTRAL Monrovia Yamoussoukro Cotonou SUDAN ETHIOPIA AFRICAN REP. SOMALIA LIBERIA Accra Lomé CAMEROON Bangui Juba Yaoundé Mogadishu From Agadez to Tripoli: From Khartoum to Tripoli: From Turkey to Izmir: West African population East African population Afghans (many going through Teheran), Iraqis and Syrians Source: Data from the EASS 2017. Note: This map was produced by the Cartography Unit of the World Bank Group. The boundaries, colors, denominations and any other information shown on this map do not imply, on the part of the World Bank Group, any judgment on the legal status of any territory, or any endorsement or acceptance of such boundaries. through the Syrian Arab Republic’s southern bor- vate cars and buses for at least part of the journey der with Turkey, Afghans consistently travelling and 34 percent used some form of public trans- overland through Iran. portation. About 36 percent traveled “off road” for part of the way, via walking or informal methods of On average, asylum seekers traveled through three smuggling; the off-road method was fairly similar relatively large transit points and spent a median in the eastern routes to Greece (34 percent) and of 26 days in each. Those reaching Italy stayed lon- the route to Italy via Libya (39 percent). ger, a median of 46 days, than those who travelled to Greece (8 days). Many survey respondents, par- The sea leg of the journey was most perilous. It ticularly those in Italian reception centers, worked took about three days to reach Italy by boat, but at the transit points, either because they expected only two hours to reach Greece. For 67 percent to settle there or because they needed more mon- of those who reached Italy and 94 percent who ey to continue their journey. reached Greece, the boats were inflatable (Figure 4.2), often necessitating a rescue at sea (Figure 4.3). Travelers typically used a combination of modes to Among those arriving in Italy, 96 percent needed reach their destination: about 53 percent used pri- to be rescued, as did 66 percent arriving in Greece. iv. the MigratiOn experience | 47 Figure 4.2. Asylum Seekers Reaching the EU via Figure 4.3. Asylum Seekers in Italy Rescued at Sea, Inflatable Boat, Percent Percent 100 100 90 90 80 80 70 70 60 60 50 50 40 40 30 30 20 20 10 10 0 0 Italy Greece Italy Greece Source: Data from the EASS 2017. Source: Data from the EASS 2017. While the Italy route took longer, both routes were The Financial Cost dangerous. In the qualitative interviews, many travelers spoke of their fear as their boats slowly The median transit cost for asylum seekers was filled with water or they spent hours adrift in the US$3,500. To put this in perspective, the median sea. “The boat was the most difficult thing,” de- payment by Sub-Saharan Africans was US$2,250, scribed one Syrian woman who travelled with her equivalent to more than three years of income for children, “I could not breathe and spent four hours a poor person (low-income-country poverty line: praying. I did not even look at the sea. In the middle US$1.90/day). For those reaching Greece from the of the journey the boat stopped working.” Middle East or South Asia (see Figure 4.4), the me- dian travel cost of US$5,000 would be over four Some described panicking when they saw the poor years of income for a poor person (low-middle-in- quality of the boats they would be traveling in, and come country poverty line: US$3.20/day). how overcrowded they were. One Nigerian man said, “We did not do up to one hour of the journey, About 42 percent of those who reached Italy and 17 the woods of the boat started breaking. After another percent of those who reached Greece worked along while, the engine stopped working. In the deep sea, the way (Figure 4.5). Most of those who worked water started entering the boat, but we poured it during transit were men (90 percent) and young- out. We were still doing this when the boat started to er than 35 (93 percent in Italy and 70 percent in compress. There were over a hundred of us…The risk Greece). The work tended to be manual labor: con- involved in the journey was beyond my wildest imag- struction for nearly 30 percent of those who worked ination.” About 5,096 people are reported to have and manufacturing for 9 percent (Figure 4.6). gone missing or died at sea in 2016 alone.34 In qualitative interviews, several West Africans re- ported leveraging personal networks to find jobs 34 https://data2.unhcr.org/en/situations/mediterranean Ac- in neighboring countries before the journey. How- cessed on May 21 2018. ever, other West Africans described travelling first 48 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Figure 4.4. Cost of the Journey, US$ Figure 4.5. Asylum Seekers Who Worked in Transit, Percent $6,000.00 50 $5,000.00 45 40 $4,000.00 35 $3,000.00 30 25 $2,000.00 20 $1,000.00 15 10 $0.00 5 Tripoli/Agadez Tripoli/Khartoum Other Italy Transitted through Izmir Other Greece Tripoli/Agadez Tripoli/Khartoum Other Italy Transitted through Izmir Other Greece 0 Italy Greece Source: Data from the EASS 2017. Figure 4.6. Employment of Those Who Worked in Transit, Percent Payment for Payment 35 all family per person 30 Source: Data from the EASS 2017. 25 20 and seeking work upon arrival—a strategy that 15 seemed to add to their vulnerabilities en route and 10 lent itself to ad hoc and precarious work. Many 5 described falling victim to criminal networks in 0 West Africa, especially as they sought work op- Agriculture Construction Manufacturing Buying/Selling Other Services portunities linked to further migration. Moreover, many reported working without pay—they worked either to get reduced transit costs or to escape imprisonment or crime syndicates. While they are particularly rampant in Libya, practices pf unpaid Source: Data from the EASS 2017. labor can be found throughout the journey from West Africa. reach Italy via Agadez (Figure 4.4) may be due to the fact that individuals financed for part of their Unpaid labor is most common on the West Afri- journey by working without pay. can Tripoli-via-Agadez route. Figure 4.7 shows the share of those who worked without pay among The Physical and Emotional Toll those who worked in transit. Of those reaching the Mediterranean through Tripoli-via-Agadez, nearly The journey imposed a physical toll on many asy- 61 percent worked without pay, versus only 25 per- lum seekers. Many who were interviewed spoke cent travelling through Khartoum and 18 percent of their exhaustion. Walking for days in the moun- through Izmir. Thus, the lower “cash price paid” to tains and crossing borders on foot in midwinter, iv. the MigratiOn experience | 49 Figure 4.7. Asylum Seekers Who Worked without Figure 4.8. Physical Violence en Route to Italy and Pay During Transit, Percent of all Who Worked En Greece, Percent Route 60 70 50 60 50 40 40 30 30 20 20 10 10 0 0 Tripoli/ Tripoli/ Tripoli/ Tripoli/ Agadez Khartoum Other Italy Transitted through Izmir Other Greece Agadez Khartoum Other Italy Transitted through Izmir Other Greece Italy Greece Italy Greece Source: Data from the EASS 2017. Source: Data from the EASS 2017. often with children, were common stories among A large group of asylum seekers, especially among Syrians and Afghans during qualitative interviews. those in Italy, described an even greater shock Many described seeing others die or suffering se- when faced with the challenges of the journey. rious injuries. Respondents in Italy also described Many had little information about the journey and deadly high-speed jeep journeys through the Sa- the situation in Libya, and it was often gleaned haran desert, hunger and dehydration, and seeing from smugglers. fellow travelers die and their bodies left on the road side. Many asylum seekers were exposed to physical violence in transit. This was a particularly strik- Almost all respondents described journeys that ing feature of the journey through Africa to Ita- were much more challenging than they expected, ly—45 percent of survey respondents reported although they had had some prior knowledge of experiencing physical violence on the way, no the risks. Yet many believed the expected bene- matter which route they took—compared to only fits still outweighed the costs. Wintah from Eritrea 5 percent of those who travelled to Greece (Fig- said, “I had heard many bad things about the jour- ure 4.8). Eritreans especially were at risk: 42 per- ney. I knew there was death: in Libya, they kill peo- cent reported being physically victimized, com- ple; in the Sahara, you can die of starvation; if you pared to 5 percent of Afghans and Syrians and 10 fall on the way, they leave you there.” She decided percent of Iraqis. to come anyway, sure that her life in Europe would be better: and “I also wanted to do it my way and Most of the difference in reported violence ap- go, even knowing nothing about Europe.” pears driven by whether a refugee came to Italy or 50 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Figure 4.9. Imprisonment on the Way, Percent Figure 4.10. Imprisonment of Primary and Secondary Movers, Percent 35 35 30 25 30 20 25 15 20 10 15 5 10 0 5 Tripoli/ Tripoli/ Agadez Khartoum Other Italy Transitted through Izmir Other Greece 0 Primary Secondary Primary Secondary Italy Greece Italy Greece Source: Data from the EASS 2017. Source: Data from the EASS 2017. lence (Box 4.1). However, some of the gender dif- Greece.35 While exposure to violence in transit was ference may be due to women under-reporting. generally higher for asylum seekers in Italy than Almost all of the women interviewed spoke of ex- Greece, within both samples those with both high periencing or witnessing violence—often sexual, and low education reported similar experiences which may be harder to talk about in quantitative with violence. However, in Italy those who had surveys. In contrast, the men interviewed in both paid less than the median price had experienced Greece and Italy often described reacting to vio- much more violence (51 percent) than those who lence as a major reason they got into trouble. Re- paid more (38 percent). For those in Greece, price acting to smuggler requests or behavior, whether was not a determinant of whether they experi- to themselves or women travelling with them, enced violence. Nor is there much difference in the was described as a cause of beatings or (on the experience of violence between primary and sec- Libya route) outright killing. ondary movers. Another difficulty along the way was detention, by Men reported more exposure to physical vio- either authorities or gangs. Detention was quite lence. Among those in Italy, 47 percent of men common along all African routes to Italy, less so and 40 percent of women had to deal with vio- along the routes to Greece (Figure 4.9). Among those traveling to Italy, 34 percent of the men and 18 percent of the women reported being detained 35 It may appear that those from high-recognition countries at some point. In both samples, primary movers confronted less violence, but that is skewed entirely by the were slightly more likely than secondary to be de- Greece sample, which met with much less violence in gener- tained (Figure 4.10). al on the way and consisted of asylum seekers primarily from high-recognition countries. Within Italy, the rates of violence during transit were similar no matter where the asylum seekers Consistent with other research and journalistic were from. reports, asylum seekers described being sold or iv. the MigratiOn experience | 51 Box 4.1. Single Women and the Journey to Italy In the Italy reception centers, 305 women (12 percent of the sample) were surveyed; of these, 5 percent were married and with their husbands, 28 percent married but traveling alone, and 67 percent unmarried. Of the single women, 53 percent came from Nigeria, 25 percent from Eritrea, and 11 percent from Côte d’Ivoire. Nearly 63 percent of them were younger than 25 and 96 percent were not yet 35. Their top reasons for migration were that their family was not safe (29 percent) and to escape conflict (21 percent). Others migrated for economic opportunity (13 percent), to avoid military service (13 percent), and to join family (6 percent). In the quantitative survey nearly, 46 percent of the single women reported being confronted by physical violence on the journey and 18 percent also reported being detained for a time, making them more vulnerable (Figure B4.1.1.). Figure B4.1.1. Violence Against It is possible, however, that these numbers underestimate Women in Transit the true level of violence against women; in interviews many 50 of them visibly struggled to talk about their experiences. It 45 appears from the qualitative interviews that sexual violence 40 affected almost all the women who reached Italy, whether 35 alone or with partners. It occurred mainly at the hands of 30 smugglers and criminal networks in or en route to Libya. 25 Women travelling alone who were interviewed also de- 20 scribed being forced into sex work in Libya, where some were 15 physically injured. Very rarely did women in Italy describe 10 sexual violence at the hands of other migrants or anywhere 5 else on the journey before Libya. To the contrary, many male 0 asylum seekers spoke up against the abuses of the women; Violence Prison some even recounted wanting or trying to protect them. Source: Data from the EASS 2017 – weighted sample. Men and women alike agreed that traveling in groups or with male family members was the most common mechanism to cope with expected risks. While sometimes effective, this did not prevent sexual violence—some women described being raped in front of husbands or co-migrants. Aside from sexual violence, some cited specific physiological disadvantages of the journey for women. Women described the difficulty of traveling while menstruating or pregnant. Miscarriages were blamed on the difficulty of the journey and were hard to manage. Ella from Eritrea described being “very ill” as she lost her baby and bled throughout her jeep journey across the Sahara. Be- coming pregnant was also a risk associated with sexual violence that some in Italy experienced. Source: EASS 2017 and qualitative surveys of this study. 52 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making handed by smugglers to criminal gangs in Libya. percent). The qualitative interviews suggest that They would be held captive until their families for many, the other countries were the first prac- could transfer money or until their labor could pay tical option or they had already settled in anoth- their “dues.” There were suggestions of differential er country, as Afghans did in Iran or West Africans treatment by nationality, in terms of what their in Libya. They were usually driven to move on by captors perceived as their ability to pay. “A Somali poor working and living conditions. Some had al- is like gold in Libya!” explained Abshir, a young So- ready been living in established refugee programs. mali man. “Somalis will pay more than the others to Eritreans and Syrians, for example, complained leave from Libya, and it has been the same for me. of poor work conditions in Ethiopia (the former) They asked me for US$7,000 per person to leave, and Turkey (the latter). As one Syrian woman ex- whereas others would pay US$800-1,000 dollars. My plained: “At first, we thought we would stay in Tur- family had to send money; if not they would have key because we are close to Syria and as soon as killed me. I had no choice, I couldn’t go back, so I the war ended, we would go back to Syria. But the paid.” situation got worse, and we lost hope of ever going back. My husband left first because there was no Intended Destinations work in Turkey, and I followed him later.” Not all refugees initially intended to settle in the Few Planned to Return Home EU. Nearly 20 percent of the asylum seekers inter- viewed in Italy stated that they originally intend- Despite all they have endured, many expressed ed to settle in countries outside of Europe (Figure hope for a future and opportunities in Europe. Re- 4.11). Rates were highest among those from Mali gardless of their reasons for leaving or how oner- (42 percent) and Guinea (30 percent) and lowest ous they found the journey or prolonged stays in for those from Somalia (13 percent) and Eritrea (6 reception centers, respondents interviewed al- most unanimously believed that their lives would Figure 4.11. Original Intention to Settle Somewhere improve within the next two years, and would be Other than the EU, Percent better than they had been at home.36 Uncertain 45 about where they would be relocated, some in 40 Greece gave a more tempered response but still foresaw improvement. Safety, employment, and 35 30 25 “Since I got here, I’ve started to 20 figure out my life, I started to 15 speak Italian. I want to have a job 10 and build my future in Italy.” 5 0 Guinean man MLI GIN GMB SEN CIV NGA SDN SOM ERI SYR IRQ AFG Italy Greece Source: Data from the EASS 2017. Country codes listed in 36 During qualitative interviews, respondents were asked to Table A1.1. Statistics for Sudan are computed from 33 grade their well-being from 0-10 at origin and at present and to observations and should thus be interpreted with caution. predict it two years ahead. iv. the MigratiOn experience | 53 (especially for the young) education were the main In both Greece and Italy the rate of those who nev- reasons for their optimism. er planned to return was about 60 percent. The numbers are much higher for those from Afghan- With hopes for a better future in Europe, a majority istan (88 percent) and Iraq (75 percent) than from expressed little willingness to return home. Figure the Syrian Arab Republic (42 percent). 4.12 plots the future plans of respondents by coun- try of origin. In Italy, nearly 60 percent of asylum The main reason of Afghans and Iraqis for dismiss- seekers interviewed—50 percent of those from ing return was the security situation: they thought high-recognition States and 66 percent of those that the conflict and uncertainty that had charac- from low-recognition States—planned never to re- terized their countries for such a long time would turn. Preferences on return were fairly similar for not be addressed in their lifetime. But other factors primary and secondary movers (Figure 4.13). may also have been operative. For instance, Af- Figure 4.12. Intention to Return to Country of Origin, Percent of Asylum Seekers 70 Intention to return: Greece 70 Intention to return: Italy 60 60 50 50 40 40 30 30 20 20 10 10 0 0 Never Once conflict Once Other Never Once conflict Once Other planning is resolved government planning is resolved government to return changes to return changes Source: Data from the EASS 2017. Figure 4.13. Intentions to Return, Primary and Secondary Movers, Percent 70 Intention to return: Italy 70 Intention to return: Greece 60 60 50 50 40 40 30 30 20 20 10 10 0 0 Never Once conflict Once Other Never Once conflict Once Other planning is resolved government planning is resolved government to return changes to return changes Primary Secondary Source: Data from the EASS 2017. 54 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making ghans interviewed who had spent their formative years living in Iran seemed particularly reluctant to go to Afghanistan: As one young man born and raised in Iran said, “What type of life am I going to have in Afghanistan . . . a country that I have never seen. A country where I don’t know anyone.” Aware- ness and fear of voluntary return programs were common among Afghan asylum seekers in ear- ly 2017 when the interviews took place and may have induced a bias in responding to the question. Asylum seekers identified many obstacles to re- turn. Some cited practical barriers, such as the lack of safety and having nothing to go back to due to conflict. This was common among those from high-recognition countries where conflicts have been recurrent. For example, one Syrian woman interviewed said, “I don’t believe Syria will go back to what it used to be. It is totally destroyed. If it goes back to what it used to be, I will go back.” However, there are other practical barriers to return, often financial, especially considering the sizable costs already incurred. The issue is reinforced when the investment was made possible by contribu- tions from relatives, friends, and members of the community with an expectation of remittances as interest payments. Practical barriers often com- bine with normative ones, generating additional perceptions of pressures to stay. A return might thus translate into both significant debt and social stigma for having failed to generate the expected return on investment. For women who traveled alone, the stigma they can expect at home is also a major deterrent to return. 55 V. Education, Skills and Work Experience An overwhelming majority of asylum seekers in both Italy and Greece view labor market integra- “The first thing is to get the doc- tion as critical for future success. In response to uments. As soon as I get them the question: “According to you, what would an asylum seeker need to have a successful integra- I’ll look for some training and I’ll tion experience in its host community? (multiple then look for a job.” answers possible).” 68 percent cited a job for at Guinean man least one household member. Others mentioned a good education for self or children. Similarly, 37 percent considered unemployment to be a major obstacle to successful integration, mainly because of the World Bank Skills Toward Employability and of language and skills deficiencies. Comparable Productivity (STEP) program. results emerged from open discussions where both men and women overwhelmingly thought of An important innovation of this study is admin- employment as the key to success, though wom- istration of a formal literacy skill assessment. In en in Greece were more likely to refer to jobs for close collaboration with the OECD, the study team husbands. For many, resolving issues relating to administered a computer-based assessment, work—the need to speak the language, getting which identifies for the test-taker a proficiency more education, or having work documents— level (below 1, levels 1, 2, and 3 and above). The were priorities. study tests were designed to determine proficien- cy at levels comparable to the OECD PIAAC results. To provide a more complete view of asylum seek- Similarly, asylum seeker results are comparable to er human capital, this report discusses not only literacy assessments obtained by the STEP pro- their education but also their technical and cog- gram. This allows for comparing asylum seekers nitive skills. The study explored their job history both with one another and with workers in host before and during migration and investigated countries. Notably, the assessment does not rely their self-reported skills, such as language, numer- on self-reported information that may be subject acy, and literacy. To allow wider comparisons, the to recall and conformity biases; nor does it assume questions on self-reported literacy and numeracy implicitly that educational attainment is compara- were borrowed directly from survey instruments ble across countries. 56 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Education and Country of Origin Figure 5.2. Education Level and Country of Origin 100 Thirty-two percent of asylum seekers in Italy and 90 29 percent in Greece had completed secondary 80 school or more. In Italy there were no marked dif- 70 60 ferences between men and women but in Greece 50 a gender gap in education was noticeable. In It- 40 aly, 33 percent of men and 28 percent of women 30 reported having completed secondary education 20 (see Figure 5.1). However, 62 percent of asylum 10 seekers in Italy, men and women alike, had com- 0 CIV GIN GMB MLI NGA SEN AFG ERI IRQ SDN SOM SYR pleted at least primary school. In Greece primary school completion rates were 74 percent for men Low recognition High recognition and 63 percent for women. The apparent absence Completed at least primary Completed tertiary of an education gender gap in Italy is mostly driven Completed at least secondary by the fact that 79 percent of women in the sample there originated from either Nigeria or Eritrea, the Source: Data from the EASS 2017. Country codes listed in two origin countries with the highest levels of edu- Table A1.1 cation for asylum seekers. When women and men Note: Proportions do not add up to 100% as individuals having completed tertiary education have also completed from the same country were compared, Eritrean primary and secondary education. and Nigerian women averaged 1.4 years of educa- tion less than their male co-nationals—about the same difference that can be seen domestically in their home countries.37 Figure 5.1. Asylum Seekers by Level of Education, Percent How educated asylum seekers are is mostly a function of schooling in their home countries (Fig- 100 ure 5.2), which can vary greatly. Eritrea, the Syrian 90 Arab Republic, Nigeria, and Iraq had the highest 80 percentages of educated individuals. Nigeria and 70 Eritrea also had the most asylum seekers who 60 completed secondary school or more. Iraqis, how- 50 ever, were less likely than others except Malians to 40 have had any secondary education. But 11 percent of Somali and Syrian asylum seekers had some 30 tertiary education. 20         10 Figure 5.3 plots the education of asylum seekers 0 against their countries of origin. Points on the Male Female Male Female Italy Greece     No educ Primary Secondary Tertiary 37 Among asylum seekers who did not complete primary ed- ucation, 16 percent of those in Italy and 8 percent of those in Source: Data from the EASS 2017. Greece had participated in a program to learn to read or write. v. educatiOn, SkillS and wOrk experience | 57 Figure 5.3. Years of Education, Asylum Seekers and 45-degree line indicate equality between the two; Males at Home, 2015 the correlation (excluding Sudan) is a very high 80 percent. 10 Average years of education of asylum seekers (EASS 2017) IRQ Among asylum seekers who completed at least NGA 8 four years of general and technical education, GMB SYR business administration and education were the main fields of specialization for both men and 6 SOM CIV women. These most educated individuals repre- sented about 12 percent of the total flow and gen- SEN SDN 4 erated 491 study observations. Among them, 12 percent studied business administration, account- MLI ing, or commerce, and 12 percent in Italy and al- 2 most 28 percent in Greece studied education (see Figure 5.4). Strikingly, education was the area of 0 specialization for 49 percent of the educated wom- 0 2 4 6 8 10 en seeking asylum in Greece. Science, mathemat- Average years of education for men age 15 and above in ics and scientific computing were the main field of origin countries in 2015 (World Bank 2018) study for 10 percent in Italy and manufacturing for 6 percent in Greece. Source: Data from the EASS 2017 and World Bank 2018. Country codes listed in Table A1.1 Note: Statistics for Sudan are computed from 33 observations The study data on training outside the formal edu- and should thus be interpreted with caution. cation system in the year before leaving for Europe Figure 5.4. Fields of Higher Education (General and Technical), Percent 30 25 20 15 10 5 -     Education Buisiness, admin Law Health Maths, computing Technical and eng Agriculture Social and behav Languages, classics General Restaurant, tourism Manufacturing Construction trade Public order and safety Appliance repair   Italy Greece                 Source: Data from the EASS 2017. 58 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making suggest that up to 8 percent of those who went to for low-skill workers past employment has been Italy and 6 percent for those who went to Greece found to increase chances of current employment prepared by enrolling in language studies, manu- (Aeberhardt and Pouget 2007; Acemoglu and Au- facturing, or science courses (including computer tor 2011). science), a result that echoes Stark, Helmenstein, and Prskawetz (1998). The differences between asylum seekers in Italy and Greece may be driven by the flow’s demo- Finally, self-reported literacy and numeracy rates graphics. First, the higher proportion of women are consistent with asylum seeker educational at- in Greece explains the lower rate of previous work: tainment. On average only 28 percent of asylum No matter what their age, only about 28 percent seekers in Italy and 20 percent in Greece reported of women from the Syrian Arab Republic, Iraq, and having read something in the past year, even very Afghanistan had any jobs before migrating. Wom- short notes.38 Of those, almost 50 percent of those en from Sub-Saharan Africa who migrated to Italy, in Italy and 62 percent in Greece mentioned having however, were almost twice as likely (46 percent) read e-mail, a demonstration of computer skills. to have had jobs than the women in Greece (28 Moreover, 24 percent in Italy and 15 percent in percent). For the men, 65 percent in Italy and 63 Greece had written something or filled out forms. percent in Greece had an employment history. For numeracy, 61 percent in Greece and almost 50 percent in Italy reported having used some math- A second difference has to do with age: While men ematics, such as estimating size or distance and aged 26–36 in both Italy and Greece were equally computing price and cost.39 About 23 percent in likely to have previously had a job, the percentag- both countries used more advanced mathematics, es for men aged 18–25 in Greece were much lower such as calculating fractions, multiplying, or using (Figure 5.5)—a finding that reflects youth unem- algebra or geometry. Figure 5.5. Past Employment by Age and Gender, Percent Work Experience 90.00 Asylum seekers in Italy had more work experience. 80.00 About 62 percent there have had jobs, compared 70.00 to 51 percent in Greece. These numbers matter: 60.00 50.00 40.00 38 The level of self-reported literacy can be inferred from the 30.00 reading, forms and writings variables—binary variables taking value 1 if the respondent has read something (including very 20.00 short notes), filled out forms, or written something in the past 10.00 year. - 39 Self-reported numeracy: the maths variable allows assess- Male Female Male Female ment of the degree of mathematical complexity applied in the Italy Greece past year. The full question is: Thinking of your activities over the past 12 months before leaving your country, have you done any 18-25 26-35 36-45 46 and + of the following ? Measure or estimate sizes, weights, distanc- es, etc ; Calculate prices or costs ; Use or calculate fractions, decimals or percentages; perform any other multiplication or Source: Data from the EASS 2017. division ; use more advanced math, such as algebra, geometry, Note: Display is of asylum seekers who reported having trigonometry, etc.; other maths. worked at least once before migration. v. educatiOn, SkillS and wOrk experience | 59 Figure 5.6. All Jobs before Migration, Percent Figure 5.8. Male Asylum Seekers who Worked While in Transit, by Country of Origin, Percent 50 45 60.00 40 35 50.00 30 25 40.00 20 15 30.00 10 5 20.00 - Services Skilled Craft Elementary & sales agriculture & trade 10.00 Italy Greece - CIV ERI GIN GMB MLI NGA SDN SEN SOM AFG IRQ Source: Data from the EASS 2017. Note: Other jobs cited occasionally were armed forces Italy Greece occupations, managers, professionals, technicians and associate professionals, clerical support workers, forestry and fishery Source: Data from the EASS 2017. Country codes listed in workers, and plant and machine operators and assemblers. Table A1.1. Statistics for Sudan are computed from 33 observations and should thus be interpreted with caution. Figure 5.7. Asylum Seekers who Had Had Elementary Jobs, by Education, Percent cohorts, they represented only a small fraction of the total flow of migrants. 60 50 Most asylum seekers who had work experience 40 before leaving for the EU had been in low-skill el- 30 ementary jobs in construction, manufacturing, or agriculture or worked as cleaners or street vendors 20 (Figure 5.6). The International Labor Organiza- 10 tion (ILO) defines elementary jobs as being based - on simple tasks, and mostly requiring the use of Italy Greece hand-held tools or physical effort. As expected, 56 No education Completed primary percent of asylum seekers who had had jobs but Completed Secondary had little education were in elementary occupa- tions (Figure 5.7). Yet 30 percent of asylum seekers who had some secondary education still worked Source: Data from the EASS 2017. in elementary jobs. ployment rates of 30.6 percent in the Syrian Arab Republic, 32.8 percent in Iraq, and 18.6 percent in A substantial fraction of asylum seekers who Afghanistan (ILOSTAT 2017). 40 Although there were migrated through Libya took elementary jobs also slight Italy-Greece differences among older while in transit (Figure 5.8). None who had not had jobs before migrating did so. Among the men who had work experience before departure, 40 https://data.worldbank.org/indicator/SL.UEM.1524.ZS. however, 69 percent of those who went to Italy 60 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Figure 5.9. Most Frequent Types of Jobs Taken in Figure 5.10. Asylum Seekers in Italy who Speak at Transit, Percent least one European language, Percent Elementary 100 Speaks at least one of the selected language Operator 80 Craft Skilled agric. 60 Services 40 Clerks 20 Technicicians Agriculture 0 Low High High recognition recognition recognition Construction (Italy sample) (Italy sample) (Greece sample) Personal services Source: Data from the EASS 2017. Manufacturing Note: The main European languages spoken are English, French, Italian, and Turkish. 0 10 20 30 40 50 60 Greece Italy Speakers of European Languages   Source: Data from the EASS 2017. A large number of asylum seekers, 80 percent in Note: The percentages of individuals who mentioned a specific job type (1,235 total, 169 of them in Greece). Operator Italy and 45 percent in Greece, speak at least one stands for Plant and machine operator, Skilled agric. stands European language, usually English (Figure 5.10). for Skilled agriculture, Services stands for Services workers, shop, market sales. The difference is attributable to the fact that En- glish and French are official languages in many of the home countries of asylum seekers in Italy but less than 20 percent of those who went to (English in Nigeria and The Gambia; French in Sen- Greece found work en route. Among them were egal, Côte d’Ivoire, Mali, and Guinea). Moreover, also 43 percent of the women in Italy and 28 per- at least two European languages are spoken by cent of those in Greece. Both men and women about 30 percent of Guineans and Malians, 20 per- worked mostly in agriculture (36 percent), con- cent of Ivorians, Senegalese, and Gambians, and struction (31 percent), and personal services (19 10 percent of Nigerians (Figure 5.11). percent), but the 11 percent of those going to Greece who worked during transit were main- Among those from the Middle East and South Asia, ly employed in manufacturing. Among men 42 percent of Afghans, 35 percent of Iraqis, and 46 migrating to Greece, 19 percent of Syrians and percent of Syrians speak English; 14 percent of Syr- 3 percent each of Afghans and Iraqis had work ians also speak Turkish and 11 percent speak two while in transit. The same destination contrast European languages. In Italy, 17 percent of male is found with women: while 19 percent worked and 4 percent of female asylum seekers report- during transit to Italy, only 6 percent did so on ed speaking Italian; language classes are offered their way to Greece. The likelihood of working in most asylum seeker centers in Italy. Finally, of during transit was not related to an individual’s 4,136 asylum seekers surveyed, only 33 (0.7 per- education. cent) reported speaking German. v. educatiOn, SkillS and wOrk experience | 61 Figure 5.11. European Languages Spoken by Male Asylum Seekers, Percent 100 90 80 70 60 50 40 30 20 10 0 CIV ERI GIN GMB MLI NGA SDN SEN SOM AFG IRQ SYR Italy Greece At least one Europ. language More than one European language Italian Turkish Note: Country codes listed in Table A1.1. Statistics for Sudan are computed from 33 observations and should thus be interpreted with caution. Literacy Proficiency Levels Compared The assessment is particularly appealing in the study context. Among common measures of hu- In addition to a survey questionnaire, a comput- man capital are education and experience. How- er-based literacy assessment was administered to ever, asylum seekers originate from a vast array of a random sub-sample of the survey respondents. countries where educational attainment and la- In both Italy and Greece, a smaller and randomly bor market experience might not be comparable. chosen group of survey respondents were invit- In addition, both variables are self-reported and ed to take a computer-based literacy assessment subject to response bias. In contrast, the literacy based on OECD’s Programme for the International assessments deliver proficiency scores (between Assessment of Adult Competencies (PIAAC). In Ita- 0 and 500) and levels (below level 1, level 1, level ly, given the asylum seeker population’s language 2 and level 3 and above) that are comparable to proficiencies, the literacy test was administered in proficiency levels assessed by PIAAC and similar either (Irish) English or (Canadian) French using assessments like the World Bank’s STEP survey, the Education & Skills (E&S) online tool.41 Given and may therefore give a better picture of where their home country languages, Eritreans, Somalis, asylum seekers actually fall in the skills distribu- and Sudanese were not part of the test. In Greece, tion of the EU or that of third countries. a shorter version of E&S online was developed for this study using similar items that were translated The literacy tests are adaptive assessments: into Arabic and Farsi. test-takers first respond to six core items that de- termine whether they can take the full test. The items have varying levels of difficulty, as measured 41 http://www.oecd.org/skills/ESonline-assessment/aboutedu- by the fraction of the population providing a cor- cationskillsonline/ rect answer. The core items are: 62 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making • “SGIH” consists of identifying a telephone 8 percent of the sample. declined to take the test. number on a business card. In Italy, however, while French and English are of- • “Bottles” consists of counting a stack of bottles. ficial languages for countries more asylum seekers • “Employment Advertisement” consists of had come from, for 26 percent of the respondents, identifying the number of job openings in a French or English was not a mother tongue or a one-page employment ad. language spoken at home or at school. • “Election Results I” consists of identifying the candidate with the fewest votes in an election The researchers invited 526 individuals to take the report. test in Greece and 286 in Italy. However, 15 percent • “Election Results II” consists of stating elec- in Greece and 5 percent in Italy refused. Those who tion turnout from the same election report. claimed they were illiterate (59 in Italy and 62 in • “Airport timetable” consists of determining Greece) were assigned a proficiency level below 1. the departure time for a flight listed on an air- For the others (20 in Italy and 79 in Greece), Tables port flight board. A4.2 and A4.4 in the Appendix 4 suggest that no specific group of asylum seekers (e.g., the less ed- Proficiency is established by calculating the aver- ucated or older) was more likely than any other to age number of correct answers on the whole litera- decline to take the literacy assessment. These cas- cy test, with more difficult questions carrying more es were dropped from the analysis, but including weight. With PIAAC and STEP, and also E&S online, those who declared themselves illiterate and giving the assessment delivers a continuous proficiency them a below-1 proficiency level does not alter the score ranging from 0 to 500. For Greece, in order conclusions significantly. Ultimately, 376 tests were to be able to administer the test in either Arabic administered in Greece and 202 in Italy. A detailed or Farsi, a dedicated assessment was developed. methodological discussion of sampling, attrition, To shorten the length of the assessment, fewer and validation is provided in the Appendix 4. items were used, and given the low levels of edu- cation observed among asylum seekers during the Asylum Seekers and Other Recent Migrants pilot phase, it was decided to retain mostly easier in Europe items so as to measure proficiency more precisely among low performers, though at the expense of The literacy assessment found low general profi- high performers. The proficiency score is there- ciency in the target languages. Giving all individuals fore a noisier measure of proficiency than PIAAC at levels below 1 a proficiency level of zero, average or STEP and for high performers, it is likely to be proficiency for asylum seekers in Greece was 0.91, biased downward. Most of the analysis thus uses with 0.61 for Afghans and 1.1 for the Syrians (Figure proficiency levels rather than scores with levels 5.12A). In Italy the average proficiency level was ranging from -1 or zero through levels 1, 2, 3 and 0.62, which, due to small sample sizes, is not broken beyond. down by country of origin. From another viewpoint, the test results indicate that the proportion of indi- Because the test was administered in one of four viduals with proficiency level 1 and above was 57 languages --Arabic, Farsi, French, and English--- percent in Greece and 41 percent in Italy. the proficiency assessment captures both indi- vidual skills and command of the target language. A STEP survey in Ghana makes it possible to com- For most test-takers in Greece, Arabic and Farsi pare asylum seekers from West Africa with urban were native languages; Iraqi Kurds, who comprise adults in Ghana. Proficiency levels in urban Ghana v. educatiOn, SkillS and wOrk experience | 63 Figure 5.12. Literacy Proficiency in Target Languages, Greece and Italy 1.2 70 1 60 50 0.8 40 0.6 30 0.4 20 0.2 10 0 0 SYR AFG IRQ AFR SYR AFG IRQ AFR Source: Data from EASS 2017-Literacy. Note: AFR represents the weighted average of proficiency levels of test-takers from Côte d’Ivoire, Guinea, The Gambia, Mali, Nigeria, and Senegal. are relatively low but not too different from those of Figure 5.13. Proficiency Levels, Urban Adults in Western African asylum seekers in Italy (Figure 5.13). Ghana and Asylum Seekers in Italy Proficiency levels of asylum seekers in both Italy 45 and Greece are similar to those of other migrants 40 in European countries, particularly those who ar- 35 rived less than 5 years ago (Figure 5.14). The com- parison focuses on the migrants who have been in 30 host countries for less than five years. The average 25 PIAAC scores for asylum track those of migrants in European countries like Finland or Sweden. For 20 validity, the asylum-seeker group in these compar- 15 isons is comprised of individuals who agreed to 10 take the assessment and performed well enough on the assessment to obtain a proficiency score. 5 Consequently, asylum seekers who declined to 0 take the test are not taken into account when mak- Level 1 and Level 2 and Level 3 and ing the comparisons. Furthermore, the proficien- above above above cy scores on assessments conducted in Greece Asylum seekers in Italy Ghana need to be interpreted with caution; the scores will be biased downward for top performers and are a much less precise measure of proficiency Source: World Bank STEP Ghana (2017) and EASS 2017-Literacy. than those achieved in E&S online, PIAAC, or STEP. 64 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Figure 5.14. Average PIAAC Scores, Asylum Seekers in Italy and Greece, and Earlier Migrants in Europe 350 300 250 200 150 100 50 0 IRL GBR ESP ITA AUT DEU BEL FRA DNK NOR NLD FIN SWE AFR SYR AFG IRQ EU EASS 2017 Native-born Foreign-born Source: OECD 2016;’results of EASS 2017-Literacy. Note: AFR is the weighted average of proficiency levels of test-takers from Côte d’Ivoire, Guinea, The Gambia, Mali, Nigeria, and Senegal. Foreign-born refers only to migrants who have been in the host country for no more than five years. Figure 5.15. Proficiency levels, by educational attainment 100 Italy sample 100 Greece sample 90 90 80 80 70 70 60 60 50 50 40 40 30 30 20 20 10 10 0 0 No Primary Secondary Tertiary No Primary Secondary Tertiary education education Source: Data from EASS 2017. However, the results do suggest that the asylum lum seekers in Greece, 25 percent of those with no seekers in Italy and Greece who obtained a profi- formal education scored at level 1 or above com- ciency score did not perform any worse than re- pared to 95 percent of those with university edu- cent migrants to a handful of European countries. cation. Education is also closely correlated with the proficiency levels of asylum seekers in Italy. Figure 5.15 suggests a strong positive correlation However, past labor force participation does not between asylum seekers’ education and their pro- affect tests results; those with work experience are ficiency in the target language. On average, of asy- no more proficient than those without. v. educatiOn, SkillS and wOrk experience | 65 Box 5.1. Effect of Validation with Core Items Analysis of core items makes it possible to assess the reliability of an instrument. Performance to the test might be driven by factors that have more to do with the test design and how it is con- ducted than the intrinsic proficiency of the test taker. For some the language of the test might not be the native language. Also, a possibly large fraction of asylum seekers might be in psychological distress that affects their ability to take a 45-minute computer-based test. Finally, despite efforts to comply with guidelines for the training of facilitators, the conditions in which the test was ad- ministered might have affected individual performance. a We made every attempt to reduce dis- turbances during testing, but having to administer tests in asylum center public spaces, indoors or outdoors, meant that not all disturbances could be minimized. The proportions of correct answers to core items for asylum seekers in both Italy and Greece exhibit patterns similar to those measured by PIAAC throughout the world and especially among Europeans. In Figure B5.1.1, the horizontal axis represents the fraction of correct answers for the test-takers for each of the six core items. The vertical axis indicates the fraction of correct answers to the same six questions for the entire worldwide population who took a PIAAC test either on paper or on computer and were given a proficiency level below 1. Figure B5.1.1. Asylum Seekers and PIAAC Worldwide Population, Scoring below Level 1 on Proficiency 1 1.0 worldwide PIAAC population (below level 1) Proportion of correct responses, for the 0.8 Bottles 0.8 SGIH SGIH Employment Advertisement 0.6 AVERAGE 0.6 Bottles Election Election Results II Employment Results I AVERAGE 0.4 Advertisement 0.4 Airport Election Results II Timetable Election Results I 0.2 0.2 Airport Timetable 0.0 0.5 1.0 0.0 0.5 1.0 Proportion of correct responses, Proportion of correct responses, for asylum seekers in Italy (EASS 2017) for asylum seekers in Greece (EASS 2017) Source: Data from EASS 2017 – Literacy and OECD PIAAC data. The alignment of the six points along the 45-degree line indicates that in both samples, the pro- portion of correct answers to an item drops as the item gets more difficult, and for both samples the drops suggest a pattern similar to that of the worldwide population. The low proficiency scores measured by this assessment for asylum seekers are thus not likely to be the result of lack of mo- tivation, the psychological distress of test takers, or the physical conditions of test administration. a http://www.oecd.org/skills/piaac/PIAAC-NPM(2014_06)PIAAC_Technical_Standards_and_Guidelines.pdf. 66 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Figure 5.16. Education, Employment Experience, and Skills 100 90 80 70 60 50 40 30 20 10 0 Completed Completed Has ever Worked Speaks EU Proficiency level primary secondary worked in transit language 1 and above Education Employment Skills Italy Low Recognition Italy High Recognition Greece High Recognition Source: Data from EASS 2017. Summary Findings Heather (2000) and Clarke and Withers (2002) found that the gender gap in labor force participa- Figure 5.16 summarizes the findings for asylum tion fades with time and over generations. These seeker education, work experience, and literacy might be reasons why some migrants are eventu- proficiency. Educational attainment and literacy ally integrated into the labor force of hosting coun- proficiency are higher for asylum seekers from high tries (see, e.g. LaLonde and Topel 1991; Baker and recognition-rate countries but prior labor force Benjamin 1994; Adsera and Chiswick 2007; Borjas participation and a higher command of some of- 2015), and why such integration often results in ficial languages in the EU are higher for those from displacing earlier waves of migrants rather than low-recognition-rate countries. the native population. While this report gives a static assessment of the supply, or potential supply, of labor, further dis- cussion on labor market integration prospects for those who are granted asylum needs to (1) char- acterize the demand side of the European labor market and (2) acknowledge that skills can be ac- quired over time. Autor and Dorn (2013) argue that demand for low-skill labor might be increasing in developed countries. On the skill acquisition side, cases of immigrants accumulating skills appropri- ate to the host country have been documented by LaLonde and Topel (1991) and Stark, Helmenstein, and Prskawetz (1998), among others. Antecol and 67 VI. Anxiety and Depression There has been growing interest in defining the one third premature mortality from noncommuni- mental health and psychosocial status of vulner- cable diseases through prevention and treatment able populations and the need for early detection and promote mental health and well-being.” and treatment to minimize debilitating or long- term effects. Those affected by unrest, violence, Mental health is defined as a state of well-being loss, separation, and drastic changes in social and in which every individual realizes his or her own living conditions are likely to have such distressing abilities, can cope with the normal stresses of life, psychological reactions as hopelessness, helpless- can work productively and fruitfully, and is able ness, anxiety, and depression, as well as behavior- to positively contribute to the community they al and social problems. The most common mental live in (WHO 2013). The wide range of mental and health diagnoses among refugees are emotional behavioral disorders is described in the Interna- problems like depressive and anxiety/depression tional Statistical Classification of Diseases and Re- disorders, such as post-traumatic stress disorder, lated Health Problems (ICD;10th and 11th edition, generalized anxiety, panic attacks, adjustment ICD-10 and ICD-11); among them are depression, disorder, and somatization. Such problems make bipolar affective disorder, schizophrenia, anxiety it difficult for people to attend to their physical disorders, dementia, substance use disorders, in- health needs, routine daily tasks, and relationships tellectual disabilities, and developmental and be- with others (World Bank 2016a). In May 2012, the havioral disorders (WHO 2001). 65th World Health Assembly adopted resolution WHA65.4 on the global burden of mental disorders The analysis discussed in this chapter focuses on and the need for a comprehensive, coordinated the determinants of anxiety and depression mea- response by national health and social sectors sured by the Global Severity Index (GSI). The index (WHO 2013). The goals of the 2013–20 WHO Action is constructed as the mean scores of the following Plan are to promote mental well-being, prevent seven indicators. mental disorders, provide care, enhance recov- ery, promote human rights and reduce mortality, • Energy: During the last week did you feel low morbidity, and disability for persons with mental in energy, slowed down? disorders. Mental health is now on the Sustainable • Guilt: During the last week did you accuse Development Agenda adopted at the United Na- yourself of different things? tions General Assembly in September 2015; Target • Sleep: During the last week did you have prob- 3.4 requests that countries “By 2030, reduce by lems falling asleep or sleeping? 68 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making • Hope: During the last week did you feel hope- those in Italy and 70 percent in Greece have severe less in terms of the future? symptoms of anxiety and depression. In addition • Melancholy: During the last week did you feel to stressors like war and violence that may have melancholic? triggered the decision to migrate, violence during • Worry: During the last week did you feel that transit caused further deterioration of their mental you worried too much about different things? condition. Moreover, prolonged uncertainty about • Effort: During the last week did you feel that approval of the asylum application may be an ad- everything was an effort? ditional source of stress. The scores range from 1 (no symptoms of anxiety/ Anxiety, Depression, and Exposure to depression) to 4 (severe symptoms of anxiety/de- Violence in Transit pression); the GSI is the average of the seven in- dicator scores. Box 6.1 documents other studies The GSI for this study sample averaged 2.25; 55 that aim at screening anxiety and depression and percent of respondents were suffering from severe relate results to certain characteristics and shocks. mental distress (Figure 6.1). Since the anxiety and depression screening module has not been validat- Among asylum seekers in general, the incidence ed, the instrument is not designed to identify anxiety of severe mental distress is high: 40 percent of disorders or clinical depression. Instead, an individ- Box 6.1. Studies of Anxiety and Depression and Demographics Taking a similar approach to the one used in this report, Das et al. (2008) studied mental health patterns and socioeconomic characteristics of individuals in Indonesia and Mexico. Using a na- tionally representative survey of more than 10,000 households they used a variation of the Gen- eral Health Questionnaire (GHQ) measure of anxiety and depression to identify socioeconomic characteristics significantly linked symptoms of anxiety and depression, such as gender, age, and physical health, and shocks that hit households. Not closely linked were poverty and education level. Based on a 12-question GHQ version of a population-wide survey of a poor and conflict-affected country, Baingana et al. (2004) analyzed links between mental health and socioeconomic out- comes in Burundi. They found that mental health problems due to psychiatric trauma and de- pression reduce both labor force participation and investment in the education of children. Starting from the 1997 Indonesian financial crisis, Friedman and Thomas (2007) studied the con- sequences of such shocks on economic welfare, physical health, and child education. Using the Indonesia Family Life Survey, they conducted a short screening survey adapted from the GHQ to measure the symptoms of depression and anxiety and found that the crisis had undermined the psychological wellbeing of the sample. They also found that the crisis had more impact on less educated groups, the rural landless, and residents of provinces heavily affected by the crisis. vi. anxiety and depreSSiOn | 69 Figure 6.1. Severe Mental Distress, GSI by Country of Origin 2.8 2.6 2.4 2.2 2 1.8 1.6 1.4 1.2 1 CIV GIN GMB MLI NGA SEN ERI SDN SOM AFG IRQ SYR Low recognition High recognition High recognition Italy Greece Source: Data from the EASS 2017. Country codes listed in Table A1.1. Statistics for Sudan are computed from 33 observations and should thus be interpreted with caution. ual was considered in severe mental distress when being older and being a woman are associated the measured GSI was above a cut-off of two, which with worse GSI (Figure 6.2), although for this group was also used by Das et al. (2009). The levels of education does not seem to be correlated with mental distress of those refugees in Greece seemed levels of anxiety and depression (Figure 6.3). These to be worse, with a mean of 2.5.42 By country of ori- patterns have been found in both developed and gin, Iraq had the highest score, close to 3, followed developing countries (Gold 1998; Andrade et al. by Afghanistan and the Syrian Arab Republic. Sene- 2002; Andrews et al. 2001; Awas et al. 1999; Kessler gal, Guinea and Côte d’Ivoire also had mean scores et al. 2005; Patel et al. 1999; Das et al. 2007, 2009). above 2 with Nigeria, Mali, Somalia, Sudan, Eritrea, While West African asylum seekers suffer more and The Gambia not far behind at just under 2. In from anxiety and depression than East Africans, sum, 55 percent of all respondents were deemed differences in education could explain the gap: in severe mental distress, among them 70 percent East African asylum seekers are on average more of respondents from the Syrian Arab Republic, Iraq, educated than West African. and Afghanistan hosted in Greece and 40 percent of respondents hosted in Italy. Women migrating to Greece have higher anxiety and depression scores than men: 70 percent of A multivariate regression analysis of risk factors for men and 80 percent of women have a GSI above anxiety/depression among asylum seekers found 2. However, unlike the population in Greece, patterns consistent with earlier studies in which women in Italy do not report more psycholog- ical distress than men, and GSI are statistically indistinguishable at 2. Of women reaching Ita- ly, 56 percent were single (against 18 percent in 42 Compare: Das et al. (2009) reported mean scores below 2 for all five countries they analyzed Bosnia and Herzegovina 1.495; Greece), and 41 percent of women in Italian re- India 1.535; Indonesia 1.413; Mexico 1.341; and Tonga 1.745. ception centers reported being single and alone. 70 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Figure 6.2. Mental Health Score by Exposure to Violence in Transit Italy sample Greece sample 4 4 3.5 3.5 3 3 2.5 2.5 2 2 1.5 1.5 1 1 0.5 0.5 0 0 No violence during Violence during transit No violence during Violence during transit transit transit Source: EASS 2017. In Greece, however, only 2 percent were women II. The transit phase covers the entire journey, single and travelling alone; the vast majority are whatever the route, from country of origin to married and travelling with spouses and children. resettlement site or host country. It refers to Moreover, 79 percent of them report not having difficult travel conditions, temporary settle- ever worked before, making them “tied” movers ment in refugee camps, or detention, and of- (Mincer 1978). ten involves further loss and traumatic stress- ors due to experiences encountered by the Other risk factors of anxiety and depression have refugee at transit points before arriving in the been decomposed into the ones associated with host country. each of the three main phases of the asylum-seek- er experience: pre-departure, transit, and resettle- III. The settlement phase covers events or experi- ment (Appendix 5 shows variables for each phase ences asylum seekers encounter daily in host used in the regression analysis). Pre-departure, countries as they compare current living con- transit, age, and gender are important drivers of ditions to earlier ones. Among these are the the mental health status of displaced people and uncertainty associated with the asylum appli- refugees (World Bank 2016b). cation process and issues related to adapting to a foreign environment. I. The pre-departure phase is defined by events that took place in the country of origin before Multivariate regression analysis (see Table A5.1 an asylum seeker departed. These could be to A5.3 in Appendix 5) indicates that respondents loss of a family member, a livelihood, or be- who experienced pre-departure shocks such as longings; or physical and emotional trauma to the death of a family member or the loss of prop- the individual or family, witnessing of tortur- erty were more anxious and depressed. Loss of a ing or murder, and social disorder. dwelling seems to have particularly deleterious vi. anxiety and depreSSiOn | 71 effects on the GSI. Note that dwelling loss is not restricted to countries at war, like the Syrian Arab “I don’t know anything about the Republic; it also affects 58 percent of Iraqi asylum future. I don’t know if they will seekers; in Italy, between 30 and 55 percent of respondents reported having had their home de- give us asylum here or deport us. stroyed. Similarly, many respondents from Mali, I don’t know if they will give us Côte d’Ivoire, Senegal, and Guinea report having a house. I don’t want to know. I lost a family member. can’t handle any more pressure.” In both Italy and Greece, asylum seekers with past Iraqi woman employment experience have less psychologi- cal distress. Accounting for other stressors in the multivariate regression, having worked before mi- seekers, even within the household, are not un- gration is found to decrease the GSI by 0.11 (see common. Fights between husbands and wives, Appendix 5). and more rarely episodes of domestic violence were mentioned explicitly by those in the Greece Asylum seekers who experienced violence and sample; a Syrian woman commented, “I am fight- other shocks in transit were in more severe mental ing more with my husband because I’m impatient. distress than those not confronted by violence on I just want to be reunited with [my son in Germany] route. Although the finding holds true for asylum but it’s not happening, and I have to take care of ev- seekers in both Italy and Greece (Figure 6.4), the erything here.” magnitudes differ: the effect of experiencing vio- lence during transit seems higher in the Greek than in the Italian sample. Finally, it was not possible to link duration of asy- lum or living conditions in reception centers with symptoms of anxiety and depression. The survey populations had been in Italy or Greece for about the same amount of time, leaving little variation to test for a correlation between duration and the GSI. In qualitative interviews, many respondents spoke spontaneously of prolonged uncertainty over the legal status and future residence as being a source of stress. In fact, 77 percent of the sample had spent more than six months since landing, often without a clear idea of what the future holds. As a young Afghan man said: “Living without knowing your future is very hard. Many are depressed here [in the camp]. My sister has lost her mind here.” Also, reports of tensions and violence between asylum 72 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making 73 VII. Conclusions and Insights into Policy The characterizations in this report of those seek- olence, mental health status, and cognitive skills. ing asylum in the EU, their backgrounds, and their However, cost-effective alternatives need to be profiles, points out their differences and similarities politically supported to promote acquisition of (1) in terms of needs and prospects, and hence the data that rigorously capture the population of in- need for tailored policy responses. While a com- terest, which often is a small fraction of the popu- prehensive policy agenda on migration and forced lation of both sending and host communities and displacement is beyond the scope of this report, also often geographically dispersed; and (2) data the evidence presented here—based on a large that are collected more often, with sampling that representative sample of the adult population of accounts for high geographical mobility. Sustain- asylum seekers in Italy and Greece as of 2017—can able approaches to collecting data on refugees, inform humanitarian and development approach- asylum seekers, other migrants, and host com- es to policymaking in the following areas: munities need to be embedded in national data collection mechanisms. Policies to improve data Data on transient and vulnerable populations should build up the quality of both administrative data and monitoring systems; they should also en- Evidence-based policies on migration, forced or courage cooperation with national statistics offic- voluntary, need data to inform their design and es to explore how to adapt existing instruments to evaluate their impact not only on the migrants cover the population of interest effectively. Inno- themselves but also on sending, transit, and re- vations in data collection can be valuable, such as ceiving communities. Collecting data for this study using mobile phones to capture time-varying dy- illuminated the methodological challenges specif- namics, though such research needs to be careful- ic to surveying transient populations in a complex ly designed because it can be difficult to manage situation. Similar challenges confront many coun- even in stable settings. tries beyond those in the EU that are dealing with large migrant populations. Given the pervasive vulnerability of the refugee population, some ethical aspects of collecting Stand-alone surveys, like this study, can provide data about them must be recognized. The follow- valuable information, particularly when compre- ing aspects addressed in this study could inform hensive information is needed on specific circum- others: taking precautions to guarantee that par- stances, here to assess the transit experience, vi- ticipation is indeed voluntary; analyzing questions 74 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making to identify any that might require respondents to mentation protocols to the population of migrants recall potentially traumatic events; and designing to maintain comparability with PIAAC test results. protocols for reporting inappropriate behavior by any party involved during the survey. Anxiety and Depression Education and Skills Training In addition to the expected evidence of the mon- etary and physical costs of migration for these About 58 percent of asylum seekers are from groups, the study findings reveal that the experi- high-recognition countries and therefore have a ence induces a severe mental health crisis: 40 per- high probability of being granted legal status in cent of the asylum seekers in Italy and 70 percent the European Union. These, and the asylum-seek- in Greece had high scores on a standard scale of ing population generally, have low educational anxiety and depression. This finding can have im- attainment, low-level cognitive skills, and often plications for policymakers, both those working language-related barriers to integration. The jobs on integration and those working on return and of those with work experience have mostly been reintegration policies. The asylum application in elementary occupations like construction, man- process provides a unique opportunity for early ufacturing, and agriculture (45 percent of those in and systematic identification and intervention, es- Italy and 37 percent in Greece). It is important to pecially for those already in asylum centers where remember that the findings are limited to those mental health support can be targeted. surveyed and cannot be generalized to, e.g., earlier waves of asylum seekers. However, it is reasonable Combating Smuggling and Trafficking to infer that a substantial proportion of asylum seekers are low-skilled, something to be consid- The persistence of migration streams and the emer- ered in designing and delivering measures to pro- gence of new routes make it clear that many still mote labor market engagement and integration. consider the benefits of migration to outweigh the Furthermore, unlike other migrants to the EU, asy- monetary, physical, and emotional costs captured lum seekers are typically concentrated in reception here. This study found, for instance, that 45 percent centers for extended periods of time while waiting of asylum seekers in Italy had been exposed to vi- for their application decisions, hence offering op- olence in transit, and that asylum seekers pay an portunities to reach them through skill upgrading average of $3,500 for transit, some even engaging programs in an efficient and cost-effective way. in unpaid work along the way to afford its cost. A unique contribution of this study was the formu- Policies directed at combating smugglers and lation, with the OECD, of a computer-based litera- human trafficking networks could help curb the cy test that gives objective information to compare flows, but they could also have unintended nega- with what is self-reported. The unique context in tive consequences: (1) They could make the jour- which the assessment was delivered – multiplicity ney even harsher and more dangerous. As mea- of native languages, difficult individual experienc- sures are being put in place to reduce flows from es, and new living environment— has contributed Libya, it appears that deaths have increased on to high refusal rates and raised the question of the the journey not only from Libya to Italy but also appropriate choice of language of administration from Morocco and Algeria to Spain (UNHCR 2018). of the test. One lesson drawn from this exercise is (2) Measures like border controls and detention the need to adapt both assessments and imple- could deter from seeking asylum those who have vi. cOncluSiOnS and inSightS intO pOlicy | 75 powerful needs for international protection—also a human cost (UNHCR 2018). Such considerations could influence how policies to combat smugglers and dismantle human trafficking networks are de- signed. Towards a Knowledge Agenda on Forced Displacement The current initiative for the World Bank–UNHCR Joint Data Center on Forced Displacement can en- sure better-quality data and promote cross-learn- ing to address both technical and nontechnical difficulties in collecting and analyzing data. The center is expected to make major contributions to the knowledge base on methodologies and norms that lead to timelier and better-quality data. Although this report helps to fill gaps in what is known about migration and forced displacement, much more needs to be done. Also useful would be additional research on (1) the impacts on host communities and their attitudes toward refugees, asylum seekers, and other migrants; (2) vulnerable groups of asylum seekers and refugees focusing on those in EU countries, particularly countries that experienced a large influx of migrants in waves previous to the current one; (3) how to capitalize on existing data collection efforts, particularly administrative, to capture more timely policy-rel- evant information; and (4) the cost-effectiveness of policies being adopted or considered, such as integration policies, measures to curb the flows in transit countries, and the roll-out of voluntary re- turn packages. 76 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making References Acemoglu, D. and D. Autor. 2011. Skills, tasks and tech- Awas, M., D. Kebede, and A. Alem. 1999. “Major Mental dis- nologies: Implications for employment and earn- orders in Butajira, Southern Ethiopia.” Acta Psychi- ings. In Handbook of labor economics 4: 1043- atrica Scandinavia (Supplement) 397: 56–64. 1171. Elsevier. Azur, M. J., E. A., Stuart, C. Frangakis, and P.J. Leaf. (2011). Adsera, A., and B. R. Chiswick. 2007. “Are There Gender “Multiple Imputation by Chained Equations: What and Country of Origin Differences in Immigrant is it and how does it work?” International Journal Labor Market Outcomes across European Coun- of Methods in Psychiatric Research. 20(1): 40–49. tries?” Journal of Population Economics 61 (2): 225 http://doi.org/10.1002/mpr.329. –30. Baingana, F., A. Dabalen, E. Menye, M. Prywes, and M. Ro- Aeberhardt, R., and J. Pouget. 2007. “National Origin sholm. 2004. “Mental Health and Socio-Economic Wage Differentials in France: Evidence from Outcomes in Burundi.” Health, Nutrition, and Pop- Matched Employer-Employee Data.” IZA Discus- ulation (HNP) discussion paper; World Bank, Wash- sion Paper No. 2779: SSRN: https://ssrn.com/ab- ington:http://documents.worldbank.org/curated/ stract=986355. en/786621468242065057/Mental-health-and-so- cio-economic-outcomes-in-Burundi. Allansson, M., E. Melander, and L. Themnér. 2017. “Orga- nized violence, 1989-2016.” Journal of Peace Re- Baker, M., and B. Dwayne. 1994. “The Performance of Im- search 54(4). migrants in the Canadian Labor Market. Journal of Labor Economics 12 (3): 369–405. Andrade L., E. E. Walters, V. Gentil, and R. Laurenti. 2002. “Prevalence of ICD-10 Mental Disorders in a Catch- Borjas, G. J. 2015. “The Slowdown in the Economic Assim- ment Area in the City of São Paulo, Brazil.” Social ilation of Immigrants: Aging and Cohort Effects Psychiatry and Psychiatric Epidemiology 37(7): Revisited Again.” Journal of Human Capital 9 (4): 316–25. 483–517. Andrews, G., S. Henderson, and W. Hall. 2001. “Prevalence, Buber-Ennser, I., J. Kohlenberger B. Rengs, Z. Al Zalak, A. Comorbidity, Disability and Service Utilisation. Goujon and E. Striessnig. 2016. “Human Capital, Overview of the Australian National Mental Health Values, and Attitudes of Persons Seeking Refuge Survey. British Journal of Psychiatry 178: 145–53. in Austria in 2015.” PLoS ONE 11 (9): e0163481. doi:10.1371/journal.pone.0163481. Antecol, H. 2000. “An examination of cross-country differ- ences in the gender gap in labor force participa- Carling, J., and F. Collins. 2017. “Aspirations, Desire and tion rates.” Labour Economics 7.4 (2000): 409-426. Drivers of Migration.” Journal of Ethnic and Migra- tion Studies 909–26. reFerenceS | 77 Clarke W., and S. Withers. 2002. “Disentangling the In- Gold, J. H. 1998. “Gender differences in psychiatric illness teraction of Migration, Mobility, and Labor-force and treatments: A critical review”. Journal of Ner- Participation.” Environment and Planning 34: 923, vous & Mental Disease. 186, 769–775. 945. Goldberg, D. P., and Y. Lecrubier. 1995. “Form and Fre- Council on Foreign Relations. 2016. Civil War in Libya. quency of Mental Disorders across Centres.” In Mental Illness in General Health Care: An Interna- Das J., Q.T. Do, J. Friedman, and D. McKenzie. 2009. “Men- tional Study (323–34). John Wiley & Sons on be- tal Health Patterns and Consequences: Results half of the World Health Organization. from Survey Data in Five Developing Countries. World Bank Economic Review 23(1): 31–55. Human Rights Council, 2015. Report of the commission of inquiry on human rights in Eritrea. Das, J., Q.T. Do, J. Friedman, D. McKenzie, and K. Scott. 2007. “Mental Health and Poverty: Revisiting Human Rights Watch, Syria: events of 2016, 2017. the Relationship”, Social Science and Medicine, International Crisis Group, Turkey’s Refugee Crisis: The Pol- 65(3):467-80, 2007. itics of Permanence, 2016. David, H. and D. Dorn. 2013. The growth of low-skill ser- IOM. 2012. “Migrants caught in crisis: the IOM experience vice jobs and the polarization of the US labor mar- in Libya.” http://publications.iom.int/system/ ket. American Economic Review, 103(5): 1553-97. files/pdf/migrationcaughtincrisis_forweb.pdf Devictor, X., and Q.T. Do. 2017. How many years have ref- ——.2015. “Report on Victims of Trafficking in Mixed Mi- ugees been in exile? Population and Development gration Flows Arriving in Italy by Sea.” http://www. Review, 43(2), 355-369. italy.iom.int/sites/default/files/documents/IOM_ Eurostat. 2018. First instance decisions on applications by Report_on_victims_of_trafficking_2015_ENG. citizenship, age and sex Annual aggregated data, pdf. updated June 14, 2017. ____.2016. “Study on Migrants’ Profiles Drivers Of Migra- ——. 2015. E. U. Trafficking in human beings. tion And Migratory Trends.” International Organi- zation for Migration: Rome. EASO. 2017. Annual Report on the Situation of Asylum in the European Union 2016, Valletta: European Asy- ——.2017. “UN Migration Agency Issues Report on Arriv- lum Support Office. als of Sexually Exploited Migrants, Chiefly from Nigeria.” https://www.iom.int/news/un-migra- _____. 2018. Latest Asylum Trends - Overview 2017, Vallet- tion-agency-issues-report-arrivals-sexuallyex- ta: European Asylum Support Office. ploited-migrants-chiefly-Nigeria. European Commission. 2015. A European Agenda On Mi- Kessler, R. C., W. T. Chiu, O. Demler, and E. E. Walters. gration. European Commission: Brussels. 2005. “Prevalence, Severity, and Comorbidity of Eurostat. 2018. Asylum quarterly report. (Data extracted 12-Month DSM-IV Disorders in the National Co- on 19 March 2018). Eurostat: Luxembourg. morbidity Survey Replication.” Archives of General Psychiatry 62(6): 617–27. Friedman, J., and D. Thomas. 2007. “Psychological Health Before, During, and After an Economic Kibreab, G. 2013. “The National Service/Warsai-Yikealo Crisis: Results from Indonesia, 1993–2000.” World Development Campaign and Forced Migration in Bank Policy Research, Working Paper No. 4386. Post-independence Eritrea.” Journal of Eastern Af- rican Studies 7 (4): 630–49. 78 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making König, R. S., and M. De Regt. 2010. “Family Dynamics in World Bank. 2016a. Poverty Reduction in Nigeria in the Transnational African Migration to Europe: An In- Last Decade. World Bank: Washington DC. troduction.” African and Black Diaspora: An Inter- ——. 2016b. “Mental Health among Displaced People and national Journal 3(1): 1–15. Refugees: Making the Case for Action at the World LaLonde, R. J., and R. H Topel. 1991. Labor market adjust- Bank Group.” Washington, D.C. Poverty Global ments to increased immigration. In Immigration, Practice. trade, and the labor market (pp. 167-199). Univer- ——. 2018. “World Developement Indicators” updated sity of Chicago Press. May 2, 2018. Mincer J. 1978. “Family Migration Decisions.” Journal of World Health Organization. 2001. The World Health Report Political Economy 86 (5): 749–73 2001: Mental Health: New Understanding, New Monsutti, A. 2008. “Afghan Migratory Strategies and the Hope. Geneva: WHO. Three Solutions to the Refugee Problem.” Refugee ——. 2013. Mental Health Action Plan: 2013–2020, Geneva: Survey Quarterly 27(1): 58–73. WHO. Patel, V., R. Araya, M. De Lima, A. Ludermir., and C. Todd. UN General Assembly, Protocol Relating to the Status of (1999). Women, poverty and common mental Refugees, 31 January 1967, United Nations, Treaty disorders in four restructuring societies. Social sci- Series, vol. 606, p. 267. ence & medicine, 49(11), 1461-1471. ——. Report of the Special Representative of the Sec- Raghunathan, T.E., J. M. Lepkowski, J. van Hoewyk, and retary-General on Migration, 3 February 2017, P. Solenberger. 2001. “A Multivariate Technique A/71/728. for Multiply Imputing Missing Values Using a Se- quence of Regression Models.” Survey Methodolo- UNHCR. 2018. Desperate Journeys. https://reliefweb. gy 27: 85–95. int/report/world/desperate-journeys-janu- ary-2017-march-2018. Stark, O., C. Helmenstein, and A. Prskawetz. 1998. “Human Capital Depletion, Human Capital Formation, and ——. 2016. Population statistics. Migration: a Blessing or a “Curse”?” Economics Let- U.S. Energy Information Administration, 2018. Spot Prices ters, 60(3): 363–67. for Crude Oil and Petroleum Products. Schuster, L., and N. Majidi. 2013. “What Happens Post-de- portation? The Experience of Deported Afghans.” Migration Studies 1(2): 221–40. Van Buuren, S., and K. Groothuis-Oudshoorn. 2011. “Mul- tivariate imputation by chained equations” in R. Journal of statistical software, 45(3). Van Buuren, S., J. P. L. Brand, C. G. M. Groothuis-Oud- shoorn, and D. B. Rubin. 2006. “Fully conditional specification in multivariate imputation”. Jour- nal of Statistical Computation and Simulation 76: 1049–1064. 79 Appendix Appendix 1: Survey sampling strategy: itrea, The Gambia, Guinea, Mali, Nigeria, Senegal, assumptions, sampling, and sample Somalia and Sudan. These nationalities represent weights 78 percent of the flow into Italy in 2016. In Greece, only three nationalities (Afghans, Iraqis, and Syri- In both Italy and Greece, the basis of the sampling ans) were included in the sampling frame. These strategy is the quasi-random allocation of asylum nonetheless represent 86 percent of the flow into seekers across reception centers throughout the Greece in 2016. country after processing by law enforcement au- thorities. Thus, in both countries, data collection The first stage of the sampling strategy entailed was carried out through a three-stage stratified sam- the selection of the reception centers in which to ple in four regions of Italy (Lazio, Lombardy, Sicily, conduct the survey in each of the regions in Italy and Puglia) and regions in Greece (Attica , Central and in Greece. In Italy, this was based on the lists of Greece, and Thessaly). Qualitative interviews were reception centers provided by the Ministry of Inter- carried out in the Lazio and Lombardy regions of nal Affairs. The selection of centers was conducted Italy only, with the addition of Central Macedonia to avoid clustering of asylum seekers in one single in Greece. The qualitative sample does not overlap location because the allocation process makes it with that of the quantitative survey, a decision made possible that individuals from the same country to avoid interview fatigue and differences in priming. and arrival time into Italy would be allocated to the same center. Furthermore, the choice of cen- The number of nationalities covered by the survey ters also aimed at ensuring some level of represen- was determined prior to field work. One reason tation of each type of centers (defined by type – was the tradeoff between coverage of the entire CARA and CAS as per Box 2.3), size and nationality, population of asylum seekers and sufficient statis- and a targeted sample size in each region. A sec- tical power to conduct country-specific analyses. ond stage created as many strata as the number Besides, the survey instrument had some nation- of countries identified for inclusion in the survey. ality-specific modules (assets and dwelling for example) that had to be included into the ques- Finally, a third stage drew a sample of interview- tionnaire beforehand. In Italy, a total of 9 nation- ees of the nationalities of interest to the survey. alities were picked for coverage. Cote d’Ivoire, Er- For the quantitative survey, these were selected 80 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making randomly from the most recent list of asylum seek- errors let to imperfectly matched proportions. Fur- ers hosted in the selected centers. The number of thermore, as illustrated in Table A1.1., some indi- respondents for each nationality was determined viduals from countries other than the 8 or 3 coun- so that the proportion of cases of a given country tries of interest still ended up being interviewed. in the survey was similar to the actual proportion Reasons include misreporting or absence of citi- as calculated using administrative data. Sampling zenship information in administrative listings. table a1.1. countries of Origin for respondents to eaSS 2017 Country Country code Total Male Female Syria SYR 976 626 350 Nigeria NGA 622 488 134 Afghanistan AFG 502 351 151 Gambia GMB 340 335 5 Eritrean ERI 327 220 107 Senegal SEN 319 318 1 Mali MLI 285 282 3 Côte d’Ivoire CIV 211 186 25 Guinea GIN 183 176 7 Iraq IRQ 173 105 68 Somalia SOM 84 65 19 Sudan SDN 32 30 2 Guinea-Bissau GNB 16 16 Iran IRN 12 9 3 Cameroon CMR 4 4 Niger NER 4 3 1 Sierra Leone SLE 4 4 Ethiopia ETH 3 2 1 Chad TCD 2 2 Lebanon LBN 2 2 Morocco MAR 2 1 1 South Sudan SSD 2 2 Burkina Faso BFA 1 1 Congo COD 1 1 Egypt EGY 1 1 Gabon GAB 1 1 Ghana GHA 1 1 Jordan JOR 1 1 1 Liberia LBR 1 1 Libya LBY 1 1 Unknown 10 5 5 Total 4,124 3,240 884 appendix | 81 Thus, the statistical analysis using quantitative data In Greece, a similar stratified random sample to re-weighs asylum seekers from countries that are un- Italy’s was drawn. Focusing on the three target der-represented in the survey to correct the propor- origin countries, i.e. Afghanistan, Iraq and Syria, tions. All statistics presented in this report are hence- sample weights were calculated by taking the forth restricted to the 8 target nationalities and are inverse of the survey ratio (see Table A1.3). computed using these sampling weights. Regression analyses however are not weighted and use all avail- able observations in the survey. For the Italian sam- table a1.3 Sample weights as the inverse of the ple, weights are designed by comparing the shares greece Survey ratio of asylum seekers’ nationalities in EASS 2017 with a Country Share in Share Greece reference population. The reference population was administra- in the Survey Ratio obtained from monthly EUROSTAT data from the Eu- tive data, survey, (survey/ad- Percent Percent min) ropean Commission, on First Time Applicants from 2012 until 2017, for Sea Arrivals to Italy from 2013 until Afghani- 35.73 31.36 .87 stan June 2017. The period studied is 2013-2017 for both Iraq 14.07 10.43 .74 the EASS 2017 sample and the reference population. Syrian 50.18 58.19 1.15 Arab Re- Since the analysis is restricted to the 9 pre-chosen public origin countries, i.e. Cote d’Ivoire, Eritrea, Guinea, Total 100 100 The Gambia, Mali, Nigeria, Senegal, Somalia and Sudan, sample weights are designed to ensure representativeness of the distribution within this sample of 9 countries and are thus the inverse of As for the case of Greece, summary statistics are the survey ratio as calculated in Table A1.2. restricted to the three countries of focus and were weighted using these sample weights. Regressions table a1.2 Sample weights as the inverse of the will however be unweighted and use the entire italy Survey ratio available sample. Share in Share in Italy Survey administra- survey Ratio (sur- On the other hand, the qualitative instrument was Country tive data, data, vey/admin- not designed to be statistically representative of Percent Percent istrative) the population of asylum seekers. Rather the qual- Côte .065 0.085 1.30 itative instrument was designed to account for d’Ivoire any social relations between interviewees and to Eritrea .265 0.172 0.65 promote an appropriate depth of responses. Thus, Guinea .065 0.072 1.10 a mixed approach was adopted, which consisted The .092 0.129 1.40 of selecting interviewees using a combination of Gambia random selection from center lists, chain referral Mali .081 0.108 1.33 sampling and purposive sampling by interviewers. Nigeria .219 0.271 1.24 Finally, while the random selection of individu- Senegal .070 0.120 1.71 als in the quantitative instrument would lead to Somalia .081 0.031 0.42 a gender mix in the survey that reflects the pop- Sudan .066 0.010 0.15 ulation mix, the qualitative instrument purposely Total 100 100 . over-sampled women. 82 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Appendix 2: Survey instrument 2.2. Qualitative instruments 2.1. Quantitative instrument Two instruments were used in the collection of qualitative data: one for in-depth individual inter- The quantitative survey instrument had several in- views (IDIs) and another for focus group discus- novative features built-in. sions (FGDs). FGDs allowed analysis of group dy- namics and provided additional contextual data Language: the survey instrument was translated to IDIs. into several languages (French, English, Bambara, Tigrinya, Arabic, and Farsi) to accommodate the The instruments covered similar themes, and were target population. designed to complement those of the primary in- strument. Additional focus was put on gender spe- Nationality-specific modules: to allow compari- cific issues, including violence encountered during son with the population in the country of origin, the journey, and gender dynamics throughout the the asset and dwelling modules were identical migration process. to those used in the most recent household sur- vey available in said country. For example, the Both instruments were semi-structured: open-end- asset and dwelling module for asylum seekers ed questions were posed, and field teams were from Nigeria was identical to the one used for trained in probing responses as relevant. the 2011 Nigerian Living Standards Measure- ment Survey. In addition to FGDs and IDIs, knowledgeable staff was interviewed within each center. They provided Standardized modules: on the other hand, details on the population hosted and other con- some modules were designed to allow com- textual information relevant to the logistics or in- parisons across asylum seekers and between terpretation of the study. asylum seekers and the world population. The education and job modules were for example Appendix 3: Data construction and data borrowed from World Bank STEP background validation questionnaires. 3.1. Data and Methodology for Constructing Context-specific modules: the migration module Welfare Aggregates was specifically designed for this specific project. The questions asked were determined after thor- The comparative analysis in this section uses the ough qualitative work conducted during the sum- EASS 2017 collected for this report, in combination mer of 2016. with the Nigeria 2011 Living Standard Measure- ment Survey; the Senegal 2011 Enquête de Suivi The survey instrument was programmed into tab- de la Pauvreté (ESPS II); and the 2013-14 Afghani- lets using the World Bank’s Computer-Assisted stan Living Conditions Survey. Personal Interview (CAPI) software.43 Given lack of information in EASS 2017 about asylum seekers’ households living standards 43 https://designer.mysurvey.solutions/account/login?Retur- before migration, survey-to-survey imputation nUrl=%2f techniques were used to, based on available wel- appendix | 83 fare aggregates in surveys of countries of origin, stroyed dwelling. Driven by this, the data captured estimate the level of welfare of asylum seekers’ included substantial missing information on those households (levels of expenditures for Nigeria and variables initially included to be relevant to the im- Senegal, and a wealth index for Afghanistan, in line putation. In some cases, the amount of observa- with information available in surveys of countries tions lost, makes traditional imputation methods of origin). impractical. The EASS 2017 was designed so that it included In some instances, the amount of missing infor- questions on socio-demographic characteristics mation is almost 50 percent (Table A3.1). This is at the individual and household levels derived most salient among questions which tried to as- from, and which are comparable to, these na- sess the quality of the respondents dwelling back tional surveys. Strategic questions were placed home, and employment. Lack of information re- in the asylum seeker’s survey which would allow garding the respondent’s location is also absent for survey to survey simulation of welfare vectors. for numerous observations. Because the goal is These questions aimed at collecting information to simulate welfare for asylum seekers, the miss- on dwelling characteristics, among others, as a ing information for these correlates can severe- key predictor of household welfare. Senegal and ly limit the total number of viable observations Nigeria, have surveys which allow for simulation in each country. Consequently, the imputation of the household’s expenditure, while Afghani- procedures rely on fully conditional specifica- stan’s survey allows for simulation of a wealth tions, also known as multivariate imputation by index. The analysis uses the Nigeria 2011 Living chained equations (MICE). Standard Measurement Survey; the Senegal 2011 Enquête de Suivi de la Pauvreté (ESPS II); and the table a3.1: number of missing observations per 2013-14 Afghanistan Living Conditions Survey. variable by country Multiple imputation techniques were used to Nige- Sene- Afghani- simulate welfare vectors on to the asylum seek- ria gal stan er’s survey. Household size 0 0 0 Share of members 248 140 209 under 15 The goal of survey-to-survey imputation tech- niques is to obtain the joint distribution of the vec- Share of members 248 140 209 [15, 65) tor of interest, expenditure in the case of Senegal Share of members 65+ 248 140 209 and Nigeria, and a set of correlates. The set of cor- relates for Nigeria and Senegal include variables Share of members 0 0 71 female on the household size, share of educated, share of Child dependency ratio 248 140 209 employed members, region, and only for Senegal Total dependency ratio 248 140 209 the number of rooms. The wealth index model for Afghanistan incorporated, in addition, variables Share employed 267 151 77 on dwelling characteristics on flooring and toilet. Share with at least 69 126 108 secondary education The dwelling characteristics included in the in- Improved toilet 150 125 243 strument could not be fully used in the imputation across all countries. This because, for ethical rea- Improved floor 172 #N/A 293 sons, the survey did not ask detailed information Improved roof 288 #N/A #N/A to individuals who reported loss of assets and de- Total observations 583 287 503 84 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Multivariate imputation by chained equations table a3.2: expenditure OlS models for Senegal (MICE), provides a method to tackle the issue of (nat. log of adult equivalized expenditure) and nigeria (nat. log of per capita expenditure) missing information on multiple vectors. MICE specifies multivariate imputation models iterative- Senegal Nigeria ly for each variable with missing information. The Share household -0.00470*** -0.00292*** members over 65 (0.00108) (0.000475) procedure operates under the assumption that Nat log. household -0.516*** -0.467*** observed information is sufficient to inform the size (0.0220) (0.0169) missingness of the specific observations, and that Share of household -0.00262*** any remaining missingness is completely random members under 15 (0.000434) (Azur et al. 2011). Share of adults em- -0.00235*** 0.000782** ployed (0.000502) (0.000337) In the MICE procedure, each variable is modeled Share of adults with 0.00848*** 0.00541*** conditional on other variables in the data.44 Im- secondary education (0.000734) (0.000223) puted values for each variable are drawn from its Dwelling has less than -0.186*** modeled distribution conditional upon other vari- 5 rooms (0.0232) ables in the data for each iteration : 14.44*** 5.740*** Constant (0.0714) (0.0430) Observations 5,891 4,907 Adjusted R-squared 0.523 0.489 Robust standard errors in parentheses ***, **, * indicate statistical significance at the 1, 5, and 10 percent level, respectively; region dummies omitted example employment),45 and thus sets all other where gi could be any type of conditional dis- variables with missing observations equal to its tribution specified (logistic, linear, Poisson, etc), mean value. This yields a simulated vector of the and are the model parameters (StataCorp, 2015). first variable which is to be used for all subse- As an example, assume we have 3 variables which quent variables’ modeling. are missing; 1) whether anyone in the asylum seeker’s household was employed, 2) the region The procedure is reminiscent to the Gibbs sam- of residence of the household, and 3) the welfare. pling algorithm, which is a Markov Chain Monte All three models can be specified differently, the Carlo (MCMC) method. In similar fashion to MCMC first as a logit, the second as a multinomial logit, methods. Despite the attractiveness of MICE, it and the third as an ordinary least squares model. may not converge to any particular distribution, The variable of interest is welfare, and thus the which has led to concerns regarding the method’s other two are used as correlates for its specifica- statistical validity.46 Nevertheless, the method is tion. The process begins with the variable that has the least number of missing observations (for This is not necessarily always the case. 45 44 For a simple and easy to follow description of MICE refer to For a thorough discussion on the concerns regarding the 46 Azur, Stuart, Frangakis, and Leaf, 2011. methodology readers should refer to van Bureen et al. (2006) appendix | 85 widely used and has proven quite useful under the most asylum seekers are in the upper parts of current circumstances. their country’s welfare distribution. This result is robust to multiple specifications and MICE as- The specific modelling choices for each country sumptions. are different, given the circumstances of missing information for each of the models, however the 3.2. Misreporting model used for expenditure in Nigeria and Sene- gal is presented in Table A3.2, and for Afghanistan Interviews only proceeded with the explicit and in Table A3.3.47 All expenditure and wealth index informed oral consent of the interviewee. Inter- models yield adjusted R squared values close to or viewees were given information about this study, above 0.5. and told that their personal information would not be passed on to authorities processing asylum table a3.3: wealth index OlS model for afghan- requests. In both quantitative and qualitative sur- istan veys, consent was obtained orally. VARIABLES AFG wealth Share household members over 65 -0.00121** As in any context, and potentially even more so Dwelling has Concrete/wood floors 1.325*** in the case of asylum seekers, responses can be Dwelling has flush toilet 0.769*** subject to conformity bias especially when the Share of household members survey is believed to have an effect one’s asylum -0.00478*** under 15 application status. While interviewers reminded Share of adults employed 0.00356*** interviewees of the independence of the project Share of adults with secondary in multiple occasions, we believe that questions 0.00815*** education pertaining to questions on triggers of migration, vi- Nat.log household size 0.0875*** olence, and anxiety and depression should receive Constant -0.530*** particular attention. Observations 19,667 Adjusted R-squared 0.633 As can be seen in Figure A3.1, an implausibly Robust standard errors in parentheses high proportion of individuals from Cote d’Ivo- ***, **, * indicate statistical significance at the 1, 5, and 10 ire, Guinea, Mali, and Senegal (labelled Group 1) percent level, respectively; region dummies omitted reports having at least one deceased household member. These numbers are particularly striking when comparing with percentages for other Af- The average welfare for Nigerian, as well as Sene- rican countries (Group 2, defined as Eritrea, The galese asylum seekers is considerably above that Gambia, Nigeria, Somalia, Sudan, alongside with of the average individual back home during the non-targeted countries)48, or even conflict-zone survey’s period. For Afghanistan, asylum seekers countries such as Afghanistan, Iraq or Syria. As are 0.55 standard deviations above the mean one cannot exclude the possibility of an over-re- wealth index for the country. Correspondingly, porting bias among the four African countries, we 48 Out of the 1431 individuals in Group 2, 98.25 percent are from 47 Models for all imputation steps are available upon request. Eritrea, The Gambia, Nigeria, Somalia, Sudan and 1.75 percent For most intermediate imputations, we use predictive mean from non-targeted countries, e.g. Cameroon, Sierra Leone or matching techniques with 10 nearest neighbors. Ethiopia. 86 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Figure A3.1. Possible over-reporting of death in household among selected countries Proportions of migrants who report having lost a family member, having had their dwelling destroyed, or undergone violence during transit, breakdown by country of origin. 80 70 60 50 40 30 20 10 0 CIV ERI GIN GMB MLI NGA SDN SEN SOM AFG IRQ SYR Italy Greece Death of a household member Destroyed dwelling Experienced violence in transit Source: EASS 2017. Statistics for Sudan are computed from 33 observations and should thus be interpreted with caution. test for correlation between reporting a death in There is little evidence of correlation between the household and other distressing event such over-reporting of death in the household and as the destruction of assets (dwelling), having over-reporting of distressing events during transit. received violence during transit, having been de- Being imprisoned during transit is not correlated tained in prison, and the overall GSI. with reporting a death in the household, through- out all samples (Group 1, Group 2, and both alto- There seems to be a positive correlation between gether). When looking at the correlation with re- reported death in the household and reported porting violence in transit, the coefficient is either destruction of dwellings for countries in Group non-significant or slightly negative (when includ- 1, contrary to other African countries of Group 2, ing center or country fixed effect), which contra- which do not exhibit any correlation. Individuals dicts the idea that migrants over-reported all type in Group 1 are more likely to say that their dwell- of distressing events. ing has been destroyed (in their original location) when they have reported a death amongst their There is little difference among Group 1 and 2 re- household members. This correlation however garding the way anxiety/depression symptoms does not hold for countries in Group 2. As there and death of household members are reported. are no obvious reason to explain why the death Low GSI levels are correlated with the death of of a member would be correlated with destroyed household member when pooling the Italian sam- asset in Group 1 but not Group 2, one should ple altogether, regardless of whether center and keep this potential bias in mind when analyzing country fixed effects are used. When decomposing outcomes that are affected by these two distress- by groups, the correlation disappears as soon as ing events. fixed effects are included, for both Group 1 and 2. appendix | 87 table a3.4. reported destroyed dwelling and death of a household Member, italy Sample Dwelling in past residence currently destroyed Group 1 and 2 Group 1 Group 2 variaBlES (1) (2) (3) (4) (5) (6) (7) (8) (9) Death in household 0.387*** 0.366*** 0.350*** 0.452*** 0.371*** 0.385*** 0.0160 -0.112 -0.135 (0.0929) (0.0859) (0.0950) (0.0812) (0.102) (0.1000) (0.180) (0.247) (0.276) Constant 0.186*** -0 -0.0719 0.144*** -0 0.0288 0.206*** -0 0.137** (0.0432) (0) (0.0646) (0.0351) (0) (0.0913) (0.0564) (0) (0.0602) Observations 864 864 864 370 370 370 494 494 494 R-squared 0.114 0.266 0.296 0.223 0.410 0.413 0.000 0.257 0.314 Center fixed effects Yes Yes Yes Yes Yes Yes Country fixed effects Yes Yes Yes Source: Authors’ calculation using EASS 2017. Note: Robust standard errors in parentheses, ***, **, * indicate statistical significance at the 1, 5, and 10 percent level, respectively. table a3.5. reported violence in transit and death of a household Member, italy Sample Has experienced violence during transit Group 1 and 2 Group 1 Group 2 variaBlES (1) (2) (3) (4) (5) (6) (7) (8) (9) Death in household -0.147* -0.102 -0.0270 -0.134** 0.0461 0.0386 -0.141 -0.0890 -0.0701 (0.0801) (0.0696) (0.0486) (0.0649) (0.0561) (0.0589) (0.0965) (0.0947) (0.101) Constant 0.488*** -0 0.389*** 0.473*** 1*** -0 0.494*** 1*** 0.649*** (0.0239) (0) (0.114) (0.0618) (0) (0) (0.0318) (0) (0.0291) Observations 1,326 1,326 1,326 520 520 520 806 806 806 R-squared 0.011 0.129 0.158 0.016 0.329 0.330 0.002 0.127 0.140 Center fixed effects Yes Yes Yes Yes Yes Yes Country fixed effects Yes Yes Yes Source: Authors’ calculation using EASS 2017. Note: Robust standard errors in parentheses, ***, **, * indicate statistical significance at the 1, 5, and 10 percent level, respectively. table a3.6. reported imprisonment during transit and death of a household Member, italy Sample Has been imprisoned during transit Group 1 and 2 Group 1 Group 2 variaBlES (1) (2) (3) (4) (5) (6) (7) (8) (9) Death in household 0.0874 0.104 0.0650 0.00663 0.0430 0.0496 0.0873 0.0699 0.134 (0.0705) (0.0688) (0.0673) (0.0712) (0.0813) (0.0827) (0.0950) (0.0948) (0.0955) Constant 0.324*** -0 -0.423*** 0.409*** 0 0.0236 0.288*** -0 -0.687*** (0.0200) (0) (0.0563) (0.0353) (0) (0.0810) (0.0242) (0) (0.121) Observations 1,340 1,340 1,340 521 521 521 819 819 819 R-squared 0.004 0.083 0.121 0.000 0.144 0.149 0.001 0.089 0.141 Center fixed effects Yes Yes Yes Yes Yes Yes Country fixed effects Yes Yes Yes Source: Authors’ calculation using EASS 2017. Note: Robust standard errors in parentheses, ***, **, * indicate statistical significance at the 1, 5, and 10 percent level, respectively. 88 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making table a3.7. anxiety/depression and death of a household Member, italy Sample gSi Group 1 and 2 Group 1 Group 2 variaBlES (1) (2) (3) (4) (5) (6) (7) (8) (9) Death in household 0.446*** 0.342*** 0.246** 0.370*** 0.0842 0.0950 0.343* 0.262 0.248 (0.0960) (0.111) (0.0921) (0.115) (0.108) (0.106) (0.176) (0.204) (0.276) Constant 1.972*** 2.167*** 1.757*** 2.065*** 2*** 2.167*** 1.932*** 3*** 2.105*** (0.0447) (0) (0.0906) (0.0804) (0) (0) (0.0471) (0) (0.330) Observations 1,389 1,389 1,389 536 536 536 853 853 853 R-squared 0.045 0.120 0.168 0.072 0.312 0.324 0.005 0.118 0.157 Center fixed effects Yes Yes Yes Yes Yes Yes Country fixed effects Yes Yes Yes Source: Authors’ calculation using EASS 2017. Note: Robust standard errors in parentheses, ***, **, * indicate statistical significance at the 1, 5, and 10 percent level, respectively. Consequently, while the answers on migration trig- of authorization code, or missing observation gers might be subject to over-reporting, the other on the literacy score). These observations were questions do not seem to have been affected. dropped from the analysis. Appendix 4: Literacy assessment Furthermore, the tests conducted in Lazio were dropped from the sample due to some oversight 4.1. Data and methodology in the administration of the assessment. The im- plementation of the literacy module of E&S online In Italy, a total of 281 individuals ended up being required manually skipping the numeracy mod- assigned a literacy proficiency level. E&S online ule. When administered in centers in Lazio, both was implemented in Canadian French and Irish numeracy and core modules were skipped, leav- English; 450 tests were completed. Attrition cate- ing no information on proficiency. gories are shown in Table A4.1. A total of 94 cases stated they were illiterate and hence did not take Finally, 20 refused to take the test. The analysis the test. For these, decision was made to take their of the characteristics of those who refused is dis- statement at face value, and a proficiency level be- played in Table A4.2 below, where regressions are low one was assigned. run on a sample of 281 individuals (202 test-takers, 59 illiterate individuals and 20 refusals). Table A4.2 table a4.1. attrition in italy sample indicates no clear patterns of selection into refus- Uptake Frequency Percent al. The analysis is now based on 202 test-takers Participated 450 78.5 and 59 illiterate individuals. Illiterate 94 16.4 Refuse 29 5.1 Four hundred and thirty-eight observations on lit- Total 573 100 eracy proficiency are available in Greece. A literacy Source: Authors’ calculations using EASS 2017 – Literacy. test specifically designed for this analysis was imple- mented. The test used a subset of items used for E&S Among the 450 individuals who participated, 28 online, usually the ones with lower levels of difficul- tests were unusable (due to repeated Id’s, absence ty. The items were then translated into Arabic and appendix | 89 table a4.2. From the attrition analysis, no clear patterns stand out amongst those who refused the test. Linear probability estimation of refusing to take the test variaBlES (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Secondary -0.0453 -0.0547 -0.0539 -0.0534 -0.0528 -0.0583 -0.0559 -0.0701** -0.0534 -0.0592* education (0.0358) (0.0356) (0.0359) (0.0359) (0.0360) (0.0362) (0.0365) (0.0319) (0.0377) (0.0331) Mental index 0.0418* 0.0394 0.0389 0.0386 0.0379 0.0359 0.00575 0.0306 -0.00465 (0.0231) (0.0241) (0.0241) (0.0242) (0.0241) (0.0244) (0.0209) (0.0268) (0.0231) Has worked -0.0164 -0.0183 -0.0162 -0.0149 -0.0161 -0.0295 -0.0137 -0.0134 (0.0324) (0.0325) (0.0328) (0.0328) (0.0329) (0.0314) (0.0352) (0.0339) Male 0.0748 0.0742 0.0766 0.0762 0.0383 0.0691 0.0179 (0.0991) (0.0994) (0.0992) (0.0994) (0.0877) (0.101) (0.0895) Age -0.00160 0.0328 0.0327 0.0279 0.0371 0.0330 (0.00321) (0.0261) (0.0261) (0.0240) (0.0265) (0.0244) Age square -0.000649 -0.000661 -0.000628 -0.000734 -0.000702 (0.000487) (0.000488) (0.000449) (0.000492) (0.000453) Single -0.0303 -0.0624 -0.0331 -0.0567 (0.0468) (0.0440) (0.0477) (0.0448) Constant 0.0874 0.00317 0.0177 -0.0530 -0.0153 -0.448 -0.409 -0.179 -0.414 -0.198 (0.0550) (0.123) (0.147) (0.277) (0.344) (0.926) (0.949) (0.345) (0.971) (0.357) Observations 278 278 277 277 277 277 277 277 277 277 Center fixed NO NO NO NO NO NO NO YES NO YES effects Country fixed NO NO NO NO NO NO NO NO YES YES effects Source: Authors’ calculation using EASS 2017. Note: Robust standard errors in parentheses, ***, **, * indicate statistical significance at the 1, 5, and 10 percent level, respective- ly. Errors clustered at the center levels, with a parametric Moulton (1986) correction factor. Farsi. The test is a computer-based that was admin- people offered to take the test refused, while oth- istered after the survey, possibly days after. A small ers indicated as justification either inability to read background questionnaire was filled in at the same or read Arabic specifically or computer-illiteracy. time to gather information on the test taker so to al- table a4.3 attrition in greece sample low matching with the rest of the data. In addition, to Frequency Percent avoid loss of information in case such matching was Participated 376 71 not possible, information on age, gender, education, Refused 79 15 and nationality was also collected. Illiterate 62 12 Language (Kurd) 9 2 An initial 526 individuals in Greece were in sam- pled to take the literacy tests. The attrition rates Total 526 100 Source: authors’ calculations using EaSS 2017-literacy. shown in Table A4.3 indicate that 15 percent of 90 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Table A4.4 looks at whether refusal to participate When the reason “Illiterate” was invoked, the in- is related to any observable characteristics. Al- dividual was assigned a proficiency level “below though a self-selection of less educated into one”, which in analyses is given value 0. Finally, refusals appears at first, this effect seems driv- when the language of the test (Arabic) was not en by the variation of education at the country appropriate for some Iraqis who spoke only Kurd, level. Indeed, the effect of secondary education these observations were simply dropped from the becomes insignificant once controlling for na- analysis. Furthermore, we drop individuals from tionality and/or center fixed effects (keeping in Syrian and Iraq who took the test, but could not mind that nationalities were clustered in some perform as they declared that they mother tongue centers). was Kurdish (37 individuals). Literacy proficiency table a4.4. attrition analysis for refusals in greece Linear probability estimation of refusing to take the test VARIABLES (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Secondary -0.0925*** -0.0928*** -0.0857** -0.0788** -0.0730** -0.0715** -0.0698** -0.0608* -0.0540 -0.0547 education (0.0336) (0.0336) (0.0343) (0.0345) (0.0345) (0.0344) (0.0345) (0.0325) (0.0353) (0.0340) Mental index -0.0259 -0.0215 -0.0244 -0.0300 -0.0306 -0.0311 -0.0270 -0.0342 -0.0268 (0.0227) (0.0231) (0.0231) (0.0233) (0.0232) (0.0232) (0.0199) (0.0217) (0.0202) Has worked 0.0383 0.0590 0.0541 0.0563 0.0553 0.0351 0.0535 0.0397 (0.0364) (0.0389) (0.0389) (0.0388) (0.0388) (0.0344) (0.0377) (0.0353) Male -0.0628 -0.0581 -0.0591 -0.0580 -0.0283 -0.0543 -0.0319 (0.0384) (0.0383) (0.0382) (0.0383) (0.0366) (0.0386) (0.0379) Age 0.00326** -0.00954 -0.0121 -0.0110 -0.0108 -0.0110 (0.00151) (0.00819) (0.00905) (0.00861) (0.00898) (0.00874) Age square 0.000170 0.000199* 0.000184* 0.000182 0.000186* (0.000106) (0.000115) (0.000110) (0.000114) (0.000111) Single -0.0275 -0.0422 -0.0211 -0.0410 (0.0436) (0.0393) (0.0423) (0.0401) Constant 0.183*** 0.251** 0.218* 0.258* 0.161 0.379 0.439 0.269 0.366 0.211 (0.0401) (0.109) (0.123) (0.132) (0.160) (0.317) (0.362) (0.259) (0.670) (0.410) Observations 516 516 513 513 513 513 513 513 509 509 Center fixed NO NO NO NO NO NO NO YES NO YES effects Country fixed NO NO NO NO NO NO NO NO YES YES effects Source: Authors’ calculation using EASS 2017. Note: Robust standard errors in parentheses, ***, **, * indicate statistical significance at the 1, 5, and 10 percent level, respec- tively. Errors clustered at the center levels, with a parametric Moulton (1986) correction factor. appendix | 91 levels are therefore available for total of 401 indi- table a4.5. Shares of education levels for the viduals in Greece. italian sample, comparing the whole italian and the 281 individuals who were offered the literacy test 4.2. Sampling weights in literacy assessment Educa- Share Share in the Literacy Test tion in the literacy test, Ratio (test/ Sampling weights are used to correct both country survey, Percent survey) and education sampling. Throughout the report, Percent sampling weights have been used to correct the No educa- 36.29 30.94 0.85 proportion of origin countries, so that they would tion match the actual proportion as calculated using Primary 34.20 34.89 1.02 administrative data. For the Italian sample, the Second- 24.58 31.65 1.29 literacy test combined another sort of selection, ary as individuals with higher level of education were Tertiary 4.93 2.52 0.51 specifically targeted. Thus, there is a need to cor- Total 100 100 . rect for such a selection, and the final weights used in the literacy section interact education weights with previous sampling weights (origin countries) in order to ensure representativeness. Weights used in the literacy section are generated as the inverse of the Italy Survey Ratio interacted with the inverse of the Literacy Test Ratio (last col- umns of Tables A1.2 and A4.5). 92 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making Appendix 5: Anxiety and Depression table a5.1. risk factors of anxiety and depression– Full sample (1) (2) (3) (4) Variable GSI GSI GSI GSI Age -0.017 -0.001 -0.005 0.002 (0.014) (0.013) (0.013) (0.013) Age squared 0.000** 0.000 0.000 0.000 (0.000) (0.000) (0.000) (0.000) Female dummy 0.217*** 0.243*** 0.240*** 0.257*** (0.058) (0.073) (0.075) (0.070) Marital status 0.034 -0.027 -0.023 -0.030 (0.055) (0.048) (0.043) (0.045) Education -0.066 -0.012 -0.042 0.017 (0.046) (0.043) (0.052) (0.046) Employed before journey -0.112** -0.072 -0.108*** -0.069* (0.048) (0.044) (0.038) (0.041) Destroyed dwelling 0.191*** 0.217*** 0.260*** 0.250*** (0.048) (0.058) (0.079) (0.066) Death HH member 0.151** 0.108 0.088 0.046 (0.076) (0.072) (0.074) (0.063) Violence during transit 0.124** 0.153* 0.191*** 0.163* (0.057) (0.082) (0.070) (0.084) Imprison. in transit -0.002 0.052 -0.029 0.016 (0.058) (0.066) (0.070) (0.072) Travel not alone 0.090* 0.066 0.061 0.068 (0.055) (0.053) (0.048) (0.054) Duration in center -0.000 0.000 -0.000 -0.000 (0.000) (0.000) (0.000) (0.000) Living conditions 0.070*** 0.043** 0.048** 0.048** (0.013) (0.020) (0.021) (0.022) Alone in center -0.153** 0.147* 0.088 0.152* (0.061) (0.074) (0.079) (0.079) Constant 2.389*** 1.838*** 2.503*** 1.940*** (0.229) (0.237) (0.218) (0.225) Adjusted R-squared 0.1411 0.2453 0.1865 0.2504 Observations 1229 1227 1227 1227 Center fixed effects NO YES NO YES Country fixed effects NO NO YES YES Source: Authors’ calculation using EASS 2017. Note: Robust standard errors clustered at the center level in parentheses; ***, **, * indicate statistical significance at the 1, 5, and 10 percent level, respectively. appendix | 93 table a5.2. risk factors of anxiety and depression – italy sample (1) (2) (3) (4) Variable GSI GSI GSI GSI Age 0.003 0.007 0.009 0.002 (0.028) (0.020) (0.022) (0.013) Age squared -0.000 -0.000 -0.000 0.000 (0.000) (0.000) (0.000) (0.000) Female dummy 0.071 0.176 0.128 0.257*** (0.087) (0.114) (0.098) (0.070) Marital status 0.033 0.022 0.010 -0.030 (0.064) (0.065) (0.054) (0.045) Education -0.078 -0.008 -0.034 0.017 (0.054) (0.053) (0.071) (0.046) Employed before journey -0.087 -0.085 -0.073 -0.069* (0.061) (0.065) (0.064) (0.041) Destroyed dwelling 0.168*** 0.222*** 0.195* 0.250*** (0.064) (0.080) (0.103) (0.066) Death HH member 0.237*** 0.131 0.163 0.046 (0.082) (0.086) (0.098) (0.063) Violence during transit 0.198*** 0.153* 0.203*** 0.163* (0.059) (0.089) (0.074) (0.084) Imprison. in transit -0.063 -0.002 -0.112 0.016 (0.062) (0.072) (0.072) (0.072) Travel not alone 0.059 0.109* 0.062 0.068 (0.077) (0.063) (0.069) (0.054) Duration in center -0.000* -0.000 -0.000** -0.000 (0.000) (0.000) (0.000) (0.000) Living conditions 0.069*** 0.067** 0.074** 0.048** (0.017) (0.027) (0.031) (0.022) Alone in center 0.245** 0.284** 0.228** 0.152* (0.106) (0.134 (0.103) (0.079) Constant 1.781*** 0.576** 1.701*** 1.940*** (0.405) (0.268) (0.374) (0.225) Adjusted R-squared 0.0683 0.1621 0.0858 0.2504 Observations 733 732 732 1227 Center fixed effects NO YES NO YES Country fixed effects NO NO YES YES Source: Authors’ calculation using EASS 2017. Note: Robust standard errors clustered at the center level in parentheses; ***, **, * indicate statistical significance at the 1, 5, and 10 percent level, respectively. 94 | ASYLUM SEEKERS IN THE EUROPEAN UNION: Building EvidEncE to inform Policy making table a5.3. risk factors of anxiety and depression – greece sample (1) (2) (3) (4) Variable GSI GSI GSI GSI Age -0.005 0.007 -0.001 0.007 (0.021) (0.017) (0.018) (0.017) Age squared 0.000 0.000 0.000 0.000 (0.000) (0.000) (0.000) (0.000) Female dummy 0.314*** 0.307*** 0.322*** 0.303*** (0.079) (0.076) (0.083) (0.078) Marital status -0.024 -0.107 -0.089 -0.111 (0.099) (0.077) (0.081) (0.077) Education -0.082 -0.002 -0.039 0.005 (0.080) (0.073) (0.073) (0.072) Employed before journey -0.109 -0.021 -0.083* -0.019 (0.077) (0.046) (0.044) (0.049) Destroyed dwelling 0.101 0.212** 0.323*** 0.254*** (0.072) (0.076) (0.086) (0.090) Death HH member -0.112 -0.059 -0.137 -0.069 (0.169) (0.122) (0.134) (0.126) Violence during transit 0.354** 0.293 0.344** 0.279 (0.168) (0.177) (0.144) (0.185) Imprison. in transit 0.269** 0.251*** 0.258** 0.253*** (0.133) (0.088) (0.093) (0.087) Travel not alone 0.047 0.028 0.065 0.032 (0.080) (0.070) (0.060) (0.069) Duration in center 0.000 0.000 0.000 0.000 (0.000) (0.000) (0.000) (0.000) Living conditions 0.016 0.012 0.012 0.013 (0.021) (0.029) (0.025) (0.029) Alone in center -0.017 0.067 0.041 0.081 (0.099) (0.102) (0.112) (0.109) Constant 2.289*** 1.638*** 2.361*** 1.766*** (0.372) (0.308) (0.344) (0.303) Adjusted R-squared 0.0808 0.1376 0.1047 0.1370 Observations 496 495 495 495 Center fixed effects NO YES NO YES Country fixed effects NO NO YES YES Source: Authors’ calculation using EASS 2017. Note: Robust standard errors clustered at the center level in parentheses; ***, **, * indicate statistical significance at the 1, 5, and 10 percent level, respectively.