International Migration

1
International Migration
2
International migration the basic numbers

• Between 1965 and 2000, the fraction of people
  living outside their countries of birth increased
  from 2.2 to 2.9 of world population
• 175 million people in 2000
• Remittances have been huge and growing

Sources UN International Migration Report 2002,
US Census Bureau.
3
Remittances vs. ODA, FDI(1991-2005)

• Source World Development Indicators 2007. Data
  are in current US.

4
Key questions

• How does emigration to the rich world affect the
  economic outcomes of migrants themselves?
• How does emigration to the rich world affect
  families left behind, and origin-country
  economies more broadly?
• What policies might help raise the development
  impact of remittances?

5
Clemens and Pritchett (2008)

• A new statistic, income per natural the mean
  annual income of persons born in a given country,
  regardless of where that person now resides
• Almost 43 million people live in a group of
  countries whose income per natural collectively
  is 50 higher than GDP per resident
• For 1.1 billion people the difference exceeds 10
• Poverty estimates are very different for national
  residents and naturals
• 26 percent of Haitian naturals who are not poor
  by the two-dollar-a-day standard live in the
  United States
• If economic development is defined as rising
  human well being, then a residence-neutral
  measure of well-being emphasizes that crossing
  international borders is not an alternative to
  economic development, it is economic
  development.

6
Table 2b
7
Table 2b
8
Causal effect of migration

• But what is the causal effect of migration?
• Observed income differences could reflect
• Labor-supply effect
• Selection effect
• Causal effect of migration
• Clemens and Pritchett (2008) review a variety of
  studies finding that at least 75 of the income
  difference is the causal effect of migration
• Next examine one important study providing
  estimated causal impact of migration on income
• Gibson, McKenzie, and Stillman (2006)

9
Gibson, McKenzie, and Stillman (2006)

• What is the impact of migration on the migrants?
• One of the most difficult problems in migration
  studies, due to selectivity of migration
• Very difficult to find an appropriate or
  convincing control group
• This paper use a lottery to obtain exogenous
  variation in migration
• Lottery for Tongans to migrate to New Zealand
• Estimate income gains from migration, using
  lottery as exogenous variation
• Compare estimates from lottery to OLS,
  single-difference, double-difference, IV,
  propensity score matching
• Results IV with a good instrument performs
  best
• D-in-D and propensity score matching do OK as
  well

10
Table 3
11
Table 4
12
Table 5
13
Table 6
14
Key questions

• How does emigration to the rich world affect the
  economic outcomes of migrants themselves?
• How does emigration to the rich world affect
  families left behind, and origin-country
  economies more broadly?
• What policies might help raise the development
  impact of remittances?

15
Yang (2008)

• What are impacts on families left behind?
• Specifically, what do remittances help pay for at
  home?
• Another natural experiment
• June 1997 6 of households in the Philippines
  had one or more members working overseas, in
  dozens of countries
• July 1997 Asian financial crisis occurs
• Large, sudden, heterogeneous changes in exchange
  rates in many locations of Filipino workers (see
  figure)
• Philippine peso also depreciates
• Examine impact of migrant exchange rate shocks on
  Philippine households

16
Distribution of Overseas Workers from
Philippines(June 1997)
17
Exchange rates over time
Exchange Rates in Selected Locations of Overseas
Filipinos (Jul 1996 - Oct 1998)
US, Hong Kong, Saudi Arabia, Kuwait
Start of Asian financial crisis (July 1997)
Japan
Singapore
Taiwan
Philippine pesos per unit of foreign currency
(July 1996 1)
Malaysia
Korea
18
Figure 1 Impact of migrant exchange rate shocks
on Philippine household remittance receipts
(1997-1998)
Percent change in mean remittances
Percent change in exchange rate
NOTESExchange rates are in Philippine pesos per
unit of foreign currency. Percent change in
exchange rate is mean exchange rate from Oct 1997
to Sep 1998 minus mean exchange rate from July
1996 to June 1997, divided by the latter. Mean
remittances are calculated among all households
with a single migrant in given overseas location.
Percent change in mean remittances is between Jan
- Jun 1997 and Apr-Sep 1998 reporting periods.
Datapoints are the top 20 locations of Philippine
overseas workers (as listed in Table 1).
19
Impact of migrant shocks on households
20
Key questions

• How does emigration to the rich world affect the
  economic outcomes of migrants themselves?
• How does emigration to the rich world affect
  families left behind, and origin-country
  economies more broadly?
• What policies might help raise the development
  impact of remittances?

21
Raising development impact of remittances

• An open area
• Increasing migrant control over remittance uses
• El Salvador Study of Migrant Families
• Remittances for microenterprise finance
• Migrant-backed loans
• Migrant co-signed loans
• Other ideas?

22
Intro slide
Remittances and The Problem of Control A Field
Experiment Among Migrants from El Salvador
Nava Ashraf, Harvard Business School Diego
Aycinena, Francisco Marroquin University Claudia
Martinez, University of Chile Dean Yang,
University of Michigan
23
DC-area Salvadorans on the problem of control

• I have many uncles and they get drunk, so I
  just send money when needed, or I send to someone
  like my sister who I trust.
• Male, 34 years old, 8 months in the U.S., works
  as roofer
• The brother of my boss sent around 50,000 to
  his mother over the years. When he thought he had
  enough money to build a house, he asked his mom
  for the money. She said she didn't have it. She
  had lent it to an uncle. When he asked for the
  money back, the uncle threatened to kill him if
  he came back to El Salvador for the money.
• Male, 30 years old, 1 year in the U.S., works as
  a carpenter

24
Some open questions about remittances

• Do migrants and recipients typically agree on the
  uses to which remittances should be put?
• If not, how do such disagreements affect
• the choice of remittance recipients?
• amounts remitted?
• remittance uses?
• If migrants were to be given more control over
  remittance uses
• would they remit more?
• how would they direct them to be used?
• what would be the impact on household-level
  development outcomes?

25
Remittances and savings

• This research focuses on the control that
  migrants have over how much of remittances are
  saved
• Migrants frequently report wanting household to
  save some fraction of remittances
• Savings potentially intended for use of either
  recipient or migrant
• Migrants report stronger preferences that
  remittances be saved, compared to recipients
• Migrants have little or no ability to control or
  monitor household savings in El Salvador
• Can only request that household save a portion of
  cash received

26
Why would migrants save in home country?

• If savings intended for regular use of family
  back home, U.S.-based accounts are inconvenient
• If savings intended for U.S.-based migrant, many
  migrants consider U.S. savings insecure
• Undocumented migrants fear they would lose
  savings in the U.S. if they were deported
• Currently an open question whether savings
  accounts are primarily for recipients or
  migrants use

27
The experiment

• We offered Salvadoran migrants in Washington,
  D.C. the ability to directly channel remittances
  into savings accounts in El Salvador
• Facilities developed for project in partnership
  with Salvadoran bank, previously not widely
  available
• We randomly varied the degree of migrant control
  over accounts offered
• Outcomes of interest
• Demand for accounts (take-up)
• Savings accumulation
• Remittances (identity of recipients, amounts)
• Later household outcomes
• E.g., consumption, schooling, entrepreneurial
  investment, housing

28
Theoretical impacts

• How should remittances change if migrants had
  more control over savings in the home country?
• Case 1 funds saved out of remittances intended
  for use of recipient household
• If migrant and recipient have similar savings
  preferences, increasing migrant control should
  have little or no effect
• If migrants prefer recipient to save more, then
  increasing migrant control should lead to higher
  remittances and savings
• Case 2 funds saved out of remittances intended
  for use of remitter (migrant)
• Funds entrusted to remittance recipient, but
  monitoring is imperfect a principal-agent
  problem
• Increasing migrant control over savings could
  lead to higher savings (by migrant), but lower
  remittance flows to family

29
Treatments

• 0. Control group
• Migrants encouraged to remit into a household
  members bank account, but no account-opening
  assistance provided
• 1. Account for remittance recipient in El
  Salvador
• Migrants encouraged to remit into an individuals
  bank account
• If no such account exists, offer to help
  household set up account
• Migrant cannot check balance or withdraw
• 2. Joint account (for migrant and household)
• Migrants encouraged to remit into shared account
• New product Cuenta Unidos
• Migrant and hh each have ATM cards migrant can
  check balance
• 3. Individual migrant account
• Migrants encouraged to remit into own account
• New product Ahorro Directo
• Only migrant has ATM card
• Not shared with household

30
Marketing brochures

• Ahorro Directo Cuenta Unidos

31
Rationales for treatment conditions

• 0. Control group Provides counterfactual for
  assessing impact of offered savings facilities on
  later household outcomes (savings, remittances,
  consumption, investment, etc.)
• Account for remittance recipient Provides
  counterfactual of demand for accounts where
  migrants have no control
• If no differences vs. 2 and 3, indicates migrants
  dont value control
• 2. Joint account (for migrant and household)
  Impact of having shared control over account
• Monitoring of account balance
• Shared ownership
• But El Salvador joint owner has full ability to
  withdraw
• 3. Individual migrant account Impact of
  exclusive control over account

32
Hypotheses to be tested

• Hypothesis 1 Take-up of new savings products
  will be
• Highest for Treatment 3 (individual migrant
  account)
• Next-highest for Treatment 2 (joint account)
• Lowest for Treatment 1 (account for remittance
  recipient)
• Difference should reflect value migrants place on
  control over savings
• Hypothesis 2 Growth in migrant savings will be
• Highest for Treatment 3
• Next-highest for Treatment 2
• Next-highest for Treatment 1
• Smallest for Treatment 0
• In later rounds, will examine effects on other
  household investment activities, such as
  schooling, health spending, entrepreneurship

33
Overview of treatment protocols

• Migrants recruited at Salvadoran consulates and
  Banagricola remittance agencies
• Must have remitted to someone in El Salvador in
  last 12 months, and have been in U.S. for lt15
  years
• Subsample of migrants and recipient households
  administered comprehensive baseline survey
• Migrants randomly assigned to one of 4
  experimental conditions
• After stratification by gender, US account
  status, years in US (3 categories), and
  relationship to recipient (4 categories)
• Marketing team member visits each migrant in
  person to administer treatment
• Visits take place in location of migrants choice
• Typically home, workplace, restaurant, nearest
  Banagricola agency
• Some visits arranged in advance, others occur on
  the spot

34
ESSMF marketing team
35
Marketing visit (1)
36
Marketing visit (2)
37
Equalizing transaction costs

• Account-opening costs
• For all account types, account opening requires
  visit by remittance recipient to a Banco Agricola
  branch in El Salvador
• Remittance transfer cost
• All accounts have equal cost of inbound
  remittance
• Inbound remittance costs also equalized with cash
  remittance

38
Remittance price randomization

• Remittance prices randomized between 4 and 9
• For a remittance up to 1500
• VIP card used to track remittances
• Will examine price elasticity of remittances

39
Baseline survey contents

• Demographics, household composition
• Employment and income
• Detailed consumption in US and El Salvador
• Use of financial services
• Savings
• Remittances
• Credit
• Financial literacy, planning
• Communication and conflict with family
• Migration history

40
Migrant survey Washington DC
41
Household survey El Salvador
42
Basic summary statistics

• Demographics
• 29 female
• Mean age 30.8
• Mean years in US 5.47
• Employment and income
• Top employment categories are construction (32),
  food services (15), cleaning services (10)
• For migrant, median annual earnings is 17,945
  for surveyed individual and 25,458 for household
• For recipient household, median annual income
  2,864
• Remittances
• Most common remittance amount sent 200 (25)
• Median annual remittances 3,900
• Median remittances as share of migrant hh income
  14.4

43
Sources of income in recipient households
44
Communication with family

• 86 of migrants have a cellphone
• 94 of migrants have not visited in the past 3
  years (proxy for legal status)
• 81 of migrants communicate at least once a week

45
Methods of communicating with family
46
Savings migrant

• 69 have a savings account
• 51 have one in US only
• 9 have one in El Salvador only
• 9 have one in both countries
• But savings are quite low for most migrants
• Median savings 700
• Median savings as of annual hh income 2.8
• Migrants express desire for more savings
• 37 very unsatisfied or unsatisfied with
  current level of savings
• 85 would open account in El Salvador if given
  opportunity to do so

47
Savings recipient household

• Only 21 have a savings account
• Savings are low
• Median savings 0 (mean savings 301)

48
Revealing preferences for remittance uses

• Goal Reveal via survey answers whether migrants
  and households differ in their preferences over
  how remittances are used
• Problem simply asking migrants and households
  about their expenditure preferences may not yield
  useful answers
• Answers may be automatic, conditioned by what
  respondents think is the right answer
• Respondents may not think carefully (in way they
  would if actual money were at stake)
• Our approach enter respondents into a
  remittance raffle
• Winning family in El Salvador will receive
  remittance of 100
• Migrants specify how they would like the money to
  be used by recipients
• Household respondent specifies how they would
  like the money to be used when received
• 13 categories of expenditures
• Cash is not an option

49
Raffle use categories

• Daily consumption
• Clothing
• Housing (includes rent, construction, mortgage)
• Medical expenditures
• Educational expenses
• Utilities bills
• Phone bills
• Agricultural inputs
• Small business expenses
• Savings
• Durable goods
• Automobile payments
• Other (specify)
• Question Do migrants allocate more money to
  certain expenditures than the corresponding
  remittance-receiving households?

50
Allocations of 100 raffle winnings
Migrant
51
Allocations of 100 raffle winnings
Migrant
Remittance recipient
52
Migrant vs. household raffle winnings allocation
(US)
Raffle use categories Migrant (in U.S.) Remittance Recipient (in El Salvador) Difference (migrant minus recipient allocation) P-value test of equality of means

Daily consumption 43.96 63.76 -19.80 0.000
Savings 17.49 2.45 15.04 0.000
Clothing 6.91 6.05 0.86 0.296
Housing 3.79 2.18 1.61 0.013
Medical expenditures 9.01 7.83 1.18 0.235
Educational expenses 4.07 6.26 -2.19 0.003
Utilities bills 3.36 3.53 -0.17 0.785
Small business expenses 2.93 1.00 1.93 0.001
Phone bills 1.36 0.55 0.81 0.023
Agricultural inputs 0.25 0.35 -0.10 0.638
Durable goods 0.41 1.11 -0.70 0.041
Automobile payments 0.08 0.00 0.08 0.318
Other (specify) 6.37 4.92 1.46 0.103

Num. obs. 1,208 1,208
53
Balance across treatment groups
54
Analysis of take-up decision

• Definition of take-up migrant fills out
  account-opening forms during marketing visit
• Focus on migrants offered Treatments 1, 2, or 3
• Account opening for Treatment 0 requires analysis
  of internal bank data or follow-up survey

55
Percentage opening accounts, by treatment
56
Regression specification

• For migrant i
• Yi a b Z2i g Z3i Xif ei
• Yi takeup indicator
• Z2i treatment 2 indicator
• Z3i treatment 3 indicator
• Xi vector of control variables (collected at
  baseline)

57
Impact of treatments on account opening
Dependent variable Migrant filled out account-opening forms during marketing visit Dependent variable Migrant filled out account-opening forms during marketing visit Dependent variable Migrant filled out account-opening forms during marketing visit Dependent variable Migrant filled out account-opening forms during marketing visit Dependent variable Migrant filled out account-opening forms during marketing visit

(1) (2) (3) (4)

Treatment 2 (joint account) 0.148 0.148 0.169 0.172
(0.044) (0.044) (0.060) (0.060)
Treatment 3 (joint account 0.200 0.196 0.149 0.150
indiv. migrant account) (0.045) (0.044) (0.060) (0.061)
Constant 0.331 0.445 0.472 0.759
(0.032) (0.046) (0.061) (0.192)

Controls for stratification variables Y Y Y
Additional migrant, recipient controls Y

Observations 732 732 408 408
R-squared 0.028 0.078 0.055 0.085

P-value of F-test of equality of 0.234 0.275 0.742 0.716
Treatment 2 and 3 coefficients

significant at 10 significant at 5 significant at 1 significant at 10 significant at 5 significant at 1 significant at 10 significant at 5 significant at 1
58
Impact of treatments on account opening
Dependent variable Migrant filled out account-opening forms during marketing visit Dependent variable Migrant filled out account-opening forms during marketing visit Dependent variable Migrant filled out account-opening forms during marketing visit Dependent variable Migrant filled out account-opening forms during marketing visit Dependent variable Migrant filled out account-opening forms during marketing visit

(1) (2) (3) (4)

Treatment 2 (joint account) 0.148 0.148 0.169 0.172
(0.044) (0.044) (0.060) (0.060)
Treatment 3 (joint account 0.200 0.196 0.149 0.150
indiv. migrant account) (0.045) (0.044) (0.060) (0.061)
Constant 0.331 0.445 0.472 0.759
(0.032) (0.046) (0.061) (0.192)

Controls for stratification variables Y Y Y
Additional migrant, recipient controls Y

Observations 732 732 408 408
R-squared 0.028 0.078 0.055 0.085

P-value of F-test of equality of 0.234 0.275 0.742 0.716
Treatment 2 and 3 coefficients

significant at 10 significant at 5 significant at 1 significant at 10 significant at 5 significant at 1 significant at 10 significant at 5 significant at 1
59
Impact of treatments on account opening
Dependent variable Migrant filled out account-opening forms during marketing visit Dependent variable Migrant filled out account-opening forms during marketing visit Dependent variable Migrant filled out account-opening forms during marketing visit Dependent variable Migrant filled out account-opening forms during marketing visit Dependent variable Migrant filled out account-opening forms during marketing visit

(1) (2) (3) (4)

Treatment 2 (joint account) 0.148 0.148 0.169 0.172
(0.044) (0.044) (0.060) (0.060)
Treatment 3 (joint account 0.200 0.196 0.149 0.150
indiv. migrant account) (0.045) (0.044) (0.060) (0.061)
Constant 0.331 0.445 0.472 0.759
(0.032) (0.046) (0.061) (0.192)

Controls for stratification variables Y Y Y
Additional migrant, recipient controls Y

Observations 732 732 408 408
R-squared 0.028 0.078 0.055 0.085

P-value of F-test of equality of 0.234 0.275 0.742 0.716
Treatment 2 and 3 coefficients

significant at 10 significant at 5 significant at 1 significant at 10 significant at 5 significant at 1 significant at 10 significant at 5 significant at 1
60
Breakdown of accounts in Treatment 3
Percentages out of 129 individuals opening
accounts in Treatment 3 as of May 19,2008.
61
Other correlates of account opening
Dependent variable Migrant filled out account-opening forms during marketing visit Dependent variable Migrant filled out account-opening forms during marketing visit Dependent variable Migrant filled out account-opening forms during marketing visit Dependent variable Migrant filled out account-opening forms during marketing visit Dependent variable Migrant filled out account-opening forms during marketing visit

Treatment 2 (joint account) 0.148
(0.044)
Treatment 3 (joint account 0.196
indiv. migrant account) (0.044)
Migrant is female -0.054
(0.040)
Migrant has US bank account -0.055
(0.039)
Migrant has been in US 6-10 years -0.129
(0.041)
Migrant has been in US 11-15 years -0.289
(0.059)
Recipient is migrant's spouse 0.010
(0.058)
Recipient is migrant's child 0.051
(0.087)
Recipient is migrant's other relative 0.048
(0.040)
Constant 0.445
(0.046)

Observations 732
R-squared 0.078

P-value of F-test of equality of 0.275
Treatment 2 and 3 coefficients

significant at 10 significant at 5 significant at 1 significant at 10 significant at 5 significant at 1 significant at 10 significant at 5 significant at 1
62
Other correlates of account opening (2)
Dependent variable Migrant filled out account-opening forms during marketing visit Dependent variable Migrant filled out account-opening forms during marketing visit Dependent variable Migrant filled out account-opening forms during marketing visit Dependent variable Migrant filled out account-opening forms during marketing visit

Treatment 2 (joint account) 0.172 Migrant's years of education 0.003
(0.060) (0.005)
Treatment 3 (joint account 0.150 Migrant's raffle allocation to savings 0.000
indiv. migrant account) (0.061) (0.001)
Migrant is female -0.037 Migrant plans for retirement 0.060
(0.059) (0.054)
Migrant has US bank account -0.028 Log (migrant's annual income) -0.009
(0.053) (0.013)
Migrant has been in US 6-10 years -0.116 Recipient's years of education 0.001
(0.055) (0.005)
Migrant has been in US 11-15 years -0.195 Recipient's raffle allocation to savings 0.000
(0.091) (0.002)
Recipient is migrant's spouse 0.035 Recipient plans for retirement 0.051
(0.088) (0.076)
Recipient is migrant's child 0.112 Log (recipient's hh income) -0.031
(0.142) (0.018)
Recipient is migrant's other relative 0.093 Recipient has bank account -0.163
(0.058) (0.067)
Constant 0.759
(0.192)

Observations 408
R-squared 0.085

P-value of F-test of equality of 0.716
Treatment 2 and 3 coefficients

significant at 10 significant at 5 significant at 1 significant at 10 significant at 5 significant at 1 significant at 10 significant at 5 significant at 1
63
Evidence for migrant control?

• Higher take-up in Treatment 2 and 3 vs. 1 may
  indicate
• Migrant desire for control, or
• Migrant desire for convenience (easy access to
  funds from U.S.)
• Take-up is not statistically significantly higher
  in Treatment 3 vs. 2
• But 95 of migrants opening accounts in treatment
  3 open Ahorro Directo (individual migrant
  account)
• Highly suggestive of migrant desire for control
• Data still to come will shed further light
  assess whether total savings balances are
  significantly higher in Treatment 3 vs. 2

64
Heterogeneity in treatment effect
65
Heterogeneity in treatment effect (2)
Dependent variable Migrant filled out account-opening forms during marketing visit Dependent variable Migrant filled out account-opening forms during marketing visit Dependent variable Migrant filled out account-opening forms during marketing visit

(2) (3)

Treatment 2 or 3 0.302 0.247
(0.106) (0.116)
(T. 2 or 3) Migrant has US bank account -0.225
(0.113)
(T. 2 or 3) Migrant has US bank account only -0.110
(0.128)
(T. 2 or 3) Migrant has ES bank account only 0.187
(0.187)
(T. 2 or 3) Migrant has both US and ES bank account -0.458
(0.198)
(T. 2 or 3) Migrant is female 0.010 0.022
(0.118) (0.118)
(T. 2 or 3) Migrant has been in US 6-10 years -0.054 -0.065
(0.118) (0.118)
(T. 2 or 3) Migrant has been in US 11-15 years 0.308 0.257
(0.202) (0.204)
(T. 2 or 3) Recipient is migrant's spouse -0.160 -0.146
(0.186) (0.188)
(T. 2 or 3) Recipient is migrant's child -0.104 -0.127
(0.268) (0.269)
(T. 2 or 3) Recipient is migrant's other relative 0.001 0.008
(0.117) (0.117)

Observations 408 408
R-squared 0.075 0.089

significant at 10 significant at 5 significant at 1 significant at 10 significant at 5 significant at 1
66
Heterogeneity in treatment effect (3)
Dependent variable Migrant filled out account-opening forms during marketing visit Dependent variable Migrant filled out account-opening forms during marketing visit Dependent variable Migrant filled out account-opening forms during marketing visit

Treatment 2 or 3 0.032
(0.391)
(T. 2 or 3) Migrant has US bank account only -0.064 (T. 2 or 3) Migrant's raffle allocation to savings -0.000
(0.135) (0.001)
(T. 2 or 3) Migrant has ES bank account only 0.194 (T. 2 or 3) Migrant plans for retirement 0.023
(0.197) (0.120)
(T. 2 or 3) Migrant has both US and ES bank account -0.432 (T. 2 or 3) Log (migrant's annual income) -0.007
(0.205) (0.027)
(T. 2 or 3) Migrant is female 0.029 (T. 2 or 3) Recipient's years of education -0.006
(0.128) (0.012)
(T. 2 or 3) Migrant has been in US 6-10 years -0.082 (T. 2 or 3) Recipient's raffle allocation to savings 0.003
(0.123) (0.004)
(T. 2 or 3) Migrant has been in US 11-15 years 0.188 (T. 2 or 3) Recipient plans for retirement 0.258
(0.212) (0.165)
(T. 2 or 3) Recipient is migrant's spouse -0.119 (T. 2 or 3) Log (recipient's hh income) 0.025
(0.202) (0.038)
(T. 2 or 3) Recipient is migrant's child 0.000 (T. 2 or 3) Recipient has bank account 0.053
(0.303) (0.153)
(T. 2 or 3) Recipient is migrant's other relative 0.057
(0.130)
(T. 2 or 3) Migrant's years of education 0.002
(0.012)

Observations 408
R-squared 0.128

significant at 10 significant at 5 significant at 1 significant at 10 significant at 5 significant at 1
67
Summary of findings so far

• Migrant demand for savings accounts in El
  Salvador rises when they are offered more
  possibility of control over them
• Demand is 60 higher when migrants have option of
  joint or individual (migrant) ownership over
  accounts, compared to accounts in the name of
  someone else in El Salvador
• Increase in take-up when migrants are
  additionally offered option of individual
  ownership (vs. joint ownership) is relatively
  small
• But when individual ownership is offered, 95 of
  migrants opening accounts avail of individual
  ownership
• Highly suggestive of demand for migrant control
• Additional tests of importance of migrant control
  await availability of savings data

68
To be continued

• Marketing visits to offer products are continuing
  through July 2008
• Savings and remittance data will be available in
  fall 2008
• Follow-up survey in early-mid 2009
• Examine effects on remittances, savings,
  consumption, investments (education, small
  enterprises), etc.

69
Extra slides
70
Principal-agent problem outline

• Remittances have two functions 1) funds for
  investment by migrant (principal), and 2)
  incentive payments for hh (agent)
• Migrant sends funds for investment k, and hh
  chooses fraction f that is spent in investment
  (consuming remainder)
• Household must conceal consumption of uninvested
  funds at cost c(f)
• Investment returns have random component, and
  migrant cannot observe f, only total investment
  returns x
• Problem for migrant is how to reduce diversion of
  funds by hh
• Migrant implements incentive scheme for hh based
  on returns x payments to hh are s(x)
• Prediction ability to directly control
  investments (e.g., savings, housing, education)
  raises remittances-as-investment (k) but reduces
  payments to hhs (s)
• Total resources transferred by migrant should
  rise, even as hh receives less

71
Summary statistics
Variable Mean Std. Dev. 10th pct. Median 90th pct. Num. Obs.

Treatment 0 (no savings facility offered) 0.23 0.42 0 0 1 956
Treatment 1 (remittance recipient account only) 0.25 0.43 0 0 1 956
Treatment 2 (joint account) 0.26 0.44 0 0 1 956
Treatment 3 (joint migrant account) 0.25 0.44 0 0 1 956
Migrant is female 0.29 0.45 0 0 1 956
Migrant has US bank account 0.63 0.48 0 1 1 956
Recipient is migrant's parent 0.51 0.50 0 1 1 956
Recipient is migrant's spouse 0.13 0.33 0 0 1 956
Recipient is migrant's child 0.05 0.22 0 0 0 956
Recipient is migrant's other relative 0.30 0.46 0 0 1 956
Migrant has been in US 0-5 years 0.51 0.50 0 1 1 956
Migrant has been in US 6-10 years 0.37 0.48 0 0 1 956
Migrant has been in US 11-15 years 0.12 0.32 0 0 1 956

Migrant's years in the US 5.47 3.51 1 5 10 692
Migrant has US bank account only 0.51 0.50 0 1 1 692
Migrant has El Salvador bank account only 0.09 0.28 0 0 0 692
Migrant has account in both US and El Salvador 0.09 0.29 0 0 0 692
Migrant allocation to savings in raffle 19.08 38.95 0 0 100 692
Migrant's annual income (US) 27,110 107,315 4,403 17,945 39,011 666
Migrant's household's annual income (US) 39,760 130,216 6,850 25,458 60,192 692
Migrant's years of education 8.43 4.94 1 9 12 648
Migrant's age 30.83 7.72 22 29 41 688
Migrant's annual remittances sent (US) 4,851 3,951 1,200 3,900 9,600 692
Migrant's remittances as share of annual hh inc. 0.683 3.836 0.033 0.161 0.688 690
Migrant's total hh savings balance (US) 3,531 19,391 0 700 8,000 631
Migrant's savings as share of annual hh inc. 0.207 1.050 0.000 0.028 0.299 629
Migrant is US citizen 0.01 0.08 0 0 0 689
Migrant hh size in U.S. 4.89 2.18 2 5 8 692
Migrant is married or partnered 0.60 0.49 0 1 1 687
Migrant is coresident with spouse/partner 0.75 0.44 0 1 1 396
Recipient allocation to savings in raffle 1.89 13.02 0 0 0 631
Recipient's annual hh income (US) 4,474 7,664 600 2,864 8,751 631
Recipient has savings account 0.21 0.41 0 0 1 630
Recipient total hh savings balance (US) 301 1,370 0 0 275 630
Recipient's years of education 5.05 5.62 0 3 12 621
Recipient's age 47.14 14.98 26 48 67 630
Recipient's annual remittances received (US) 2,645 2,834 250 1,800 5,900 631
Recipient hh size 4.59 2.38 2 4 8 631