Foreign Affiliates in the French Manufacturing Industry: Source or Recipient of Technology Spillovers?

Transcription

1 Foreign Affiliates in the French Manufacturing Industry: Source or Recipient of Technology Spillovers? Liza JABBOUR * Jean Louis MUCCHIELLI ** Abstract Technological sourcing is becoming an important motivation for multinational enterprises to locate affiliates abroad, especially in technologically leading countries. In this paper we try to verify if foreign affiliates located in France are acquiring technology from the French economy and if they are transmitting some of their knowhow to French firms. We define three channels of technology spillovers, Horizontal, backward linkages and forward linkages. Our empirical analysis is based on a firm level data base of French manufacturing industries covering the period Our results show that foreign affiliates benefit from the proximity of French firms and from the acquisition of their inputs from local suppliers. We also find that foreign affiliates transmit technology to local firms through backward and forward linkages. Keywords: Technology spillovers, Technology sourcing, Vertical Linkages. JEL Classification: F23. * University of Paris I Panthéon-Sorbonne and TEAM-CNRS, ljabbour@univ-paris1.fr. Address: Maison des Sciences Economiques (TEAM Pôle Internationale) Boulevard de l Hôpital Paris France. ** University of Paris I Panthéon-Sorbonne and TEAM-CNRS, jlmuc@univ-paris1.fr. Address: Maison des Sciences Economiques (TEAM Pôle Internationale) Boulevard de l Hôpital Paris France. 1

2 I-Introduction: Traditionally, the economic literature analyzed the international diffusion of technology by focusing on the role of multinational firms in host countries. The major hypothesis behind this analysis are, first, the fact that multinational firms rely generally on their ownership specific advantages in order to be competitive internationally, second, the fact that multinational firms transfer know-how and technology to their affiliates located abroad and third, the fact that technology brought in by foreign affiliates will diffuse through the host economy. This analysis is relevant in the case of developing countries but not necessarily in the case of advanced and developed ones. In fact, domestic firms in developed countries generally lack the capacity to innovate and to successfully carry out research and development (R&D) activity, thus it is pertinent to suppose that those firms rely on foreign sources of technology. The case of developed countries is different. Domestic firms in those countries possess the necessary infrastructure 1, at the individual and national level, to be innovative and highly competitive on their market. The technological capacity 2 of firms in developed countries makes it more pertinent to suppose a diffusion of technology in both directions 3 between domestic firms and foreign affiliates. Many recent studies have shown that technology sourcing is an important motivation for firms to locate abroad especially in technologically leading countries, particularly the United-States. In fact, since technology is generally geographically localised [ Jaffe, Trajtenberg & Henderson (1993) and Keller (2002)], firms locate some of their activities, particularly R&D activity, close to technological clusters to benefit from the technology and know-how spilling over through the cluster. Generally those affiliates transmit the technology acquired to their parent company. 1 By infrastructure we mean the human capital available in the labor market, the research institutions present in the country, the technological infrastructure defined as the access to internet, telephony and electricity. (Archibugi & Cocco 2004) 2 Econometric studies of technology transfer through FDI usually find significance evidence on the existence of technology transfer in the case of developed countries and insignificant evidence in the case of developing countries. The lack of absorptive capacity on the part of developing countries firms is the most important reason explaining the absence of technology transfer. 3 From and toward foreign affiliates. 2

3 Serapio & Dalton (1999) certify that Foreign parent companies, particularly in drugs/ biotechnology and electronics industries, have established or acquired foreign R&D laboratories in the US in order to gain access to science and technology, and enhance their global capabilities for technology and innovation. Fosfuri and Motta (1999) and Siotis (1999), build theoretical models to analyse the FDI decision motivated by Technology sourcing. Fosfuri and Motta (1999) construct a model with two home country firms, with different technologies of production and targeting the same foreign market. They can decide to enter the foreign market through export or foreign direct investment (FDI) or not to enter. Their model shows that, the less productive firm will be motivated to engage in FDI activity if productivity spillovers, and thus the possibilities to acquire the other firm technology, are high. Siotis (1999) constructs a similar model and add the possibility of two-ways technology spillovers between foreign and domestic firms. He presents theoretical evidence of technology sourcing as a motivation for FDI. Empirical evidence on technology spillovres through outward FDI has been presented through many papers. For example, Griffith, Harrison & Reenen (2003) analyse the case of UK firms locating innovative activity in the USA and find some evidence on the existence of knowledge spillovers associated with technology sourcing. Singh (2003) uses patent citation data from six innovative countries -United States, Japan, Germany, France, United Kingdom and Canada- and finds strong evidence on technology spillovers associated with outward investment. Globerman, Kokko & Sjöholm (2002) analyse the case of 220 Swedish patents and find higher patent citation associated with outward investment. In this paper we study the case of the French manufacturing industry. Our aim is to verify if foreign affiliates installed in France are motivated by technology sourcing and/or if they help the diffusion of new technologies in the French industry. We define three channels of technology diffusion, Horizontal, Vertical-Backward, and Vertical-Forward and try to determine the direction of technology diffusion (i.e. from or toward foreign affiliates located in France). We use firm level data extracted from the firm annual survey realized by the French Ministry of Industry. We estimate the effect of foreign presence, at the horizontal and vertical levels, on the productivity of domestic firms and find significant evidence of technology spillovers only at the vertical levels. We also estimate the effect of domestic firms on the productivity of 3

4 foreign affiliates and find a significant and positive effect at the horizontal and Vertical-Forward levels but an insignificant effect at the Vertical-Backward one. The main contribution of the paper is twofold; first, the analysis of technology sourcing through vertical linkages. In the prior literature, technology sourcing has been mainly linked to the horizontal presence of foreign affiliates as well as to R&D activity of foreign affiliates in the host country. Second, the use of firm level data on vertical linkages. Previous work on spillovers through vertical linkages have suffered from unavailability of firm level data on linkages and had to construct sector level proxies for backward and forward linkages using input-output tables. The use of firm level data to construct vertical linkages increases the accuracy of the results. The rest of the paper is structured as follows; in the second section, we present a detailed analysis of the objectives of the paper. In the third section, we present data and methodology. In the fourth section, we present the results and in section five we present conclusions. II-Objectives of the paper: The objective of this paper is to analyse the structure of technology spillovers in the French manufacturing industry. More precisely the paper aims at answering the following questions: Do foreign affiliates investing in France, introduce new technologies and Know-how to the French economy? Does this technology spill over to French firms? And/or do foreign affiliates invest in France to benefit from technology and Know-how developed by French firms? As we mentioned earlier, the assumption of international technology diffusion through FDI is motivated by the ownership specific advantages theory of multinational firms [Dunning (1988)]. This theory argues that multinational firms rely on specific advantages such as superior technology or know how and that they use those advantages in their affiliates located abroad. Thus, through their affiliates, multinational firms introduce new technologies to host countries. 4

5 Nevertheless, recent theoretical and empirical works argue that ownership advantages exploitation may not be the only motivation for FDI. In the presence of technology spillovers, firms may choose to enter the foreign market through FDI, even when they lack ownership advantages. The existence of spillovers allows them to access advanced technology and by consequence to fill up their competitiveness gap [Fosfuri and Motta (1999) and Siotis (1999)]. Empirical evidence supporting the technology sourcing assumption has been presented by many recent studies. For example, Neven and Siotis (1996) find evidence of technology sourcing by Japanese affiliates installed in the United States and by American affiliates in the European Union. Driffield and Love (2003) find that the technology produced by the domestic industries in the United Kingdom spills over to foreign affiliates. The reasons that push us to assume that foreign affiliates in France may be motivated by technology sourcing are numerous. France is a highly developed country, fifth largest economy in the world in term of GDP and second in term of outward FDI. Compared to other European countries, France is a leading country in terms of R&D expenditures (public and private) and in terms of innovative activity (number of patents). Moreover, the French economy is endowed with a highly qualified, educated and productive labour force 4. In order to investigate the presence of technology spillovers in French industries and their structure, we define three channels of technology spillovers, Horizontal, Vertical-Backward and Vertical-Forward and estimate the impact of those channels on the productivity of firms 5. Horizontal technology spillovers stands for the technology, knowledge and know-how that are implicitly and involuntary diffused between firms investing in the same sector and/or region and are canalized by mechanisms like employees turnover [Fosfuri et al. (2001)] and demonstration effect [Glass and Saggi (1998, 1999)]. Horizontal technology spillovers have been analysed frequently in empirical studies and results differ significantly from country to country but generally significant evidence is found 4 Figures and Sources are presented in appendix A. 5 Illustrations of horizontal and vertical spillovers are presented in figures I and II. 5

6 in the case of developed countries 6. The absence of spillovers in the case of developing countries is mostly explained by the weak absorptive capacity of domestic firms [Haddad & Harrison (1993), Aitken & Harrison (1999)]. In the case of the French industry, we assume that both domestic firms and foreign affiliate have sufficient absorptive capacity that allows them to benefit from the technology developed by other firms. Thus we expect the horizontal channel to be effective for technology spillovers. By vertical technology spillovers, we mean technology and know-how that are transferred between firms and their suppliers. Vertical-Backward spillovers refer to the technology transferred from firms to their suppliers. The efficiency of backward linkages as a channel of technology transfer is justified by the idea that firms are willing to transfer some of their knowledge to their suppliers in order to guarantee the quality of their inputs. At the same time, firms also impose quality requirement and quality control that urge the suppliers to upgrade their technological and managerial capacities [Kugler (2000), Moran (2001)]. Evidence on technology transfer through backward linkages has been provided by recent studies like those of Smarzynska (2002) on Lithuania, Blalock and Gertler (2003) on Indonesia and Jabbour and Mucchielli (2003) on Spain. However, those studies suffer from the absence of firm level data on backward linkages. Authors compensate this lack of data by constructing proxies for backward linkages between foreign affiliates and local firms at the sectoral level using input-output tables. Such sectoral proxies may not give an exact representation of the relationship between firms and suppliers. In this paper, we overcome this problem by using firm level data on backward and forward linkages 7. The use of firm level data gives a more accurate estimation of the impact of vertical linkages on productivity. Vertical-Forward spillovers refer to the technology transferred from suppliers to their customers. In fact we suppose that the technology and Know-How of a firm is reflected in the quality of its products. When those products are used as inputs by another firm, their quality affects positively its productivity. In other words, technology and Know-How of suppliers are transmitted to their customers through the quality of inputs. 6 Görg and Strobl (2001) and Görg and Greenaway (2004) present an exhaustive survey of the literature on technology spillovers. 7 Details of the construction of backward linkages and forward linkages variables are presented in the next section. 6

7 The intensity of technology spillovers through backward and forward linkages depends on several elements, particularly the nature of those linkages. More precisely we assume that firms are more willing to share their know-how and their technology with their suppliers if the intermediate product supplied is specific to their production process. The more specific and specialised the supplied product is, the higher the quality requirement of the buyer will be and the more specialised the transferred technology will be. If the supplier produces a non-specific intermediate output, (not strategically related to the production process of the buyer), its possibility of learning is weak and limited to general techniques of production. Similarly, if the input is specific to the production process of the firm that buys it, it will have a greater effect on the productivity of this firm. We associate specialised linkages to backward and forward linkages between firms belonging to the same industry, and non-specific linkages to linkages between firms belonging to different industries. To verify the existence of technology spillovers, we estimate the effect of Horizontal and Vertical channels on the total factor productivity of firms. We consider a positive and significant effect as evidence of technology spillovers. To determine the direction of technology diffusion, we split the data into two samples, a sample of foreign affiliates and a sample of domestic firms. We estimate the horizontal and vertical effects of domestic firms on foreign affiliates to verify if foreign affiliates source technology by investing in France, and we estimate the horizontal and vertical effects of foreign affiliates on domestic firms to verify if foreign affiliates in France are a source of technology spillovers. III- Data description and Methodology: The data base used in this paper is derived from the firm annual survey conducted by the French Ministry of Industry. This survey is exhaustive, obligatory and concerns all firms with more than twenty employees. The data base we use covers 7

8 the period and sixteen manufacturing sectors 8. The sectoral classification of firms follows the three digits French classification NAF36. The data base is an unbalanced panel with a number of firms per year varying from firms in 1990 to firms in The survey provides data on the productive activity of firms: output, exports, number of employees, stock of capital, investment, use of intermediate inputs and subcontracting activity conducted and confided by the firm. Realised subcontracting activity stands for the production realized by a firm to the benefit of other firms, thus the firm in question is a supplier. We use this data to construct the backward linkages variable. The confined subcontracting activity refers to the production realized by other firms to the benefit of the firm in question, which is, in this case, a buyer. We use this data to construct the forward linkages variable. All the variables are deflated using sectoral price indices. To distinguish between foreign affiliates and domestic firms we completed our data with financial data provided by the financial liaisons survey conducted by the French national statistic office INSEE. This is an obligatory survey that gives information on the ownership of an enterprise and on relations between an enterprise and its parent company and identifies the home country of the enterprises. The financial liaisons survey is addressed to firms satisfying at least one of these criteria: more than 500 employees, value of stocks higher than 1.2 millions euros, parent company in the previous year, directly controlled by foreign company in the previous year, turnover value higher than 30 millions euros. Since the criteria of the financial liaisons survey are different from those of the firm annual survey and concern relatively large scale firms, some of the small and medium enterprises, present at the firm annual survey, may not be covered by the financial liaisons survey. We define foreign affiliates as firms with thirty percent or more of their capital controlled by foreigners. Generally, and following the definition of the OECD and the 8 Textile, Leather, Wood and paper, Publishing and printing, Chemicals rubber and plastic products, Metallurgy, Motor vehicles, Pharmaceutical, Mechanical equipments, Electric and electronic component, Electric and electronic equipment, Mineral products, Home equipments, Water gas and electricity, Combustibles and fuel, Naval railway and aeronautical construction. 8

9 IMF, the cut-off capital participation rate for the definition of foreign affiliates is set at ten percent. We consider that the ten percent cut-off rate does not represent sufficient control over the activity of a firm. We consider all the firms not covered by the financial liaisons survey, and for which we do not possess information on the ownership structure, as local firms. This lack of information and the criteria of the financial liaisons data base may lead to biased estimation of the presence of foreign affiliates. Table IV presents the employment shares of foreign affiliates in each sector through the panel period. On average, foreign affiliates control twenty per cent of the employment in the French industry and their presence has been growing in the majority of sectors. They are mostly present in the Pharmaceutical and Chemical industries with employment shares of more than forty per cent. To verify the existence of technology transfer through horizontal and vertical channels we estimate the following equation: TFP it = α +β 1 X it + β 1 Z it + d i + d t + ε ijrt. (Eq 1) i, j, and t represent respectively firms, sectors, and time and ε ijrt is a error term, X is a vector of technology spillovers variables, horizontal, backward and forward linkages and Z is a vector of control variables. As control variables, we use the wage bill paid by the firm measured by the natural logarithm of wages to control for the scale of the firm as well as for the qualification of the employees, the Herfindahl concentration index to control for the intensity of competition faced by the firm and the market share of the firm. TFP it represents total factor productivity of firm I at time t. We estimated total factor productivity using the semi parametric estimation suggested by Olley and Pakes (1996) 9, which allows avoiding the selection and the simultaneity problems. In order to estimate the TFP we used data on output, on 9 Details of the semi parametric estimation are presented in Appendix B. 9

10 employment, measured by the number of employees, on the stock of capital, which is equal to the value of fixed assets, on the use of intermediates, which it is equal to the purchased value of intermediates adjusted for changes in stock, and on investment. Horizontal jt is a sector specific variable that represents the foreign (local) presence in sector j at time t and it is defined as the share of foreign (local) firms in the total employment of the sector. Horizontal jt = ( for it lit / i j ) i l j it Where for it (local it ) is a dummy that takes the value one if a firm is a foreign (local) and zero otherwise. The variable Horizontal jt captures the effect of foreign (local) affiliates on their local (Foreign) competitors. A positive coefficient on this variable reflects the existence of horizontal technology transfer. Backward-linkages: these variables capture the effect of firms on their suppliers. We use the amount of subcontracting realized by each firm and reported in the firm annual survey to calculate two Backward-linkages variables: The foreign Backward-linkages variable, which estimates the subcontracting activity addressed to foreign affiliates, is calculated as follows: Foreign Backward-linkages 10 ijt= log[( α k jk Foreign ) Subcontracting kt Where α jk is equal to the proportion of sector j output that is supplied to sector k. The proportions are taken from the input-output matrix at the three digit level of the NAF. Foreign 11 kt represents the foreign presence in sector k. We include the amount of linkages within a sector, in the calculation of the Backward-linkages variables. In fact a significant part of the subcontracting activity takes place in the same sector of activity 12. If we exclude subcontracting activity within the same sector, their effect will be captured by the Horizontal variable and the coefficient on this variable will be biased. A it ] 10 The firm i is a firm of the sector j. 11 Foreign and Local are respectively Foreign Horizontal and Local Horizontal. 12 Particularly when the aggregation of the data is limited at the three digit level. 10

11 significant and positive effect of this variable on the productivity of domestic firms is evidence on the existence of technology transfer from foreign affiliates to local suppliers. In other words, vertical spillovers are not exclusively inter-industrial; they are related to the supplier-buyer relationship. Thus vertical spillovers can take place within the same industry if they are channelled by linkages between suppliers and buyers. The variable Horizontal jt captures technology spillovers channelled through demonstration effects and employees turnover. Since those mechanisms are more frequent within the same industry, the Horizontal jt variable has an intra-industrial dimension. The local Backward-linkages variable, which estimates the subcontracting activity realized to the benefit of local firms, is calculated as follows: Local Backward-linkages ijt = log[( α jk Localkt ) Subcontracting it ] k A positive and significant effect of this variable on the productivity of foreign firms is evidence of the technology diffusion toward foreign suppliers. Forward linkages: these variables measure the extent of technology contained in intermediate products and transferred from suppliers to their customers. We use the confided subcontracting activity reported in the firm annual survey to estimate two backward linkages variables, a foreign backward-linkages variable and a local foreign backward-linkages variable. The foreign Forward-linkages variable is calculated as follows: Foreign Forward-linkages ijt = log[( β k jk Foreign kt ) confided Subcontracting Β jk represents the share of the total inputs of sector j that is supplied by sector k. The β jk proportions are derived from the Input-Output Table. A positive coefficient on this variable is evidence on technology spillovers through backward linkages from foreign suppliers to local buyers. it ] 11

12 The local Forward-linkages variable is calculated as follows: Local forward-linkages ijt = log[( β k jk Local kt ) confided Subcontracting it * Foreign A positive and significant effect of this variable on the productivity of firms is synonymous of technology spillovers through forward linkages from local suppliers to foreign buyers. jt ] Specific Linkages: to test the difference between specific linkages and non-specific linkages we constructed four specific linkages variables, two for backward linkages and two for forward linkages which are defined as follows: Foreign Backward-linkages ijt (intra) = log[( α Foreign ) Subcontracting jj jt it ] Local Backward-linkages ijt (intra) = log[( α Local ) Subcontracting ] jj jt it Foreign Forward-linkages ijt (intra)= log[ β Foreign * confided Subcontracting jj jt it ] Local Forward-linkages ijt (intra)= log[ β Local * confided Subcontracting jj jt it ] Non-specific linkages: We construct four non-specific variables, two for backward linkages and two for forward linkages, which are defined as follows: Foreign Backward-linkages ijt (inter) = log[( k α j kj Foreign ) Subcontracting kt it * Local jt ] Local Backward-linkages ijt (inter) = log[( kj Local kt) Subcontracting k α j it * Foreign jt ] 12

13 Foreign Forward-linkages ijt (inter)= log[( k β j kj Foreign kt ) confided Subcontracting it * Local jt ] Local Forward-linkages ijt (inter)= log[( k β j kj Local kt ) confided Subcontracting it * Foreign The specific linkages variables are not analogous to the Horizontal one; they aim at capturing intra-industrial technology spillovers taking place through subcontracting activities while the horizontal variable is related to intraindustrial spillovers taking place through imitation and employees mobility. jt ] IV- Results: Table one presents the results for the sample of domestic firms. These results show that the presence of foreign affiliates has a significant negative impact on the productivity of domestic firms. This result is contrary to our expectations, however similar impact has been found for the United Kingdom by Driffield and Love (2003). Generally we explained this kind of negative impact as the result of the competition between domestic firms and foreign affiliates 13 [Aitken & Harrison (1999)]. We find no significant impact of foreign affiliates on domestic firms through backward linkages. However when we control for the composition of backward linkages, we find that technology spillovers exists only in the case of inter-industry linkages. Firms may be reticent to transfer knowledge to suppliers from the same industry that may be in the same time close competitors. Moreover, intra-industry spillovers may not need significant investment and learning effort on the part of the supplier thus limiting the spillovers. We also find strong evidence on technology spillovers through forward linkages with foreign affiliates. Table two presents the results for the sample of foreign affiliates. These results are evidence on horizontal spillovers from local firms towards foreign affiliates. The difference between this result and that found in table one may be related to the fact 13 We need to keep in mind that we estimate total factor productivity using data on values not on quantities, thus estimated TFP is very close to firm s mark-up. 13

14 that the entry of foreign affiliates increase the degree of the competition in the industry and the fact that foreign affiliates are more prepared to face this competition given their specific advantages. Moreover, since we measure the foreign presence by employment shares, the negative coefficient in table one may reflect the qualified labour mobility from domestic firms toward foreign affiliates. Backward linkages with domestic firms seem to have a negative impact on the productivity of foreign affiliates mainly in the case of intra-industrial linkages. This result may also be related to competition effects. Domestic firms usually have their established network of suppliers before the entry of foreign affiliates. In this case, foreign affiliates face strong barriers of entry into networks of suppliers and domestic buyers use the competition between suppliers to get the lower possible price [Pack and Saggi (2001)]. Similarly to the case of domestic firms, these results show significant evidence on technology spillovers toward foreign affiliates through forward linkages with domestic firms. In the analysis of technology spillovers adsorptive capacity is an important element. We have assumed that domestic firms in our sample as well as foreign affiliates have sufficient absorptive capacity in order to capture technology spilling over the French economy, however our results show limited extend of such spillovers. In order to take account of the absorptive capacity of firms we create a technology gap variable. This variable is defined as the difference between the average TFP of a sector and the TFP of firms in that sector: Technology gap ijt = AverageTFP jt - TFP ijt We then create a dummy (Gap) taking the value one if the technology gap variable is positive and zero otherwise. This variable allows us to distinguish firms that are less productive than the average of the sector. We consider that those firms have a small absorptive capacity. In tables III and IV we interact the Gap dummy with our variables of interest in order to estimate the impact of absorptive capacity on the extent of spillovers. Table III presents the results for the sample of domestic firms. The results in this table show a considerable impact of absorptive capacity and that spillovers are exclusive to highly productive firms. Table VI presents the results for the sample of foreign affiliate. These results are similar to that in table III. However, even when we control for absorptive capacity, 14

15 the coefficient on Local Backward Linkages remains significantly negative. Probably spillovers from backward linkages cannot overcome the effect of competition faced by foreign supplier. The control for technology gap offers a clarification for the difference of the coefficient on the Horizontal variable between table one and table two. In fact on average, foreign affiliates are more productive that the sector s mean while on average foreign affiliates are less productive 14. V- Conclusion: In this paper we tried to answer the following question: Are foreign affiliates investing in France diffusing technology in the French economy or are they outsourcing technology? We define three channels of technology spillovers, the horizontal channel which refers to technology implicitly spilling over between firms investing in the same sector, the backward linkages channel which refers to the technology transferred from firms to their suppliers and forward linkages channels which refers to the technology transferred from suppliers to their customers and canalised by the intermediate products. We find that foreign firms benefit from their presence in the French industry and from the proximity with domestic firms. We also find that foreign affiliates benefit from buying their inputs from domestic suppliers. However we find negative impact of their supplying activity. Our results also show that domestic firms benefit from supplying goods to foreign affiliates and from buying intermediated products from them. Spillovers through linkages are more present between industries than within an industry since firms tend to limit diffusion of their knowledge to potential competitors. Moreover, we find that absorptive capacity is an essential condition for firms being able to capture knowledge spillovers. For domestic firms as well as foreign affiliates, forward linkages seem to the most efficient channel for technology transfer. In fact, firms operating in the French industry are highly productive and efficient, thus it is normal for their efficiency to be 14 Summary statistics for this variable are presented in table V. 15

16 reflected in the quality of their products. In the same time, considering this efficiency, firms do not need to transfer great amount of knowledge to their suppliers to guarantee the quality of inputs. We expect backward linkages to have greater effect in the case of less competitive suppliers. For the theorists on technology outsourcing, locating R&D activity in the host country is essential to acquire the technology and knowledge developed in that country. Unfortunately, our data base does not provide data on the R&D activity. Thus we were not able to test the effect of R&D activity on technology diffusion towards foreign affiliates and further research is certainly necessary to bring clarifications to that question. 16

17 Reference: Aitken B and A Harrison (1999): Do Domestic Firms Benefit from Direct Foreign Investment American Economic Review 89: Blalock G and P Gertler (2003): Technology from Foreign Direct Investment and Welfare Gains through the Supply Chain mimeo, Cornell University. Dunning J. H (1988): Explaining International Production, London Unwin Hyman. Driffield N and J Love (2003): Foreign Direct Investment, Technology Sourcing and Reverse Spillovers The Manchester School 71(6). Fosfuri A and M Motta (1999): Multinationals without Advantages Scandinavian Journal of Economics 101: Fosfuri A, M Motta and T Rønde (2001): Foreign Direct Investment and Spillovers through Workers Mobility CEPR Working Paper No Glass A and K Saggi (1998): International Technology Transfer and the Technology Gap Journal of Development Economics 55: Glass A and K Saggi (1999): Foreign Direct Investment and the nature of R&D Canadian Journal of Economics 31(1): Görg H and D Greenway (2004): Much Ado about Nothing? Do Domestic Firms Really Benefit from Foreign Direct Investment World Bank Research Observer 19(2): Görg H and E Strobl (2001): Multinational companies and productivity spillovers: A meta-analysis Economic Journal, 111: F723-F739. Griffith R, R Harrison and J V Reenen (2003): Technology Sourcing by UK Manufacturing Firms: an Empirical analysis using Firm-Level Patent Data. Haddad M and A Harrison (1993): Are there Positive Spillovers from Direct Foreign Investment? Evidence from Panel Data for Morocco Journal of Development Economics 42: Jabbour L and J L M Mucchielli (2003): Technology Transfer through Backward Linkages: The Case of the Spanish Manufacturing Industry Jaffe A, M Trajtenberg and R Henderson (1993): Geographic Localization of Knowledge Spillovers as Evidenced by Patent Citations Quarterly Journal of Economics 434:

18 Keller W (2002): Geographic Localization of International Technology Diffusion American Economic Review 92(1). Kugler M (2000): The Diffusion of Externalities from Foreign Direct Investment: Theory ahead of Measurement Discussion Papers in Economics and Econometrics, University of Southampton, U.K. Moran T (2001): Parental Supervision: the New Paradigm for Foreign direct Investment and Development Institute for International Economics, Washington D.C. Neven D and Siotis G (1996): Technology Sourcing and FDI in the EC: an Empirical Evaluation, International Journal of Industrial Organization 14: Pack H and K Saggi (2001): Vertical Technology Transfer via International Outsourcing Journal of Development Economics 46: Olley S and A Pakes (1996): The Dynamics of Productivity in the Telecommunications Equipment Industry Econometrica 64 (6): Serapio M G Jr and D H Dalton (1999): Globalization of Industrial R&D: an Examination of Foreign Direct Investment in R&D in the United States Research Policy 28: Siotis G (1999): Foreign Direct Investment Strategies and Firms Capabilities Journal of Economics and Management Strategy 8: Singh J (2002): Knowledge Diffusion and the Role of Multinational Subsidiaries Unpublished Paper, Harvard University. Smarzynska B (1999): Technological Leadership and the Choice of Entry Mode by Foreign Investors: Evidence from Transition Economies World Bank Working Paper: Smarzynska B (2002): Does Foreign Direct Investment Increase the Productivity of Domestic Firms? In Search of Spillovers through Backward Linkages World Bank Policy Research Working Paper

19 Table I: Sample of domestic firms I II III IV V VI VII VIII IX Horizontal -.29 * * * * * * * * * (Foreign) (.031) (.031) (.031) (.031) (.032) (.032) (.031) (.031) (.031) Foreign Backward linkages (.0003) (.0003) Foreign.0239 *.0231 * Forward (.0004) (.0004) linkages Foreign Backward linkages (intra) 8e-06 (.0004) * (.0023) Foreign Backward linkages (inter) (.0003).0136 * (.0023) Foreign Forward linkages (intra).027 * (.0004).0029 (.0018) Foreign forward.028 * (.0004).025 * (.0018) linkages (inter) Wages.179 * (.0026).163 * (.0026).160 * (.0026).179 * (.0026).179 * (.0026).179 * (.0026).160 * (.0026).160 * (.0026).160 * (.0026) Market share * (.629) * (.63) * (.624) * (.629) * (.629) * (.629) * (.629) * (.628) * (.628) Herfindahl * (.145) * (.144) * (.144) * (.145) * (.145) * (.145) * (.144) * (.144) * (.144) Time Yes Yes Yes Yes Yes Yes Yes Yes Yes dummies R Nb of observations *: indicates difference in means is significant at the 1% level **: indicates difference in means is significant at the 5% level ***: indicates difference in means is significant at the 10% level. 19

20 Table II: Sample of foreign affiliates I II III IV V VI VII VIII IX Horizontal.538 *.497 *.525 *.539 *.536 *.547 *.487 *.506 *.537 * (Local) (.080) (.077) (.08) (.08) (.08) (.08) (.077) (.077) (.078) Local Backward linkages ** (.0008) ** (.0008) Local.0119 * * Forward (.001) (.001) linkages Local Backward linkages (intra) ** (.0008) ** (.0038) Local Backward linkages (inter) *** (.0009).0064 (.0039) Local Forward linkages (intra).013 * (.0011) ** (.008) Local forward.014 * (.0011).036 * (.008) linkages (inter) Wages.048 * (.006).04 * (.0062).038 * (.0064).048 * (.006).048 * (.0063) 048 * (.0063).038 * (.0062).038 * (.0062).038 * (.0062) Market share * (1.23) * (1.21) * (1.23) * (1.23) * (1.23) * (1.23) * (1.21) * (1.21) * (1.21) Herfindahl * (.423) * (.414) * (.422) * (.423) * (.424) * (.424) * (.414) -3.2 * (.413) * (.414) Time Yes Yes Yes Yes Yes Yes Yes Yes Yes dummies R Nb of observations

21 Table III: Sample of domestic firms I II III IV Horizontal.566 * (Foreign) (.031) Gap*Horizontal * (.007) Foreign Backward.0025 * (.0004) linkages Gap* Foreign * Bwd Foreign Forward linkages Gap* Foreign Fwd Foreign Backward linkages (intra) Gap* Foreign Bwd (intra) Foreign Backward linkages (inter) Gap* Foreign Bwd (inter) Foreign Forward linkages (intra) Gap* Foreign Fwd (intra) Foreign forward linkages (inter) (.0004).047 * (.0004) * (.0004) ** (.0034) ** (.0033).0289 * (.0033) * (.0035) * (.0025).009 * (.002) * (.002) Gap* Foreign Fwd (inter) * (0.002) Wages.126 * (.0024).119 * (.0026).164 * (.0026).121 * (.0026) Market share * (.588) * (.588) * (.629) * (.628) Herfindahl * (.135) * (.135) * (.145) * (.137) Time dummies Yes Yes Yes Yes R Nb of observations

22 Table IV: Sample of foreign affiliates: I II III IV Horizontal (Local).472 * (.031) Gap*Horizontal * (.008) Local - Backward.0023 ** linkages (.0008) Gap* Local Bwd (.001) Local Forward.0231 * linkages Gap *Local Fwd Local Backward linkages (intra) Gap* Local Bwd (intra) Local Backward linkages (inter) (.001) * (.0009).004 (.004) * (.0067).0018 (.004) Gap* Local Bwd (inter).004 (.001) Local Forward linkages (intra) (.008) Gap* Foreign Fwd (intra) (.007) Local forward linkages (inter).039 * (.008) Gap* Local Fwd (inter) * (.007) Wages.009 * (.005).011 * (.006).04 * (.006).012 * (.0058) Market share 33.5 * (1.09) 31.6 * (1.14) * (1.21) * (1.12) Herfindahl * (.37) * (.39) * (.41) -3.4 * (.38) Time dummies Yes Yes Yes Yes R Nb of observations

23 Table V: Summary statistics of the main variables : Sample of foreign affiliates Variables Obs Mean Std Min Max Deviation Output e+07 Investment Capital e+07 Labour TFP Subcontracting e+07 Confined subcontracting Technology gap Sample of local firms Output e+08 Investment e+08 Capital e+08 Labour TFP Subcontracting e+07 Confined e+07 Subcontracting Technology gap The unit measure of the variables is thousands of French francs. 23

24 Table VI: Foreign presence at the sectoral level: Sector Textile Leather Wood and paper Publishing and printing Chemicals, rubber and plastic products Metallurgy Motor vehicles Pharmaceutical Mechanical equipments Electric and electronic components Electric and electronic equipments Home equipments Mineral products Water, gas and electricity Combustibles and Fuel Naval railway and aeronautical construction

25 Figure 1: Horizontal technology spillovers Firm A introduces modern technologies into its industry through innovation. Demonstration effect and imitation effort by competitor firms allow the diffusion of those technologies into the industry. Moreover, through professional training, firm A transmits part of its knowledge to its employees. This knowledge becomes accessible to competitor firms if they hire employees previously trained by firm A (Employees turnover). 25

26 Figure II: Vertical spillovers Backward spillovers Forward spillovers Technology spillovers through backward linkages take place when the buyer (Firm A) facilitates the access, for its supplier, to modern technologies and when it assists the supplier with professional training and human capital formation. Moreover, backward linkages may affect positively the productivity of the supplier as the buyer imposes quality requirements and exercises quality control, thus pushing the supplier to invest in the upgrading of its technologies and production process. Forward linkages help the diffusion of technology from the supplier to the buyer (Firm A) as the technology of the supplier, embodied in the inputs, become accessible to the buyer. 26

27 Appendix A: The main aggregates of the French economy and the French innovative activity: Top 10 GDP countries (Billions of Euros) United States 9728 Japan Germany United Kingdom France Italy Spain Netherlands Belgium Sweden Source: Eurostat 2004 Top five R&D expenditures European Countries (Billion of Euros) Patents Number (2002) Germany France United Kingdom 14.6 Italy 7.2 Spain Germany France United Kingdom Netherlands Sweeden Spain Source: Eurostat

28 Labour force qualification (2003) Labor productivity index (2002) Ireland France United Kingdom Finland Sw eden France Netherlands United States Italy Germany Source: OECD 2003 * Source: European Commission 2002 FDI Inflows 2003 (Billions of Dollars) Germany United Kingdom Italy Netherlands Ireland Spain Belgium United States France 47 Luxembourg Source: Unctad ( World Investment Report) 2004 * Labour qualification is measured as the number of years sciences and technology graduates per thousands of people in the same age category. 28

29 FDI Outflows 2003 (Billions of Dollars) Sweden Canada Spain Japan Netherlands Belgium United Kingdom France Luxembourg 96 United States Source: Unctad (World Investment Report)

30 Appendix B: The Olley-Pakes Semiparamatric Estimation Olley and Pakes (1996) propose a model of firm behaviour to control for the selection and the simultaneity problems. The simultaneity problem arises because productivity shocks are unobservable for the econometrician but are known to the firms when they choose their inputs (Marschak and Andrews 1944). The firms knowledge of their productivity makes it more appropriate to consider inputs as endogenous variables (Griliches and Mairesse 1995). The estimation of productivity by ordinary least squares (OLS) considers labour, capital and other inputs as exogenous variables and may lead to a biased estimated coefficient. The selection problem is related to the entry and exit of firms. Firms decide whether to exit the market or to continue their activity after considering their expected productivity and profitability. Expectations of productivity are partially determined by current productivity. The result is that an exit decision depends on the firms perception of their productivity. Traditionally, econometricians have dealt with entry and exit of firms by reducing the data set to a balanced panel. Restricting the analysis to a balanced panel does not take into account the endogeneity of the exit decision and generate a selection bias. At the beginning of every period, a firm has to decide whether to exit the market or continue in operation. If it decides to continue, it chooses variable inputs (labour and intermediates) and a level of investment. The investment and the current capital value determine the capital stock at the beginning of the next period. K t+1 = (1- δ) K t + I t. (1) Where δ represents the rate of depreciation of capital, K represents capital and I represents investment. Firms will exit the market if the expected value of continuing operations is less than the market value of the assets. The expected value of continuing operations depends, among other things, on firm s own state variables. State variables consist of capital stock K and productivity ω. Olley and Pakes (1996) show that firms will continue in operation if the productivity exceeds a threshold level ϖ t (k t ). X t = 1 if ω t ϖ t (k t ) and 0 otherwise. (2) 30

31 We assume the following production function: Y it = α + β L L it + β K K it + β M M it + ω it + η it. (3) Whereas Y it is the log of output of firm i at time t, L it the log of its labour input, K it the log of its capital input and M it the log of its materials input. ω it represents productivity and η it is either a measurement error or a shock to productivity. Labour and materials represent variable factors and their amounts are affected by the current level of productivity. Capital is a fixed factor and it is only affected by the distribution of productivity, conditional on information at time t-1 and passed values of`ω. Pakes (1994, theorem 27) shows that investment is strictly increasing in ω for each K; thus investment can be used as a proxy for the shock conditional on fixed variables. As a consequence, the investment function, I t = I t (ω t, K t ), can be inverted to ω t = h t (I t, K t ). (4) The estimation process consists of three stages; in the first stage, we estimate the coefficient on the variable inputs (labour and materials), in the second stage we use a probit model to control for the selection bias and in the third stage we estimate the coefficient on fixed factors (capital) conditional on the prior period s shock and the probability of exit. Stage 1: Substituting (4) in (3), we have: Y it = β L L it + β M M it + φ t (K it, I it ) + η it. (5) Where φ t (K it, I it ) = α+ β K K it + h t (I t, K t ). To estimate the partially linear model in (5), we regress Y it on labour, materials and a third order polynomial in (K it, I it ) [with a full set of interaction]. Since the error term η it is not correlated with the variable inputs, the estimation of equation (3) gives unbiased coefficient on labour and materials. Stage 2: At this stage we estimate the survival probabilities: Pr {X it+1 = 1 ϖ it+1 (k it+1 )} =Pr {ω it+1 ϖ it+1 (k it+1 ) ϖ it+1 (k it+1 ), ω it+1 } = ρ it {ϖ it+1 (k it+1 ), ω it+1 } = ρ it (k it, I it ) 31