Socioeconomic Effects of Remittances and Migration

1
Socioeconomic Effects of Remittances and
Migration

• David McKenzie
• World Bank

2
Growing importance of remittances worldwide

• Developing countries recorded remittances have
  surged in recent years
  
• 73 increase between 2001 and 2005
• More than five times as large as in 1990
• Reflects
• Better measurement, shift from informal to formal
  channels
  
• Less regulations and falling costs
• Some growth in migration
• This has led to a surge in interest in the impact
  of remittances on development outcomes

3
Traditional approaches to finding the impact of
remittances

• Ask migrants what remittances are for, or
  families what remittances are spent on.
  
• E.g. Mexican Migration Project

4
result

• Early studies came to largely pessimistic
  conclusions about potential of remittances to
  promote economic growth
  
• View remittances as leading to a cycle of
  dependency, money as being wasted on food,
  drinks, fiestas and conspicuous consumption.

5
Newer studies

• Realize money is fungible, so even if spend
  remittances on parties, this allows you to spend
  some of your other income on other uses,
  including productive investments.
• Use regression approach, e.g.
• Outcome a bRemittances cX e

6
How do we know what the impact of remittances
really is?

• Approach 1 Treat remittances as manna from
  heaven
  
• - assume some households just happen to receive
  remittances, see what they spend the additional
  income on
  
• - Problem suppose I only send remittances to my
  family when someone is sick then I would see
  households which receive remittances have worse
  health!

7
Are remittances any different from other income?

• We know that more income lowers poverty, improves
  HD outcomes
  
• So should we expect anything different from
  remittances?

8
Are remittances any different from other income?

• One reason remittances might be spent differently
  is that they are only sent for specific events,
  or conditional on certain actions occurring
  
• 66 of remittances received in Tonga were for a
  special purpose.
  
• Main purposes are misinale (33), payment of
  school fees (28), funeral expenses (14)
  
• Money is fungible, so earmarking only changes
  consumption if conditions are binding, or if
  families receiving remittances face different
  prices.

9
Why else might remittances be spent differently?

• Households may view remittances as being more
  temporary in nature
  
• Permanent income theory suggests households will
  save a larger fraction of temporary income.
  
• But cross-sectional surveys provide very little
  information on sustainability.
  
• PINZMS asks migrants and their families
  expectations for remittances.

10
The expected chance of remittances decays

• Expectation of receiving remittances declines
  over time, and declines for almost every single
  family
  

11
Implications

• The chance of receiving remittances is expected
  to decay over time for both migrants and their
  families
  
• Should expect to see receiving households save or
  invest more of it than they would for wage
  income.
  
• Overall amount of remittances received in Tonga
  may not decay if more established migrants start
  sending larger amounts when they do send (or if
  sufficient new migrants each year)

12
Approach Two Look at the overall effect of
migration

• Remittances arent just manna from heaven, they
  are accompanied by other events many of which
  are tied to the migration decision.
  
• Need to think of reasons why one household may
  have a migrant and another similar household
  would not (instrument).

13
Example 1 Impact on Schooling

• Remittance effect
• Alleviates credit constraints buy more
  schooling if you are constrained
  
• Income is a normal good buy more of everything
  including schooling
  
• But what else is going on?

14
Example 1 Impact on Schooling

• But what else is going on?
• Absent parents may require kids to do things
  for parents
  
• Incentives to migrate seeing parents do well
  abroad with low schooling lowers incentives to
  get education.

15
How important are these other effects?

• Survey of Zacatecas students found those with
  migrant parents less likely to want to continue
  their studies.
  
• McKenzie and Rapoport (2006)

16
Overall effect

• After controlling for other factors, find
  children in migrant households attain less
  schooling than children in non-migrant
  households
• Boys migrate instead of continuing in school
• Girls do housework
• This is the opposite from what we would predict
  just looking at remittances!

17
Our identification strategy

• Use historic state-level migration rates as
  instruments for current migration
  
• Rates are for 1924, and reflect pattern of
  arrival of the railroad into Mexico
  
• Initial networks lower cost of subsequent
  migration, resulting in self-reinforcing process
  
• A household living in a community with high
  levels of early 20th century migration therefore
  more likely to have a migrant member than an
  otherwise identical household in a community with
  low historic migration

18
Example 2 Impact on child health outcomes

• Hildebrandt and McKenzie (2006)
• Remittance effect more income means households
  spend more on health inputs, improving health
  outcomes.
  
• Additional migration effect mothers in migrant
  households have better health knowledge, allowing
  them to get better health outcomes out of the
  same inputs.
• Source of exogeneity historic networks.

19
Results

• Find 3 to 4.5 lower infant mortality rate in
  migrant households
  
• Being in a migrant household raises birthweight
  by 364 grams, and lowers probability of being
  underweight by 6.9
  
• Only part of this effect is due to greater
  wealth/income, some is due to improvements in
  health knowledge
  
• But some costs less likely to be breastfed,
  receive all vaccines on schedule in first year.

20
Example 3 Increase in income from Migration

• McKenzie, Gibson and Stillman (2006).
• Look at migration from Tonga to New Zealand
• Have a great way of identifying the effect of
  migration New Zealand has a quota to allow a
  certain number of Tongans in each year, uses a
  lottery to decide who can come in
• We surveyed winners and losers in the lottery.
• We also surveyed people who didnt apply for the
  lottery, and use this to calculate
  non-experimental measures of the income gain

21
Monthly income gain from migration

• Difference in GDP per capita would suggest a gain
  of NZ546 per week
  
• Comparing winners and losers in the migration
  lottery gives an experimental estimate of NZ274
  per week
  
• How well do different non-experimental estimators
  do at approaching this?

22
Non-experimental approaches

• If you can, find a good instrument
• We use distance in Tonga to the NZ embassy
• - this affects likelihood of applying to
  migrate
  
• - but shouldnt affect incomes in NZ
• Get an estimate of NZ 279 (only 1.8 different
  from experimental estimator)

23
Non-experimental estimators

• but a bad instrument can really hurt
• Use migrant network in New Zealand
• Seems reasonable that you would be more likely to
  migrate if you have more family there (first
  stage F-stat is 14)
  
• But exclusion restriction violated
• Estimate is 499 (82 higher than experimental!)

24
What if a good instrument is not available

• OLS regression 31-40 too high
• Single difference 25 too high
• Difference-in-differences 20 too high
• Note both single-difference and diff-in-diff
  require panel data on retrospective information
  on outcome of interest, before and after
  migration

25
Matching

• Results are about 26 too high in a simple match
• Improvement to 19 too high when using past
  income in matching, and bias-adjustment
  
• Results not that sensitive to trimming or to
  number of matches here

26
Conclusions

• It is really hard to separate the effect of
  remittances from the overall effect of migration
  
• To do so requires thinking of a exogenous reason
  why one migrant might send more remittances than
  another migrants (exchange rate shocks studied by
  Dean Yang in Philippines come close)
• In general then, want to look at overall effect
  of migration, of which remittances is an
  important part
  
• Migration has a number of socioeconomic effects
  which differ from the pure effect of remittances

27
Do the poor benefit from remittances?

• Typical cross-sectional survey has total income,
  remittances
  
• Questions of interest are remittances going
  mainly to poor households? Do remittances lift
  households out of poverty?
  
• Naïve answer 1 look at where households
  receiving remittances lie in the income
  distribution.
  
• Problem this is AFTER remittances

28
Poverty and Remittances

• Naïve answer 2 subtract remittances from total
  income, and see where those receiving remittances
  lie in the distribution
  
• Problem treats remittances as manna from
  heaven
  
• Solution try and calculate what household
  income would have been in absence of remittances

29
Considerations

• Lost income of migrant some approaches try and
  calculate this by predicting income for migrant
  if s/he had stayed
  
• Receiving remittances might change labor supply
  of other household members, might allow them to
  overcome liquidity constraints on
  entrepreneurship, etc.
• Absence of a member might change labor supply,
  income-earning opportunities in household
  
• i.e. income of other household members affected
  by receipt of remittances and by migration
  
• Household resources per person is higher due to
  absence of an eater
  
• Need to consider all these factors when
  attempting to look at poverty impact

30
So what should we do

• Try and think of possible good instruments
• E.g. exchange rate shocks (Philippines), labor
  market conditions in receiving country, historic
  networks, policy changes, cost of sending money
  changes,
• Use pre- and post-migration data for
  difference-in-differences or matching
  
• Think about possible directions of biases, and
  whether non-experimental methods are likely to
  give upper or lower estimate.