Introduction. Holger Gorg 1. Kiel Institute for the Wof'ld Economy and Christian-Albrechts-University of Kiel, Germany

Transcription

1 Introduction Holger Gorg 1 Kiel Institute for the Wof'ld Economy and Christian-Albrechts-University of Kiel, Germany "If a prince bases the def ence of his state on mercenaries he will never achieve stability or security. For mercenaries are disunited, thirsty for power, undisciplined, and disloyal; they are bmve among their friends and cowa1 ds before the enemy; they have no fear of God, they do not keep faith with their f ellow men; they avoid defeat just so long as they avoid battle; in peacetim e you are despoiled by them, and in wartime by the enemy. 'l'he reason for all this is that there is no loyalty or inducement to keep them on the field apart from the little they are paid, and this is not enough to make them want to die f or you. They are only too ready to serve in your army when you are not at war; but when war com es they either desert or disperse." Niccolo Machiavelli: The Prince (tmnslated by George Bull, Penguin Publishers, 1995, p. 38) Foreign direct investment (FDI) by multinational enterprises (MNEs) is arguably one of the most visible and important aspects of economic globalisation. As the latest United Nations' World Investment Report shows, annual flows of FDI amounted to around $1.3 trillion in 2014 with the total stock rising to around $26 trillion. To put this in 1 The a ut hor is also affiliated with the Tuborg Centre for Research on Globalisat ion and Firms at Aarhus University, t he Centre for Research on Globalisation and Economic Policy at Nottingham University, a nd IZA. He is grateful to Aoife Hanley for helpful comments on a n earlier draft. XV

2 xvi Multinational Enterprises and Host Country Development perspective, FDI stocks stood at "merely" $7.2 trillion in 2000 (UN 2015). One may wonder whether multinationals behave like the mercenary troops described by Machiavelli. Are they as unreliable for host countries as mercenaries are for a prince? Many globalization critics may be prone to answer this question in the affirmative. Many governments, however, seem to contradict this negative view assuming that inward FDI can bestow large potential benefits to the economy. A fully satisfactory answer to this question would, of course, encompass detailed analyses of many economic, environmental, political and sociological issues, well beyond the scope of a single discipline, let alone this volume_ From an economic point of view, FDI by multinational enterprises channels not only new investment to the host country thus boosting national income. There is an additional important benefit, namely the inflow of new foreign knowledge and technology. This latter aspect of FDI may lead to spillovers to the local economy, resulting in higher productivity and thus accelerating overall economic development of the host country. These external effects from multinationals to domestic firms are generally referred to as "productivity spillovers" or "technology spillovers" (see Paper II.2 in this volume for a detailed review). 2 On the other hand, of course, multinationals may crowd out domestic entrepreneurship with potential negative effects for the economy. Host count ry governments seem to assume that there are generally positive effects on economic development. T his sanguine view of FDI is evidenced in the effort made by governments to actively attract multinational companies and encourage foreign direct investment into their economies. Policies applied by governments range from creating more liberal investment environments to providing substantial public subsidies. While it is difficult to obtain reliable data on such financial incentives, Girma et al. (2001) report that the UK 2 There is a related literature which investigates whether inward FDI fosters economic growth using macro-level data. For example, Borensztein et al. (1998) a nd Alfaro et al. (2003) show that inward FDI can have positive growth effects if countries have a certain level of human capita l or well developed financial markets.

3 Introduction xvii government provided an estimated $30,000 per employee to attract Samsung to the northeast of England in the late 1990s and $50,000 per employee to attract the German multinationals Siemens. It seems reasonable to ask whether such policy intervention can be justified. Does the assumption that FDI brings with it growth and development enhancing spillovers really stand up to scrutiny? This volume brings together a number of empirical studies showing that, contrary to Machiavelli's denunciation of mercenaries, multinational enterprises can indeed carry benefits to their host country. Outlining the Issues To orientate the reader, this review commences by presenting the basic intuitive framework as to why multinationals can be predicted to benefit host country firms or economic growth. In general, the availability of new foreign knowledge through inward FDI may benefit domestic firms as these can learn the technology from the multinationals, allowing them to upgrade their own production process and as a result to increase their productivity. The theoretical argument for why one may expect such productivity spillovers from multinationals is straightforward. Given the foreign multinationals' limited knowledge of the local market, and the distance from their parent firm, they are generally at a disadvantage compared with local firms in the host country (Caves, 1996). Hence, multinationals will only be able to locate profitably abroad if they possess some sort of offsetting advantage. This takes the form of a "firm specific asset" (FSA), such as a superior production technique, know-how or management strategy. The impacts of FSAs are somewhat similar to those of a public good and the possession of such assets enables the firm to locate profitably abroad (Markusen, 2002). Most importantly, these FSA can be transferred at low or zero cost between subsidiaries of the same firm. 3 3 Examples of such firm specific assets are know-how, brand names or other intangible assets that have the characteristics of public goods, i.e., t hey can be easily transferred

4 xviii Multinational Enterprises and Host Country Development The possibility of productivity spillovers then arises because multinationals may find it difficult to protect a leakage of the FSA in the host country, as other firms may somehow "learn" about, for example, the production technology or management strategy. They can then apply it in their own firm, thus allowing them to improve their technology and ultimately productivity. The public good characteristics imply that once the FSA is out in the open it can be used by other firms, due to its de facto non-rivalrous and non-excludable characteristics. 4 The inability of the multinationals to protect these intangible assets is due to a number of reasons. Firstly, domestic firms may just imitate the multinationals in terms of products, production or management techniques. Secondly, labour may move from multinationals to domestic firms, taking with them some of the knowledge of the FSA. Thirdly, domestic firms supplying to or purchasing inputs from multinationals may be exposed to the superior technology and, hence, be able to upgrade their own production techniques. Fourthly, competition from multinationals may force domestic rivals to update production techniques to become more productive and competitive. This is frequently referred to as a "competition effect." As Aitken and Harrison (1999) point out, however, this competition effect may also reduce productivity in domestic firms, if multinationals siphon off demand from their domestic competitors, forcing domestic firms to reduce output and, hence, productivity. within the firm, from the parent to subsidiaries abroad. The argument about FSA is in line with recent heterogeneous firms models which assume that multinationals face specific sunk costs and hence, only the most productive will choose to become multinat ionals, less product ive firms export a nd the least productive ones serve purely the domestic market (Helpman et al., 2004). 4 T his is the t heoretical argument genera lly referred to in the literature. Markusen a nd Venables (1999) point out a different theoretical channel for spillovers. They highlight the role of pecuniary externa lities, rather tha n technological externa li ties. T he former do not affect t he production function of the benefit ing firm, rather they affect the profit function via cost reductions or increased revenues. In a nutshell, increases in output by multinationals lead to an expansion of dema nd for intermediate products supplied by domestic suppliers. T his increase allows these suppliers to produce at a more efficient scale, thereby reducing average costs. T his will ultimately reduce the price t hat multinationals and other domestically based final good producers pay for intermediates.

5 Introduction xix While this appears to be a very intuitive and plausible theoretical proposition, t he question is whether it also stands up to empirical scrutiny_ The first step in evaluating this is to start off by considering the importance of firm specific assets. If these are indeed crucial then one should observe that multinationals apply a comparatively "superior" technology vis-a-vis that applied by their domestic competitors. The use of technology is, of course, difficult to observe and measure, but it should arguably be correlated with a number of other measurable firm characteristics, and in particular with productivity. If a firm uses a superior technology one would, all other things equal, expect the firm to demonstrate higher levels of labour or total factor productivity than its rivals. The chapters in Part I look at this issue and consider productivity, technology and wages in multinational enterprises compared to purely domestic host country firms. What we can take from this discussion thus far is that there are plausible theoretical reasons for why there may be spillovers from multinationals, and that the evidence generally also points to the prevalence of "superior characteristics" in multinationals - hence, the potential for spillovers should be taken as given. The next step is then to look for evidence that domestic firms do indeed appropriate productivity spillovers from multinationals. When it comes to empirical analysis, it is important to point out that spillovers are difficult to measure since, as Krugman (1991) puts it, "knowledge flows [... ] leave no paper trail by which they may be measured and tracked" (p. 53). The approach adopted in the empirical literature therefore largely avoids the (arguably difficult to answer) question as to how precisely productivity spillovers actually take place, but focuses on the simpler issue of whether or not the presence of multinationals affects productivity in domestic firms. A large empirical literature has arisen that attempts to gauge the extent of spillovers (summarized in Papers 11.1 and 11.2 in this volume and, more recently in Irsova and Havranek, 2013 and Havranek and Irsova, 2011). Most econometric analyses are undertaken in a framework in which labour productivity or total factor productivity of domestic firms is regressed on a range of independent variables. To measure productivity spillovers from foreign multinationals

6 xx Multinational Enterprises and Host Country Development a variable is included which proxies the extent of foreign firms' presence in the same or vertically related industries in order to gauge the importance of multinationals in the industry. This variable is usually calculated as the share of employment or sales in multinationals over total industry employment/sales in a given sector. If the regression analysis yields a positive and statistically significant coefficient on the foreign presence variable, this is taken as evidence that domestic firms have appropriated positive spillovers from multinationals. The papers collected in P art II of this volume present various approaches towards looking for spillovers using firm level data. While most of the literature on host country effects of multinationals has focused on measuring productivity spillovers using this "traditional approach," there are also some attempts to explore other issues. Part III of this volume presents a number of papers that look specifically at the mechanisms through which spillovers may work, namely, worker movement and linkages at the micro level. Also, Part III contains papers presenting alternative ways of assessing the impact of multinationals on host country development, looking at the impact of multinationals on entry and survival of domestic firms. We now provide a brief overview of the chapters included in this volume. Part I - Firms? Are Multinationals "Better" Than Domestic P art I includes chapters that attempt to evaluate whether affiliates of multinationals are indeed performing better than comparable domestic firms. One would expect this to be the case if the multinational possesses some sort of firm specific asset and, hence is able to apply a superior technology. The chapters in this section look at superior performance in terms of productivity, innovation, survival and wages. An evaluation of such performance differences is not trivial. Quite to the contrary, there is a fundamental problem in identifying the performance difference that is attributable to a firm's multinationality status, per se. Multinationals may be attracted to more technology intensive industries. These technology intensive industries may be

7 Introduction xxi simultaneously more productive and also be associated with wage premia. It should therefore be evident that any regressions that attempt to explore performance differentials are subject to an endogeneity problem. Consequently, it is difficult to impute and interpret performance differentials between foreign and domestic firms. Controlling for some industry and firm characteristics might go some way towarcls reducing this bias, though the inclusion of all possible relevant control variables is a difficult, if not an impossible, task. Chapter 1 attempts to overcome this problem by analyzing the effects of an acquisition of a domestic establishment by a foreign multinational enterprise on the resulting productivity growth in the acquisition target. Assuming that an acquisition does not change any of the main characteristics of the takeover target (at least in the short run) a possible effect of the foreign acquisition on productivity in the domestic target can be attributed to the change in ownership from domestic to foreign. The acquisition decision is unlikely to be random, however, which may again bias the estimated effect. In order to deal with this problem, the chapter uses a difference-in-differences propensity score matching approach to identify the average effect of foreign acquisitions on productivity growth in the domestic target. Chapter 1 decomposes the possible productivity advantage into a technology and a scale effect. While the former represents the assumed technological advantage of a foreign multinational, the scale effect is due to the fact that, as output expands, workers and firms gain proficiency at producing particular products and are, consequently, able to improve their productivity. A distinction of these two effects is not only academically interesting but also highly policy relevant. Government policies are generally informed by the assumption that multinationals bring in new technologies. They are, hence, predicated on the existence of technical efficiency in multinationals. If, however, scale efficiency is the dominant component of foreign firms' productivity premia, such policies may be misguided. The analysis uses establishment level data for the UK, focusing on selected manufacturing sectors over the period 1980 to He results of the econometric estimation indicate substantial sectoral heterogeneity but also clearly show that there is a positive impact

8 xxii Multinational Enterprises and Host Country Development of ownership change on productivity growth. This is predominantly due to changes in technical efficiency and not in scale effects. This finding underlines the importance of firm specific assets being transferred to the takeover target. The analysis also shows that the preacquisition TFP level of the takeover target plays a role in mediating the rate of technology transfer from the multinational parent companies. Furthermore, the estimated productivity growth effects are not merely limited to the year of acquisition but tend to persist through time. Chapter 2 uses a similar identification strategy of analysing the effects of foreign acquisitions of domestic targets with a differencein-differences propensity score matching approach. The focus of the paper is on the wage differential between foreign and domestic firms, however. Specifically, the paper estimates the impact of a foreign acquisition on average wages in the takeover target. The paper distinguishes the effects for the wages of skilled and unskilled workers, and allows for different effects of acquisitions depending on the nationality of the acquirer - namely, whether it is from the US, the EU or the rest of the world. The analysis is carried out using establishment level data for the UK, using data for the whole manufacturing sector over the period 1980 to The econometric exercise shows robust and economically significant wage effects for both skilled and unskilled workers acquired by US multinationals. For skilled workers, the estimated wage effect is roughly 8 percent - i.e., the average wage for skilled workers increases by 8 percent as a result of an acquisition. The impact on unskilled wages is nearly 13 percent two years after the acquisition. In stark contrast, no evidence is found for any effect on wages, skilled or unskilled, following acquisition by EU based multinationals. Finally the paper also finds some evidence for positive effects on wages for unskilled workers (not for skilled workers) following acquisition by multinationals from the rest of the world. The estimated magnitudes are lower, however, than for US acquisitions. Many papers in the literature document a wage premium for foreign owned multinationals. The causes of these premia are largely unexplored, however. Chapter 3 looks at a potential mechanism,

9 I ntroduction xxiii arguing that differences in the provision of training by multinationals and domestic firms may explain wage differentials. Multinationals may have stronger incentives to provide training in order to equip workers with the necessary knowledge to implement the superior technology available in the firm. In a simple theoretical model, the chapter shows that differences in the provision of training my lead to higher wage growth for workers in foreign compared to domestic firms, while starting wages of workers do not necessarily differ. The empirical analysis is carried out using a small linked employer-employee dataset for manufacturing firms in Ghana for the year The data was collected as part of the Regional Programme for Enterprise Development ( RP ED), an extensive data collection effort for a number of African countries, organised initially by the World Bank. The empirical analysis shows that there are no statistically significant differences in starting wages of trained and untrained workers in foreign and domestic firms, once other individual and firm characteristics are controlled for. However, workers receiving on-the-job training in foreign firms experience higher wage growth than workers being trained in domestic firms. By contrast, there does not appear to be any difference in wage growth between workers not receiving training in either foreign or domestic firms. Training, hence, seems to be crucial for determining wage differences. This suggests that host country workers may benefit from the presence of foreign owned firms in terms of higher wages, particularly if the foreign multinationals provide training. Chapter 4 looks at a more direct measure of technology, namely R&D expenditure by multinationals. This chapter investigates what happens to the R&D activity of firms in the host country once they are acquired by foreign multinationals. If foreign multinationals upgrade technology in the host country, one may expect a positive post-acquisition effect. However, it may also be the case that multinationals tend to siphon off R&D activity to their home country. In this case, there may be a negative effect on R&D activity in the host country. The chapter also considers whether the postacquisition effect depends on whether the take-over target is itself a

10 xxiv Multinational Enterprises and Host Country Development multinational firm, or whether it operates only on the host country market. The empirical analysis is carried out using firm level data for Swedish manufacturing, a country that has a large number of domestic multinationals, many of which were taken over by foreign owners in recent decades. The data relates to the period The empirical strategy is similar to that used in Chapters 1 and 2, combining propensity score matching and difference-in-differences in order to identify the effect of a foreign acquisition on R&D activity in the target firm. The estimation results provide robust evidence that there are, on average, strong positive effects from foreign acquisitions on the R&D performance of Swedish acquisition targets. These effects are stronger for the acquisition of domestic non-mnes than for Swedish MNEs. The point estimates show increases in R&D intensity by between 5 to 10 percent after a foreign acquisition. This suggests that foreign acquisitions may be an important way to generate new knowledge and to contribute to boosting the level of technology in the host economy. Chapter 5 examines a different proxy for technology, measuring innovation output as the share of output due to new processes or technology in relation to total output. The aim of the paper is to establish whether this share is higher in firms with foreign ownership compared to domestic firms. Anticipating the ideas that will be further discussed in Part II of this volume, the paper also allows for spillovers from the presence of foreign multinationals in a sector/ region on such innovation activity in firms. The estimation strategy is an instrumental variable Tobit model, allowing for the bounded nature of the dependent variable (output share lies between zero and one) and the possible endogeneity of the foreign ownership variables. The empirical estimation is implemented using firm level panel data for Chinese manufacturing, covering the period 1999 to a period that witnessed a strong influx of foreign direct investment into the Chinese economy. Results show that firms with foreign capital participation innovate more than others. This is in line with the idea that they have

11 Introduction xxv superior technology and a firm specific asset. Inward FDI at the sectoral level is positively associated with domestic innovative activity only if firms engage in their own R&D activities or have good access to domestic finance. These results again suggest that an influx of foreign multinationals may lead to technology upgrades in the host country. If multinationals apply superior technology, they may also be expected to be better able to compete in the market. This, ceteris paribus, should enable them to have better survival prospects than competitors with lower technology levels (i.e., domestic firms). The corollary to this is that if one compares the survival probabilities of foreign and domestic firms in a host country, the former would be expected to exhibit higher survival probabilities. However, critics of multinationals frequently argue that these firms are inherently "footloose" and ready to move their facilities to other countries in the wake of an adverse shock. If this were true, one would expect the survival probabilities of foreign multinationals to be lower than those of comparable domestic firms. Chapter 6 looks at this question, using firm level data for manufacturing in the Republic of Ireland. The dataset is for the period 1973 to Given that the data covers virtually the whole population of establishments, firm entry, exit and survival can be measured quite precisely. Chapter 6 conducts a survival analysis using Cox Proportional Hazard models. The results show that, controlling for a multitude of firm level characteristics, foreign firms have higher exit probabilities, i.e., lower rates of survival than comparable domestic firms. The point estimates suggest that the exit probability is about 40 percent higher for multinationals. While this suggests that multinationals may indeed be more 'footloose' than comparable domestic firms, an extension of the analysis shows that employment generated in surviving multinationals is more likely to persist than in domestic firms. Since the identification strategy in Chapter 6 really only amounts to a simple comparison of survival rates between foreign and domestic firms, it cannot say much about a causal effect of foreign ownership. Chapter 7 digs somewhat deeper into this issue. It considers

12 xxvi Multinational Enterprises and Host Country Development foreign acquisitions (i.e., a change in ownership) and estimates the effect of the foreign acquisition on the survival probability of the takeover target. Possible endogeneity of the acquisition decision is allowed for in an instrumental variables set up. The analysis is conducted using Swedish firm level data, similar to the data used in Chapter 4. The estimation results show that foreign acquisition has an effect on plant survival only if the takeover target is a domestic exporter. If the target was not operating on international export markets, no effect can be estimated. One possible explanation for this is that firms that are already active on export markets use higher levels of technology and are more productive. Accordingly, domestic exporters are better able to absorb any new knowledge introduced by the foreign acquirer and improve their performance. Depending on whether the acquisition is horizontal (i.e., within the same industry) or vertical, acquired exporters have a percent or 6-8 percent higher survival probability, respectively, compared with plants of non-acquired firms. In an extension to the analysis, where the paper investigates the impact of the acquisition on employment growth, it is found that employment growth is also higher in takeover targets that were exporters prior to acquisition. T here is no evidence of any negative effects for foreign acquisitions, neither on survival nor on growth. Accordingly, the evidence using the Swedish data demonstrates that fears for the sustainability of domestic industry with increasing foreign acquisition rates appear unfounded. Taken together, the evidence in Part I of the volume indicates that multinationals can have positive direct effects for the host country. Multinationals tend to be more productive - hence, an influx of FDI may improve the host country's aggregate productivity performance. They also tend to pay higher wages than domestic firms, are more likely to conduct H.&D and innovate, and may improve the survival and growth performance of domestic firms in the event of a takeover. All of this is in line with the theoretical idea that multinationals possess a firm specific asset, conferring an advantage vis-a-vis domestic competitors. Whether this firm specific asset confers further benefits

13 Introduction xxvii for the host country through indirect "spillover" effects on domestic firms is an issue that is investigated in Part II of the volume. P art II - Searching for Productivity Spillovers: The "Traditional Approach" T he empirical literature on measuring productivity spillovers from foreign direct investment started with Caves (1974), Globerman (1979) and Blomstrom and Persson (1983). They used cross section data for Australia, Canada and Mexico, respectively, and found positive associations between the presence of FDI in an industry and domestic firms' productivity in that industry. Since then, numerous studies have appeared that have refined and extended this "traditional approach," using mostly firm level data for different countries and time periods. Chapter 8 presents an early meta-analysis of studies on productivity spillovers, including papers up to the late 1990s. The metaanalysis shows that papers based on cross-section data tend to find that FDI is positively associated with domestic productivity, while studies using panel data are less likely to find such evidence. A possible explanation is that cross section data do not allow researchers to adequately control for unobserved industry specific effects. A positive association between industry level FDI and domestic productivity in the sector may just be due to multinationals locating in highproductivity sectors (as would be expected given their firm specific assets) even in the absence of any productivity spillovers taking place. The use of panel data helps to mitigate this problem, by controlling for unobserved (time invariant) industry or firm characteristics. A more traditional verbal review of the literature is supplied in Chapter 9 which also includes papers up to the early 2000s. T he review also concludes that evidence for positive spillovers is, at best, mixed, and considers policy options that governments may use to foster spillovers. A reason for not finding productivity spillovers in studies using panel data may be that they are searching for spillovers in the wrong place. The traditional approach is to look at intra-industry

14 xxviii Multinational Enterprises and Host Country Development (horizontal) spillovers. In other words, the question is whether the productivity of domestic firms depends on FDI in the same sector. However, in the same sector, negative effects of competition may well outweigh any potentially positive spillover effects. It may be more fruitful to investigate whether the productivity of domestic firms is affected by multinationals in the context of a vertical linkage. Specifically, a linkage describes the case where multinationals buy intermediates from domestic firms or, alternatively, where multinationals sell intermediates to domestic firms. In other words, it may be important that there is a vertical linkage, either backward (if the local firm supplies to multinationals) or forward (if the local firm buys inputs from multinationals). These linkages have increasingly been investigated, starting with Driffield et al. (2002) and J avorcik (2004). Chapter 10 conducts such a "traditional" spillover study, allowing for horizontal as well as vertical spillovers through FDI. In line with the literature, the latter are measured by the presence of multinationals in vertically related (backward and forward) industries. The chapter also considers two other issues. Firstly, it allows the effect of spillovers to differ across domestic firms by taking into account firms' absorptive capacity. This is measured as a firm's productivity relative to the industry leader, similar to the approach adopted by Girma (2005). It also distinguishes the effects of FDI on domestic exporters and non-exporters separately. The rationale for this distinction is the expectation that competition effects from multinationals are different for domestic firms in these two exporter categories. Secondly, the chapter allows for the heterogeneity of FDI by distinguishing spillovers from primarily export-oriented multinationals and those whose sales are aimed at local markets. Those "domestic market oriented multinationals" may pose stronger competition for local firms than their export-oriented counterparts. The empirical analysis is based on firm level data for the United Kingdom and covers the years 1992 to Results point to the general conclusion that the export orientation of both domestic and foreign multinationals is relevant for productivity spillovers. Specifically, new insights are gained when one considers the export orientation of foreign MNEs. There is robust evidence for positive horizontal

15 Introduction xxix spillovers from export-oriented multinationals only. However, only domestic-market-oriented MNEs generate positive spillovers through backward linkages for both domestic exporters and non-exporters. In general, the evidence presented in the paper underlines the importance of buyer-supplier linkages for productivity spillovers. The reliability of the evidence on spillovers from traditional spillovers studies, be they horizontal or vertical, crucially depends on the plausibility of the spillover proxies employed. The empirical strategy for measuring backward linkage spillovers has generally consisted of using foreign multinationals' output shares in each sector as a measure of the total multinational demand for inputs in that sector. Moreover, host country input- output tables can be used to proxy how this demand is distributed across sectors. Such a measure is based on a number of implicit assumptions, such as (i) that multinationals use domestically produced inputs in the same proportion as imported inputs, (ii) that foreign firms have the same input sourcing behaviour as domestic firms, (iii) that all MNEs have the same input sourcing behaviour irrespective of their origin, and (iv) that multinationals' demand for locally produced inputs is proportional to the share of output produced by MNEs in a given sector. Chapter 11 discusses each of these assumptions - in particular their plausibility - and proposes alternative spillover proxies that do not rely on such strong assumptions. These alternative measures are then used in an analysis of spillovers from backward linkages using firm level panel data for Ireland over the period The empirical results in the paper show clearly that the choice of backward linkage measure matters greatly in determining whether or not FDI is beneficial to the host economy. The preferred measure for backward linkages, which uses the input-output table for the multinational's home country rather than its host country, provides strong evidence for backward spillovers. No such evidence is found using the standard measure generally employed in the literature. The results are robust to controlling for the potential endogeneity of the spillover variables using instrumental variables/ GMM techniques. The results point to the conclusion that the discovery of spillovers from backward linkages is highly sensitive to assumptions made

16 xxx Multinational Enterprises and Host Country Development about the similarity of sourcing patterns for domestic and multinational firms. The assumption of perfect comparability in sourcing patterns goes against the very premises underlying the search for spillovers arising from FDI, namely that foreign multinationals, by virtue of their firm specific asset, differ from their domestic counterparts. Analogously, it is likely that results for other studies might a.lso ~hange once proxies of backward linkages are used that relax this strong assumption of similar sourcing patterns. After all, in any host country there are multinationals from different home countries with arguably different technologies and input sourcing behaviour. These differences may impinge on potential backward spillovers arising from FDI. Chapter 12 links Part I and Part II by addressing two crucial problems related to the identification of direct effects and the traditional way of estimating spillovers. Firstly, from Part I we know that multinationals perform better than domestic firms. However, the identification strategies generally used in papers estimating such direct effects are usually based on the assumption that there are no indirect effects. In other words, when comparing domestic and foreign firms, estimation approaches typically assume that the presence of foreign firms exerts no indirect effects on domestic firms, and vice versa. Yet, Part II and the related literature reveal that indirect effects do matter. This casts some doubt on the identification strategies used when estimating direct effects. Secondly, the spillovers literature in Prut II is generally quite silent about potential selection problems, i.e., the fact that multinationals select their location and that this location selection aspect might be correlated with the productivity of domestic firms. Chapter 12 proposes and implements a unified framework to estimate direct and indirect effects from foreign ownership on firm level productivity. Specifically, the framework allows for interactions between foreign and domestic, and foreign and foreign firms. The approach follows recent econometric advances, such as Hudgens and Halloran (2008) and allows the consistent estimation of a number of different treatment effects, taking into account the problems mentioned in the previous paragraph. In particular, one direct effect of

17 Introduction xxxi foreign ownership on firms in receipt of foreign capital (i.e., the topic of Part I) and two types of indirect effects, namely the indirect effects on other foreign and domestic firms. The latter comprises the topic of Part II. The estimated effects are all allowed to be moderated by the level of foreign ownership in an economic cluster. This novel approach is illustrated using firm level data for Chinese manufacturing from the same source as the data used in Chapter 5. An important finding is that the direct effect of foreign ownership is not homogenous across industry-region clusters. It is generally positive and increases strongly with the overall level of foreign-owned firms in a cluster. The paper also finds consistently negative indirect effects of foreign-ownership on domestic firms, indicating negative spillovers from foreign presence on domestic firms. Spillovers are more pronouncedly negative with increasing presence of foreign firms up to a threshold of around 40 percent foreign presence in a sector, after which they become less negative. Part III - Alternative Ways of Looking for Spillovers An important drawback of the traditional spillover studies is that they treat the specific mechanisms by which the spillovers are supposed to occur as a "black box." As pointed out above, they usually regress total factor or labour productivity of domestic firms on a number of covariates, including a measure of the extent of multinational presence in an industry. A positive and statistically significant coefficient on that variable is then interpreted as evidence for positive productivity spillovers. T he underlying mechanisms that empirically drive such correlations are generally not explored in the empirical analyses. The movement of workers from multinationals to domestic firms is one such potential mechanism, where workers take with them some of the knowledge about production or management processes (the firm specific asset) which can then be gainfully employed in the domestic firm and raise its productivity. This channel is also described in theoretical models such as Fosfuri et al. (2001). Chapter 13 presents an early empirical study focusing on this particular channel. The paper exploits a small linked-employer employee

18 xxxii Multinational Enterprises and Host Country Development dataset for Ghana from the same source as that used in Chapter 3. The data comprises five waves of an annual sample of Ghanaian manufacturing firms over the years Information is provided on whether or not the entrepreneur, i.e. the owner or chairperson, of the domestic firms worked for a foreign multinational before joining or setting up his/ her own domestic firm, and whether this work experience was gained in the same or in different industries. While the data do not provide information on all workers in a firm, they do relate to the entrepreneur, the main decision maker whose influence is particularly important for firm performance. Using these data, the paper investigates whether domestic firms employing entrepreneurs with foreign experience have a productivity advantage compared to other firms. The empirical analysis proceeds in two steps. First, it is an estimation of production functions allowing for firm fixed effects. These are estimated separately for different industries. The residuals from these estimations are used as proxies for total factor productivity. TFP is then regressed on a number of firm level controls, including information pertaining to the entrepreneur. A potential criticism with this approach is that highly productive domestic firms may attract entrepreneurs that have gained experience in multinationals within the same industry. This criticism, however, would not apply if these domestic firms were established by the owners and were not already in existence prior to the owner having joined. Hence, an alternative estimation strategy only uses data on domestic firms that were newly established by the current owner. Results show that firms whose entrepreneur worked in multinationals in the same industry are more productive than other domestic firms. No such evidence is found for firms run by entrepreneurs who worked for multinationals in other industries. This suggests that some of the multinationals' knowledge is industry specific and cannot be easily transferred to firms in other industries. Another potential mechanism for spillovers that has been investigated in more detail is backward linkages. The traditional studies in the spirit of Driffield et al. (2002) and Javorcik (2004) - as in Chapters 10 and 11 - regress domestic firm productivity on the

19 Introduction xxxiii presence of multinationals in vertically related industries. There is no way of knowing, though, whether domestic firms actually do possess a linkage, i.e., whether they do actually supply multinationals. All the conventional approach allows one to do is to proxy whether there is a potential for backward linkages. Chapter 14looks directly at firm level evidence of actual backward linkages by multinationals. It discusses the policy focus on linkages and investigates the factors that determine the extent of multinationals' local sourcing using firm level panel data for Ireland for the period 1983 to The data base used provides firm level information on local purchases of intermediate inputs. The paper focuses on investigating whether the extent of such a local linkage is impacted upon by the amount of subsidies paid to the firm. The analysis is conducted using fixed effects estimation as well as a GMM estimator to account for endogeneity of subsidies received by firms. The descriptive analysis in the paper shows that multinationals from different home countries, on average, source around 20 to 30 percent of their material inputs in Ireland. The main finding of the econometric analysis for the determinants of linkages, is that multinationals from the US and Europe develop backward linkages quite independently of government grants. This finding is most interesting from a policy perspective. By contrast, multinationals from the rest of the world respond positively to government subsidies in terms of linkage creation. It seems plausible that factors other than financial incentives override government efforts to influence linkages in the case of US and European multinationals. Even if there are linkages, the mechanism through which backward linkages may generate positive effects for their suppliers is not clear. Do multinationals voluntarily share knowledge and cooperate with their suppliers or do they force their customers to adopt new practices- and punish suppliers if these new practices are not implemented to their satisfaction? Chapter 15 tackles these issues by using firm level information on linkages, but focusing on the part of local suppliers. T he first question to be addressed is whether domestic firms that supply multinationals are more productive than other domestic firms. The second question

20 xxxiv Multinational Enterprises and Host Country Development is then whether domestic firms gain more or less if they report that they experience "pressure from their customers to reduce production costs or create new products." The empirical analysis is carried out using firm level data for emerging market economies in Central and Eastern Europe and Central Asia, from the Business Environment and Enterprise Performance Survey (BEEPS) provided jointly by the World Bank and the European Bank for Reconstruction and Development. The paper combines data from the 2005 and 2002 surveys. The difficulty is again posed when it comes to ascertaining the direction of causality. Is it the case that domestic firms improve their productivity if they supply multinationals (which is what the spillovers argument suggests) or is it simply the case that multinationals pick better performing firms as suppliers (J avorcik and Spatareanu, 2009)? In an attempt to deal with this problem, the estimation strategy adopted in the paper is to regress domestic firms' productivity growth between 2002 and 2005 on the supplier status in 2002, thus exploiting the panel dimension available in the data. This set up allows one to be more confident about identifying an effect of the supplier status on productivity growth. There is also an explicit test for the exogeneity of the supplier status, using an instrumental variables approach. The empirical estimations show that productivity gains only materialize for suppliers of multinationals when they are pressured by their customers to reduce production costs or create new products. This adds a new angle to the literature. Most of the studies on spillovers through backward linkages, at least implicitly, are based on the assumption of a benevolent multinational that is willing to share knowledge voluntarily and instruct willing suppliers. While the evidence in the paper does not suggest otherwise, the results hints that the customer-supplier relationship may not always be so harmonious. The remaining chapters in Part III move away from the standard approach of measuring spillovers on domestic firms' productivity. Rather, they are concerned with the implications of the presence of multinationals for domestic firms' entry, exit and survival. The approach used in Chapter 16 is based on the idea that an increase in productivity through spillovers will, ceteris paribus, reduce a host

21 Introduction xxxv country firm's average cost of production and, hence, increase profitability. Profitability, in turn, has long been regarded as a main determinant of firm survival. Hence, technology spillovers from multinationals and the associated increase in productivity should lead to a higher probability of survival for host country firms. This hypothesis is tested using establishment level data for Ireland covering virtually the entire population of manufacturing plants over the period 1973 to 1996 and is similar to the data used in Chapter 6. In the empirical analysis, using a Cox Proportional Hazard model, the paper finds that, controlling for other plant and sector specific effects, the presence of multinationals (measured as the foreign employment share) has a survival enhancing effect only on domestic plants operating in high technology sectors. This suggests that there may be productivity spillovers taking place in these industries. There is, however, no evidence for such effects on the survival of domestic plants in low technology industries. This may indicate a lack of absorptive capacity in such low tech domestic firms, hence, they are not able to benefit from spillovers. Chapters 17 and 18 develop another alternative approach to gauge the effect of multinationals on the development of domestic plants. They apply and extend a theoretical approach implemented by Markusen and Venables (1999), which argues that multinationals, through developing backward linkages with domestic suppliers increase markets for these firms and, accordingly, increase their profits in the short run. This can lead to further entry of new domestic firms into supplier industries, a fact that both drives down the price of supplies and also increases the scope for the entry of new domestic firms into final good industries. In line with this idea, Chapter 17 investigates econometrically whether the entry rate of domestic firms in an industry is positively related to the presence of multinationals (again measured as foreign employment share) in the same industry. Similar to Chapter 16, the study uses data on the population of manufacturing plants in Ireland from 1974 to In line with the theoretical expectation, the chapter reports evidence that the presence of multinationals indeed exerts a positive effect on the entry rate of domestic firms. This is taken to

22 xxxvi Multinational Enterprises and Host Country Development indicate that spillovers from multinationals encourage entrepreneurs to set up new firms, either in supplier or in final good industries. Chapter 18 extends the Markusen and Venables (1999) model and allows for the interaction between MNEs and domestic firms to take place through several channels. The first channel works through factor markets. Multinationals lead to a capital inflow and modify thp. host country capital endowment. In addition, foreign affiliates use differentiated intermediate products which indirectly affect the production conditions for domestic firms. The second channel is a competition effect, whereby multinationals compete with local producers in their product markets as well as in the market for factors. The model shows that the evolution of the number of local firms as a function of the presence of foreign firms can be depicted as a u-shaped relationship where the competition effect first dominates but is gradually outweighed by positive externality effects. Empirical evidence, also based on the micro level data for Ireland, support this theoretical prediction. Conclusion The evidence amassed in the literature points towards the conclusion that multinationals can have positive effects on the economic development of the host country. Multinationals bring with them superior technology which manifests itself in multinationals having, on average, higher productivity and technology levels, and paying higher wages than comparable domestic firms. The evidence in Part I of this volume supports this conclusion. The superior technology of multinationals may spill over to domestic firms, a phenomenon that can increase the productivity performance of domestic firms as well as improving the survival and the development of local industry. These aspects are explored in Parts II and III. There are still a number of remaining issues which the extant literature has not addressed satisfactorily. Do all domestic firms gain equally from multinationals, or are there certain characteristics that firms need to possess in order to reap these benefits? Do all multinationals have the same potential for spillovers? And,