Accounting for Migration and Remittance Effects

Accounting for Migration and Remittance Effects

Susan PozoPrepared for Conference on Regional Trade Agreements, Migration

and Remittances with Special Focus on CAFTA and Latin AmericaSam Houston State University

April 12, 2008

Much more attention paid to the migratory process in the past 5 years

1998 1999 2000 2001 2002 2003 2004 2005 2006 20070

20

40

60

80

100

120

140

Econ Lit Hits

1. Is this a research fad?

Source: Econ Lit database, 2008

10.47.9

6.24.7

5.46.9

8.811.6

13.214.7

13.614.8

13.314.4

13.29.7

0 5 10 15

2000199019801970196019501940193019201910190018901880187018601850

United StatesPercent Foreign Born

2. Growth in the number of persons affected by the migratory process?

Sour

ce: U

.S. B

urea

u of

the

Cens

us, 2

008

050

100

150

200

Mig

rant

s

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005

(in millions)World Migrants

Source: Data from UN (2008)

01

23

wor

ld m

igra

nts/

wor

ld p

opul

atio

n

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005

(percent of world population)World Migrant Stock

Source: Data from UN (2008)

0

1000

2000

3000

4000

5000

6000

7000

1996 1998 2000 2002 2004 2006

REMITTANCES

Remittances to Mexico (quarterly frequency, in millions of US dollars)

Source: Data from Banco Central de Mexico, 2008

.000

.005

.010

.015

.020

.025

.030

1980 1985 1990 1995 2000 2005

WR_GDP

Source: World Development Indicators, 2008

Remittances to Mexico (yearly frequency, Percent of GDP)

1

2

3

4

5

6

7

80 85 90 95 00 05 10 15 20

PERCENT

Remittances to Italy as a percent of Italian GDP(1880-1910)

Source: Computed by the author with data from Cinel (1991) and from Flandreau & Zumer (2004)

1990

2006

Source: US Census Bureau, http://factfinder.census.gov

3. Increased dispersion of the foreign born?

0.0

5.1

.15

.2.2

5.3

Den

sity

0 5 10 15 20 25 30p1990

1990Percent population foreign born

Computed by the author from Census Bureau

0.0

5.1

.15

.2.2

5.3

Den

sity

0 5 10 15 20 25 30p2000

2000Percent population foreign born

Computed by the author from Census Bureau

0.0

5.1

.15

.2.2

5.3

Den

sity

0 5 10 15 20 25 30p2006

2006Percent population foreign born

Computed by the author from Census Bureau

0.0

5.1

.15

.2.2

5.3

Den

sity

0 5 10 15 20 25 30p1990

1990Percent population foreign born

0.0

5.1

.15

.2.2

5.3

Den

sity

0 5 10 15 20 25 30p2006

2006Percent population foreign born

Increased spread of the foreign-born in 2006 relative to 1990

1990

2006

3. Increased dispersion of the foreign-born?

1990 2000 20060

5

10

15

20

25

30

Source: Computed by author from 1990, 2000 Decennial Censuses and 2006 American Community Survey, US Census.

Economic Development Effects of the Migratory Process on

Labor supply Health

Education Happiness

Poverty levels

Business Investments

Tend to focus on only one facet of the migratory process…

Poverty -- remittancesLabor force participation – remittancesEducation—remittancesBusiness Investment—(return) migrationHealth – emigrationHappiness - migration

Migratory Process

Remittances Migration

Economic Development Effects of the Migratory Process on

Labor supply Health

Education Happiness

Poverty levels

Business Investments

Migrant HH and Remittance Receipt

40%

4%

38%

17%

neithermigrant onlyremit onlyremit and migrant

Haiti

Source: Amuedo-Dorantes, Georges and Pozo, (2007)

Haiti

DR

Mexico

38

65

74

3.5

13

9

33

10

4

25

12

14

HH type in %R&M Remit migrant neither

Source: Computed by author from

LAMP and M

MP databases

18%

14%

4%64%

Cuenca, Ecuador

M&R Remit Migrant Neither

Computed by the author from : Discrimination and Economic Outcomes Survey Database, IADB, 2006

Too large

Too small

We miss out on the story when we focus on one or the other alone

In the modeling of education a typical strategy might be to estimate:

Education = βRemit +δX +ЄSeveral problems: i) endogeneity due to reverse causalityii) endogeneity due to omitted variable bias

Type of Household All

Model Specification ProbitVariables M.E.Remittance Receipt .0067HH Currently Employed -0.0199Assets 0.0494***% dependent age 0.3121Ed 17+ -0.2857*Ed female adult 0.0979% kids school age -0.3581**Own Child 0.1090*Boy -0.0210Child’s Age 0.0075Firstborn Child -0.0326**Urban -0.1263No. of Observations 327Wald Chi2-test 23.71Prob>Chi2 0.0222Log pseudolikelihood -104.4399

Source: Amuedo-Dorantes, Georges and Pozo, (2007)

Typical solution

Instrument for remittances:Using migration or variables linked to long-standing migratory patterns, such as the mapping of railroads. Essentially migration networks.

Problems with this Approach…

1. An instrument can’t be something that should be in the equation in the first place, i.e. migration and variables proxying for long-standing migratory patterns are likely to impact educational attainment via: A disruptive effect, in the case of family migration A network effect, in the case of both family and broadly

defined migration networks

Education

Migratory Process

Remittances

Migration

Migration K/networks

Everything else

Migration capital/networks

Expected value of additional education varies with the probability of future migration

EVH = (pH) RH,H + ( 1 - pH) RH,US

Type of Household All Model Specification Probit

CoefficientMigration networks/capital 0.4827**Household Head Currently Employed 0.0037Current Household Assets 0.2743***Percent of Non-working Age Household Members 1.8011Mean Potential Education of 17 Years + -1.7777**Potential Ed Attainment of Spouse or Head 0.3882Percent of School-age Children in the HH -2.1341***Own Child 0.4865Boy -0.1973*Child’s Age 0.0196Firstborn Child -0.1239

Problems with this Approach…

1. An instrument can’t be something that should be in the equation in the first place, i.e. migration and variables proxying for long-standing migratory patterns are likely to impact educational attainment via: A disruptive effect, in the case of family migration A network effect, in the case of both family and broadly

defined migration networks

2. We notice significant differences in selectivity with respect to different types of HHs. HHs without migrants receiving remittances are very different from HHs with migrants receiving remittances.

Conclusions

1. Redesign of surveys to take into account the diversity in the incidence of migration and remittances.

2. Redesign of econometric methodology to recognize differential “migration,” “remittance” and “migration capital” effects.

Type of Household All Non-migrant

Model Specification Probit IV-ProbitVariables M.E. M.E.Remittance Receipt .0067 0.6791***HH Currently Employed -0.0199 -0.2073*Assets 0.0494*** 0.0213% dependent age 0.3121 0.0223Ed 17+ -0.2857* 0.0182Ed female adult 0.0979 -0.2607% kids school age -0.3581** -0.2329Own Child 0.1090* 0.1594**Boy -0.0210 0.0214Child’s Age 0.0075 -0.0067Firstborn Child -0.0326** 0.0402Urban -0.1263 0.0216No. of Observations 327 258Wald Chi2-test 23.71 1181.35Prob>Chi2 0.0222 0.0000Log pseudolikelihood -104.4399 -243.2202IV Exogeneity Testa n.a. 0 < = 5.99Wald Test of Exogeneity n.a. Chi2(1)=19.85

Prob>Chi2=0.0000

Source: Amuedo-Dorantes, Georges and Pozo, (2007)

Sources: Population Division of the Department of Economic and Social Affairs of the United Nations Secretariat, Trends in Total Migrant Stock: The 2005 Revision http://esa.un.org/migration, Saturday, April 05, 2008; 8:31:39 AM.

Marc Flandreau and Frédréric Zumer, The Making of Global Finance, 1880-1913, OECD 2004. (Italian GDP data)

Cinel, Dino, “The national integration of Italian return migration, 1870-1929.Cambridge, Cambridge University Press, 1991.