1) The document analyzes data from surveys of Syrian refugees in Turkey to understand their choices regarding returning to Syria, staying in Turkey, or migrating elsewhere. 2) It finds that as the duration of living as a refugee increases, the probability of returning to Syria decreases significantly, while the likelihood of migrating to another country increases. 3) Refugees who experienced greater damage, losses, or deaths due to the war in Syria are less likely to return and more likely to migrate internationally in search of asylum.

1. Making Hard Choices
An Analysis of Settlement Choices and Willingness to Return
of Syrian Refugees
Mehmet Balcılara
and Fuat Oktayb
IDRC Davos 2016, 28 August - 1 September, Davos, Swittzerland
a
Eastern Mediterranean University,Famagusta, T.R. Northern Cyprus; University of Pretoria, Preto-
ria, South Africa; IPAG Business School, Paris, France
b
Repuclic of Turkey, Prime Ministry, and Disaster and Emergency Management Authority (former),
Ankara, Turkey

2. Outline
1. Background
2. Hard Choices for Refugees and Politicians
3. Motivation
4. Contribution
5. Population and Sampling Design
6. Models
7. Demographics & War Impacts
Demographics
War Impacts
8. Empirical Results
9. Conclusion
1

3. Background

4. Forced Migration
Largest Crises since WW II
• highest levels of forced displacement since World War II
• more than 60 million people are forcibly displaced
• Reasons: violence, war and civil war, human rights violations, and
genocide
• Top Five Refugee Outflow Countries: Syria, Afghanistan, Somalia,
Sudan, and South Sudan
• 4.84 million Syrians are registered as refugees outside Syria (July
2016).
• a toll on millions of lives
• around half million died, and
• more than half of the country’s population is displaced
2

5. Syrian Refugee Crises
• Millions refuged to Turkey, Iraq, Jordan, Lebanon, Libya, and Egypt
• Turkey hosts 2.74 million refugees (July 2016)
• Devastating effects not only for Syria, but also for Turkey, Lebanon,
more recently for most EU countries due to refugee inflow
• The economic cost on Turkey, Lebanon, Jordan, Iraq, and Egypt is
estimated as $35 billion in output at the beginning of 2016 and it
will climb
• $10 billion will be needed through 2020 for the financing needs of
Syrian crises
3

6. Hard Choices for Refugees and
Politicians

7. Re-migration
• Starting in 2014, a large number of Syrian asylum seekers in Turkey
re-migrated to EU countries
• Migration routes: Mediterranean Sea and Aegean Sea on boats, at
face of a high risk of dying due to boat sinks
• Approximately 2% of the those attempting to migrate to Europe
from Turkey die by drowning
• Yet, the thousands attempt to migrate through this dangerous way
every day.
4

8. Hard Choices
” ... it is only in face to
face meetings with people
seeking asylum and with
refugees that it is possible
to comprehend the
subtlety of these stories
and the huge range of
emotions behind them;
anger, fear, anxiety,
jubilation, hope, guilt and
mistrust are just a few of
the more obvious ones.
–Alison Jeffers, Refugees, Theatre and Crisis:
Performing Global Identities, 2012.
5

9. Hard Choices
Hard Choices for Refugee
• Stay in the initial migration location, but living conditions are harsh
• Re-migrate to west, but costly, risky, likelihood of deport
• Return, but not safe, no place to live, may not even able to run the
kitchen
Hard Choices for Politicians
• Political solution has trade offs for all parties
• Refugee size ise to big for everybody
• Financing gap is large
6

10. Motivation

11. Motivation-I
• Existing literature on conflict-driven or forced migration offers only
limited knowledge on hard choices facing Syrian refugees
• Further migration intentions
• Return to homeland
• Stay in their present location
• Migrate to alternative locations (within the same country)
• These choices are likely to be affected by the physical and
non-physical damage caused by the conflict, living conditions of
refugees, and demographic and social-economic characteristics of the
refugees.
• Not much is known on these, which requires individual level data
7

12. Motivation-II
• Most of the literature on forced migration is based on aggregate
(mostly state- level) data (large-n studied)
• Large-n studies uses state level data
• Economic factors: gross domestic product (GDP) per capita, GDP
growth rate, unemployment, income distribution, economic
discrimination, energy consumption
• Geographic and environmental factors: Natural disasters, climate
change, geography, transportation, border controls, distance. etc.
• Conflict -related factors: war and civil war, internal conflicts,
genocide, civil rights, etc.
• Large-n studies draw inferences about individual decisions on
whether to migrate or stay, arguing that people can make rational
choices even under the extremely violent conditions of war
• However, the data used in these studies are not individual level data
8

13. Contribution

14. Contribution
• Uses three-wave survey data collected by AFAD in 2013, 2014, and
2015
• Unique data, capturing more detail on the various different
dimensions of the violence experienced in Syria and the effects of
these and other factors on their plans for further migration
• Considers migration, re-migration, re-location, and return decisions
• Capture a great deal in the way of heterogeneity among different
groups of Syrian refugees
• Obtains significant information on who, when, where and why moves
9

15. Population and Sampling Design

16. Geography
10

17. Population & Sampling Design-I
• Population: Syrian refugees in Turkey.
• Design: Multistage random sampling.
• Survey Waves: Wave 1 (2013), Wave 2 (2014), Wave 3 (2015), all
conducted by AFAD.1
.
Geographic Distribution of the Sample by Wave and Province
11

18. Population & Sampling Design-II
Geographic Distribution of the Sample
12

19. Population by Time and Settlement Type
13

20. Models

21. Research Questions: Refugee Migration Choices
1. RETURN: What determines the intention to return to homeland.
Variable: takes a value of 1, if the respondents stated they would
return to Syria, 0 otherwise.
2. MIGRATE: Indicator of migrating to another location. Variable:
takes a value of 1 if the respondent stated he (she) is planning to
migrate, and a value of 0, otherwise.
3. RTNASAP: Return to the home country “as soon as possible”.
Variable: takes a value 1, if the respondents state they will return to
Syria “as soon as possible”, and 0 otherwise.
4. MIGRATEINT. Indicator of re-migration. Variable: takes a value of
1, if the respondents stated they are planning to migrate to another
country, and a 0 otherwise.
14

22. Individual Level Explanatory Variables-I
1. INCOME: Household income (INCOME) while living in Syria before
migrating to Turkey, measured in Syrian pounds and classified into 6
classes, 0-10000, 10001-20000, 20001-30000, 30001-40000,
40001-50000, and 50001+.
2. SHELTER: A binary variable, taking a value of 1 if the respondents
state availability of shelter in the home country, and 0 if the shelter
is not available.
3. DAMAGE: Measures property damage, takes a value of 1 if the
refugee stated his(her) home to be “collapsed”, “heavily damaged”,
or “partially damaged”; takes on a value of 0, if the refugee stated
his home is “not damaged”.
15

23. Individual Level Explanatory Variables-II
4. DEATH: Existence of deaths of the family members, takes a value of
1, if the respondents stated one or more families died due to war,
and a value of 0, if no family member died.
5. RFGTIME: Time lived as a refugee in the current host country
(Turkey), defined as 1-12 months, 13-24 months, 25-36 months,
35-48 months, and 49-60 months.
6. RFGMONTHS: Number of months lived as refugee, used when it
gives a better fit than RFGTIME in terms of the Bayesian
Information Criterion.
16

24. Demographics & War Impacts

25. Distribution by Home Governorate & Sex
17

26. Geographic Distribution by Home Governorate
18

27. Time Lived as Refugee
19

28. Month of Leaving Syria by Governorate
20

29. Verified War Casualties by Month
21

30. Verified War Casualties by Governorate
22

31. Damage to Property
23

32. Empirical Results

33. Empirical Models
Logistic regression for response yi :
Pr(yi = 1|Xi , β) =
exp{Xi β}
1 + exp{Xi β}
(1)
where yi is one of the response variables: RETURNi , MIGRATEi ,
RTNASAPi , or MIGRATEINTi
and
Xi β = β0 + β1INCOMEi + β2SHELTERi
+ β3DAMAGEi + β4DEATHi + β4RFGTIMEi
(2)
24

34. Descriptive Statistics
25

35. Full Model Estimates
26

36. Models Selected by General Specific Approach
27

37. Predictive Margins of Return for Time Lived As Refugee
28

38. Predictive Margins of International Migration for Time Passed
After Leaving Syria
29

39. Conclusion

40. Conclusion-I
• Duration of the violence in Syria increases the probability of
permanent settlement in another country.
• As the duration of the violence in Syria is increased the probability
of return to Syria is drastically reduced.
• One important determinant of decision to move from Turkey (initial
refuge location) is the duration of time lived as a refugee.
• As the cost of the conflict (measured by damage to property,
casualties to family members, job loss, etc.) increases for the
conflict-affected households, it is less likely that these refugees will
return to Syria.
• As the cost of the conflict increases, the more likely that the
refugees will move to higher income countries in search of asylum.
30

41. Conclusion-II
• Refugees who would not have a shelter in their home country if they
were to return home, are less likely to return to Syria and more likely
to migrate or re-migrate.
• Refugees who have deaths in their family due to war have lower
probability of return to Syria and a higher probability of international
migration.
• Women refugees are more likely to return back to Syria.
• Protraction of the conflict reduces the probability of returning to
Syria and raises the probability of re-migration.
• Higher income refugees are more likely to move to another country
from Turkey.
• There is a time threshold, which when exceeded, the refugees lose
any hope for return to Syria and search for permanent settlement
elsewhere.
31

42. Thank you ...
32