You are on page 1of 61

Selection in

immigration:
Who moves?
Introduction

“How do changes in
Utility- or income- (economic) conditions in
Last lecture: maximization model and origin or destination
gravity model locations affect the number
of people who migrate?”

“How do changes in
(economic) conditions in
This lecture: Roy model origin or destination
locations affect who
migrates?”
Selection in migration

Immigrants are unlikely to be perfectly representative of the population in the origin


They may be predominately high skilled or low skilled; workers; students; family
members; retirees; ...

Why do not all migrants from an origin country go to the same destination country?
People in the same country typically share a culture and language, and they are about
the same distance from other countries. So why do some migrants go to one place
while others go elsewhere?
What are ”desirable” characteristics?

Origin and destination countries can have different (conflicting) perspectives on what
characteristics are “desirable” among migrants:
Origin countries may be concerned that emigration of skilled workers will slow
economic growth, while destination countries may want to attract skilled workers in
order to boost economic growth
Origin countries may want students to go abroad to receive an education but then
return home to contribute to the economy. Destination countries may not want to
incur the cost of educating foreign students unless they will stay and join the
workforce
Selection models

The economics of immigration uses selection models to look at immigrants’


characteristics
Main focus on the relationship between immigrants’ skill levels and returns to
skill in the origin and destination, though the model can be applied to other
characteristics as well

Key prediction: people sort across potential destinations or decide to not migrate
at all based on their characteristics and the relative returns to those characteristics
I. The Roy model
The Roy model
Borjas (1987) developed an influential model of skill selection in
immigration, based on a canonical model by Andrew Roy (1951)
that examined how self-selection into occupations affects the
distribution of income
Borjas, G. J. (1989).
Economic theory and
international migration. 
International Migration
Review, 23(3), 457-485.
Lifetime earnings

The (log) earnings for residents of the source (0) and host (1) country are characterized by the earnings functions:
Log w0 = X δ0 + ε0 (1)
Log w1 = X δ1 + ε1 (2)
with:
• w: individual lifetime earnings
• X: a vector of observable socioeconomic characteristics (such as education and age)
• δ: rate of return to socioeconomic characteristics X
• ε: random variable ∼ N(0, σ2)
• Xδ: mean earnings for person with characteristics X
• ρ = corr(ε0,ε1): if ρ > 0 and close to 1 the host and source country value unobserved ability in the same way
εO
• εO is a measure of how much wages vary across individuals relative to the
average
 εO >0: earn more than the average
 εO <0: earn less than the average

• Interpretation: measure of the return to unobserved characteristics (like


ambition or willingness to work hard)
• Everyone knows X δ0 and X δ1  and the cost of migrating, and people know
their own εO and ε1
ρ
• This correlation coefficient can range from −1 to +1. How countries value
unobserverd skills might thus differ
• If ρ > 0: above-average earnings in the origin  above-average earnings in
the destination, below-average earnings in the origin  below-average
earnings in the destination
unobserved skills that are valuable in the origin also are valuable in the destination
most pairs of countries ρ  is quite large and positive. But there are also other cases
• If ρ small or <0: people who have higher-than-average earnings in the
origin have lower-than-average earnings in the destination
The migration decision

We can define the index function:

with:       
• C: mobility costs to migrate from country 0 to 1
• π = C/w0: “time-equivalent” measure of the costs of migration ,
assumed to be the same for all individuals in the source country
Should I stay or Should I go?

An individual will decide to migrate to the host country when I>0,


and to remain in the source country when I≤ 0

Probability that an individual with characteristics X born in country


0 migrates to country 1:
George Borjas: Captain Obvious?

The emigration
rate of a country • the higher the mean income in
populated by the host country
income- • the lower the mean income in the
maximizing source country
individuals is • the lower the migration costs
higher...
Composition of the migration flow

The framework allows also to identify the types of sorting generated


by the immigration market:
A. Selection in observed characteristics (focus on skills)
B. Selection in unobserved characteristics
The direction of selection

Borjas developed the selection model to examine where immigrants are likely to be
in the distributions of wages in the origin and the destination, or the “direction of
selection”

The model assumes that


• the destination country has higher wages than the origin country
• but the destination country may have a higher or lower return to observed and
unobserved characteristics than the origin country
A. Selection in observed characteristics

For simplicity, assume that X consists of only one factor: education, which is normally
distributed
Then the source country’s educational attainment can be described by:
X = μX + εX (5)
with
• μX: mean education in the source country
• εX: random variable describing the heterogeneity of educational attainment in source
country ∼ N(0, σ2X)
A. Selection in observed characteristics

The conditional mean of education among migrants:


E(X|I>0)=μX +k(δ1−δ0) (6)
where k is a positive number
The direction of selection in a normal distribution of
skill
B. Selection in unobserved characteristics

The variance in earnings among individuals is explained predominantly by


differences in unobservables like skills, ability or luck, and only to a limited extent
by (an exhaustive set of observable variables) X
⇒ Given X, what kind of persons migrate? Consider the conditional expectations:

where σv : standard deviation of v = ε1 − ε0


B. Selection in unobserved characteristics

The values of these conditional expectations (Q0&Q1) determine the extent to which the self-
selection of immigrants results in the unmeasured skills of the foreign born population to deviate
from the average:

1. Q0 = 0: migrants have average unmeasured skills or ability of the source country


2. Q1 = 0: migrants have the same unmeasured skills or ability as natives in the host country
3. Q0 > 0 and Q1 > 0: positive selection
4. Q0 < 0 and Q1 < 0: negative selection
5. Q0 < 0 and Q1 > 0: refugee sorting
B.3. Positive selection in unobserved
characteristics

Necessary and sufficient conditions:


• Correlation coefficient in the earnings across the two countries is sufficiently high:
a positive constant
• Income in the host country is more dispersed than in the source country (higher inequality): σ1 >
σ0

Immigrants have above average abilities in both the source and host country
High income workers benefit relatively more from migrating than low income workers →
brain drain
B.4. Negative selection in unobserved
characteristics

Necessary and sufficient conditions:


• Correlation coefficient in the earnings across the two countries is sufficiently high:
ρ>ρ̄
• Income in the host country is less dispersed than in the source country (lower
inequality): σ1 < σ0

Low income workers can improve their situation in the host country
Low income workers benefit relatively more from migrating than high income
workers
B.5. Refugee sorting in unobserved
characteristics

Necessary and sufficient condition:


• Correlation coefficient in the earnings across the two countries is small or negative:
ρ<ρ̄

Special case: country shifts from market economy to communist regime (confiscation
and redistribution of income → original distribution of income is reversed)
Migrants from such countries have below average income (they saw their assets vanish)
but perform relatively well in the host country’s market economy labor market
Holds also for refugees: changes in political regimes tend to devalue skills, but ability is
once again valuable when they migrate to a market economy
Refugees and selection: nuance

But not all refugees are successful: many have low skill levels and do poorly in both the origin and the
destination
• Refugees may be negatively selected relative to the destination if they would not migrate absent some
adverse event, such as a natural disaster or a civil war
• Because their migration is not motivated by potential economic gains, they are unlikely to be selected on
characteristics that are valued in the destination
• Refugees also may have little choice of destination country but simply flee to the nearest safe country or to
a country that will take them

“Ultimately, the direction of selection among refugees depends on the nature of the refugee-producing event
and other idiosyncratic factors” (Chin and Cortes, 2014)
Implications: solving empirical mysteries

The theory helps to explain some empirical mysteries such as the drop in the
standardized earnings of migrants after the 1965 Amendments to the US
Immigration and Nationality Act
• The Amendment abolished restrictions on migration from non-European countries
to the US, which are more likely to have greater income inequality than the US
• The Amendment led to a shift from visa allocation based on observable skills and
occupational characteristics to kinship relationships between potential migrants
and persons currently residing in the US (i.e. social networks as “safety net”)
What if migration costs vary with skills?

If migration costs depend on skill, immigrants are from the middle of the skill (and wage) distribution 
“Intermediate selection” rather than positive or negative selection

Suppose the return to skill is higher in the origin than in the destination and skill is still perfectly
transferable across countries
Low-skilled: bigger gains from migrating but also higher migration costs  no benefit from migrating
Intermediate skilled: still gains from migrating and lower migration costs than low-skilled  benefit
from migrating (positive net gains)
Highest-skilled: low migration costs but returns to skill higher at home  do not migrate
Intermediate selection

The origin has a higher return to skill


than the destination, but migration
costs decrease with skill. People with
skill levels between s and s’ migrate
Why would migration costs decrease with skills?

Several reasons:
• People with low skill levels (and hence low incomes) may face liquidity
constraints, situations where they cannot save or borrow enough to pay the
costs of migrating
• Immigration policies that favor skilled migrants may make migration costs
lower for skilled migrants than for unskilled migrants
• Migrant networks may increase with skill and having a bigger migrant
network reduces migration costs
Summing up the model

Conditions determining Selection on the basis of unobserved ability depends on


selection in terms of • income inequality in the host and source countries
unobserved ability have • the correlation between earnings in the two countries
nothing to do with the
conditions determining Selection on the basis of observed demographic
selection in terms of characteristics depends on
observed demographic • parameters measuring the price attached by each country's
characteristics labor market to that particular characteristic
Summing up the model

The immigrant flow may be Important insight:


Any combination of selection composed of relatively highly If the migrant flow to any given
mechanisms in unobserved and educated persons, but these host country is composed of
observed characteristics is possible: highly educated persons do not highly educated individual, this
negative selection in unmeasured skill do well in the host country and does not imply that these
(or ability) may be occurring did not do well in the source highly educated persons are the
simultaneously with positive selection country (compared to other most productive highly
highly educated persons) prior educated persons in the country
in education, or vice versa to their migration of origin
Why are not all poor people moving then?

In both the selection model and the utility- or income-maximization model, people decide whether to move
based on their income in the origin and in the destination
They decide to migrate if their income will be higher in the destination, net of migration costs and adjusted
for the cost of living
We should observe almost no one living in poor countries

Why then do we observe so few people migrating from poor countries?


• Barriers to immigration in rich countries, such as quotas
• Borrowing constraints
• Limited migrant networks
• Lack of information about the gains to migrating
• But also the indirect, or psychic, costs of migration may be very high!
Why are not all poor people moving then?

In both the selection model and the utility- or income-maximization model, people decide whether to move based on
their income in the origin and in the destination
They decide to migrate if their income will be higher in the destination, net of migration costs and adjusted for the cost
of living
We should observe almost no one living in poor countries

Why then do we observe so few people migrating from poor countries?


• Barriers to immigration in rich countries, such as quotas
• Borrowing constraints
• Limited migrant networks
• Lack of information about the gains to migrating
• But also the indirect, or psychic, costs of migration may be very high!
II. Empirical evidence on selection
Empirically testing the Roy model

The Roy model gives a clear prediction:


“Immigrants will be positively selected on skill if the return to skill is higher in the
destination than in the origin, and negatively selected if the return to skill is lower in
the destination than in the origin”

Empirically testing this model requires having data on


• the returns to skill in destination and origin countries
• data on migrants’ skill levels relative to the population in the destination or the origin
Measuring the return to skill I

Potential measures:
1. Gini index:
• measures how much the distribution of income deviates from a perfectly
equal distribution
• ranges from zero, which indicates perfect equality of incomes, to 100,
which indicates that one person or household receives all of the income in
that country
Gini index
Mexico 48.2
Gini index for 20 major China 46.5
immigrant origin or Saudi Arabia 45.9
United States 45.0
destination countries Philippines 44.4
Russia 41.2
Turkey 40.2
Source: U.S. Central Intelligence New Zealand 36.2
Agency World Factbook. India 35.2
United Kingdom 32.4
Bangladesh 32.1
Canada 32.1
Poland 30.8
Pakistan 30.7
Australia 30.3
Afghanistan 29.4
Germany 27.0
Kazakhstan 26.3
Ukraine 25.5
Sweden 24.9
Measuring the return to skill II

2. Income shares: measure the fraction of total income that accrues to a


specific portion of the income distribution
E.g. the 75/25 income share: the ratio of the % of total income that accrues to the top
25% of households to the % of total income that accrues to the bottom 25% of
households
3. Income ratio:
E.g. the 75/25 income ratio: the ratio of income earned by a household at the 75th
percentile of the income distribution to income earned by a household at the 25th
percentile of the income distribution
Measuring the return to skill III

4. Compare the earnings of workers in different skill groups:


E.g. compare the earnings of workers who have a bachelor’s degree and
workers who have a master diploma  the return to a bachelor’s degree
relative to a master diploma
5. Compare the earnings of workers by type of job
E.g. compare the earnings of workers in white-collar (office) jobs with the
earnings of workers in blue-collar (manual labor) jobs
Global patterns of selection

Globally, immigrants tend to be positively selected from source countries

Figure: Out of the 195 countries displayed, the emigration rate is higher among
tertiary-educated adults than among primary-educated adults for 177 countries
At odds with the selection model (we should observe a mix)

Studies that control for origin- and destination-country characteristics likewise find
that immigrants tend to be positively selected (e.g., Grogger and Hanson, 2011)
Emigration rates for adults
by source country and
education, 2010

Source: Emigration rate from


Brücker, H., Capuano, S. and
Marfouk, A. (2013) “Education,
gender and international migration:
Insights from a panel-dataset 1980–
2010.”
Migration to one single destination

But the data combine 20 destination countries, some of which have higher returns to skill than
many origin countries and some of which have lower returns to skill
 Looking at a single destination country may be more useful (see next figure for US)

The selection model predicts that the next figure should show a negative relationship: the
more high-skilled workers earn relative to low-skilled workers in the origin country, the lower
the emigration rate should be among high-skilled workers relative to low-skilled workers
Yet, the figure shows no relationship between the return to skill and the emigration rate
among these origin countries
Relative earnings and
emigration rates to the US for
tertiary- and primary-educated
adults, by origin country

Source: Emigration rate from


Brücker, H., Capuano, S. and
Marfouk, A. (2013) “Education,
gender and international migration:
Insights from a panel-dataset 1980–
2010.”
Other early research findings

Some other early research finds results that appear consistent with the selection
model, though
For example, Borjas (1987) analyzes data on male immigrants from 41 countries to
the United States. He concludes that male immigrants’ wages are negatively related
to income inequality in their origin country, as measured by the 90/20 income share
This finding is consistent with the selection model
However, the finding is sensitive to what other variables are included in the
model (also case in other studies e.g. by Cobb-Clark, 1993)
Limitations of earlier studies

Early studies of the selection model were hampered by data limitations:


• Lacked data on people who did not migrate  their analysis is not a true
test of the selection model because it says nothing about the skills of
immigrants relative to the population of the origin country
E.g. an immigrant from a developing country might look low skilled compared
with a U.S. native or an immigrant from an industrialized country but might be
from the top portion of the developing country’s skill distribution
• Data available for only a few destination countries
More recent findings

More recently, better evalutation possible due to availability of estimates on emigration rates by education
from a large number of origin countries to OECD destination countries

The results provide support for the model, but only under certain conditions
Belot and Hatton (2012), e.g. indeed find that as the return to skill increases in the destination relative to
the origin, selection becomes more positive. However, they only find this result when they control for the
poverty rate in the source country hypothesize that poverty prevents low-skilled people from migrating
from countries with high returns to skill
Other research finds evidence at odds with the Roy model and attributes it to migration costs or barriers to
immigration for the low skilled (e.g. ; Brücker and Defoort, 2009; Grogger and Hanson, 2011)
Limitations facing both older and newer studies

• Most available data combine all types of immigrants


Ideally, researchers would have separate data depending on the channel of migration (work,
study, family reunion, asylum)
The model has a different prediction for refugees than for immigrant workers, and it does not
apply well to family-based immigrants who do not plan to work
• Researchers typically do not know immigrants’ characteristics prior to migration
Immigrants’ earnings in the origin before migrating are usually not known, nor are their
education levels prior to migrating
If immigrants acquire education in the destination, it may not be a surprise that immigrants have
more education than people who do not migrate
Effects of migration costs and other factors

Migration costs and cultural factors shape immigrant selectivity as well:

POSITIVE SELECTION is higher NEGATIVE SELECTION is higher


• the bigger the distance between the origin and the • when countries of origin and destination share
destination colonial past
• the more similar are official languages in origin • when migrant networks are larger
and destination (transferability) • when admission policies that emphasize family ties
• the stricter is immigration policy (particularly to residents
admission policies that emphasize skills, such as
education and ability to speak the destination
country’s language)
Example: Turkey
• is roughly equidistant from Germany, Luxembourg and Spain
• has no colonial ties to any of those Western European countries
• nor is its language similar to the languages spoken in those countries
The role of  Germany recruited workers from Turkey during the 1960s and early 1970s. As a

networks in result, some 1.2 million Turks lived in Germany in 2000. About 6% of them had a
tertiary education, and 86% had a primary education
 About 1,040 Turks lived in Spain; 33% of them had a tertiary education, and 29%
Turkish had a primary education
 About 194 Turks lived in Luxembourg in 2000; 44% of them had a tertiary

migration
education, and 26% had a primary education.

 The larger the network, the more immigrants and the lower their average education
level. Beine, Docquier and Özden (2011) note, “This simple example highlights the
striking relationship between migrants’ networks and both the size and the skill
composition of migration flows” (p. 32).
III. Selection on health
The “healthy immigrant effect”

Most economic research on immigrant selection focuses on


immigrants’ education or earnings, but a few studies examine
selection on health

Studies typically find that immigrants are positively selected on


health relative to both non-migrants in the origin and natives in the
destination (e.g. Kennedy, McDonald and Biddle, 2015)
Explanations for the effect

Differences in education or income: Immigrants tend to be positively selected on


education relative to non-migrants and must have enough income to bear migration
costs  immigrants are healthier than non-migrants in the origin as health tends to be
positively related to education and income
BUT immigrants also healthier than natives in the destination
Could be due to immigration policies that screen immigrants on health, income or skill
BUT immigrants tend to be healthier than natives even in destination countries where
immigrants have less education and lower incomes, on average, than natives
 Points to a more nuanced explanation than differences in education or income
Self-selection on health

Self-selection may help explain the healthy immigrant effect!


Characteristics that increase the likelihood a person becomes an immigrant, such as
having a low discount rate, may be associated with healthy behaviors like exercising,
eating a healthy diet and not smoking
Immigrants also tend to be relatively young, so adverse health conditions may have not
yet manifested for many immigrants

The healthy immigrant effect tends to decrease as duration of residence in the destination
increases (see next lecture)
Positive selection on health: Irish immigrants to England who
were born before 1920 or after 1960 tend to be healthier than
their counterparts who remained in Ireland and than the
English

Irish immigrants Negative selection on health: The opposite is true for those
who were born between 1920 and 1960. That group of
to England migrants tended to be negatively selected on education as well,
and they had experienced relatively high rates of child abuse.
(Delaney et al., 2013) This created high psychic costs of staying in Ireland for many
of them. Although these migrants have relatively poor physical
and mental health, they probably benefited from the fact that
health care quality tended to be higher in England than in
Ireland when they migrated
IV. Selection in return migration
Selection in return or onwards migration

Immigrants who choose to emigrate from the destination country,


either to return to the origin country or to migrate to yet another
country, may not be randomly selected from the population of
immigrants in the destination

Understanding whether out-migrants are positively or negatively


selected is important for several reasons
Selective emigration affects the wage profile of an
arrival cohort (i.e. a group of immigrants who arrive
A. Impact on during a certain period, such as 2010 to 2020) of

wage profile an immigrants over time:


 If emigration is positively selected, the average wage
immigrant among a cohort of immigrants will increase less over

cohort time than it would if emigration was random or


negatively selected
Selective emigration also affects immigration’s labor
market and fiscal balance:
 Positively selected emigration means that remaining
immigrants will compete more with low-skilled
B. Impact on natives, and they will pay less in taxes while
receiving more in government transfers
labor market
and fiscal The direction of selection in return migration also
balance affects labor market and fiscal impacts in the origin
country:
 Likely to be the opposite of the effects in the
destination country
Finally, selection in immigration and in return
migration can have long-run effects:
 How well immigrants’ children and later descendants
do is likely to depend on immigrants’ characteristics
and success in the destination
C. Long-run  Political power and fiscal policies: Over time,
immigrants may become eligible to vote in the
effects destination. Their preferences regarding taxes,
government spending and other government
functions and institutions can affect political
decisions that in turn affect long-run economic
growth
V. Final thoughts on selection
Who cares about what?
Selection can be measured relative to either the origin or the destination:
Origin countries are mainly interested in the characteristics of the people who leave relative to those who stay
Destination countries are mainly interested in the characteristics of immigrants relative to natives and earlier
immigrants, not relative to the origin population
Both origin and destination countries care about whether return migrants are positively or negatively selected
among immigrants in the destination

The next lecture discusses another way of thinking about immigrant selection: assimilation, or how well immigrants
do in the destination, both initially and over time.

You might also like