THE EFFECTS OF OUTWARD FDI ON DOMESTIC EMPLOYMENT Cesare Imbriani 1, Filippo Reganati 2, Rosanna Pittiglio 3 1 University of Roma La Sapienza, P.le Aldo Moro, 5; 00100 Roma, Italy, e-mail: cesare.imbriani@uniroma1.it 2, 3 University of Foggia, L.go Papa Giovanni Paolo II, 1; 71100 Foggia, Italy, e-mail: 2 f.reganati@unifg.it; 3 r.pittiglio@unifg.it Abstract. One of the channels through which home jobs may be affected by the increased degree of economic integration in the EU-25 is through the employment (re)-allocation choices of multinational enterprises. Using data on Italian multinationals, this paper examines the impact of Italian outward FDI on local employment between 2001 and 2005. The authors found some evidence of a complementary relationship between parent firm employment and affiliate employment located in Central and Eastern countries. They found no statistically significant effect of a reduction in wages on parent employment of affiliates located in old European Union members. Keywords: multinational enterprises, foreign and domestic employment. 1. Introduction 1 There are concerns that firms opening new foreign subsidiaries also transfer jobs abroad. Between 1985 and 1997 Italian manufacturing firms created 360,000 new foreign jobs and lost 350.000 domestic jobs. Is there any relation between these two trends? Are foreign jobs created at the expense of domestic ones? This paper addresses these questions by analysing trends in foreign and domestic employment for a sample of Italian multinationals between 2001 and 2005. Which are the factors that might affect the nature of the relationship between domestic and foreign employment? The general idea that the relationship is one of substitution derives from straightforward reasoning. If a firm invests to save labour costs, it will substitute expensive domestic labour for cheap foreign labour. If it invests to enter a foreign market it will be likely to stop exporting to that market and domestic employment will fall along with domestic output. There are however also good arguments supporting a relationship of complementarity that may offset this expected loss of domestic jobs. Foreign activities may require more headquarters personnel; reductions in production costs because of cheap labour or the exploitation of scale economies make MNEs 1 This work was jointly conceived and produced by the three authors. However, sections 1 and 6 were written by Cesare Imbriani, sections 2 and 4 by Filippo Reganati and sections 3 and 5 by Rosanna Pittiglio. more competitive and their domestic output and employment may rise; if sales of the foreign subsidiary complement exports (e.g. by extending the product range), then parent output and employment will also increase. In substance there are two channels at work. The first one (labour intensity effect) is that for given domestic output foreign investments affect the labour intensity of production, the number of employees per unit of output. The second one (output effect) that for given labour intensity foreign investments affect domestic output and thus domestic employment. The question, though, is broader. It is not just how foreign and domestic employment are related, but also what would have happened to domestic employment of MNEs had they not invested abroad. Even though we may observe that domestic jobs are replaced by foreign ones, this loss could have been greater with no foreign investments. From this perspective, this paper carries out a counterfactual analysis: it uses firm level data to look at the links between home and foreign employment of a sample of Italian MNEs, also taking also account how the domestic demand for labour would have changed if these firms had not invested abroad. As the 'right' counterfactual does not exist (we cannot observe what would happen to firms as a consequence of what they have not done), we proxy it with a sample of non investing national firms. As far as we know, this is the first work that makes use of this counterfactual analysis for outward investments. As an increasing number of firms expand operations abroad, there are many fears that domestic jobs 527
are being exported to foreign countries. The relocation of labour-intensive activities towards low-wage countries is often evoked in the public debate as a major determinant of job losses at home, at the expenses of unskilled labour. Foreign Direct Investment (FDI) towards advanced economies could also negatively affect domestic employment, as firms choose to serve foreign markets by local production rather than by exports from the home country. There could even be a dynamic effect through which FDI indirectly determines lower employment, as in the model of Basevi and Ottaviano [1], where firms' relocation abroad implies lower externalities and a higher cost of innovation at home: the domestic location will therefore become less attractive to new firms, ending up with a smaller than optimal amount of firms and employment. At a closer analysis, however, the effects of FDI on home-country employment appear to be less clearcut. First of all, there is not necessarily a perfect investment-substitution between the home and the foreign country: firms may invest abroad in order to diversify or expand in foreign markets, without reducing at the same time the domestic capital stock. Second, FDI, through lower operating costs or improved access to distant markets, may be the only way to expand firms' scale of output: in other words, FDI amounts to pick an investment opportunity that otherwise would have been taken by other competitors. Finally, co-ordinating and supervising the activities of foreign affiliates may require more labour in the home-based headquarters. Since theoretical predictions about the effects of FDI on employment vary greatly, it is very useful to address the question at an empirical level. So far, the literature has focused on multinational companies, looking mainly at how parent employment responds to changes in foreign affiliate's wages [2, 3, 4]. However, as convincingly argued in Barba Navaretti and Venables [5], the results of these studies are conditional on the multinational having already invested abroad. In other words, this approach is not able to deal with the potential substitution effect which takes place when a company moves production activities away from home for the first time. Another important limitation is that it ignores the external effects of FDI on non-multinational companies, such as local suppliers. Earlier works on the home country effects of outward FDI have primarily addressed the issue of whether employment or production abroad complements or substitutes employment in parent companies. Overall, the results do not seem to support the fear that MNEs are exporting domestic jobs, particularly to low-wage countries. In fact, looking at the United States and focusing on how the multinational firm set employment in different international locations in response to changes in relative wages, Brainard and Riker [3] found little evidence of substitutability between parent and affiliates employment. Braconier and Ekholm [2] for Swedish MNEs, and Konings and Murphy [4] for a sample of European MNEs, found a relationship of substitution but only between parent firms and their affiliates in high- income countries. These results are not completely satisfactory for two reasons. First, as already noted, they lack a counterfactual analysis on what would have happened in the absence of multinational expansion. Secondly, all previous studies focus on how MNEs set employment in parents and already existing foreign affiliates, while no attention is given to the decision of setting up new foreign activities. As recently shown by Braconier and Ekholm [2], the impact on parent's labour demand may be very different, the opening of new affiliates abroad often implying a completely new organisation of the activities within the MNE. The rest of the paper is structured as follows. The next section reviews the empirical literature, while section 3 presents the data. The econometric specification is described in the following section, and the results are illustrated in section 5. Section 6 concludes. 2. A brief survey of the literature In the empirical literature one strand has studied the role of multinationals activities by testing the effect of producing abroad on parent labour demand per unit of output. Using firm level data, Blomström, Fors and Lipsey [6] found that US multinationals have allocated some of their more labour intensive operations to affiliates in developing countries. They compared their results with those of Swedish multinationals and found no evidence in Swedish multinationals of labour substitution between home and foreign employment. Another strand has tested employment linkages by estimating translog cost functions, in which the coefficients of relative wages can be regarded as substitution elasticities. Using panel data for US multinationals and their affiliates in 90 countries, Brainard and Riker [3, 7] estimate substitution elasticities between employees in parent companies and their foreign affiliates as well as between employees in affiliates disaggregated by the geographical location and the development stage of the host country. Among their main findings is a very low degree of substitution between parent and affiliate employment. In contrast, employees in affiliates in different emerging market countries are found to have a very high degree of substitution, whereas the relationship between employees in affiliates in industrial countries and emerging market countries is one of complementarity. Slaughter [8] uses a very similar methodology, except that he confines his analysis to production workers; i.e. those employees who are most exposed to potential displacement effects. He also finds a very low degree of substitution between employees in parent companies and their foreign affiliates; in fact, when the capital stock is allowed to vary, the relationship is 528
one of complementarity. As further proof against the claim that MNEs have been exporting jobs, he notes that affiliate employment of production workers does not seem to be systematically related to relative parent/affiliate wages and has actually fallen over the period 1977 89. Hatzius [9] poses the question of whether Swedish MNEs vary their distribution of labour demand in response to wages in Sweden and in a range of other host countries. The results indicate that labour demand is highly responsive to wages in actual and potential host countries but not to wages in Sweden. This asymmetry is somewhat odd but might suggest that, as in the United States, substitution elasticities are rather low. Bruno and Falzoni [10] argue that once adjustment costs are taken into account, strong substitution effects can be found, especially between parent firms and their affiliates in developed countries, while they find complementarity in the long run. Using a firm level panel data set of more than 1,200 European multinational enterprises and their subsidiaries, located in either the European Union, Central and Eastern Europe or both, Konings and Murphy [4] investigate whether employment in the MNEs subsidiaries are substitutes for home employment. They found evidence for substitution effects between parent and foreign employment, operating in the manufacturing sector, taking place mainly between EU parents and their affiliates located within the EU, rather than affiliates located in Central and Eastern Europe. Analysing the employment effects of the relocation of production by Swedish multinational enterprises towards Central and Eastern Europe, Braconier e Ekholm [2] found a relationship of substitution but only between parent firms and their affiliates in highincome Western Europe. One of the first empirical assessments of the effect of FDI on employment was made by Brainard and Riker [7], who estimate an equation of U.S. multinationals labour demand across different plant locations. The coefficients on cross-elasticity of substitution provide then information on whether foreign affiliate labour complements or substitutes parent labor. They find that the cross-elasticity between the parents and the affiliates is less than one, implying only partial substitution. Substitution between affiliates in different countries is instead markedly higher, especially for low value-added industries and for affiliates located in countries with similar levels of development. They conclude that labour in the U.S. does compete only at the margin with labour abroad, and that employment shifting takes place predominantly between foreign affiliates in less developed countries. Other studies based on the same methodology find similar results: contrary to conventional wisdom, employment in foreign affiliates located in low-wage countries appears to be complementary to home employment, while there is substitution between the latter and employment in advanced countries. As mentioned earlier, this literature does not assess the impact of investments abroad made for the first time. A different approach is taken by Barba Navaretti and Castellani [11], who compare the employment performance of firms investing abroad for the first time with an appropriate counterfactual of national firms. Using data on Italian companies, they find that becoming a multinational firm has no significant effect on employment performance, while it is associated with a better performance in terms of output and total factor productivity growth. This work does not distinguish across investment destination countries, which may significantly influence its effect on employment. Our work is also related to another two strands of literature. One looks at the effects of FDI on the labour intensity of home-country production, showing that they depend not only on the location of the affiliates, but also on certain structural features of the home country. As outlined by Blomström, Fors and Lipsey [6], for instance, larger affiliate production implies a lower labour intensity in the U.S., while the opposite is observed for Sweden. This difference presumably reflects different investment strategies, with US firms allocating production activities across countries in order to exploit factor price differences, and Swedish affiliates more engaged in selling to local customers. The other branch of the literature focuses on the skill composition of domestic employment. Using various measures of affiliate activity, Slaughter [12] finds that they do not appear to influence the share of non-production worker wages in the total wage bills of 32 US manufacturing industries. Replicating his industry-level estimates with data on Japanese firms, Head and Ries [13] obtain similar results. However, once they move to a firm-level analysis, higher affiliate employment implies a higher nonproduction worker wage share in the parent firm. There are also differences depending on the destination of FDI: the positive effect is associated with affiliate employment in low-wage countries, while more employment in the US appears to have the opposite effect. 3. Preliminary data analysis Our dataset is related to a sample of 298 Italian parent firms with 299 affiliates located in the Western Europe (WE) and 104 affiliates located in Central and Eastern Europe (CEE). The data covers the period 2001-2005 and it is in the form of a balanced panel data. The data source is the Amadeus (Analyse MAjor Databases from EUropean Sources) database, provided by Bureau Van Dijk; in its most concise version, it contains detailed reports on the 200,000 largest European companies including the Eastern European ones (turnover exceeding 15 million Euro) 2 2 The authors are very grateful to Umberto De Marco (Senior Sales Executive of Bureau Van Dijk Electronic Editions) for his assistance with the Amadeus database. 529
Each company report contains descriptive information and up to 10 years history of consolidated and unconsolidated annual accounts. Hence, the authors extracted from the Amadeus database the sample of Italian companies with their affiliates located both in the Western and in the Central Eastern European (Bulgaria, Czech Republic, Estonia, Hungary, Poland, Romania, Slovak Republic, Latvia, Lituania, Slovenia). Following Konings and Murphy [14], the 50 % participation rate benchmark is adopted in the analysis to determine a subsidiary. So, we consider parent an Italian firm holding a direct ownership share of at least 50 % in one or more firms located in an other country (different from Italy). We refer to these firms as affiliates. For the selected companies, there were retrieved data on turnover (operating revenue), number of employees and cost of employees. From this data, the authors obtained the firms' real output (proxied by turnover), employment and average labour cost (cost of employees divided by number of employees). As in Brainard and Riker [15], the subsidiary companies were assumed to be homogeneous and were aggregated at the parent firm level. Hence, the authors constructed the average values of the variables of interest taken over all subsidiaries in Western Europe and the average values of the variables of interest taken over all affiliates in Central and Eastern Europe. Descriptive statistics on characteristics of our sample are reported in Table 1 and Table 2. Table 1 shows the distribution of Italian affiliates across the different European. In our sample nearly 75 % of Italian parents have affiliates in Western European, most of which are located in France and Spain (24.8 % and 20.3 % of the whole sample, respectively). Central and Eastern European host almost 26 % of affiliates of Italian firms. Romania and Poland are two countries that host most of the affiliates located in Central and Eastern Europe with 13.9 % and 6.9 % respectively. Table 1. Distribution of Affiliates N % N % Western Central Eastern 299 74,2 104 25,8 Austria 2 1.0 Bulgaria 3 0,7 Belgium 18 5.0 Czech Republic 9 2,2 Denmark 3 1.0 Hungary 1 0,2 Finland 2 1.0 Poland 28 6,9 France 100 25.0 Romania 56 13,9 Germany 24 6.0 Slovacchia 1 0,2 Ireland 1 0.0 Slovenia 6 1,5 Netherlands 13 3.0 Portugal 4 1.0 Spain 82 20.0 Sweden 10 3.0 United Kingdom 40 10.0 Whole sample 403 100 Source: own elaborations on Amadeus database The information above is complemented by Table 2, giving the same information, by sector. In the specific case, the table shows the distribution of parents and affiliates between two classes of sectors manufacturing and non manufacturing relating to two areas considered. First of all, it can be observed that most of Italian parent firms have affiliates in the same group of sectors. So, in our sample, most of manufacturing parent firms have affiliates in manufacturing sectors (57 %) mainly located in Western European. A similar situation can be observed when we consider the non-manufacturing firms. In fact, in this case, nearly 74 % of non-manufacturing parent firms control affiliates in non-manufacturing sectors (for the most part located in West Europe) while only 26.12 of parent firms control affiliates in manufacturing sectors (more or less in the same percentage between West and CEE affiliates). Table 2. Sectorial Distribution of Parents and Affiliates Affiliates Manufacturing Non-manufacturing Western Source: own elaborations on Amadeus database Central Eastern Western Central Eastern Manufacturing 35.36% 21.43% 38.21% 5.00% Italian Parents Non-manufacturing 12.69% 13.43% 64.18% 9.70% 4. Methodology and economic specification Let us consider a MNE that produces global output using a production function which depends only on labour input in different locations: Y = F( L, L, L ) (1) P WE CEE where Y is the total output of the MNE, L p employment in parent firm, L CEE, is employment in CEE affiliates and L WE is employment in Western European (WE) affiliates. Total cost minimisation under constraint (1) yields to the following conditional demand for employment in the parent firm: N P = f P ( wp, wwe, wcee, Y ) (2) where W P, W WE and W CEE, respectively indicate the wage cost per worker in the parent firm, in WE affiliates and in CEE affiliates. Assuming that labour productivity remains constant, one would expect the following partial derivatives of equation (2) L P < 0 ; L P > 0 ; L P 0 ; > 0 w w w Y with k =CEE, WE. P k k L P (3) 530
From (3) it is expected that the own wage is negatively related to home parent labour demand: an increase in the parent unit wage cost should lead to a decrease in the demand for parent employment. If parent and subsidiary employment are substitutes, it is L expected that P > 0, but when parent and affiliate wk employment are complements it is expected that LP 0. wk Equation (2) is the basis of the empirical specifications used in this paper. In particular, following Brainard and Riken [15] and Konings and Murphy [4], the paper estimates (2) by accounting for firm-level fixed effects and assuming a log-linear approximation of equation (2): ln Nipt = αip + α1 lnwipt + α2 lnwiceet + α3 lnwiwet + α3 lnypt + εit (4) In equation (4) Y is the weighted sum of value added in the parent company and the value added of its affiliates, where the weight represents the importance of value added in total value added for the parent versus the affiliates; W ip is the total wage bill of the i-th parent company divided by its total employment; W icee is the average total wage bill of all CEEC affiliates of the i-th parent firm divided by the average total employment of all CEEC affiliates of that particular MNE; W iwe is the average total wage bill of all WE based affiliates of the i-th parent firm divided by the average total employment of all WE affiliates of that particular MNE; α ip is a firm specific fixed effect that is not observable and ε ip is a white noise error term. Furthermore, there are included in (4) year dummies to check for unobserved aggregate shocks, which are common to all parent firms. 5. Estimation results Table 3 shows firm level fixed effect estimates for equation (4). Column (1) gives the results for the whole sample, while column (2) and (3) for parent firms operating in manufacturing and non manufacturing sectors respectively. The first point worth noting is that the own wage elasticity is negative and significant, as expected: the coefficient of W P has a value ( 0.40) which is well within the range of estimated labour demand elasticities reported in the literature. The cross-wage elasticities give an indication of the responsiveness of parent employment to wage changes in affiliates. These elasticities are given by the coefficients that are associated with W WE, W CEE, reflecting the effect of wage changes in affiliates located in the Western Europe and Central and Eastern Europe, respectively. The two coefficients are estimated negatively, however, only the wage effect on parent employment of affiliates located in Central and Eastern Europe is statistically significant at 1 % level, with a coefficient of 0.14. This would suggest that a reduction in labour costs by 10 % in affiliate wages located in Central and Eastern European countries is associated with an increase in home (parent) employment of 14 % on average. In all probability, the main reason of this result is due to the fact that the foreign activities may require more headquarters personnel. However, although the negative signs of both variables would suggest complementarity between parent and affiliate employment, their coefficients are not statistically significant. Table 3. Estimation results All sectors Manufacturing Non- Manufacturing (1) (2) (3) -3.368687-6.205444-0.9839845 COSTANT [-7.27]*** [-10.17 ]*** [-1.32 ] Ln Y Ln Wp Ln W WE Ln W CEE Number of obs. 0.977702 1.328934 0.6385827 [23.91]*** [24.75 ]*** [10.25 ]*** -0.4031758-0.5216709-0.3170543 [-12.50]*** [-10.98]*** [-6.96]*** -0.0368628-0.0537804 0.0183191 [-0.70] [ -0.99] [0.15] -0.1416733-0.1029562-0.1705198 [-1.84]* [-1.14] [-1.27] 1500 1060 440 Number of groups 300 212 88 R squared 0.3787 0.4703 0.3034 Notes: standard errors are reported in parentheses. *significant at 10% level, ** significant at 5% level, ***significant at 1% level. The regressions include time dummies. A similar analysis has been conducted for the sub-sample of parent companies operating in the manufacturing and non-manufacturing sector. Also in these cases, we don t find statistically significant substitution elasticities for both sub-samples of parent companies. The only sign of employment relocation effect, giving an indication of the technological substitution possibilities between parent and affiliate employment, could be identified in the positive sign of variable W WE in the non-manufacturing sectors (column 3), however, also in this case, its sign not statistically significant doesn t allow us to conclude anything about possible relocation effects. 6. Conclusions This paper tested to what extent labour employed in the home country is substitutable or complementary to foreign employment. Using data on Italian multinationals, the authors estimated a conditional demand function for parent employment. They didn t find any evidence of a substitutionary relationship between parent firm employment in Italy and affiliate employment located in Central and Eastern countries. 531
Moreover the authors find no statistically significant effect of a reduction in wages of affiliates located in old European Union members. These results seem to suggest that the foreign activities do not bring any reduction in home (parent) employment. In other words and in contrast to the expectations, it would suggest that competition from low wage locations (on average) does not constitute a threat to Italian parent employment. However, this paper has not investigated the employment impact of outward FDI at a sectoral level. Further analyses on this direction, might be important in order to assess the complete impact of actual investment on employment relocation between parents and their affiliates. References 1. BASEVI, G.; OTTAVIANO, G. L.P. The District and the Global Economy: Exportation vs Foreign Location. Journal of Regional Science, Vol 42, No 1, 2002, p. 47 51. 2. BRACONIER H.; EKHOLM, K. Foreign Direct Investment in Central and Eastern Europe: Employment Effects in the EU. CEPR, Discussion Papers, No 3052, 2001. 47 p. 3. BRAINARD, S. L.; RIKER, D. Are U.S. Multinationals Exporting U.S. Jobs? NBER Working Paper, No 5958, 1997. 59 p. 4. KONINGS, J.; MURPHY, A. Do Multinational Enterprises Substitute Parent Jobs for Foreign Ones? Evidence from Level Panel Data. LICOS Discussion Paper, No 100, 2001. 37 p. 5. BARBA NAVARETTI, G.; VENABLES, A. J. Multinational Firms in the World Economy. Princeton University Press, 2004. 233 p. 6. BLOMSTRÖM, M.; FORS, G.; LIPSEY, R. E. Foreign Direct Investment and Employment: Home Country Experience in the United States and Sweden. Economic Journal, 107, 1997, p. 51 63. 7. BRAINARD, S. L.; RIKER, D. U.S. Multinationals and Competition from Low-Wage. NBER Working Paper, No 5959, 1997b. 59 p. 8. SLAUGHTER, M. Multinational Corporations, Outsourcing and American Wage Diversion. NBER Working Paper, No 5253, 1995. 63 p. 9. HATZIUS, J. Domestic Jobs and Foreign Wages. The Scandinavian Journal of Economics, 100, 4, 1998, p. 4 13. 10. BRUNO, G.; FALZONI, A..M. Multinational Corporations, Wages and Employment: Do Adjustment Costs Matter? Centre for Economic Policy Research, No 2471, 2000. 97 p. 11. BARBA NAVARETTI, G.; CASTELLANI, D. Does Investing Abroad Affect Performance at Home? Comparing Italian Multinational and National Enterprises. CEPR Working Paper, 2004. 76 p. 12. SLAUGHTER, M. Production Transfer within Multinational Enterprises and American Wages. Journal of International Economics, Vol 50 (2), 2000, p. 43 54. 13. HEAD, K..; RIES, J. Offshore Production and Skill Upgrading by Japanese Manufacturing Firms. Journal of International Economics, Vol 58, No 1, 2002, p. 23 35. 14. KONINGS, J.; MURPHY, A. Do Multinational Enterprises Relocate Employment to Low Wage Regions? Evidence from European Multinationals. LICOS, 2004, p. 13 19. 15. BRAINARD, S. L. An Empirical Assessment of the Proximity-Concentration Trade-off Between Multinational Sales and Trade. The American Economic Review, 87, 4, 1997, p. 21 29. 532