Master Thesis Double Degree Program:
M.Sc. International Economics and Business – University of Groningen M.A. International Economics – University of Göttingen
Does It Matter Where You Go To?
Home Country TFP Effects of Outward FDI and its Destination
Student:
Name: Christina Pötzsch Std. Number: 2238330 (Groningen)
20705694 (Göttingen) Email: c.potzsch@student.rug.nl
Affiliation: University of Groningen, Faculty of Economics and Business University of Göttingen, Faculty of Economic Sciences
Supervisor: Dr. Bart Los, University of Groningen
Abstract:
Based on country-level panel data for 24 economies between 1991 and 2010 the present study examines the effect of outward foreign direct investments (OFDI) on home country total factor productivity (TFP). A key contribution to the extant literature is to assess this relationship via a distinction between different target and sending country groups, while in addition taking moderating influences of international trade activities into account. The empirical analysis provides evidence of advanced sending countries to benefit from investments in both developed and less developed host economies. Nevertheless, the latter effects appear to be more easily realized. Furthermore, results show emerging home countries to only experience TFP increases from OFDI directed towards less developed target economies, while compared to advanced sending countries requiring additional constraints to be overcome. In sum, theoretically based inferences indicate vertical investment patterns to be the most effective form of OFDI to increase home country TFP.
Keywords:
1. Introduction
“In the search for the secrets of long-run economic growth, a high priority should go to rigorously defining TFP, empirically dissecting it, and identifying the policies and institutions most conducive to its growth.”
(Easterly and Levine, 2001, p.179)
What accounts for growth differences across countries? What explains cross-country income differences? According to most of the empirical literature the overwhelming answer is total factor productivity (TFP) (see e.g. Klenow and Rodríguez-Clare, 1997; Hall and Jones, 1999; Caselli, 2005) – an answer which Easterly and Levine (2001) even subsume among the heading of ‘new stylized facts of growth’. Nevertheless, the consensus regarding the significance of productivity differences in accounting for substantial parts of cross-country income variation only represents a proximate explanation. However, in line with the above stated objective formulated by Easterly and Levine (2001), establishing the ultimate causes requires unraveling the determinants of productivity growth – and the current study intends to contribute to it. Specifically, the potential role of outward foreign direct investments (OFDI) for home country TFP advances is being analyzed, motivated by this prospective channel having hitherto received relatively little attention in the extant academic literature. In fact, it has been primarily evaluated for individual advanced sending countries, whereas emerging or developing home economies have only been considered sporadically. In contrast, the current analysis is to the best of my knowledge the first to apply a diverse dataset consisting of both advanced and emerging home economies so as to investigate the potential relationship between OFDI and home country TFP. Importantly, OFDI is distinguished according to its destination, i.e. whether it is directed towards developed or less developed countries, so as to account for the possibility that different forms of OFDI have different impacts on home country TFP. Furthermore, the possibility that advanced and emerging home countries display different relationships is considered. More precisely, the central question whether ‘it matters where you go to’ is considered separately from the perspectives of both advanced and emerging sending economies.
The remainder of the study is organized as follows. Section 1 continues by formulating both precise research questions and a general economic model. Section 2 introduces target and sending country considerations into the theoretical link between OFDI and home country TFP while Section 3 provides a short review of extant empirical insights. Section 4 describes the data and the empirical methodology. Section 5 reports and discusses the estimation results while Section 6 concludes.
1.1. Research Questions
investments are usually accompanied by large costs, which may only be covered with appropriate preexistent productivity levels. In fact, the literature on heterogeneous firms suggests only the most productive enterprises to engage in international activities such as trade and FDI (see e.g. Helpman et al., 2004). In addition, consistent with traditional FDI theory OFDI from emerging country firms may be assumed to be unattainable for they may lack sustainable firm-specific advantages yielding a competitive edge in foreign markets.
Notwithstanding, OFDI from emerging and transition economies has risen significantly in recent years.1 Specifically, between 1991 and 2010 OFDI flows from emerging and transition countries as a percentage of GDP have risen from 0.35 to 1.73 and from 0.23 to 2.89 percent, respectively (UNCTADstat, 2012). Furthermore, as can be seen from Figure 1, since the aftermath of the global economic crisis and its significant negative impact on international activities transition economies are characterized by a higher OFDI intensity than their developed counterparts.
Figure 1: OFDI as Percentage of GDP (1991-2010)
Data source: UNCTADstat, April 2012.
Hence, recent trends in global OFDI flows indicate the hitherto widespread confinement of the potential link between OFDI and home country TFP developments to advanced sending countries to be of decreasing relevance. While productivity advances may be of particular interest for middle-income and developing countries, those economies are increasingly engaged in global OFDI thereby rendering its analysis suitable for a large set of sending countries. Thus, the current study intends to exploit these increasingly diverse patterns of OFDI activities in an attempt to provide further insights into the link between OFDI and home country TFP. Specifically, the crucial contribution to the existing literature is the application of a locational distinction of OFDI
1
from both a target and sending country perspective so as to determine potentially diverging relationships for different country groups. This is made feasible via the diverse set of countries included in the data sample and results in the objective to answer two central questions:
(1) Do home country TFP effects of OFDI differ with respect to its destination? Specifically, do investments in developed and less developed countries yield differential effects? (2) Do home country TFP effects of location-distinguished OFDI differ for different home
economies, i.e. advanced and emerging sending countries?
To resolve these questions, data for 15 advanced and 9 emerging economies between 1991 and 2010 is employed. In total, investments are directed towards 222 target countries yielding 2629 home-host combinations.
Importantly, the answers to the above stated questions may be of specific applicability for policy considerations from an emerging country perspective. The subject matter as to whether countries should further or discourage the internationalization of home country firms or whether policy should be neutral is especially relevant for emerging economies where firms may lack both appropriate internal and external financial means, particularly regarding investments in advanced and technologically leading countries. Nevertheless, a thorough resolution of this question also requires considering potential alternatives to OFDI and prospective interrelations regarding resultant TFP effects. Specifically, previous literature generally highlights the importance of trade as a channel for international knowledge and technology diffusion especially for less developed countries usually characterized by an inherent larger gap with respect to the world technology frontier (Criscuolo and Narula, 2008). As a result, the current study additionally emphasizes and explicitly analyzes potential synergetic TFP effects between OFDI and trade, which has to the best of my knowledge not been done before.
1.2. Economic Model
With the purpose of elucidating the above stated research questions a more formal framework of analysis is introduced. In particular, the above stated research agenda can be analyzed within the subsequent general reduced-form economic model
( ∑ ∑ ) ( ∑ ∑ )
and TFP. Importantly, the separate functions and allow this relationship to differ for the two sending groups and .
2. Theoretical Background
Nevertheless, before empirically analyzing the relationship between OFDI and home country TFP via both a target and sending country perspective, insights into the theoretical factors underlying such a potential link should be obtained. In other words, why should foreign investment activities by domestic firms influence home country productivity?
2.1. Total Factor Productivity and Its Components
In order to answer this question it is essential to first formally describe the economic construct of TFP and its components. In particular, TFP is commonly defined as the portion of output not explained by the amount of inputs employed in production (Comin, 2008). In other words, it denotes the efficiency with which all input factors are used so as to generate output (Liu and Wang, 2003). This definition of TFP is usually formalized via the Solow residual method (Solow, 1957). In particular, it is based on the assumption a Hicksian production function of the general form , where is real output of country at time and is a Hicks-neutral shift parameter of the production function specified for input factors capital and labor (Hulten, 2001). Crucially, within this framework output growth can be achieved via increases in both inputs and . In particular, captures any variations in output over and above changes in inputs, thereby giving rise to the definition of as TFP.
Importantly, productivity advances as measured by increases TFP can be broadly decomposed into technical efficiency improvements and technological progress. The former refers to a more proficient exploitation of given means of production and levels of technology via the application and imitation of existing knowledge as well as changes in resource allocations, while the latter captures prospective advances in production opportunities due to innovation and enhancements in technological capabilities (Zhao et al., 2010). Building upon the growth accounting concept proposed by Nelson and Pack (1999) and their forwarded incorporation of capital assimilation in TFP analysis, these two distinct components can be further elucidated by employing a dynamic production function framework.2 Specifically, Figure 2 depicts two production functions of countries 1 and 2. Since for given capital-intensities function implies higher levels of output per worker, it can be interpreted as representing more advanced technologies. Assuming countries 1 and 2 to conduct production at points respectively, corresponding per-capita output levels are given by . The disparity in these output levels can be classified as a
2
difference in TFP which can in turn be decomposed into the central components introduced above.
Figure 2: Decomposition of TFP
Source: Own depiction.
First, given that country 1 initially produces below its best-practice production function a more efficient exploitation of its extant technology allows achieving an increased output level of . Specifically, operating on its own production function country 1 has a maximum efficiency level of 1, equivalent to country 2 producing at . Thus, the in Figure 2 depicted output increase denoted by is to be attained via efficiency improvements. Yet, further output growth can be achieved via advances in technology. Specifically, referring to the growth accounting framework proposed by Nelson and Pack (1999), country 1 can invest in the acquisition of more advanced technology while its subsequent assimilation via domestic efforts in terms of education, sound entrepreneurial environments and resultant own build-upon innovations allow the attainment of a higher production function and accompanying productivity levels. Thus, shifting its function to the level of while at the same time maintaining a maximum production efficiency, country 1 can realize a per-capita output of . In other words, engaging in technological progress via technology assimilation country 1 can increase its output by the amount of . The remaining output difference between countries 1 and 2 is then merely due to different capital-labor ratios employed in production. However, engaging in this process of capital accumulation results in rather negligible productivity increases as compared to technology assimilation, specifically since factors of production are assumed to run into diminishing returns for given levels of technological knowledge. Thus in sum, decomposing TFP via a dynamic production function analysis allows efficiency improvements to be depicted as movements towards a given domestic best-practice production function, whereas technological progress implies its shift to more advanced levels of technology.
or research and development (R&D) facilities in technologically advanced host countries (Zhao et al., 2010). Thus, technology sourcing FDI may lead to international technology transfers from target to investing economies thereby increasing home country productivity. Second, efficiency-seeking or vertical FDI may result in efficiency gains, for it allows the optimization of production structures on an international basis thereby exploiting the global distribution of comparative advantages. Consequently, both key elements of TFP may be positively affected by OFDI thus turning it into a vital source for the initially described essential increases in home country productivities.
2.2. Productivity Effects and Different Forms of Outward FDI
Importantly, this rather roughly outlined potential role of OFDI in affecting both TFP components is subsequently depicted in more detail. In particular, while foreign investment activities continue to be categorized according to different forms and motivations, target and sending country perspectives are introduced in their respective analyses. As a result, more profound insights into the specific underlying mechanisms are derived, which in turn permits inferring hypotheses regarding the central research questions stated above.
2.2.1. Technology Sourcing FDI
First, target country considerations shape potential effects of technology sourcing FDI on home country TFP, for host countries differ in their technological developments and therefore in the scope allowing for the retrieval of advanced technologies. Particularly, technological development tends to be concentrated in a small number of developed economies (Fu and Gong, 2011). Nevertheless, technology sourcing FDI need not be confined to investment activities in these so called ‘centers of excellence’, notably the United States, Europe and Japan (von Zedtwitz and Gassmann, 2002). Instead, technology sourcing can also be achieved by investing in relatively technologically advanced emerging or even developing economies.
Employing a sending country perspective this might be a particularly vital option for emerging country firms, for they might often lack the required absorptive capabilities so as to engage in successful technology sourcing in advanced host countries. Specifically, larger pre-existing technology gaps between home and host economies have been suggested to decrease the scope for foreign knowledge acquisition (Glass and Saggi, 1998), for the required abilities to interpret it and to engage in the above introduced important process of assimilation by adapting foreign sourced technology and linking it to domestic environments and local conditions is reduced (Criscuolo and Narula, 2008).3 On the other hand, ‘South-South’ technology sourcing is likely to entail a higher technological proximity between investing and target country firms, for the latter do usually not operate at the global technology frontier, thereby increasing the scope for successful technology
3 Underlying these considerations is a distinction between absorptive capacity and absorptive capability. While
transfer (Herzer, 2011a). This idea can further be elucidated via the appropriate technology paradigm (see e.g. Basu and Weil, 1998; Acemoglu and Zilibotti, 2001), which assumes developed economies to invent technologies that primarily match their own current production structures and factor endowments. However, these technologies are depicted to not be compatible with the different production techniques and factor mixes of other less developed countries. Put simply, the most productive technologies created and employed by advanced economies may be inappropriate for their emerging counterparts (Jerzmanowski, 2007). Including the notion of appropriate technology in the dynamic production function analysis described above the potential importance of ‘South-South’ investment patterns for home country TFP advances can be explicitly illustrated. Particularly, Figure 3 depicts the previously in Figure 2 discussed production functions and for countries 1 and 2, respectively. Yet, the in Figure 2 implicitly imposed idea that country 2’s technology can be achieved by every other economy regardless of its current technological knowledge is removed. Rather, the newly incorporated production function is assumed to be the uppermost available being compatible with or ‘appropriate’ for country 1’s current production function .
Figure 3: TFP and Appropriate Technology
Source: Own depiction.
second step, ‘South-North’ investments may successfully overcome technology differences between countries 1 and 2 for the underlying gap has been previously reduced, thereby resulting in a further shift and accompanying productivity increases .
Thus, summing up the just described considerations both a target and sending country distinction appear to be relevant so as to properly assess the relationship between technology sourcing FDI and home country TFP. First, a target country perspective is important since host countries differ in their technological developments and therefore in the scope for technology sourcing. However, this scope has to be evaluated relative to existing home country technological positions, for successful technology sourcing FDI depends on the ability of investing firms to interpret foreign technologies as well as to adapt and assimilate them to domestic environments. In other words, effective technology sourcing requires target countries that are relatively advanced while at the same time displaying some technological proximity or ‘appropriateness’ thereby increasing the required absorptive capabilities due to a sufficient pre-existing knowledge base. The realization of both conditions can be assumed to vary for different sending countries. In particular, developed sending countries may most successfully engage in technology sourcing in other advanced economies, while emerging countries may also benefit from investing in other emerging or even developing economies.
Nevertheless, target countries may not only differ with regard to technological developments but also in terms of competitive and institutional environments influencing the success of technology sourcing FDI and hence successive home country TFP effects. More precisely, developing country markets may be characterized by weaker institutional provisions of property rights protection and enforcement mechanisms than their developed counterparts (Makino et al., 2004), thereby facilitating the acquisition of foreign technologies. Yet on the other hand, developing country markets may often lack sophisticated local supplier industries (Makino et al., 2004) and hence important vertical linkages that constitute a potential source of reverse engineering and thus technology sourcing activities.
2.2.2. Efficiency-Seeking FDI
However, technology sourcing is not the only form of OFDI that might affect home country TFP. Specifically, efficiency-seeking or vertical FDI activities might lead to a more proficient and thus productivity increasing exploitation of domestic production opportunities. The underlying reasoning rests on the definition of vertical FDI as the international fragmentation of production on the basis of low-cost considerations (Markusen and Maskus, 2002). In other words, firms seek to utilize existing technologies and production processes more efficiently by locating each stage of production in the country where it can be carried out with the lowest costs (Herzer, 2011b). Consequently, target country distinguished foreign investment activities might reveal differential impacts on home country TFP for host economies differ in comparative advantages thereby displaying different scopes for efficiency-seeking FDI.
advanced home countries can be assumed to have a comparative advantage in high-skill labor intensive stages of production, thereby primarily benefiting from vertical investment patterns that transfer low-skill labor intensive processes abroad (Barba Navaretti et al., 2010). In contrast, emerging economies may have different home country comparative advantages leading to distinctive and probably more diverse patterns of international production structures. Efficiency-seeking FDI may for example also entail locating higher value-adding activities such as R&D or marketing services in advanced host countries, while simultaneously exploiting a domestically available relatively cheap unskilled labor force. On the other hand, although being compatible with the theory of comparative advantage such production structures may in reality be more reasonably assumed to be subject to outsourcing than intra-firm organization. Explicitly, emerging country firms may lack firm-specific advantages in high-skill intensive activities worthy of protection in order to motivate the organization of production via FDI.
Nevertheless, in sum the just depicted considerations indicate efficiency-augmenting international production patterns to depend on comparative advantages of both home and host countries (Kokko, 2006), thus leading to potentially different relationships between OFDI and home country TFP from both a target and sending country perspective. Specifically, advanced sending countries can be assumed to benefit relatively more from vertical investments conducted in developing host economies, for the resulting international production structures strengthen the domestic comparative advantage in skill-intensive activities. Similarly, emerging home countries may also benefit relatively more from efficiency-seeking FDI in emerging and developing target economies. In particular, despite potential similarities in factor endowments this target country group might reveal a higher scope for vertical investments as their comparative advantage patterns may be more diverse than the ones displayed by their advanced host country counterparts. In other words, vertical FDI in emerging and developing countries may entail more flexible investment and production structures thereby allowing for possibly more pronounced efficiency increases in home activities than vertical investments directed towards advanced economies.
Nevertheless, despite these similar target country positions advanced sending countries can be assumed to have a higher potential for TFP increases due to vertical investments than their emerging counterparts. Specifically, the domestically performed stages of production are usually high value-adding activities with a high potential for productive dynamism. In contrast, labor-intensive production processes presumed to be relatively more conducted in emerging home economies offer fewer opportunities to lead to continuous productivity increases.
2.2.3. Market-Seeking FDI
resultant fortifications of home country TFP. Specifically, their usually larger markets may lead to particularly pronounced increases in competitiveness and thus intensifications of home activities. If these are characterized by either economies of scale or imply a reinforcement of home country comparative advantages the resulting productivity increases may be essential. Nevertheless, advanced target country markets may at the same time display rather established market structures thereby complicating the objective to obtain improved positions. As a result, they may be more suitable so as to strengthen and expand already existing market shares. In contrast, emerging and developing country markets frequently exhibit growing selling potentials and weak competitive structures thus offering opportunities for early positioning so as to ensure long-term success and market shares.
Consequently, taking into consideration a sending country perspective probably more established advanced home country firms may benefit from horizontal investments in both developed and less developed country markets. While the former may improve already existing market shares, the latter may establish new positions thereby fostering future profits in growing markets. Emerging country firms on the other hand may primarily benefit from market-seeking FDI in other emerging and developing host economies. Specifically, gaining significant market shares in advanced countries may be challenging due to pre-existing competitive structures. Thus, newly positioning firms may mainly be successful via niche products. However, this might not be profitable for emerging country firms for niche products might only have a limited selling potential in their respective home markets. In contrast, emerging and developing country markets may offer superior opportunities so as to position firm-established products thereby securing future profits via increasing market shares.
Additionally, sending country considerations indicate advanced home economies to enjoy a generally higher potential for home country TFP increases due to market-seeking investments. In particular, market-seeking FDI usually entails an increased need for headquarter services at home. This in turn might imply both the realization of firm-level economies of scale and a more efficient exploitation of home country factor configurations thereby increasing productivity. Nevertheless, this warrants a domestic comparative advantage in skill-intensive headquarter services, which might primarily be the case for advanced home country firms.
2.3. Home Country Productivity Effects
In sum, the previous sections reveal different forms of OFDI to potentially positively affect home country TFP. Furthermore, the underlying relationships are shown to depend on both home and host country characteristics thus validating the importance of both a target and sending country perspective in an empirical analysis.
productivity effects thus yielding benefits from OFDI at a home country level. The main mechanisms via which firm-internal TFP increases can affect overall home country productivity are compositional effects and productivity spillovers. Unfortunately however, literature on OFDI related home country compositional and spillover effects is scarce. Nevertheless, the theoretical channels are comparable to potential inward FDI related effects in host economies (Vahter and Masso, 2006). Hence, the subsequent discussion of home country productivity effects draws in part on the extensive literature on host country impacts and refines considerations if applicable. 2.3.1. Compositional Effects
Compositional effects imply productivity-improving foreign investing firms to raise average home country TFP due to their sheer presence (for inward FDI see Barba Navaretti and Venables, 2006). However, prerequisite firm-internal productivity increases may be jeopardized if foreign investments replace domestic ones. This might be especially likely for emerging country firms since firm-internal financing opportunities may be limited and external financing may be costly. Moreover, this potential substitution effect is particularly detrimental if domestic investment activities are required so as to develop or sustain absorptive and innovative capabilities at home. In that case OFDI may lead to a reduction in home country productivity in the long-run (Herzer, 2011b). Yet what is more, the replacement of domestic investments need not be confined to appearing inside the firm but can also emerge at both the industry- and country-level, especially in middle-income countries with potentially limited external financing opportunities. More precisely, OFDI may divert investment resources away from domestic firms thereby constraining their capacities for productivity development or sustenance. In addition, the impending relocation of resources to foreign countries via OFDI may reduce domestic capital deepening (Herzer, 2011b) thereby constraining the process of home country technological sophistication.
On the other hand, OFDI may also be critical so as to ensure firm survival notably in industries that face an increasingly global competition (Fors and Kokko, 2001). Specifically, an enhanced openness to trade and inward FDI may intensify home country competition thereby threatening local firms’ market shares and positions. As a result, if foreign investments increase firm-internal productivities and thus both national and international competitiveness the exit of domestic firms may be prevented thereby positively affecting indigenous home country productivity.
2.3.2. Productivity Spillovers
Demonstration and imitation refer to the idea that indigenous firms learn from foreign investing and probably TFP increasing firms by observing their products and processes.4 In particular, reverse engineering is often cited as a promising way so as to acquire knowledge from home country multinationals, yet the prospective scope crucially depends on the complexity of the products and technology involved (for inward FDI see Görg and Greenaway, 2004). Labor mobility on the other hand implies that workers employed by foreign investing firms acquire some of the firm-specific superior knowledge and transfer it to local enterprises when they leave and work for domestic firms or open up their own companies (for inward FDI see Görg and Strobl, 2005). Alternatively, indigenous firms may be forced to enhance their productivity and efficiency so as to survive the increasingly intense competition from firms that attain foreign knowledge and efficiency gains by investing abroad (for inward FDI see Lipsey, 2004). However, if local competitors cannot defy intensified competitive structures and are rather forced to exit the market the number of home country firms that can absorb and exploit potential productivity spillovers is reduced (Kokko, 2006). Horizontal linkages on the other hand imply intra-industry productivity increases in indigenous firms that may inter alia be caused by external economies of scale. More precisely, if home country multinationals manage to expand their remaining home country production domestic firms may benefit as well if the resulting economies of scale are external to the firm. In contrast, vertical linkages refer to potential inter-industry productivity spillovers. An OFDI related increase in domestic production may e.g. enhance the demand for local intermediate products and inputs from home country suppliers. Thus, foreign investments may confer an economies-of-scale effect to both upstream and downstream domestic producers which might lead to TFP improvements. Moreover, home country multinationals may increase the supply of new and superior intermediate inputs due to foreign technology sourcing thus giving rise to technology transfer via forward linkages (for inward FDI see Rodríguez-Clare, 1996). Similarly, foreign investing firms may also raise the demand for higher-quality products from local suppliers (for inward FDI see Keller and Yeaple, 2003). Specifically, due to potential foreign technology sourcing and the resulting upgrading of products and technological processes, indigenous multinationals are likely to demand more sophisticated and superior knowledge intensive intermediates thereby implying a knowledge transfer to local suppliers via backward linkages. Moreover, firms engaging in OFDI can be assumed to willingly convey technological know-how to local suppliers so as to ensure the proper provision of required inputs (for inward FDI see Blalock and Gertler, 2008). Therefore, similar to inward FDI related spillovers, backward linkages can be assumed to yield a particularly high potential for technology and productivity transfer from home country multinationals to domestic national firms (for inward FDI see Javorcik, 2004), especially since upstream sectors can usually be assumed to be more technology intensive than downstream activities that cover mostly wholesale or logistic activities. Nevertheless, OFDI may also cause home country multinationals to switch from local to host
4 Note however that even in the absence of productivity increases due to OFDI, foreign investing firms can be
country suppliers, either so as to reduce disintegration costs or acquire foreign technologies via reverse engineering.
In sum however, there are various channels via which productivity spillovers may enhance productivities of local firms, thereby raising overall home country TFP. Hence, OFDI might be a viable option, especially for emerging countries, so as to achieve technological catch-up and productivity growth if these spillovers to the rest of the economy exist. Specifically, economy-wide productivity effects might also serve as a justification for government support so as to further OFDI by home country firms. Nevertheless, the scope for home country productivity spillovers is decisively mediated by domestic industry and structural effects (Kokko, 2006). In particular, as mentioned above, critical factors are whether domestic competitors are able to compete with foreign investing firms thereby staying in the market and whether home country multinationals strengthen local suppliers and upstream sectors or whether they switch to host country firms.
2.4. Hypotheses
Overall, the theoretical considerations concerning the prospective relationship between OFDI and home country TFP suggest a generally positive association. Nonetheless, owing to the preceding detailed analyses regarding the effects of different forms of OFDI more explicit predictions can be derived from both a target and sending country perspective. Resuming the notation of the economic model introduced in Section 1.2. and employing arguments explained in Section 2.2., Table 1 summarizes the expected relative target country benefits of different forms of OFDI for advanced and emerging home countries and , respectively.
Table 1: Target and Sending Country Hypotheses Advanced Sending Countries
Technology Sourcing ( ∑ ) ( ∑ ) Efficiency-Seeking ( ∑ ) ( ∑ ) Market-Seeking ( ∑ ) ( ∑ )
Emerging Sending Countries
Technology Sourcing ( ∑ ) ( ∑ ) Efficiency-Seeking ( ∑ ) ( ∑ ) Market-Seeking ( ∑ ) ( ∑ )
2.4.1. Advanced Sending Countries
As can instantly be seen, advanced sending countries are expected to experience mixed relative benefits from target country distinguished OFDI depending on the particular investment form. First, technology sourcing may be more effective when conducted in developed host economies. In particular, these can be more reasonably assumed to be in part relatively more technologically advanced than their less developed target country counterparts thereby revealing a higher scope for technology sourcing activities. Efficiency-seeking FDI on the other hand may reveal higher TFP effects when directed towards less developed countries, for these host economies usually display different comparative advantages than advanced sending countries thereby allowing efficiency gains to be realized. In contrast, target country effects of market-seeking FDI may be difficult to distinguish for advanced home economies. Specifically, while developed host countries are usually characterized by larger markets, less developed target economies have a larger growing potential and weaker competitive structures thereby facilitating the establishment of market positions and the security of long-term market shares.
Thus in sum, for advanced sending countries the relative benefits of OFDI directed towards the two target country groups may be difficult to distinguish. However, it could be assumed that the investment form of efficiency-seeking FDI is in general most prevalent. Furthermore, advanced sending countries can be assumed to have a specifically high potential for TFP increases due to vertical investments for the remaining domestic stages of production are usually high-value adding activities that have a high potential for productive dynamism. As a result, advanced sending countries could be hypothesized to benefit marginally more from OFDI directed towards less developed countries.
Furthermore, productivity spillovers have a high potential to be attained in advanced sending countries. Competitors and intermediate product producers can be assumed to be sophisticated enough so as to acquire both technology and knowledge advances from foreign investing firms thereby allowing both intra- and inter-industry effects. In addition, external financing opportunities can be assumed to be sufficient so as to not divert investment opportunities away from domestic firms thereby inhibiting their productivity development or sustenance.
2.4.2. Emerging Sending Countries
in sum, emerging countries may benefit relatively more by investing in other emerging and developing economies.
However, in contrast to advanced sending countries, home country productivity spillovers in emerging countries may be more constrained if upstream and downstream sectors are not sophisticated enough. Furthermore, limited external financing opportunities may divert resources away from domestic investments thereby increasing the likelihood of an exit of domestic competitors.
Nevertheless, it should be noted that the overall effects from OFDI directed towards developed and less developed economies depend on the relative weights of the different forms of foreign investment activities as well as on the comparative sizes of the individual effects. As a result, theoretical effects of total OFDI activities are to some extent still ambiguous thereby validating the significance of an empirical analysis.
3. Related Empirical Literature
Nevertheless, before delving into the empirical investigation conducted in the current study central insights obtained from the extant literature are reviewed. Specific emphasis will be placed on findings relating to target and sending country perspectives so as to conform to both the theoretical insights presented above as well as the empirical analysis conducted in Section 5. Importantly, the explicit analysis of OFDI related home country productivity effects has to a large extent focused on the specific forms of technology sourcing and efficiency-seeking FDI. For example, employing a UK industry-level panel dataset over the period 1987-1996, Driffield et al. (2009) analyze whether these investment activities affect home country sector-specific TFP. Decisively, the two motives are distinguished by employing a target country perspective; i.e. technology sourcing FDI is proxied for by UK OFDI flows in high cost, high R&D-intensive countries while the efficiency-seeking rationale is identified as OFDI directed towards low cost, low R&D-intensive economies. The results reveal that both forms of OFDI positively influence industry-level TFP. Moreover, both estimated coefficients are similar in size and statistical significance.
be important so as to properly assess potential home country productivity effects. In other words, a simple correlation between aggregate OFDI flows and domestic TFP does not allow a theoretical inference about potential underlying mechanisms and hence about the sustainability of the process. As a result, a target country perspective so as to distinguish potential rationales of OFDI might generate useful insights into the causal relationship.
Proceeding along similar lines, Sun et al. (2010) attempt to distinguish technology- and efficiency-seeking FDI related productivity effects by analyzing industry-level panel data for Taiwanese manufacturing sectors between 1993 and 2001. OFDI flows are distinguished as being directed towards advanced and emerging economies on the one hand and developing countries on the other. Importantly, in contrast to Driffield et al. (2009) TFP is calculated according to the Malmquist productivity index, implying that Sun et al. (2010) are able to explicitly decompose productivity changes in its components of efficiency change and technological progress. In other words, while the analysis due to Driffield et al. (2009) only allows a theoretical inference of the distinct impacts implied by movements towards existing domestic production opportunities and shifts to higher-technology incorporating production functions, Sun et al. (2010) are able to directly measure these differential effects of OFDI. First, looking at aggregate TFP reveals foreign investments in advanced and emerging countries to not yield statistically significant effects, whereas OFDI in developing countries reduces industry-level productivity. Hence, assuming that foreign investments in developing countries are primarily motivated by efficiency-seeking FDI, the potentially resulting improvements in proficient factor usage do not appear to be realized. Yet, disaggregating TFP into its two distinct components reveals that investments in developing countries result in negative effects on technological progress while efficiency advances are indeed positively affected, thus being in line with theory. In contrast, OFDI to advanced and emerging countries do not yield statistically significant effects even for estimates on disaggregated TFP components. Consequently, the theoretical prediction that middle-income countries may primarily benefit from foreign investments in developing countries is partly confirmed, while at the same time the just described results support the general hypothesis that TFP effects of OFDI differ according to its destination. Furthermore, considering that Driffield et al. (2009) reveal positive UK TFP effects for all target country groups whereas Sun et al. (2010) find differential host country effects for the emerging economy Taiwan confirms the supposed importance of a sending country perspective.
Nevertheless, while Bitzer and Görg (2009) derive useful insights from a sending country distinction they do not consider a target country perspective the importance of which has however been demonstrated in the formerly cited studies. The aim of the current analysis is to fill this gap. Specifically, the following empirical investigation intends to contribute to the existing literature on OFDI related home county TFP effects in the following way. While most studies only apply a target country distinction for single home economies, analyzing productivity effects for a large set of home countries allows the combination of both a target and sending country perspective that has so far not been done in the extant academic literature. Importantly, in contrast to Bitzer and Görg (2009) home economies are categorized so as to see whether there are structural differences across distinctive sending country groups.
4. Empirical Methodology and Data
Specifically, the current empirical analysis employs a panel dataset for 15 advanced and 9 emerging home countries between 1991 and 2010. Foreign investments of these focal economies are directed towards a total of 222 target countries yielding 2629 home-host combinations. The following section presents a thorough description of the data, introduces the econometric framework and discusses the estimation techniques applied.
4.1. Data
Data for the implementation of empirical estimations are obtained from two sources. In particular, data for TFP calculations and additional control variables are acquired from the World Bank Development Indicators; data on OFDI are obtained from the statistical database of the OECD (OECD.Stat).
4.1.1. TFP
Following the hitherto most widespread and thus commonly established approach TFP is calculated according to the Solow residual method introduced in Section 2.1. Importantly, its empirical application requires substituting a specific production function for the general notion
. In line with the conventional assumption made in the prevalent literature a Cobb-Douglas production function with constant output elasticities and constant returns to scale is applied for all countries and time periods (Hulten, 2001). More precisely, this yields
drawback of omitting labor income due to self-employment (Aiyar and Dalgaard, 2005). However, implementing various corresponding calculation adjustments for the just described factor share approximations via national accounts data, Gollin (2002) concludes labor income shares to be essentially constant across both countries and time. The validity of this result is inter alia supported by Bernanke and Gürkaynak (2001) who find labor income shares to not vary systematically with either real GDP per capita or time. As a result, assuming identical factor income shares across both countries and time has become the standard practice in the extant empirical literature (Aiyar and Dalgaard, 2005; Woo, 2009). Consequently, in line with the estimated average labor income share due to both Gollin (2002) and Bernanke and Gürkaynak (2001) the present analysis follows the prevailing modus operandi of most existing studies in setting labor income shares to be 0.65 (see e.g. Aiyar and Dalgaard, 2005; Klenow and Rodríguez-Clare, 2005; Krammer, 2010; Herzer, 2011b) thus leaving capital income shares to be 0.35 due to the above introduced assumption of constant returns to scale. Performing a simple diagnostic check these values are in addition empirically validated for the current dataset (see Appendix A).
As a result, TFP approximations are calculated by a simple transformation of the Cobb-Douglas production function, yielding
In order to calculate TFP via this equation data on real output , as well as on input factors labor and capital is required. The World Bank Development Indicators provide readily available data for measured by real GDP in constant dollars of 2000, and labor quantified as the total economically active labor force. In contrast however, capital stocks have to be constructed as the World Bank Development Indicators only include data on gross capital investments . Consequently, capital stocks are computed via the commonly applied perpetual inventory method (see e.g. Hall and Jones, 1999; Timmer, 1999; Caselli, 2005), which derives the capital stock as the sum of net capital additions (Hulten, 2001). In particular,
where is the capital stock of period , denotes a deprecation rate of the capital stock and measures capital investments (Kamps, 2006). Data on gross capital investments come in constant dollars of 2000. The annual depreciation rate is assumed to be 10 percent, in line with most of the extant academic literature (see e.g. Bitzer and Kerekes, 2008; Bitzer and Görg, 2009; Tuan et al., 2009; Krammer, 2010).
(Caselli, 2005). Specifically, using equation (3) so as to obtain the growth rate of the capital stock and solving for yields
Given the steady state conjecture of the Solow framework the unknown capital growth rate equals the growth rate of investments (Kamps, 2006). Thus, assuming investment to grow in the first period at the same average rate as in the first ten years of the sample (Jones and Olken, 2005), values of the initial capital stocks can be obtained by dividing the initial capital investments by the sum of the depreciation rate and the average of the ten subsequent growth rates of capital investments, i.e.
(∑ ) ⁄
where is the capital investment growth rate between and for country (Hall and Jones, 1999; for subsequent applications see e.g. Bonfiglioli, 2008; Woo, 2009; Zhao et al., 2010). Thus, combining calculations depicted in formulas (3) and (4) capital stock data is obtained so as to subsequently estimate TFP via equation (2).
4.1.2. Outward FDI
OFDI is measured in stocks so as to allow for medium-and longer-run effects on TFP (Bitzer and Görg, 2009). In particular, TFP effects might consequently be more visible or pronounced since firms need time to both absorb foreign knowledge and technologies as well as to realize potential efficiency gains. In addition, for effects to be observable at the country-level technology spillovers from foreign investing to domestic national firms play a crucial role, thereby further enhancing the time requirement for TFP effects to occur. The OECD provides data on OFDI stocks in millions of current US dollars. So as to be compatible with the TFP estimates obtained from World Bank data, FDI is deflated to constant dollars of 2000 using the consumer price index (CPI) provided by the OECD which allows correcting for the general economy-wide level of inflation (for further empirical studies deflating FDI data via the CPI see e.g. Schneider, 2005; Yao, 2006; Eichengreen and Tong, 2007).
Country classifications into advanced and emerging sending countries as well as developed and less developed target economies are conducted on the basis of per capita GDP taken from the World Bank. Specifically, both emerging home and less developed host countries are categorized according to having an average sample period GDP per capita below 11000 US dollars.5 This definition generally follows the country group classification scheme employed by the World Bank, which defines the upper per capita income limit for middle-income countries to be 12275 US dollars (World Bank, 2012). Yet, in order to obtain a sample with significantly large income
5
More precisely, average sample period GDPs per capita ̅̅̅̅̅ are calculated as ̅̅̅̅̅ ∑ ⁄
differences between advanced and emerging sending economies a minor downward adjustment has been conducted in the current research setting leading to the application of a slightly stricter emerging country definition.
4.1.3. Control Variables
In addition to the main variables of interest TFP and OFDI the empirical analysis includes control variables so as to avoid an omitted variable bias. Specifically, this implies incorporating potential influential factors on TFP that can at the same time be assumed to be correlated with OFDI. Here, control variables include inter alia home country total merchandise trade for it represents another channel that may lead to the international diffusion of knowledge and technology and thus TFP advances (Saggi, 2002).6 In addition, engaging in vertical OFDI and thus an international organization of production is usually accompanied by an increasing involvement in transnational trade structures, thereby implying a correlation between these two forms of international activity (Keller, 2010). Thus, in order to correctly estimate the effect of OFDI on TFP the inclusion of a measure of international trade activity is crucial (cf. Krammer, 2010). Importantly however, while trade is a direct determinant of TFP, it can also be assumed to have a moderating influence on the relationship of interest between OFDI and TFP. Specifically, all in previous sections depicted potentially TFP enhancing forms of OFDI are connected to trade. First, TFP advances due to market-seeking investments may be positively affected by trade for it usually involves higher prior and contemporary market access facilitating the establishment of foreign affiliates and leading to a faster and increased realization of productivity effects due to existing market knowledge and experience. Similarly, higher trade volumes may encourage TFP advances due to efficiency-seeking FDI. In particular, vertical investments and potential subsequent productivity effects require an increased engagement in trade. As a result, prior experience and established trade structures as well as high concurrent trade volumes may facilitate productivity advances to be reaped. Furthermore, TFP effects due to technology sourcing FDI may be increased via international trade activities. In particular, trade may allow access to more sophisticated foreign intermediate products and motivate reverse engineering activities with regard to both technologically advanced final and intermediate goods. As a result, trade may enhance home country technological capabilities that may facilitate the acquisition of foreign advanced technologies via OFDI.
Additionally, OFDI can in general be presumed to be correlated with further potential sources of technology and TFP development, which are thus also to be included in the current empirical set-up. Specifically, R&D expenditures (cf. Bitzer and Görg, 2009) and domestically provided credit control for national prospects of technology and productivity advances. Furthermore, the
6
labor ratio is added so as to proxy for home country technological sophistication. As such it might determine both the propensity to engage in OFDI and the scope to successfully source foreign technology and assimilate it domestically. Moreover, in line with the notion of ‘appropriate technology’ introduced in Section 2.2.1. the capital-labor ratio may also be crucial to include so as to obtain unbiased target and sending country results. More precisely, home countries with similar capital-labor ratios and thus similar technologies as their corresponding target economies may be induced with a higher capability for successful technology sourcing. As a last control variable, the percentage of the total labor force with tertiary education is included so as to control for human capital (cf. Driffield et al., 2009). In general, human capital may directly affect productivity levels while at the same time increasing a country’s absorptive capabilities (Krammer, 2010). In the present context this amounts to the idea that higher levels of human capital increase the propensity to engage in foreign investments and influences the scope so as to realize productivity advances due to OFDI.
In general, all variables are measured in US dollars and constant prices of 2000. Data on trade, R&D expenditures, and domestic credit are provided as the percentage of current GDP. Hence, they are multiplied by current GDPs and converted into constant dollars of 2000 using the consumer price index from the OECD. In addition, they are scaled to millions so as to have the same dimension as OFDI. In contrast, tertiary educated labor force is quantified as the percentage of total labor force.
Including these control variables a general-to-specific estimation approach is applied. Starting form a general estimation including all control variables these are subsequently removed if depicted to be statistically insignificant. The underlying motivation is the fact that the current research setting does not derive regression equations from an exactly identified theoretical relationship or precisely specified economic model. Thus, the validity of control variables can only be assessed by including them in the regression equation. The consequential idea of the general-to-specific methodology is to first define a general empirical model that contains all potentially relevant information. By subsequently eliminating insignificant control variables in parsimonious regression equations this information is refined and perfected (Hoover and Perez, 2004). Specifically, the presence of irrelevant variables can be shown to reduce the accuracy of coefficient estimates of other included variables (Hill et al., 2012), thereby motivating the exclusion of insignificant controls so as to obtain precise estimation results and efficient standard errors for the variables of interest.
4.2. Econometric Model
In order to formally analyze and answer the research questions introduced in 1.1. these data are applied in various econometric models.
( ) ( ∑ ∑ ) [ ( ∑ ∑ ) ( )] [ ( ∑ ∑ ) ( )] ( ) ( ) ( )
So as to capture the previously established proposition that trade has a moderating influence on the relationship between OFDI and TFP interaction terms between aggregate foreign investments and total merchandise trade are incorporated. Specifically, since the above depicted line of argumentation refers to both concurrent and prior trade effects, aggregate OFDI is interacted with both present and past trade volumes at times and , respectively, thereby allowing for a more dynamic relationship to be estimated. Furthermore, extant trade and its one-year lag are in addition included individually so as to control for direct TFP influences. The remaining previously introduced control variables are included in and enter the model in additive form. The calculation of both dependent and independent variables in natural logarithms represents the common functional form used in the extant academic literature that deals with the effects of OFDI on TFP (see e.g. Bitzer and Kerekes, 2008; Bitzer and Görg, 2009; Zhao et al., 2010), thereby ensuring a general comparability to previous results.7 It leads to the estimation of constant
elasticities thus depicting constant relative changes between these two variables of interest. In addition, marginal effects of OFDI are estimated to depend both on the level of foreign investments and on the level of TFP which denotes a reasonable assumption. Particularly, potential advances in TFP due to increasing levels of OFDI may be subject to decreasing returns, for the scope to realize both home country technology and efficiency improvements is reduced. Furthermore, the log-functional form encompasses the additional advantage of moderating the severity of heteroskedasticity (Liu and Wang, 2003). Specifically, log-transformed variables come closer to being normally distributed (Hill et al., 2012); a characteristic which is conveyed to the estimated error terms thereby increasing the prospect of homoskedasticity.
Importantly, as coefficient estimates obtained from equation (5) serve as pure benchmark results, they do not in themselves allow economic conclusions with respect to the research questions and hypotheses outlined above. Rather potential changes in coefficients’ signs, sizes or significance induced by subsequent target and sending country perspectives indicate that aggregate effects obscure differential impacts and thus further insights into the relationship between OFDI and TFP, thereby validating the current research framework.
7 The only variable not included in logarithmic form is human capital, for it is measured as the percentage of
In order to first incorporate a target country perspective in equation (5), home countries and continue to be pooled together while OFDI is distinguished according to investments in developed and less developed countries, thus yielding the following regression equation
( ) ( ∑ ) [ ( ∑ ) ( )] [ ( ∑ ) ( )] ( ∑ ) [ ( ∑ ) ( )] [ ( ∑ ) ( )] ( ) ( ) ( )
Coefficients to and to now capture aggregate home country TFP effects of OFDI directed towards developed and less developed host countries and , respectively. Lastly, these home country TFP effects are disaggregated for different sending country groups
and via a sample division. In other words, equation (6) is estimated separately for advanced and emerging home economies.
4.3. Estimation
Estimating these empirical models different econometric issues and techniques are considered. In particular, while first applying standard panel data estimation methods, a formal assessment of the possibly endogenous relationship between OFDI and TFP is conducted.
4.3.1. Fixed and Random Effects Estimation
to formally test whether random or fixed effects are more appropriate to use. To preview results, consistent with the previously depicted intuitive notion random effects are in each case found to yield inconsistent coefficient estimates (see Table B1 in Appendix B). Thus in all regression equations a fixed effects model is employed via Least Squares Dummy Variables (LSDV) estimation.
Regardless however of either fixed or random effects applications, taking time-constant individual heterogeneity into account has the additional advantage of reducing within-individual error-autocorrelation thereby further mitigating the potential problem of heteroskedasticity. Nevertheless, as heteroskedasticity leads to incorrect standard error calculations its presence is formally tested via applying a Breusch-Pagan test. In case error variances cannot be shown to be constant heteroskedasticity-robust standard errors are obtained so as to derive valid inferences about statistical significance and hypotheses.
4.3.2. Endogeneity and 2SLS Estimation
Nevertheless, another issue in the current research setting is whether OFDI and trade are indeed exogenous. In particular, while OFDI and trade may affect TFP by various previously depicted mechanisms, the literature on heterogeneous firms suggests the most productive firms to self-select into international activities (see e.g. Helpman et al., 2004). Specifically, entering foreign markets via direct investments or exports is assumed to be costly thus requiring certain prior firm-specific productivity levels so as to be able to meet these costs. In other words, the causality between TFP and OFDI as well as TFP and trade might be bidirectional. High levels of home country TFP might reflect a high density of productive firms and thus an eminent propensity of high country-level international activities in terms of both OFDI and trade. Thus, in the above described empirical set-up both forms of international activity may be subject to endogeneity due to simultaneity.
promising choices for valid instruments in terms of orthogonality.8 In addition, following Blundell and Bond (2000) OFDI and trade from countries and may also be instrumented with their respective lagged levels and lagged differences. Thus, instrumental overidentification can be achieved for both OFDI and trade of each cross-sectional unit.
Yet, before applying the DWH test the validity of each set of instrumental variables is formally assessed. First, given that each regression equation has multiple potentially endogenous variables the predictive power is tested via calculating first-stage Angrist-Pischke (AP) F-statistics for the joint significance of the excluded instruments.9 Importantly, a Monte Carlo simulation due to Stock, Wright and Yogo (2002) indicates that the respective values should be larger than 10 so as to ensure a sufficient predictive power (Angrist and Pischke, 2009). Secondly, in all econometric models the orthogonality condition is examined via the Sargan statistic. The corresponding null hypothesis being tested assumes the required validity of all instrumental variables in terms of zero error-correlation. As a result, a sufficiently high p-value is necessitated so as to not be able to reject it. Nevertheless, given that these test statistics are invalid under heteroskedasticity, a Pagan-Hall test is conducted in each case so as to formally test for constant error variances. In case this underlying null hypothesis can be rejected heteroskedasticity-robust test statistics are obtained. In particular, the heteroskedasticity-robust equivalence of the Sargan statistic is the Hansen-J statistic.
Having formally assessed instrument validity for each regression equation, the DWH test for endogeneity is applied. To preview results, OFDI and trade are in each case found to be endogenous. (The corresponding IV and DWH test statistics are summarized in Appendix C.) As a result, two-stage least squares (2SLS) estimation results are obtained. Importantly however, for these to yield consistent coefficient estimates the just described verification of instrument validity is crucial. Specifically, if instruments are weak 2SLS estimates can be shown to be biased (Angrist and Pischke, 2009). Moreover, instrumental variable estimation is generally less efficient than OLS (Baum et al., 2002). Therefore, in each case both LSDV as well as 2SLS results are reported.
5. Estimation Results
Applying the just described estimation techniques to the economic models introduced above, various insights into the relationship between OFDI and home country TFP are obtained.
8
Nevertheless, it should be noted that if countries and invest in or trade with and , respectively, OFDI and trade of and may well exert influences on TFPs in countries and . Yet, countries’ and investments and trade volumes may only be a small part of total inward investments and trade relationships of and , thereby not crucially determining ’s and ’s TFPs individually.
9
5.1. Descriptive Statistics
Shortly summarizing the main insights, emerging sending countries firstly display significantly lower levels of TFP as to be expected. At the same time however, they also experience higher average TFP growth rates over the sample period under consideration. Regarding OFDI activities advanced home economies exhibit substantially larger OFDI stocks in both target country groups than their emerging counterparts. Examining the respective within sending country group target distributions both advanced and emerging home countries maintain relatively higher OFDI stocks in developed host economies. Yet, the differences in these target distinguished OFDI activities is lower for emerging sending countries, thereby displaying a comparatively more equal distribution. Nevertheless, OFDI growth rates suggest advanced home economies to reduce their target country gap, for they experience on average higher increases in OFDI stocks in less developed target countries. In contrast, for emerging sending economies the respective growth rates in target-distinguished investment activities are similar for both developed and less developed countries. 5.2 Benchmark Results
To begin with, aggregate benchmark results obtained from estimating equation (5) are depicted in Table 3. In general however, it should be noted that high R-squared values observed for all subsequently reported regressions should not be overrated. Specifically, the primary cause may be the LSDV-inherent estimation of home country indicator variables adding significant explanatory power to the model specifications.
5.2.1. Benchmark LSDV Estimations
The first two columns of Table 3 display fixed effects LSDV results. According to the specific estimation presented in column (2) the marginal effect of OFDI on TFP is calculated as
10 Thus, both concurrent and prior trade volumes negatively affect otherwise positive TFP advances to be achieved via OFDI. Yet, so as to allow for a substantively more meaningful interpretation the marginal effect of OFDI is evaluated at average sample values of the relevant trade statistics, the result of which is reported at the bottom of Table 3. Specifically, a one-percent increase in OFDI is revealed to enhance home country TFP on average by 0.002 percent. However, this effect is statistically insignificant.
5.2.2. Benchmark 2SLS Estimations
Comparing this result to the specific 2SLS estimation displayed in column (4) the marginal effect of OFDI becomes
10Notably, despite the inclusion of interaction terms in the logarithmic model specification, the marginal effect of
