“Civil Wars Beyond their Borders: The Human Capital and Health

Consequences of Hosting Refugees”

Javier E. BaezHuman Development Department (LCSHS)

World Bank

Third Annual Research Conference on Population, Reproductive Health and Economic Development

Dublin, IrelandJanuary 16-18, 2009

2 CK / WB / 081206

Motivation

Worldwide stock of refugees(numbers in thousands)

JB / WB / 01-17-09

-

2,000

4,000

6,000

8,000

10,000

12,000

14,000

16,000

18,000

20,000

1970

1972

1974

1976

1978

1980

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

3 CK / WB / 081206

Motivation

But yet, little evidence on the impacts of refugees from wars on the well-being of host communities

Two works:

Alix-Garcia (2006): positive effects on household-level expenditures and proxies of welfare (micro data)

Montalvo & Reynal-Querol (2007): arrival of refugees increased number of cases of malaria (macro data)

My goal: identify the effects of hosting refugees on local populations, in particular on children’s well-being

JB / WB / 01-17-09

4 CK / WB / 081206

My approach

Context: a rural region in sub-Saharan Africa

Scope: identify impacts and examine their dynamics (i.e. short and long run analysis)

Strategy: exploit two past genocides as a natural experiment → D-D & D-D-D

Outcomes: anthropometrics, morbidity & infant mortality

(short run) height, health and human capital

accumulation (long run)

JB / WB / 01-17-09

5 CK / WB / 081206

Empirical implementation?

Ideal for evaluation purposes: random allocation of refugees among communities. But, is it feasible?

Need a big and exogenous population shock + an instrument for refugee intensity

Exploits population flows from the genocides in Burundi (late 1993’s) and Rwanda (early 1994’s) + geography as a natural experiment

JB / WB / 01-17-09

6 CK / WB / 081206

Flows of refugees into Tanzania

JB / WB / 01-17-09

Source: United Nations High Commission for Refugees (UNHCR)

7 CK / WB / 081206

Area of this study: Kagera

JB / WB / 01-17-09Source: U.S. Central Intelligence Agency

8 CK / WB / 081206

Kagera

Remotest region from Dar es Salaam

Mostly rural, major economic activity is agriculture

In 1993 (pre-shock time): Poorest among all the 20 regions in Tanzania

(GDP ≈ US$280) More than 50% living below the US$1 poverty

line Around 1,550,000 inhabitants

Refugee crisis: Kagera hosted over 500,000 refugees (≈ 33% of

local population, ≈ 80% of population in two main recipient provinces)

JB / WB / 01-17-09

9 CK / WB / 081206

Topographic characteristics

JB / WB / 01-17-09

10 CK / WB / 081206

Land use

JB / WB / 01-17-09

11 CK / WB / 081206

Rainfall

JB / WB / 01-17-09

12 CK / WB / 081206

Data

For short-run analysis: Tanzania Demographic and Health Surveys

(TDHS) Pool of cross-sections (1992, 96), used same

sample frame

For long-run analysis: Kagera Health and Development Survey (KHDS) A 10-13 year panel of ≈ 900 hh’s (1991-94,

2004)

Both datasets merged with their own GIS modules and 25 years of monthly rainfall data

JB / WB / 01-17-09

13 CK / WB / 081206

Evidence supporting the IV story?

Anecdotal evidence

Case studies & reports Personal talk with former Prime Minister

of Tanzania

Quantitative evidence

5th round of KHDS asked community leaders to assess the intensity of the influx of refugees in their villages (in a scale from 1 to 4)

JB / WB / 01-17-09

14 CK / WB / 081206

Location of refugees

JB / WB / 01-17-09

IndicatorNgara Karagwe Biharamulo Muleba

Bukoba Rural

Bukoba Urban

Refugee influx intensity 83.3% 60.0% 20.0% 14.3% 5.9% 0.0%

High Treatment Regions

Low Treatment Regions

Districts of Kagera Region by the Intensity of the Influx of Refugees

Notes: The indicator of intensity is calculated using retrospective information from the fifth round of the Kagera Health and Development Survey (2004) and shows the percentage of communities that reported having refugees’ resettlements close by in the village or the ward after 1993.

15 CK / WB / 081206

Administrative divisions

JB / WB / 01-17-09

16 CK / WB / 081206

Location of refugees

JB / WB / 01-17-09

0.5

1

1.5

2

2.5

3

3.5

4

4.5

0 20 40 60 80 100 120 140

Distance to the border with Rwanda (kms)

Ind

ex o

f in

ten

sity

of

refu

gee

s re

sett

lem

ents

Intensity of the Influx of Refugees and Distance from the Village to the Border with Rwanda

17 CK / WB / 081206

Research design

D-D Two specifications: binary and continuous

treatment Two strategies: intra- & intra-regional

variation

JB / WB / 01-17-09

18 CK / WB / 081206

Empirical specification (short-run analysis)

dtiidtitd RXY 43201

JB / WB / 01-17-09

itditddtdt ZxRxRx )()( 65

Y = outcome variable (anthropometrics, morbidity and mortality)

X = set of pre-shock covariates (individual & household)

, = pre-shock village covariates, village and year fixed effects

= treatment indicator (both binary and continuous)

Errors clustered at the village-year level

Note: pre-shock balance in observables (not shown)

R

19 CK / WB / 081206

Strategy 1: Short-run local impacts

JB / WB / 01-17-09

OutcomeHigh treat areas

Low treat areas

(i) (ii) (i) (ii) N

Nutritional Indicators

Weight-for-height z-score -0.446 -0.490 -0.542 ** -0.596 ** 0.0072 ** 0.0077 ** 457[1.107] [1.203] [0.222] [0.247] [0.0025] [0.0034]

Height-for-age z-score -1.666 -1.746 -0.522 -0.660 ** 0.0075 * 0.0094 ** 457[1.356] [1.391] [0.307] [0.302] [0.0043] [0.0041]

Morbidity

Children with diarrhea 0.036 0.046 0.135 0.183 * -0.0024 * -0.0028 ** 504[0.186] [0.211] [0.090] [0.093] [0.0014] [0.0012]

Children with fever 0.250 0.267 0.217 0.270 ** -0.0037 * -0.0040 * 504[0.434] [0.443] [0.125] [0.100] [0.0017] [0.0021]

Infant Mortality

Death in the first 5 years of life 0.084 0.163 0.083 0.108 * -0.0016 ** -0.0018 ** 647[0.279] [0.370] [0.061] [0.054] [0.0007] [0.0007]

Controls? No Yes No Yes

D-D: Distance to the border

Pre-shock mean D-D: West=1, East=0

Sample: Villages in Kagera Region

Reduced form regressions

20 CK / WB / 081206

Strategy 2: Short-run local impacts

JB / WB / 01-17-09

Outcome T C (i) (ii) (i) (ii) N

Nutritional Indicators

Weight-for-height z-score -0.474 -0.252 -0.291 * -0.333 * 0.0009 * 0.0011 ** 1,619 [1.165] [1.188] [0.170] [0.175] [0.0005] [0.0005]

Height-for-age z-score -1.715 -1.695 -0.304 * -0.303 * 0.0011 ** 0.0011 ** 1,619 [1.376] [1.258] [0.183] [0.184] [0.0005] [0.0005]

Morbidity

Children with diarrhea 0.042 0.100 0.151 *** 0.171 *** -0.0004 *** -0.0005 *** 1,796 [0.202] [0.300] [0.044] [0.047] [0.0001] [0.0001]

Children with fever 0.260 0.373 0.179 ** 0.201 *** -0.0005 *** -0.0006 *** 1,784 [0.439] [0.483] [0.074] [0.079] [0.0002] [0.0002]

Infant Mortality

Death in the first 5 years of life 0.136 0.128 0.047 0.076 * -0.0002 *** -0.0003 *** 2,286 [0.343] [0.334] [0.039] [0.041] [0.0001] [0.0001]

Controls? No Yes No Yes

Sample: Villages in Kagera, Mara and Mwanza regions

Pre-shock meanD-D: Kagera=1,

Kagera=0D-D: Distance to the

border

Reduced form regressions

21 CK / WB / 081206

Distributional changes0

.1.2

.3.4

den

sity

-5 -4 -3 -2 -1 0 1 2 3 4 5wh_zscore

treatment control

JB / WB / 01-17-09

0.1

.2.3

.4d

ensi

ty-5 -4 -3 -2 -1 0 1 2 3 4 5

wh_zscore

treatment control

Pre- and post-shock distributions of the W/H z-score between treatment and control villages (within Kagera)

22 CK / WB / 081206

Distributional changes0

.1.2

.3.4

den

sity

-6 -4 -2 0 2 4 6ha_zscore

treatment control

JB / WB / 01-17-09

Pre- and post-shock distributions of the H/A z-score between treatment and control villages (within Kagera)

0.1

.2.3

.4d

ensi

ty-6 -4 -2 0 2 4 6

ha_zscore

treatment control

23 CK / WB / 081206

Empirical strategy (long-run analysis)

JB / WB / 01-17-09

D-D & D-D-D:

Strategy: use intra-regional and intra- and inter- cohort variation in exposure to the shock

Exposed cohort: children 0-5 years old in 01/1994, (10-15 years in 2004) & Western=1,

Less exposed cohorts: children 0-5 years old in 01/1994 & Western=0,

Less responsive cohorts: people 15-20 years old in 01/1994 both at Western=1 & Western=0.Specifications A & B

24 CK / WB / 081206

Empirical specification

dtiidtitd RXY 43201

JB / WB / 01-17-09

D-D & D-D-D:

itditddtdt ZxRxRx )()( 65

1 2 3 4 5 6( ) ( )ijt idt i t d i t i dY X R R

7 8( ) ( )t d t i d itdR R

Y = outcome variable (height, chronic morbidity , schooling and literacy)

25 CK / WB / 081206

Results

JB / WB / 01-17-09

Outcome West=1 East=0 (i) (ii) (i) (ii) (i) (ii) (i) (ii)

Anthropometrics

Height (centimeters) 84.90 86.31 -1.943 -1.719 ** -1.777 * -2.184 0.0208 0.0196 ** 0.0280 ** 0.0289 **

[1.030] [0.639] [2.068] [0.680] [1.051] [1.135] [0.0229] [0.0094] [0.0131] [0.0142]

Observations 1,861 1,785 1,159 1,078 1,861 1,785 1159 1078

Morbidity

Chronic disease 0.088 0.110 -0.022 -0.005 -0.066 -0.069 0.0003 0.0001 0.0006 0.0007

[0.034] [0.042] [0.041] [0.037] [0.130] [0.147] [0.0005] [0.0006] [0.0017] [0.0019]

Observations 1,548 1,340 855 722 1,548 1,340 855 722

Education

School attainment (years) --- --- -0.179 -0.172 -0.272 -0.200 0.0040 * 0.0045 *** 0.0064 *** 0.0051 **

[0.163] [0.120] [0.193] [0.204] [0.0022] [0.0014] [0.0023] [0.0025]

Observations 2,125 2,026 1,281 1,190 2,125 2,026 1281 1190

Does read/write? --- --- -0.057 -0.068 -0.075 -0.065 0.0011 0.0013 ** 0.0013 ** 0.0012 **

[0.067] [0.048] [0.0467] 0.051 [0.0007] [0.0005] [0.0005] [0.0006]

Observations 2,127 2,028 1,284 1,193 2,127 2,028 1284 1193

Controls? No Yes No Yes No Yes No Yes

Reduced form regressions

D-D: West=1, East=0 D-D: Distance to the border

Pre-shock mean Pooled Panel Pooled Panel

26 CK / WB / 081206

Robust analysis

JB / WB / 01-17-09

1. Pre-trends? No, based on placebo test with pre-shock data

2. Misspecification bias? D-D PSM replicate results

3. Endogenous migration/attrition? Models suggest not an issue

4. Selection for survival? fixed (comparison of parental height distribution), variable (lower bound)

5. Measurement error in real level of exposure/non-exogeneity of international and local assistance? (lower bound)

27 CK / WB / 081206

Interpretation: why negative effects?

JB / WB / 01-17-09

Threats posed to public health Emergence and spread of existing infectious

diseases by food, water and environmental contaminating factors

Multiplicative effects in the transmission/infection of malaria

Overpopulation

Competition for resources (food, land, wood)

Collapse of health and educational services

28 CK / WB / 081206

Conclusions

JB / WB / 01-17-09

Evidence of a previously undocumented indirect effect of civil wars on the health and human capital of children in hosting communities.

External validity? More than 15 civil wars + internally displaced +

refugees from political oppression

Persistence of the effects Health shocks in childhood correlated with cognitive

deficits and attained body size later in life Effects on future productivity (2.8% fall in their

earnings)

Policy implications

“Civil Wars Beyond their Borders: The Human Capital and Health

Consequences of Hosting Refugees”

Javier E. BaezHuman Development Department (LCSHS)

World Bank

Third Annual Research Conference on Population, Reproductive Health and Economic Development

Dublin, IrelandJanuary 16-18, 2009