Is there such a thing as Migration of Poverty in Albania?

Is there such a thing as Migration Is there such a thing as Migration of Poverty in Albania?of Poverty in Albania?

Jessica Hagen-ZankerCarlo Azzarri

ABCDE ConferenceTirana, June 10-11, 2008

Introduction

• Migration most important social, political & economic phenomenon in Albania since 1990

• Internal migration also important, but understudied• Internal migration mainly rural to urban/peri-urban

areas• In ‘90s urban population increased by 14%, but not

much known on living conditions of migrants What is the impact of internal migration on migrant

households?• Since 1990 poverty decreases, especially in rural Has poverty relocated from rural to urban areas?

Novelty of the paper

• Focus on impact of internal migration

• Albania as quasi-experimental case: no internal or international migration before 1990

• Unique dataset- Households over-sampled in peri-urban areas- Retrospective information on migration- Information on households in 1990 (controls)

Data

• Data- LSMS 2005 (nationally representative)- 3840 households- 200 peri-urban households oversampled • Groups- RNM = Rural household, head did not migrate

internally- PNM = Peri-urban household, head did not migrate

internally- PM = Peri-urban household, head did migrate

internally

Descriptive statistics I

• Migrants to peri-urban younger & less educated• Migrants to peri-urban more likely to be unemployed

& working fewer hours Employed in casual construction sector

• Rural households with more international migrants than peri-urban specialization?

Descriptive statistics II: Income and consumption

RNM = Rural household, head did not migrate internallyPNM = Peri-urban household, head did not migrate internallyPM = Peri-urban household, head did migrate internally

02,0

00

4,0

00

6,0

00

8,0

00

10

,000

Month

ly a

moun

t in

new

Leks

RNM PNM PM

Per capita income Per capita consumption

Descriptive statistics III

• Peri-urban migrants show worse housing condition, both compared to rural non-migrant households & own situation in 1990 (in terms of house type, number of rooms, water access & quality)

• Peri-urban migrant children least likely to be sent to primary school (70%) & as unlikely to secondary school as rural households (33%)

- Schools far- Teenagers work as much as in rural areas- Households do not consider education as important

as peri-urban non-migrants do

Descriptive statistics IV

• Comparison over time (whether moved 90-94, 95-99, 00-04)

• Internal migrants move for different reasons1) Pioneers: to improve life more likely to send

children to school2) Crisis movers: out of need (pyramid savings scheme

crisis) poor housing & employment3) Followers: to make money highest income gains

• Different expectations different impacts

Econometric analysis• Aim: measuring impact of internal migration on outcome of

interest (e.g. income)

Two Solutions:• Propensity Score Matching

Compare peri-urban internal migrant households to very similar non-migrant rural household

• Instrumental Variable Analysis

Replace explanatory variable with another variable (IV) correlated with explanatory variable only

XmigrationY **

Propensity Score Matching Results

Dependent variable

Income/ capita Consumption/ capita

ATT 2257.37*** 658.34

s.e. 575.19 500.73

ATU 2525.05 334.48

ATE 2498.28 366.76

Confirms descriptive statistics

ATT=Average treatment effect for treated; ATU=Average treatment effect for un-treatedATE=Average treatment effect for population

Instrumental Variables

• Instrumental variables used:

1) Wealth in 1990 influences decision to move, but unlikely to affect current income due to the rapid changes that took place in Albania

2) Housing variables 1990 impacts decision to move, but not current income

Instrumental Variables Results

Treat_year= number of years since the household has moved, 0 for rural non-moversOther explanatory variables omitted for space reason

Dependent variable: Log income per capita

Instruments Coefficient & p-value of treat_year

No of obs. Anderson-Rubin Wald test p-value

Hansen J statistic p-value

Kleibergen-Paap F-statistic

Wealth 1990House type 1990

0.24(0.033)

1623 0.039 0.71 5.51

Wealth 1990Water source 1990

0.25(0.049)

1623 0.043 0.82 4.20

Wealth 1990Toilet type 1990

0.35(0.014)

1623 0.000 0.48 3.37

Wealth 1990Rooms/ capita 1990

0.18(0.065)

1623 0.070 0.37 4.30

• All the tests successful, although instruments could be stronger

Conclusions

• Migrants are better off in terms of income• In peri-urban monetary poverty amongst

migrants still high compared to non-migrants migration of poverty?

• Migrants are worse off in terms of housing, education, health, access to utilities, access to stable employmentLiving expenses increase > income gain

Faleminderit!