Brain Drain and Firm Productivity: Evidence from the Sequential Opening of EU Labour Markets a (Preliminary draft)

Brain Drain and Firm Productivity: Evidence from the Sequential Opening of EU Labour Markets a (Preliminary draft) Yvonne Giesing b, Nadzeya Laurentsyeva c Abstract This paper establishes a causal link between brain drain and firm performance by exploiting changes in EU labour mobility legislation from 2004 to 2014. These changes induced higher emigration rates from new member states, thus lowering human capital supply in these countries. The industry, country and time variation in the opening of EU labour markets allows us to construct an instrument for the brain drain. Using firm-level panel data, we show that exposure to higher emigration rates has reduced firm total factor productivity in new member countries, and thus could have contributed to the persistence of the productivity gap between old and new EU member states. JEL classification: O15, D24, F22, J24 Keywords: Brain Drain, Productivity, Human Capital, Migration, EU Enlargement a We would like to thank Carsten Eckel, Florian Englmaier, Lisandra Flach, David A. Jaeger, Monika Schnitzer, Davide Suverato and participants of COPE 2015, GEP/CEPR Postgraduate Conference 2015 and Munich seminars for helpful comments and suggestions. We would like to thank Vanya Bodurska for excellent research assistance. Financial support from the DFG Research Training Group 1928: Microeconomic Determinants of Labour Productivity is gratefully acknowledged. b University of Munich - Contact: Yvonne.Giesing@econ.lmu.de c University of Munich - Contact: Nadzeya.Laurentsyeva@econ.lmu.de / Ludwigstr. 33, Room 439, 80539 Munich / Tel: +49 (0)89 / 2180 2943 1

1 Introduction In 2013, 232 million people were international migrants, representing 3.2 % of the world population. 72.4 million migrants lived in Europe, half of them having immigrated from other European countries (UN 2013). Migration outflows have especially increased for Central and Eastern European countries after they joined the EU. Average annual labour migration from new EU member countries to old EU member states doubled in the four years after EU accession, compared with the four years before, reaching a record high of 708 000 migrants in 2007. Young and highly educated workers were most likely to migrate (OECD 2012). For example, in 2010/11, 18.4 % of all Romanian and 15.5 % of all Polish citizens with tertiary education resided in other OECD countries. To compare, total migrant stocks from Romania and Poland constituted respectively 12.7 and 8.9 % of these countries populations (OECD 2013b). Although there are important positive consequences of free labour mobility in terms of lower unemployment, better skill match, and faster wage convergence between states, there have been growing concerns that the emigration of highly qualified individuals has created a severe challenge for source countries. The aim of this paper is to establish a causal link between brain drain and firm performance. We exploit changes in EU labour mobility legislation from 2004 to 2014 and consequent increases in emigration rates from new EU member states as a shock to human capital supply in these countries. We find that firms in industries subject to higher labour mobility experience a drop in total factor productivity and an increase in their personnel costs. We further show that the EU opening has created binding human capital constraints 1 for some firms in new member countries. An emigration-driven one standard deviation increase in skill shortages leads to a 6.1 % drop in firm total factor productivity and a 5.2 % increase in personnel costs. Panel firm-level data allows us to account for firm heterogeneity and explore the link between firms characteristics and their sensitivity and adaptation to brain drain. We find that innovating and foreign-owned firms increase personnel costs by more and their productivity is less affected by brain drain. To our knowledge, no paper has so far directly evaluated emigration effects on firm performance, nor made use of legislation variation at the industry level. Furthermore, while previous papers have mostly provided single country analysis, we are able to exploit cross-country variation due to the availability of harmonized data. 1 In the text and regression output, we use the notions skill (labour) shortages and human capital constraints as synonyms. 2

Following Docquier and Rapoport (2008), we define brain drain as the emigration of high-skilled individuals (typically, university graduates above the age of 25) mainly from developing to developed countries. Research on brain drain has started in the 1960s and has since then revolved around the question if brain drain is necessarily hindering a source country s development or if there can be positive effects due to remittances, return migration, incentives for education and trade, FDI and technology spillovers. 2 Freeman (2006) summarizes these contrasting effects and states that the empirical analysis is inconclusive about the sign of the effect. Docquier and Rapoport (2012) provide an insightful overview over the last 40 years of brain drain literature. They conclude that most macroeconomic studies do not identify the causal effect of brain drain on a country s development and that microeconomic analysis should exploit existing natural and policy experiments to enrich the debate. 3 The transitional provisions on labour immigration laws of old EU members provide us with such natural experiment to establish causality and to analyse the consequences of brain drain on source countries. Our microeconomic approach will focus on the firm as the unit of analysis. Similarly to the trade literature that benefited greatly by the introduction of the firm (Melitz 2003), we expect that the migration literature can gain richer insights into the consequences of brain drain by investigating firm performance. Accounting for firm heterogeneity, firm entry and exit is important if one intends to shed light on microeconomic mechanisms, which shape the observable effect of migration on macro outcomes. William Kerr and co-authors, for instance, are encouraging this approach for the analysis of immigration (Kerr et al. (2014), Kerr et al. (2013), Kerr (2013)). One of the reasons why, until recently, the brain drain literature has not looked directly at firm outcomes, has been the lack of representative firm-level datasets. In the present paper, we are able to use harmonized balance-sheet and profit-loss panel data, covering up to 80% of all firms in the new EU member states. We combine this information with the firmlevel BEEPS survey, conducted by the European Bank of Reconstruction and Development (EBRD) throughout 2002-2013. Another major challenge for the migration studies has been the endogeneity problem. Migration outflows respond to negative productivity shocks, resulting in reverse causality problem. Furthermore, omitted variable bias may 2 Research on international migration is still scarce compared to similar disciplines, such as international trade (13 times more frequent in all published article abstracts in Research Papers in Economics (RePEc). Moreover, research on immigration is four times as frequent as research on emigration (Clemens 2011). 3 Clemens (2013) analyses the first random experiment of the income gains from migration for Indian H1-B visa lottery winners. 3

hinder from identifying the effect of interest. The least productive firms may be more likely to lose their qualified people, for instance, due to poor management. In both cases, simple OLS would indicate a negative correlation between emigration rates and firm productivity but without causality. To circumvent this endogeneity problem, we exploit an exogenous shock to brain drain. Following the EU enlargements in 2004 and 2007, labour emigration from new EU member countries has increased significantly. Yearly emigration rates of up to 5 % (OECD 2013a) of the labour force decreased the labour supply in the source countries. Importantly, the citizens of new member countries did not receive universal labour mobility within the EU immediately upon accession, as old EU member countries could apply transitional provisions. In particular, Germany and Austria restricted labour market access for EU8 4 and EU2 5 citizens for up to seven years. France, Italy and Spain opened up gradually, providing easier labour market access for migrants in specific sectors, while Sweden, Ireland and the UK opened their labour markets directly at the moment of accession. In addition, the reaction of EU8 and EU2 citizens to labour market reforms in old EU member states 6 was influenced by bilateral distance, language proximity, and historic ties as well as employment prospects that varied across industries of the destination countries. Hence, we are able to exploit the EU labour market opening as a quasi-natural experiment and construct an instrument of brain drain, which varies on the country, industry, and time level. Our instrumental variable approach is similar to the one adopted by Llull (2014). This paper looks at the effect of immigration on wages and uses interaction of push factors, bilateral distances, and skill cell dummies to control for endogenous allocation of immigrants across skill groups in the U.S. and Canada. There are several studies that have used changes in labour mobility laws in the EU as exogenous variation to labour supply. Dustmann et al. (2012) and Elsner (2013) estimated the effects of emigration on wages in Poland and Lithuania respectively and found that wages increased for the stayers. 7 Mayr 4 Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovakia, Slovenia. Malta and Cyprus also joined in 2004 but are excluded from the analysis because transitional agreements were not applied to them. 5 Bulgaria, Romania. 6 Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherlands, Portugal, Spain, Sweden, United Kingdom - in the text we also refer to these countries as EU15. Other EEA members: Iceland, Liechtenstein, Norway, and Switzerland also applied transitional provisions; however, due to missing data, we exclude them for the moment from our analysis. 7 Mishra (2014) provides a global overview of studies analysing the effects of emigration on the wages of stayers and finds that estimates range from positive and significant 2-5.5 % following a 10 % emigrant supply shock. 4

and Peri (2009) develop a model to study the consequences of European free labour mobility on human capital in the source countries and differentiate between brain drain and brain gain due to return migration and increased incentives to invest in education. However, the literature focusing on source countries and emigration so far has not looked at firm performance outcomes. In destination countries, the consequences of labour mobility for firms have very recently started to become of research interest. Peri (2012), Kerr and Kerr (2013), Paserman (2013) and Mitaritonna et al. (2014) study the effects of immigration on firm productivity in the US, Israel, and France. They find that increase in the supply of foreign-born workers positively affects firm productivity due to faster growth of capital and the specialisation of natives in more complex tasks. Lewis (2013) furthermore finds that besides increased investment, firms also adapt technology due to higher immigration levels. Using firm-level German data, Dustmann and Glitz (2015) analyse how industries and firms respond to changes in the local labour supply. They find that immigration alters the local skill composition and investigate three adaptation mechanisms: change in factor prices, within-firm change in skill intensity, and adjustment through entry and exit of firms. This research can be seen as complementary to what we aim to achieve. While the authors look at the effects of immigration on firms, we focus on the consequences of emigration. We complement the analysis with data on skill shortages (human capital constraints) as reported by firms in EU8 and EU2 countries. Variation in emigration rates created by the EU opening predicts changes in shortages of qualified workforce, thus, conforming the presence of the brain drain problem. In this way, our paper also contributes to the literature relating to skill shortages and firm productivity 8. The paper is organized as follows. The next section provides some background information on the EU opening, its free labour mobility and the resulting migration flows, which helps to understand the mechanism underlying our instrument. Section 3 then discusses skill shortages as a potential channel that link brain drain and firms productivity. Section 4 describes the data, followed by section 5 that shows the empirical specification. Section 6 8 Haskel and Martin (1993) were, to our knowledge, the first to specifically analyse the effect of skill shortages on firm productivity. Using UK survey data of 81 industries from 1980-1986, they found that a 1% increase in labour shortages was associated with a 1.8%-2.1% decrease in aggregate industry productivity. Adding to this, Wallis (2002) found that for every percentage point increase in skill shortages, wages increased by 0.09 percentage points. More recent evidence from Europe by Brixiova et al. (2009) shows that skill misallocation across sectors and resulting skill shortages reduce the attractiveness of entry for new firms. 5

illustrates the results, while section 7 provides robustness checks. Section 8 concludes. 2 Transitional Arrangements for the free movement of workers during EU Enlargement Our paper exploits the gradual opening of the EU labour market to the citizens from the new member states (NMS) as an exogenous shock to labour supply due to emigration. This section provides background information on the details of the EU enlargement to explain how different countries allowed for free labour mobility in particular sectors at different times. In 2004, ten Eastern and Southern European Countries joined the 15 old member states of the European Union (EU15): Cyprus, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Malta, Poland, Slovakia and Slovenia. While free capital mobility was introduced instantaneously by all countries, free labour mobility was initially restricted as some EU15 countries feared a tremendous inflow of cheap labour. The EU Commission thus allowed the old member states to unilaterally restrict their labour markets by national laws for a period of seven years. These transitional arrangements were used against all new members (EU8) in the same way but excluded Malta and Cyprus. In 2007, Bulgaria and Romania (EU2) joined the European Union, facing the same transitional agreement rules. The option to unilaterally restrict labour markets generated many different rules within the EU. While Ireland, Sweden, and UK decided to open their labour markets immediately in 2004 without sector restrictions, other countries delayed the access or applied special job schemes in certain sectors. Denmark, Greece, Spain and Portugal, for instance, removed restrictions only in 2009. France, Belgium, Netherlands and Austria opened their labour markets gradually, allowing only workers from certain sectors and introducing quotas. Germany kept the labour market almost completely closed until the expiration of the transitional agreements (2011/2014). Table 8 in the annex provides an overview of precise opening dates and sectoral details per country. This sequential opening by country, year and sector had a significant effect on migration rates. Constant (2011) and Kahanec (2012) provide descriptive evidence of EU migration flows following the enlargement and the transitional agreements. They show that the transitional agreements influenced the migrants in their movement. The UK and Ireland, for example, have become the main EU destination country for Polish, Slovakian and Latvian 6

workers. Kahanec et al. (2014) applying difference-in-differences analysis confirm that outward migration from NMS increased with the EU entry, but its full potential was hampered by the presence of transitional arrangements. One might argue that the restriction of a country s labour market is endogenous and is related to local labour market conditions. Germany, for instance, was experiencing high unemployment rates during the mid-2000s and this was one of the reasons for its labour market restrictions. However, while the transitional arrangements are endogenous to labour market conditions and firm productivity in the respective receiving country, they are exogenous to firm outcomes in the new member (sending) states. Another argument against our identification is that the decisions to open a particular sector by EU15 countries were to some extent endogenous to conditions in the new member countries. For example, mobility restrictions might have been directed at the NMS citizens working in countries and sectors with high volumes of EU15 FDIs. However, we are convinced that this is not the case for the following reasons. First, EU15 could not differentiate transitional provisions across countries in EU8 and EU2 groups. Second, this proposition is hard to reconcile with significant time-variation in the removal of provisions. One might further suggest that the sector-specific timing of labour market openings coincided with trade liberalization. Yet, all new EU member countries had signed and enforced Free Trade Agreements with the EU prior to their accession. Hence, it is plausible to conclude that the application of transitional provisions by EU15 was driven mainly by their own economic conditions and is thus exogenous to firm outcomes in NMS. 3 Channels This section sheds some light on three potential channels: vacancy duration, change in technology, and training costs, which could explain how emigration might affect firms productivity. Firms in industries that experience higher emigration rates are unable to fill their vacancies in due time. Thus, capacity utilization cannot be optimised and production potential lies idle. This channel can theoretically be derived from the literature on search and matching models pioneered by Mortensen and Pissarides. In this setting, high emigration rates would reduce the matching probability as the pool of suitable workers is reduced. A lower matching efficiency ceteris paribus causes longer vacancy duration and thus lower productivity. Further, firms in industries that face high emigration rates adapt their technology, which in turn affects their productivity. Here, a change in tech- 7

nology includes several adaptation mechanisms, for instance a change in relative input quantities. Consistent with the international trade literature on relative comparative advantages, an exogenous reduction in human capital supply might cause firms to reduce their labour input and instead focus on a capital intensive production technology. Another mechanisms might be a reduced innovation capacity and a reduced probability to adopt the newest and most productive technology. Firms that face human capital shortages might not be able to innovate enough as they lack the required labour force. This consequently affects productivity in a negative way. A third channel can be increased training costs. If firms cannot find adequately educated workers to fill their vacancies, they might hire less skilled or lower educated candidates, but alternatively provide more on-the-job training in the required skills. This will result in increased training costs in the short run. 4 Data For our analysis we use firm-level financial and survey data, aggregate industry and country indicators, detailed migration data, and information from EU labour legislation. We obtain firm-level data from Bureau Van Dijk s AMADEUS database that provides standardised annual balance-sheet and profit information for European public and private companies. We work with an unbalanced panel of about 110 000 firms located in the new member states. The period covered ranges from 2000 to 2013, and, on average, there are 5 annual observations for each firm. The sample includes companies in manufacturing, construction, retail trade and services. Apart from financial reports, the dataset provides information on firms patenting activity, ownership structure, export markets, and exit status (such as bankruptcy or liquidation). We complement this data with firm-level information from the Business Environment and Enterprise Performance Surveys (BEEPS) administered by the European Bank for Reconstruction and Development (EBRD) in all new member states. The survey was conducted in 2002, 2005, 2009 and 2012 and contains an extensive questionnaire on firms financial performance (selfreported), workforce composition, management practices, innovation, and perceptions of the business environment (including availability and quality of human capital, financial constraints, and corruption). The survey data provides a representative sample of manufacturing, construction, services, and retail trade firms. In total, there are 13 972 firm-year observations, of 8

which 2556 (1293 firms) make up an unbalanced panel. Migration data is taken from Eurostat Labour Force Surveys, which take place annually in all EU member countries and cover about 5% of national populations. The survey provides demographic information on individuals, including their country of birth, education level, occupation, and employing industry. To construct the instrumental variable, we use the Labour Reforms database (section on labour mobility) of the EU Commission and complement it with information from the national legislations of the old EU member states. A measure of human capital constraints (HCCs) is taken from the EU Commission Business Survey, which is conducted quarterly in all EU member countries by the Directorate General for Economic and Financial Affairs (DG ECFIN). The survey addresses representatives of the manufacturing, services, retail trade, and construction sectors and asks for firms assessment and expectations of the business development. Among other questions, the survey s participants are asked to evaluate factors limiting their production (such as labour, access to finance, demand, and equipment). The EU commission publishes information on a two-digit industry level, thus the obtained HCCs measure is equal to the share of firms in a given industry reporting to be constrained by labour. To match the data to other datasets, we aggregated quarterly indicators to annual level, As an alternative HCCs measure, we consider firms replies from the BEEPS survey, which asks respondents to evaluate importance of inadequately educated labour as an obstacle for business. Following the EU Commission Survey, we also aggregate individual firms replies on a two-digit industry level. As additional covariates, we use aggregated (two- and four-digit) industry level data, which is available for all EU member states and is harmonized by Eurostat. The structural business statistics database contains annual information on industries performance, including output, investment, employment, and personnel costs. Macroeconomic controls (GDP, FDI, unemployment, interest rates) are taken from the Worldbank statistical database. 5 Econometric Specification The aim of the empirical analysis is to determine how EU2 and EU8 firm performance has changed in response to industry-specific negative shocks to human capital supply (brain drain). For identification we exploit legislation changes as exogenous shocks to human capital supply and control for other possible channels, which could have simultaneously affected firm outcomes and industry-specific emigration rates. 9

5.1 Baseline Model The general empirical specification is represented below: Y fict = β 1 EM ict + β 2 X fict + β 3 I ict + β 4 C ct + φ t + ν fic + ɛ fict (1) where Y fict are different outcomes of a firm (f) in industry (i), country (c) and year (t). EM ict are annual emigration rates, specific to the citizens of a country (c) qualified to work in an industry (i), proxy measure of the brain drain. X fict is a set of firm time-varying controls. I ict includes controls (such as total investment) on a two-digit industry level. It accounts for variation in emigration rates and firm outcomes, which could have been caused by labour demand or technology shifts within an industry of a particular country. C ct is a vector of macroeconomic covariates, controlling for country-wide changes: GDP growth rate, FDI, real interest rates. φ t are time dummies. ν fic represents firm fixed effects, and ɛ fict is an error term. In the baseline empirical model, we consider only within-firm variation in outcomes over shifts in emigration rates. Such a specification allows us to take care of firm unobserved time-invariant heterogeneity (as initial management ability or quality of business ideas) and other constant characteristics of firm location or industry-specific production technology. Firm productivity is the main focus of our analysis. We estimate the effect of brain drain on labour productivity as well as total factor productivity (TFP) of firms. The details on the TFP calculation are provided in the Appendix (9.3). As discussed in section 3, an exogenous negative shock to labour supply can result in a productivity drop (at least, in the short-term) due to longer vacancy duration, higher personnel costs (both in monetary and time terms), lower productivity of available workers and consequent underinvestment. At the same time, firms could adjust technology and capital to compensate for labour shortages. Unequal possibilities to bear these adjustment costs could widen the gap between more and less productive firms. Therefore, one may expect heterogeneous effects of brain drain depending on firm characteristics. To account for this, we estimate the model for several sub-samples of firms. We look separately at the outcomes of foreign-owned firms and patenting (innovating) firms, which are likely to be more productive. In addition, we estimate the brain drain effect for a sample of incumbent enterprises, which existed in EU8 and EU2 countries prior to the EU accession (we set the threshold at 2002). Hence, we exclude new firms that made their entry decisions after the labour market conditions had changed and therefore could come from a different productivity distribution. We look at other outcomes to shed some light on the mechanisms that transmit the effect of brain drain to productivity, in particular: vacancy 10

duration, training expenses and technology adjustment. Although vacancy duration is not directly observable in our data, lower firm profitability and higher personnel costs could serve as indirect evidence. In addition, in the BEEPS data we can observe if firms train their employees. Adjustment of technology could be reflected in the change of firms capital intensity and innovation activities. Decreases in innovation capacity or outsourcing of innovation abroad could be likely outcomes of the brain drain problem. Reducing innovation expenditures could also be an optimal response of firms, which face a reduced supply of qualified labour and hence higher hiring and training costs. 5.2 Instrumental Variable Even in the presence of fixed effects and a number of industry- and countryspecific covariates, there might be unobservables (such as industry-countryyear demand, technology shocks or policy changes), which at the same time alter emigration rates and firm productivity. 9 To resolve the endogeneity problem and to ensure that our explanatory variable of interest is exogenous, we exploit a natural experiment setting, which was created by the gradual opening of the EU labour markets to the citizens of the new member states (for details, see section 2). We construct an instrument for the emigration rates of NMS citizens, qualified to work in a particular industry. The main identifying assumptions are 1) the IV is a good predictor of emigration rates, but is otherwise exogenous to the productivity shocks of firms, once we control for various fixed effects and observables; 2) there are no other channels, through which the IV impacts productivity other than through increasing emigration. In addition, we assume that migrating workers look for a new job in the same sector as their previous employment. An industry-year-country cell makes up one observation. Industries are represented at the NACE two-digit level. The main period under consideration is from 2004 to 2014 (from the accession of EU8 countries to the termination of all transitional provisions applied to all new member states) 10. First, for each observation we construct a set of 15 dummies, with each dummy corresponding to one of the 15 old EU member countries. A dummy takes the value of 1 if according to the legislation of an old EU member, its corre- 9 Reverse causality could be another issue; however it is less problematic in our setting, since it is unlikely that firm-specific productivity shocks systematically influenced emigration rates, which are measured on the aggregate industry level. 10 For robustness checks and better estimates of the fixed effects, we also include the data from 2000-2003. 11

sponding industry is open to labour migrants from a given new member state. For example, the UK completely opened up its labour market for the EU8 group in 2004. Therefore UK dummies for all industries for all EU8 countries equal 1 starting from 2004. In contrast, France held transitional provisions for 2004-entrants until 2008. Prior to 2008, the French government applied a special job scheme, which allowed for free labour market access only in specific industries, such as construction, tourism, and catering. Hence France dummies for EU8 industries take a value of 0 until 2008, except for the three mentioned sectors. To summarize a set of 15 dummies in a single measure, we apply special weights that reflect how strongly the opening of a particular EU15 labour market affects the citizens of a given new member state. It is reasonable to assume that labour migrants, for example, from Estonia were more sensitive to the opening of the Finnish labour market than of the Portuguese one. One approach is to use EU2 and EU8 migration stocks as of year 2000 in each of the EU15 countries to obtain such weights. w c,cd = Stock2000 c,c d 15 j=1 Stock2000 c,cj (2) Namely, a weight for each pair of a new member state (c, source country) and an old member (c d, destination country) is equal to the share of migrants from this source country living in the destination country relative to the total number of migrants from this source country in EU15. Such weights reflect historical and geographic ties between EU members and account for network effects, which facilitate migration decisions. In addition, we construct two alternative sets of weights: first, by using the distances between the two largest cities of each source and destination countries (the shorter the distance, the larger is the weight); second, by combining these distance-based weights with stock-based ones (we take an average of the two weights). The legislation information is summarized in one instrument measure: 15 IV cit = w c,cj D ccj it (3) j=1 IV cit is the instrument s value for one observation (source country-industryyear). D ccj it - legislation dummy for openness of the labour market in a j th old EU member s corresponding industry for the citizens of a given source country in a given year. w c,cj - weights. By construction, IV cit is in the range [0;1]. 12

One of the limitations of the legislation dummies is the low industry-level variation. Many old EU members did not explicitly specify which industries are open to labour migrants from EU10, but rather allowed for special job schemes in sectors that experienced labour shortages. To control for this, we multiply the legislation dummy D ccj it by a measure of labour shortages in a given industry of a j th old EU member state. As such measure, we use the share of firms (in the destination industries) reporting to be constrained by the labour factor. This information is available from the EU Commission Business Survey. Such modification allows not only to account for implicit legislation changes but also to control for differences in labour market conditions across industries of the old EU members. Easiness to find a job, which increases in sectors experiencing labour shortages, can be another important criteria for mobile workers. 11 The baseline first stage regression takes the following form: EM ict = γ 1 IV ict + γ 2 I ict + γ 3 C ct + φ t + κ ic + u ict (4) EM ict is an endogenous measure of emigration rates from (1). γ 1 - the coefficient of interest - reflects the marginal contribution of our IV given industry- and country-specific time-varying covariates (I ict, C ct ) and time dummies (φ t ). By including industry fixed effects κ ic, we identify the IV effect only from within-industry variation in propensity to emigrate. 6 Empirical Results The section presents and discusses empirical results. For all reported models, the instrument was constructed using combined distance- and migrationbased weights 12. As discussed in the section 5.2, we interact the legislation dummies with a measure of labour shortages in a given industry of a particular old EU member. All regressions include firm fixed effects and thus capture within-firm variation in performance as a response to changes in an industry s human capital supply. In our preferred specification, we control for country s GDP growth rate, lagged FDIs, lagged industry-level investment, average labour shortages in EU15, and time fixed effects. The sample spans over 2000-2013 and includes firms with at least two years of available 11 A possible concern with such a modification is that labour shortages in the old EU member states might not be fully exogenous to firm productivity in EU10 countries, due, for example, to common technology shocks. We control for this by including industryspecific time dummies or an average measure of labour shortages in a given industry for all EU15 members. 12 The results qualitatively hold with only distance-based and migration-based weights. 13

financial data to calculate the TFP index. As a note of caution, we might not capture companies at the lower tail of the productivity distribution, if they are also less likely to be included in the sample. Based on observables, though, firms in the regression sample are not statistically different from ones in the full sample (see table 1). Note: We are currently in process of acquiring detailed migration data from the Eurostat, which we need to run the main specification (1). In this preliminary draft, we present reduced form estimations and 2SLS results, where instead of actual emigration rates we use firm-reported skill shortages. 6.1 Reduced Form Regressions Tables 2-4 present the reduced form estimations: we regress firm outcomes directly on the instrument value. We use one-period lag for the IV to account for some inertia between the legislation change and migration decisions. All dependent variables are in natural logarithms, and the IV ranges from 0 to 1. The coefficients may be interpreted as the log point change in dependent variables, when the IV increases from 0 (no free labour mobility within EU for workers qualified to work in a particular industry) to 1 (complete free labour mobility). For the main sample of firms (table 2), the direct effect of the EU opening is significant for firm total factor productivity, personnel costs, and assets. Firms in industries subject to higher labour mobility experience a drop in TFP, while increasing their spendings on the employees and capital investments. In an extreme case of the IV jump from 0 to 1, firm TFP drops by 0.15 log point (13.9%), personnel costs rise by 0.088 log point (9.1%), and investments in fixed assets grow by 0.19 log point (20.9%). To check whether these effects are driven by particular firms, we estimate the model for several sub-samples. First, we restrict the sample to Incumbents - firms, which were incorporated prior to 2002. We drop 2003-2012 market entrants that could have adjusted their entry decisions to new labour market conditions. The only precisely estimated effect for the incumbent firms is the TFP drop. Insignificance of the IV coefficient for other dependent variables might be related either to large remaining heterogeneity within the sub-sample or weaker adaptation possibilities of the incumbent firms to changes in human capital supply. In addition, we run the regressions on two sub-samples, which are likely to contain the most productive firms: Innovators - firms with registered patents and Foreign - firms with foreign capital. For innovating and foreign-owned firms, the productivity drop is imprecisely estimated, while personnel costs in the extreme case would increase by 0.17 log points (18.7%) and by 0.14 log points (14.8%) respectively. More productive firms also increase their capital 14

investments (unreported) and raise their capital intensity (capital/labour ratio) while facing adverse shocks to human capital supply. These results suggest that innovating and foreign-owned firms have been more successful in adapting to harsher labour market conditions caused by the EU opening. Personnel costs could have increased due to additional hiring and training expenses as well as higher wage bill if more productive firms (unlike other companies) have been able to retain qualified employees. In addition, more productive firms experience higher opportunity costs of vacancy duration and could be more willing to increase wages to circumvent the brain drain problem. We further explore the brain drain effect using firm-level data from the BEEPS survey. Table 5 presents the reduced form estimations. BEEPS contains only a limited number of firms with available panel data, therefore, in the reported specification we pulled firm observations together, adding firm-level covariates: lagged sales, capital, quadratic terms for firm age and lagged number of employees, share of foreign capital, share of export in sales. All regressions are estimated with country*year (c*y), country*industry(c*i), and industry*year(i*y) fixed effects. The remaining variation in dependent variables should come from country-industry-year changes in the value of the instrument. As with the Amadeus data, we find negative effect of the EU labour markets opening on firm total factor productivity. The effect on wage is imprecisely estimated, this could be related to the fact that it does not include other components of personnel costs. As an additional insight, we report significant increase in employee training by firms in industries, which have potentially experienced higher labour emigration. Further, these firms have been less likely to introduce a new product (innovate). The latter result should be interpreted with caution, since, due to data availability, many observation were dropped. 6.2 Human Capital Constraints as the Brain Drain Consequence, 2SLS Regressions The reduced form regressions represent the intention-to-treat effect. It shall be of interest to estimate changes in outcomes for firms, which have effectively experienced the brain drain problem. We consider skill shortages (human capital constraints - HCCs) as an indicator of the brain drain problem. If changes in EU15 labour mobility legislation have indeed induced higher emigration rates of the qualified workforce, we would observe increasing skill shortages as reported by firms in the new member states. The 15

measure of HCCs is described in the section 4 above. Table 6 presents the first-stage estimation results for three modifications of the IV. IV1 contains only legislation dummies weighted by migration stocks; IV2 contains legislation dummies but uses weights that combine migration stocks and bilateral distances between source and destination countries; IV3, in addition, controls for labour shortages in old EU member states, weighted by migration stocks and bilateral distances. We include a number of covariates to switch-off demand-driven changes in the reported HCCs. Country GDP growth rate and FDI inflows (GDP ct, F DI ct 1 ) control for general country-specific shocks. Lagged investment (invest ict 1 ) accounts for country-industry-specific increase in HCCs due to the expansion of existing companies or new entries. Average HCCs in a given industry in EU15 countries (meanhcc it ) control for industry-specific labour-demand shocks, which are common across all EU members. Our IV coefficient thus captures residual variation in reported HCCs, which arises due to the emigration of people qualified to work in particular industries. All three IV modifications return similar results: a complete opening (IV=1) of all EU15 (industryspecific) labour markets would have resulted in a 15-21% increase 13 in skill shortages for firms (within the corresponding industry) from a new member state. IV coefficients are statistically significant, and the F-test rejects the null hypothesis of insignificance for all three modifications. Table 7 presents 2SLS estimates with human capital constraints as an instrumented independent variable. We estimate the reported models using the instrument (IV3), which accounts for legislation changes as well as labour shortages across industries of destination countries. The first-stage details (IV coefficient with standard error) are presented below the main regressions results 14. HCCs measure (share of firms in an industry, reporting to be constrained by labour) ranges from 0 to 1. The coefficient of interest thus represents log point change in dependent variables when HCCs increase by 1 unit (100%). It is more convenient to use realistic changes in HCCs to characterise the estimated effects. A one st. deviation (0.11) increase in HCCs caused by the EU15 labour market opening leads to a 6.1% drop in firm TFP and a 5.2% increase in personnel costs. Again as in the reduced-form estimations, more productive innovating and foreign-owned companies do not experience significant decreases in TFP, but raise their employee expenses and increase capital intensity. 13 relative to the case when all EU15 labour markers are closed, i.e. IV =0 14 The reported first-stage coefficients might differ slightly from those reported in Table 6, since some industry-year observations were dropped due to missing firm-level data. 16

7 Conclusion This paper uses firm-level panel data to evaluate the effect of brain drain on firm outcomes. To overcome the endogeneity bias, we exploit the natural experiment setting of the EU enlargements in 2004 and 2007. We argue that the gradual and industry-specific opening of EU labour markets to citizens from new member states throughout 2004-2014 can serve as an instrument for emigration rates experienced by EU2 and EU8 countries. Our findings suggest that changes in labour mobility laws created a large increase in emigration rates and binding human capital constraints for firms in the most affected industries. We show that an emigration-driven reduction in labour supply resulted in lower total factor productivity of EU2 and EU8 firms. We also document an increase in personnel costs and training of employees. Furthermore, we find that more productive innovating and foreign-owned firms increased their personnel costs by more and experienced smaller drops in productivity. These firms have been more successful in adjusting to higher labour mobility, in particular, in retaining and attracting better qualified workers. Our results are important both for firms and for policy makers. Being aware of the brain drain helps firms to react timely and in an adequate way. Firms can benefit from active human resource strategies, focusing, for instance, on providing training and retention measures. For policy makers, the effects of migration are not a matter of fate, to a large extent, they depend on the public policies adopted in the receiving and sending countries 15. The prevalence of skill shortages, for instance, justifies the need to invest in the skills of their local labour force and to mitigate search frictions. A skill upgrading of the local labour force can in the short-term be addressed by providing specific training courses by public institutions and in the long-term by adjusting the education system to labour market needs. Knowing that those skilled people are needed can justify the investment. Labour market frictions could be reduced by an efficient labour agency and especially by harmonizing EU-wide labour agencies, so that workers within the EU are aware of all possibilities. This might encourage unemployed in other EU states to search for a job in countries and industries that experience shortages. By attracting workers from other EU countries and incentivised return migration, firms in new member states could also reap the benefits of labour migration in an enlarged Europe. In further research, we will evaluate the effects of brain drain along firms initial productivity distribution to see if there are heterogeneous effects de- 15 Docquier and Rapoport (2012). 17

pending on the initial productivity of firms. We also plan to use other TFP measures, i.e. Olley&Pakes and Levinson&Petrin. Besides this, we intend to further exploit the constructed instrument and empirically evaluate whether and how migration-driven skilled labour shortages affect firms practices. In particular, we will work more with the BEEPS firm-level survey data 16 to estimate the effect of brain drain on changes in firms training provision, outsourcing, and human capital management. 16 The 2012-13 wave became available in December 2014. 18

References Brixiova, Zuzana, Wenli Li, and Tarik Yousef (2009). Skill shortages and labor market outcomes in Central Europe. Economic Systems 33 (1), pp. 45 59. Clemens, Michael A. (2011). Economics and Emigration: Trillion-Dollar Bills on the Sidewalk? Journal of Economic Perspectives 25 (3), pp. 83 106. (2013). Why Do Programmers Earn More in Houston Than Hyderabad? Evidence from Randomized Processing of US Visas. American Economic Review 103 (3), pp. 198 202. Constant, Amelie F. (2011). Sizing It Up: Labor Migration Lessons of the EU Enlargement to 27. IZA Discussion Papers 6119. Institute for the Study of Labor (IZA). Docquier, Frdric and Hillel Rapoport (2008). brain drain. The New Palgrave Dictionary of Economics. Ed. by Steven N. Durlauf and Lawrence E. Blume. Basingstoke: Palgrave Macmillan. (2012). Globalization, Brain Drain, and Development. Journal of Economic Literature 50 (3), pp. 681 730. Dustmann, Christian and Albrecht Glitz (2015). How Do Industries and Firms Respond to Changes in Local Labor Supply? Journal of Labor Economics (forthcoming). Dustmann, Christian, Tommaso Frattini, and Anna Rosso (2012). The Effect of Emigration from Poland on Polish Wages. CReAM Discussion Paper Series 1229. Centre for Research and Analysis of Migration (CReAM), Department of Economics, University College London. Elsner, Benjamin (2013). Does emigration benefit the stayers? Evidence from EU enlargement. Journal of Population Economics 26 (2), pp. 531 553. Freeman, Richard B. (2006). People Flows in Globalization. Journal of Economic Perspectives 20 (2), pp. 145 170. Gorodnichenko, Yuriy and Monika Schnitzer (2013). Financial Constraints And Innovation: Why Poor Countries Don t Catch Up. Journal of the European Economic Association 11 (5), pp. 1115 1152. Haskel, Jonathan and Christopher Martin (1993). The Causes of Skill Shortages in Britain. Oxford Economic Papers 45 (4), pp. 573 88. Kahanec, Martin (2012). Labor Mobility in an Enlarged European Union. IZA Discussion Papers 6485. Institute for the Study of Labor (IZA). Kahanec, Martin, Mariola Pytlikova, and Klaus F. Zimmermann (2014). The Free Movement of Workers in an Enlarged European Union: Institutional 19

Underpinnings of Economic Adjustment. IZA Discussion Papers 8456. Institute for the Study of Labor (IZA). Kerr, Sari Pekkala and William R. Kerr (2013). Immigration and Employer Transitions for STEM Workers. American Economic Review 103 (3), pp. 193 97. Kerr, Sari Pekkala, William R. Kerr, and William F. Lincoln (2013). Skilled Immigration and the Employment Structures of U.S. Firms. NBER Working Papers 19658. National Bureau of Economic Research, Inc. (2014). Firms and the Economics of Skilled Immigration. Innovation Policy and the Economy, Volume 15. NBER Chapters. National Bureau of Economic Research, Inc. Kerr, William R. (2013). U.S. High-Skilled Immigration, Innovation, and Entrepreneurship: Empirical Approaches and Evidence. NBER Working Papers 19377. National Bureau of Economic Research, Inc. Lewis, Lafortune Tessada (2013). People and Machines A Look at The Evolving Relationship Between Capital and Skill In Manufacturing 1850-1940 Using Immigration Shocks. Working paper. Llull, Joan (2014). The Effect of Immigration on Wages: Exploiting Exogenous Variation at the National Level. Working Papers 783. Barcelona Graduate School of Economics. Mayr and Peri (2009). Brain Drain and Brain Return: Theory and Application to Eastern-Western Europe. The B.E. Journal of Economic Analysis & Policy 9 (1), pp. 1 52. Melitz, Marc J. (2003). The Impact of Trade on Intra-Industry Reallocations and Aggregate Industry Productivity. Econometrica 71 (6), pp. 1695 1725. Mishra, Prachi (2014). Emigration and wages in source countries: a survey of the empirical literature. International Handbook on Migration and Economic Development. Chapters. Edward Elgar. Chap. 9, pp. i iii. Mitaritonna, Cristina, Gianluca Orefice, and Giovanni Peri (2014). Immigrants and Firms Productivity: Evidence from France. IZA Discussion Papers 8063. Institute for the Study of Labor (IZA). OECD (2012). Free Movement of Workers and Labour Market Adjustment. OECD. (2013a). Coping with Emigration in Baltic and East European Countries. OECD Publishing. (2013b). World Migration in Figures. A joint contribution by UN-DESA and the OECD to the United Nations High-Level Dialogue on Migration and Development, 3-4 October 2013. 20