Download - International Migration and Remittances: Assessing the Impact on Rural Households in El Salvador
International Migration and Remittances: International Migration and Remittances:
Assessing the Impact on Rural Households Assessing the Impact on Rural Households
in El Salvadorin El Salvador
by
Amy Damon
SSEF
July, 2008
Source: http://www.elsalvador.com/noticias/2006/01/12/portada/img/portada3.jpgSource: WDI, 2007
Remittances as a % of GDP 2006
Guatemala 10.2%
Honduras 19.4%
Mexico 2.9%
Nicaragua 12.2%
Panama 0.9%
El Salvador 18.1%
Regional Importance of RemittancesRegional Importance of Remittances
Percent of Households that Receive
Remittances by Municipality, 2004
Source: EHPM 2001 – 2004, Chapter 5 UNDP Human Development Report El Salvador
Previous Literature: Migration TheoryPrevious Literature: Migration Theory
• Migration and development – the Harris-Todaro approach:
– Two sector model where rural to urban labor migration is a result of expected income differences between two sectors.
– Assumes migrants maximize their individual utility by migrating to labor market with highest expected income.
• The new economics of labor migration (NELM):– Addressed assumption that migration is an individualistic
process.– Migration is rational behavior of a group.– Migration is a response not just to wage differentials, but also
relative deprivation.– Migration is a function of missing credit and capital markets.
Research QuestionsResearch Questions
(1) Which households choose to migrate and what determines remittance amounts?
(2) How are household labor decisions affected by migration and remittances?
(3) How are agricultural production, crop choice, and agricultural assets affected by the receipt of remittances?
Data• Four year (1996, 1998, 2000, 2002) panel in El Salvador.
• Collected by Ohio State University and FUSADES.
• 450 households that have information for each year.
• Information on migration, migrants, household characteristics, household production activities, and detailed individual time allocation data.
• Cumulative attrition rate of 28 percent.
• Also EHPM data for community level data and CPS data for US wages and unemployment
Empirical Model for Migration DecisionResearch Question 1
The equation used to predict migration is:
xit is a set of exogenous community and household characteristics
including:
(1) % of households that receive remittances in community
(2) distance of households from a paved road.
(3) other household characteristics
• Estimation Procedure:
– Random effects probit
itiitit xY *
Empirical Results For Determinants of Migration
Explanatory Variables Random Effects Probit Estimates
% of households that receive remittances 0.016***
Distance to Paved Road from the HH (in km) 0.001
Age of HH Head 0.064***
Age of HH Head Squared -0.001***
Number of Senior Citizen Present in HH 0.457***
Land Area (in Ha) 0.021*
Value of Livestock holdings/1000 0.043
Constant -2.618***
Total Sample over 4 years 1303
Number of Households in each year 449
Standard errors excluded for presentation – see paper.
* significant at 10%; ** significant at 5%; *** significant at 1%
Explaining Remittances (with panel data)
Remittance Equation:
Jit = Xitα1 + Zitα1 + εit
Xit is a set of household characteristics that influence the level of remittances
Zit is the wage rate and the unemployment rate in the destination U.S.A. city
εit is a normally distributed error term
Estimation Procedures: (1) Household Fixed Effects Model (2) Heckman Model
Empirical Results – Explaining Remittance Amounts Household Fixed Effects Heckman
Regression Selection Equation
Unemployment rate in destination city -9,455.365* -4029.88
US Wage in destination city 8.60*** 8.01***
Age of HH Head 86.7 23.9
Age of HH Head Squared -0.81 -0.21
Dependency Ratio -121.84 -46.62
Number of Senior Citizen Present in HH 9.03 25.22
Female Headed HH 803.30** 550.58***
Number of HH Members -89.89 -48.81
Number of Children Present in HH 138.48 59.24
Land Area in HA 8.16 13.87
Value of Livestock holdings/1000 -28.953* -12.11
ES wage -66.85 46.32
ES transfers 0.065* 0.071***
Constant -2668.24 -1122.88 -0.652***
% of households that receive remittances 0.016***
Distance to Paved Road from the HH (in km) 0
Observations 502 1528 1528
Number of Households 268
Diagnostics
LR test of independent equations (prob > chi2) 0.0004
* significant at 10%; ** significant at 5%; *** significant at 1%
Explaining Remittances 2002 Cross-Section
• Objective: to look at gender and relationship to the household head effects using 2002 data
The Remittance equation is:
Ji = α1wusai + α1Nusa
i + α2Xi + ui
Xi is a set of household characteristics
Estimation Procedure:• Heckman Selection Model
Cross-Sectional Remittance Results-2002 (1) OLS (2) OLS
US Wage 7.564 10.859**
US Unemployment Rate 38.365 62.777
Migrant is a Female (=1 if migrant is female) 312.806 768.446**
Migrant is Son of HH Head 1,229.727***
Migrant is Daughter of HH Head 1,445.923***
Migrant is Brother of HH Head 243.409
Migrant is Sister of HH Head 1,469.658***
Migrant is HH Head 2,553.755***
Age of HH Head -18.257* -7.328
Dependency Ratio 326.056 -681.660*
Number of Senior Citizen Present in HH -290.915 -35.73
Female Headed HH -143.23 781.652***
Number of HH Members 109.53 -3.406
Number of Children Present in HH -80.238 143.834
Constant -2,161.75 -2,093.93
Observations 413 413
R-squared 0.26 0.05
Intuition for Question 2:Work Hours and Remittances
• If a household operates in a perfectly functioning market environment (complete credit and labor markets:
– An increase in remittances will increase consumption– Separability holds (production and consumption decisions are
independent of one another)– Remittances will not affect labor allocation outcomes.
• But if a household is credit constrained:– Migration and remittances may substitute for missing credit or
insurance markets.– Separability no longer holds and migration and remittances will
impact on-farm and off-farm labor allocation decisions and investment decisions.
Labor and Migration: Theoretical Model
)];();([);();( 2211 ZCUZCUZCUZCU FFMMFFMM },,,,{ BKMRC ttt
it
MMusa
FFF CTNwBMwGKKRfrKC 111110111 )1(),,(
MMusa
FFF CTNwMwGKKKRfBrrKC 2222210222 )1(),,()1(
FFF MRT 111
FFF MRT 222
Max
subject to:
)( 11 BB
In the credit constrained version
Comparative Static ResultsComparative Static ResultsHow choice variables change with an increase in remittances
No Credit Constraint
Consumption
On-Farm Work
Off-Farm Work
Capital
Credit Constrained
01 usdw
dC
01 usdw
dR
01 usdw
dM
01 dJ
dC
01 dJ
dR
01 dJ
dM
01 usdw
dK 01 dJ
dK
Labor Supply EstimationLabor Supply EstimationThe labor supply equation of interest is:
Hit: measure of (change in) labor hours Xit: set of household demographic change variables Jit: (change in) predicted level of remittances a household receives Migr: (change in) predicted migration εi: aggregate error term assumed to be white noise But…..Mig and Jit is endogenous so we use an instrumental variable (2sls) approach.
Instruments are: (1) % of hh that receive remittances in community (2) in USA wage rate (3) Unemployment rate in USA and (4) Household distance to a paved road
Estimation Procedures: First Differences Model (also household fixed effects estimation - see paper)
1,1,31,211,1, tttittittittjt MigrJXH
Types of Labor Examined
• Total Household Labor
• Total Farm labor
• On-Farm– Male, Female, Child, Hired
• Off-Farm Wage Labor– Male, Female, Child
• Non-agricultural Self-Employment– Male, Female
• Household Work
Total Labor and On-Farm Labor First Differences Model On-Farm Work
Total Hours Total Farm Hours Female Male Child Hired
Remittances 0.588 0.02 -0.008 0.042 0.035 -0.051
Migration Status -46.583 2,090.687*** 226.989* 994.693** 206.615* 707.411
Land Area -1.003 18.909 1.194 7.298 -0.017 10.731
No. of Senior Citizens in HH 39.345 -162.803 -20.836 -4.925 -24.796 -117.417
Female Head Status -1,493.411* -663.814 64.344 -347.248 -48.853 -333.75
Number of HH Members 340.032** 119.556 0.334 92.435 0.201 29.511
No. of HH Children 105.903 -183.797 -39.579 -45.614 1.964 -107.867
Livestock Value -0.099*** -0.067*** 0 -0.003 0.003 -0.067***
Dependency Ratio -86.563 -82.855 -8.73 -114.268 3.475 42.416
ES Wage 148.476** 5.312 -4.424 -1.525 1.133 9.721
ES Transfers -0.07 -0.028 -0.006 -0.016 -0.007 0.001
Constant -642.775 -506.308** -72.225* -286.165* -71.476* -81.982
Observations 180 181 180 181 181 181
* significant at 10%; ** significant at 5%; *** significant at 1%
Off-Farm Work Results First Differences Model Female Male Child
Remittances 0.192 0.444 0.06
Migration Status -531.47 -1,981.864** 678.682
Land Area 25.817 -42.047 5.79
No. of Senior Citizens in HH 185.444* 107.268 110.804
Female Head Status 219.067 -1,017.054* -253.132
Number of HH Members 35.717 181.948 65.19
No. of HH Children 128.106 131.897 -30.431
Livestock Value -0.003 -0.021 -0.005
Dependency Ratio -121.874 -105.934 -197.56
ES Wage 50.032** 86.483** 35.955
ES Transfers -0.017 -0.033 -0.014
Constant 171.401 -161.757 20.976
Observations 181 181 181
* significant at 10%; ** significant at 5%; *** significant at 1%
Other Work Non-Agricultural Self-Employment Housework
First Differences Model Female Male
Remittances -0.019 -0.046 0.138
Migration Status 282.747 -160.807 1,923.663**
Land Area 0.485 -3.946 -1.95
No. of Senior Citizens in HH -32.564 32.569 -47.074
Female Head Status -49.612 109.035 -540.245
Number of HH Members -50.53 13.215 139.699
No. of HH Children -0.185 2.201 -51.773
Livestock Value -0.008 0.002 -0.017
Dependency Ratio 146.236 1.249 -256.25
ES Wage 12.935 12.126 23.937
ES Transfers -0.002 0.009 -0.031
Constant -60.699 42.013 -607.293**
Observations 181 181 181
* significant at 10%; ** significant at 5%; *** significant at 1%
Question 3.
How are agricultural production activities affected by migration and remittances?
Intuition and Literature
• Many studies have examined the relationship between farm income safety nets (migration and remittances) and agricultural outcomes such as cropping patterns (Smith and Goodwin, 1996; and Babcock and Hennessey, 1996).
• Chavas and Holt (1990) examine how farmers allocate acreage to different crops under risk and find that both risk and wealth are important in corn-soybean acreage decisions.
• Since migration and remittances are a form of insurance (Stark and Lucas, 1988; Gubert, 2002; Stark and Lucas, 1988; Cox et al., 1998); do migration and remittances affect risk behavior or crop acreage decisions?
TheoryTheoryTheoretical model suggests that the change in
land use in response to wealth depends on the risk preferences of the household.
– Constant absolute risk aversion implies no change in acreage with a change in wealth
– Decreasing absolute risk aversion means they will move into riskier crops as wealth increases
– Increasing absolute risk aversion households would move into less risky crops as wealth increases.
Empirical Approach
• Risk is measured by coefficient of variation (CV) for crop and livestock revenue:
• And explained using household characteristics and remittances (migration):
CV
tjtjtj RCV 21X
σ is standard deviation of total farm revenue for farm j across four years of data
μ is the mean of total farm revenue
Explanatory Variables Remittances Migration
Remittances / 1000 0.272
Migration Status (0/1) -0.38
Total Land Area (ha) -0.022* -0.028***
Livestock Value -0.095 -0.034
Number of household senior citizens -0.138 0.086
Female headed household 0.049 0.297***
Age of household head -0.004 -0.006
Number of household members 0.054 0.005
Number of household children -0.13 -0.021
Dependency Ratio 0.242 -0.016
Salvadoran agricultural wage rate 0.003 0.029**
Minutes to a paved road 0 -0.001
Constant 1.139*** 1.514***
Observations 248 391
R-squared .13
* significant at 10%; ** significant at 5%; *** significant at 1%
IV Regression explaining agricultural revenue coefficient of variation
Explaining acreage decisions
titititijt RYXA 431
α is the household fixed effect,
Xit is a vector of household demographic characteristics,
Yit is total land area,
Rit is remittances (replaced by the dichotomous variable, MIGRit, in the migration version of this regression),
εt is an independently distributed error term
Estimated using a household fixed effects model instrumenting for migration and remittances.
Land use in hectares
House
lot PastureFallow/Forest Cultivation
Basic grains Coffee
Other cash crops
Migrant Status (0/1) 0.494** 0.191 -0.023 -0.278 1.461*** 0.24 -1.226**
Total Land Area (ha) 0.002 0.468*** 0.291*** 0.078*** 0.025***0.005*
* 0.038***
Number of HH Senior Citizens 0.032 -0.08 0.138 0.061 -0.106 -0.018 0.074
Female Headed HH -0.083 -0.062 0.238 -0.331 -0.449*** -0.081 0.084
Number of HH Members 0.016 -0.052 0.072 0.024 0.077*** 0.013 -0.052
Number of HH Children -0.018 0.033 -0.02 0.008 -0.062** -0.011 0.041
Value of Total Livestock 0 0*** 0 0 0** 0 0***
Salvadoran Wage Rate -0.002 -0.032 0.01 0.012 0.009 0 -0.004
Constant -0.174 -0.09 -0.719 0.434 -0.57** -0.083 0.895***
Observations 1279 1727 1727 1727 1727 1727 1727
Number of households 449 449 449 449 449 449 449
Fixed-effects instrumental variable regression explaining land use by migration status.
* significant at 10%; ** significant at 5%; *** significant at 1%
Asset Holdings and Land Rental Markets
Land Area
Land Rented In
Land Rented Out
Migrant Status 3.675** 0.652* 1.712**
Number of Senior Citizens -0.525 -0.124 -0.215
Female Headed Household -1.654** -0.262* -0.265
Years of Education of the Head 0.005 0.015 -0.001
Age of the HH Head 0.024 0.007 -0.007
Number of HH members 0.197 0.053* 0.067
Number of HH children -0.489* -0.091 -0.219*
Dependency Ratio 1.360* 0.195 0.699**
Constant -1.661 -0.566* -0.535
Observations 1253 1253 1253
Number of households 448 448 448
* significant at 10%; ** significant at 5%; *** significant at 1%
Concluding Points• It is the act of migration rather than remittances that change
household behavior.
• Migrant households allocated their labor back to the farm when they send out a migrant.
• When female migrants’ wages increase they send more money. Males appear to send less.
• Migrant households allocated more land to “food security” crops rather than other crops or cash crops.
• Migrant households do not appear to undertake riskier crops (in terms of revenue).
• Migrant household have larger land holdings and have larger land areas involved in rental markets.