The economic e ects of immigration: evidence from European regions.

The economic e ects of immigration: evidence from European regions. Gianluca Ore ce* University of Milano and Centro Studi Luca D Agliano June 26, 2009 Abstract This paper investigates the e ects of immigration ows on the economic performance of European regions. Literature provides a substantial (if ambigous) evidence on the labor market e ects of immigration, but precious little is known about the relation between immigration and economic performance. In order to account for the heterogeneity among regions in both immigrants endowment and economic performaces, data on regional level (NUTS2) have been used; data cover 260 European regions in the period 2002-2005. The main results are that immigration is positively correlated with both an expansion of per-capita GDP and per worker gross value added (GVA), as well as positively correlated with inward and outward FDI. Keywords: Immigration, FDI, regional economic growth, factor mobility. JEL Classi cation: F22, J61, R11. (very preliminary, please do not circulate without the authors consent) * Gianluca Ore ce, department of Economics, Business and Statistics, University of Milan, via Conservatorio 7, 20122 Milano. E-mail: gianluca.ore ce@unimi.it 1

1 Introduction The aim of the paper is to analyse the e ects of immigration on both regional economic performances and FDI. In particular this paper tries to investigate the e ects of immigration on: (i) per capita GDP, (ii) per worker gross value added, (iii) inward FDI and (iv) outward FDI. The growing international labor migration suggests the importance of this topic in international economics: the percentage of foreign-born population over the total population residing (legally) in USA has increased by 3.6% from 1995 to 2005 moreover, the percentage of foreign-born over USA total population in 2005 was more than 15% 1. In Europe the immigrants endowment over the total population increased from less than 4% in 1960 to more than 8% in 2005 (Figure 1). Figure 2 shows the immigrants endowment of Eu15 countries from 1960 to 2005: in Germany there are about 10 mil. of foreign born individuals who accounts for 12.3% of the total population, while Finland hosts only 0.17 mil. immigrants that is about 3% of the total population. In Italy in 2005 lived 2.52 mil. of immigrants (4.3%of the total population). Thus migration has, potentially, a crucial role for the comprehension of future economic development: does immigration increase GDP for the hosting countries? Do immigrants contribute signi cantly to the economic performance of the hosting regions? Do FDI go where a wide immigrant workers supply exists? Do immigrant workers make local rms more competitive and thus push the rm s internationalization? These are the main questions that the paper wants to investigate. The importance of this topic is straightforward if one looks at the policy implication but, since few works have been written on this topic, it also tries to complete a lack in literature. The underlying idea is that immigrants by increasing the low wage workers endowment of a region, may push up the overall production level of such region. Moreover, assuming imperfect substitution between native and foreign born workers, immigrantion can induce a better task specialization (within home and foreign born workers) and thus the labor productivity is expected to increase (Peri and Sparber 2008). As a consequence, the increased labor productivity may drive rms to internationalize the production 2 (outward FDI). Finally, the 1 United Nations, Department of Social a airs "Trend in total migrants stock: the 2005 revision" http://esa.un.org/migration 2 The increase in labor productivity promotes outward FDI because it will be easier for local rms to overcome the productivity threshold to internationalize the production (Helpman, Melitz and Yeaple 2004). 2

increase in labor endowment may push inward FDI looking for cost reduction or labor supply. In providing empirical evidence of the previous correlations, this paper tries to keep together the features shown in Ortega (2008) and Ortega and Peri (2009) by using as an instrumental variable the estimated gravity-push immigration without economic determinants. The last feature in data regards the high region s heterogeneity in both economic performces and immigrants endowment; to take into account this feature the analysis will concern 260 European regions (NUTS2). In order to understand the economic e ects of immigration, this paper provides an econometric estimation (by using both OLS and IV models) of the impact of immigration on: (i) economics performaces (measured as per capita GDP both in level and percentage variation); (ii) labor productivity (measured as per worker gross value added both in level and percentage variation); (iii) inward FDI (as number of project), (iv) outward FDI (as number of projects). The rest of the paper is organized as follows: section 2 provides a short review of the existing literature on the e ects of immigration on the host economy; section 3 reports some descriptive statistics; Section 4 presents the empirical model, econometric strategy and results. Section 5 concludes. Figure 1: Historical immigrants endowments as percentage of total poplation Immigrants as % of total population 20 18 16 14 12 10 8 6 4 2 0 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 Africa Europe Northern America Asia Latin America and the Caribbean Oceania Source: United Nations, Department of Economic and Social Affairs 3

Figure 2: Immigrants endowment in the Eu-15 countries 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 Austria 0.10 0.13 0.17 0.22 0.28 0.28 0.47 0.72 0.93 1.23 Belgium 0.44 0.54 0.68 0.77 0.87 0.89 0.90 0.91 0.88 0.72 Denmark 0.08 0.10 0.12 0.14 0.16 0.19 0.22 0.25 0.30 0.39 Finland 0.03 0.03 0.03 0.04 0.04 0.05 0.06 0.10 0.13 0.16 France 3.51 4.45 5.21 5.57 5.89 5.96 5.91 6.09 6.28 6.47 Germany 5.94 9.09 9.80 10.14 Number of Greece 0.05 0.07 0.09 0.12 0.17 0.31 0.41 0.55 0.73 0.97 international migrants Ireland 0.07 0.09 0.13 0.17 0.22 0.23 0.23 0.26 0.38 0.59 (mil.) Italy 0.75 0.83 0.91 1.01 1.11 1.22 1.35 1.48 1.63 2.52 Luxembourg 0.04 0.05 0.06 0.08 0.09 0.10 0.11 0.14 0.16 0.17 Netherlands 0.45 0.34 0.26 0.34 0.49 0.76 1.19 1.39 1.56 1.64 Portugal 0.04 0.06 0.10 0.16 0.27 0.35 0.44 0.53 0.63 0.76 Spain 0.24 0.29 0.37 0.30 0.24 0.41 0.77 1.01 1.63 4.79 Sweden 0.29 0.39 0.54 0.57 0.61 0.65 0.78 0.91 0.99 1.12 United Kingdom 1.66 2.54 2.95 3.20 3.47 3.62 3.75 4.20 4.76 5.41 International migrants as a percentage of the population 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 Austria 1.5 1.8 2.3 2.9 3.7 3.7 6.1 8.9 11.4 15.1 Belgium 4.8 5.7 7 7.9 8.8 9 9 9 8.5 6.9 Denmark 1.8 2.1 2.4 2.7 3.2 3.7 4.3 4.8 5.7 7.2 Finland 0.7 0.7 0.7 0.8 0.8 1 1.2 2 2.6 3 France 7.7 9.1 10.3 10.6 10.9 10.8 10.4 10.5 10.6 10.7 Germany 7.5 11.1 11.9 12.3 Greece 0.6 0.8 1 1.4 1.8 3.1 4.1 5.1 6.7 8.8 Ireland 2.6 3.3 4.4 5.4 6.6 6.4 6.5 7.3 10.1 14.1 Italy 1.5 1.6 1.7 1.8 2 2.2 2.4 2.6 2.8 4.3 Luxembourg 13 15.8 18.2 21 25.5 28.3 30.2 33.4 36.9 37.4 Netherlands 3.9 2.8 2 2.5 3.5 5.3 8 9 9.8 10.1 Portugal 0.4 0.7 1.2 1.8 2.7 3.5 4.4 5.3 6.2 7.3 Spain 0.8 0.9 1.1 0.8 0.6 1.1 1.9 2.5 4 11.1 Sweden 3.9 5.1 6.7 7 7.3 7.8 9.1 10.3 11.2 12.4 United Kingdom 3.2 4.7 5.4 5.8 6.3 6.5 6.6 7.3 8.1 9.1 Source: United Nations, Department of Economic and Social Affairs 2 Literature Review From a theoretical viewpoint, the economic e ects of migration on receiving countries can be analysed beginning with the extension of the traditional Slow-Swan model. The simplest thing is to assume immigrants transporting no human capital, in this case immigration is like an increase in country s population so that immigration leads to a slower per capita income growth (because of the local capital dilution). But if we assume that immigrants transport some kind of human capital, this may o set the dilution of local physical capital and some economic growth in per capita terms is allowed for. A rst estimation on the economic e ects of immigration by Dolado, Goria and Ichino (1993) found a negative e ect of immigration on per capita income growth, so they argued that this was due to the fact that immigrants in OECD countries have lower human capital than natives. Borjas (1995) set a model in which "immigration surplus 3 " is de ned as the overall receiving 3 In Borjas (1995) the immigrants surplus is de ned as follows: immigrants surplus = [w(0) w(1)]m = 1 2 2 sem2 Where w is the wage level, s is labor s share of national income, e is the elasticity of factor price for labor, m is 4

country gain from immigration. Let s assume a traditional production function with capital (K) and labor (L), where labor is composed by natives and foreign born individuals; when immigration is allowed, the national labor endowment rises as far as L: The new internal equilibrium is now characterized by lower national wage and higher employment, as a result the national income increases and the di erence with respect to the initial equilibrium is the so called "immigrants surplus". Borjas (2006) uses data from 1960 to 2000 to calculate the immigration surplus, in the simulation he assumes 0.7 labor s share of nationl income and a 10% increase in the supply of workers in a skill group reduces the wage of that group by 3.5% (elasticity of factor price for labor). He nds that the immigration surplus in USA was 1 billion dollars in 1960 and 21.5 billion dollars in 2000. But, immigration doesn t just increase the cake (GDP), it also a ects the size of the slices: immigration reduced the total earnign of natives by 2.8% of GDP. In a recent paper Hanson (2008) studies the welfare consequences of immigration. Assuming perfect substitutability between native and foreign-born workers and assuming two kinds of labor input (skilled and unskilled workers) Hanson (2008) shows that when low-skilled workers are allowed to freely move between countries, there will be migration from low wage country to high wage country until the wage will equalize. As a result the wage of sending country will increase, while the wage level in receiving country will decrease, thus home rms can gain from migration. In receiving country home-born workers lose while the native high-skilled workers win in terms of surplus. The immigration surplus in the receiving country is positive and it coincides with the gain in terms of GNP, while the gain in terms of GDP would be greater because it includes the income that migrants receive in the host country. Empirical research on the economic e ects of immigration is scarse (economists have been working mostly on labor market e ects of immigration 4 ), among them Ortega (2008) investigates the short run e ects of immigration on Spanish regions in the period 1998-2008. The main empirical the fraction of the workforce that is foreign born. The elasticity of factor price is small when the labor demand is elastic, so the immigration surplus is small when labor and capital are easily substitutable. The immigration surplus, therefore, arises because of the complementarities that exist between immigrants ad native-owned capital 4 Card (2001, 2005), Borjas (2003), Aydemir and Borjas(2007), Borjas Grogger and Hanson (2008) nd negative link between native low-skilled wages and immigrants. On the contrary Ottaviano and Peri (2008), Peri and Sparber (2009) nd positive link between native wage and immigration. 5

problem concerning the estimation of the economic e ects of immigration is the endogeneity problem. In order to solve this kind of problem Ortega (2008), following the literature from Card (2001) on, uses the estimated immigration in ows as instrumental variable 5 : [1] Z r;t = P c F Br;c;t0 F B c;t0 F B c;t where F B r;c;t0 is the number of individuals born in foreign country c and residing in region r at t 0, F B c;t0 is the number of individuals born in foreign country c and residing in the host country which region r is part of; F B c;t is the immigration in ows from c at t. This instrument is based on a well known feature of immigration: the presence of ethnic networks in immigration localization decision. So the estimated immigrants in ows in region r at time t;as in [1], depends on the historical (with t 0 << t) regional localization of migrants. Assuming that the total population varies only because of the immigrants in ows, Ortega estimates the e ects of population changes on economic performances using the estimated immigration in ows (Z) as an instrument for the population variable. In this way, one can infer the e ects of immigration by looking at the e ects of population changes on the economic performance. He nds that a 10% growth in total population leads to a 6.7% increase in GDP, to a 3.3% reduction in per capita GDP and to a 4% reduction in per hour GDP. Because the estimated immigration in ows (the instrument Z) is a strong predictor (with positive relation) of population growth rates, he looks at the population coe cient in order to evaluate the e ects of immigration. The weak point of this analysis is to consider the e ects of population growth to assert something on immigration e ects. The paper by Ortega and Peri (2009) looks both at the determinants and consequences of immigration. They estimate a gravity style model to predict the immigrants in ows using wage di erential, distance, land border dummy, language dummy and colonial dummy. Thus, they used this kind of estimation without wage di erential, to build their instrumental variable obtaining estimated immigration in ows well correlated to immigration in ows and not correlated with economic performance. They test the signi cance of the instrument with di erent samples of data and in each case the coe cient of the instrument is positive and very signi cant. The 2SLS 5 This instrument is widely used in the empirical literature concerning the labor market e ect of immigration. It was introduced for the rst time by Card (2001). 6

estimates show that an increasing immigration leads to: (i) an increasing employment growth, (ii) a decreasing hour per worker growth, (iii) an increasing in GDP and capital growth. The idea that immigrants may in uence the in ows of foreign capital (FDI) borns with respect to the paper by Grossman (1982). In this empirical paper a strong complementarity was found between capital and foreign born workers. Grossman estimated that in the short run, when some rigidities in labor market make native s wages downwardly in exible, a 10% increase in immigrants endowment leads to a 2.2% increase of the return on capital, wich in turn attract foreign capital. In the long run, when all wages are exible, a 10% increase in immigration leads to a 4.2% increase in the return on capital. This paper tries to keep together the features shown in Ortega (2008) and Ortega and Peri (2009). The instrumental variable will be an estimated gravity-push immigration, that consider also the agglomeration e ect in historical immigrants localization (in order to take into account the ethnic network e ects). The other point of the paper is to nd empirical evidence of the relation between immigration and FDI; this link is debated in literature and unambiguous evidence does not exist. From a traditional Hecksher Ohlin approach, labor in ow will reduce the capital relative endowment and thus increases capital remuneration, this will deter outward FDI but stimulates inward FDI. Up to now some theoretical studies state that a kind of complemetarity between immigrants and o shoring exist: immigrants provide additional information about their country of origin, this reduces the rm s risk to produce abroad pushing in this way outward FDI (business network literature). Empirical evidence was not able to solve this ambiguity: Bandyopadhyay and Wall (2005) and Barba Navaretti Bertola and Sembenelli (2008) nd negative relation between immigrants and o shoring, while Barry (2002), Javorcik (2006) nd complementarity between immigrants and outward FDI. Probably in assessing the relation between immigration and FDI the skill intensity of migrants and the sector of FDI are important. Kugler and Rapoport (2005) nd positive correlation between FDI and skilled migration in service sector but negative correlation between FDI in manufacturing and skilled immigration. El Yaman, Kugler and Rapoport (2007) nd complementarity between outward FDI and skilled migration but substitutability between unskilled migrants and outward FDI. 7

3 Data and Stylized Facts 3.1 Data The data set consists of four variables: stock of immigrants, inward and outward FDI and some proxies of regional economic performances (per capita GDP, per worker gross value added). The unit of observation is the region (NUTS 2). Data on immigrants, per capita GDP and gross value added are from Eurostat data base, in particular data on immigration comes from the Census 2001, while data on economic performances comes from the regional statistics data base. Data on inward and outward FDI are taken from the OCOMONITOR data base, where it is possible to obtain data on inward FDI in a very detailed geographic localization. The data set includes regions from: Austria, Cyprus, Czech Republic, Estonia, Finland, France, Germany, Greece, Hungary, Italy, Latvia, Lithuania, Luxembourg, Netherlands, Norway, Portugal, Poland, Slovakia, Slovenia, Spain, Sweden, Switzerland, United Kingdom. Since data on the in ow of immigrants at regional level are not available, it has been approximated as follows: [2] imm_flows region;time = stock_imm2001;region stock_imm 2001;country imm_flows country;time Assuming that new immigrants go to regions where historically immigrants are localized, we can allocate new immigrants in ows (imm_flows country;time ) given at country level, on the base of recent regional immigrant distribution (stock_imm 2001;region = stock_imm 2001;country ). The underlying assumption is that from 2001 Census to recent years regional immigrants distribution did not change. 3.2 Stylized Facts Figures 3-14 reports some descriptive evidence concerning the relation between the variables considered in this paper 6. In particular gure 3 and 4 show a positive relation between immigrants endowment (stock of foreign born individuals in 2001) and the regional economics performaces both 6 For a more clear evidence, in the paper has been reported gures for a subsample of the data set where the poorest regions are excluded. 8

in term of per capita gdp and in terms of per worker gross value added, this kind of relation is valid along the considered time horizon (2002-2005). Figure 5 and 6 show an interesting feature: positive relation between the endowment of immigrants in 2001 and the number of both inward and outward FDI projects. This descriptive evidence nds a theoretical backgroud looking at two propositions: (i) immigrants increases the endowment of low wage labor force wich in turn attracts FDI; (ii) immigrants induce a better tasks specialization increasing the home rm s labor productivity, then, following the well known Helpman, Melitz and Yeaple approach, rms will be more competitive and will overcome the threshold to internationalize the production (outward FDI). Figures 7-10 show the same kind of evidence of the former gures but using the immigrant s in ows 7 rather than the stock of immigrants. Again, immigrant s in ows are positively correlated with the economic performances (per capita gdp and per worker gross value added) and both inward and outward FDI projects. In gure 11 are reported the kernel density estimations for per capita GDP, per worker gross value added, inward and outward FDI; the dashed lines in the gures are the density for the low-intensive immigrants regions 8, while the unbroken lines are densities for high-intensive immigrants regions. It can be seen that high-intensive immigrants regions have an higher mean value of per capita gdp, per worker gross value added, number of inward and outward FDI. Figure 12 shows the strong positive relation between inward and outward FDI projects, in particular the overall number of projects from 2003 to 2006 is reported in the gure. Figure 13 shows the Lorenz curves for population, per capita gdp, per worker gross value added, immigrants in ows, inward and outward FDI. Outward FDI is the more concentrated variable among regions (but also inward FDI and immigrants are concentrated as well), on the contrary per capita GDP and per worker gross value added seem not very concentrated. This paragraph provides some descriptive evidence of the positive correlation between: (i) immigration and economic performance, (ii) immigration and both inward and outward FDI. The following paragraph tries to give a strong econometric evidence to these correlations. 7 The immigrants in ows have been computed as shown in section 4.1 8 Low intensive immigrants regions are regions with an immigrants endowment under the mean value. At the same way, high-intensive immigrats regions are region with an immigrants endowment over the mean value 9

4 Empirical Strategy The aim of the paper is to asses whether a correlation exists between immigration and regional economic performances, in particular correlations are between immigration and: (i) per capita GDP, (ii) labor productivity (per worker gross value added), (iii) inward and outward FDI. To this end the empirical strategy can be implemented into two steps: (i) building the instrumental variable by estimating the expected immigration in ows in order to solve the endogeneity problem; (ii) estimating the e ects of immigration on some regional economic performances and FDI projects 4.1 The empirical approach: problems and solutions A likely explanation of why such ambiguity in empirical evidence about the economic consequences of immigration is concerning empirical methods and strategies. A lot of approach has been used in this literature, this is due to some problems that arise when one tries to estimate immigration s e ects on hosting economies: (i) endogeneity from immigrants localization choose, (ii) internal migration and factor price equalization, (iv) data availability. Endogeneity arises if immigrants choose where to stay on the basis of regional wage or GDP di erentials. In this instance it is true not only that immigration drives economic performances (or labor market changes), but also that local economic performaces drive immigration. This problem leads to a biased estimation of the e ects of immigration on economic performances. The endogeneity problem can be solved by using instrumental variables: if one can nd a variable correlated with the change in immigrants presence and uncorrelated with the local economic performance, the bias due to immigration choice can be removed. When immigrants choose the region where to stay, they can take into account also other aspects of a region, such as existing networks and the presence of a community with the same culture and language. Thus, besides economic performance reasons, immigrants may tend to settle in regions with high density of immigrants. Since the stock of existing immigrants in a region is unlikely to be correlated with current economic shocks (notice that a su cient time lag is necessary), historic settlement pattern may solve the endogeneity problem. Altonji and Card (2001) used the stock of immigrants in 1970 as an instrumental variable for the change in immigrant 10

population between 1970 and 1980 in USA cities. The logic is the following: new immigrants tend to go where other immigrants already reside, but this variable is uncorrelated with local economic outcomes or wages. Unfortunately data on immigrants endowment at regional level among Europe countries (Eurostat) are available only for 2001 so there isn t su cient time lag with in ows in 2003-2005. For this reason has been followed Ortega and Peri 2009 in building the instrumental variable. A gravity-push immigration volume without economic determinants was estimated, and the t of this regression used as instrumental variable. In this way the instrumental variable results to be well correlated with immigration ows and uncorrelated with recent economic shocks. The problem of internal migration concerns the factor price equalization within the country in country level analysis. This problem is here solved by using a regional level data analysis. Low quality data problem can be solved by providing some reasons for caution in using the foreign born by total residents: (i) a considerable number of foreign born workers in manufacturing industries are skilled;(ii) not all native born workers are skilled and (iii) not all immigrants participate in the labor market, particularly following an intense process of family regrouping in recent years (Friedberg and Hunt 1995). 4.2 Econometric Speci cation The rst step is to build the instrumental variable. In order to do this, a gravity style equation has been estimated: [3] immi_flow r;t = t + 1 Bord r + 2 Curr r + 3 Eu r + 4 pop 1992;r where immi_flow r;t is the computed immigration in ows as in [2], t are the time xed e ects,bord r is a dummy variable equal to one if the region is a border region, Curr r;t is a dummy variable equal to one if the region belongs to a EMU member country, Eu r is a dummy variable equal to one if the region belongs to a UE27 member country; immi_pop 1992;r is the total population (native plus foreign born individuals) in each region in 1992 (this variable keeps the gravity e ect 11

by agglomeration 9 ). Estimating this kind of regression (following Ortega and Peri 2009), one may use the t of the regression as an estimated immigrants in ows without any economic determinant (instrumental variable). This estimated immigrants in ows (from now on gravity-push estimated immigrants ows) will be the instrumental variable to solve the endogeneity problem. To be a good instrument, it has to be: (i) well correlated with the instrumented variable (the immigrants in ows in this case) and (ii) uncorrelated with the dependent variable (economic performces and FDI in this case). As shown in gure 14, the economic performance is not correlated with the gravity-push estimated immigrants in ows, while they are well correlated with the immigrants in ows ( gure 13). Using the 2SLS estimation, the main estimated equation is: [4] ln y r;t = t + ln immi_flows r;t + " r;t. The left hand side of the equation represents the economic performances: per capita GDP (both in level and in percentage growth), per worker gross value added (both in level and in percentage gowth) and both inward and outward FDI (as number of investment in/from the region). The right hand side of the equation is composed by the time xed e ects t and by the immigration in ows in the region ln immi_flows r;t : 4.3 Results Figure 16 shows the results of the OLS estimation used to buil the instrumental variable. All the determinants used to estimate the immigrants in ows are signi cant with a good explained variance. This suggests that the t of the OLS estimation is a good proxy for the immigrants in ows without economic determinants. The coe cients associated to regional characteristics dummies (border land, EMU, UE27) and to resident population in 1992 are signi cant. In order to test the robustness of the instrument, three di erent subsamples have been used (overall, EMU region only, without poorest regions 10 ). In each case the coe cient of the instrument is very signi cant, thus 9 The underlying idea is that new immigration waves localize in bigger cities where probably exists a large immigrant s comunity. A su cient time lag between the population resident and the immigrants in ows is needed in order to avoid any economic role in the immigrants localization. 10 The lower 15% of the per capita GDP regional distribution was dropped. 12

the instrument seems to be powerful and captures only the immigration in ows due to gravity-push factors. Figure 17 shows the OLS estimation of the impact of the immigrants in ows to some economic performances and FDI. Coe cients associated to per capita GDP and per worker gross value added are positive and signi cant. In particular a 10% increase in the ln of immigration in ows leads to 1.9% increase in the ln of per capita GDP; while it leads to a 3.8% increase in the ln of per worker gross value added. Coe cient associated to percentage variation in per capita GDP and per worker GVA are close to zero even if very signi cant. Immigrants also lead to an increase in the number of inward and outward FDI. Figure 18 shows the rst stage estimation for the 2SLS model and it con rms the positive strong relation between the gravity-push estimated immigrants in ows and the "true" immigrants in ows 11. Figure 17 reports the results of the IV estimation in the form of equation 6 12, notice that coe cients are quite similar to OLS estimation. In particular a 10% increase in the ln of immigrants supply leads to a 2.5% increase in ln per capita GDP, to a 4% increase in ln per worker gross value added. The e ects of immigration on the percentage change in per capita GDP and on the gross value added is close to zero (when signi cant). A robust positive e ects of immigrantion on both inward and outward FDI has been found: a 10% increase in immigrants in ow (in ln) leads to a 2.4% increase in inward FDI and to a 2.5% increase in outward FDI projects (in ln). Finally, gure 19 shows the results for the robusti ed Durbin-Wu-Hausman test fo the endogeneity. On the base of this test one can conclude that in 11 over 18 estimations there was a problem of endogeneity (hopefully solved by using the IV model). As a last step in the analyses, gure 20 shows the e ects of immigration on the number of both inward and outward FDI in manufacture and services: immigration has a positive e ects both on inward and outward number of FDI in manufacture and services. This result is coherent with the following intuitions: (i) immigrations provides additional low wage labor supply attracting in this way inward FDI, (ii) immigration reduces the marginal cost for the rms and increase the labor productivity (per capita gross value added), so rms may overcome the productivity threshold to internationalize the production. 11 Notice that even if the instrument and the instrumented variable are the same for each estimation, coe cients associated to instrument are di erent because of the di erent number of observations 12 For the 2SLS estiation the heteroskedasticity-robust standard errors option has been usen 13

5 Conclusions The aim of the paper is to study the e ects of immigration on both economic performances and inward and outward FDI. The basic intuition is that immigrants by increasing the low wage workers endowment for some regions, it may bring to an expansion of output level. Moreover immigration can induce a better task specialization (within home and foreign born workers) increasing the overall labor productivity (this is the reason why was estimated the immigration e ect on the per worker gross value added). If immigrants increase the overall labor productivity, it will be easier for local rms to overcome the productivity threshold to internationalize the production by outwad FDI (Helpman, Melitz and Yeaple 2004). Finally, additional low wage workers may attract new foreign capital (inward FDI) looking for cost reduction (or simply for labor supply). The main di culty was to nd a good instrumental variable in order to solve the endogeneity problem that usually arises in this topic. The idea for the instrument used in this paper comes both from the network esternalities in immigration settlement (Card 2001) and from a gravity based equation for estimated immigrants in ows (Ortega and Peri 2009). The instrument was built by regressing the immigrants in ows with some regional characteristics and with the population using a su cient time lag with the economic perfomances. The t of this regression is a gravity-push estimated immigrantion without any economic determinants. This kind of instrument is very robust and signi cant and allows to estimate the economic e ects of immigration. In particular, immigration ows lead to better economic performances in terms of per capita GDP and per worker gross value added but it has a negligible e ect on growth in GDP and gross value added (annual percentage change). Moreover, immigration has a robust and positive e ects on both inward and outward FDI. This suggests that immigration provides additional low wage workers attracting new FDI from abroad; but also immigration by increasing the labor productivity (per worker gross value added), boosts the rm s competitiveness pushing rms to internationalize the production abroad. These results are mainly coherent with results in Ortega and Peri 2009 for OECD countries, the di erence lies on the instrumental variable and on the fact that this paper tries to avoid internal migration problems in estimations by using regional level data (taking into account the heterogeneity 14

of regions). The weak point of the estimation lies on the immigrants in ows variable because, as shown in section 3.1, it is approximated by assuming unchanged distribution among regions of new immigration in ows at country level. Thus,one possible further step in this research is to use the "true" immigrants in ows at regional level. An other possible future step is to go in deep in the e ects of immigration on the internationalization of rms by using data on intermediates good trade. References [1] Aydemir A. G. and Borjas, G. (2007) "A Comparative Analysis of the Labor Market Impact of International Migration: Canada, Mexico and the United States", in Jornal of the European Economic Association 5, pp.663-708. [2] Altonji, J. and D.Card (2001) "The e ects of Immigration on the Labor Market Outcomes of Less-Skilled Natives", in Abowd J. and R.Freeman (eds.) "Immigration Trade and The Labor Market", Chicago: University of Chicago Press. [3] Bandyopadhyay S. and H. Wall (2007), Immigration and Outsourcing: A General Equilibrium Analysis, Federal Reserve Bank of St. Louis working paper series. [4] Barba Navaretti G., Bertola G. and A. Sembenelli (2008) O shoring and Immigrant Employment: Firm Level Theory and Evidence, CEPR Discussion paper n. 6743. [5] Barry F. (2002) FDI, Infrastructure and the Welfare E ects of Labour Migration, CEPR DP 3380 and The Manchester School, 70 (3). [6] Borjas, G. (1995) The Economic Bene ts from Immigration Journal of Economics Perspectives, vol.9 n.2. [7] Borjas, G. (1999) The Economic Analysis of Immigration, in Orley C. Ashenfelter and David Card, eds., Handbook of Labor Economics, Amsterdam: North Holland, pp. 1697 1760. 15

[8] Borjas, G. (2003) "The Labor Demand Curve is Downward Sloping: Re-examining the Impact of Immigration on the labor market", in Quarterly Journal of Economics, 118, pp.1335-1374. [9] Borjas, G. (2006) "The Impact of Immigration on the Labor Market", paper prepared for the IMF conference in Labor and Capital in Europe following enlargement. [10] Borjas, G., J. Grogger and G.H.Hanson (2008) "Imperfect Substitution between Immigrants and Native: a reappraisal", NBER working paper n.13887. [11] Card (2001) "Immigrants In ows, Native Out ows, and the Local Labor Market Impacts of Higher Immigration", Journal of Labor Economics, vol. 19 n.1, pp.22-64. [12] Card (2005) Is the new immigration really so bad? The economic Journal, 115. [13] El Yaman S., Kugler M. And H.Rapoport (2007) Migrations et investssements directs etrangers dns l espace europeen, in Revue Economique, vol.58, pp.725-33. [14] Friedberg R.H. and J.Hunt (1995) "The impact of immigrants on host country wages, employment and growth" in Journal of Economics Perspectives, vol.9, n.2. [15] Grossman J.B. (1982) "The Substitutability of Natives and Immigrants in Production", in The Review of Econmics and Statistics, vol.64, n. 4, [16] Hanson G.H. (2008) "The Economic Consequence of International Migration of Labor" NBER working paper n. 14490. [17] Helpman E., Melitz M.J. and S.R. Yeaple (2004) "Export Versus FDI with Heterogeneous Firms", in The American Economic Review, vol.94, n.1. [18] Javorcik B.S., Özden Ç., Spatareanu M. and C. Neagu (2006), Migrant Network and Foreign Direct Investment, Rutgers University Newark Working Papers n. 2006-003. [19] Kugler M. and H. Rapoport (2005), Skilled Emigration, Business Networks and Foreign Direct Investment, CESifo working paper. 1455. [20] Ortega F. (2008) "The short run e ects of a large immigration wave: Spain 1998-2008", mimeo. 16

[21] Ortega F. and G.Peri (2009) "The Causes and E ects of International Migration: evidence from OECD countries 1980-2005", NBER working paper 14833. [22] Ottaviano G.I and G. Peri (2005), "Rethinking the gains from immigration: theory and evidence from the US", CEPR discussion paper series, n.5226. [23] Ottaviano G.I and G. Peri (2008), "Immigration and national wages_clarifying the theory and empirics", mimeo. [24] Paserman, M.D. (208), "Do High Skill Immigrants Raise Productivity? Evidence from Israeli Manufacturing Firms 1990-1999", IZA discussion paper n. 3572. [25] Peri G. and C. Sparber (2008) "Task specialization, Immigration and wages" CreAM discussion paper 02/08, march 2008. 6 Tables and Figures 17

Figure 3: relation between per capita GDP from 2002 to 2005 and the stock of immigrants in 2001 per capita gdp 2002 in ln 9 9.5 10 10.5 11 per capita gdp 2003 in ln 9 9.5 10 10.5 11 6 8 10 12 14 stock of immigrants in ln from census 2001 ln per capita gdp 6 8 10 12 14 stock of immigrants in ln from census 2001 ln per capita gdp per capita gdp 2004 in ln 9 9.5 10 10.5 11 per capita gdp 2005 in ln 9 9.5 10 10.5 11 11.5 6 8 10 12 14 stock of immigrants in ln from census 2001 ln per capita gdp 6 8 10 12 14 stock of immigrants in ln from census 2001 ln per capita gdp Source: author on Eurostat 18

Figure 4: relation between per worker gross value added from 2002 to 2005 and the stock of immigrants in 2001 per worker gva 2002 in ln 2 2.5 3 3.5 4 4.5 per worker gva 2003 in ln 2 2.5 3 3.5 4 4.5 6 8 10 12 14 stock of immigrants in ln from census 2001 ln per worker gross value added 6 8 10 12 14 stock of immigrants in ln from census 2001 ln per worker gross value added per worker gva 2004 in ln 2 2.5 3 3.5 4 4.5 per worker gva 2005 in ln 2 2.5 3 3.5 4 4.5 6 8 10 12 14 stock of immigrants in ln from census 2001 ln per worker gross value added 6 8 10 12 14 stock of immigrants in ln from census 2001 ln per worker gross value added Source: author on Eurosat 19

Figure 5: relation between the number of inward FDI projects (yearly from 2003 to 2005 and overall 2003-2006) and the stock of immigrants in 2001 number of outward FDI 2003 in ln 2 0 2 4 6 number of outward FDI 2004 in ln 0 2 4 6 6 8 10 12 14 stock of immigrants in ln from census 2001 6 8 10 12 14 stock of immigrants in ln from census 2001 ln_out_fdi_tot ln_out_fdi_tot number of outward FDI 2005 in ln 0 2 4 6 6 8 10 12 14 stock of immigrants in ln from census 2001 total number of outward FDI from 2003 to 2006 in ln 0 2 4 6 8 6 8 10 12 14 stock of immigrants in ln from census 2001 ln_out_fdi_tot ln_out_fdi0306 Source: author on Eurostat and OCOMONITOR 20

Figure 6: relation between the number of outward FDI projects (yearly from 2003 to 2005 and overall 2003-2006) and the stock of immigrants in 2001 number of inward FDI 2003 in ln 0 1 2 3 4 5 number of inward FDI 2004 in ln 0 1 2 3 4 5 6 8 10 12 14 stock of immigrants in ln from census 2001 6 8 10 12 14 stock of immigrants in ln from census 2001 ln number of inward FDI projects ln number of inward FDI projects number of inward FDI 2005 in ln 0 2 4 6 total number of inward FDI from 2003 to 2006 in ln 0 2 4 6 8 6 8 10 12 14 stock of immigrants in ln from census 2001 6 8 10 12 14 stock of immigrants in ln from census 2001 ln number of inward FDI projects ln_in_fdi0306 Source: author on Eurostat and OCOMONITOR 21

Figure 7: relation between per capita GDP and the immigrants in ows from 2002 to 2005 per capita gdp 2002 in ln 9 9.5 10 10.5 11 per capita gdp 2003 in ln 9 9.5 10 10.5 11 4 6 8 10 12 flow of immigrants in ln 2002 ln per capita gdp 6 8 10 12 flow of immigrants in ln 2003 ln per capita gdp per capita gdp 2004 in ln 9 9.5 10 10.5 11 per capita gdp 2005 in ln 9 9.5 10 10.5 11 11.5 6 8 10 12 flow of immigrants in ln 2004 ln per capita gdp 6 8 10 12 flow of immigrants in ln 2005 ln per capita gdp Source: author on Eurostat 22

Figure 8: relation between per worker gross value added and immigrants in ows from 2002 to 2005 per worker gva 2002 in ln 2 2.5 3 3.5 4 4.5 per worker gva 2003 in ln 2 2.5 3 3.5 4 4.5 4 6 8 10 12 flow of immigrants in ln 2002 ln per worker gross value added 6 8 10 12 flow of immigrants in ln 2003 ln per worker gross value added per worker gva 2004 in ln 2 2.5 3 3.5 4 4.5 per worker gva 2005 in ln 2 2.5 3 3.5 4 4.5 6 8 10 12 flow of immigrants in ln 2004 ln per worker gross value added 6 8 10 12 flow of immigrants in ln 2005 ln per worker gross value added Source: author on Eurostat 23

Figure 9: relation between the number of inward FDI projects and immigrants in ows number of inward FDI 2003 in ln 0 1 2 3 4 5 number of inward FDI 2004 in ln 0 1 2 3 4 5 6 8 10 12 flow of immigrants in ln 2003 6 8 10 12 flow of immigrants in ln 2004 ln number of inward FDI projects ln number of inward FDI projects number of inward FDI 2005 in ln 0 2 4 6 6 8 10 12 flow of immigrants in ln 2005 total number of inward FDI from 2003 to 2006 in ln 0 2 4 6 8 6 8 10 12 mean flow of immigrants in ln from 2002 to 2006 ln number of inward FDI projects ln_in_fdi0306 Source: author on Eurostat and OCOMONITOR 24

Figure 10: relation between the number of outward FDI and the immigrants in ows number of outward FDI 2003 in ln 0 2 4 6 number of outward FDI 2004 in ln 0 2 4 6 6 8 10 12 flow of immigrants in ln 2003 6 8 10 12 flow of immigrants in ln 2004 ln_out_fdi_tot ln_out_fdi_tot number of outward FDI 2005 in ln 0 2 4 6 6 8 10 12 flow of immigrants in ln 2005 total number of outward FDI from 2003 to 2006 in ln 0 2 4 6 8 6 8 10 12 mean flow of immigrants in ln from 2002 to 2006 ln_out_fdi_tot ln_out_fdi0306 Source: author on Eurostat and OCOMONITOR 25

Figure 11: kernel density estmation of: (i) per capita GDP, (ii) per worker gross value added, (iii) inward FDI in number of projects, (iv) outward FDI in number of projects. The dashed lines stand for the low-intensive immigrants regions, the unbroker lines stand for high-intensive immigrants regions Density 0.5 1 1.5 2 Density 0.5 1 1.5 9 9.5 10 10.5 11 11.5 mean value of per capita gdp from 2002 to 2006 in ln Kernel density estimate kdensity ln_mean_pc_gdp 2 2.5 3 3.5 4 4.5 mean value of per worker gva from 2002 to 2006 in ln Kernel density estimate kdensity ln_mean_pw_gva Density 0.1.2.3.4.5 0 2 4 6 8 total number of inward fdi projects from 2003 to 2006 Kernel density estimate kdensity ln_in_fdi0306 Density 0.1.2.3 0 2 4 6 8 total number of outward fdi projects from 2003 to 2006 Kernel density estimate kdensity ln_out_fdi0306 Source: author on Eurostat and OCOMONITOR 26

Figure 12: relation between inward and outward overall (from 2003 to 2006) number of FDI projects total number of inward FDI from 2003 to 2006 in ln 0 2 4 6 0 2 4 6 8 total number of outward FDI from 2003 o 2006 in ln ln_in_fdi0306 Source: author on OCOMONITOR 27

Figure 13: lorenz curves for the concentration of population, per capita gdp, per worker gross vaue added, immigrants in ows, inward FDI and outward FDI Lorenz(pop) 0.2.4.6.8 1 Lorenz(pc_gdp) 0.2.4.6.8 1 0.2.4.6.8 1 Cumulative population proportion 0.2.4.6.8 1 Cumulative population proportion Lorenz(immi_flow) 0.2.4.6.8 1 Lorenz(gva_per_work) 0.2.4.6.8 1 0.2.4.6.8 1 Cumulative population proportion 0.2.4.6.8 1 Cumulative population proportion Lorenz(in_fdi_tot) 0.2.4.6.8 1 Lorenz(out_fdi_tot) 0.2.4.6.8 1 0.2.4.6.8 1 Cumulative population proportion 0.2.4.6.8 1 Cumulative population proportion Source: author on Eurostat and OCOMONITOR 28

Figure 14: relation between the instrument (gravity-push estimated immigration in ows) and the regional economic performances (per capita GDP) per capita gdp 2002 in ln 8 9 10 11 4 6 8 10 12 gravity push estimated flow of immigrants 2002 in ln per capita gdp 2003 in ln 8 9 10 11 4 6 8 10 12 gravity push estimated flow of immigrants 2003 in ln per capita gdp 2004 in ln 8 9 10 11 per capita gdp 2005 in ln 8 9 10 11 4 6 8 10 12 gravity push estimated flow of immigrants 2004 in ln 4 6 8 10 12 gravity push estimated flow of immigrants 2005 in ln Source: author on Eurostat 29

Figure 15: relation between the instrument (gravity-push estimated immigrants in ows) and the computed immigrants ows flow of immigrants 2002 in ln 4 6 8 10 12 flow of immigrants 2003 in ln 6 8 10 12 4 6 8 10 12 gravity push estimated flow of immigrants 2002 in ln 4 6 8 10 12 gravity push estimated flow of immigrants 2003 in ln flow of immigrants 2004 in ln 4 6 8 10 12 flow of immigrants 2005 in ln 4 6 8 10 12 4 6 8 10 12 gravity push estimated flow of immigrants 2004 in ln 4 6 8 10 12 gravity push estimated flow of immigrants 2005 in ln Source: author on Eurostat Figure 16: OLS estimation of the non-economic determinants of immigration ows (instrumental variable) Determinant of immigration flows D_border D_currency D_Eu27 Total population 1992 Basic EMU sample No poorest regions Coefficient 5087.84 12906.3 4963.2 (p value) 0.003 0.000 0.009 Coefficient 6503.340 4729.330 (p value) 0.000 0.007 Coefficient 10184.150 8951.740 (p value) 0.000 0.001 Coefficient 0.008 0.009 0.008 (p value) 0.000 0.000 0.000 R square 0.337 0.386 0.352 F stat 45.550 36.130 40.890 Observations 692 351 611 30

Figure 17: e ects of immigration on economic performances and FDI projects: OLS and IV estimations Dependent variable Basic (OLS) EMU sample (OLS) No poorest regions (OLS) Basic (IV) EMU sample (IV) No poorest regions (IV) ln per capita GDP ln per worker GVA Δ ln per capita GDP Δ ln per worker GVA ln number inward FDI ln number outward FDI Coefficient 0.196 0.017 0.046 0.247 0.02 0.301 (p value) 0.000 0.127 0.000 0.000 0.424 0.219 Coefficient 0.384 0.386 0.379 0.408 0.415 0.399 (p value) 0.000 0.000 0.000 0.000 0.000 0.000 Coefficient 0.005 0.003 0.002 0.005 0.001 0.003 (p value) 0.000 0.000 0.077 0.205 0.626 0.195 Coefficient 0.001 0.001 0.001 0.001 0.001 0.001 (p value) 0.000 0.000 0.000 0.000 0.000 0.000 Coefficient 0.232 0.217 0.237 0.239 0.218 0.244 (p value) 0.000 0.000 0.000 0.000 0.000 0.000 Coefficient 0.252 0.259 0.250 0.245 0.248 0.243 (p value) 0.000 0.000 0.000 0.000 0.000 0.000 Figure 18: rst stage estimation of the 2SLS Instrument: Predicted inflows of immigrants from gravity push equation ln per capita GDP ln per worker GVA Δ ln per capita GDP Δ ln per worker GVA ln number inward FDI ln number outward FDI Basic (IV) EMU sample (IV) No poorest regions (IV) Coefficient 0.463 0.48 0.566 (p value) 0.000 0.000 0.000 Coefficient 0.890 0.918 0.904 (p value) 0.000 0.000 0.000 Coefficient 0.447 0.487 0.577 (p value) 0.000 0.000 0.000 Coefficient 0.894 0.923 0.907 (p value) 0.000 0.000 0.000 Coefficient 0.892 0.915 0.903 (p value) 0.000 0.000 0.000 Coefficient 0.916 0.916 0.923 (p value) 0.000 0.000 0.000 31

Figure 19: results of the robusti ed Durbin-Wu-Hausman test for the endogeneity Dependent variable Basic EMU sample No poorest regions ln per capita GDP ln per worker GVA Δ ln per capita GDP Δ ln per worker GVA ln number inward FDI ln number outward FDI F stat F stat F stat F stat F stat F stat 1.29 5.84 18.00 27.52 0.04 5.31 0.01 1.60 34.07 5.08 5.76 1.48 10.01 8.44 65.31 13.50 0.12 0.24 (p value) (p value) (p value) (p value) (p value) (p value) 0.26 0.02 0.00 0.00 0.85 0.02 0.94 0.20 0.00 0.02 0.01 0.22 0.00 0.00 0.00 0.00 0.73 0.62 Figure 20: e ects of immigration on number of inward and outward FDI per sectors: OLS and IV estimations Dependent variable ln number inward FDI manufacturing ln number inward FDI services ln number outward FDI manufacturing ln number outward FDI services Basic (OLS) EMU sample (OLS) No poorest regions (OLS) Basic (IV) EMU sample (IV) No poorest regions (IV) Coefficient 0.180 0.178 0.186 0.186 0.117 0.192 (p value) 0.000 0.000 0.000 0.000 0.000 0.000 Coefficient 0.147 0.142 0.148 0.153 0.139 0.152 (p value) 0.000 0.000 0.000 0.000 0.000 0.000 Coefficient 0.194 0.210 0.192 0.193 0.207 0.191 (p value) 0.000 0.000 0.000 0.000 0.000 0.000 Coefficient 0.197 0.208 0.195 0.188 0.197 0.187 (p value) 0.000 0.000 0.000 0.000 0.000 0.000 32