David McKenzie, World Bank
(with John Gibson)




Temporary migration programs seen as offering
a “triple-win” – good for migrants, the sending
country, and the receiving country.
Seasonal workers are the largest single
temporary migration category in the OECD.
Remain controversial – with critics arguing that
workers will overstay, compete down wages of
native workers, and that workers may be
exploited or not earn enough to make short-term
migration worthwhile.
What is lacking is evidence on development
impacts.



Launched in 2007 with the dual goals of aiding
economic development in the Pacific Islands
while easing labor shortages in New Zealand’s
horticulture and viticulture industries.
Designed taking account of lessons from other
countries, and now seen as possible model
policy.
E.g. ILO good practices database states
◦ “The comprehensive approach of the RSE scheme towards filling labour
shortages in the horticulture and viticulture industries in New Zealand and
the system of checks to ensure that the migration process is orderly, fair,
and circular could service as a model for other destination countries.”


“First and foremost it will help alleviate
poverty directly by providing jobs for rural
and outer island workers who often lack
income-generating work. The earnings they
send home will support families, help pay for
education and health, and sometimes provide
capital for those wanting to start a small
business”.
Winston Peters, New Zealand Minister of Foreign
Affairs, at the approval of the RSE program,
October 2006.
The Evolution of
the RSE Program
Over 5000 workers
recruited in first 15
months (to June 2008)
•
•
4000 from Pacific,
1000 from Others (SE
Asia)
quota then increased to
8000 per year
•
•
total 2nd season almost
met the new quota
down slightly in third
year partly due to effects
of the recession on NZ
labour market
•
at the same time,
Australian pilot scheme
announced in Aug 2008
and started in Feb 2009
has recruited just over
100 workers
•
RSE
Quota





Quota of 5000 (now 8000) seasonal workers,
allowed to come for maximum of 7 months in
11 month period.
Work in horticulture and viticulture.
Workers can return again in next season
Several mechanisms to low risk of workers
overstaying – overstay rate <1%
Mostly male workers >80%



Multi-year prospective evaluation launched
alongside the roll-out of the policy.
Close collaboration between World Bank, New
Zealand Government, and Governments in Tonga
and Vanuatu.
Conducted baseline surveys of over 450
households (>2000 individuals) in Tonga and
Vanuatu before workers left
◦ Sample contains households with workers selected for
RSE, households where workers applied but not selected,
and non-applicant households.

Then follow-up surveys of these same
households at 6 months, 1 year and 2 years after
RSE program had begun

Two largest suppliers of RSE workers
Vanuatu
Tonga
Samoa
Solomons
Tuvalu
Kiribati
NonPacific

Substantial difference between the two in previous migration experience
◦

Different recruitment approaches
◦

High emigration from Tonga almost none from Vanuatu
Government-led, pro-poor in Tonga, private sector in Vanuatu
These differences should increase the external validity of extrapolating from our
results to other contexts

Tonga -- “negative selection” on income
◦ RSE workers drawn from households in the poorer
parts of the income distribution
 Any positive household-level impacts likely to be
pro-poor

Samoa – “neutral selectivity”
◦ RSE workers drawn from households whose income
appears indistinguishable from average households
 (relying on a retrospective comparison)

Vanuatu – “positive selection”
◦ RSE workers drawn from richer households
*after year 1 for Samoa

Tonga -- low in terms of lost wages at least
◦ Very few (8%) RSE workers in wage employment prior to
leaving for work in New Zealand
 Remaining family mainly need to replace a home
production contribution rather than a cash contribution

Samoa – moderate
◦ One fifth of RSE workers employed prior to going to
New Zealand
 (Based on first six months of 2007 rather than six months
prior to leaving for NZ)

Vanuatu – higher
◦ Two-fifths of sampled RSE workers had been employed
prior to leaving for New Zealand
*Unconditional average over all (both employed and unemployed)
who became RSE workers
12


Panel difference-in-differences and panel fixed
effects models after trimming sample to households
with propensity scores in the range [0.1, 0.9]
Surveys contain rich data and many of elements
which lead matching to be more successful
◦ Same questionnaire given to treatment and control at same
point in time from same labor markets
◦ Have good knowledge of the characteristics which affect
both supply and demand side of participation and can
control for them.
◦ Ask retrospective earnings history, so can match on more
than one pre-treatment period of earnings.
◦ Natural reason why some people migrated and others with
similar characteristics did not – there was excess demand
for RSE employment, so not all households who wanted to
participate could do so.
Match on:
PS-1:
- household demographic variables pre-RSE
- characteristics of the 18-50 year old males
(english literacy, education, health, alcohol use)
- household’s previous experience and network
in New Zealand
- Household baseline assets and housing
- Geography (island)
- Wage and Salary history
PS-2: all this plus whether they applied to work in
RSE.

Difference-in-differences Specification
4
𝑌𝑖,𝑡 = 𝛼 + 𝛽𝐸𝑣𝑒𝑟𝑅𝑆𝐸𝑖 +
𝛿𝑡 + 𝛾𝑅𝑆𝐸𝑖,𝑡 + 𝜀𝑖,𝑡
𝑡=2
Fixed Effects Specification:
4
𝑌𝑖,𝑡 = 𝜇𝑖 +
𝛿𝑡 + 𝛾𝑅𝑆𝐸𝑖,𝑡 + 𝜀𝑖,𝑡
𝑡=2
Two measures of exposure to policy:
- Ever work in New Zealand in RSE by time t
- Cumulative number of months in RSE at time t
The exposure to
the ‘treatment’
varied widely
0.14
Tonga
0.12
Vanuatu
Cumulative
duration of RSE
months worked at
wave 4 of our survey
•
duration ranges
from 1 RSE month to
over 20 RSE months
0.08
Density
variation due to
variable spell
lengths, only some
workers having
multiple spells and a
few households with
multiple workers
•
0.1
0.06
0.04
•
0.02
0
0
5
10
15
Months
20
25

Remarkably low attrition in Tonga
◦ Only 1% per round (sample sizes 448 in round 1, 442 in round 2,
444 in round 3, 440 in round 4).

Higher attrition in Vanuatu
◦ 10-15% per round, but not cumulative since rounds 3 and 4 also
tried to track baseline households not in round 2
◦ Sample sizes 456 in round 1, 382 in round 2, 388 in round 3, 348
in round 4.
◦ Slightly higher among non-RSE households
◦ Partly due to mobility of individuals amongst extended family and
lower cost residential mobility
 Greater reliance on informal/shanty housing in urban settlements in
Vanuatu than in Tonga
 More difficult to track households over survey rounds

Appendix shows main results are robust to bounding
approaches to deal with this attrition.
40
participating in RSE
Percentage Increase from
45
35
30
25
20
15
10
5
0
Income -DD
Income -FE
Tonga
Expenditure -DD
Vanuatu
Expenditure -FE
Table 2: Average Impact of RSE Migration on Household Income and Expenditure in Tonga
Difference-in-Differences
Fixed Effects
Baseline Mean
(1)
(2)
(3)
(5)
(6)
Outcome Variable:
for RSE households
All
PS-1
PS-2
All
PS-1
PANEL A: IMPACT OF EVER BEING IN THE RSE
Per Capita Income
979
331.0*** 278.4*** 233.1*
347.4***
271.0***
248.7**
Log per Capita Income
6.57
0.355***
Per Capita Expenditure
829
224.1**
127.1
Log per Capita Expenditure
6.58
0.124**
PANEL B: IMPACT PER MONTH'S DURATION IN RSE
Per Capita Income
979
24.56***
0.346*** 0.290***
(7)
PS-2
0.383***
0.355***
0.324***
104.6
249.8**
117.8*
133.0
0.117**
0.0834
0.128***
0.0926*
0.0900
23.49**
20.03*
39.48***
35.06***
31.19***
Log per Capita Income
6.57
Per Capita Expenditure
829
9.765
3.476
-0.126
15.78
2.632
1.170
Log per Capita Expenditure
6.58
0.00257
0.00227
-0.00178
0.00339
-0.00106
-0.00277
1,774
448
1,499
379
1,092
274
1,774
448
1,499
379
1,092
274
Number of Observations
Number of Households
0.0326*** 0.0350*** 0.0329*** 0.0459***
0.0469*** 0.0449***
Table 3: Average Impact of RSE Migration on Household Income and Expenditure in Vanuatu
Difference-in-Differences
Fixed Effects
Baseline Mean
(1)
(2)
(3)
(5)
(6)
(7)
Outcome Variable:
for RSE households
All
PS-1
PS-2
All
PS-1
PS-2
PANEL A: IMPACT OF EVER BEING IN THE RSE
Per Capita Income
85282
42,861*** 44,441*** 48,241*** 29,522* 32,760** 37,717**
Log per Capita Income
10.73
Per Capita Expenditure
65872
Log per Capita Expenditure
10.63
0.320*** 0.301*** 0.364***
0.186*
0.167
0.267**
12,353** 13,020**
1,093
2,228
2,978
0.240*** 0.261*** 0.254***
0.134*
0.137*
0.132*
3,391
5,066**
5,964**
8,495
PANEL B: IMPACT PER MONTH'S DURATION IN RSE
Per Capita Income
85282
3,382*
Log per Capita Income
10.73
Per Capita Expenditure
65872
Log per Capita Expenditure
10.63
Number of Observations
Number of Households
4,992**
5,219**
0.0366*** 0.0466*** 0.0501*** 0.0284** 0.0364*** 0.0488***
1,006
2,225**
2,023**
660.3
1,531**
1,504**
0.0233** 0.0351*** 0.0321*** 0.0162* 0.0258** 0.0236**
1,574
456
1,225
360
977
269
1,574
456
1,225
360
977
269


Results show a 31-42 percent increase in
income per capita over the 2 years for the
average household participating in the RSE
This makes it one of, if not the most,
successful development intervention for
which rigorous impact evaluation is available:
◦ Progresa/Oportunidades – only 8 percent increase
in per capita expenditure, and this involved handing
out lots of money
◦ Microfinance – recent Spandana randomized trial by
Banerjee et al. finds no impact on average income.


But migrants earned NZ$12,000 in NZ, compared
to baseline household incomes of $NZ1400 per
capita in Tonga and NZ$2500 per capita in
Vanuatu.
So if you can earn 5-7 times per capita income,
why is increase “only” 30-40%?
◦ 1) Migrants have costs in NZ and of airfare, so amount
remitted + repatriated = NZ$5,500.
◦ 2) This increase is only for one individual per household.
Since average household is 5-6 members, this makes
PER CAPITA increase around NZ$1,100.
◦ 3) Only half the households went for 2 years, so average
per capita increase over 2 years requires dividing by 1.5
◦ 4) Then households face opportunity cost in terms of
what migrant would have done at home.


Imagine a 10-step ladder, where on the bottom step were the
poorest people and the top step the richest people, state
which step of the ladder they thought their household was on
today, and on which step their household was on two years
ago.
We find 0.43 step increase in Tonga and 0.65-0.77 increase
in Vanuatu – approximately 0.5 standard deviation increase in
both countries.
Find Tongan households 10-11 percentage points more likely to
have improved dwelling over 2 years if in RSE.
Vanuatu 6-7 percentage points more likely.




Households with Bank account: 10-14 percentage
point increase in Tonga from RSE, 17-19 percentage
point increase in Vanuatu
Tongan RSE households are significantly more likely
to have acquired a cellphone, television, DVD player,
and bicycle over the two-year period than similar
non-RSE households.
RSE households also seem to have sold their
kerosene ovens and purchased gas or electric ovens
instead.
The ni-Vanuatu RSE households are significantly
more likely to have acquired a radio or stereo, a DVD
player, a computer, a gas or electric oven and a boat
over the two-year period than similar non-RSE
households.
- School fees identified as one of most important uses of money
earned
-Find a 10-14 percentage point increase in proportion of 16-18
year olds attending school in Tonga; no significant effect in
Vanuatu
-Not much non-farm self-employment


Relatively modest – villages got about $500 in
remittances to village on average – used in
Tonga mostly to improve water supply.
In Tonga 92 percent of leaders say that it has
had positive effects after two years, and in
Vanuatu, even at 6 months, 72 percent of
leaders say the overall impact is positive.




RSE designed with promoting development in the
Pacific as an explicit goal.
We find it has largely met this goal.
Results make seasonal migration one of the most
cost-effective and largest impact development
interventions for which rigorous evidence available
Caveats:
◦ Impacts may change over longer-term
◦ Gains still pale in comparison to permanent migration
◦ Open question how far one can extrapolate to other country
contexts, but scale and best-practice procedures along
with large baseline differences between Tonga and Vanuatu
suggest reasonable external validity.