Wage Effects of High-Skilled Migration

Transcription

1 Public Disclosure Authorized Policy Research Working Paper 6317 WP6317 Public Disclosure Authorized Public Disclosure Authorized Wage Effects of igh-killed Migration International Evidence Volker Grossann David tadelann Public Disclosure Authorized The World Bank Developent Econoics Vice Presidency Partnerships, Capacity Building nit January 2013

2 Policy Research Working Paper 6317 Abstract The international igration of high-skilled workers ay trigger productivity effects at the acro level such that the wage rate of skilled workers increases in host countries and decrease in source countries. The authors exploit data on international bilateral igration flows and provide evidence consistent with this theoretical hypothesis. They propose various instruentation strategies to identify the causal effect of skilled igration on log differences of GDP per capita, total factor productivity, and the wages of skilled workers between pairs of source and destination countries. These strategies ai to address the endogeneity proble that arises when international wage differences affect igration decisions. This paper is a product of the Partnerships, Capacity Building nit, Developent Econoics Vice Presidency. It is part of a larger effort by the World Bank to provide open access to its research and ake a contribution to developent policy discussions around the world. Policy Research Working Papers are also posted on the Web at The authors ay be contacted at volker.grossann@unifr.ch and david.stadelann@unifr.ch. The Policy Research Working Paper eries disseinates the findings of work in progress to encourage the exchange of ideas about developent issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the naes of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Developent/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governents they represent. Produced by the Research upport Tea

3 Wage Effects of igh-killed Migration: International Evidence Volker Grossann and David tadelann JEL classification codes: F22; O30. Keywords: International high-skilled igration; Wage effects; Total factor productivity. Volker Grossan: niversity of Fribourg; CEifo, Munich; Institute for the tudy of Labor (IZA, Bonn. Postal address: niversity of Fribourg, Bd. de Pérolles 90, G424, 1700 Fribourg, witzerland. Tel.: +41 ( Eail: David tadelann (corresponding author: niversity of Fribourg; CREMA - Center for Research in Econoics, Manageent and the Arts, witzerland. Postal address: niversity of Fribourg, Bd. de Pérolles 90, F408, 1700 Fribourg, witzerland, +41 ( , david.stadelann@unifr.ch Acknowledgeents: We are grateful to three anonyous referees for valuable coents and suggestions. Moreover, we thank Michel Beine, Bruno. Frey, Mark Gradstein, illel Rapoport, Avi ihon, and John Wilson for coents on an earlier draft. We also benefited fro discussion with seinar participants at the Ben-Gurion niversity, niversity of Zurich, niversity of Geneva, niversity of iegen, the Annual Meeting of the European Econoic Association in Milan, and the conference Globalization and the Brain Drain: Theory, Evidence and Policy in Jerusale and Raat Gan. 1

4 The recent surge in the international igration of high-skilled workers not only raised standard concerns about adverse brain-drain effects for developing countries but also led to worries about native high-skilled workers in advanced destination countries. 1 Doestic workers with higher education levels fear that their wages will decline in response to increased copetition fro siilarly qualified igrants. Whereas debates on igration in the past have centered on asylu rights and low-skilled igrants, over the years, politicians and the ass edia have discovered the issue of high-skilled iigration. For instance, in witzerland and Austria, the discussion has recently becoe eotionally charged owing to significant inflows of tertiary-educated workers, particularly fro Gerany. 2 For the nited tates, anson, cheve and laughter (2009 find that skilled natives tend to oppose iigration in states with a relatively skilled ix of iigrants ore than in states in which the skill coposition of iigrants features a high proportion of low-skilled iigrants. iilarly, a recent panel study by Müller and Tai (2010 for Europe suggests that higher-skilled workers have less favorable attitudes toward iigration when iigrants are ore skilled relative to the average skill level in the destination country. This paper exaines whether doestic skilled workers have reason to oppose high-skilled iigration and, vice versa, whether nonigrating high-skilled workers win or lose fro brain drain in source countries. We argue that the international igration of high-skilled workers triggers productivity effects at the acro level such that the wage rate of skilled workers ay rise in host countries and decline in source countries. By exploiting data on international bilateral igration flows fro Docquier, Marfouk and Lowell (2007, we epirically exaine the ipact of an increase in high-skilled 2

5 eigration rates on log differences in GDP per capita, total factor productivity (TFP, and the wage incoe of skilled workers between pairs of source and destination countries. We propose a range of instruental variables to address the potential reverse causality proble that arises when international wage differences affect individual igration decisions (e.g., Lucas, 2005; Egger and Radulescu, 2009; Grogger and anson, Our theoretical odel suggests that even when considering adjustents in educational decisions, an increase in high-skilled eigration (iigration lowers (raises the doestic skill intensity in production. 3 This relationship has two effects on the relative wages of high-skilled workers between destination and source econoies. First, for a given TFP and as a consequence of the declining arginal productivity of a certain type of labor, high-skilled workers lose in the destination econoy and win in the source econoy. owever, external effects of igration on TFP (positive in destination, adverse in source ay reverse this result. The net effect of high-skilled igration on international wage differences is thus theoretically abiguous. This theoretical approach akes the relationship between high-skilled igration and wages an epirical question. Our analysis suggests that, if anything, the external productivity effect is likely to doinate. Moreover, because of copleentarity between high-skilled and low-skilled labor, an increase in low-skilled igration unabiguously benefits high-skilled workers in the receiving country. Our findings are consistent with the recent literature on the wage effects of highskilled iigration in single countries. Borjas (2003 and Dustann, Fabbri and Preston (2005 provide evidence for a sall but positive ipact of an inflow of iigrants with a college degree on wages for college-educated natives in the nited tates and nited 3

6 Kingdo, respectively. iilarly, Friedberg (2001 suggests that native wages ay rise after iigrants enter high-skilled occupations in the Israeli labor arket. Our epirical contribution is to provide international evidence for the theoretical possibility of positive wage effects in destination countries relative to source countries. We exploit data on bilateral igration between country pairs, thereby copleenting single-country studies on labor arket effects of iigration. Another strand of literature has ephasized the positive effects of brain drain on arket incoe in the source econoy (e.g., Mountford, 1997; tark, elenstein and Prskawetz, 1997; Beine, Docquier and Rapoport, 2001, This possibility arises fro the idea that an increase in iigration quotas in advanced countries iproves iigration prospects for skilled workers in developing countries and thereby raises incentives to acquire education. owever, epirically, the net effect on the size of the skilled labor force appears to be positive, except for very poor countries and/or countries with low levels of huan capital (Beine et al., 2001, In our theoretical fraework, brain drain reduces the skill intensity in the source country, even when educational decisions are adjusted. Because our epirical fraework investigates the effect of skilled igration on relative outcoes between destination and source, we do not test the alternative hypothesis advanced in the brain gain literature. We can conclude, however, that the destination country tends to gain ore fro skilled igration than the source country. The reainder of this paper is organized as follows. ection I presents a siple theoretical odel. The odel provides the basis for the epirical analysis in section II on the effects of higher eigration on relative GDP per capita, relative TFP, and the relative 4

7 wage incoe of skilled workers between the source and the destination. The last section provides concluding rearks. I. TEORETICAL CONIDERATION Our theoretical analysis shows that the presence of the external productivity effects of skilled labor iplies that in response to an increase in high-skilled igration, the wage level of educated workers ay increase in the host country relative to the source country. et p Consider two econoies, hoe and foreign. There is a hoogenous consuption good, which is chosen as the nueraire. Output Y is produced under perfect copetition according to the technology (1 Y = AF(, L ALf ( k, where and L denote the high-skilled and low-skilled labor inputs, respectively, A is TFP, the function F is linearly hoogenous, k / L denotes the skill intensity of production, and f ( k F( k,1. Furtherore, f is increasing, strictly concave, and fulfills the standard boundary conditions. Before igration, there is (for siplicity the sae nuber N of individuals/workers in both countries. There is a positive external effect of a higher concentration of skilled labor, h / N, on TFP, (2 A = a(h, 5

8 where a is an increasing function. This assuption captures huan capital externalities as foralized, for instance, by Lucas (1988 in the context of endogenous growth. These huan capital externalities ay arise fro learning spillover effects across workers, increased innovation activity in firs, and better institutional quality in a country, which ay be associated with a ore highly skilled doestic population. The epirical literature on huan capital externalities is soewhat inconclusive but is ostly supportive. For instance, Aceoglu and Angrist (2000 find odest evidence in favor of huan capital externalities fro secondary schooling, whereas Ciccone and Peri (2006 find no evidence. Iranzo and Peri (2009 argue in favor of strong huan capital externalities fro college graduates in the nited tates but not fro an increased share of high school graduates. In a recent study, Gennaioli et al. (2011 find strong epirical evidence of huan capital externalities. They eploy a new data set with 1569 subnational regions fro 110 countries and argue that huan capital is the priary driver of regional developent. Moreover, they copleent their finding with fir-level evidence on regional education levels for productivity and find large effects. The authors conclude that the previous epirical literature has underestiated the agnitude of huan capital externalities. iilarly, unt (2011 eploys a.. state panel data set for the period fro 1940 to 2000 to show that an increase in the share of the iigrant college graduate population of one percentage point increases the nuber of patents per capita by 9 18 percent. This is strong evidence in favor of the hypothesis that skilled iigration increases TFP. Each individual decides whether to becoe skilled and whether to igrate. Both skilled and unskilled individuals are internationally obile, but they ay differ in 6

9 igration costs. Forally, let c i denote the consuption level of individual i. tility level u i is given by (3 u i ci if i stays at hoe, = ci / θi if i igrates, where θ = θ > 1 if i is skilled and θ = θ L > 1 if i is unskilled. The odeling of i i igration costs as discounted consuption follows tark et al. (1997, aong others. Education coes at tie cost e 0, which ay be interpreted to be a learning cost. i Whereas an unskilled individual supplies one unit of tie to a perfect labor arket, a skilled individual i supplies only 1 ei units of tie. The wage rate per unit of tie of high-skilled and low-skilled individuals at hoe is denoted by w and w L, respectively. Moreover, denote all foreign variables and functions with superscript *. Therefore, the consuption of individual i born at hoe is given by (4 c i (1 ei w if i is skilled and stays at hoe, wl if i is unskilled and stays at hoe, = (1 ei w if i is skilled and eigrates, wl if i is unskilled and eigrates. Denote by G (e the cuulative distribution function of the learning cost e in the population at hoe. For convenience, suppose that G is continuously differentiable. We allow functions G, F, and functions G, F, and a, respectively. a (characterizing the foreign country to be different to As will becoe apparent, the equilibriu outcoe is the sae regardless of whether we assue that igration possibilities are already considered in the education 7

10 decisions of individuals. This condition is an iplication of the siplifying assuptions that (i learning abilities and igration costs are uncorrelated and (ii individual igration costs are the sae for all workers within a skill group. Derivation of Testable ypotheses We will now derive the testable hypotheses. For this purpose, we treat igration as exogenous. According to equations (1 and (2, copetitive factor prices are as follows: (5 w = a( h f ( k, (6 w L = a( h [ f ( k kf ( k ]. According to equations (3 and (4, an individual of skill type j {, L} chooses to igrate if w j j / θ w j ; therefore, in an interior equilibriu, w wl (7 = w, = wl. L θ θ A nonigrating individual i chooses education whenever ( 1 e w w. i L Moreover, staying at hoe and being educated yields higher utility than igrating and reaining unskilled if ( 1 w, which is the sae condition. iilarly, L e i w wl /θ = L we find that a igrating individual chooses education if (1 e w i /θ wl / L θ, which, in view of equation (7, again gives us condition ( 1 e w w. Moreover, igrating i L and being educated yields higher utility than not igrating and reaining unskilled if ( 1 ei w / θ = (1 ei w wl. 8

11 Therefore, all individuals with learning costs below soe endogenous threshold level, e, which depends on doestic wages only, becoe skilled: wl f ( k kf ( k (8 ei 1 = 1 e( k. w f ( k Because f < 0, we have e < 0. As the skill intensity, k, increases, the wage rate of unskilled individuals relative to skilled individuals, w / w L, increases; consequently, ore individuals reain unskilled, indicating that the threshold learning cost e is lower. The fraction of doestically born unskilled workers,, is given by ~ (9 = 1 G( e( k ( k, where ~ > 0. The effective units of skilled labor in the hoe country per native, before igration, are given by 4 e ( k ~ (10 = (1 e dg( e ( k. 0 ~ Therefore, < 0. Denote by and the fraction of skilled and unskilled labor units eigrating to the foreign country ( eigration rates, respectively. After igration, we have h : = / N = and l = L / N = :, respectively. Therefore, using equations (9 and (10, the skill intensity at hoe, k = / L, is iplicitly given by (11 ~ k = ~ ( k ( k. 9

12 ~ sing > 0 and ~ < 0, we see that the right-hand side of equation (11 is decreasing in k. Therefore, in an interior labor arket equilibriu, the skill intensity given by equation (11, denoted by k k ~ (,, is unique. The function k~ is decreasing in the eigration rate of skilled labor, rate of unskilled labor,., and increasing in the eigration In a two-country world, eigrants of one country are iigrants of the other country. Therefore, the foreign skill intensity k is uniquely given by 5 ~ ( k (12 k = ~ ( k + +. We write k ~ k (,. The function k ~ is increasing in and decreasing in. sing h = and h = +, TFP in the foreign (host country relative to the hoe (source country can be written as 6 ~ ~ A a ( ( (13 ~ ( ( ~ k (, = A a k (, + + α ~ α ( according to equation (2. Moreover, according to equation (5, the relative wage rate for skilled workers is ~ ~ w a ( ( k (14 ~ ( ( ~ (, = w a k (,, ~ + ( f ( k (, ω ~ ~ ω ( + f ( k (,,,. 10

13 11 Define the elasticities of the skill intensity at hoe and in the foreign country with respect to the igration of skilled and unskilled labor fro the hoe country to the foreign country: (15, ~ ~, ~ ~ k k k k κ κ (16. ~ ~, ~ ~ k k k k κ κ Note that the elasticities are defined such that they are positive: 0,,, > κ κ κ κ. Moreover, define by (17, ( ( ( h a h ha h ε (18, ( ( ( k f k kf k η the elasticity of TFP with respect to skilled labor per native h and the elasticity of f with respect to skill intensity k. (We define ε and η analogously. It is siple to show the following results. First, the elasticity of relative destination-to-source TFP ( A A / = α with respect to the eigration rate of the skilled ( and unskilled ( is given by ( = h l k h h l k h κ ε κ ε α α ( ~ ( ( ( ~ ( ~ ~, (20, ( ~ ( ( ( ~ ( ~ ~ = l k h l k h κ ε κ ε α α

14 respectively. Therefore, if the effect of a change in the skill intensity (triggered by igration on the education decision is sall (i.e., the agnitude of derivatives ~ ~,( < 0 are sall, the odel predicts that an increase in the igration rate of skilled labor ( has a positive effect on relative destination-to-source TFP (α. Moreover, an increase in the igration rate of unskilled labor ( has a positive but sall effect on α because the igration of unskilled labor only has an indirect TFP effect by lowering education incentives in the source country (and vice versa in the destination country. By contrast, as a result of huan capital externalities ( ε, ε > 0, the eigration of skilled labor also has a direct TFP effect on skilled labor input per native ( h in the source country (and, again, vice versa in the destination country. The effect is itigated because an increase in country (and provides disincentives in the destination country. fosters education incentives in the source econd, the elasticity of the destination-to-source relative wage incoe of skilled labor ( ω = w / w with respect to the eigration rate of skilled and unskilled labor is given by ~ ~ ω α (21 ~ = ~ η( k κ η ( k κ, ω α ~ ~ ω α (22 ~ = ~ + η( k κ + η ( k κ, ω α respectively. Therefore, the ipact of the igration of unskilled labor (increase in on the relative destination-to-source wage incoe of skilled labor is unabiguously positive. The relative TFP increases as a result of education effects, and the resulting 12

15 increase in skill intensity k reduces the wages of skilled labor in the source country (and vice versa in the destination country, where the skill intensity decreases. By contrast, for a given TFP, the wage rate of skilled labor decreases with the skill intensity; therefore, the ipact of the igration of skilled labor (increase in on the relative destinationto-source wage incoe of skilled labor ( ω is abiguous, even if the relative destination-to-source TFP (α increases. Only if the TFP effects are sufficiently large owing to huan capital externalities does an increase in increase ω. In su, we predict that an increase in the eigration rate of high-skilled labor ( increases relative TFP α = A / A, whereas the ipact of the eigration of unskilled labor ( on α ay be sall. Moreover, an increase in has a positive and possibly large effect on the relative wages of the skilled, ω = w / w. Finally, an increase in ay lead to an increase in ω if TFP effects are sufficiently large. These are iportant theoretical results for political debate in soe destination countries of skilled workers. We have focused the theoretical analysis on the predictions of the effects of igration, although we allowed individuals to consider the igration decision when choosing education. Because igration is endogenous according to the odel and depends (inter alia on international wage differences, the odel also indicates an epirical endogeneity issue, which we try to address by using instruentation strategies. 13

16 II. EMPIRICAL ANALYI Our theoretical analysis has highlighted the effect of the eigration of highskilled and low-skilled labor on TFP differences and the wage incoe gap of skilled labor to potential host econoies of expatriates. We have seen that there ay be counteracting channels by which skilled igration affects the wages of skilled workers: the external TFP effects of igration and the effect on the arginal productivity of skilled labor when TFP is held constant. The direction fro (wage incoe differences to igration flows has been exained epirically elsewhere. Two recent papers are notable. First, Grogger and anson (2011 provide convincing evidence for the critical role of wage differences between country pairs on the eigration patterns of tertiary educated workers. 7 econd, Beine et al. (2011 show that in addition to wage differences, network effects are iportant for the igration decisions of both high-skilled and low-skilled workers. The authors show that eigrants already living in the destination country positively affect igration flows. 8 Our analysis copleents research on the interaction between wage differences and skilled igration by focusing on the opposite direction, the ipact of igration on both international (wage incoe differences for skilled workers and TFP differences between country pairs. Inter alia, we instruent skilled igration with past igration stocks, as suggested by Beine et al. (2011. Data and Estiation trategy The eigration rate of highly skilled individuals is our ain explanatory variable. Docquier and Marfouk (2006 established a dataset of eigration stocks and rates by 14

17 educational attainent for the years 1990 and The authors count as eigrants all foreign-born individuals at least 25 years old who live in an OECD country and class the by educational attainent and country of origin. Thereby, eigration into OECD countries is captured, representing approxiately 90 percent of educated igrants in the world. 9 Because we are interested in bilateral igration patterns, we eploy an extended dataset by Docquier et al. (2007. We construct the high-skilled eigration rate fro country i to j in year t, denoted by Mig ij,t, as the stock of skilled eigrants fro country i living in (OECD country j divided by the stock of skilled residents in (source country i. In soe regressions, we also control for the (lagged low-skilled eigration rate, Mig ij,t-1, which is constructed analogously. Denote by y i, t the outcoe easure in country i in year t. We consider GDP per capita, TFP, and the wage incoe of skilled workers. For a country pair ( i, j, we estiate y j, t ' (23 log = β 0 + β1migij, t + β2migij, t 1 + xij, t 1 x + uij. y β i, t Equation (23 is theoretically otivated by relationships w w = ~ ω (, / and A A = ~ α (, ; see equations (14 and (13 derived in section I, respectively. / According to equation (19, the theoretical odel suggests that β 1 > 0 when the log difference in TFP, log( A / A, is the dependent variable. When the log difference of wages for skilled workers, log( w / w, is the dependent variable, then we predict β 1 > 0 if and only if the TFP effects of igration are sufficiently high, according to 15

18 equation (21. Moreover, we predict β 2 > 0 when log( w / w is the dependent variable. x ij, t 1 is a vector of other (lagged controls that potentially affect log incoe differences between i and j, such as relative school enrollent rates, relative investent rates, relative urban population shares, and fixed effects for the source country to capture institutional differences to OECD destination countries. With respect to the dependent outcoe easures, we focus on the year 2000 and usually easure controls other than skilled igration in lagged for (for 1990 to reduce endogeneity bias. u ij is an error ter. To construct a easure of log( w / w, we would like to use (log wages differences for high-skilled individuals. owever, because wage incoe by education category is not available, we construct several epirical proxy easures. Freean and Oostendorp (2000 analyze inforation on earnings by occupation and industry fro the ILO October Inquiry urvey fro 1983 to 1998 for a nuber of countries. 10 For each country, we use Freean and Oostendorp s earnings easures to calculate the 80th and the 90th percentiles as two easures for wages of high-skilled workers. For ost countries, data are available for only a few years. Therefore, for each country, we take the ean across the period fro 1995 to 2003 to obtain wage data for the year The two constructed (log relative wage variables for the 80th and the 90th percentiles are denoted by RelWage80 ij,t and RelWage90 ij,t. One ay argue that igrating skilled workers do not receive wage incoe in the sae percentile as they do at hoe. In particular, high-skilled workers fro developing countries ay not be considered highly skilled in the destination country. Therefore, as a 16

19 robustness check, we assue that soeone working in the 80th percentile at hoe earns only the edian wage incoe abroad. The corresponding relative wage easure is denoted by RelWage80to50 ij,t. For relative GDP and relative TFP between destination and source countries, denoted by RelGDP ij,t and RelTFP ij,t, respectively, we use Penn World Tables and the NIDO World productivity database. In particular, GDP data have better availability than wage data such that the nuber of observations increases. Details on variable definitions, data sources, and the suary statistics of the eployed variables are presented in the appendix (table A1. As indicated, although recent epirical literature has focused on the ipact of incoe differences on igration patterns, we ai to exaine the opposite channel. In a first attept to address endogeneity, we replace the high-skilled eigration rate in 2000 by the lagged one in 1990, denoted by Mig ij,t-1, in OL regressions. Doing so allows for the possibility that the TFP effects of the igration flows of skilled workers (for instance, through innovation activity take tie to coe into effect. econd, we explore potential instruents for the high-skilled eigration rate for We use the lagged rate of expatriates in 1990 eigrating fro country i to j, denoted by TotalMig ij,t-1, as an instruent for Mig ij,t, thereby predicting the rate of highskilled eigrants by the lagged rate of all eigrants. This approach is otivated by the notion that a larger percentage of eigrants fro a certain source country already living abroad act as a signal to potential high-skilled igrants regarding the destination country s openness and its adinistrative bodies treatent of foreigners. The presence of ore eigrants to a certain destination creates obility cost-reducing network effects 17

20 for potential eigrants (e.g., Massey et al., 1993; Beine, Docquier and Ozden, Past igration also easures other intangible factors unrelated to incoe, such as trust, cultural proxiity, and social openness to igrants of the destination as perceived by eigrants of the source country. Moreover, we eploy indicators for geographical factors (Dist ij, Contig ij and linguistic proxiity (CoLang ij, which are typically used in the literature on igration as additional instruents. To further address potential endogeneity bias, we use the total eigration rate in 1960 instead of TotalMig ij,t-1 as an instruent, which, however, cannot be readily observed at the tie of analysis. We thus construct a proxy for the total eigration rate. Denote by NetMig i,1960 the total net eigration rate (the nuber of eigrants inus the nuber of iigrants divided by population size in country i in 1960, provided by the nited Nations Population Division. 13 Our easure of bilateral total eigration rates in 1960 is defined by NetMigi,1960 Pop j,1960 (24 TotalMig ij,1960 =, Pop 100 i, 1960 where Pop i,1960 is population size in the source i and Pop j,1960 is the population size in the destination j in the year As suggested by Beine, Docquier and Rapoport (2001, one ay use countries population sizes to reflect iigration quotas. NetMig Pop is thus a proxy for the net stock of eigrants fro country i i, 1960 j,1960 received in country j in Because our epirical strategy focuses on eigration rates rather than stocks, we divide this easure by (100 ties the population size of source country i to obtain an estiate for the past bilateral eigration rate. 15 The fraction of high-skilled igrants before 1960 was coparatively low; therefore, potential 18

21 effects of past igration should only work through induced high-skilled eigration. In other words, the instruent should be uncorrelated with the dependent variable, which is supported by J-tests. Results Reported standard errors fro all estiates account for destination clusters, following Grogger and anson (2011, aong others. 16 <<table 1 about here>> Table 1 presents OL estiates of equation (23. We first oit the low-skilled igration rate. We observe that the estiated effects of an increase in the high-skilled igration rate on relative GDP (RelGDP ij,t, relative TFP (RelTFP ij,t, and relative wages (RelWage80 ij,t and RelWage90 ij,t between destination and source countries are positive and significant. sing the lagged high-skilled igration rate (Mig ij,t-1 rather than the conteporaneous one (Mig ij,t only slightly decreases the coefficient. Therefore, an increase in the high-skilled eigration rate increases (log incoe differences between countries. The control variables of all estiates include the lagged relative school enrolent (priary and tertiary, the relative capital investent, and the relative urban population share as well as source fixed effects. Except for (lagged priary school enrollent, which is never significant, the controls have the expected signs. The (lagged relative investent rate and the (lagged relative urban population share are typically significantly different fro zero. To consider the effect quantitatively, we use a coefficient β 1 of about 0. 2 in the wage regressions presented in coluns (5 (8. Doubling the high-skilled eigration rate 19

22 (Mig ij,t fro its ean level of thus iplies that the relative wage for high-skilled workers between the destination and the source increases by approxiately 0.5 percent ( = This effect is sall, which is consistent with the icroeconoic estiates of the effect of high-skilled iigration on wages for highly skilled individuals inside the nited tates by Borjas (2003 and for the nited Kingdo by Dustann et al. (2005. <<table 2 about here>> Tables 2 4 address the potential proble of reverse causality by providing instruental variable estiations of (23. The upper panels report second-stage results, whereas the lower panels in tables 2 and 3 report the partial correlations of the instruents in the first stage. We start with the results for relative GDP as a dependent variable in table 2. In coluns (1 and (2, we use the total eigration rate fro country i to j in 1990 (TotalMig ij,t-1 as a single instruent. In coluns (3 (6, the bilateral geographical distance between i and j (Dist ij, an indicator for a coon border (Contig ij, and an indicator for the coon language of the source and destination countries (CoLang ij are used as instruents in addition to the total eigration rate. We use TotalMig ij,t-1 in coluns (3 and (4 and our proxy for the total eigration rate for 1960, TotalMig ij,1960, in coluns (5 and (6. As in table 1, we control for the lagged relative values of school enrollent, private investent, and urbanization and include source country fixed effects (results not shown. The effect of high-skilled igration on log GDP differences between the destination and the source country is positive, as in the OL estiations. All estiates suggest a significant and higher effect of skilled igration on relative GDP 20

23 copared to the OL estiates in table 1. Coluns (2, (4, and (6 also control for the (lagged low-skilled igration rate in 1990, Mig ij,t-1. We observe that the coefficient on Mig ij,t-1, β 2 in equation (23, is not significantly different fro zero and does not alter the coefficient of the instruented variable Mig ij,t in an iportant way. Coluns (7 (12 in table 2 present the results for relative TFP analogously to coluns (1 (6. The results are siilar to those for relative GDP: the estiated effect of high-skilled igration is always positive and increases copared with OL estiates, whereas low-skilled igration is not significant. In particular, the estiates of β 1 in coluns (7 (12 of table 2 confir our theoretical prediction that α = A / A is increasing in as a result of huan capital externalities. Again, the coefficient of the (lagged low-skilled igration rate, β 2, is not significantly different fro zero and is soeties positive, in line with the theoretical odel. An F-test for the first-stage results shows that the instruents are significantly related to the eigration rate. In particular, past igration appears to be an iportant deterinant of high-skilled igration. 18 None of the J-statistics suggest probles with the instruents. <<table 3 about here>> In table 3, we present the results analogous to table 2 for relative wages in the 80th and 90th percentiles instead of relative GDP and relative TFP, respectively. Again, coluns (1 (2 and (7 (8 use the total eigration rate in 1990 as a single instruent for the high-skilled eigration rate. The first-stage results indicate that the total eigration rate in 1990 is well correlated with Mig ij,t. β 1 is again positive and 21

24 significantly different fro zero. According to the other estiations in table 3, the results are siilar when using the easure for the total igration rate in 1960, geographical variables, and linguistic proxiity as instruents. According to the theoretical prediction in equation (21, β 1 should be higher when relative TFP (α rather than the relative wages of skilled labor ( ω is the dependent variable. In the estiates presented in tables 2 and 3, this is not the case. It is iportant to note, however, that saple sizes are very different because wage data are available for less (and, on average, richer countries than TFP. Estiated coefficients of the instruented high-skilled igration rate in 2000, Mig ij,t, becoe saller when we also control for the lagged low-skilled igration rate in 1990, Mig ij,t-1. Moreover, coefficient β 2 of Mig ij,t-1 is positive and typically significant (and is higher than β 1. This finding is in line with the theoretical prediction and is due to the copleentarity between skilled and unskilled labor. Only in coluns (6 and (12 does β 1 becoe insignificant; in these cases, it reains positive and quantitatively sizable. In su, we ay conclude that the effect of skilled igration on international wage differences, albeit liited in agnitude, is positive and also often significant. The results of relative TFP in table 2 and those in table 3 in connection with our theoretical considerations appear to suggest that the possible positive effects of skilled iigration on the wages of skilled workers are derived fro the positive TFP effects of skilled iigration. Moreover, low-skilled igration tends to benefit the skilled labor force in the receiving country. 22

25 The first-stage results in table 3 suggest that factors that are potentially unrelated to incoe, such as network effects, language, and geography, drive the high-skilled eigration rate. Interestingly, the coefficients of the instruented variable Mig ij,t in table 3 are often ore than twice as high as in OL regressions (table 1. This finding suggests that igrants who arrive through social networks have a particularly high ipact on international differences in (log wages of skilled workers. Migrants who arrive through social networks appear to find it easier to integrate into the host country and thus have a larger effect on TFP (possibly being eployed in jobs that are ore suitable to their qualifications than workers without social networks. In fact, we cannot rule out that skilled iigrants work in different jobs than they do in the source country, often earning wages that are within a lower percentile of the wage distribution than at hoe. For instance, a university degree in a developing source country ay reflect a lower acquired skill level than a university degree in an OECD destination country. Moreover, a skilled iigrant ay occupy a low-skilled job briefly after arrival owing to language probles in the destination country. We account for these possibilities by using as the dependent variable the log difference between the wage of the edian in the destination country and the 80th percentile in the source country, RelWage80to50 ij,t. <<table 4 about here>> The results are reported in table 4. Coluns (1 and (2 are analogous to the OL estiations in table 1 and show siilar results as the wage regressions (5 (8 in table 1. Coluns (3 (8 are instruental variable estiations, which are analogous, for instance, to coluns (1 (6 of table 3 with respect to the use of instruents. The instruental 23

26 variable estiates are siilar in significance and agnitude to the results of the wage regressions in table 3. We conduct a further sensitivity analysis (see table.1 to This analysis suggests that our conclusions are fairly robust overall. First, we include destination fixed effects rather than source fixed effects. The results with destination fixed effects are siilar to those with source fixed effects. 20 We also exaine whether results are sensitive to a specific destination country. We run rolling regressions, in which we oit one destination country each tie, to confir that the results are basically unchanged. econd, we include regional duies and a duy variable that indicates whether the source country also belongs to the OECD 21 instead of fixed effects as controls to consider institutional differences, which ay affect incoe differences, in an alternative way. Third, we eploy an eigration data set by Defoort (2006 to construct an alternative proxy for the total eigration rate. The data set contains eigration to six iportant destination countries in the year The proxy is constructed analogously to equation (24 and is used as an instruent for the skilled igration rate. Finally, we use the stock of high-skilled and low-skilled igrants rather than igration rates as regressors. Our ain conclusions reain qualitatively unchanged and overall robust. III. CONCLDING REMARK In this paper, we analyzed the ipact of an increase in the international bilateral igration of high-skilled and low-skilled workers on relative incoe and relative TFP between pairs of source and destination countries of expatriates. Our theoretical odel suggested that an increase in the nuber of skilled igrants increases international wage 24

27 inequality by adversely affecting TFP in the source econoy and increasing it in the host econoy. Our epirical analysis provided evidence which is consistent with this hypothesis. sing a data set on the bilateral eigration of skilled workers, our results suggested that an increase in high-skilled eigration rates tends to slightly increases TFP differences and therefore (albeit also slightly wage incoe for skilled workers in destination countries relative to source countries in a causal way. None of our estiations suggested that skilled workers in the destination country lose fro skilled igration relative to the source country. Finally, skilled workers in the receiving countries unabiguously gain fro low-skilled igration. APPENDIX <<table A.1 about here>> REFERENCE Aceoglu, Daron, and Joshua Angrist ow Large are uan-capital Externalities? Evidence fro Copulsory chooling Laws. NBER Macroeconoics Annual 15: Beine, Michel, Frédéric Docquier, and illel Rapoport Brain Drain and Econoic Growth: Theory And Evidence. Journal of Developent Econoics 64: Beine, Michel, Fréderic Docquier, and illel Rapoport Brain Drain and uan Capital Foration in Developing Countries: Winners and Losers. Econoic Journal 118:

28 Beine, Michel, Frédéric Docquier, and Caglar Ozden Diasporas. Journal of Developent Econoics 95: Borjas, George J The Labor Deand Curve is Downward loping: Reexaining the Ipact of Iigration on the Labor Market. Quarterly Journal of Econoics 118: Defoort, Cecile Tendances de Long Tere en Migrations Internationales: Analyse à Partir de 6 Pays Receveurs. niversité Catholique de Louvain (ieo. Docquier, Frédéric, and Aldesla Marfouk International Migration by Educational Attainent ( Release 1.1. In International Migration, Reittances and Developent. Ed. C. Ozden and M. chiff, pp New York: Palgrave Macillan. Docquier, Frédéric, B. Lindsay Lowell, and Abdesla Marfouk A Gendered Assessent of the Brain Drain. IZA Discussion Papers No Dustann, Christian, Francesca Fabbri, and Ian Preston The Ipact of Iigration on the British Labour Market. Econoic Journal 115: F Egger, Peter, and Doina Maria Radulescu The Influence of Labour Taxes on the Migration of killed Workers. The World Econoy 32: Friedberg, Rachel M The Ipact Of Mass Migration On The Israeli Labor Market. Quarterly Journal of Econoics 116: Freean, Richard B., and Reco. Oostendorp Wages Around the World: Pay Across Occupations and Countries. NBER Working Paper No

29 Gennaioli, Nicola, Rafael La Porta, Florencio Lopez-de-ilanes, and Andrei hleifer uan Capital and Regional Developent. NBER Working Paper No Grogger, Jeffrey, and Gordon. anson Incoe axiization and the selection and sorting of international igrants. Journal of Developent Econoics 95: Grossann, Volker, and David tadelann International obility of the highly skilled, endogenous R&D, and public infrastructure investent. IZA Discussion Paper No Grossann, Volker, and David tadelann Does international obility of highskilled workers aggravate between-country inequality? Journal of Developent Econoics 95: anson, Gordon., Kenneth cheve, and Matthew J. laughter Individual Preferences over igh-killed Iigration in the nited tates. In killed Iigration Today: Probles, Prospects, and Policies, eds. Jagdish Bhagwati and Gordon anson, Oxford niversity Press. unt, Jennifer, and Marjolaine Gauthier-Loiselle ow Much Does Iigration Boost Innovation? Aerican Econoic Journal: Macroeconoics 2: Iranzo, usana, and Giovanni Peri (2009. chooling Externalities, Technology, and Productivity: Theory and Evidence fro.. tates, Review of Econoics and tatistics 91,

30 Lucas, Robert E. (1988. On the Mechanics of Econoic Developent, Journal of Monetary Econoics 22, Lucas, Robert E.B International Migration and Econoic Developent: Lessons fro Low-incoe Countries. Cheltenha, Edward Elgar Publishing. Massey, Douglas., Joaquín Arango, Graee ugo, Ali Kouaouci, Adela Pellegrino, and J. Edward Taylor Theories of International Migration: A Review and Appraisal. Population and Developent Review 19: Mayer, Thierry, and oledad Zignago A Note on CEPII's Distances Measures, Explanatory note, CEPII, Paris. Mountford, Andrew Can a Brain Drain be Good for Growth in the ource Econoy? Journal of Developent Econoics 53: Müller, Tobias, and ilvio.t. Tai Individual attitudes towards igration: a reexaination of the evidence. niversity of Geneva, ieo. tark, Oded, Christian elenstein, and Alexia Prskawetz A Brain Gain with a Brain Drain. Econoics Letters 55: NOTE 1 The nuber of tertiary-educated iigrants living in OECD countries increased fro 12.5 illion in 1990 to 20.4 illion in 2000 (Docquier and Marfouk, alf of the skilled igrants resided in the nited tates, and approxiately one-quarter resided in other Anglo-axon countries. 28

31 2 igh-skilled iigration surged in witzerland after the enactent of a bilateral agreeent between witzerland and the European nion on the free oveent of labor in June Grossann and tadelann (2011 develop an overlapping-generations odel with endogenous education choice that shows how igration is triggered by a decrease in the obility costs of high-skilled workers and how it ay evolve over tie. In the present paper, we focus epirically on the effect of higher international igration. 4 Recall that individual i provides 1 ei units of skilled labor when e i e(k. 5 Functions ~ and ~ are defined analogously to equations (9 and (10, respectively. 6 Without a loss of generality, we label the foreign country the host country. 7 In Grossann and tadelann (2008, we presented evidence for the interaction between eigration flows and incoe changes using a structural equation odel. owever, we exained the ipact of a higher aggregate eigration stock of a country on its per capita incoe. That is, we did not consider bilateral relationships. 8 This finding suggests that there exist obility-cost reducing network effects fro counities of people fro the sae nation and fro friends and relatives already living abroad (see also Massey et al., ee Docquier and Marfouk (2006 for a detailed discussion concerning data collection and construction issues. 29

32 10 To correct for differences in how countries report earnings, Freean and Oostendorp (2000 use a standardization procedure to ake the data coparable across countries and tie. In 2005, they provided an update for their earnings easures for the ILO October Inquiry data using an iproved version of the standardization procedure and the application of country-specific data type correction factors. A detailed technical docuentation of the standardization procedure for the ILO October Inquiry data is available online at 11 We also included data for Turkey for the year Another way to capture the effect of obility cost-reducing network effects is to use the past total nuber of igrants instead of the past eigration rate as the instruent for conteporaneous igration. We confir that the results do not change. 13 Countries with negative net eigration are coded to have an eigration rate equal to zero. 14 The easure is inspired by Beine, Docquier and Ozden (2011. They use a siilarly constructed proxy as an instruent for the total diaspora of igrants in 1990 (rather than the high-skilled eigration rate. 15 Calculating partial correlations confirs that the past total eigration rate is well correlated with the high-skilled eigration rate in We use the uber-white ethod to adjust the variance-covariance atrix fro our least squares results. 30

33 17 In fact, between 1990 and 2000, the nuber of tertiary-educated iigrants living in OECD countries alost doubled (Docquier and Marfouk, That contiguity (variable Contig ij has a negative effect on high-skilled eigration in our first-stage estiate parallels a finding siilar to Grogger and anson (2011. They explain the result by selection and sorting effects. 19 The results are reported in an online appendix available at 20 We cannot include both siultaneously because they would, by construction, fully explain the different relative incoe variables due to ulticollinearity. 21 Recall that all destination countries are OECD countries. 31

34 TABLE 1. Effect of high-killed Eigration Rates on Wage, GDP, and TFP Differences between Countries Dependent variable: RelGDP ij,t Dependent variable: RelTFP ij,t Dependent variable: RelWage80 ij,t Dependent variable: RelWage90 ij,t (1 (2 (3 (4 (5 (6 (7 (8 Mig ij,t *** *** *** *** ( ( ( ( Mig ij,t *** ( RelInvest ij,t * ( Relrban ij,t *** ( RelPrichool ij,t ( RelTertchool ij,t ( (Intercept ( * ( *** ( ( * ( ( ( ( ( * ( ( *** ( ( ( ( * ( ( ** ( ** ( ( ( ( ** ( ** ( ** ( ( ( ( * ( * ( ( ( ( Origin FE YE YE YE YE YE YE YE YE ** ( * ( * ( ( ( ( Adj. R N Note: All dependent variables are expressed in logs and represent relative differences between countries j and i. Mig ij,t denotes the stock of high-skilled eigrants fro country i living in country j divided by the stock of high-skilled residents in i. RelInvest ij,t-1, Relrban ij,t-1, RelPrichool ij,t-1 and RelTertchool ij,t-1 denote the lagged relative investent share, relative urbanization share, relative priary school enrollent, and relative tertiary school enrollent between j and i. Table A1 in the appendix provides additional inforation on all variables. Robust standard errors are in parentheses and clustered for igration destinations. *** indicates a significance level below 1 percent; ** indicates a significance level between 1 and 5 percent; * indicates a significance level between 5 and 10 percent. 32

35 TABLE 2. Effect of igh-killed Eigration Rates on GDP and TFP Differences between Countries (Instruental Variables Estiations Mig ij,t * ( Mig ij,t ( Dependent variable: RelGDP ij,t Dependent variable: RelTPF ij,t (1 (2 (3 (4 (5 (6 (7 (8 (9 (10 (11 ( *** ** *** * ** ** *** ** *** *** ( ( ( ( ( ( ( ( ( ( ( Other controls YE YE YE YE YE YE YE YE YE YE YE YE Origin FE YE YE YE YE YE YE YE YE YE YE YE YE Adj. R N F-Test (first stage J-Test Instruents used TotalMigij,t-1 TotalMigij,t-1 TotalMigij,t-1 + Distij + CoLangij + Contigij TotalMigij,t-1 + Distij + CoLangij + Contigij TotalMigij, Disti + CoLangij + Contigij ( TotalMigij, Disti + CoLangij + Contigij First stage (partial correlations ( TotalMigij,t-1 TotalMigij,t-1 TotalMigij,t-1+ Distij + CoLangij + Contigij ( TotalMigij,t-1+ Distij + CoLangij + Contigij TotalMigij, Disti + CoLangij + Contigij *** ( ( TotalMigij, Disti + CoLangij + Contigij TotalMigij,t *** (3.7e *** (7.9e *** (3.8e *** (8.0e 04 TotalMigij, e 04 *** (1.1e 05 Distij *** ( CoLangij ** ( Contigij *** ( *** ( ( *** ( *** ( *** ( ** ( e 04 *** (1.0e *** ( *** ( *** ( *** (5.8e *** ( *** (6.0e *** ( ( *** ( *** ( *** ( * ( ** ( e 04 *** (3.1e *** ( *** ( * ( e 04 *** (3.0e *** ( *** ( *** ( Note: All dependent variables are expressed in logs and represent relative differences between countries j and i. Mig ij,t (Mig ij,t-1 denotes the stock of high- (low- skilled eigrants fro country i living in country j divided by the stock of high- (low- skilled residents in i. All estiations include RelInvest ij,t-1, Relrban ij,t-1, RelPrichool ij,t-1 and RelTertchool ij,t-1 as additional control variables. TotalMig ij,t-1, Dist ij, CoLang ij, and Contig ij represent the share of the eigrant population fro country i living in country j, the distance between i and j, whether i and j share a coon language, and whether i and j have a coon border, respectively. Table A1 in the appendix provides additional inforation on all variables and instruents. Robust standard errors are in parentheses and clustered for igration destinations.*** indicates a significance level below 1 percent; ** indicates a significance level between 1 and 5 percent; * indicates a significance level between 5 and 10 percent. 33