Using Survey Data to Improve Registry Data The Case of Syrian Refugees Paolo Verme World Bank.
-
Upload
kenneth-cobb -
Category
Documents
-
view
212 -
download
0
Transcript of Using Survey Data to Improve Registry Data The Case of Syrian Refugees Paolo Verme World Bank.
Using Survey Data to Improve Registry Data
The Case of Syrian Refugees
Paolo VermeWorld Bank
Motivation• UNHCR keeps records of all refugees registered with the proGres database
• UNHCR collects a rich set of additional information via home visits’ surveys
• This information is little exploited for analysis
• UNHCR needs to target refugees with its cash assistance program due to funding shortages
• How good is the UNHCR targeting capacity?
• Can UNHCR improve its targeting capacity by making better use of available information and reduce the cost of surveys?
Background• A WB-UNHCR partnership: WB analytical expertise on poverty and welfare and UNHCR expertise on refugees
• Pilot study in March 2014
• Two countries (Jordan and Lebanon)
• 6 Data sets (3 Jordan and 3 Lebanon): • UNHCR registry for Jordan and Lebanon (PG: 650,000 records Jordan; 1.2 m records Lebanon); • UNHCR and WFP home visits (45,000 cases in Jordan, 2 rounds; 30,000 cases Lebanon, 1 round)• UNHCR and WFP survey (1,700 cases Lebanon, 1 round)
• Focus on refugees living outside camps
Þ How Poor Are Refugees? A Welfare Assessment of Syrian Refugees Living in Jordan and Lebanon(16 December, 2015)
Basic Idea
1. Construct welfare aggregates
2. Construct a welfare and poverty model with PG+HV data
3. Tests best proxies for welfare and poverty
4. Tests composite indexes for welfare
5. Derive the best welfare and poverty models (PG+HV data)
6. Find the best variables to add to the proGres data
7. Predict welfare and poverty using PG data
8. Use the new PG model for targeting assistance programs
1) Construct Welfare Aggregate
Income per capita Expenditure per capita Expenditure per capita net of UNHCR assistance0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
16.1 16.1 16.1
39.2
0.0 2.1
44.7
83.9 81.8
Missing Zeroes Positive
1) Construct Welfare Aggregate
0.2
.4.6
.8
0 2 4 6 8x
Exp/cap with UNHCR cash Inc/cap with UNHCR cashExp/cap without UNHCR cash
2) Construct Welfare ModelsPHASE I :
Where W=welfare measure (income, expenditure or poverty); HP=vector of case characteristics present in both the PG and HV databases; H=vector of case characteristics present in the HV data but not in the PG data; P=vector of case characteristics present in the PG data but not in the HV data; h=Subset of H significant in Phase I; = normally distributed error term with zero means; i=household (case number in UNHCR data).
PHASE II : ,PHASE III :
3) Test best proxies of Welfare• Basic tabulations
• Univariate regressions on 185 variables
• Ranking by R2
• Multivariate tests: Backward and forward regressions
• Manual tests of key explanatory variables
• Tests repeated for composite welfare indicators
• Maximizing R squared
3) Test best proxies of Welfare
Male
Female
1/11/2013-24/2/2014
25/2/2014-1/10/2014
Non Muslim
Muslim
Non Arab
Arab
Edu 8 years of less
Edu less than 8 years
Blue Collars
White Collars
Low Skills
High Skills
gend
ertim
ere
ligio
net
hnic
educ
atoc
cup
skill
s
0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0
3) Test best proxies of WelfareVariable Description Obs W R2 lincsize_hv Individuals in case (HV) 15975 0.470121638dem_p_child Proportion of children 15975 0.312118415edu_p_attend Proportion of children in school 15975 0.152996621pov_inc_unhcr UNHCR Monthly Financial Assistance 15787 0.150577538
cash_largeLarge Family &/or Family with Babies, Toddlers or Children Attending School 5000 0.135079263
prot_bail_date 708 0.100568917pov_cop_aid Humanitarian assistance 15788 0.092286375cash_smother Single Women 5000 0.058084132prot_aggr Score for work and residence permit, MOI and bail out doc 3863 0.050228105cash_elderly Elderly 60 and Above 5000 0.048847739prot_moi_diff_apolice Not comfortable approaching a police station 187 0.041721867pov_inc_aggr Total number of kinds of income 15787 0.03765949prot_bail BailedOutFromCamp 13212 0.035152565cash_decision Decision_for_cash_Assistance 15811 0.034241792pov_cop_host Living together with host family 15788 0.029711534ref_elderly Elderly alone 1111 0.028308115prot_know_school School 14949 0.022848084edu_p_notattend Proportion of children not in school 15975 0.02232405house_sanitary SanitaryFacilitiesStatus 15602 0.021637243dem_pafemale Female Principal Applicant 15835 0.021228419
4) Test Composite Indexes of Welfare
Variable Obs Mean Std. Dev. Min Max R2 %
i_rent 15975 0.91518 0.278622 0 1 0.017427 1.742672
i_latrine 15975 0.773083 0.418852 0 1 0.014387 1.438695
i_good_liv~d 15975 0.476557 0.499466 0 1 0.013578 1.357762
i_housecon~n 15975 0.86723 0.339337 0 1 0.007125 0.712506
i_pipewater 15975 0.878685 0.326503 0 1 0.007003 0.700316
i_good_san~y 15975 0.138717 0.345662 0 1 0.006762 0.676154
i_good_ven~n 15975 0.28626 0.452026 0 1 0.006047 0.604679
i_waste 15975 0.746792 0.434863 0 1 0.005852 0.585157
i_water 15975 0.797684 0.401739 0 1 0.005125 0.512468
i_good_ele~y 15975 0.281189 0.449594 0 1 0.005017 0.501678
4) Test Composite Indexes of Welfare
Variable Obs Mean Std. Dev. Min Max R2
ind_house_crowd 15975 1.781887 1.364509 0 16 0.267
ind_house_crowd1 15975 2.551506 1.697571 0 58 0.022
ind_wash_water 15975 3.196244 1.168634 0 4 0.014
ind_nfi 15975 0.16169 0.381374 0 7 0.011
ind_house_subjective 15975 1.736588 1.619757 0 6 0.009
ind_house_assets 15975 8.3682 3.006631 0 13 0.008
ind_cope_index 15975 2.448013 1.727429 0 5 0.007
ind_wash_hygiene 15975 4.192363 1.150333 0 5 0.007
ind_cope_wfp 15975 1.665477 1.481963 0 8 0.006
ind_food_wfp 15975 42.55236 16.63382 0 112 0.003
ind_house_quality 15975 1.685383 0.57311 0 2 0.003
ind_food_score 15975 22.13459 8.49884 0 56 0.002
ind_food_variety 15975 7.101659 1.576538 0 8 0.001
5) Derive the optimal Welfare Model (PG+HV)
Coef. t Coef. t Coef. tVariable
importanceCase size (Ref. Csize=1) 2 -0.47 -34.9 -0.47 -36.5 -0.47 -35.8 18.09%
3 -0.83 -48.5 -0.83 -50.0 -0.83 -49.04 -1.04 -61.3 -1.04 -62.9 -1.05 -61.95 -1.19 -67.6 -1.19 -68.9 -1.20 -68.06 -1.42 -77.2 -1.40 -77.9 -1.41 -76.47 -1.45 -68.4 -1.43 -68.8 -1.44 -67.9 8 - 11 -1.66 -80.4 -1.62 -79.9 -1.63 -78.8>=12 -2.36 -36.6 -2.32 -37.3 -2.32 -36.9
Proportion of children <18 years (Ref. =0) 0 - 50% -0.05 -2.9 -0.06 -4.0 -0.06 -3.7 0.05%50% - 75% -0.06 -4.2 -0.07 -4.9 -0.06 -4.4>75% -0.09 -5.4 -0.08 -5.1 -0.07 -4.4
Employment of PA ( Ref. None) Low Skilled 0.00 0.1 0.03 2.4 0.03 2.2 0.07%Skilled 0.01 1.0 0.02 1.6 0.01 1.1High Skilled 0.02 1.6 0.01 0.9 0.01 0.7Professional 0.07 5.1 0.06 4.4 0.06 4.4
Age of PA ( Ref: <=34 years) 35-54 years 0.07 8.7 0.06 7.8 0.06 8.0 0.15%>=55 years 0.07 5.8 0.04 3.7 0.05 4.3
Marital status of PA (Ref. Married or engaged) Divorced or separated. -0.13 -5.5 -0.10 -4.3 -0.10 -4.2 0.20%Single -0.12 -9.7 -0.10 -8.6 -0.10 -8.2Widowed -0.10 -7.0 -0.07 -5.3 -0.06 -4.3
Highest education of PA ( Ref. Below 6 years) 6-8 years 0.06 7.1 0.02 2.2 0.02 2.2 0.10%9-11 years 0.11 10.1 0.05 5.0 0.05 4.612-14 years 0.13 10.6 0.07 5.8 0.06 5.2At least university 0.25 14.9 0.18 10.9 0.16 9.9
Origin (Ref. Damascus) Al-hasakeh -0.11 -2.8 -0.01 -0.2 -0.03 -0.7 0.17%Aleppo -0.03 -2.2 -0.05 -3.2 -0.06 -4.0Ar-raqqa -0.04 -1.6 -0.02 -0.9 -0.04 -1.6Dar'a 0.00 0.2 -0.03 -2.2 -0.02 -1.7Hama -0.33 -19.3 -0.02 -1.2 -0.02 -1.4Homs -0.06 -4.7 -0.11 -8.5 -0.09 -7.3 Idleb -0.32 -11.8 -0.04 -1.3 -0.05 -1.7Rural Damascus -0.03 -2.2 -0.03 -1.9 -0.02 -1.7Tartous -0.10 -0.9 -0.13 -1.1 -0.13 -1.1As-sweida -0.05 -0.6 0.04 0.5 0.06 0.8Deir-ez-zor -0.17 -4.3 -0.14 -3.6 -0.14 -3.6Lattakia 0.07 1.2 0.05 1.0 0.06 1.1Quneitra 0.05 1.1 -0.01 -0.2 0.00 -0.1
Only PG variables PG + house + wash All variables
Welfare Model(Cont.)
Legal arrival 0.12 13.3 0.09 9.9 0.08 9.0 0.16%Destination (Ref. Amman) Ajloun -0.12 -4.6 -0.14 -5.6 -0.16 -6.2 1.12%
Aqabah 0.06 1.0 0.04 0.7 0.02 0.4Balqa -0.08 -4.9 -0.07 -4.2 -0.07 -4.3Irbid -0.08 -8.7 -0.08 -9.0 -0.09 -9.7Jarash -0.13 -5.6 -0.15 -6.5 -0.19 -8.0Karak -0.14 -6.3 -0.15 -7.2 -0.15 -6.9Maan -0.14 -4.5 -0.14 -4.7 -0.13 -4.2Madaba -0.16 -7.2 -0.12 -5.4 -0.11 -5.1Mafraq -0.21 -19.8 -0.10 -9.3 -0.08 -7.0Tafilah -0.40 -9.5 -0.45 -11.1 -0.48 -11.8Zarqa -0.25 -22.6 -0.21 -20.0 -0.20 -18.6
Border crossing point (Ref. Airport) Ruwaished - Hadallat -0.01 -0.6 0.06 4.1 0.06 3.6 0.19%Tal Shihab -0.06 -3.7 -0.05 -3.2 -0.04 -2.8Nasib-official or unofficial. -0.07 -6.2 -0.05 -5.1 -0.04 -4.0other or no data. -0.04 -3.0 -0.02 -1.8 -0.02 -1.3
house_kitchen_d 0.10 11.6 0.11 11.5 0.39%house_electricity_d 0.04 3.3 0.04 3.5 0.07%house_ventilation_d 0.05 5.5 0.05 5.3 0.07%house for rent or owned 0.59 39.8 0.60 39.9 3.55%concrete_house 0.10 6.8 0.12 8.1 0.16%House area:sq m per capita (Ref. <10 sq meters) 10-15 sq meters 0.02 3.0 0.02 2.7 0.31%
>15 sq meters 0.11 14.4 0.09 11.5wash_piped (water through piped AND piped sewerage) 0.04 5.3 0.04 4.5 0.05%nfi_1_dummy (Receiving NFIs) -0.05 -5.9 0.08%pov_cop_aid (Coping strategy=humanitarian assistance) -0.15 -16.0 0.58%pov_cop_host (Coping strategy=host family) -0.12 -17.5 0.70%pov_cop_comm (Coping strategy=host community) -0.04 -6.0 0.09%prot_cert_valid (Is certificate valid) 0.06 7.1 0.13%pov_inc_unhcr (UNHCR financial assistance) -0.35 -34.4 -0.40 -40.2 -0.34 -31.7 2.28%
_cons 4.88 212.5 4.05 152.4 4.05 142.5F statistic 721.20 727.92 651.11Adjusted R-squared 0.48 0.53 0.53N 42217 40541 38694
6) Select the Best Variables to Add to PG DataThe poverty model predicts poverty correctly 90.1% of the times
• PG-HV predictors:• Case size • Rent• Place of destination (country and region)• Official entry and point of entry• Place of origin (Damascus vs other regions)
• HV predictors (h)• Principal applicant characteristics (age and marital status)• Selected assets (latrine, piped water, kitchen)
• Education and former occupation are less important than expected
7) Predict Welfare using new PG model
Wexp_unhcr_lncap
Coef t
Individuals in case (HV) -0.212*** -69.811
Proportion of children -0.611*** -25.245
Concrete House 0.195*** 8.017
Santitation average or above 0.109*** 7.244
Ventilation average or above 0.100*** 6.194
Free Housing -0.705*** -24.419
Proportion school-aged children 0.113*** 8.919
Proportion of children in school -0.207*** -15.267
Sharing costs with host family -0.095*** -6.628
Living together with host family 0.114*** 8.987
IsCertificateValid 0.124*** 9.190
_cons 4.715*** 162.779
Number of observations 14,150
R2 0.555
Adjusted/Pseudo R2 0.554
8) Use New PG Model to Target Assistance
Income
Expenditure Non poor Poor Total
Non poor 8.5 37.2 45.6
Poor 4.6 49.8 54.4
Total 13.1 86.9 100.0
Conclusion
• World Bank and UNHCR have complementary skills and resources
• Existing UNHCR data can be used to improve targeting:• Shift from income to expenditure• Reduce leaking and costs• Improve coverage and targeting
• The analysis highlighted areas to improve:• Data collection (survey design, sampling, questionnaires)• Data management (PG management system, survey administration)• Data analysis (economists and social protection officers)