hat is the greatest single class of distortions in the global economy? One at is the greatest single class of distortions in the global economy? One
contender for this title is the tightly binding constraints on emigration ontender for this title is the tightly binding constraints on emigration
from poor countries. Yet the effects of these distortions are little studied rom poor countries. Yet the effects of these distortions are little studied
in economics. Migration economics has focused elsewhere—on n economics. Migration economics has focused elsewhere—on immigration, how
the movement of people affects the economies that receive migrants—while the he movement of people affects the economies that receive migrants—while the
effects of ffects of emigration go relatively neglected. go relatively neglected.
Vast numbers of people in low-income countries want to emigrate from those ast numbers of people in low-income countries want to emigrate from those
countries but cannot. The Gallup World Poll fi ountries but cannot. The Gallup World Poll fi nds that more than 40 percent of nds that more than 40 percent of
adults in the poorest quartile of countries “would like to move permanently to dults in the poorest quartile of countries “would like to move permanently to
another country” if they had the opportunity, including 60 percent or more of adults nother country” if they had the opportunity, including 60 percent or more of adults
in Guyana and Sierra Leone (Pelham and Torres, 2008; Torres and Pelham, 2008). n Guyana and Sierra Leone (Pelham and Torres, 2008; Torres and Pelham, 2008).
Emigration is constrained by many forces, including credit constraints and limited migration is constrained by many forces, including credit constraints and limited
information at the origin (Hatton and Williamson, 2006). However, policy barriers nformation at the origin (Hatton and Williamson, 2006). However, policy barriers
in the destination countries surely play a major role in constraining emigration. The n the destination countries surely play a major role in constraining emigration. The
size of these constraints is apparent in the annual U.S. Diversity Visa Lottery, which ize of these constraints is apparent in the annual U.S. Diversity Visa Lottery, which
allocates permanent emigration slots mainly to developing countries. In fi llocates permanent emigration slots mainly to developing countries. In fi scal year scal year
2010, this lottery had 13.6 million applications for 50,000 visas (U.S. Department 010, this lottery had 13.6 million applications for 50,000 visas (U.S. Department
of State, 2011)—272 applicants per slot. Many other
84 Journal of Economic Perspectives
to policies that restrict emigration, there appear to be trillion-dollar bills on the o policies that restrict emigration, there appear to be trillion-dollar bills on the
sidewalk. The fi idewalk. The fi rst section of this paper reviews existing estimates of the global gains rst section of this paper reviews existing estimates of the global gains
from the reduction of migration barriers. The gains to eliminating those barriers rom the reduction of migration barriers. The gains to eliminating those barriers
amount to large fractions of world GDP—one or two orders of magnitude larger mount to large fractions of world GDP—one or two orders of magnitude larger
than the gains from dropping all remaining restrictions on international fl han the gains from dropping all remaining restrictions on international fl ows of
goods and capital. These estimates are sensitive to assumptions, and in the following oods and capital. These estimates are sensitive to assumptions, and in the following
sections I discuss the (limited) available research on four kinds of assumptions that ections I discuss the (limited) available research on four kinds of assumptions that
underlie these estimates: how migrants affect nonmigrants, the shape of labor nderlie these estimates: how migrants affect nonmigrants, the shape of labor
demand, the effect of location on productivity, and the feasibility of greater migra- emand, the effect of location on productivity, and the feasibility of greater migration
fl ows. These kinds of questions are not the primary focus of the traditional ows. These kinds of questions are not the primary focus of the traditional
research agenda in migration economics; at the end of the paper, I speculate about esearch agenda in migration economics; at the end of the paper, I speculate about
why, and propose a new research agenda. hy, and propose a new research agenda.
Estimates of the Gains From Reducing Migration Barriers stimates of the Gains From Reducing Migration Barriers
Researchers have built models of the world economy to estimate the gains from esearchers have built models of the world economy to estimate the gains from
eliminating various barriers to trade, capital fl liminating various barriers to trade, capital fl ows, and migration. Table 1 summa- ows, and migration. Table 1 summarizes
several recent estimates for policy barriers to trade, and (to my knowledge) all izes several recent estimates for policy barriers to trade, and (to my knowledge) all
existing estimates for barriers to capital fl xisting estimates for barriers to capital fl ows and migration. Even without delving ows and migration. Even without delving
into the details of these studies, the overall pattern is unmistakable and remarkable: nto the details of these studies, the overall pattern is unmistakable and remarkable:
The gains from eliminating migration barriers dwarf—by an order of a magnitude he gains from eliminating migration barriers dwarf—by an order of a magnitude
or two—the gains from eliminating other types of barriers. For the elimination of r two—the gains from eliminating other types of barriers. For the elimination of
trade policy barriers and capital fl rade policy barriers and capital fl ow barriers, the estimated gains amount to less ow barriers, the estimated gains amount to less
than a few percent of world GDP. For labor mobility barriers, the estimated gains are han a few percent of world GDP. For labor mobility barriers, the estimated gains are
often in the range of 50–150 percent of world GDP. ften in the range of 50–150 percent of world GDP.
In fact, existing estimates suggest that even small reductions in the barriers n fact, existing estimates suggest that even small reductions in the barriers
to labor mobility bring enormous gains. In the studies of Table 1, the gains from o labor mobility bring enormous gains. In the studies of Table 1, the gains from
complete elimination of migration barriers are only realized with epic movements omplete elimination of migration barriers are only realized with epic movements
of people—at least half the population of poor countries would need to move to f people—at least half the population of poor countries would need to move to
rich countries. But migration need not be that large in order to bring vast gains. ich countries. But migration need not be that large in order to bring vast gains.
A conservative reading of the evidence in Table 2, which provides an overview of conservative reading of the evidence in Table 2, which provides an overview of
effi ciency gains from partial elimination of barriers to labor mobility, suggests that ciency gains from partial elimination of barriers to labor mobility, suggests that
the emigration of less than 5 percent of the population of poor regions would bring he emigration of less than 5 percent of the population of poor regions would bring
global gains exceeding the gains from total elimination of all policy barriers to lobal gains exceeding the gains from total elimination of all policy barriers to
merchandise trade and all barriers to capital fl erchandise trade and all barriers to capital fl ows. For comparison, currently about ows. For comparison, currently about
200 million people—3 percent of the world—live outside their countries of birth 00 million people—3 percent of the world—live outside their countries of birth
(United Nations, 2009). United Nations, 2009).
Should these large estimated gains from an expansion of international migra- hould these large estimated gains from an expansion of international migration
outrage our economic intuition, or after some consideration, are they at least ion outrage our economic intuition, or after some consideration, are they at least
plausible? We can check these calculations on the back of the metaphorical envelope. lausible? We can check these calculations on the back of the metaphorical envelope.
Divide the world into a “rich” region, where one billion people earn $30,000 per year, ivide the world into a “rich” region, where one billion people earn $30,000 per year,
and a “poor” region, where six billion earn $5,000 per year. Suppose emigrants from nd a “poor” region, where six billion earn $5,000 per year. Suppose emigrants from
the poor region have lower productivity, so each gains just 60 percent of the simple he poor region have lower productivity, so each gains just 60 percent of the simple
Michael A. Clemens 85
earnings gap upon emigrating—that is, $15,000 per year. This marginal gain shrinks arnings gap upon emigrating—that is, $15,000 per year. This marginal gain shrinks
as emigration proceeds, so suppose that the s emigration proceeds, so suppose that the average gain is just $7,500 per year. gain is just $7,500 per year.
If half the population of the poor region emigrates, migrants would gain $23 tril- f half the population of the poor region emigrates, migrants would gain $23 trillion—which
is 38 percent of global GDP. For nonmigrants, the outcome of such a ion—which is 38 percent of global GDP. For nonmigrants, the outcome of such a
wave of migration would have complicated effects: presumably, average wages would ave of migration would have complicated effects: presumably, average wages would
rise in the poor region and fall in the rich region, while returns to capital rise in the ise in the poor region and fall in the rich region, while returns to capital rise in the
rich region and fall in the poor region. The net effect of these other changes could ich region and fall in the poor region. The net effect of these other changes could
theoretically be negative, zero, or positive. But when combining these factors with heoretically be negative, zero, or positive. But when combining these factors with
the gains to migrants, we might plausibly imagine overall gains of 20–60 percent of he gains to migrants, we might plausibly imagine overall gains of 20–60 percent of
global GDP. This accords with the gasp-inducing numbers in Tables 1 and 2. lobal GDP. This accords with the gasp-inducing numbers in Tables 1 and 2.
This calculation suggests a different kind of sanity check on the global estimates: his calculation suggests a different kind ofsanity check on the global estimates:
comparing the price wedges caused by different types of international barriers. If omparing the price wedges caused by different types of international barriers. If
the gains from eliminating barriers to labor mobility are greater than all remaining he gains from eliminating barriers to labor mobility are greater than all remaining
Table 1
Effi ciency Gain from Elimination of International Barriers
(percent of world GDP)
All policy barriers to merchandise trade
1.8 Goldin, Knudsen, and van der Mensbrugghe (1993)
4.1 Dessus, Fukasaku, and Safadi (1999)a
0.9 Anderson, Francois, Hertel, Hoekman, and Martin (2000)
1.2 World Bank (2001)
2.8 World Bank (2001)a
0.7 Anderson and Martin (2005)
0.3 Hertel and Keeney (2006, table 2.9)
All barriers to capital fl ows
1.7 Gourinchas and Jeanne (2006)b
0.1 Caselli and Feyrer (2007)
All barriers to labor mobility
147.3 Hamilton and Whalley (1984, table 4, row 2) c
96.5 Moses and Letnes (2004, table 5, row 4) c
67 Iregui (2005, table 10.3) c,d
122 Klein and Ventura (2007, table 3) e
a These studies assume a positive effect of trade on productivity; the other
trade studies assume no effect.
b Change in consumption rather than GDP.
c Assumes two factors of production, immobile capital, and no differences
in total factor productivity. Estimates from Hamilton and Whalley and from
Moses and Letnes cited here assume no differences in inherent productivity
of migrants and nonmigrants. Some much smaller estimates in Moses and
Letnes assume that poor-country emigrants at the destination are 1
/5 as
productive as nonmigrants at the destination, which (as the authors note in
their footnote 12) is certainly extremely conservative.
d Computable general equilibrium (CGE) model.
e Assumes three factors of production and international differences in total
factor productivity in a dynamic growth model.
86 Journal of Economic Perspectives
gains from eliminating barriers to trade and capital fl ains from eliminating barriers to trade and capital fl ows, we should expect to see ows, we should expect to see
proportionately greater international price wedges between different labor markets roportionately greater international price wedges between different labor markets
than between different goods and capital markets. In fact, this pattern is exactly what han between different goods and capital markets. In fact, this pattern is exactly what
we see. Typical international trade costs, up to and including the border—not just e see. Typical international trade costs, up to and including the border—not just
policy barriers but olicy barriers but all barriers, including distance, language, currency, and infor- barriers, including distance, language, currency, and information—are
the rough equivalent of a 74 percent ad valorem tariff, according to ation—are the rough equivalent of a 74 percent ad valorem tariff, according to
Anderson and van Wincoop (2004, p. 692) nderson and van Wincoop (2004, p. 692)1
; price wedges between the same goods price wedges between the same goods
in different national markets are also of this magnitude (for example, Bradford and n different national markets are also of this magnitude (for example, Bradford and
Lawrence, 2004). For identical fi awrence, 2004). For identical fi nancial instruments, Lamont and Thaler (2003) nancial instruments, Lamont and Thaler (2003)
fi nd that the price rarely differs across the globe by more than 15 percent. Both nd that the price rarely differs across the globe by more than 15 percent. Both
these wedges look small next to the global price wedges for equivalent labor. In hese wedges look small next to the global price wedges for equivalent labor. In
Clemens, Montenegro, and Pritchett (2008), we document gaps in real earnings for lemens, Montenegro, and Pritchett (2008), we document gaps in real earnings for
1 This includes only international trade costs and excludes domestic distribution and retailing costs
behind the border.
Table 2
Effi ciency Gain from Partial Elimination of Barriers to Labor Mobility
Removal of barriers
Net emigration rate
(% origin-region population)
Effi ciency gain
(% world GDP)
Moses and Letnes (2004, 2005) Complete 73.6 96.5
Partial 29.3 54.8
Partial 10.3 22.0
Iregui (2005) Complete 53 67
Partial 24 31
Klein and Ventura (2007) Complete > 99 122
Partial 14.8 20
Partial 7.3 10
Walmsley and Winters (2005) Partial 0.8 0.6
Partial 1.6 1.2
van der Mensbrugghe and Partial 0.8 0.9
 Roland-Holst (2009) Partial 2.0 2.3
Notes: The Moses and Letnes fi gures on emigration rates from are from Moses and Letnes (2005) table 9.3;
fi gures on effi ciency gains are from Moses and Letnes (2004) table 9, scaled to assume equal inherent
labor productivity across countries (for example, 10 percent elimination of wage gap gives $774 billion
gain in table 9, multiplied by the ratio 96.5/9.6 in table 5 to equalize inherent labor productivity, and
divided by world GDP gives 22 percent). Iregui (2005) fi gures are from tables 10.3, 10.6, 10.8, and 10.9.
Klein and Ventura (2007) fi gures are from tables 2 and 7 (emigration rates calculated from population
allocations given 80 percent initial population allocation to poor region). Walmsley and Winters (2005)
fi gures from tables 4 and 11, assuming 80 percent of world population starts out in (net) migrant-sending
countries. Van der Mensbrugghe and Roland-Holst (2009) fi gures come from tables 6 and 7, and likewise
assume 80 percent of world population starts out in (net) migrant-sending countries. World GDP in 2001
is taken to be $32 trillion, doubling (in 2001 dollars) to $64 trillion by 2025.
Economics and Emigration: Trillion-Dollar Bills on the Sidewalk? 87
observably identical, low-skill workers exceeding 1,000 percent between the United bservably identical, low-skill workers exceeding 1,000 percent between the United
States and countries like Haiti, Nigeria, and Egypt. tates and countries like Haiti, Nigeria, and Egypt.
2
 Our analysis suggests that no Our analysis suggests that no
plausible degree of unobservable differences between those who migrate and those lausible degree of unobservable differences between those who migrate and those
who do not migrate comes close to explaining wage gaps that large. ho do not migrate comes close to explaining wage gaps that large.
All of this suggests that the gains from reducing emigration barriers are likely ll of this suggests that the gains from reducing emigration barriers are likely
to be enormous, measured in tens of trillions of dollars. But of course, the exact o be enormous, measured in tens of trillions of dollars. But of course, the exact
magnitudes of the estimates in Tables 1 and 2 are highly sensitive to modeling agnitudes of the estimates in Tables 1 and 2 are highly sensitive to modeling
assumptions. For convenience, I will refer to the st
88 Journal of Economic Perspectives
Figure 1 illustrates the importance of the answers to these questions. Following igure 1 illustrates the importance of the answers to these questions. Following
Bhagwati (1984), consider a world with two countries—one with low wages and one hagwati (1984), consider a world with two countries—one with low wages and one
with high wages. Wages for the low-wage country are on the left-hand vertical axis, ith high wages. Wages for the low-wage country are on the left-hand vertical axis,
and the quantity of labor in that country is measured from left to right, from nd the quantity of labor in that country is measured from left to right, from O toL.
The labor demand curve there is he labor demand curve there is D. Wages for the high-wage country are measured ages for the high-wage country are measured
on the right-hand vertical axis, and quantity of labor in the high-wage country is n the right-hand vertical axis, and quantity of labor in the high-wage country is
measured right to left from easured right to left from O*
 toL. The labor demand curve there is The labor demand curve there is D*
. Thus,
world labor supply is the entire length of the horizontal axis orld labor supply is the entire length of the horizontal axis OO*
. Initial wages in Initial wages in
each country are ach country are w0 andw 0
*. If completely free migration were allowed, the wage If completely free migration were allowed, the wage
rate between the two countries would equalize at the point where the labor demand ate between the two countries would equalize at the point where the labor demand
Figure 1
Determinants of the Gains from Emigration
Note:The migrants gain welfare corresponding to area a + b. In the low-income country, labor gains area c;
owners of other factors (say, capital and land) lose area b + c. In the high-income country, labor loses area
d; owners of other factors gain area d + e. In sum, the global welfare gain is a + e, the shaded area.
D (low-wage country)
D*
 (high-wage country)
w w*
O L L O*
a
w0
*
b c
e d
w0
Wage in low-wage country
Wage in high-wage country
Quantity of labor
Michael A. Clemens 89
curves intersect. An increase in migration—say by having the workers between urves intersect. An increase in migration—say by having the workers between L
and L′ migrate from the low-income to the high-income country—leads to a rela- migrate from the low-income to the high-income country—leads to a relatively
small decrease in the wage rate for the high-income country, a relatively small ively small decrease in the wage rate for the high-income country, a relatively small
rise in the wage rate for the low-income country, and a large rise in income for ise in the wage rate for the low-income country, and a large rise in income for
the migrants themselves. Migrant workers gain; nonmigrant workers gain in the he migrants themselves. Migrant workers gain; nonmigrant workers gain in the
low-income country and lose in the high-income country; owners of other factors ow-income country and lose in the high-income country; owners of other factors
(such as capital and land) lose in the low-income country and gain in the high- such as capital and land) lose in the low-income country and gain in the highincome
country. Adding up these welfare effects, global welfare rises by an amount ncome country. Adding up these welfare effects, global welfare rises by an amount
corresponding to the shaded area of the fi orresponding to the shaded area of the fi gure.
The fi gure makes it clear how the answers to the four questions above shape the gure makes it clear how the answers to the four questions above shape the
size of the global welfare gain. If emigration exerts negative externalities on nonmi- ize of the global welfare gain. If emigration exerts negative externalities on nonmigrants,
the gain could be offset by consequent downward shifts in both demand rants, the gain could be offset by consequent downward shifts in both demand
curves. If labor demand becomes highly elastic at the origin or at the destination, urves. If labor demand becomes highly elastic at the origin or at the destination,
the gain he gain shrinks. To the extent that emigrants are hrinks. To the extent that emigrants are inherently less productive than less productive than
nonmigrant workers at the destination, the true demand curve for their labor lies onmigrant workers at the destination, the true demand curve for their labor lies
further below urther below D*, and the gain also shrinks. Finally, as the size of feasible migration and the gain also shrinks. Finally, as the size of feasible migration
shrinks, hrinks, L′ gets closer to gets closer to L, and once again, the gain shrinks. and once again, the gain shrinks.
In the following sections, I will consider each question in turn. I argue that, n the following sections, I will consider each question in turn. I argue that,
for most of them, economists need much more evidence than we have, but that the or most of them, economists need much more evidence than we have, but that the
existing evidence gives us little reason to believe that the numbers in Tables 1 and 2 xisting evidence gives us little reason to believe that the numbers in Tables 1 and 2
greatly overstate the gains to lowering migration barriers. reatly overstate the gains to lowering migration barriers.
Question 1: What are the External Effects of Migrants on uestion 1: What are the External Effects of Migrants on
Nonmigrants? onmigrants?
Begin with the country of origin. The departure of some people—such egin with the country of origin. The departure of some people—such
as the skilled or talented—from a poor country might reduce the productivity s the skilled or talented—from a poor country might reduce the productivity
of others in that country. Such an effect would tend to offset the gains from f others in that country. Such an effect would tend to offset the gains from
emigration. Externalities like these are often assumed to be so pervasive that migration. Externalities like these are often assumed to be so pervasive that
the literature refers to skilled migration with a pejorative catchphrase—“brain he literature refers to skilled migration with a pejorative catchphrase—“brain
drain”—embodying the assumption. (To see why economists should avoid this rain”—embodying the assumption. (To see why economists should avoid this
term, picture reading a journal article on female labor force participation that erm, picture reading a journal article on female labor force participation that
calls it the “family abandonment rate.”) In this issue, Gibson and McKenzie review alls it the “family abandonment rate.”) In this issue, Gibson and McKenzie review
and critique this literature. nd critique this literature.
But it is not well-established under what conditions the emigration of skilled ut it is not well-established under what conditions the emigration of skilled
workers results in a net depletion, in equilibrium, of the stock of skilled workers in orkers results in a net depletion, in equilibrium, of the stock of skilled workers in
the origin country. Mountford (1997), Stark, Helmenstein, and Prskawetz (1997), he origin country. Mountford (1997), Stark, Helmenstein, and Prskawetz (1997),
and a subsequent literature theorize that when emigration to high-wage countries nd a subsequent literature theorize that when emigration to high-wage countries
becomes possible, even when it is costly and uncertain, the expected value of ecomes possible, even when it is costly and uncertain, the expected value of
human capital rises for all potential migrants. Because not all of those who were uman capital rises for all potential migrants. Because not all of those who were
thus encouraged to invest will leave, the existence of an emigration option for some hus encouraged to invest will leave, the existence of an emigration option for some
people can tend to raise the human capital stock at home. Macro and micro studies eople can tend to raise the human capital stock at home. Macro and micro studies
suggest that this effect is real and large enough to substantially offset the departures uggest that this effect is real and large enough to substantially offset the departures
90 Journal of Economic Perspectives
in some settings (Beine, Docquier, and Rapoport, 2008; Chand and Clemens, 2008; n some settings (Beine, Docquier, and Rapoport, 2008; Chand and Clemens, 2008;
Batista, Lacuesta, and Vicente, 2011; Docquier and Rapoport, forthcoming). atista, Lacuesta, and Vicente, 2011; Docquier and Rapoport, forthcoming).
But for the sake of argument, consider cases where skilled emigration unambigu- ut for the sake of argument, consider cases where skilled emigration unambiguously
lowers the stock of human capital at the origin. A broad theoretical literature usly lowers the stock of human capital at the origin. A broad theoretical literature
posits that human capital externalities shape the development of poor countries (for osits that human capital externalities shape the development of poor countries (for
example, Romer, 1990; Kremer, 1993; Lucas, 1988). If positive human capital exter- xample, Romer, 1990; Kremer, 1993; Lucas, 1988). If positive human capital externalities
are real and large, it is possible that the depletion of human capital stock via alities are real and large, it is possible that the depletion of human capital stock via
emigration infl migration infl icts negative externalities on nonmigrants. However, these externali- icts negative externalities on nonmigrants. However, these externalities
have proven diffi ies have proven diffi cult to observe, their theoretical basis remains unclear, and their cult to observe, their theoretical basis remains unclear, and their
use to justify policy remains shaky. I will discuss each of these in turn. se to justify policy remains shaky. I will discuss each of these in turn.
Human capital externalities are, it turns out, hard to locate and measure in the uman capital externalities are, it turns out, hard to locate and measure in the
wild. The most commonly cited example of externalities that emigrants might impose ild. The most commonly cited example of externalities that emigrants might impose
on those remaining in the origin country involves healthcare workers. But if human n those remaining in the origin country involves healthcare workers. But if human
capital externalities from health workers were a fi apital externalities from health workers were a fi rst-order determinant of basic health rst-order determinant of basic health
conditions, African countries experiencing the largest outfl onditions, African countries experiencing the largest outfl ows of doctors and nurses ows of doctors and nurses
would have systematically worse health conditions than other parts of Africa. In fact, ould have systematically worse health conditions than other parts of Africa. In fact,
those countries have systematically better health conditions (Clemens, 2007). More hose countries have systematically better health conditions (Clemens, 2007). More
broadly, if the external effects of schooling were major and straightforward determi- roadly, if the external effects of schooling were major and straightforward determinants
of economic development, the vast increases in schooling levels across the world ants of economic development, the vast increases in schooling levels across the world
since 1960 would have been accompanied by a substantial rise in total factor produc- ince 1960 would have been accompanied by a substantial rise in total factor productivity.
As Pritchett (2001) points out, nothing like that happened in poor countries. ivity. As Pritchett (2001) points out, nothing like that happened in poor countries.
These facts do not negate the existence of human capital externalities. But they hese facts do not negate the existence of human capital externalities. But they
do suggest that externalities from national stocks of human capital per se—all else o suggest that externalities from national stocks of human capital per se—all else
equal—might be small enough for their effects to be swamped by other forces. qual—might be small enough for their effects to be swamped by other forces.
Furthermore, the theoretical mechanisms of human capital externalities are urthermore, the theoretical mechanisms of human capital externalities are
poorly established in the literature, which has few conclusions in general and even oorly established in the literature, which has few conclusions in general and even
fewer for developing countries (Ciccone and Peri, 2006). Potential mechanisms for ewer for developing countries (Ciccone and Peri, 2006). Potential mechanisms for
human capital externalities include knowledge spillovers, research and develop- uman capital externalities include knowledge spillovers, research and development,
physical health, political leadership, fertility, and capital accumulation (for ent, physical health, political leadership, fertility, and capital accumulation (for
example, Meng and Ye, 2009; Canton, 2009; Spilimbergo, 2009). Docquier, Özden, xample, Meng and Ye, 2009; Canton, 2009; Spilimbergo, 2009). Docquier, Özden,
and Peri (2010) fi nd Peri (2010) fi nd that the effects of emigration on nonmigrants depend critically nd that the effects of emigration on nonmigrants depend critically
on the assumed mechanisms of human capital externalities. If economists are to n the assumed mechanisms of human capital externalities. If economists are to
understand, measure, or predict the external effects of emigration, they require nderstand, measure, or predict the external effects of emigration, they require
greater clarity about these mechanisms. reater clarity about these mechanisms.
For example, the external effects of emigration on nonmigrants depend or example, the external effects of emigration on nonmigrants depend
crucially on the spatial extent of human capital externalities, which is poorly under- rucially on the spatial extent of human capital externalities, which is poorly understood.
If the external effects of human capital act over short distances—such as a tood. If the external effects of human capital act over short distances—such as a
doctor’s care—the effects on nonmigrants from the emigration of human capital octor’s care—the effects on nonmigrants from the emigration of human capital
per se might be small. Most doctors in many African countries already work in cities er se might be small. Most doctors in many African countries already work in cities
(and in nicer neighborhoods), so the marginal effect of their international emigra- and in nicer neighborhoods), so the marginal effect of their international emigration
on people in rural areas and slums could be limited (Clemens, 2009). On the ion on people in rural areas and slums could be limited (Clemens, 2009). On the
other hand, if the external effects of human capital act over large distances—such ther hand, if the external effects of human capital act over large distances—such
as by the generation of public goods like new ideas that can fl s by the generation of public goods like new ideas that can fl ow back home—this ow back home—this
too diminishes the external effects of emigration per se (Kerr, 2008). oo diminishes the external effects of emigration per se (Kerr, 2008).
Economics and Emigration: Trillion-Dollar Bills on the Sidewalk? 91
Here is another example of the need for greater clarity: Let us assume that ere is another example of the need for greater clarity: Let us assume that
greater inequality has negative external effects, perhaps by fostering crime and reater inequality has negative external effects, perhaps by fostering crime and
mistrust. It is conceivable that emigration could exert external effects on non istrust.Itis conceivable that emigrationcouldexert external effectsonnonmigrants migrants
by raising inequality—if for instance a few families with relatives abroad get big y raising inequality—if for instance a few families with relatives abroad get big
remittance fl emittance fl ows. But if this is an argument for restricting migration, it proves a ows. But if this is an argument for restricting migration, it proves a
little too much. Inequality of remittance income would be lower if no families had ittle too much. Inequality of remittance income would be lower if no families had
this opportunity his opportunity or if all families had this opportunity. So it is not clear whether a if all families had this opportunity. So it is not clear whether a
little emigration is “too much” or “too little” by this criterion. Furthermore, if a man ittle emigration is “too much” or “too little” by this criterion. Furthermore, if a man
from Morocco triples his income by moving to France, the effects on inequality rom Morocco triples his income by moving to France, the effects on inequality
of outcomes within Morocco and within France are both contingent: they depend f outcomes within Morocco and within France are both contingent: they depend
on where in Morocco’s income distribution he came from, and where in France’s n where in Morocco’s income distribution he came from, and where in France’s
income distribution he goes to. But the effects on inequality of outcomes for France ncome distribution he goes to. But the effects on inequality of outcomes for France
and Morocco nd Morocco collectively almost certainly declines. Does international inequality almost certainly declines. Does international inequality
exert worse external effects than domestic inequality? Both theory and evidence are xert worse external effects than domestic inequality? Both theory and evidence are
lacking here. acking here.
Even if the mechanism and magnitude of these external effects were crystal- ven if the mechanism and magnitude of these external effects were crystalclear,
there are important concerns about policy instruments that would address lear, there are important concerns about policy instruments that would address
the externalities by limiting or taxing migration fl he externalities by limiting or taxing migration fl ows: whether it is practical to set ows: whether it is practical to set
correct Pigovian taxes, statically or dynamically, and the extent to which the assump- orrect Pigovian taxes, statically or dynamically, and the extent to which the assumptions
justifying Pigovian taxes hold here. ions justifying Pigovian taxes hold here.
An economic case for emigration taxes or restrictions based on human capital n economic case for emigration taxes or restrictions based on human capital
externalities would require fabulous amounts of information. A social planner selec- xternalities would require fabulous amounts of information. A social planner selectively
restricting skilled emigration from each country would need a vast database ively restricting skilled emigration from each country would need a vast database
of domestic labor market conditions for hundreds of skilled occupations, as well as f domestic labor market conditions for hundreds of skilled occupations, as well as
reliable estimates of the socioeconomic externalities conveyed by those of different eliable estimates of the socioeconomic externalities conveyed by those of different
professions in different locations at different times: entrepreneurs, nurses, engineers, rofessions in different locations at different times: entrepreneurs, nurses, engineers,
and others. nd others.
4
 In practice such restrictions end up indiscriminate—such as the policy In practice such restrictions end up indiscriminate—such as the policy
of the United Kingdom’s National Health Service to ban recruitment from most f the United Kingdom’s National Health Service to ban recruitment from most
developing countries, regardless of conditions there. The optimal future timepath of eveloping countries, regardless of conditions there. The optimal future timepath of
such restrictions is even less clear. Should the emigration of skilled workers from the uch restrictions is even less clear. Should the emigration of skilled workers from the
Democratic Republic of the Congo, one of the poorest countries on Earth, be “tempo- emocratic Republic of the Congo, one of the poorest countries on Earth, be “temporarily”
restricted so that those workers’ positive externalities somehow spark growth arily” restricted so that those workers’ positive externalities somehow spark growth
4 Ng (2004) makes this argument formally: In the presence of an externality that cannot be directly
removed, achieving a second-best welfare outcome via intervention (such as a Pigovian emigration tax)
can be impossible if: 1) there are multiple substitute/complement relationships of unknown direction
and sign among different goods in the market; and 2) the costs of administering the intervention are
large. Ng (p. 202) does argue that a probabilistic “third-best” outcome may be achievable if there is at
least good information on the size and sign of the original externality, but poor information on the
size and sign of other relationships in the economy. Thus, if we held diffuse priors about the harm to
emigrants from restricting emigration, but narrow priors about the positive effects on non-emigrants
from restricting emigration, a probabilistic third-best outcome might be achievable by emigration restrictions.
But regulating emigration based on human capital externalities faces the opposite situation: the
magnitude of the harm to potential emigrants from emigration restrictions is clearly negative and can be
large, whereas the size and sign of the human capital externalities that the intervention seeks to correct
are theoretically and empirically uncertain.
92 Journal of Economic Perspectives
there? Even if that happened—for which there is no clear economic evidence—such here? Even if that happened—for which there is no clear economic evidence—such
“temporary” restrictions might need to last for centuries before Congo caught up to temporary” restrictions might need to last for centuries before Congo caught up to
the destination countries and the restrictions became unnecessary. he destination countries and the restrictions became unnecessary.
5
What is more, policy measures to tax or limit emigration often rest on assump- hat is more, policy measures to tax or limit emigration often rest on assumptions
that bear closer examination. In one of the best-known proposals, Bhagwati ions that bear closer examination. In one of the best-known proposals, Bhagwati
and Dellalfar (1973) argue for a Pigovian tax on skilled emigration, to compensate nd Dellalfar (1973) argue for a Pigovian tax on skilled emigration, to compensate
their countries of origin (for additional discussion, see Bhagwati and Hamada, 1974; heir countries of origin (for additional discussion, see Bhagwati and Hamada, 1974;
Wilson, 2008). Their argument is that skilled workers convey a positive externality ilson, 2008). Their argument is that skilled workers convey a positive externality
on other workers in the same country, such as facilitating the adoption of foreign n other workers in the same country, such as facilitating the adoption of foreign
technologies. Their argument does not rest on fi echnologies. Their argument does not rest on fi scal effects, such as the loss of scal effects, such as the loss of
public subsidies for tuition spent on those who later emigrate, but rather on the ublic subsidies for tuition spent on those who later emigrate, but rather on the
pure external effects of having smart and talented people nearby. ure external effects of having smart and talented people nearby.
There are at least two fundamental problems with this idea. First, it assumes here are at least two fundamental problems with this idea. First, it assumes
that skilled labor emigration is not already taxed. But many skilled workers face hat skilled labor emigration is not already taxed. But many skilled workers face
binding migration restrictions that are the economic equivalent of large taxes. The inding migration restrictions that are the economic equivalent of large taxes. The
United States strictly rations its visas for temporary and permanent employment- nited States strictly rations its visas for temporary and permanent employmentbased
skilled migration, especially from large countries like India, and most ased skilled migration, especially from large countries like India, and most
physicians from the developing world face large nonvisa migration barriers such hysicians from the developing world face large nonvisa migration barriers such
as the requirement to repeat medical residency for U.S. licensing. Just as nontariff s the requirement to repeat medical residency for U.S. licensing. Just as nontariff
trade barriers have a tariff equivalent, quotas and licensing restrictions on the rade barriers have a tariff equivalent, quotas and licensing restrictions on the
movement of skilled workers have a migration tax equivalent. International gaps ovement of skilled workers have a migration tax equivalent. International gaps
in real earnings for high-skill workers are very high: 500–1,000 percent for some n real earnings for high-skill workers are very high: 500–1,000 percent for some
professors, computer programmers, and health workers (Clemens, 2009). Even rofessors, computer programmers, and health workers (Clemens, 2009). Even
if only a small fraction of these gaps is due to policy restrictions, the economic f only a small fraction of these gaps is due to policy restrictions, the economic
equivalent of a large emigration tax is already broadly applied. quivalent of a large emigration tax is already broadly applied.
Second, Coase (1960) taught us that the mere existence of an externality does econd, Coase (1960) taught us that the mere existence of an externality does
not imply that a Pigovian tax maximizes welfare. In a world of transaction costs and ot imply that a Pigovian tax maximizes welfare. In a world of transaction costs and
externalities, welfare is maximized if the property right is assigned to the party with xternalities, welfare is maximized if the property right is assigned to the party with
a higher cost of reducing the externality. Requiring skilled emigrants to pay a tax, higher cost of reducing the externality. Requiring skilled emigrants to pay a tax,
rather than requiring non-emigrants to pay potential emigrants to stay, assigns owner- ather than requiring non-emigrants to pay potential emigrants to stay, assigns ownership
of emigrants’ positive externalities to non-emigrants. But in settings of skilled hip of emigrants’ positive externalities to non-emigrants. But in settings of skilled
emigration, it is not obvious who bears the higher cost. Taxing an emigrant Filipino migration, it is not obvious who bears the higher cost. Taxing an emigrant Filipino
registered nurse at 10 percent of foreign income per year for 10 years might raise egistered nurse at 10 percent of foreign income per year for 10 years might raise
$40,000. 40,000.
6
 But the all-inclusive cost of eliminating the external effect by a different But the all-inclusive cost of eliminating the external effect by a different
5 Real per capita income in the United States is now 150 times real per capital income in the Democratic
Republic of the Congo (as measured by the World Bank at purchasing power parity). Assume (heroically!)
that human capital externalities are enormous, so that restricting emigration from Congo could
raise its real per capita growth rate from roughly zero to about 4 percent per year. Given that real per
capita growth in the United States is historically about 2 percent per year, it would take about two and a
half centuries before emigration restrictions became unnecessary.
6 Bhagwati and Dellalfar (1973) suggest a tax of 10 percent on the after-U.S.-tax income of skilled
emigrants from low-income countries working in the United States. Assuming an annual income in the
United States of $60,000 and U.S. tax rate of 30 percent, such an emigration tax would yield $42,000
over 10 years.
Michael A. Clemens 93
route—training a new registered nurse in the Philippines—can be less than $12,000. oute—training a new registered nurse in the Philippines—can be less than $12,000.
The former policy might be globally impoverishing relative to the latter. he former policy might be globally impoverishing relative to the latter.
Further complications arise from the implicit assumption that non-emigrants urther complications arise from the implicit assumption that non-emigrants
hold property rights in the positive externalities of skilled migrants. If non- old property rights in the positive externalities of skilled migrants. If nonemigrants
own these rights, do they also own any negative externalities the migrants own these rights, do they also own any negative externalities the
emigrants would have provided by staying—like contributions to urban conges- migrants would have provided by staying—like contributions to urban congestion
or to pollution? And who decides whose positive externalities are owned by ion or to pollution? And who decides whose positive externalities are owned by
whom? Presumably, an American doctor’s decision not to provide care in Haiti hom? Presumably, an American doctor’s decision not to provide care in Haiti
causes the same loss of positive externality to Haitians as a Haitian doctor’s deci- auses the same loss of positive externality to Haitians as a Haitian doctor’s decision
to leave Haiti, but few would consider taxing the American doctor’s decision. ion to leave Haiti, but few would consider taxing the American doctor’s decision.
So far I have discussed externalities at the origin, as this piece seeks to shine o far I have discussed externalities at the origin, as this piece seeks to shine
light on emigration rather than immigration. But I now turn briefl ight on emigration rather than immigration. But I now turn briefl y to the issue y to the issue
of negative externalities imposed on people who already live at the destination f negative externalities imposed on people who already live at the destination
when migrants arrive there. The arrival of migrants could, for example, decrease hen migrants arrive there. The arrival of migrants could, for example, decrease
the availability of unpriced public goods at the destination like open space, clean he availability of unpriced public goods at the destination like open space, clean
air, publicly-funded amenities, and a degree of cultural homogeneity that may be ir, publicly-funded amenities, and a degree of cultural homogeneity that may be
valued by nonmigrants. These too would tend to reduce the global welfare gains of alued by nonmigrants. These too would tend to reduce the global welfare gains of
greater mobility. reater mobility.
Economics knows little about the mechanisms and magnitudes of such exter- conomics knows little about the mechanisms and magnitudes of such externalities
at the destination, particularly under large-scale emigration. These deserve alities at the destination, particularly under large-scale emigration. These deserve
study. But there is little reason at present to think that they would greatly alter the tudy. But there is little reason at present to think that they would greatly alter the
message of Tables 1 and 2. First, the literature contains no documented case of large essage of Tables 1 and 2. First, the literature contains no documented case of large
declines in GDP or massive declines in public-service provision at the destination eclines in GDP or massive declines in public-service provision at the destination
caused by immigration. Second, century-old issues of the aused by immigration. Second, century-old issues of the American Economic Review
and theJournal of Political Economy extensively discuss concerns that any further extensively discuss concerns that any further
emigration might degrade the American economy and society (for example, Hall, migration might degrade the American economy and society (for example, Hall,
1913; Kohler, 1914). Since then the American population has quadrupled—with 913; Kohler, 1914). Since then the American population has quadrupled—with
much of the rise coming from increasingly diverse immigration to already settled uch of the rise coming from increasingly diverse immigration to already settled
areas—and the United States remains the world’s leading economy, with much reas—and the United States remains the world’s leading economy, with much
greater availability of publicly-funded amenities than a century ago. Third, there reater availability of publicly-funded amenities than a century ago. Third, there
are also many plausible positive externalities from increased immigration. These re also many plausible positive externalities from increased immigration. These
include spatial aggregation economies in high-skill labor (for example, Glaeser nclude spatial aggregation economies in high-skill labor (for example, Glaeser
and Maré, 2001) and the effects of low-skill labor availability on the productivity of nd Maré, 2001) and the effects of low-skill labor availability on the productivity of
high-skill labor, particularly women’s labor (for example, Kremer and Watt, 2009; igh-skill labor, particularly women’s labor (for example, Kremer and Watt, 2009;
Cortes and Tessada, forthcoming). Fourth, all serious economic studies of the ortes and Tessada, forthcoming). Fourth, all serious economic studies of the
aggregate fi ggregate fi scal effects of immigration have found them to be very small overall— scal effects of immigration have found them to be very small overall—
small and positive at the federal level (Auerbach and Oreopoulos, 1999; Lee and mall and positive at the federal level (Auerbach and Oreopoulos, 1999; Lee and
Miller, 2000), small and negative at the state and local level (Congressional Budget iller, 2000), small and negative at the state and local level (Congressional Budget
Offi ce, 2007). ce, 2007).
Here again, even if we had solid evidence that immigration exerted clear and ere again, even if we had solid evidence that immigration exerted clear and
large net negative externalities to those at the destination, an economic justifi arge net negative externalities to those at the destination, an economic justifi cation
for internalizing those externalities with quotas or taxes would face unanswered or internalizing those externalities with quotas or taxes would face unanswered
questions. If people’s taste for cultural homogeneity justifi uestions. If people’s taste for cultural homogeneity justifi es limits on immigration es limits on immigration
from abroad, could a taste for cultural homogeneity also justify blocking certain rom abroad, could a taste for cultural homogeneity also justify blocking certain
94 Journal of Economic Perspectives
kinds of internal migration to a neighborhood or city that has, to its current resi- inds of internal migration to a neighborhood or city that has, to its current residents,
a pleasing degree of cultural homogeneity? This raises further diffi ents, a pleasing degree of cultural homogeneity? This raises further diffi cult issues cult issues
of the extent to which preferences for homogeneity are endogenous to exposure, f the extent to which preferences for homogeneity are endogenous to exposure,
so that greater immigration might alter the pre-existing preference. And returning o that greater immigration might alter the pre-existing preference. And returning
to the issue of property rights, any taxes or quotas to internalize the negative exter- o the issue of property rights, any taxes or quotas to internalize the negative externalities
of immigration at the destination require an assumption: that migrants are alities of immigration at the destination require an assumption: that migrants are
responsible for their esponsible for their negative externalities at the destination. But as I have discussed externalities at the destination. But as I have discussed
above, some economists assume that migrants’ home countries—not the migrants bove, some economists assume that migrants’ home countries—not the migrants
themselves—own migrants’ hemselves—own migrants’ positive externalities at the origin. It is doubtless possible, externalities at the origin. It is doubtless possible,
with some effort, to lay out a theory in which societies own the positive externali- ith some effort, to lay out a theory in which societies own the positive externalities
of migrants in their origin country but societies are not the owners of ies of migrants in their origin country but societies are not the owners of negative
externalities of migrants in their destination country. The economics literature has xternalities of migrants in their destination country. The economics literature has
barely begun to address these issues. arely begun to address these issues.
In short, there is little in the admittedly scanty literature so far to support the n short, there is little in the admittedly scanty literature so far to support the
notion that externalities from labor mobility would greatly affect the global welfare otion that externalities from labor mobility would greatly affect the global welfare
estimates presented earlier in this paper. stimates presented earlier in this paper.
Question 2: What is the Elasticity of Labor Demand at the Origin uestion 2: What is the Elasticity of Labor Demand at the Origin
and Destination? nd Destination?
Economists studying international migration have given much of their energy conomists studying international migration have given much of their energy
to estimating how the movement of emigrants affects the wages of nonmigrants. For o estimating how the movement of emigrants affects the wages of nonmigrants. For
example, does the arrival of immigrants lower wages for incumbent workers? In a U.S. xample, does the arrival of immigrants lower wages for incumbent workers? In a U.S.
context, Borjas (2003) and Borjas and Katz (2007) argue that low-wage workers do ontext, Borjas (2003) and Borjas and Katz (2007) argue that low-wage workers do
experience a modest decline in nominal wages from immigration. On the other side, xperience a modest decline in nominal wages from immigration. On the other side,
Card (2009) and Ottaviano and Peri (forthcoming) fi ard (2009) and Ottaviano and Peri (forthcoming) fi nd that millions of recent immi- nd that millions of recent immigrants
to the United States have caused the average worker’s nominal wages to decline rants to the United States have caused the average worker’s nominal wages to decline
a few percent—if at all—while Cortes (2008) fi few percent—if at all—while Cortes (2008) fi nds that immigration lowered the price nds that immigration lowered the price
of a typical consumption basket about half of 1 percent. The mass migrations of the f a typical consumption basket about half of 1 percent. The mass migrations of the
nineteenth century likely caused a cumulative decline of 1 or 2 percentage points ineteenth century likely caused a cumulative decline of 1 or 2 percentage points
each decade in wages at the destination (Hatton and Williamson, 1994). ach decade in wages at the destination (Hatton and Williamson, 1994).
Conversely, does the departure of emigrants raise the wages of non-emigrants onversely, does the departure of emigrants raise the wages of non-emigrants
in the origin country? Mishra (2007) fi n the origin country? Mishra (2007) fi nds that the vast emigration of Mexicans to nds that the vast emigration of Mexicans to
the United States between 1970 and 2000 may have caused an 8 percent increase he United States between 1970 and 2000 may have caused an 8 percent increase
in Mexicans’ nominal wages in Mexico. Economic historians have evidence that n Mexicans’ nominal wages in Mexico. Economic historians have evidence that
comparable increases in home wages were caused by mass emigration from Sweden omparable increases in home wages were caused by mass emigration from Sweden
(Karlström, 1985) and Ireland (Hatton and Williamson, 1993; O’Rourke, 1995). Karlström, 1985) and Ireland (Hatton and Williamson, 1993; O’Rourke, 1995).
These estimates are roughly in line with the elasticities used in the global welfare hese estimates are roughly in line with the elasticities used in the global welfare
estimates of Tables 1 and 2. For example, in the model of Moses and Letnes (2004), stimates of Tables 1 and 2. For example, in the model of Moses and Letnes (2004),
a 10 percent removal of emigration barriers generates a 3–4 percent increase in 10 percent removal of emigration barriers generates a 3–4 percent increase in
wages for non-emigrants at the origin, and a 2.5 percent decline at the destination. ages for non-emigrants at the origin, and a 2.5 percent decline at the destination.
Even substantial adjustment of these elasticities is unlikely to alter the estimated ven substantial adjustment of these elasticities is unlikely to alter the estimated
effi ciency gains a great deal. ciency gains a great deal.
Economics and Emigration: Trillion-Dollar Bills on the Sidewalk? 95
Of course, these elasticities could be different at much higher levels of emigra- f course, these elasticities could be different at much higher levels of emigration.
The literature gives no clear support for such a pattern, however, even under ion. The literature gives no clear support for such a pattern, however, even under
greatly increased migration. In historical cases of large reductions in barriers to reatly increased migration. In historical cases of large reductions in barriers to
labor mobility between high-income and low-income populations or regions, abor mobility between high-income and low-income populations or regions,
those with high wages have not experienced a large decline. For example, wages hose with high wages have not experienced a large decline. For example, wages
of whites in South Africa have not shown important declines since the end of the f whites in South Africa have not shown important declines since the end of the
apartheid regime (Leibbrandt and Levinsohn, 2011), despite the total removal of partheid regime (Leibbrandt and Levinsohn, 2011), despite the total removal of
very large barriers to the physical movement and occupational choice of a poor ery large barriers to the physical movement and occupational choice of a poor
population that outnumbered the rich population six to one. The recent advent opulation that outnumbered the rich population six to one. The recent advent
of unlimited labor mobility between some Eastern European countries and Great f unlimited labor mobility between some Eastern European countries and Great
Britain, though accompanied by large and sudden migration fl ritain, though accompanied by large and sudden migration fl ows, has not caused ows, has not caused
important declines in British wages (Blanchfl mportant declines in British wages (Blanchfl ower and Shadforth, 2009). ower and Shadforth, 2009).
Further, even if emigrants modestly depress wages when they arrive at the desti- urther, even if emigrants modestly depress wages when they arrive at the destination,
this does not justify restricting movement by the standard welfare economics ation, this does not justify restricting movement by the standard welfare economics
analysis. Such effects represent “pecuniary” externalities rather than “technical” nalysis. Such effects represent “pecuniary” externalities rather than “technical”
externalities. The human capital externalities discussed in the previous section, xternalities. The human capital externalities discussed in the previous section,
along with common examples like belching smokestacks, are examples of technical long with common examples like belching smokestacks, are examples of technical
externalities. Pecuniary externalities, in contrast, operate through the price mecha- xternalities. Pecuniary externalities, in contrast, operate through the price mechanism:
for example, my decision not to place a bid on the house you are selling may ism: for example, my decision not to place a bid on the house you are selling may
lower the price you can receive from an alternative buyer. Pecuniary externalities ower the price you can receive from an alternative buyer. Pecuniary externalities
are a near-universal feature of economic decisions. In standard economic analysis, re a near-universal feature of economic decisions. In standard economic analysis,
they offer no welfare justifi hey offer no welfare justifi cation for taxation or regulation of those decisions. cation for taxation or regulation of those decisions.
7
For example, research on domestic labor movements has found—to the or example, research on domestic labor movements has found—to the
surprise of few—that movement of labor from one city to another tends to modestly urprise of few—that movement of labor from one city to another tends to modestly
lower wages at the destination (Boustan, Fishback, and Cantor, 2010), and that the ower wages at the destination (Boustan, Fishback, and Cantor, 2010), and that the
entry of women into the labor force can modestly lower men’s wages (Acemo ntry of women into the labor force can modestly lower men’s wages (Acemog ˘lu,
Autor, and Lyle, 2004). However, no economist would argue that these facts alone utor, and Lyle, 2004). However, no economist would argue that these facts alone
signify negative externalities that reduce social welfare and should be adjusted with ignify negative externalities that reduce social welfare and should be adjusted with
a Pigovian tax on those who move between cities or on women entering the work- Pigovian tax on those who move between cities or on women entering the workforce,
because these externalities seem to be almost purely pecuniary. Similarly, orce, because these externalities seem to be almost purely pecuniary. Similarly,
economists would be virtually unanimous against imposing a tax on new domestic conomists would be virtually unanimous against imposing a tax on new domestic
competitors on the grounds that they imposed costs on existing fi ompetitors on the grounds that they imposed costs on existing fi rms, because rms, because
again such externalities are pecuniary. Of course, this argument need not imply gain such externalities are pecuniary. Of course, this argument need not imply
that policies to help low-wage U.S. workers in some manner are socially undesirable, hat policies to help low-wage U.S. workers in some manner are socially undesirable,
only that such policies should be based on concerns over equity or building human nly that such policies should be based on concerns over equity or building human
capital, rather than on standard effi apital, rather than on standard effi ciency justifi ciency justifi cations. cations.
7 This classifi cation can be subtle. For example, McKenzie and Rapoport (forthcoming) fi nd that some
children in Mexico drop out of high school as a result of the emigration of a household member. Should
we treat this fi nding as an externality? If the behavior occurs because those children see that the returns
to emigration are much higher than the returns to education, then their behavior need not impose
an externality. However, a negative externality could arise if high school graduates in Mexico convey
substantial positive technical externalities on other workers. The literature has not established the theoretical
extent or empirical magnitude of such externalities.
96 Journal of Economic Perspectives
Question 3: Is Labor Productivity Mostly about Who You Are, Or uestion 3: Is Labor Productivity Mostly about Who You Are, Or
Where You Are? here You Are?
Existing estimates of the effi xisting estimates of the effi ciency gains from greater emigration hinge on ciency gains from greater emigration hinge on
a critical assumption: How productive will migrants be at the destination? Many critical assumption: How productive will migrants be at the destination? Many
have low productivity where they now are, in poor countries. How much of that ave low productivity where they now are, in poor countries. How much of that
low productivity moves with them? Klein and Ventura (2007) assume that migrants’ ow productivity moves with them? Klein and Ventura (2007) assume that migrants’
productivity is about 68 percent of the productivity of nonmigrants at the destination; roductivity is about 68 percent of the productivity of nonmigrants at the destination;
Moses and Letnes (2004) estimate scenarios where this number is anything from 20 oses and Letnes (2004) estimate scenarios where this number is anything from 20
to 100 percent. The assumption has fi o 100 percent. The assumption has fi rst-order effects on their effi rst-order effects on their effi ciency estimates. ciency estimates.
We can observe the earnings of today’s migrants in destination countries to e can observe the earnings of today’s migrants in destination countries to
learn something about their productivity. The question is how these marginal earn something about their productivity. The question is how these marginal
migrants would differ from average migrants under larger-scale migration. Future igrants would differ from average migrants under larger-scale migration. Future
migrants could be more positively or more negatively selected than today’s migrants, igrants could be more positively or more negatively selected than today’s migrants,
with regard to observable or unobservable determinants of productivity. Selection ith regard to observable or unobservable determinants of productivity. Selection
on observable traits is easier to measure; selection on unobservables is harder. n observable traits is easier to measure; selection on unobservables is harder.
This question is closely linked to the question of whether international differ- his question is closely linked to the question of whether international differences
in productivity are explained by differences in people or differences in places. nces in productivity are explained by differences in people or differences in places.
If an emigrant to a rich country is more productive than an observably identical f an emigrant to a rich country is more productive than an observably identical
worker who stayed in a poor country, there are two broad explanations. One is that orker who stayed in a poor country, there are two broad explanations. One is that
the emigrant is different in unobservable ways from the non-emigrant; this argument he emigrant is different in unobservable ways from the non-emigrant; this argument
is compatible with a high degree of positive selection in migration and little infl s compatible with a high degree of positive selection in migration and little infl uence
of location itself on productivity. The other is that the emigrant is not that different f location itself on productivity. The other is that the emigrant is not that different
in unobservable ways from the non-emigrant; this argument is compatible with less n unobservable ways from the non-emigrant; this argument is compatible with less
positive selection in emigration, and a large effect of location on productivity. ositive selection in emigration, and a large effect of location on productivity.
Empirical work is beginning to attack this problem. One approach is to use mpirical work is beginning to attack this problem. One approach is to use
more sophisticated structural models to account for selection in macroeconomic ore sophisticated structural models to account for selection in macroeconomic
estimations (Ortega and Peri, 2009; Grogger and Hanson, 2011). A complemen- stimations (Ortega and Peri, 2009; Grogger and Hanson, 2011). A complementary
approach is to fi ary approach is to fi nd natural experiments that identify the extent of selection nd natural experiments that identify the extent of selection
on unobservables in microeconomic settings (summarized by McKenzie and Yang, n unobservables in microeconomic settings (summarized by McKenzie and Yang,
2010). Examples of the latter include McKenzie, Gibson, and Stillman (2010), who 010). Examples of the latter include McKenzie, Gibson, and Stillman (2010), who
use a naturally randomized visa lottery to show that the gains from emigrating from se a naturally randomized visa lottery to show that the gains from emigrating from
Tonga to New Zealand are only somewhat lower than the simple wage difference onga to New Zealand are only somewhat lower than the simple wage difference
for observably identical workers inside and outside Tonga—in other words, there is or observably identical workers inside and outside Tonga—in other words, there is
little unobserved self-selection in those who emigrated. In Clemens (2010), I also ittle unobserved self-selection in those who emigrated. In Clemens (2010), I also
use a naturally randomized visa lottery to show that large gains to overseas work se a naturally randomized visa lottery to show that large gains to overseas work
experienced by Indian software workers cannot be primarily the result of unob- xperienced by Indian software workers cannot be primarily the result of unobserved
positive self-selection in those workers. erved positive self-selection in those workers.
While this literature is actively evolving, in no case has one of these recent and hile this literature is actively evolving, in no case has one of these recent and
rigorous studies identifi igorous studies identifi ed a country pair for which large differences in earnings ed a country pair for which large differences in earnings
across the border can be mostly accounted for by self-selection of workers (migrant cross the border can be mostly accounted for by self-selection of workers (migrant
or otherwise) who cross the border. Numerical simulations using U.S. census micro- r otherwise) who cross the border. Numerical simulations using U.S. census microdata
on immigrants (Hendricks, 2002) and combined U.S. and foreign microdata ata on immigrants (Hendricks, 2002) and combined U.S. and foreign microdata
(Clemens, Montenegro, and Pritchett, 2008) have shown that plausible degrees of Clemens, Montenegro, and Pritchett, 2008) have shown that plausible degrees of
Michael A. Clemens 97
positive selection are insuffi ositive selection are insuffi cient to explain more than half the earnings gap between cient to explain more than half the earnings gap between
workers in the United States and observably identical potential migrants abroad. orkers in the United States and observably identical potential migrants abroad.
Sophisticated survey data do not fi ophisticated survey data do not fi nd positive emigrant selection on unobservable nd positive emigrant selection on unobservable
determinants of earnings to much exceed 50 percent in any case that has been eterminants of earnings to much exceed 50 percent in any case that has been
studied ( Jasso and Rosenzweig, 2009; Kaestner and Malamud, 2010; Fernández- tudied ( Jasso and Rosenzweig, 2009; Kaestner and Malamud, 2010; FernándezHuertas,
2011; Ambrosini, Mayr, Peri, and Radu, 2011). uertas, 2011; Ambrosini, Mayr, Peri, and Radu, 2011).
These initial results accord well with an entirely separate macroeconomic liter- hese initial results accord well with an entirely separate macroeconomic literature
(for example, Hall and Jones, 1999) which fi ture (for example, Hall and Jones, 1999) which fi nds that most of the productivity nds that most of the productivity
gap between rich and poor countries is accounted for by place-specifi ap between rich and poor countries is accounted for by place-specifi c total factor c total factor
productivity, not by productivity differences inherent to workers. Large differences roductivity, not by productivity differences inherent to workers. Large differences
in location-specifi n location-specifi c total factor productivity mean that free movement of goods and c total factor productivity mean that free movement of goods and
capital cannot by themselves achieve the global equalization of wages, as they can apital cannot by themselves achieve the global equalization of wages, as they can
in the most abstract trade models (O’Rourke and Sinott, 2004; Freeman, 2006, n the most abstract trade models (O’Rourke and Sinott, 2004; Freeman, 2006,
Kremer, 2006). remer, 2006).
In other words, the existing evidence, preliminary and spotty though it is, n other words, the existing evidence, preliminary and spotty though it is,
gives no reason to believe that a better accounting for unobserved differences in ives no reason to believe that a better accounting for unobserved differences in
the determinants of productivity between migrants and nonmigrants would greatly he determinants of productivity between migrants and nonmigrants would greatly
alter the preceding estimates of effi lter the preceding estimates of effi ciency gains from greater labor mobility. ciency gains from greater labor mobility.
8
Question 4: What Future Level of Emigration Is Feasible? uestion 4: What Future Level of Emigration Is Feasible?
The extent of feasible emigration depends on the willingness of politicians he extent of feasible emigration depends on the willingness of politicians
at potential destinations, acting as agents for their electorates, to allow immigra- t potential destinations, acting as agents for their electorates, to allow immigration.
About 59 percent of people in the United Kingdom tell pollsters that there ion. About 59 percent of people in the United Kingdom tell pollsters that there
are currently “too many” immigrants; in Italy it is 53 percent, in the United States re currently “too many” immigrants; in Italy it is 53 percent, in the United States
37 percent, in France 33 percent, in Germany and the Netherlands 27 percent, and 7 percent, in France 33 percent, in Germany and the Netherlands 27 percent, and
in Canada 17 percent (German Marshall Fund, 2010). n Canada 17 percent (German Marshall Fund, 2010).
The global gains in aggregate economic welfare in Figure 1 mask the conse- he global gains in aggregate economic welfare in Figure 1 mask the consequent
redistribution between labor and other factors, and between labor at different uent redistribution between labor and other factors, and between labor at different
levels of skill. If the median voter at the destination holds relatively little capital or evels of skill. If the median voter at the destination holds relatively little capital or
skill, this could limit the willingness of citizens and politicians in that country to kill, this could limit the willingness of citizens and politicians in that country to
reduce impediments to emigration from poor countries. Noneconomic attitudes educe impediments to emigration from poor countries. Noneconomic attitudes
such as nationalism can also play an important role. Mayda (2006) fi uch as nationalism can also play an important role. Mayda (2006) fi nds that it is nds that it is
the wealthier, better-educated, and less-nationalist individuals in rich destination he wealthier, better-educated, and less-nationalist individuals in rich destination
countries who have more favorable attitudes toward immigration. ountries who have more favorable attitudes toward immigration.
8 In fact, the emerging evidence on selection suggests that some of the estimates of gains from emigration
are small because they are too conservative. Walmsley and Winters (2005), for example, assume in their
base scenarios that migrants from low-productivity countries to high-productivity countries acquire only
half of the difference in productivity between the two countries. That is, they assume that the difference
in productivity between observed non-emigrants at the destination and observed non-emigrants at the
origin overstates by 100 percent the true productivity effect of emigration on new emigrants. None of
the existing estimates suggests that the magnitude of selection—in the limited cases that have been
studied—is close to that large.
98 Journal of Economic Perspectives
A substantial expansion of emigration may thus seem politically impractical, substantial expansion of emigration may thus seem politically impractical,
and within a time frame of a few years, this objection holds some force. But the nd within a time frame of a few years, this objection holds some force. But the
global effi lobal effi ciency gains from even small relaxations of existing barriers to emigration ciency gains from even small relaxations of existing barriers to emigration
are large relative to the gains from further relaxation of barriers to trade and capital re large relative to the gains from further relaxation of barriers to trade and capital
fl ows (as shown earlier in Table 2). ows (as shown earlier in Table 2).
Furthermore, economists should be open to the possibility that dramatic urthermore, economists should be open to the possibility that dramatic
changes in what is practical can happen over several decades. After all, changes hanges in what is practical can happen over several decades. After all, changes
in geographic labor mobility that were unthinkable only a few decades ago have n geographic labor mobility that were unthinkable only a few decades ago have
come to pass. Through the 1980s, a Polish national attempting to emigrate to West ome to pass. Through the 1980s, a Polish national attempting to emigrate to West
Germany could be shot by soldiers sealing the Inner German border from the east. ermany could be shot by soldiers sealing the Inner German border from the east.
Today, Polish jobseekers may move freely throughout Germany. The world has oday, Polish jobseekers may move freely throughout Germany. The world has
summarily discarded vast systems of restrictions on the labor mobility of medieval ummarily discarded vast systems of restrictions on the labor mobility of medieval
serfs, slaves, women, South African blacks, indigenous Australians, and a long list erfs, slaves, women, South African blacks, indigenous Australians, and a long list
of others. f others.
Even modest expansions of emigration have provided great benefi ven modest expansions of emigration have provided great benefi ts for large ts for large
numbers of people. In Clemens and Pritchett (2008), my coauthor and I show umbers of people. In Clemens and Pritchett (2008), my coauthor and I show
that among the people born in Haiti, Mexico, or India who live above a (uniform, hat among the people born in Haiti, Mexico, or India who live above a (uniform,
purchasing power parity–adjusted) international poverty line, large fractions live in urchasing power parity–adjusted) international poverty line, large fractions live in
the United States. For example, among Haitians who live either in the United States he United States. For example, among Haitians who live either in the United States
or Haiti and live on more than $10/day measured at U.S. prices—about a third of r Haiti and live on more than $10/day measured at U.S. prices—about a third of
the U.S. “poverty” line—four out of fi he U.S. “poverty” line—four out of fi ve live in the United States. ve live in the United States.
9
 Emigration from Emigration from
Haiti, as a force for Haitians’ poverty reduction, may be at least as important as any aiti, as a force for Haitians’ poverty reduction, may be at least as important as any
economic change that has occurred within Haiti. conomic change that has occurred within Haiti.
Economic policy might help ease political constraints. Several economists have conomic policy might help ease political constraints. Several economists have
proposed policy mechanisms to compensate nonmigrants at the destination for roposed policy mechanisms to compensate nonmigrants at the destination for
declines in welfare, by charging immigrants a tax—the fi eclines in welfare, by charging immigrants a tax—the fi gure $50,000 often comes gure $50,000 often comes
up—or auctioning work permits (including Becker and Becker, 1997; Freeman, p—or auctioning work permits (including Becker and Becker, 1997; Freeman,
2006; Orrenius and Zavodny, 2010; Fernández-Huertas and Rapoport, 2010). All of 006; Orrenius and Zavodny, 2010; Fernández-Huertas and Rapoport, 2010). All of
these, in one way or another, seek to minimize the number of “losers” from labor hese, in one way or another, seek to minimize the number of “losers” from labor
mobility by moving from the Kaldor–Hicks concept of optimality—which is that obility by moving from the Kaldor–Hicks concept of optimality—which is that
social gains are suffi ocial gains are suffi cient so that it would be potentially possible to compensate cient so that it would be potentially possible to compensate
losers—toward Pareto optimality, in which parties that would otherwise be losers osers—toward Pareto optimality, in which parties that would otherwise be losers
receive actual compensation. Mechanisms like this might alleviate some of the eceive actual compensation. Mechanisms like this might alleviate some of the
political constraints to greater labor mobility. olitical constraints to greater labor mobility.
9 That fi gure probably overestimates the effect of migration on Haitians’ poverty, since there is evidence
of some positive selection of Haitian emigrants on observable and unobservable determinants of earnings
(Clemens, Montenegro, and Pritchett, 2008). But even strong assumptions about highly positive
selection would still leave emigration as the cause of a large fraction of the poverty reduction that has
occurred for people born in Haiti. In addition, this estimate tends to understate the effect of emigration
on poverty reduction—because it does not count Haitians who emigrated to countries other than
the United States; because part of the poverty reduction for non-emigrants could have been caused by
remittances, in turn caused by emigration; and because the estimate was made before a 2010 earthquake
crushed Haiti’s economy.
Economics and Emigration: Trillion-Dollar Bills on the Sidewalk? 99
A Research Agenda Whose Time Has Gone Research Agenda Whose Time Has Gone
The four questions above outline a research agenda on emigration with which he four questions above outline a research agenda on emigration with which
I will conclude this paper. But fi will conclude this paper. But fi rst I want to contrast that new agenda with the old rst I want to contrast that new agenda with the old
agenda in the literature we have now, and speculate about why that old agenda has genda in the literature we have now, and speculate about why that old agenda has
dominated so far. ominated so far.
Start talking about the welfare benefi tart talking about the welfare benefi ts and costs of emigration at any research ts and costs of emigration at any research
or policy meeting on migration and economic development, and the conversation r policy meeting on migration and economic development, and the conversation
turns quickly to two topics: “brain drain” and workers’ remittances. These have urns quickly to two topics: “brain drain” and workers’ remittances. These have
been the primary focus of research on the relationship between emigration and een the primary focus of research on the relationship between emigration and
global economic development, and not coincidentally, they are the focus of the lobal economic development, and not coincidentally, they are the focus of the
other two papers in this symposium. Both issues are important. Yet neither of ther two papers in this symposium. Both issues are important. Yet neither of
these approaches shows much promise as a way of better understanding the global hese approaches shows much promise as a way of better understanding the global
effi ciency gains from greater emigration. I have already discussed the reasons why ciency gains from greater emigration. I have already discussed the reasons why
the literature on migration and human capital externalities has not yielded good he literature on migration and human capital externalities has not yielded good
reasons to alter existing estimates of the gains to greater emigration. I turn now to easons to alter existing estimates of the gains to greater emigration. I turn now to
the literature on remittances. he literature on remittances.
Global fl lobal fl ows of remittances are rising toward $400 billion per year (Mohapatra, ows of remittances are rising toward $400 billion per year (Mohapatra,
Ratha, and Silwal, 2011). This trend has helped to launch a large and valuable atha, and Silwal, 2011). This trend has helped to launch a large and valuable
research literature, presented and discussed by Dean Yang in this issue. But remit- esearch literature, presented and discussed by Dean Yang in this issue. But remittances
are typically a small fraction of emigrants’ foreign wage, especially for ances are typically a small fraction of emigrants’ foreign wage, especially for
permanent emigrants (van der Mensbrugghe and Roland-Holst, 2009). To a fi ermanent emigrants (van der Mensbrugghe and Roland-Holst, 2009). To a fi rst
approximation, remittances are intrahousehold transfers that cross borders, and the pproximation, remittances are intrahousehold transfers that cross borders, and the
reasons that people send remittances (Rapoport and Docquier, 2006) are broadly easons that people send remittances (Rapoport and Docquier, 2006) are broadly
the same as the reasons people make other intrahousehold transfers (Laferrère he same as the reasons people make other intrahousehold transfers (Laferrère
and Wolff, 2006). If a Mexican woman experiences an income gain from working nd Wolff, 2006). If a Mexican woman experiences an income gain from working
in Mexico, the whole value of that gain adds to her household’s welfare—both the n Mexico, the whole value of that gain adds to her household’s welfare—both the
portion she consumes and the portion she shares with her husband. This social ortion she consumes and the portion she shares with her husband. This social
welfare calculation is unaffected if she experiences an income gain by stepping over elfare calculation is unaffected if she experiences an income gain by stepping over
the Mexican border into Texas. he Mexican border into Texas.
In short, barriers to emigration have a fi n short, barriers to emigration have a fi rst-order effect on welfare; any rst-order effect on welfare; any
barriers to fl arriers to fl ows of remittances have only a second- or third-order effect on ows of remittances have only a second- or third-order effect on
welfare. But the literature has gone into great detail about the smaller effects, elfare. But the literature has gone into great detail about the smaller effects,
frequently investigating whether the welfare effects of remittances are attenuated requently investigating whether the welfare effects of remittances are attenuated
by any withdrawal of remittance recipients from the labor force (for example, y any withdrawal of remittance recipients from the labor force (for example,
Cox-Edwards and Rodríguez-Oreggia, 2008) or by any consequent appreciation ox-Edwards and Rodríguez-Oreggia, 2008) or by any consequent appreciation
of the sending-country exchange rate (for example, Amuedo-Dorantes and Pozo, f the sending-country exchange rate (for example, Amuedo-Dorantes and Pozo,
2004; Vargas-Silva, 2009). 004; Vargas-Silva, 2009).
Why has the literature focused so much more on the relatively small and uncer- hy has the literature focused so much more on the relatively small and uncertain
effects of remittances and “brain drain” than on the relatively massive and likely ain effects of remittances and “brain drain” than on the relatively massive and likely
global effects of migration—including the benefi lobal effects of migration—including the benefi ts for the migrants themselves? ts for the migrants themselves?
Perhaps many economists consider the estimates of effi erhaps many economists consider the estimates of effi ciency gains in Table 1 to ciency gains in Table 1 to
be self-evident and thus not worthy of much study. But the review above suggests e self-evident and thus not worthy of much study. But the review above suggests
we have much to learn about those estimates. I suspect the reason lies elsewhere. e have much to learn about those estimates. I suspect the reason lies elsewhere.
100 Journal of Economic Perspectives
Perhaps the literature focuses on remittances and “brain drain” because those erhaps the literature focuses on remittances and “brain drain” because those
effects more obviously pertain to national welfare than individual welfare. ffects more obviously pertain to national welfare than individual welfare.
Focusing on national welfare is a grand old tradition in economics—older, ocusing on national welfare is a grand old tradition in economics—older,
indeed, than the work of Adam Smith, who prominently inserted “the wealth ndeed, than the work of Adam Smith, who prominently inserted “the wealth
of nations” into the title of his great book. Economists in the long-ago mercan- f nations” into the title of his great book. Economists in the long-ago mercantilist
tradition largely agreed that the goal of economic policy was to encourage ilist tradition largely agreed that the goal of economic policy was to encourage
national production and exports, to discourage imports, and thus to bring specie ational production and exports, to discourage imports, and thus to bring specie
into the country. To this end, they advised encouraging manufactured exports and nto the country. To this end, they advised encouraging manufactured exports and
discouraging raw material exports, to generate domestic employment (Heckscher, iscouraging raw material exports, to generate domestic employment (Heckscher,
1935[1955]; Irwin, 1996, p. 34). The classic work of Furniss (1920, p. 8) draws 935[1955]; Irwin, 1996, p. 34). The classic work of Furniss (1920, p. 8) draws
out the implications of mercantilist assumptions for labor: Raising export revenue ut the implications of mercantilist assumptions for labor: Raising export revenue
requires mass production at low cost. High manufacturing employment at low equires mass production at low cost. High manufacturing employment at low
wages requires a moral duty for low-income people to work in manufacturing ages requires a moral duty for low-income people to work in manufacturing
and a moral duty for high-income people to preferentially consume domestic nd a moral duty for high-income people to preferentially consume domestic
manufactures. In this view, the emigration of labor only affects national welfare to anufactures. In this view, the emigration of labor only affects national welfare to
the extent that it encourages or discourages home production and exports. The he extent that it encourages or discourages home production and exports. The
mercantilist writer Josiah Child (1668 [1751], pp. 146–7), for example, argued that ercantilist writer Josiah Child (1668 [1751], pp. 146–7), for example, argued that
any emigration to colonial plantations “is certainly a damage, except the employ- ny emigration to colonial plantations “is certainly a damage, except the employment
of those people abroad, do cause the employment of so many more at home ent of those people abroad, do cause the employment of so many more at home
in their mother kingdoms, and that can never be, except the trade be restrained to n their mother kingdoms, and that can never be, except the trade be restrained to
their mother kingdom . . .” heir mother kingdom . . .”
Economic research carried out under this set of working assumptions would conomic research carried out under this set of working assumptions would
focus on estimating the extent to which emigration tends to raise origin-country ocus on estimating the extent to which emigration tends to raise origin-country
production (such as through remittances, minus the consequent labor force roduction (such as through remittances, minus the consequent labor force
withdrawal) or tends to lower origin-country production (such as through “brain ithdrawal) or tends to lower origin-country production (such as through “brain
drain”). Little attention would be reserved for the gains to migrants. This, in very rain”). Little attention would be reserved for the gains to migrants. This, in very
broad strokes, describes the bulk of extant literature on the economic effects of road strokes, describes the bulk of extant literature on the economic effects of
emigration. I am far from the fi migration. I am far from the fi rst to claim a link between the mercantilist tradition rst to claim a link between the mercantilist tradition
and economics’ preoccupation with suspected negative, within-country effects from nd economics’ preoccupation with suspected negative, within-country effects from
emigration. migration.
10 After Furniss (1920, p. 54), this link has been explicitly made by Nobel After Furniss (1920, p. 54), this link has been explicitly made by Nobel
laureate Ted Schultz (1978); Charles Kindleberger (1986) when he was president of aureate Ted Schultz (1978); Charles Kindleberger (1986) when he was president of
the American Economic Association; and one of the great scholars of migration and he American Economic Association; and one of the great scholars of migration and
economic growth, Brinley Thomas (1973, pp. 1–6); among many others. conomic growth, Brinley Thomas (1973, pp. 1–6); among many others.
11
10 Incidentally, the leading economic research on immigration—not the focus of this essay—looks quite
different. Under mercantilist assumptions, any wage-depressing effects of immigration at the destination
would raise that country’s welfare by increasing the competitiveness of its manufactured exports. But the
centerpiece of the immigration literature is the anti-mercantilist suspicion of immigration’s deleterious
effects on labor. I have no explanation for the difference other than to point out that the modern
immigration literature was founded by microeconomic labor economists such as George Borjas and
David Card. The modern literature on economic growth and development overseas, in contrast, was
founded by researchers with a primarily macroeconomic focus such as Alexander Gerschenkron, Paul
Rosenstein-Rodin, and Robert E. Lucas.
11 Others include Hovde (1934), Hamilton (1940), Shepperson (1953), Middendorf (1960), O’Brien
(1966), Letiche (1969), Jeremy (1977), Dowty (1986), Backhaus and Wagner (1987), Davis (1988),
de Soto (1989, p. 201), and Mokyr and Nye (2007).
Michael A. Clemens 101
A New Research Agenda On Emigration New Research Agenda On Emigration
The available evidence suggests that the gains to lowering barriers to emigra- he available evidence suggests that the gains to lowering barriers to emigration
appear much larger than gains from further reductions in barriers to goods ion appear much larger than gains from further reductions in barriers to goods
trade or capital fl rade or capital fl ows—and may be much larger than those available through any ows—and may be much larger than those available through any
other shift in a single class of global economic policy. Indeed, “some big bills have ther shift in a single class of global economic policy. Indeed, “some big bills have
not been picked up on the routes that lead from poor to rich countries” (Olson, ot been picked up on the routes that lead from poor to rich countries” (Olson,
1996). Research economists, however, write relatively little about emigration. The 996). Research economists, however, write relatively little about emigration. The
term “international trade” is 13 times more frequent than “international migration” erm “international trade” is 13 times more frequent than “international migration”
in all the published article abstracts contained in the Research Papers in Economics n all the published article abstracts contained in the Research Papers in Economics
(RePEc) archive. Furthermore, economists focus on arrival, not departure: in RePEc) archive. Furthermore, economists focus on arrival, not departure: in
RePEc, “immigration” is four times as frequent as “emigration.” ePEc, “immigration” is four times as frequent as “emigration.”
It should be a priority of economic research to seek a better characterization t should be a priority of economic research to seek a better characterization
of the gains to global labor mobility and to investigate policy instruments to realize f the gains to global labor mobility and to investigate policy instruments to realize
a portion of those gains. The four questions in this paper suggest one structure portion of those gains. The four questions in this paper suggest one structure
for that agenda. We clearly need a better theoretical and empirical understanding or that agenda. We clearly need a better theoretical and empirical understanding
of human capital externalities; the dynamics of labor demand under large-scale f human capital externalities; the dynamics of labor demand under large-scale
migration fl igration fl ows; the magnitude and mechanisms of the effect of workers’ loca- ows; the magnitude and mechanisms of the effect of workers’ location
on their productivity, relative to the effect of workers’ inherent traits on their ion on their productivity, relative to the effect of workers’ inherent traits on their
productivity; and the policy instruments that might make greater labor mobility roductivity; and the policy instruments that might make greater labor mobility
possible. Many of the outstanding questions are discussed by Pritchett (2006), ossible. Many of the outstanding questions are discussed by Pritchett (2006),
Rosenzweig (2006), Hanson (2009), and Docquier and Rapoport (forthcoming). osenzweig (2006), Hanson (2009), and Docquier and Rapoport (forthcoming).
Complicating the empirical portion of this agenda is the fact that even basic omplicating the empirical portion of this agenda is the fact that even basic
statistics on international migration are often unavailable to economists (Commis- tatistics on international migration are often unavailable to economists (Commission
on International Migration Data for Development Research and Policy, 2009). ion on International Migration Data for Development Research and Policy, 2009).
Detailed statistics are either held confi etailed statistics are either held confi dential by governments or not collected at all, dential by governments or not collected at all,
and publicly-released data can be a mess of incomparable time periods, modes of nd publicly-released data can be a mess of incomparable time periods, modes of
migration (temporary vs. permanent, entries vs. individuals, and other categoriza- igration (temporary vs. permanent, entries vs. individuals, and other categorizations),
and defi ions), and defi nitions of occupations. Just estimating bilateral stocks of migrants at nitions of occupations. Just estimating bilateral stocks of migrants at
a single point in time, even without any other information about those migrants, is single point in time, even without any other information about those migrants, is
a costly enterprise requiring heroic assumptions and massive imputation (Parsons, costly enterprise requiring heroic assumptions and massive imputation (Parsons,
Skeldon, Walmsley, and Winters, 2007). Publicly available international migration keldon, Walmsley, and Winters, 2007). Publicly available international migration
statistics have roughly the quality of international trade statistics in the 1960s. As tatistics have roughly the quality of international trade statistics in the 1960s. As
occurred in international trade, investing in migration data collection and compila- ccurred in international trade, investing in migration data collection and compilation
must be part of the research agenda. ion must be part of the research agenda.
If this additional research tends to confi f this additional research tends to confi rm that barriers to emigration place rm that barriers to emigration place
one of the fattest of all wedges between humankind’s current welfare and its ne of the fattest of all wedges between humankind’s current welfare and its
potential welfare—no doubt with a number of useful caveats—then understanding otential welfare—no doubt with a number of useful caveats—then understanding
and realizing the gains from emigration deserve much more research priority. nd realizing the gains from emigration deserve much more research priority.
Emigration’s literature remains scattered; emigration’s migration’s literature remains scattered; emigration’s Wealth of Nations unwritten; unwritten;
emigration’s Ricardo undiscovered.