Prison or Sanctuary? An Evaluation of Camps for Syrian Refugees - - PowerPoint PPT Presentation
Prison or Sanctuary? An Evaluation of Camps for Syrian Refugees Thomas Ginn Center for Global Development April 2020 Motivation 68.5 million people displaced by conflict worldwide 10 million displaced people live in official camps
Center for Global Development
◮ 68.5 million people displaced by conflict worldwide ◮ ≈ 10 million displaced people live in official camps or settlements
◮ Otherwise live in cities, towns, informal settlements, etc: “urban”
◮ 68.5 million people displaced by conflict worldwide ◮ ≈ 10 million displaced people live in official camps or settlements
◮ Otherwise live in cities, towns, informal settlements, etc: “urban”
◮ Perception of camps:
◮ UN High Commissioner for Refugees 2014 policy change: “camps
◮ 68.5 million people displaced by conflict worldwide ◮ ≈ 10 million displaced people live in official camps or settlements
◮ Otherwise live in cities, towns, informal settlements, etc: “urban”
◮ Perception of camps:
◮ UN High Commissioner for Refugees 2014 policy change: “camps
◮ What are the effects of creating a camp on well-being?
◮ Including camp residents, urban refugees & citizens
◮ How does living in a camp, instead of the local community, affect:
◮ Labor market outcomes ◮ Income relative to cost of living ◮ Amenities: education, health care, social networks, safety, etc. ◮ Overall satisfaction
◮ How does living in a camp, instead of the local community, affect:
◮ Labor market outcomes ◮ Income relative to cost of living ◮ Amenities: education, health care, social networks, safety, etc. ◮ Overall satisfaction
◮ How does the difference across locations evolve over time?
◮ How does living in a camp, instead of the local community, affect:
◮ Labor market outcomes ◮ Income relative to cost of living ◮ Amenities: education, health care, social networks, safety, etc. ◮ Overall satisfaction
◮ How does the difference across locations evolve over time? ◮ Are camps cost effective?
◮ Would camp residents prefer the additional aid expenses in cash? ◮ Do camps in Jordan generate a deadweight loss or efficiency gain?
◮ Present Jordan first
◮ Cost estimates disaggregated by location (gov’t & humanitarian) ◮ Existing literature on effects of urban refugees on Jordanians ◮ 18% “choose” to live in camps; direct aid similar across locations
◮ Extend to Iraqi Kurdistan for comparison ◮ Collected in 2016; recall outcomes for 2010 (pre-conflict) & 2013
◮ Exploit different variation and require different assumptions ◮ Similarity of estimates suggests they are causal
◮ Camp residence reduces income:
◮ Camp residence reduces income:
◮ Camp residents save on rent:
◮ Camp residence reduces income:
◮ Camp residents save on rent:
◮ Minimal differences between camp & urban satisfaction, amenities
◮ Camp residence reduces income:
◮ Camp residents save on rent:
◮ Minimal differences between camp & urban satisfaction, amenities ◮ Services & aid cost more in camps:
◮ Camp residence reduces income:
◮ Camp residents save on rent:
◮ Minimal differences between camp & urban satisfaction, amenities ◮ Services & aid cost more in camps:
◮ Net efficiency gain from camps:
◮ Camp residence reduces income:
◮ Camp residents save on rent:
◮ Minimal differences between camp & urban satisfaction, amenities ◮ Services & aid cost more in camps:
◮ Net efficiency gain from camps:
◮ Urban refugees increase rent for Jordanians, few other net effects
◮ Camp residence reduces income:
◮ Camp residents save on rent:
◮ Minimal differences between camp & urban satisfaction, amenities ◮ Services & aid cost more in camps:
◮ Net efficiency gain from camps:
◮ Urban refugees increase rent for Jordanians, few other net effects ◮ After 4 years, camps in Jordan are efficiently subsidizing
◮ Forced Migration
◮ Krishnan et al (2017, internal World Bank) summarizing these data ◮ Outcomes for Displaced: Lehrer (2009), Betts (2014), Kondylis (2007),
◮ Place-Specific Effects
◮ Immigrant Enclaves & Camps: Borjas (2000), Edin & Fredriksson (2001),
◮ General: Chetty & Hendren (2018a,b), Bryan & Morten (2018), Franklin
◮ Framework ◮ Setting ◮ Data ◮ Empirical Strategy & Selection ◮ Results ◮ Cost Effectiveness ◮ Policy Discussion
◮ How are people allocated across space? ◮ Geographic concentration: returns to scale vs. congestion costs
◮ How are people allocated across space? ◮ Geographic concentration: returns to scale vs. congestion costs ◮ Setup: one existing city, & a wave of forced migrants arrive:
◮ Market: households choose city or desert based on private returns ◮ Social planner: can choose city sizes based on total social welfare
◮ How are people allocated across space? ◮ Geographic concentration: returns to scale vs. congestion costs ◮ Setup: one existing city, & a wave of forced migrants arrive:
◮ Market: households choose city or desert based on private returns ◮ Social planner: can choose city sizes based on total social welfare
◮ How are people allocated across space? ◮ Geographic concentration: returns to scale vs. congestion costs ◮ Setup: one existing city, & a wave of forced migrants arrive:
◮ Market: households choose city or desert based on private returns ◮ Social planner: can choose city sizes based on total social welfare
◮ How are people allocated across space? ◮ Geographic concentration: returns to scale vs. congestion costs ◮ Setup: one existing city, & a wave of forced migrants arrive:
◮ Market: households choose city or desert based on private returns ◮ Social planner: can choose city sizes based on total social welfare
◮ How are people allocated across space? ◮ Geographic concentration: returns to scale vs. congestion costs ◮ Setup: one existing city, & a wave of forced migrants arrive:
◮ Market: households choose city or desert based on private returns ◮ Social planner: can choose city sizes based on total social welfare
◮ How are people allocated across space? ◮ Geographic concentration: returns to scale vs. congestion costs ◮ Setup: one existing city, & a wave of forced migrants arrive:
◮ Market: households choose city or desert based on private returns ◮ Social planner: can choose city sizes based on total social welfare
◮ How are people allocated across space? ◮ Geographic concentration: returns to scale vs. congestion costs ◮ Setup: one existing city, & a wave of forced migrants arrive:
◮ Market: households choose city or desert based on private returns ◮ Social planner: can choose city sizes based on total social welfare
◮ Camps potentially act as a coordinating mechanism for new arrivals
◮ 1 in 14 people is a Syrian refugee (2nd highest globally) ◮ No access to formal labor market (changed after survey) ◮ Direct assistance for both camp & urban residents
◮ Food vouchers, food in-kind, cash
◮ Free education, some free health care ◮ Limitations on movement
◮ Camp residents: need permit to leave temporarily ◮ Urban residents: worry about arrest over paperwork
◮ Two main UNHCR camps: 80,000 and 30,000 residents in 2016 ◮ After 2012 opening, residence required on arrival; ≈ 200, 000 left ◮ Employment: 62% with NGOs, 80% inside camps ◮ “Nothing permanent” restrictions
◮ With thanks to partners at the World Bank, UNHCR, and Danida ◮ Sample’s 1st stage: randomly select areas, stratify by % Syrian ◮ Sample’s 2nd stage: randomly select households, by host/Syrian
◮ With thanks to partners at the World Bank, UNHCR, and Danida ◮ Sample’s 1st stage: randomly select areas, stratify by % Syrian ◮ Sample’s 2nd stage: randomly select households, by host/Syrian ◮ Representative* sample (incl unregistered), use sampling weights ◮ Response rates above 85% ◮ Two 15-64 year-olds per household randomly selected for labor
◮ Recall outcomes for Sept 2010 (pre-conflict) & Sept 2013
◮ Goal: estimate how camp residents would have done outside camps
◮ Treatment on the treated estimate ◮ Partial equilibrium: one resident leaves, camps remain open
◮ Identifying assumption: conditional on observables, choice to live
◮ Control for 85 baseline (2010) covariates:
◮ Household: demographics, housing, assets, exposure to violence ◮ Individual: employment status, occupation, wages, education
◮ Location choice (selection into camps) interesting on its own
◮ Perception that vulnerable - poor, women, children - go to camps
Means
◮ Canvas tents until mid-2013 ◮ Black market for extra tents and caravans ◮ More space/person upon arrival to camps, more likely to stay
Graphs
Means
◮ Movers more likely to cite housing and insecurity than jobs:
◮ Generalized Random Forest to generate propensity score (Athey et
◮ Flexibly captures interactions, higher orders ◮ 81% successful predictions ◮ Arrival month, housing, & origin are best predictors
◮ Propensity score balances on 2010 covariates
◮ Stratify (blocking): Imbens & Rubin (2015) ◮ Ordinary Least Squares
◮ Control for head’s age, age2, gender, 2010 education, months
◮ No governorate fixed effects: wrt avg of potential destinations ◮ Governorate fixed effects: wrt surrounding areas
◮ Setting ◮ Data ◮ Empirical Strategy & Selection ◮ Results ◮ Cost Effectiveness ◮ Policy Discussion
◮ Total wage & self-employment income 30 days prior to survey ◮ Inv Hyp Sin: log including 0’s (60% camp, 40% urban)
(1) (2) (3) (4) Inverse Gov FEs Blocking OLS Hyp Sin Blocking Camp
(20.22) (20.30) (0.26) (20.75) Linear Propensity
Score (15.43) (0.20) Observations 1197 1235 1235 1197 Governorate FEs No No No Yes Clustered SEs No Yes Yes No Number of Blocks 6 1 1 6 R2 .17 .16 Oster δ for β = 0 1.25 1.53 Means (USD) Camps 82 82 82 82 All Urban 192 193 193 192 ◮ Driven by lower male employment (45 vs 32%) ◮ ↑ female employment (9 vs 2%, NGOs) not enough to compensate
◮ Other income (i.e. remittances) & aid similar across locations ◮ Minimum expenses standardized to 0 in camps
Details ◮ Conservative estimate of rent saved ◮ Median reported food prices equal
◮ Convert to per capita to standardize with cost estimates
Regressions
◮ No difference in access to credit, food consumption
Graphs
◮ Fewer assets - TVs, radios, appliances - in camps
Graphs
◮ Sample: individuals who have not moved since 2013 ◮ P-values for different trends: 0.72, 0.42 ◮ Causal under parallel trends assumption
◮ P-values for different trends: 0.04, 0.66
◮ Education in camps:
Graph ◮ Primary school age children more likely to be in school in 2016 ◮ Upon arrival, less likely to be in school; some catching up
◮ Equal utilization and similar reasons for last health clinic visit ◮ Public services (garbage, electricity, etc.): less satisfied in camps ◮ 97% of sample reports feeling safe or very safe
◮ Men in camps feel less safe than urban males
◮ Women in camps list more regular contacts than urban women ◮ Highest satisfaction in Za’atari Camp, lowest in Azraq Camp
◮ As in Jordan camps: lower male employment, more self-employed ◮ Different from Jordan: no effect on females
Camp Characteristics
◮ Setting ◮ Data ◮ Empirical Strategy & Selection ◮ Results ◮ Cost Effectiveness ◮ Policy Discussion
◮ Spending by 64 non-profits & government not tracked by location
◮ Compiled 1,265 budgets of proposed projects ◮ Scaled to actual funding & arrivals (76%)
◮ Spending by 64 non-profits & government not tracked by location
◮ Compiled 1,265 budgets of proposed projects ◮ Scaled to actual funding & arrivals (76%)
◮ Sector-level estimates of government expenditures
◮ Sectors: education, health, electricity, municipal services
◮ Reports spending net of RRP (gross budget support)
◮ Spending by 64 non-profits & government not tracked by location
◮ Compiled 1,265 budgets of proposed projects ◮ Scaled to actual funding & arrivals (76%)
◮ Sector-level estimates of government expenditures
◮ Sectors: education, health, electricity, municipal services
◮ Reports spending net of RRP (gross budget support)
◮ 22% increase in urban refugee population ◮ Refugees: likely crowd each other out of jobs & housing
◮ Current estimates would understate camp surplus
◮ Jordanians: growing literature on effects of urban refugees
◮ Income: Minimal effects ◮ Krishnan et al (2017), Fallah et al (2018), Fakih & Ibrahim (2016) ◮ Cost of Living: Inelastic supply of housing ⇒ rent increases ◮ Rozo & Sviatschi (2018), Al-Hawarin et al. (2018) ◮ Public Services: Minimal effects, aid mitigates crowd out ◮ Rozo & Sviatschi (2018), Assaad et al (2018)
◮ Total aid budget potentially affected
◮ Returns to in-kind shelter possibly higher than cash equivalent
◮ Housing supply inelastic in short-run; risky for builders ◮ Shelter is a key constraint for both hosts and refugees
◮ Providing shelter does not have to imply other restrictions
◮ Isolation, limits on movement & building are likely lose-lose ◮ Fewer refugees stay, higher costs, no gains to hosts ◮ Examples: Azraq’s design, Za’atari’s initial phases ◮ Incentivize camp residence to minimize citizen crowd out
◮ Counterfactual labor and housing markets matter
◮ Weak Jordanian markets, NGOs can only employ in camps
◮ Long-run policy and outcomes uncertain and important
◮ Housing subsidy, potential tax on labor, alternate service provider
◮ Housing subsidy, potential tax on labor, alternate service provider
◮ Housing subsidy, potential tax on labor, alternate service provider
◮ Residents: Income, expenses, amenities
◮ Housing subsidy, potential tax on labor, alternate service provider
◮ Residents: Income, expenses, amenities ◮ Costs: Efficiency of public goods production
◮ Housing subsidy, potential tax on labor, alternate service provider
◮ Residents: Income, expenses, amenities ◮ Costs: Efficiency of public goods production ◮ Non-residents: Citizens & refugees
◮ Housing subsidy, potential tax on labor, alternate service provider
◮ Residents: Income, expenses, amenities ◮ Costs: Efficiency of public goods production ◮ Non-residents: Citizens & refugees ◮ Dynamics: Initial decisions could be difficult to reverse
◮ Housing subsidy, potential tax on labor, alternate service provider
◮ Residents: Income, expenses, amenities
◮ Little previous data or evidence
◮ Costs: Efficiency of public goods production ◮ Non-residents: Citizens & refugees ◮ Dynamics: Initial decisions could be difficult to reverse
◮ Housing subsidy, potential tax on labor, alternate service provider
◮ Residents: Income, expenses, amenities
◮ Little previous data or evidence
◮ Costs: Efficiency of public goods production
◮ Assemble and standardize government & non-profit estimates
◮ Non-residents: Citizens & refugees ◮ Dynamics: Initial decisions could be difficult to reverse
◮ Housing subsidy, potential tax on labor, alternate service provider
◮ Residents: Income, expenses, amenities
◮ Little previous data or evidence
◮ Costs: Efficiency of public goods production
◮ Assemble and standardize government & non-profit estimates
◮ Non-residents: Citizens & refugees
◮ Other literature on effects of refugees outside of camps
◮ Dynamics: Initial decisions could be difficult to reverse
(1) (2) (3) (4) (5) (6) Inverse Governorate Fixed Effects Blocking OLS Hyp Sin Blocking OLS IHS Camp
(20.22) (20.30) (0.26) (20.75) (11.84) (0.19) Linear Propensity
6.10 0.17 Score (15.43) (0.20) (12.68) (0.16) Observations 1197 1235 1235 1197 1235 1235 Governorate FEs No No No Yes Yes Yes Clustered SEs No Yes Yes No Yes Yes Number of Blocks 6 1 1 6 1 1 R2 .17 .16 .31 .23 Oster δ for β = 0 1.25 1.53 .69 2.7 Means (USD) Camps 82 82 82 82 82 82 All Urban 192 193 193 192 193 193 Govs near Camps 117 116 116 117 116 116 Capital City 293 293 293 293 293 293
Back
(1) (2) (3) (4) (5) (6) Inverse Governorate Fixed Effects Blocking OLS Hyp Sin Blocking OLS IHS Camp
(4.14) (5.33) (0.08) (4.31) (2.45) (0.05) Linear Propensity
2.22 0.01 Score (3.15) (0.04) (1.78) (0.03) Observations 1197 1235 1235 1197 1235 1235 Governorate FEs No No No Yes Yes Yes Clustered SEs No Yes Yes No Yes Yes Number of Blocks 6 1 1 6 1 1 R2 .11 .13 .21 .25 Oster δ for β = 0 1.08 1.19 .61 .65 Means (USD) Camps 43 43 43 43 43 43 All Urban 62 62 62 62 62 62 Govs near Camps 51 51 51 51 51 51 Capital City 76 76 76 76 76 76
Back
(1) (2) (3) (4) (5) (6) Inverse Governorate Fixed Effects Blocking OLS Hyp Sin Blocking OLS IHS Camp 10.99*** 9.40** 0.95*** 12.84*** 13.01*** 1.07*** (4.02) (3.75) (0.18) (4.27) (2.38) (0.13) Linear Propensity
2.16
Score (2.14) (0.12) (1.81) (0.11) Observations 1197 1235 1235 1197 1235 1235 Governorate FEs No No No Yes Yes Yes Clustered SEs No Yes Yes No Yes Yes Number of Blocks 6 1 1 6 1 1 R2 .05 .09 .11 .11 Oster δ for β = 0 8.14 2.16 3.8 2.1 Means (USD) Camps 43 43 43 43 43 43 All Urban 39 39 39 39 39 39 Govs near Camps 32 32 32 32 32 32 Capital City 47 47 47 47 47 47
Back
Back
Back
Back
Back
Back
Camp - Urban Difference Category Selected 2010 Variables Mean 2016 Arrival Housing Rooms 4.85
More
Apartment 0.20
Concrete Floor 0.13 0.18*** 0.10*** Household Household Members 4.74
0.33** Demographics Age (Head) 35.15
More
Dependents 2.12 0.04 0.37*** Human and Years Education (Head) 6.95
0.38 Physical Capital Met Basic Needs 0.87
0.01
More
Vehicle 0.23 0.08*** 0.05* Conflict and Months in Jordan 37.93
Displacement Dwelling Destroyed 0.38 0.00 0.06*
More
< 1 Day to Prepare 0.36 0.01 0.04 Observations 1,239
Back
Camp - Urban Difference Variables from 2010 Mean 2016 Arrival Labor Market: Males Randomly Selected Employed 0.53 0.06* 0.06 Skilled Job (ISCO Classification, 1-4) 2.05 0.24*** 0.01 Monthly Income (Median) 308 30 30 Observations 1,033 Labor Market: Females Randomly Selected Employed 0.03 0.05*** 0.02* Skilled Job (ISCO Classification, 1-4) 2.76
Monthly Income (Median) 253
Observations 1,292
Back
Back
(1) (2) (3) (4) (5) (6) (7) (8) (9) Linearized Propensity Score Observations Linearized Propensity Score OLS: HH Earnings Min Max Camp Urban Total Difference P-Value Coef SE Block 1
8 291 299 0.49 0.00
45.9 Block 2
33 116 149 0.10 0.00
42.2 Block 3
0.15 64 86 150 0.01 0.77
22.3 Block 4 0.15 0.46 114 35 149 0.03 0.10
37.3 Block 5 0.46 0.74 128 22 150 0.01 0.67
34.4 Block 6 0.75 1.98 275 25 300 0.20 0.00
39.3 All Blocks
1.98 622 575 1,197 1.42 0.00
16.7 Avg Treatment Effect on Treated
20.2 Avg Treatment Effect
20.9
Back
Back
◮ Also insignificant with controls and employment or wages as DV ◮ Suggests commuting costs not first-order
Back
◮ Also insignificant with controls and employment or wages as DV ◮ Suggests minimal differences in agglomeration in this range
Back