The Economic Journal, 118 (April), 631–652. Ó The Author(s). Journal compilation Ó Royal Economic Society 2008.

Published by
Blackwell Publishing, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA.

BRAIN DRAIN AND HUMAN CAPITAL FORMATION IN

DEVELOPING COUNTRIES: WINNERS AND LOSERS*

Michel Beine, Fre´deric Docquier and Hillel Rapoport

Using new data on emigration rates by education level, we examine the impact of brain drain
migration on human capital formation in developing countries. We find evidence of a positive effect
of skilled migration prospects on gross human capital formation in a cross-section of 127 countries.
For each country of the sample we then estimate the net effect of the brain drain using counter-
factual simulations. Countries combining relatively low levels of human capital and low emigration
rates are shown to experience a Ôbeneficial brain drainÕ, and conversely, there are more losers than
winners, and the former tend to lose relatively more than what the latter gain.

The term Ôbrain drainÕ designates the international transfer of resources in the form of
human capital and mainly applies to the migration of relatively highly educated indi-
viduals from developing to developed countries. Recent comparative data reveal that by
2000 there were 20 million highly skilled immigrants (i.e., foreign-born workers with
tertiary schooling) living in the OECD member countries, a 63.7% increase in ten years
against only a 14.4% increase for unskilled immigrants (Docquier and Marfouk, 2006).
The vast majority of these highly skilled immigrants come from developing countries
and now represent more than a third of total immigration to the OECD. The causes of
this growing brain drain are well known. On the supply-side, the globalisation of the
world economy has strengthened the tendency for human capital to agglomerate
where it is already abundant and contributed to increase positive self-selection among
international migrants. And on the demand side, starting with Australia and Canada in
the 1980s, host countries have gradually introduced quality-selective immigration
policies and are now engaged in what appears as an international competition to attract
global talent (ILO, 2006).
The consequences for source countries, on the other hand, are less obvious. Early
contributions (Grubel and Scott, 1966; Bhagwati and Hamada, 1974; McCulloch and
Yellen, 1977) identified a range of positive feedback effects (e.g., remittances, return
migration with additional skills acquired abroad, creation of scientific and business
networks) but concluded that the welfare of those left behind would still fall given that
the social return to education exceeds its private return.1
In contrast, a series of recent papers (Mountford, 1997; Vidal, 1998; Beine et al.,
2001) suggested instead that in a context of probabilistic migration, the brain drain

* A previous version of this article received the 2003 Milken Institute Award for Distinguished Economic
Research (Beine et al., 2003). Remarks and suggestions from two referees were very helpful. We thank for
comments Andrea Bassanini, John Baude, François Bourguignon, Serge Coulombe, José-Antonio Gonzales,
Hubert Jayet, David McKenzie, Abdul Noury, Sergio Perelman, Pierre Pestieau, Maurice Schiff and Thomas
Piketti. The second author is grateful for the financial support from the Belgian French-speaking Commu-
nity’s programme ÔAction de recherches concertéesÕ (ARC 03/08 -302) and from the Belgian Federal Gov-
ernment (PAI grant P6/07 Economic Policy and Finance in the Global Equilibrium Analysis and Social
Evaluation). The usual disclaimer applies.
1
The first papers investigating the effects of the brain drain in an endogenous growth framework also
emphasised its negative impact. See Miyagiwa (1991), Haque and Kim (1995) or Wong and Yip (1999).

[ 631 ]
632 THE ECONOMIC JOURNAL [APRIL
may ultimately contribute to human capital formation in the sending countries. The
essence of the argument is that since the return to education is higher abroad,
migration prospects can raise the expected return to human capital and induce more
people to invest in education at home.2 Under certain theoretical conditions explored
in these models, this incentive effect (or brain gain) can dominate that of actual
emigration, in which case there is a net gain for the source country (i.e., a beneficial
brain drain).3
In the absence of reliable comparative data on international migration by skill level,
the debate on the consequences of the brain drain for developing countries has long
remained purely theoretical.4 This article takes advantage of a recent data set on
emigration rates by education levels (Docquier and Marfouk, 2006) to investigate
empirically how the positive and negative effects of the brain drain just described
balance out. We first estimate the effect of skilled migration prospects on gross (or pre-
migration or ex ante) human capital levels. We find that doubling the emigration rate of
the highly skilled induces a 5% increase in human capital formation among the native
population (residents and emigrants together). The coefficient is very stable across
specifications and estimation methods. For each country of the sample we then use
counterfactual simulations to estimate the net effect of the brain drain (i.e., once
skilled emigration is netted out). We find that most countries combining low levels of
human capital and low migration rates of skilled workers end up with a positive net
effect. In contrast, the brain drain appears to have negative effects in countries where
the migration rate of the highly educated is above 20% and/or the proportion of
people with higher education is above 5%. There appears to be more losers than
winners and, in addition, the former incur relatively high losses. However, the gains of
the latter dominate in absolute terms, resulting in an overall gain for developing
countries.
The remainder of this article is organised as follows. Section 1 presents the the-
oretical framework and derives the main testable implications of the analysis. Section
2 summarises the migration data. The empirical analysis is divided between Section
3, which discusses a number of econometric issues and then presents the cross-
sectional results, and Section 4, dedicated to country-specific calculations. Section 5
concludes.

2
For this incentive effect to operate, education must not only increase one’s chances of migration but
also allow for accessing to legal, high-skill jobs. In a context where immigration is illegal and migrants can
only access unskilled jobs, the prospect of migration can instead reduce education investment. See
McKenzie and Rapoport (2006) for Mexico and De Brauw and Giles (2006) for rural–urban migration in
China.
3
Using a slightly different perspective, Stark et al. (1997) also elaborated on the possibility of a brain
gain associated with a brain drain in a context of imperfect information with return migration. See also
Katz and Rapoport (2005) on migration imparting education with an option value, and McCormick and
Wahba (2000), who obtain the result that more highly skilled migration may benefit those left behind in a
model where migration, remittances and domestic labour-market outcomes are jointly determined and
multiple equilibria arise, with the high-migration equilibrium Pareto-dominating the low-migration equi-
librium. Commander et al. (2004) and Docquier and Rapoport (2008) survey the recent brain drain
literature.
4
An exception is Beine et al. (2001), who found a positive and significant effect of migration prospects
on human capital formation in a cross-section of 37 developing countries. However, their study suffers
from the fact that due to data constraints, they used gross migration rates as a proxy measure for the brain
drain.

Ó The Author(s). Journal compilation Ó Royal Economic Society 2008

2008 ] BRAIN DRAIN AND HUMAN CAPITAL FORMATION 633
1. Theoretical and Empirical Framework
1.1. Theoretical Background
Consider a stylised small open developing economy producing goods and human
capital. The amount of goods produced is proportional to labour measured in effi-
ciency units: Yt ¼ wtLt, where wt is the equilibrium wage rate in this economy. At birth,
individuals are endowed with a given level of human capital normalised to one. Indi-
viduals live for two periods and make two decisions: whether to invest in education
during their youth, and whether to migrate in adulthood. In particular, increasing
human capital requires private spending in education. There is a unique education
programme e. For an individual opting for education, the number of efficiency units
once adult is given by h > 1, while the cost of education, which is decreasing in per-
sonal ability, is denoted by c, a variable with cumulative distribution F(c) and density
function f(c) defined on Rþ.
Once adult, people can emigrate to a high-wage destination with probability p for
skilled workers and p for unskilled workers. As explained in our introduction, selective
immigration policies, together with the tendency for migrants to positively self-select
out of the general population, explain why emigration rates are much higher among
the highly educated and skilled.5 We will therefore assume that p > p. For analytical
simplicity, we normalise p to zero. Also, in what follows, we treat p as exogenous, as if it
were the result of a relative quota set by immigration authorities independently of the
number of applicants. However, we could equally assume that a given number of visas is
attributed, which can be translated into a probability of receiving an entry visa by agents
with perfect (in which case the adjustment is immediate)6 or adaptative (in which case
the subjective and objective probabilities only coincide at the steady state) expectations
with respect to othersÕ education decisions.7
Individuals are assumed to be risk-neutral and maximise lifetime income. There is no
intertemporal discounting of income. As explained, unskilled workers are assumed to
remain in the home country and therefore earn the domestic wage w in both periods. In
contrast, skilled workers have the possibility of migrating to a technologically more
advanced country where the wage rate per efficiency unit of human capital is w > w.
They earn w  c in the first period and then either w h if they migrate or wh if they do not.
For a given migration probability p, the condition for investing in education is therefore:

wt  c þ ð1  pÞwtþ1 h þ pwtþ1 h > wt þ wtþ1

and individuals will opt for education if


c < cp;t  wtþ1 ðh  1Þ þ phðwtþ1  wtþ1 Þ: ð1Þ

Clearly, migration prospects raise the expected return to human capital in the
developing country, thus inducing more people to invest in education. The critical
threshold cp,t is increasing in the probability of migration and in the wage differential.

5
For example, Docquier and Marfouk (2006) find that emigration propensities are five to ten times higher
for workers with more than twelve years of education than for workers with less than twelve years of education.
6
Formally, p can be a decreasing function of cp(p) in (1), defining an implicit solution for p.
7
In the empirical analysis, however, it will be important to assess the exogeneity of the migration prob-
ability.

Ó The Author(s). Journal compilation Ó Royal Economic Society 2008

634 THE ECONOMIC JOURNAL [APRIL
This suggests that the incentive effect of migration will be stronger in poor countries.
However, credit constraints on education investment are likely to be more binding in
poor countries. To take this into account, we introduce a minimum threshold of first-
period consumption, lt, which must be financed out of first-period earnings. Hence,
for an educated individual, it must be the case that wt  c > lt or, equivalently, that:
c < cl;t  wt  lt : ð2Þ
Liquidity constraints are binding if cl,t <c p,t, that is, if wt  lt < wtþ1 ðh  1Þ þ
phðwt  wt Þ. At the steady state (i.e., for wt ¼ wtþ1), the binding liquidity constraints
condition may be written as:
wð2  hÞ  phðw   wÞ < l:
We therefore impose the restriction that h 2 [1,2] to allow for the possibility of
either binding or non-binding constraints,8 depending on the value of w. It is clear
from the last expression that liquidity constraints are more likely to be binding in poor
countries (low w) facing high emigration rates (high p).
We denote by Ha,t and Hp,t respectively the gross or ex ante (i.e., before migration
occurs) and the net or ex post (i.e., once emigration is netted out) proportions of
educated in the population, which we take as a measure of the country’s human capital
level. The proportion of young agents opting for education is given by Ha;t ¼ F ðct Þ
where ct ¼ Minðcp;t ; cl;t Þ while the proportion of skilled adults remaining in the
country is given by:
ð1  pÞHa;t1
Hp;t ¼ : ð3Þ
1  pHa;t1
At the steady state, we have
@Hp ð1  pÞ@Ha =@p  Ha ð1  Ha Þ
¼ :
@p ð1  pHa Þ2
Using the above expression, it appears that:
 There is a possibility of beneficial brain drain over some ranges of p providing
that oHp/op is positive at p ¼ 0. This first requires that oHa/op is positive (i.e.,
there is an incentive effect), which implies that liquidity constraints are not
binding in the closed economy;
 At the margin, an increase in the rate of skilled emigration is good for human
capital formation if oHp/op is positive at the current emigration rate. Again, this
first requires that liquidity constraint are not binding, but this time at the cur-
rent level of p;
 Finally, the total or net effect of migration on human capital formation can be
obtained by comparing the ex post (or net) level of human capital with its
counterfactual level in the closed economy solution, Hpjp¼0 ¼ Hajp¼0  H~ .
There is a beneficial brain drain if the net effect is positive, that is, if Hp > H~ .

8
The hypothesis h < 2 is required to obtain internal solutions with non-binding constraints. Using a non-
linear utility function with risk aversion would enable us to consider higher values for h.

Ó The Author(s). Journal compilation Ó Royal Economic Society 2008

2008 ] BRAIN DRAIN AND HUMAN CAPITAL FORMATION 635
As explained, the realisation of these conditions depends on whether liquidity
constraints are binding as well as on the ability distribution. For illustrative purposes,
let us consider the case of a uniform distribution: c ! U[0,1] and assume l < w  1
to avoid corner solutions. With a uniform distribution, Ha ¼ c ¼ Min(cp,cl). Starting
from a closed economy equilibrium, three configurations arise.
The most pessimistic one occurs when liquidity constraints are binding in the closed
economy. In this case, when w(2  h) < l (i.e., when the domestic wage rate is low),
there can be no incentive effect: oHa/op ¼ 0. Hence, any marginal increase in the
skilled migration probability would generate a net loss:
@Hp ðw  lÞð1  w þ lÞ
¼ < 0:
@p ½1  p ðw  lÞ2
Obviously, in this case the brain drain can only be detrimental (Hp < H~ Þ.
An intermediate configuration arises when liquidity constraints are not binding in
the closed economy but become binding once migration prospects are introduced. In
this case, when w(2  h) > l > w(2  h)  ph(w  w) (i.e., when the domestic wage
rate is not too low and the migration rate is relatively high), a sufficiently small degree
of openness can foster ex post (or net) human capital if oHp/op is positive at p ¼ 0, that
is if
hðw   wÞ > wðh  1Þ½1  wðh  1Þ: ð4Þ
However, at the current migration rate, a marginal increase in p reduces the proportion
of educated remaining in the economy as binding credit constraints do not allow for
the incentive effect to operate further (oHa/op ¼ 0). The net effect is positive
(Hp > H~ ) if the skilled emigration probability does not exceed the following critical
value:
wð2  hÞ  l
p< :
ðw  lÞð2  hÞ
The most optimistic case arises when liquidity constraints are never binding, thus
allowing for the incentive effect to fully operate. In this case, obtained when
w(2  h)  ph(w  w) > l (i.e., when the domestic wage rate is high enough and the
skilled emigration rate is sufficiently low), the condition for a sufficiently small degree
of openness to foster net human capital formation is the same as in (4) and the net
effect is positive (Hp > H~ ) when the skilled emigration rate does not exceed the
following critical value:
hðw   wÞ  wðh  1Þ½1  wðh  1Þ
p< :
hðw   wÞ½1  wðh  1Þ
Finally, the sign of oHp/op evaluated at the current migration rate can be positive or
negative depending on the wage differential and on the magnitude of emigration.
When p tends to one, clearly, oHp/op is more likely to become negative.
On the whole, our simple theoretical model predicts that migration prospects can
stimulate the accumulation of human capital in developing countries under certain
conditions: first, there must be an incentive effect (or brain gain) and, second, the latter

Ó The Author(s). Journal compilation Ó Royal Economic Society 2008

636 THE ECONOMIC JOURNAL [APRIL
must be greater than actual skilled emigration (or brain drain). The incentive effect
would seem to be potentially stronger in poor countries but may be limited there if
liquidity constraints are binding. It is therefore unclear a priori whether poor or inter-
mediate income countries experience the strongest incentive effects and, consequently,
it is also unclear which type of countries gain or lose more from the brain drain.

1.2. Related Empirical Model

To evaluate the incentive hypothesis described theoretically in (1), we use a b-conver-
gence empirical model and regress the growth rate of the ex ante stock of human capital
(i.e., including emigrants) between 1990 and 2000, D ln(Ha)  ln (Ha,00)  ln(Ha,90),
on a set of explanatory variables. It is this human capital formation equation, (5), that
we estimate econometrically in Section 3:
D lnðHa;9000 Þ ¼ a0 þ a1 lnðHa;90 Þ þ a2 lnðp90 Þ þ a3 lnðp90 ÞGNID90
ð5Þ
þ a4 DENS90 þ a5 SSAD þ a6 LATD þ a7 REM90 þ 

The following explanatory variables enter in the estimation of (5):9

 The log of the initial level of ex ante human capital, ln(Ha,90), to capture
potential catching-up effects. A negative sign for the coefficient a1 would indi-
cate convergence in nativesÕ (residents plus emigrants) human capital among
the countries sampled.
 The log of the skilled migration rate at the beginning of the period, ln(p90), as a
proxy for the migration incentives faced by educated individuals. Ideally, the
incentive effect of migration on human capital investment should be identified
through the impact of migration prospects on expected returns to education.
However, these cannot be computed directly as there are no comparative data on
education premia in developing countries. Using differences in GNI per capita,
on the other hand, raises endogeneity concerns as this variable is strongly cor-
related with human capital. In our benchmark model, we will thus let aside wage
differentials and differences in GNI per capita and use instead ln(p90). A positive
sign for the coefficient a2 indicates that the incentive effect operates (i.e., there
is a brain gain). Still, one may be concerned about possible non-linearities in the
relationship between migration prospects and human capital formation at dif-
ferent income levels. In alternative specifications, we allow for this possibility by
interacting this initial skilled emigration rate, ln(p90), with dummy variables for
whether the country’s income per capita was lower than a given threshold at the
beginning of the period, GNID90. A negative sign for the coefficient a3 would
suggest that the impact of higher liquidity constraints more than offsets that of
higher wage differentials, resulting in a weaker incentive effect in poor coun-
tries. Obviously, robustness checks imply the use of different possible thresholds.
 The population density in 1990, DENS90, as a proxy for the cost of acquiring
education. Clearly, education costs depend on a host of factors such as public
expenditures on general and higher education, distances to schools etc.

9
The data sources are given in the Appendix.

Ó The Author(s). Journal compilation Ó Royal Economic Society 2008

2008 ] BRAIN DRAIN AND HUMAN CAPITAL FORMATION 637
However, public expenditures on education at the beginning of the sample
period (in 1990) are statistically very highly correlated in our sample with the
initial level of human capital H90. This certainly suggests that such expenditures
are effective but the magnitude of the correlation (0.72) precludes any correct
joint estimation of the impact of public expenditures and of possible conver-
gence effects. Population density is likely to reduce distances to schools and,
therefore, to decrease the opportunity cost of education.
 WorkersÕ remittances as a share of GDP, REM90, first because they can relax
credit constraints on human capital investment, and second, because in the
absence of statistics on return migration, they provide an indirect means of
controlling for possible returns in subsequent periods.10
 Regional dummies for sub-Saharan Africa (SSAD) and Latin America (LATD).

2. Data on Human Capital and Migration Rates

Our empirical analysis is based on a new data set on international migration by educa-
tional attainment, namely, on the World Bank sponsored Docquier and Marfouk (2006)
(henceforth DM) data set.11 DM collected census, register and survey data reporting
immigrantsÕ educational levels and countries of birth from 27 OECD countries in 2000
(which accounts for 98% of the OECD immigration stock) and 24 countries in 1990
(91%). For the few remaining countries for which census data were not available, existing
data by country of birth were split across educational levels on the basis of the regional
structure or of the OECD average. They use these data to compute emigration rates by
education level for 195 emigration countries in 2000 and 174 emigration countries in
1990. South–South migration is not taken into account but, on the basis of census data
collected from selected non-OECD countries, DM estimate that about 90% of all highly
skilled migrants live in the OECD area. Descriptive statistics (Docquier and Marfouk,
2006) show a clear decreasing relationship between emigration rates and country size,
with average emigration rates being about 7 times higher for small countries (with
population lower than 2.5 million) than for large countries (with population higher than
25 million). The highest emigration rates are observed in middle income countries
where people have both the incentives and means to emigrate. High income countries
(low incentives) and low income countries (where liquidity constraints are likely to be
more binding) exhibit the lowest rates. This holds true for both total and skilled
migration. Regarding the regional distribution of the brain drain, the most affected
regions appear to be the Caribbean and the Pacific, which consist of relatively small
islands, Sub-Saharan Africa, and Central America. The difference between skilled and
total emigration rates is particularly striking in Africa.
The method used by DM is to rely on receiving country rÕs census or population
register to extract information on immigrants’ country of birth, age, and skill level. Let

10
Indeed, preparing one’s return is known to be a central motivation for remitting and remittances tend
to decline over time as migrants become better integrated in the host country. See Rapoport and Docquier
(2006) for a comprehensive survey of migrants remittances.
11
In an earlier version of this article we used the Carrington and Detragiache (1998) data and found very
similar results. See Beine et al. (2003).

Ó The Author(s). Journal compilation Ó Royal Economic Society 2008

638 THE ECONOMIC JOURNAL [APRIL
r
Mt;sdenote the stock of working-age individuals born in a given country, of skill level s,
s ¼ l,m,h (for low, medium and high skill)12 and living in country r at time
P t. The stock
r
of emigrants from a given country for a given education level, Mt;s ¼ r Mt;s , is then
obtained by summing over receiving countries. Emigration rates by education levels are
then obtained by comparing the number of emigrants to the population at origin with
similar characteristics, Nt,s. Emigration rates are given by
Mt;h Mt;l
pt;h ¼ ; pt ¼
Nt;h þ Mt;h Nt;l þ Mt;l
and the share of high skilled among the total native population (residents and
emigrants included) by
Nt;h þ Mt;h
Ha;t ¼ P  :
s Nt;s þ Mt;s

These steps require collecting data on the size and skill structure of the working-age
population in the origin countries. Population data by age are provided by the United
Nations.13 Data are missing for a small number of countries but can be estimated using
the CIA world factbook.14 Population data are split across educational groups using
international human capital indicators. The DM data set is based on the Barro and Lee
(2001) estimates for most countries. For countries where the Barro and Lee measures
are missing, DM transposed the skill structure of the neighbouring country with the
closest human development index regarding education.
The variables Hp,t and Ha,t are two outputs of the data set. Implicitly, these two
variables are connected in the following way:
ð1  pt ÞHa;t
Hp;t  ð6Þ
1  pt Ha;t  p ð1  Ha;t Þ
t

where p is the average emigration rate of workers without tertiary education.

t
This equation will be useful for our counterfactual experiments. Note that while we
ignored unskilled migration in the theoretical model by setting unskilled migrantsÕ
probability to zero, this is clearly not satisfactory from an empirical perspective. We
therefore include unskilled migration in our computation of the post-migration
human capital stock in (6). This variable will play an important role when we will
introduce counterfactual simulations to estimate the net effect of skilled migration on
human capital formation in Section 5.
To conduct the empirical analysis, and given that we focus on the brain drain impact
on developing countries, our sample excludes high-income countries as well as coun-
tries from the former USSR, Yugoslavia and Czechoslovakia (for consistency between
the 1990 and the 2000 data points), which gives a total sample of 127 developing
countries. We measure the emigration rate of skilled workers as the emigration rate
among individuals with tertiary education: pt ¼ pt,h. As emigration rates are strongly

12
We define high-skill workers as those with tertiary (i.e., post-secondary) education, medium-skill workers as
those with upper-secondary education and low-skill workers as those with less than upper-secondary education.
13
See http://esa.un.org/unpp.
14
See http://www.cia.gov/cia/publications/factbook.

Ó The Author(s). Journal compilation Ó Royal Economic Society 2008

2008 ] BRAIN DRAIN AND HUMAN CAPITAL FORMATION 639
increasing in human capital, we also assume that the minimal or incompressible emi-
gration rate is the one observed among people with primary education: p ¼ pt;l .
t

3. Results
3.1. Econometric Issues
Before we carry out the estimation, we first address some specification issues. A first
important question concerns the exogeneity of the migration rate. When trying to
determine the impact of migration on education, one has to control for the reverse
effect since, on average, the proportion of educated is likely to affect the rate of skilled
migration. This is due to a number of reasons.15 In an attempt to cope with this
endogeneity issue, recent empirical growth analyses (Barro and Sala-I-Martin, 1995;
Hall and Jones, 1999) have been concerned with the use of truly exogenous instru-
ments. In these studies, the following variables have been suggested as candidate
instruments for a first-stage migration equation:
 Life expectancy at birth (LE90), as a proxy for general living conditions;
 The country’s population size (POP90), as small countries tend to be more open
to migration (this is also very clear from the DM data);
 Racial tensions (RAC), a key traditional ÔpushÕ factor;
 The number of emigrants living in the OECD area at the beginning of the period
(MT), to capture the size of the migration network on which prospective emi-
grants can count on;16
 The GDP per capita of the source country, as a proxy for wage differentials –
clearly a driving force of migration.
We retain only two out of these five candidate instruments in our first-stage migration
equation as we have to eliminate the variables for which there is a strong presumption
of a correlation with human capital. This is the case for wage differentials, for obvious
reasons,17 and for life expectancy, the exogeneity of which is questionable given the
fact that longer-lived individuals can enjoy the benefits of education over a longer
period of time. We also exclude racial tensions, for both technical and substantive
reasons. Technically, their introduction would result in a significant drop in the size of
the sample and would therefore lower the comparability with the OLS results.18 More

15
Standard neoclassical models would suggest that a larger stock of human capital may reduce the skill
premium and thus increase skilled migration incentives through higher international wage differentials. On
the other hand, a larger stock of human capital may also generate positive externalities on wages through a
variety of channels emphasised in new growth and new economic geography models (Klenow and Rodriguez-
Clare, 2005).
16
As is well known, larger networks are associated with lower migration costs (especially information-
related ones) and higher expected wages; all else equal, they should act to increase the number of future
migrants. See for example Carrington et al. (1996), Munshi (2003) and Kanbur and Rapoport (2005).
17
As a crude test, the correlation between wage differentials and human capital levels is indeed higher
than 0.5.
18
More precisely, the sample size falls to 59 countries when racial tensions are added to the set of
instruments. We still obtain a positive incentive effect (of a higher magnitude) and conclude in favour of the
exogeneity of the three instruments. The first-stage estimation also supports racial tensions as a strong
instrument at the 10% significance level. The results with this specification are available from the authors
upon request.

Ó The Author(s). Journal compilation Ó Royal Economic Society 2008

640 THE ECONOMIC JOURNAL [APRIL
substantially, it could well be that racial tensions impact on human capital formation,
especially if ethnic discrimination is a serious issue.19 We are therefore left with two
instruments: total population size, and migration stocks at the beginning of the period.
At a theoretical level, there is no obvious reason why the demographic size of a country
should be correlated with its education level. Likewise, once we control for remittances,
there is no a priori reason why migration networks at destination should impact on
human capital formation beyond their effect on migration prospects and incentives
(captured by our instrumentation equation). Since we have only one endogenous
explanatory variable, the number of instruments is large enough to test for exogeneity
of the retained instruments using a traditional overidentification test.
At an empirical level, the validity of our instruments rests on two conditions: the
instruments should first be significantly correlated with the migration rate, and the
exogeneity condition requires that they should be uncorrelated with the error term in
(5).
Equation (7) reports the results of an OLS regression of the migration equation for
the full sample on the two selected instruments (t-statistics in parenthesis):
p ¼ 1:20 þ 0:454 lnðMT Þ  0:518 lnðPOP Þ
ð2:24Þ ð8:46Þ ð13:92Þ ð7Þ
R 2 ¼ 0:509; Nobs ¼ 127; F ¼ 97:14:

The two instruments are significant at the 1% significance level and are therefore
kept throughout the analysis. As expected, population size enters with a negative sign.
The sign of ln(MT) is also in line with intuition: a higher initial stock of migrants
stimulates future emigration. Together, the variables ln(MT) and ln(POP) account for
more than 50% of the migration variability, which is quite satisfactory for a cross-section
analysis. A more formal test relies on the value of the F statistics testing the null
hypothesis that all coefficients in (7) jointly equal zero. The test reveals that this null
hypothesis is clearly rejected, suggesting that the two instruments are strong. Finally,
given that we have more instruments than endogenous variables, a J-test of overiden-
tification was also run to assess the exogeneity property of the retained instruments, the
p-values of which are reported in the result tables below. For the parsimonious speci-
fication, the test supports the exogeneity assumption of the two instruments, thus
providing additional confidence that our instruments are indeed uncorrelated with the
human capital variable.

3.2. Testing for Incentive Effects

We now turn to the estimation of (5). Table 1 reports the estimation results for the full
specification and for a more parsimonious model from which insignificant variables
such as LATD, DENS90 and REM90 were excluded. Exclusion of these variables leads to
a significant increase in the number of countries included (from 103 to 127) as the
variable REM90 displays many missing values. The results appear to be very robust
across specifications and estimation techniques (OLS and IV). Skilled migration
appears to increase gross (or ex ante, or pre-migration) human capital stocks
19
See Tremblay (2001) and Docquier and Rapoport (2003).

Ó The Author(s). Journal compilation Ó Royal Economic Society 2008

2008 ] BRAIN DRAIN AND HUMAN CAPITAL FORMATION 641
Table 1
Estimation Results: Benchmark Regressions
Dependent variable ¼ gross investment in human capital.

Variable (1) (2) (3) (4) (5)

constant 0.0035 0.0798 – 0.0214 0.0798
[0.04] [1.12] – [0.20] [1.12]
ln(p90) 0.0487* 0.0421** 0.0481** 0.0573** 0.0499**
[1.86] [2.03] [2.29] [2.22] [2.30]
ln(H90) 0.2240*** 0.2211*** 0.1990*** 0.2238*** 0.2216***
[6.37] [6.30] [9.12] [6.38] [6.37]
SSAD 0.382*** 0.325*** 0.299*** 0.386*** 0.323***
[3.98] [3.91] [4.36] [4.13] [3.90]
LATD 0.0258 – – 0.0351 –
[0.45] – – [0.59] –
DENS90 0.0998 – – 0.1085 –
[0.92] – – [0.99] –
REM90 0.0051 – – 0.0053 –
[1.12] – – [1.14] –
R2 0.410 0.353 0.763 0.409 0.352
Hausman – – – 0.552 0.484
J test – – – 0.056 0.163
Nobs 103 127 127 103 127

Variables: p90 ¼ skilled emigration rate in 1990. H90 ¼ lagged dependent variable (ex ante proportion of
educated). SSAD ¼ sub-Saharan African dummy.
LATD ¼ dummy for Latin American countries. DENS90 ¼ population density in 1990. REM90 ¼ workersÕ
remittances as % of GDP.
Notes. t-statistics in brackets. White corrections for heteroscedasticity.
Columns 1, 2 and 3: OLS regressions. Columns 4 and 5 : variable instrumental regressions; instruments:
populatition size and stock of migrants in OECD countries.
Hausman and J test report the p-values for respectively the null of no endogeneity of migration rates and the
null of valid instruments (no correlation with error term).
*, ** and *** denote significance at 10, 5 and 1% levels respectively.

significantly. The value of the migration coefficient lies between 0.042 and 0.050 for the
OLS estimate (depending on whether the constant and the insignificant explanatory
variables are included) and is slightly higher (0.050) after instrumenting.20 Taken
literally, this means that doubling the migration propensity of the highly skilled
increases gross human capital formation by 5%. This is not negligible in countries
where the proportion of highly educated typically lies in the 2–8% range and higher
education significantly increases the chance of emigration (by a factor of 5 to 10).
Regarding the other control variables, we find evidence of convergence in human
capital levels among the developing countries sampled. Indeed, the coefficient on the
lagged human capital stock is negative and significant at the 1% threshold in all
specifications. Moreover, in line with the findings of Easterly and Levine (1997), we
find that Sub-Saharan countries display poor performances in terms of human capital
formation. In contrast, population density and the dummy variable for Latin-America
do not seem to exert any significant impact and are therefore omitted in the parsi-
monious specifications. Finally, workersÕ remittances are also insignificant in all
20
The IV results obtained without a constant are not reported here to save space. In this regression, the
estimated incentive effect amounts to 0.057. We obtain similar results with respect to the Hausman test and
the over-identification test.

Ó The Author(s). Journal compilation Ó Royal Economic Society 2008

642 THE ECONOMIC JOURNAL [APRIL
Table 2
Estimation Results: Conditional Effects
Dependent variable ¼ gross investment in human capital.

Variable (1) (2) (3)

constant 0.128* 0.089 0.100
[1.69] [1.08] [1.33]
ln(p90) 0.031* 0.040** 0.036**
[1.86] [2.74] [2.53]
ln(p90)GNID 0.037 0.005 0.012
[1.09] [0.17] [0.47]
ln(H90) 0.237*** 0.224*** 0.228***
[6.08] [5.34] [5.83]
SSAD 0.322*** 0.327*** 0.326***
[3.93] [3.96] [3.95]
R2 0.370 0.353 0.355
Nobs 127 127 127

Variables: p90 ¼ skilled emigration rate in 1990. H90 ¼ lagged dependent variable
(ex ante proportion of educated).
SSAD ¼ sub-Saharan African dummy. GNID ¼ low-income dummy.
Notes. t-statistics in brackets. White Corrections for heteroscedasticity.
In columns (1) (2) and (3), the low income dummies are defined using thresholds
of income per head in 1990 equal respectively to 500, 750 and 900 US$.
*, ** and *** denote significance at respectively 10, 5 and 1% levels.

regressions and are therefore left aside throughout the rest of the empirical analysis.
While the overidenfication test supports the exogeneity of the two instruments in the
parsimonious specification, the Hausman test does not support the need for
accounting for reverse causality. The p-values associated with this test for the two
specifications considered are indeed above the usual significance levels.
Regardless of the retained specification and the estimation method, we always find a
positive incentive effect in the sense that the coefficient of the rate of skilled migration
is significantly positive at a 5% level (10% in column (1)). The benchmark elasticity of
human capital formation to skilled migration is obtained in column (3) of Table 2. In
this best parsimonious specification, we have a2 ¼ 4.81%. Using the standard error of
the coefficient, we can also provide an interval of confidence at 90% for the elasticity.
The lower bound for a2 is equal to 1.37% and the upper bound to 8.25%. Hence, the
incentive effect is definitely positive.

3.3. Testing for Non-linearities

To test for possible non-linearities in (5), we interact skilled migration rates with a
dummy variable for low-income status. To define a Ôpoor countryÕ we use three alter-
native threshold values of the 1990 GNI per head (500, 750 and 900 US$). We augment
the benchmark specification by adding the interaction term ln (p90)GNID to the set
of explanatory variables in (5), where GNID is a dummy variable equal to 1 if country i is
a low income country. The advantage of this specification is that the correlation
between the raw migration rate and the interaction term remains modest, which
moderates the statistical effects of collinearity. Table 2 reports the results with this
specification. As the Hausman test conducted above confirmed the exogeneity of the
Ó The Author(s). Journal compilation Ó Royal Economic Society 2008
2008 ] BRAIN DRAIN AND HUMAN CAPITAL FORMATION 643
migration rate, we only present the OLS results for the specification with interaction
terms.21
On the whole, the results do not provide any evidence of a different impact for the
poorest countries. In all regressions, the interaction term ln (p90)GNID is insignifi-
cant at usual significance levels. Interestingly, the value of the migration coefficient,
ln (p90), seems unaffected by the inclusion of interaction terms. However, one may be
concerned that in the absence of information on income distribution, average income
levels may only imperfectly capture the extent of liquidity constraints. In unreported
regressions, we also interacted skilled migration with a dummy variable POOR for
whether more than 40% of the country’s population live on less than one dollar per
day. As with the previous definition, no significant differences were found between
poor and richer countries, leading us to conclude to the absence of non-linearities in
the skilled migration–human capital formation relationship.

4. Country-specific Results
The cross-sectional results just derived show that migration prospects have a signifi-
cant positive impact on gross human capital formation. From the perspective of
source countries however, what matters is not the number of people who acquire
education but the number of educated who remain in the country after education is
acquired. To evaluate whether the country has experienced a beneficial or a detri-
mental brain drain, one must compare its observed human capital level to some
relevant counterfactual. Since the incentive effect emphasised above relies on skill-
biased migration prospects, a natural counterfactual experiment to make is to com-
pare current human capital levels to their erstwhile value had skilled workers been
allowed to emigrate at the same rate as unskilled workers in 1990 and 2000, i.e.
p90 ¼ p and p00 ¼ p .22 We consider the initial stock of human capital, Ha,1990, as
90 00
given. In other words, people who were educated prior to 1990 are considered as
having done so independently of their chances of migration. It is important to
emphasise that this assumption increases the likelihood of our counterfactual
experiment yielding a negative (detrimental) outcome.
Our simulations are based on the coefficient obtained in the best parsimonious
specification presented in column 3 of Table 1 (i.e., a2 ¼ 0.0481). Since p < p90 , the
cf 90
counterfactual proportion of tertiary educated natives, Ha;2000 , is always lower than the
actual proportion, Ha,2000. Assuming p00 ¼ p , (8) is directly derived from (6). Using
00
(6), it clearly appears that the ex post and ex ante human capital stocks are equal when
skilled workers are allowed to emigrate at the same rate as unskilled workers: this gives
(9). The simulation system is given by the following equations:
cf
Ha;2000 ¼ Ha;2000  a2 lnðp90 =p Þ ð8Þ
90

21
Using the values of p90 predicted by the first stage migration regression leads to similar estimates. These
results are available upon request.
22
For a small number of countries where the unskilled emigration rate is close to zero and, given that we
use a log specification, we impose a lower bound equal to 10% of the skilled emigration rate.

Ó The Author(s). Journal compilation Ó Royal Economic Society 2008

644 THE ECONOMIC JOURNAL [APRIL
cf cf
Hp;2000 ¼ Ha;2000 : ð9Þ

Consequently, for each country in the sample, we measure the human capital gain/
loss associated to the brain drain as the difference between the current and counter-
cf
factual proportions of skilled, that is, by Hp;2000  Hp;2000 . The results of this counter-
factual experiment are apparent from Table 3 which gives the net effect of the brain
drain on the labour force, on the number of skilled workers and on the proportion of
high skill workers residing in their home country. As the latter variable is the
cornerstone of our analysis, we provide a confidence interval evaluated at the 90% rate
(i.e., using the lower and upper bounds of the coefficient a2 from the previous
Section).
The countries are ranked in Table 3 by decreasing gain. As may be seen from the
Table, there are slightly more losers than winners. More importantly, the gains of the
winners are relatively small and exceed 1% of the country’s skilled labour force only in
a handful of cases. In contrast, the losses of the losers can be substantial and exceed
10% in many small Caribbean and Pacific countries.
A more general pattern emerges when the gains and losses in terms of human capital
formation are plotted against two key characteristics: the skilled emigration rate in 1990
and the observed proportion of educated natives in 1990. It appears that the countries
experiencing a beneficial brain drain (the ÔwinnersÕ) generally combine low levels of
human capital and low skilled migration rates, whereas the ÔlosersÕ are typically char-
acterised by high skilled migration rates (above 20%) and/or high proportions of
highly educated in the adult population (above 5%). Figures 1 and 2 give the reduced-
form relationship between the human capital impact of the brain drain and these two
variables. For each relationship, we estimate a quadratric reduced-form adjustment.
The relationships are very significant and exhibit high R2 (respectively 61% and 37%).
Finally, it is striking from Table 3 that the most populated countries (China, India,
Indonesia, Brazil, Egypt, Bangladesh) are all among the winners. Once translated into
absolute numbers, their relatively modest gains more than offset the losses of the many
small countries hard hit by the brain drain. This is more apparent from Table 4, which
gives the results for country groups defined according to demographic size, income
level, and region. In aggregate, there were 116.5 million skilled workers living in the
127 developing countries of our sample in 2000 (representing about 5% of the sam-
ple’s labour force). This number would fall to 113.2 million under the counterfactual
scenario, meaning that according to our computations the brain drain generates a 3%
increase in the total number of skilled workers living in the developing world.
Desegregating by demographic size, income level and region, it is noteworthy that
large countries (with population higher than 25 million) form the only group to
experience a net gain while losses are concentrated on the relatively small countries
(with a population lower than 10 million). For the smallest countries (with a popula-
tion lower than 1 million), the losses are substantial once expressed in relative terms as
they represent a 33% net loss. In contrast, there is no clear pattern for the decom-
position by income levels (2000 classification). Finally, at a regional level, the brain
drain appears to be extremely detrimental in Central America (especially in the
Caribbean), the Pacific region, and to a lower extent in Sub-Saharan Africa, while Asia
and South America experience significant gains.
Ó The Author(s). Journal compilation Ó Royal Economic Society 2008
2008 ] BRAIN DRAIN AND HUMAN CAPITAL FORMATION 645
Table 3
Country-specific Impact of Skilled Migration on Human Capital Counterfactual experiment:
skilled emigration rate ¼ unskilled emigration rate

Countries experiencing Effect on Effect on the Effect on the Confidence

a beneficial brain the labour skilled labour proportion of interval
drain force: DLF force: DSLF skilled: BG on BG (90%)
Argentina 89827 292215 1.5% (0.3%–2.8%)
Venezuela 64675 131002 1.3% (0.1%–2.4%)
Saudi Arabia 9720 113487 1.2% (0.4%–2.1%)
Mongolia 2225 12668 1.2% (0.3%–2.1%)
Maldives 128 1102 1.1% (0.3%–1.9%)
Libya 9186 22575 1.1% (0.2%–2.0%)
Costa Rica 24903 15304 1.0%  (0.3%–2.4%)
Thailand 83572 318506 1.0% (0.2%–1.8%)
Bolivia 27614 26067 0.9% (0.1%–1.9%)
Albania 955 14390 0.9% (0.3%–1.4%)
Oman 713 9331 0.8% (0.3%–1.5%)
Chile 76311 59461 0.8% (0.2%–1.9%)
Bahrain 2095 2589 0.8% (0.0%–1.6%)
Egypt 135204 202416 0.7% (0.0%–1.5%)
Brazil 152218 625298 0.7% (0.2%–1.3%)
Jordan 28054 7439 0.7% (0.5%–1.8%)
Paraguay 6788 13063 0.6% (0.0%–1.2%)
Syria 44301 31541 0.6% (0.2%–1.4%)
Ecuador 79255 17925 0.6% (0.5%–1.7%)
South Africa 152228 74385 0.4% (0.3%–1.2%)
Indonesia 99302 451452 0.4% (0.1%–0.8%)
Swaziland 955 21987 0.4% (0.1%–0.7%)
Bulgaria 28998 28998 0.4% (0.4%–0.4%)
Uruguay 19474 5619 0.4% (0.4%–1.2%)
Solomon Islands 692 531 0.4% (0.1%–0.9%)
India 942212 1513624 0.3% (0.0%–0.7%)
Namibia 930 2118 0.3% (0.0%–0.6%)
Botswana 853 1812 0.3% (0.0%–0.6%)
Bhutan 104 1933 0.2% (0.1%–0.4%)
Burma (Myanmar) 28033 49958 0.2% (0.0%–0.5%)
Bangladesh 75739 122289 0.2% (0.0%–0.5%)
Cote d’Ivoire 10916 10775 0.2% (0.0%–0.4%)
China 741293 1440794 0.2% (0.0%–0.4%)
Colombia 211071 15774 0.2% (0.5%–0.9%)
Turkey 3522 58858 0.2% (0.1%–0.3%)
Burkina Faso 1744 6032 0.2% (0.0%–0.3%)
Chad 1200 4371 0.2% (0.0%–0.3%)
Philippines 1008357 176017 0.1% (1.5%–1.8%)
Nepal 11906 13083 0.1% (0.0%–0.3%)
Iraq 83960 4330 0.1% (0.5%–0.8%)
Yemen 6554 5472 0.1% (0.0%–0.2%)
Madagascar 10964 4998 0.1% (0.1%–0.3%)
Sudan 17086 9840 0.1% (0.1%–0.2%)
Central African Republic 1720 949 0.1% (0.1%–0.2%)
Lesotho 269 423 0.1% (0.0%–0.1%)
Malaysia 92619 815 0.1% (0.5%–0.6%)
Burundi 3234 869 0.0% (0.1%–0.2%)
Niger 949 797 0.0% (0.0%–0.1%)
Vanuatu 657 30 0.0% (0.5%–0.6%)
Ethiopia 46732 2765 0.0% (0.1%–0.2%)
Nigeria 135982 1811 0.0% (0.2%–0.2%)
Djibouti 558 8 0.0% (0.1%–0.2%)
Guinea 3331 175 0.0% (0.1%–0.1%)

Ó The Author(s). Journal compilation Ó Royal Economic Society 2008

646 THE ECONOMIC JOURNAL [APRIL
Table 3
Continued

Countries experiencing Effect on Effect on the Effect on the Confidence

a beneficial brain the labour skilled labour proportion of interval
drain force: DLF force: DSLF skilled: BG on BG (90%)
Equatorial Guinea 851 50 0.0% (0.3%–0.3%)
Benin 4351 292 0.0% (0.1%–0.1%)
Mali 3487 973 0.0% (0.1%–0.0%)
Tanzania 29329 4517 0.0% (0.2%–0.1%)
Pakistan 201568 28980 0.0% (0.3%–0.2%)
Congo, Dem. Rep. 30061 7170 0.0% (0.1%–0.1%)

Countries experiencing Effect on Effect on the Effect on the Confidence

a detrimental brain the labour skilled labour proportion of interval on
drain force: DLF force: DSLF skilled: BG BG (90%)
Malawi 4981 2383 0.1% (0.1%–0.0%)
Mauritania 2306 883 0.1% (0.2%–0.0%)
Zimbabwe 29708 5280 0.1% (0.5%–0.3%)
Rwanda 4121 2685 0.1% (0.1%–0.1%)
Gabon 1971 569 0.1% (0.3%–0.1%)
Mozambique 9725 8087 0.1% (0.1%–0.1%)
Zambia 12489 4958 0.1% (0.3%–0.0%)
Cameroon 19833 8158 0.1% (0.3%–0.0%)
Tunisia 23298 8637 0.2% (0.4%–0.1%)
Guinea-Bissau 1377 777 0.2% (0.2%–0.1%)
Senegal 13889 5724 0.2% (0.3%–0.0%)
Togo 7143 3230 0.2% (0.4%–0.0%)
Hungary 98959 25187 0.2% (0.9%–0.5%)
Iran 280075 74908 0.2% (0.7%–0.3%)
Algeria 43766 31182 0.2% (0.3%–0.1%)
Morocco 84703 40772 0.3% (0.5%–0.0%)
Comoros 1130 769 0.3% (0.4%–0.2%)
Mexico 302138 179516 0.3% (0.5%–0.1%)
Afghanistan 48244 27984 0.3% (0.5%–0.1%)
Papua New Guinea 10581 7277 0.3% (0.5%–0.2%)
Uganda 31811 24376 0.3% (0.4%–0.2%)
Panama 49890 15899 0.4% (1.9%–1.2%)
Angola 18426 17753 0.4% (0.4%–0.4%)
Cambodia 45513 23192 0.4% (0.7%–0.1%)
Congo, Rep. 13246 6755 0.5% (0.9%–0.1%)
Kenya 70493 55544 0.5% (0.7%–0.4%)
Gambia 3310 2942 0.6% (0.6%–0.5%)
Somalia 25277 17720 0.6% (0.8%–0.4%)
Sri Lanka 105462 69618 0.6% (0.9%–0.4%)
Ghana 64804 54217 0.7% (0.9%–0.6%)
Vietnam 458807 289465 0.7% (1.1%–0.4%)
Honduras 43364 22237 0.8% (1.3%–0.2%)
Guatemala 59056 36179 0.8% (1.1%–0.4%)
Sierra Leone 16382 14255 0.9% (1.0%–0.7%)
Kiribati 972 516 0.9% (1.5%–0.3%)
Sao Tome and Principe 452 537 1.0% (0.9%–1.1%)
Dominican Republic 111922 65695 1.2% (2.0%–0.4%)
Palau 290 188 1.3% (1.9%–0.6%)
Liberia 18950 15693 1.5% (1.8%–1.2%)
Laos 48145 37361 1.7% (2.0%–1.4%)
El Salvador 81164 56829 1.7% (2.3%–1.1%)
Nicaragua 61669 38884 1.8% (2.6%–0.9%)
Cuba 273935 187232 2.1% (2.8%–1.3%)

Ó The Author(s). Journal compilation Ó Royal Economic Society 2008

2008 ] BRAIN DRAIN AND HUMAN CAPITAL FORMATION 647
Table 3
Continued

Countries experiencing Effect on Effect on the Effect on the Confidence

a detrimental brain the labour skilled labour proportion of interval on
drain force: DLF force: DSLF skilled: BG BG (90%)
Suriname 6144 5711 2.2% (2.4%–2.1%)
Mauritius 19957 16512 2.3% (2.6%–2.0%)
Micronesia, Fed. States 1595 1136 2.4% (3.1%–1.7%)
Marshall Islands 1216 967 2.9% (3.7%–2.1%)
Cape Verde 5880 5456 3.2% (3.4%–3.0%)
Lebanon 104570 83527 3.8% (4.5%–3.0%)
Haiti 138488 128385 4.0% (4.4%–3.7%)
Seychelles 1951 1667 5.3% (5.9%–4.6%)
Fiji 36598 30356 6.7% (7.7%–5.7%)
St Lucia 6420 5701 7.1% (7.7%–6.4%)
Samoa 6361 5763 7.4% (8.0%–6.9%)
Tonga 4825 4242 8.3% (9.1%–7.5%)
Belize 14090 12004 9.9% (11.2%–8.7%)
Barbados 25201 22986 10.3% (11.0%–9.5%)
Trinidad and Tobago 108326 93869 11.0% (12.2%–9.8%)
Dominica 5954 5560 12.3% (13.0%–11.6%)
Jamaica 238038 217245 14.0% (14.9%–13.0%)
St Vincent & Grenadines 10403 9522 14.1% (14.9%–13.1%)
Antigua and Barbuda 8881 7816 14.7% (16.2%–13.1%)
St Kitts & Nevis 4728 4481 16.9% (17.6%–16.1%)
Guyana 94604 85811 17.8% (19.0%–16.5%)
Grenada 11309 10583 21.5% (22.6%–20.4%)

Effect on the labour force (population aged 25 and more): observed labour force minus counterfactual
labour force
Effect on the skilled labour force (with post-secondary education): observed skilled labour force minus
counterfactual skilled labour force
Effect on the proportion of skilled (BG): observed proportion minus counterfactual proportion (brain gain)
Source: Own calculations

2.00
1.50
1.00
0.50
0.00
%

0 5 10 15 20 25 30 35 40 %
–0.50
–1.00
–1.50 R2 = 0.6142
–2.00
–2.50
–3.00

Fig. 1. Brain Drain Effect and Skilled Migration Rate

Ó The Author(s). Journal compilation Ó Royal Economic Society 2008
648 THE ECONOMIC JOURNAL [APRIL
2.00
0.00
0 5 10 15 20 25 30 %
–2.00
–4.00
–6.00
%

–8.00 R2 = 0.3657
–10.00
–12.00
–14.00
–16.00

Fig. 2. Brain Drain Effect and ResidentsÕ Human Capital

5. Conclusion
The brain drain has long been viewed as a serious constraint on poor countriesÕ
development. However, recent theoretical literature suggests that migration prospects
can raise the expected return to human capital and foster investment in education at
home. This article investigates how these positive and negative effects balance out.
Using recent data on emigration rates by education levels (Docquier and Marfouk,
2006), we find evidence of a positive effect of skilled migration prospects on gross
(pre-migration) human capital levels in a cross-section of 127 developing countries.
More precisely we find that the elasticity of human capital formation to skilled
migration is equal to about 5% and is very stable across specifications and estimation
methods. For each country we then estimate the net effect of the brain drain using
counterfactual simulations. We find that countries combining relatively low levels of
human capital and low skilled emigration rates are more likely to experience a bene-
ficial brain drain (net positive effect) and conversely. There appear to be slightly more
losers than winners and, more importantly the former tend to lose relatively more than
what the latter gain. The situation of many small countries in Sub-Saharan Africa and
Central America, in particular, is extremely worrisome. In contrast, the main globalisers
(China, India, Brazil) all seem to experience non-negligible gains. Once translated into
numbers, these gains outweigh the losersÕ losses, resulting in an overall gain for
developing countries as a whole.
Two central conclusions emerge from the above analysis. First, brain drain migration
contributes to an increase in the number of skilled workers living in the developing
countries. This suggests that the traditionally pessimistic view of the brain drain has no
empirical justification at an aggregate level. Second, the brain drain has important dis-
tributional effects among developing countries, a dimension that has so far been absent
from policy debates.
This article offers initial insights on the general circumstances under which a
beneficial or a detrimental brain drain is obtained. However, further empirical
Ó The Author(s). Journal compilation Ó Royal Economic Society 2008
 Table 4
Results by Country Group
2008 ]

Counterfactual experiment: skilled emigration rate ¼ unskilled emigration rate

Observations in 2000 Counterfactual experiment Brain drain effect

Number In % of Number of In % of Change in Change in the

Labour of skilled the labour Labour skilled the labour the number Change proportion
Force workers force Force workers force of skilled in % of skilled
(LF  1000) (Y  1000) (y ¼ Y/LF) (LF 0  1000) (Y 0  1000) (y 0 ¼ Y 0 /LF 0 ) (Y  Y 0 ) (Y 0 ) (y  y 0 )
By country size (in 2000)
Large (>25 million) 2,001,110 97,370 4.9 2,006,533 93,081 4.6 4,288 4.6 0.2
Upper-Middle (from 10 to 25) 181,152 11,968 6.6 182,472 12,066 6.6 97 0.8 0.0
Lower-Middle (from 2.5 to 10) 80,638 6,525 8.1 81,752 7,104 8.7 578 8.1 0.6
Small (<2.5 million) 10,026 632 6.3 10,419 946 9.1 313 33.1 2.8
By Income Group (in 2000)
Upper-Middle 244,175 26,917 11.0 245,441 26,064 10.6 853 3.3 0.4
Lower-Middle 274,867 29,990 10.9 278,272 30,356 10.9 367 1.2 0.0
Low-Income 1,753,884 59,589 3.4 1,757,464 56,776 3.2 2,813 5.0 0.2
Least Developed 278,320 6,801 2.4 279,192 6,939 2.5 137 2.0 0.0
By region
China 759,550 20,508 2.7 760,291 19,067 2.5 1,441 7.6 0.2
India 480,422 23,060 4.8 481,364 21,547 4.5 1,514 7.0 0.3
Indonesia 103,980 5,199 5.0 104,079 4,748 4.6 451 9.5 0.4
Turkey 33,130 2,816 8.5 33,134 2,757 8.3 59 2.1 0.2
Other Middle East 62,404 5,494 8.8 62,964 5,478 8.7 16 0.3 0.1

Ó The Author(s). Journal compilation Ó Royal Economic Society 2008

Other Asia 344,538 23,927 6.9 347,308 24,045 6.9 118 0.5 0.0
Asia 1,721,620 75,510 4.4 1,726,177 72,163 4.2 3,347 4.6 0.2
Egypt 29,266 3,131 10.7 29,401 2,929 10.0 202 6.9 0.7
Other Northern Africa 33,560 2,264 6.7 33,722 2,322 6.9 58 2.5 0.1
Nigeria 40,174 1,245 3.1 40,310 1,247 3.1 2 0.1 0.0
South Africa 19,914 2,071 10.4 20,066 1,997 10.0 74 3.7 0.4
Other sub-Saharan Africa 174,178 3,164 1.8 174,747 3,387 1.9 222 6.6 0.1
BRAIN DRAIN AND HUMAN CAPITAL FORMATION

Sub-Saharan Africa 234,266 6,480 2.8 235,123 6,630 2.8 150 2.3 0.1
Africa 296,842 11,870 4.0 297,995 11,876 4.0 5 0.0 0.0
Pacific Islands 849 60 7.1 903 103 11.4 43 41.4 4.3
Mexico 45,226 5,111 11.3 45,528 5,290 11.6 180 3.4 0.3
Carribbean 16,577 1,545 9.3 17,520 2,304 13.1 759 32.9 3.8
Other Central America 14,499 1,498 10.3 14,833 1,665 11.2 167 10.0 0.9
Central America 76,302 8,154 10.7 77,882 9,259 11.9 1105 11.9 1.2
649
 650

Table 4
(Continued)

Observations in 2000 Counterfactual experiment Brain drain effect

Number In % of Number of In % of Change in Change in the

Labour of skilled the labour Labour skilled the labour the number Change proportion
Force workers force Force workers force of skilled in % of skilled
0
(LF  1000) (Y  1000) (y ¼ Y/LF) (LF 0  1000) (Y  1000) (y 0 ¼ Y 0 /LF 0 ) (Y  Y 0 ) (Y 0 ) (y  y 0 )
Brazil 87,063 7,313 8.4 87,215 6,688 7.7 625 9.3 0.7
Argentina 20,151 3,970 19.7 20,241 3,678 18.2 292 7.9 1.5
Other South America 53,887 7,410 13.8 54,473 7,232 13.3 177 2.5 0.5
South America 161,101 18,693 11.6 161,929 17,598 10.9 1,095 6.2 0.7
Latin America 237,403 26,846 11.3 239,811 26,856 11.2 10 0.0 0.1

Ó The Author(s). Journal compilation Ó Royal Economic Society 2008

Total 2,272,926 116,495 5.1 2,281,177 113,196 5.0 3,299 2.9 0.2
THE ECONOMIC JOURNAL

Source: Own calculations.

[APRIL
2008 ] BRAIN DRAIN AND HUMAN CAPITAL FORMATION 651
research is needed before policy conclusions can be derived with more confidence.
We see three main directions for future empirical research. First and most obviously,
panel data over longer time periods are needed to confirm the evidence. Second, it is
also important to control for immigrants’ age of entry since only people who
acquired education in their home country can truly be defined as Ôhighly skilled
emigrantsÕ. And third, the sectorial composition of emigration can be of interest,
especially if the brain drain disproportionately affects specific professions (e.g., health
professionals, engineers) whose presence at home strongly conditions the productive
potential of others.

Appendix: Data Sources

 Data on human capital levels (Ha,t and Hp,t), emigration rates (pt and pt) and total stocks
of emigrants (MTt) are taken from Docquier and Marfouk (2006).
 Data on GNI and GDP per capita, population size (POPt) and population density (DENSt),
life expectancy at birth (LEt) and workersÕ remittances (REMt) are taken from the World
Development Indicators (World Bank, 2005). The GNI per capita is measured in US$,
using the Atlas method. The GDP per capita is measured in constant 2000 US$.
 Data on racial tensions (RAC) come from the International Country Risk Guide (1984)
 Regional dummies SSAD and LATD are according to the commonly used World Bank
classification
 Dummies based on poverty rates (POOR) are taken from the United Nations. We use the
1900–2003 average proportion of the population living with less than $1 a day.

University of Luxembourg, Universite´ Libre de Bruxelles and CESifo,

FNRS and IRES, Universite´ Catholique de Louvain,
Bar-Ilan University and EQUIPPE, Universite´s de Lille
Submitted: 24 January 2004
Accepted: 8 February 2007

References
Barro, R.J. and Lee, J.W. (2001). ÔInternational data on educational attainment: updates and implicationsÕ,
Oxford Economic Papers, vol. 53(3), pp. 541–63.
Barro, R. and Sala-I-Martin, X. (1995). Economic Growth, New York: McGraw-Hill.
Beine, M., Docquier, F. and Rapoport, H. (2001). ÔBrain drain and economic growth: theory and evidenceÕ,
Journal of Development Economics, vol. 64 (1), pp. 275–89.
Beine, M., Docquier, F. and Rapoport, H. (2003). ÔBrain drain and LDCsÕ growth: winners and losersÕ, IZA
Discussion Paper No 819, July.
Bhagwati, J.N. and Hamada, K. (1974). ÔThe brain drain, international integration of markets for profes-
sionals and unemploymentÕ, Journal of Development Economics, vol. 1 (1), pp. 19–42.
Carrington, W.J. and Detragiache, E. (1998). ÔHow big is the brain drain?Õ, International Monetary Fund
Working Paper, No 98.
Carrington, W.J., Detragiache, E. and Vishwanath, T. (1996). ÔMigration with endogenous moving costsÕ,
American Economic Review, vol. 86 (4), pp. 909–30.
Commander, S., Kangasniemi, M. and Winters, L.A. (2004). ÔThe brain drain: curse or boon? A survey of the
literatureÕ, in (R. Baldwin and L.A. Winters, eds.), Challenges to Globalization, pp. 235–72, Chicago:
University of Chicago Press.
de Brauw, A. and Giles, J. (2006). ÔMigrant opportunity and the educational attainment of youth in rural
ChinaÕ, IZA Discussion Paper No 2326, September.

Ó The Author(s). Journal compilation Ó Royal Economic Society 2008

652 THE ECONOMIC JOURNAL [ A P R I L 2008 ]
Docquier, F. and Marfouk, A. (2006). ÔInternational migration by education attainment, 1990-2000Õ, in
(C. Ozden and M. Schiff, eds.), International Migration, Brain Drain and Remittances, pp. 151–99, New York:
Palgrave Macmillan.
Docquier, F. and H. Rapoport (2003). ÔEthnic discrimination and the migration of skilled laborÕ, Journal of
Development Economics, vol. 70 (1), pp. 159–72.
Docquier, F. and Rapoport, H. (2008). ÔSkilled migration: the perspective of developing countriesÕ, in
(Baghwati, J. and Hanson, G., eds.) Skilled Migration: Problems and Prospects, New York: Russell Sage
Foundation.
Easterly, W. and Levine, R. (1997). ÔAfrica’s growth tragedy: policies and ethnic divisionsÕ, Quarterly Journal of
Economics, vol. 112 (4), pp. 1203–50.
Grubel, H. G. and Scott, A. (1966). ÔThe international flow of human capitalÕ, American Economic Review,
vol. 56, pp. 268–74.
Hall, R.E. and Jones, C.I. (1999). ÔWhy do some countries produce so much more output per worker than
others?Õ, Quarterly Journal of Economics, vol. 114 (1), pp. 83–116.
Haque, N.U. and Kim, S.-J. (1995). ÔHuman capital flight: impact of migration on income and growthÕ, IMF
Staff Papers, vol. 42 (3), pp. 577–607.
ILO (2006). Competing for Global Talent, Geneva: ILO.
Kanbur, R. and Rapoport, H. (2005). ÔMigration selectivity and the evolution of spatial inequalityÕ, Journal of
Economic Geography, vol. 5 (1), pp. 43–57.
Katz, E. and Rapoport, H. (2005). ÔOn human capital formation with exit optionsÕ, Journal of Population
Economics, vol. 18 (2), pp. 267–74.
Klenow, P.J. and Rodriguez-Clare, A. (2005). ÔExternalities and growthÕ, in (P. Aghion and S. Durlauf, eds.),
Handbook of Economic Growth, Vol. 1, pp. 817–61, Amsterdam: Elsevier-North Holland.
McCormick, B. and Wahba, J. (2000). ÔOverseas unemployment and remittances to a dual economyÕ,
Economic Journal, vol. 110(463), pp. 509–34.
McCulloch, R. and Yellen, J.T. (1977). ÔFactor mobility, regional development and the distribution of incomeÕ,
Journal of Political Economy, vol. 85 (1), pp. 79–96.
McKenzie, D.J. and Rapoport, H. (2006). ÔCan migration reduce educational attainment? Evidence from
MexicoÕ, World Bank Policy Research Working Paper No 3952, June.
Miyagiwa, K. (1991). ÔScale economies in education and the brain drain problemÕ, International Economic
Review, vol. 32 (3), pp. 743–59.
Mountford, A. (1997). ÔCan a brain drain be good for growth in the source economy?Õ, Journal of Development
Economics, vol. 53 (2), pp. 287–303.
Munshi, K. (2003). ÔNetworks in the modern economy: Mexican migrants in the US labour marketÕ, Quarterly
Journal of Economics, vol. 118 (2), pp. 549–99.
Rapoport, H. and Docquier, F. (2006). ÔThe economics of migrantsÕ remittancesÕ, in (S.-C. Kolm and
J. Mercier Ythier, eds.), Handbook of the Economics of Giving, Altruism and Reciprocity, Vol 2, pp. 1135–98,
Amsterdam: Elsevier-North Holland.
Stark, O., Helmenstein, C. and Prskawetz, A. (1997). ÔA brain gain with a brain drainÕ, Economics Letters, vol. 55,
pp. 227–34.
Tremblay, K. (2001). ÔEthnic conflicts, migration and developmentÕ, unpublished PhD dissertation, University
of Paris I-Sorbonne.
Vidal, J.-P. (1998). ÔThe effect of emigration on human capital formationÕ, Journal of Population Economics,
vol. 11 (4), pp. 589–600.
Wong, K.-Y. and Yip, C.K. (1999). ÔEducation, economic growth, and brain drainÕ, Journal of Economic Dynamics
and Control, vol. 23 (5–6), pp. 699–726.
World Bank (2005). World Development Indicators, Washington: The World Bank.

Ó The Author(s). Journal compilation Ó Royal Economic Society 2008