the impact of migration on the sending …...motivation (1) agricultural households in the...
TRANSCRIPT
THE IMPACT OF MIGRATION ON THE SENDING COUNTRY: EVIDENCE FROM THE PHILIPPINES MARJORIE PAJARON
Shorenstein APARC, AHPP Seminar
Introduction
More than 215 million people (3% of world’s population) now live outside their country of birth (UN, 2009)
Introduction
Migration of temporary international Filipino workers or Overseas Filipino Workers (OFWs)
v 2 Million OFWs abroad or 2% of population (Survey on Overseas Filipinos, 2008)
v 51% male
Top Destination of OFWs in 2008
20% Saudi Arabia, 14% in UAE, Singapore, HK, Japan, Qatar, Taiwan 9% Europe; 8% North & South America (Source: SOF, 2008)
Introduction
International remittances – familial transfers
v developing countries received about $372 billion of remittances (World Bank, 2011)
v Philippines ranked 4th most remittance-dependent country - next to India, China and Mexico (WB, 2009)
v major source of foreign exchange reserves (BSP, 2008)
v Source of primary income for most households (FIES, 2003)
v In 2009, Philippines ranked 4th next to India, China, and Mexico (World Bank, 2011)
6
Top developing-country recepients of workers' mittances, 1999-2005
0
5
10
15
20
25
1999 2000 2001 2002 2003 2004 2005
Billi
ons o
f dol
lars
Bangladesh
China
Colombia
El Salvador
Egypt
India
Lebanon
Mexico
Morocco
Pakistan
Philippines
India
China
Mexico
Philippines
7
Source: Bangko Sentral ng Pilipinas, 2007
Sources of Foreign Exchange
-4%-2%0%2%4%6%8%
10%12%14%
2002 2003 2004 2005 2006
Rem as % of GDPFDI as % of GDPPortfoloio I as %GDP
Remi%ances as % of GDP in the Philippines, 2007
Introduction
OFWs referred to as “bagong bayani” or new heroes
30% laborers and unskilled workers (helpers, laborers in construction, mining, and manufacturing)
Research Questions
I. Do farming households depend on their network of family and friends (loans, sale of assets, and transfers) when they encounter a natural disaster?
II. Does migration affect the bargaining power
within the household?
REMITTANCES, MIGRATION, INFORMAL LOANS, AND ASSETS AS RISK-COPING MECHANISMS: EVIDENCE FROM AGRICULTURAL HOUSEHOLDS IN RURAL PHILIPPINES
Research Questions:
- Do agricultural households in rural Philippines insure their consumption against adverse income shocks (rainfall shocks)? (Extended to other types of households)
- If yes, do they use remittances, migration, informal loans, and assets as risk-coping mechanisms?
- Other outcome variables (entrepreneurial activity, HH L supply, children’s L supply, schooling, education, health, and durable goods expenditures)
11
Motivation
(1) Agricultural households in the Philippines face extreme income variation due to natural disasters and non-irrigated land
v 46% of the total irrigable land in the Philippines was irrigated, the remaining still depend on rainfall (NIA, 2006)
v Limited access to formal credit, capital, and insurance markets.
v Poorest among the marginalized and poor groups in the Philippines -- 43% incidence of poverty in 2003 (NSCB, 2003); average household income below that needed by a five-member to meet basic and non-food needs (FIES, 2003)
12
14
- 1948-2004, an average of 20 typhoons occurred in the Philippine Area of Responsibility annually
- Six drought-causing El Niño episodes since 1968 (1968-1969; 1972-1973; 1976-1977; 1982-1983; 1986-1987; 1990-1995; and
1997-1998) Source: Philippine Atmospheric Geophysical and Astronomical Services Administration (PAGASA), 2011
Motivation
(2) Domestic and International Migration v Internal migration unexplored (most data from 1980s or 1990s)
v 16% of household income of agricultural households receiving international remittances; 50% of income of households receiving domestic transfers (FIES, 2006)
15
Existing literature
1) Tests of full consumption insurance: v rejected (Cochrane (1991), Mace (1991), and Townsend (1994)
2) Risk-coping mechanisms v Remittances as insurance (Lucas & Stark 1985, Paulson 2000, Clarke
& Wallsten 2003, Yang & Choi 2007)
v Informal loans as insurance (Platteau & Abraham 1987, Udry1990)
v Assets as insurance (Deaton 1992, Rosenzweig & Wolpin, 1993)
v Combination of different mechanisms (Fafchamps & Lund 2003)
v 77% of agricultural households use one or a combination of these.
16
I. Consumption insurance (Theory)
v Following (Cochrane (1991), Mace (1991), and Townsend (1994) full consumption insurance equation is derived
v Pareto efficient allocation of risk across households. v Household i's consumption depends only on the
networks’ average consumption and sum of Pareto weights.
v Household income does not affect household consumption.
1
1 ln( ) (1/ ) ln( )N
ist st i j
jc c Nλ λ
γ =
⎡ ⎤= + −⎢ ⎥
⎢ ⎥⎣ ⎦∑
18
Empirical Model 1) Consumption insurance (after decomposing income into
permanent and transitory component)
cist= dt + yiT
st + ω i + uit (2)
dt – proxy for average C ω i – household fixed effect proxy for permanent component of income and Pareto weights
OLS biased estimate – reverse causation (IV) and fixed effects (FD) Rainfall shocks = (RF 2006 - RF 2003) in 1000 mm. assigned using
Great Circle Distance.
19
Empirical Model
1) Consumption insurance
Δci2006 = δ0+β1ΔyiT
2006 + β2 ΔXi2006 + β3 d2006*Vi + Δui
2006 (3)
where δ0 = δ0Δd2006 given that dt is equal to one if t = 2006 and zero otherwise.
v P.E. allocation of risk: household income should not affect
household consumption 0<β1< 1 (some degree of insurance) β1 = 0 (full consumption insurance)
21
Data Description
1) Household data Ø Source: Family Income and Expenditure Survey (FIES) Ø 2003 and 2006 panel data Ø 721 agricultural households
2) Rainfall data Ø Philippine Atmospheric Geophysical and Astronomical
Services Administration (PAGASA) Ø 2006 – 2003 rainfall (in 1,000mm.) Ø From 45 weather stations Ø Assigned using GCD (50km-distance from WS, similar
climate type)
22
Outcome Variable Description Agricultural Households
Total Income Change in total annual household income (from 2003 to 2006) as a percentage of income in 2003
0.33 (0.89)
Total Expenditure Change in total annual expenditures of households (from 2003 to 2006) as a percentage of expenditures in 2003
0.28 (0.59)
Rainfall Variables RF 2006 - RF 2003 (1000 mm.)
Change in annual rainfall from 2003 to 2006 (in 1,000 mm.) assigned to households based on their municipalities' climate type and their distance from the nearest weather station, computed using great circle distance.
0.12 (0.60)
RF 2006 - RF mean (1000 mm.)
2006 annual rainfall (assigned to households based on their municipalities' climate type and their distance from the nearest weather station using great circle distance) less mean annual rainfall from 1974 to 2000.
0.15 (0.34)
RF 2003 - RF mean (1000 mm.)
2003 annual rainfall (assigned to households based on their municipalities' climate type and their distance from the nearest weather station using great circle distance) less mean annual rainfall from 1974 to 2000.
0.05 (0.52)
Number of Households 721
Definition, Mean (Standard Deviation) of Outcome and Rainfall Variables
23
2003 Rainfall Deviation from Average Rainfall (1974-2000)
24
25
2006 Rainfall Deviation from Average Rainfall (1974-2000)
Reduced-form
Δyi2006 =ξ1 +ξ2ΔRFi
2006+ξ3ΔXi2006+ξ4d2006*Vi+Δvi
2006 Δci
2006 =δ1 +δ2ΔRFi2006+δ3ΔXi
2006+δ4d2006*Vi+Δui2006
ξ2 ,δ2 < 0 and ξ2> δ2 both adversely affected by RF (negative effect higher for income) then some degree of consumption insurance exists
28
Variables Income ConsumptionRF 2006 - RF 2003 -0.116* -0.098**(1000 mm.) -0.061 -0.046Household Size 0.147*** 0.115***
-0.024 -0.011Age -0.006 -0.001
-0.005 -0.007Education (Elementary omitted) a/
High School -0.044 -0.024-0.087 -0.066
College 0.118 -0.059-0.155 -0.071
Marital Status (Married omitted)Single b/ -0.046 -0.113***
-0.076 -0.036Type of Job (Agriculture omitted) Other Agriculture c/ -0.069 -0.127**
-0.105 -0.056Constant 0.408*** 0.355***
-0.07 -0.053Number of households 721 721
Effect of Rainfall on Household Consumption in Agricultural Households (OLS)
Income decreases by 5.8% while expenditures drop by 4.9% for a 500mm. increase in RF
29
This suggests ~ 16% of C is
insured
Risk-coping mechanisms ci
st = yist + ri
st + tist + li
st + aist
ri
st + tist + li
st + aist = dt + yiT
st + ω i + uit
1
1 ln( ) (1/ ) ln( )N
i i i i ist st st st st st i j
jr t l a y c Nλ λ
γ =
⎡ ⎤+ + + = − + + −⎢ ⎥
⎢ ⎥⎣ ⎦∑
Reduced-form
Δoi2006=π1+π2ΔRFi
2006+π3ΔXi2006+ π4d2006*Vi+Δεi
2006
Δoi2006 is the change in the outcome variable
(international remittances, domestic transfers, loans, net loans, and net assets)
π2> 0 outcome variable used as risk-coping tool
32
Outcome Variable Description Agricultural Households
International Remittance
Change in total annual international remittances (from 2003 to 2006) received as a percentage of income in 2003
-0.01 (0.36)
Domestic Remittance
Change in total annual domestic remittances (from 2003 to 2006) as a percentage of income in 2003
0.01 (0.29)
Net Asset Change in net assets as a percentage of initial annual income in 2003. Net Assets defined as sale less purchase of assets. Assets are either: (a) real assets, which encompass land, real estate, and other personal assets such as jewelry; or (b) financial assets, which include profits from sale of stocks and real assets.
-0.01 (0.39)
Loans Change in total annual loans (from 2003 to 2006) from other families as a percentage of income in 2003.
0.002 (0.11)
Net loans Change in total annual net loans (defined as loans received from other families less loans given to other families as) from 2003 to 2006 as a percentage of income in 2003.
0.002 (0.13)
Number of Households 721
Definition, Mean (Standard Deviation) of Outcome and Rainfall Variables
33
Variables RF 2006 - RF 2003(1,000 mm.)Household Size
Age
High School
College
Single b/
Other Agriculture c/
Constant-0.023 -0.016 -0.011 -0.011
-0.038 -0.026 -0.016 -0.017-0.007 0.023 -0.005 -0.008
Type of Job (Agriculture omitted)0.002 -0.054** 0.027* 0.030*
0.036 -0.023 0.003 0.012-0.064 -0.036 -0.017 -0.018
-0.059 -0.037 -0.019 -0.021Marital Status (Married omitted)
-0.034 -0.014 -0.009 -0.010.015 -0.088** 0.005 0.018
Education (Elementary omitted) a/
-0.016 -0.012 0.006 0.012
0 0.001 0.001 0.001-0.001 -0.001 -0.002 -0.002
0.02 -0.012** 0.001 -0.001-0.012 -0.006 -0.002 -0.002
-0.027 0.059** -0.018** -0.022***-0.018 -0.024 -0.007 -0.007
International Transfers
Domestic Transfers
Loans Net Loans
Effect of Rainfall on Remittances, Loans, and Net Assets in Agricultural Households (OLS )
Domestic transfers increase by about 3%, loans decrease by 2% for a 500-mm increase in RF
34
Compared to estimated coefficient of income, domestic transfers replace 51% of Y decline; loans constitute about 34% of Y decline.
Type of Household Obs. Income Consumption International Remittances
Domestic Transfers Loans Net Loans
(1) (2) (3) (4) (5) (6)Rural Households 1,236 -0.095* -0.080** 0.019 0.033** -0.013** -0.015**
-0.053 -0.036 -0.035 -0.014 -0.005 -0.006Rural Non-Agriculture 515 -0.074 -0.065 0.077 0.003 -0.007 -0.005
-0.054 -0.044 -0.077 -0.011 -0.007 -0.008Urban Households 1,170 0.064 0.004 0.017 -0.009 0 -0.004
-0.085 -0.053 -0.016 -0.009 -0.003 -0.004Urban Agricultural 204 0.096 -0.001 0.012 -0.041 -0.006 -0.007
-0.088 -0.058 -0.062 -0.026 -0.024 -0.026Migrant 124 0.074 0.011 -0.112 0.175 -0.009 -0.017
-0.103 -0.047 -0.081 -0.108 -0.018 -0.019Non-Migrant 597 -0.169** -0.147*** -0.028 0.029 -0.015 -0.019*
-0.067 -0.053 -0.026 -0.024 -0.01 -0.011
Effects of Rainfall on Income, Consumption, and Risk-coping Strategies in Other Types of Households (OLS)
35
Explaining Decrease in Loans
(1) Creditworthiness of borrowers and risk aversion of lenders
- sample split into wealthy and unwealthy
36
Loans Net Loans Loans Net LoansVariables (1) (2) (3) (4)
-0.017** -0.020** -0.001 -0.019(1,000 mm.) -0.007 -0.008 -0.02 -0.021
616 616 105 105R-squared 0.013 0.015 0.275 0.237
Unwealthy Wealthy
RF 2006 - RF 2003
Number of households
(2) Incomes of lenders and borrowers have high covariance, which reduces the effectiveness of local risk-sharing arrangements
- no data available
(3) Loans used instead as ex ante risk-coping mechanism (investment in technologies & crops not susceptible to weather variation)
- test with another period to capture effect of RF on loans during good state of the nature
37
Explaining Decrease in Loans
Robustness Checks
v Summary: partial consumption insurance exits; domestic remittances replace 51% of income decline; loans’ decrease is 34% (inefficient risk-sharing)
v Regress sum of risk-mitigating strategies on RF shocks, insignificant
v Although there exists a change in risk-coping mechanisms the net available resources remain unchanged
38
Other Outcomes
v Because farmers cannot share efficiently share their risks, other sources of income and possible risk-coping strategies are explored:
39
www.textually.org
Other Outcomes
(1) Entrepreneurial activity – wholesale and retail activity (entry and net entry) increased (peddling, sidewalk vending)
(2) L hours (of all members) increase and work more without pay on own family-operated farms or businesses
(3) Children’s L hours and schooling unaffected
(4) Expenditures on education increased by 1% and expenditures on health decreased by 0.5%; durable goods unaffected
40
41
Effects of Rainfall on Entrepreneurial Activities in Agricultural Households (OLS)
Effects of Rainfall on Household Labor Supply in Agricultural Households (OLS)
42
Effects of Rainfall on Children’s Labor Supply and Schooling in Agricultural Households (OLS)
43
Conclusion
(1) Some degree of consumption insurance exists (though small)
(2) Domestic remittances used by agricultural households when rainfall shocks increase
(3) Loans decrease may be due to creditworthiness of borrowers or loans used instead as ex ante risk-coping mechanisms (used during production/planting)
45
Conclusion
(4) Entrepreneurial activity and expenditures on education increase when rainfall increased; health decreased
(5) Because net change in resources is small, still need public transfers and better infrastructure
(6) Improve banking system and informal channels to facilitate international transfers in rural households
46
Limitations and future research
(1) Measure for assets (2) Idiosyncratic shocks (3) Ex ante risk-coping mechanisms
47
Current revision
v 2 Super typhoons in 2006 – Reming (one of 5 strongest since 1948) and Milenyo (one of 5 typhoons that caused the most damage since 1948).
48
www.dilg.gov.ph
THE ROLES OF GENDER AND EDUCATION OF THE HOUSEHOLD HEAD ON THE INTRA-HOUSEHOLD ALLOCATIONS OF REMITTANCES OF FILIPINO MIGRANT WORKERS
(1) In the intrahousehold bargaining literature, how resources are allocated depends on who has “bargaining power” or “say”
à Higher relaFve resources controlled by women results to:
v Higher expenditure shares on food, children’s clothing and educaFon; lower on alcohol and cigare%es
(Hoddino& and Haddad, 1995; Quisumbing and de la Briere, 2000; Quisumbing and Maluccio, 2003; Rubalceva et al. 2004)
v Greater effects on the family’s health and child survival
probabilities in Brazil; improve the health status of children in South Africa; (Thomas, 1990; Duflo, 2003)
Introduction
(2) Policy implicaFons à Affects efficacy of public transfers – PROGRESA (government
anF-‐poverty strategy in Mexico since 1997) (Skoufias and McClafferty, 2001) à Monetary and in-‐kind benefits transferred directly to
mothers (Adato et al., 2000)
ContribuOon: To add to intrahousehold allocaOon literature by incorporaOng migraOon and remi&ances using Philippine data
Introduction
(1) Does gender of the household head (as proxy for bargaining power) ma%er in allocaFon of resources (remi5ances)?
(2) Does educa7on of the household head shows
heterogeneity in allocaFon of remi%ances, keeping gender constant?
Research Questions
(1) Migration à power structure in the household (Chen, 2006)
- women working abroad, higher income - de facto female heads more say in the allocation
(2) Importance of remittances and migration in the Philippines
Motivation
(3) Gender Differentials and Intrahousehold Allocation in the Philippines
à Relatively egalitarian compared to other societies but women still at a disadvantage compared to Filipino men (Eder, 2006)
à Husbands hand wages to wives but wives have limited access to economic assets and limited power in allocaFng resources
v Women’s access to economic assets is indirect (Eviota, 1986) v In poor households, small amount of money to allocate v Unable to refuse requests from husbands for money to drink and gamble (Chant and McIlwain, 1995)
Motivation
Methodology
Empirical Model
cih – expenditure share on ith good by household h
Xh – pce, pce2, hh size, edu of hhh, age of hhh, proportion of demographic groups in the hh, location dummies
rh * gh - captures the importance of gender of household head in allocating remittances H0: β3i = 0, which essentially tests β3im = β3if
cih = βoi +β1irh +β2igh +β3irh *gh +β4iXh +uih
i j h 0i j 1i j j h 2i j j h i j hc r X uβ β β= + + +
H0 : β1im = β1i f
Average Budget Shares for Female-Headed Households with Migrant Spouse
Other Goods, 40%
Food, 37%
Clothing, 8%
Education, 7%
Health, 2%
Household Oper., 3% Non-durable, 0.2% Alcohol and Tobacco, 0.4%
Durable, 3%
Figure 1: Female-headed N = 644
Average Budget Shares for Male-Headed Households with Migrant Spouse
Food , 43%
Other Goods, 34%
Clothing, 8%
Education, 6%
Household Oper., 2%
Alcohol and Tobacco, 3%
Durable, 3% Health, 2%
Non-durable, 0.2%
Figure 2: Male-headed N = 394
Figure 3: Percentage change in budget shares for a 10% change in remittances (households with migrant spouse), Philippines 2003
-0.1%
0.4% 0.3%
0.9% 0.7%
-0.7%
-0.1%
-0.8
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
Male heads (N=394) Female heads (N=644)
(33%)
(194%)
(8%) (13%)
(0.2%) (0.2%)
(13%)
Food Education Health Alcohol & Household Others Tobacco Oper.
Prediction #1: More relative resources controlled by women à more budget shares on
food and children’s health, clothing and education; less on alcohol and cigarettes
2.2%
-5.6% 0.2%
1.6%
-0.3% -1.5% -0.1%
-6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0
Male Highschool (N=394) Female Highschool (N=644)
(1%)
(68%)
(> ave.) (4%)
(23%)
(> ave.) (0.4%)
Food Education Clothing Children’s Durable Non- Other Clothing durable
Figure 4: Predicted difference in budget shares by gender and educational attainment (households with migrant spouse), Philippines 2003
Prediction #2: Better-educated parents invest more in their children’s education; better-educated female heads allocate more on food, education, health, less on A&T.
Figure 5: Percentage change in budget shares for households with other migrant members (non-spouse), Philippines 2003
-0.1%
0.6%
0.1% 0.3%
1.0% 1.0%
-0.5% -0.1%
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
Male Heads, Wife Present (N=730) Female Heads, No Spouse (N=287)
(0.1%)
(14%)
(2%) (47%)
(20%) (28%)
(56%) (0.3%)
Food Education Clothing Children’s Health Alcohol & Other Clothing Tobacco
Predictions #3, #4: The presence of wife affects the allocation decision of male heads. The allocations of female heads who are divorced, widowed, or separated are consistent with those of the de facto female heads.
Additional Test: Less-formally educated female heads who
are divorced, widowed, or separated and less-formally
educated male heads, with wife present, allocate more on
education.
Result: With the presence of wife, male heads allocate more on
children’s clothing.
v Gender of the household head and not gender of the remitter, because of moral
hazard, matters in intrahousehold allocation of remittances
v Consistent with intrahousehold bargaining
literature, female heads increase shares on
education and health; less on alcohol and tobacco
v Regardless of the gender of the head, education shares increase
v The existence of spouse matters in allocation (at least for clothing and
personal items)
v Educational attainment of the head captures heterogeneity in the allocation of
remittances among male-and among female-headed households: less-formally
educated heads value education more.
Conclusion