The economic impact of east–west migration on the European Union

This study contributes to the literature on destination-country consequences of international migration, with investigations on the effects of immigration from new EU member states and Eastern Partnership countries on the economies of old EU member states during the years 1995–2010. Using a rich international migration dataset and an empirical model accounting for the endogeneity of migration flows, we find positive and significant effects of post-enlargement migration flows from new EU member states on old member states’ GDP, GDP per capita, and employment rate, and a negative effect on output per worker. We also find small, but statistically significant negative effects of migration from Eastern Partnership countries on receiving countries’ GDP, GDP per capita, employment rate, and capital stock, but a positive significant effect on capital-to-labor ratio. These results mark an economic success of the EU’s eastern enlargements and free movement of workers in an enlarged EU.

a factor of 2.5, over just 6 years. 5 As this large-scale policy experiment can certainly provide a number of interesting insights into the labor market effects of migration, quite naturally, a significant body of literature studying the repercussions of such migration flows mainly for the receiving but also the sending labor markets has emerged. 6 This literature has mainly looked at the effects on wages, employment and unemployment, and welfare take up in individual member states separately. Generally speaking, besides some local effects, the available evidence is that the receiving labor markets absorbed post-enlargement immigrants rather seamlessly with statistically or economically insignificant effects on labor market indicators.
This evidence may, however, mask broader consequences of post-enlargement mobility. Migration in general facilitates cross-border social and economic ties, leading to an increased mobility of ideas and technologies, capital, and goods and services and thus a better allocation of production factors and improved total factor productivity, as well as gains from trade. 7 Although inherently difficult to detect, such effects may significantly affect EU member states, and thus their measurement is important for the debate about EU's migration policy.
The aim of this study is to analyze the effects of recent east-west mobility on economic outcomes across the EU and in the EU as a whole. Using an empirical model accounting for the problem of endogeneity of migration flows, we look at a range of indicators, in particular at GDP per capita, employment rates, capital stock and total factor productivity (TFP). The analysis is based on a rich dataset of immigration flows and stocks of foreigners, which has been collected by writing to selected national statistical offices, in 42 destination countries from virtually all source countries from around the globe for the years 1980-2010. 8 We comparatively evaluate the effects of post-enlargement intra-EU mobility (after the 2004 and 2007 enlargements) as well as immigration from the Eastern Partnership (EaP) countries on a subsample consisting of EU destination countries. 9 The main contribution of this study is twofold. First, the massive postenlargement migration flows over a relatively short period of time offer a unique framework that is worth exploring in order to inform the academic debate about the broader economic effects of migration and migration policy. Second, a comparative analysis of the costs and benefits of mobility under various migration regimes is much needed, in light of the heated policy debates surrounding migration policy in the EU. This agenda has become ever more urgent in view of the EU's plans to upgrade its mobility framework within its Eastern Partnership program and an increased migration potential from some of the key source countries as a consequence of the recent events in EU's neighborhood including the Arab Spring events, the Syrian civil war of the 2010s, and the Ukrainian crisis that started in 2014.
The rest of the paper is organized as follows. Section 2 presents the theoretical and empirical literature relevant to our study. Section 3 briefly describes the novel international migration database and other variables important for our analyses and provides some descriptive statistics. Section 4 presents an empirical model on the impact of immigration on the destination country's economy, on which we base our analysis, and our identification strategy. We discuss the results of econometric analyses in Sect. 5. Finally, Sect. 6 concludes, providing a discussion for future steps in our research.

Literature review
The effects of immigration on receiving countries has been a much debated issue in economics for a long time. Early theoretical models on the effects of labor mobility considered immigration in an extended version of the traditional Solow-Swan model, where immigrants are assumed to increase a country's unskilled population, which ceteris paribus leads to a lower per capita income because of a reduction in capital. Benhabib (1996) relaxes the assumption of the Solow-Swan model that immigrants do not provide any capital, which leads to some economic gain from immigration in terms of per capita GDP. Borjas (1995) argues that immigrants increase labor endowment in receiving countries and the new internal equilibrium is then characterized by a lower national wage, higher employment and higher national income. The difference with respect to the initial equilibrium is the so called ''immigrants surplus'' (Borjas 1995). A study by Hanson (2008) analyzes welfare consequences of immigration by assuming heterogeneity of workers in terms of skills, and perfect substitutability between native and foreign-born workers. The author shows that when low-skilled workers are allowed to freely move between countries, there will be migration from low-wage countries to high-wage countries until the wages will equalize. In the receiving country, home-born unskilled workers lose while the native high-skilled workers win in terms of welfare (surplus). Thus, so far, the theory says that the effect of migration depends on the type and selectivity of immigrants. Besides substitutability or complementarity of immigrant and native labor, capital endowments play an important role: if the physical capital endowment provided by immigrants is lower than the average native capital endowment, the effect of immigration will be negative in terms of per capita GDP. From the empirical point of view, the question of immigration's economic impact is thus still open.
Most of the existing empirical papers examine the impact of immigration by focusing only on labor market implications and on one or only a few receiving countries (e.g. Aydemir and Borjas 2007;Borjas 2003;Ottaviano and Peri 2008;Manacorda et al. 2012). Angrist and Kugler (2003) use a panel of European countries and analyze the labor market effects of immigration. Related to this paper, Peri (2008) and Gonzalez and Ortega (2011) analyze the effects of immigration on employment, capital accumulation and productivity, respectively, across US states and Spanish regions. The literature on the aggregate effects of migration using cross-country panel analysis is very scant. From earlier contributions, Dolado et al. (1994) found a negative effect 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. Recently, the aggregate effects of immigration have been discussed by a number of studies by Giovanni Peri, and a general overview can be found in Peri (2016). For instance, Peri (2012) analyzes the effects of immigration on each input of production function and on total factor productivity (TFP) for U.S. states' economies. The author also discusses the potential endogeneity problem, which he solves by using the instrumental variable (IV) technique, with past settlement patterns of immigrants driven by proximity to the border as an instrument for gross migration rates. In particular he shows that an increasing immigration leads to: (1) zero crowding out of the employment of natives (2) an increasing TFP growth. Felbermayr, Hiller and Sala (2010) investigate the effect of immigrants (by using the stock of immigrants in a destination country) on per capita GDP in the host countries. Using an IV cross-sectional approach and controlling for institutional quality and trade and financial openness, they find a positive effect of immigration on per capita GDP: a 10% increase in the migrants stock leads to a 2.2% increase in per capita GDP. Similarly, Bellini et al. (2013) find that the share of foreigners in the total population has a positive effect on per capita GDP in EU destination regions.
Further, Peri (2007) argues that immigrants' and natives' skills are not perfectly substitutable, 10 which creates the incentive for natives to specialize in more skilled jobs (e.g. more intensive in communication and language tasks) 11 and let the immigrants to do the manual tasks (Peri and Sparber 2009). This finding is consistent with other immigration studies that show immigration does not crowd out natives, but in fact it has a positive effect on employment and investment (Ortega and Peri 2009;, while total factor productivity is increased by optimizing task specialization and by encouraging the adoption of unskilled-efficient technologies (Peri 2012).
In an earlier paper, Peri (2006) argues that although immigration increases employment for the natives with complementary skills, it has a negative effect on those with substitutable skills. Previous research also shows that immigrants are substitutes for work performed by migrants that came in earlier migration waves. In particular, using data from different countries and different econometric methods, they highlight that immigration increases the overall wages for natives in the host country, but reduces the wages of previous immigrants (Ottaviano and Peri 2012;D'Amuri et al. 2010;Docquier et al. 2013;Longhi et al. 2010). A recent study by Foged and Peri (2016), however, shows that even if immigrants may be imperfect substitutes to low-skilled workers, they still improve their labor market position. The reason is that, as a reaction to the migrant inflow, low-skilled native workers had moved to complementary job market areas and started to specialize in nonmanual skills. This leads to an increase in their wages and employment opportunities (Foged and Peri 2016). However, in contrast to the hypothesis of imperfect substitutability of immigrants and natives, Docquier et al. (2013) find that immigration increases wages. On average, it has a negative effect for highly educated workers (except for the US) and has a positive effect for the wages of lowskilled workers.
From other outcome variables, it is worth mentioning that immigration appears to have a positive effect on trade creation, by reducing the fixed costs of trade, through network effects, and stimulates the trade of differentiated products (Peri and Requena 2010), and on increasing foreign direct investment (Javorcik et al. 2011;Gormsen and Pytlikova 2012). The effect on services is also positive, in the sense that it decreases the prices for low-skilled services (e.g. gardening, house-cleaning), which benefits the natives (Longhi et al. 2010). Regarding the effects of immigration on education, some previous studies suggest that the increase in the number of foreign students has a negative effect on the education of natives, while it increases the knowledge creation for universities (Hanson 2008;Kato and Sparber 2013). Using a panel of EU member states, industries and skill-groups, Guzi et al. (2015), document that immigrants are more responsive to labor and skill shortages than natives, contributing to economic efficiency in the receiving countries. Kahanec and Zimmermann (2014), argue that immigration tends to reduce income inequality.
When it comes to the effects of post-enlargement migration on receiving countries, the consensus in the literature appears to be that there was very limited, if any, effect on wages or unemployment rates (see Kahanec andZimmermann 2010, 2016;Gilpin et al. 2006;Blanchflower et al. 2007;Lemos and Portes 2008). Doyle et al. (2006), Hughes (2007) and Barrett (2010) report that even in Ireland, with the highest relative inflows from the new member states, effects on the aggregate unemployment rate could not be detected, although some substitution might have occurred. Brenke et al. (2010) point at competition for low-skilled jobs between the immigrants from Central and Eastern European (CEE) states that entered the EU in 2004 (EU8), and immigrants from outside of Europe. Similarly, Blanchflower and Lawton (2010) report some substitution in low skilled sectors. Shadforth (2009) andBlanchflower et al. (2007) argue that it was the fear of unemployment that resulted in some wage moderation in the UK prior to the 2004 enlargement. Several authors, including Kahanec andZimmermann (2010, 2016), Kahanec et al. (2013), Giulietti et al. (2013), or Barrett (2010) have proposed some positive macroeconomic effects of post-enlargement mobility within the EU. The latter study, for example, argues that increased immigration from the new member states fueled the Irish economy and boosted its GNP growth during the boom preceding the Great Recession. However, empirical analyses using more general multi-country data to investigate this hypothesis are missing. Even less is known about the possible effects of immigration from EaP countries. This paper contributes to the literature by providing empirical estimates on the effects of immigration on total GDP and GDP per capita, aggregate employment, capital stock, productivity and, consequently, income per capita at the country level by focusing on the recent large immigration flows from Central and Eastern Europe to the EU15.

Data description
The dataset on international migration used for the analyses was collected by Mariola Pytlikova and encompasses information on bilateral flows and stocks of immigrants from 42 destination countries over the period 1980-2010. 12 The dataset had been gathered by requesting detailed information on migration inflows and foreign population stocks by source country from selected national statistical offices in 27 countries. For six other countries-Chile, Israel, Korea, Mexico, the Russian Federation and Turkey-the migration data comes from the OECD International Migration Database. For nine other destinations-Bulgaria, Croatia, Cyprus, Estonia, Latvia, Lithuania, Malta, Romania and Slovenia-the data is collected from Eurostat. For the purpose of our analysis, we have used data on foreign population stocks 13 and have focused on EU15 and EU27 as destination countries and the EU12 and EaP as sending countries, for a time period ranging from 1995 to 2010. 14 As concerns the number of observations on foreign population stocks across all EU27 destination countries, the data has become more comprehensive over time, and thus missing observations have become less of a problem in more recent years. Compared to other migration data sources, our data is more comprehensive, for most countries have annual information on current stock of migrants and had kept such records for a relatively long time-period. In our dataset, as in the other existing datasets, different countries use different definitions for an ''immigrant'' and draw their foreign population statistics from different sources. While some countries report country of birth definition, which is preferred in our data, other countries use the definition by citizenship or country of origin, which includes the second or third generations of immigrants, excluding the naturalized ones. In the process of data gathering, the definition of country of birth was prioritized whenever possible. The main reason was to avoid problems related to the naturalization of foreigners, which can range significantly across countries, and therefore relying on definition by citizenship would lead to measurement issues. Although for the vast majority of destinations in our data we use country of birth definitions, for some destinations we have only data by citizenship or country of origin, see Table 4 in Appendix 1 for an overview of definitions and sources of the foreign population data. This may induce some measurement issues; see Pedersen et al. (2008), Adserà et al. (2015) and Cai et al. (2016) for a discussion. Unfortunately, two important European destinations, Germany and the Netherlands, did not provide data by country of birth. 15 We acknowledge that using migration data by country of nationality for these two countries may induce some measurement issues, which we address in the analyses, at least partly, by including country fixed effects. The information on other economic and social factors for these countries has come mostly from the World Development Indicators (WDI) by the World Bank, and part from sources such as OECD, ILO, or IMF.

Descriptive statistics
Compared to other advanced economies, labor mobility is relatively low in the European Union. Gill and Raiser (2012) report that the annual interstate mobility of working-age population in the EU15 was about 1% before the 2004 enlargement. The corresponding rate for the US was 3%, Australia and Canada 2%, and even the Russian Federation exhibited only 1.7%. In southern Europe, mobility rates are even lower at about 0.5% annually, whereas countries like France, Ireland, the Netherlands or the UK report mobility rates of around 2% (Bonin et al. 2008).
Most migration in Europe happens among EU member states; inflows from Eastern Partnership (EaP) countries to the EU had been increasing before the onset of the Great Recession, but remain much below those from other source regions. Figure 1 describes migration flows into EU countries, by continent of source countries. As can be seen, the biggest migration flows come from Europe, followed by Asia and Africa. Figure 2 allows for a closer look at the migration flows from Europe. We divide the source countries of foreigners into the ''old'' EEA/EFTA18 countries, EaP countries and EU 2004 and EU 2007 entrants to the EU. Figure 2 shows that the highest numbers of immigrants come from the ''old'' EU/ EEA/EFTA18 source countries and their inflows are relatively stable over time, whereas the lowest immigration into EU27 destinations stems from the EaP source countries. Figure 2 also shows the consequences of happenings in the European history such as the break up of the former Soviet Union and the EU enlargements towards the East. The 1992 peak of migration from ''Other European source countries'' region corresponds to the development in migration surrounding the fall of the USSR, and also partly covers the flow of refugees from the former Yugoslavia following its ethnic conflicts starting in 1991. Also, one can observe a gradual but considerable increase in migration flows for the new EU 2004 entrants after the first wave of EU's eastern enlargement in 2004. Similarly, migration from Bulgaria and Romania increased sharply after the 2007 EU enlargement. The decline after 2008 for all countries most likely corresponded to the financial crisis, which began to affect Europe in that year.
Looking at the evolution of migration stocks by continents of origin, we may observe that migration trends follow the development of the migration flows closely. European countries provide the highest number of migrants, followed by Asia and Africa, see Fig. 3.  (2011) Empirica (2017) 44:407-434  415 Similarly as in the case of immigrant flows, we divided the foreign population stocks stemming from Europe into more detailed regions of origin (see Fig. 4). We observed that the highest number of migrants living in EU27 countries originally came from the ''old'' EU15 countries and Norway, Iceland and Switzerland (''old'' EEA/EFTA18), whereas foreigners stemming from the EaP countries have the lowest numbers. Still, it can be seen an upward trend, suggesting future increases in the stock of migrants from EaP countries.
Transitional arrangements applied differently across the EU towards citizens of new member states. This and other factors such as linguistic proximity or labor  (2011) market performance resulted in significant variation in terms of the intensity of migration flows across destination countries. As consequence, stocks of foreign population vary across EU destinations markedly. Whereas circa 2010 the main target countries for EU8 citizens were the UK and Germany, 16 relatively few of them lived in Malta, Bulgaria or Slovenia (see Table 1). Italy and Spain dominated as the most attractive destinations for the immigrants from Bulgaria and Romania (EU2), while the EU8 countries were mostly at the other end of the range. Migrants from EaP countries predominantly live in Italy, Germany, but also in Poland and the Czech Republic. Countries such as Malta, Finland, Slovenia and the Netherlands are the least popular destinations among the EaP migrants (see Table 1). We may observe that there was only a slight increase in the share of immigrants from the EaP countries in the EU destination, from 3.36 to 3.58% immigrants from the EaP in total immigration in 1995 and 2010, respectively. The effects of immigrant inflows very much depend on the skill composition of immigrant inflows. Although the data does not generally permit a detailed account of the variation in skill composition across destination countries, previous literature using micro-data indicates that migrants from the new EU member states appear to have been predominantly medium skilled, but with rather high proportions of high skilled individuals Brücker and Damelang 2009). Brücker and Damelang (2009) reported that the share of high skilled individuals was 27% among EU15 natives, 22% among EU8 immigrants, and 18% among EU2 immigrants. The corresponding figures for low-skilled migrants were 27, 17, and 29% respectively. Although EU8 migrants appear to be relatively skilled, we should note that many of them have worked in occupations below their level of formal education, which probably affected their impact on the labor market . As for the cross-country variation, Holland et al. (2011) report that Luxembourg, Denmark, Sweden, and Ireland exhibit the highest shares of highskilled workers from the new member states, whereas Portugal, Spain, Belgium, Netherlands, and Finland disproportionally attracted their lower-skilled colleagues. According to Kahanec (2012) migrants from the EaP countries appear to have been the least educated of the three immigrant groups considered in this study, and have been similarly pushed to downskill into lower skilled jobs. Table 2 provides baseline statistics for the key variables used in our analysis. We observe the well-known gap in GDP per capita between the EU15, EU8, and EU2 countries, with the EU15 being the most affluent and the EU2 the least affluent regions, with EU8 countries in between.
Similar patterns can be observed for output per worker. As for the employment rates, 17 the gaps seem to much less significant. The table also shows, that EU15 is the most capital-abundant region of the EU, with EU8 and especially EU2 workers being equipped with significantly lower capital. On the other hand, growth in 16 However, the numbers for Germany do not include so called ''Aussiedlers'', or in other words German resettlers or ethnic Germans, who moved in large numbers from CEECs, in particular from Poland, Czechoslovakia and Romania, to Germany during the nineties. Obviously, if the numbers were included, the share of EU8 and EU2 migrant stocks in Germany would be much higher. 17 The employment rates are calculated as 1-unemployment rate. Capital to labor ratio capital-to-labor ratio is the highest in the EU2, and the lowest in the EU15, with the EU8 in between. Figure 5 documents the evolution of GDP per capita during the period 1995-2010 for five major blocs of EU members: northern, southern and western EU15 members, EU10, and EU2. We observe significant gaps between the five blocs, but also convergence of EU10 new member states with respect to the EU15, growing gaps within the EU15, and growing but lagging EU2. The effect of the Great Recession after 2008 is clearly visible.
In Fig. 6 we visualize unemployment patterns across the same five blocs of countries. The variation across the blocs and over time is much larger than for GDP per capita. Western and Northern European blocs exhibit the lowest unemployment rates in the EU. Southern Europe, the EU10, and EU2 countries, on the other hand, exchange positions in the ranking of blocs by unemployment rate several times over the studied period of time. All blocs share the same pattern of decreasing unemployment rates before the Great Recession, and increasing unemployment during the Great Recession. One exception is the EU2, which went a period turbulent times and increasing unemployment in the late 2000s. The EU10 and Southern Europe exhibit the steepest increase in unemployment rate since 2008.

Methodology
To determine the effects of immigration from new EU member states and from Eastern Partnership Countries on the receiving EU economies, we depart from an aggregate production function framework in our analyses, similarly as in Peri (2012), Ottaviano and Peri (2012) and Docquier et al. (2013). In our analyses, we  1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008  investigate effects that immigration has on wages (as proxied by GDP per capita, PPP, given the usage of aggregate data) and economic growth rate, as well as on total employment, physical capital, total factor productivity and the capital to labor ratio. In other words, we are estimating the following set of models: where X represents one of the following: employment rate and labour force participation (to account for the labor input), capital services and capital to labor ratio (to account for the capital input), total factor productivity (calculated as the Solow residual), output per worker (as a proxy for the average wage) and output per capita.
To capture other factors determining the economic outcomes of our interest that cannot be attributed to the changes in stock of foreigners per population, we account for country-specific time-invariant characteristics, represented by the term t j , time fixed effects h t , as well as time fixed effects interacted with region dummies 18 in our main specifications, d r h t . Finally, e jt represents the robust error term clustered by country. The explanatory variable of our interest is foreign population stock S from particular regions of origin relative to the total population P in destination country j, s jt ¼ S jt P jt . Thus, the effects of immigration on the destination country economies are captured by coefficient c.  1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008  We hypothesize that foreign population can affect the aggregate production of the receiving country. In particular we expect that, first, immigrants increase the total labor supply and may at the same time either crowd-out some natives or attract them into employment (especially if they provide jobs complementary to those of natives and stimulate productivity and specialization, or enable natives to enter the labor market by providing household services). We therefore estimate immigration's total effect on employment, which combines their direct contribution and the effect on native employment. Second, we expect immigration to affect investment, as the marginal product of capital may be increased due to the increase in labour supply. In addition, depending on the skill composition of immigrants, the effect on capital accumulation and capital intensity can be positive, as highly educated immigrants may work in more capital-intensive sectors, or may use capitalcomplementary techniques. On the other hand low-skilled immigrants can have a negative effect on capital, or leave it unaffected. Thus, the impact on capital accumulation and capital intensity in the short and long run depends on the composition of immigrants. Finally, immigrants may either give rise to crowding out effects; given fixed factors of production (acting as substitutes), and/or they may add to the varieties of ideas and products in the receiving economy (acting as complements); depending on which effect prevails, this may result in a higher or lower total factor productivity.

Identification
A methodological problem that arises from the models described above is the problem of simultaneity or reverse causality. It may well be the case that immigration rates are influenced by the dependent variables (low employment, or low GDP triggering migration flows), and not the other way around. To deal with the potential endogeneity problems, we apply the instrumental variable (IV) technique in our analyses, in which identification of causal effects rests on the instrumental variable. To qualify as a good instrument, a variable has to meet two conditions. First, it must be uncorrelated with the error term of the structural model and, second, it must be correlated with the endogenous variable.
As an instrument, we use a predicted foreign population rates, which we construct from predicted stocks of migrants obtained based on using a model of determinants of bilateral migration. In our two-stage strategy, the first-stage model of migration determinants has the following form: where s ijt stands for the share of foreign population originating from country i and living in country j at time t. On the right hand side we include an interaction of origin country fixed effects and time dummies, k i h t , to account for any economic, demographic or social changes in origin countries in each year and a set of bilateral country-pair specific effects, d ij . Based on the model we predict foreign population stocks, which are then summed by each destination country and adjusted for the population size of each particular destination country. The resulting variable is used Empirica (2017) 44:407-434 423 as an instrument for the structural equation in the second stage. Hence, for our identification strategy, we assume that development in home countries represented by the interaction of the origin country dummies and time is uncorrelated with economic conditions in destination countries (i.e. with our dependent variables that we use in the second step), and at the same time those push factors represent strong predictors of international migration (Adserà et al. 2015;Palmer et al. 2015).

Results
The results of our analyses of the effect of immigration on the EU15 economies are presented in Table 3. We report each model estimated by the OLS method with country fixed effects (FE) and by the instrumental variable technique with country fixed effects (2SLS-FE), which accounts for possible endogeneity of migration flows. The rows correspond to models with the GDP per capita (to account for the average wage) and total GDP (both PPP adjusted), the employment rate and labor force participation (to account for the labor input), capital services and capital to labor ratio (to account for the capital input), and total factor productivity (calculated as the Solow residual), and output per worker (to account for productivity) as dependent variables. To account for possible differences across immigrant categories, as defined by their origins, we distinguish the results for foreigners stemming from the 2004 EU entrants, 2007 EU entrants, and EaP countries. A number of notable results emerge in Table 3. Whereas fixed-effects OLS models (FE) generally produce insignificant results, relatively small, but negative and statistically significant, effects on GDP, GDP per capita, capital-to-labor ratio, and output per worker emerge for immigration from the EaP countries. Due to possible endogeneity of migration flows, our preferred specification is the instrumental variable (2SLS-FE) model. Here, we believe that the variation in the expected stocks of migrants from EU 2004, EU 2007 and EaP countries is reasonably exogenous conditional on origin country fixed effects and time dummies, and bilateral country-pair specific effects as noted in Eq. (2) above. In Table 3 we report the weak-identification Kleibergen and Paap (2006) F-statistic. As we can see from Table 1, the value of F-statistics exceeds the Stock and Yogo (2005) weakidentification critical value of 16.38 (for 10% maximal size distortion) for immigration from EU 2004 entrants and from EaP countries, whereas it is lower for immigration from EU 2007 entrants. This suggests that our instruments are strong for model specifications for immigration from EU 2004 and EaP group of countries, whereas the instrument is weak for effects of immigration from EU 2006 entrants. 19 In 2SLS-FE regressions, we observe a statistically significant positive effect of immigration from the new EU countries on GDP and GDP per capita in the EU15   Period of analyses: 1995-2010 Each cell shows the coefficient from a different regression with the dependent variable described in the first cell of the row and the explanatory variable equal to the total flow of immigrants as a share of the initial population of the receiving country. All regressions includes year, country fixed effects and interaction of region dummy and time. Robust standard errors clustered by country are in parentheses. The 2SLS estimation method uses the predicted flow of immigrants from the gravity push factors as instruments, in particular we use (xi: xtreg lnflowstocks i. from*i. year, fe) model: ln(sijt) = a ? b(country FE*year) ? v(country FE); the predicted share of foreign population per destination population are then summed on the destination country level and used as an IV. The Number of observations, The LM-and Wald F-statistics at the bottom of the table corresponds to the number of observations for each regression-they are the same for all regressions for particular groups of countries ***, **, * Imply significance at the 1, 5 and 10% level destination countries, whereas the coefficient for immigrants coming from EaP countries is negative. The estimated effect on GDP per capita is quite large as the coefficients imply that a 10% increase in the number of immigrants coming from the 2004 and 2007 EU member countries per destination population increases the destination's GDP per capita by 0.3 and 0.55%, respectively. In contrast, a 10% increase in share of immigrants coming from the EaP lowers GDP per capita in the EU15 countries by 0.13%. Whereas in the FE regressions, there is some evidence that an increase in the shares of foreigners from new EU member states increases labor force participation (at a 10% level of significance), in the 2SLS-FE regressions, the coefficients are no longer significant. The positive effect of immigration from new member states on the employment rates is documented in the 2SLS-FE regressions; however, a small, but negative and statistically significant, coefficient emerges for immigrants from EaP countries.
No statistically significant results emerge in the 2SLS-FE models for the effects on total factor productivity. The same applies to the impacts on capital stock and the capital-to-labor ratio for immigration from the new EU member states; however, for immigrants from the EaP countries, a small negative effect on capital stock and a positive impact on the capital-to-labor ratio emerge as statistically significant. Interestingly, the latter result contradicts the one found in the FE model, indicating that countries with increasing capital-to-labor ratio might be substituting capital for immigrant labor from the EaP countries. Finally, negative effects on output per worker are found for immigrants from new EU member states, but the corresponding results for those from EaP countries are insignificant. 20

Conclusions
In this study we contribute to the literature on destination-country consequences of international migration. In particular we look at the effects of immigration from the new EU member states and Eastern Partnership countries on the EUseparately for old EU member states (EU15) and on the EU as a whole (EU27)between the years 1995 and 2010. Taking into account possible reverse causality from economic indicators to migration flows, our results show positive and significant effects of post-enlargement migration flows from the new EU member states on GDP, GDP per capita, and employment rate and a negative effect on output per worker. Regarding immigration from EaP countries, we find small but statistically significant negative effects on GDP, GDP per capita, employment rate, and capital stock, but a positive significant effect on capital-to-labor ratio, in EU countries. 20 We run similar analyses using immigration to the EU27 countries. It turns out that the results are generally very similar to those estimated for the EU15 countries, except that the coefficients are, as a rule, estimated less precisely. This indicates that the results we observe are primarily driven by the EU15 countries. This is not surprising, given that immigration to the EU15 is considerably larger and has a longer history than migration flows to the rest of the EU. The results are available in Table 5 in Appendix 2. Empirica (2017) 44:407-434 427 Our results for intra-EU mobility are in line with the previous literature; complementing it by showing that the generally neutral-to-positive positive effects found at the micro level, or at various levels of aggregation, also show up at the macro, EU-wide, level, and for a number of, but not all, economic indicators. On the other hand, small negative effects are found for immigration from EaP origins. Further research is needed to better understand why EaP immigration differs from mobility from new EU member states. Besides the possibility that this difference emerges due to different composition of immigrant inflows from the two clusters of origins, an alternative hypothesis is that it is an artifact of the different legal status of immigrants from new EU member states and those from EaP countries. One plausible explanation is that free labor mobility contributes to the positive effects of intra-EU migration on the receiving countries by enabling immigrants to allocate and integrate more efficiently. As a corollary, it may well be that legal barriers to immigration from the EaP and their integration hamper positive economic effects of their immigration.
These findings underscore the positive economic effects of intra-EU mobility as a pillar of economic efficiency of the single market in the EU, and provide an economic argument for eliminating, or at least reducing, barriers to labor mobility and immigrant integration. They also highlight the unfortunate gap between what hard data show about labor market impacts of migration, on the one hand and public perceptions and beliefs about free mobility in the EU on the other hand, as also demonstrated by the public debates surrounding Brexit.
Appendix 2: Impacts of foreign population in the EU27 See Table 5.   [1995][1996][1997][1998][1999][2000][2001][2002][2003][2004][2005][2006][2007][2008][2009][2010] Each cell shows the coefficient from a separate regression with the dependent variable described in the first cell of the row and the explanatory variable equal to the total flow of immigrants as a share of the initial population of the receiving country. All regressions includes year, country fixed effects and interaction of region dummy and time. Robust standard errors clustered by country are in parentheses. The 2SLS estimation method uses the predicted flow of immigrants from the gravity push factors as instruments, in particular we use (xi: xtreg lnstocksperpop i. from*i. year, fe) model: ln(sijt) = a ? b(country FE*year) ? v(country FE); the predicted share of foreign population per destination population are then summed on the destination country level and used as an IV ***, **, * Imply significance at the 1, 5 and 10% level Empirica (2017) 44:407-434 431