Do R&D investments in weak IPR countries destroy market value? The role of internal linkages
Abstract
Research Summary
The growth of emerging economies has attracted R&D investments by multinational enterprises, but firms have struggled to protect their knowledge assets in these environments with weak intellectual property rights protection. Knowledge misappropriation may be reduced if firms use cross-unit R&D teams to strengthen intra-firm interdependencies and control. We examine the relationship between such internal linkage strategies, foreign R&D investments, and firm market valuation in a dynamic market valuation model for 117 leading multinational firms. While foreign R&D investments are positively associated with market value, IP risks in host countries reduce it. The latter effect disappears if firms have developed a pronounced internal linkage strategy in weak IP environments. Linkage strategy bears costs and, in the absence of IP risks, is negatively associated with market value.
Managerial Summary
The growing market potential of emerging economies has led to an increase in research & development (R&D) activities there by multinational firms. A known challenge to multinational R&D investors in these economies is the weak protection of intellectual property rights (IPR). Firms may seek to reduce the risks of local knowledge spillovers and IPR infringement by actively embedding internal linkages in the organization of R&D and relying on cross-unit international R&D teams. We examine the consequences of such an internal linkage strategy for the performance effects of firms’ investments in weak IPR countries. Based on an analysis of 1763 cross-border R&D investments by 117 leading multinational firms, we find that a firm's market value is negatively affected by new R&D investments in weak IPR countries, but that such a negative influence can indeed be mitigated if the firm's R&D organization embeds internal linkages. Consistent with this observation, we demonstrate that firms with an internal linkage strategy are able to limit local knowledge outflows in weak IPR environments. However, if an internal linkage strategy is applied in strong, rather than weak, IPR environments, it bears substantial costs and negatively affects performance.
1 INTRODUCTION
The globalization of Research and Development (R&D) by multinational enterprises (MNEs) has grown significantly in recent years, with emerging economies featuring as major recipients of R&D investments (Branstetter, Glennon, & Jensen, 2019; Lewin, Massini, & Peeters, 2009; OECD, 2016).11
In the current study, we document that more than one third of global cross-border R&D investment projects were directed at the BRICS countries (Brazil, Russia, India, China, and South Africa), between 2003 and 2014.
Foreign R&D investments in emerging economies are motivated by the need to support local manufacturing operations, to adapt products to growing local markets, and to benefit from lower wage costs and an abundant supply of researchers and engineers (e.g., Belderbos, 2003; Kuemmerle, 1999).
At the same time, however, a known challenge to MNEs investing in emerging economies is the weak protection of Intellectual Property Rights (IPR) they are often confronted with, as shown in several surveys (e.g., EIU, 2004; Potters, Grassano, & Tübke, 2017; Schmiele, 2013). For instance, the US Trade Representative continues to voice concerns about weak IPR protection in China—despite a series of legal revisions (USTR, 2017, p. 102), while Schmiele (2013) observed that foreign R&D operations increase the probability of IPR infringement. Such infringement and misappropriation by local rivals can seriously threaten the MNE's market position and future profitability. An example is the experience of the SI Group, a leading U.S. rubber resin producer, that claimed that the theft of its resin formula in China had allowed a local rival, Sino Legend, to build a dominant market share on the Chinese market, replacing the SI Group (APFC, 2014).
Recent studies have suggested that one way that firms can reduce the risk of knowledge spillovers and misappropriation by rivals is to develop capabilities to integrate and build on technologies developed in different R&D units across weak and strong IPR regimes (Zhao, 2006: 1186–1188). Multi-location firms, and a fortiori MNEs, have the possibility to partition the knowledge generation process and distribute it across multiple locations, with internal linkage across units ensuring knowledge integration (Alcácer & Zhao, 2012; Belderbos & Somers, 2015; Nandkumar & Srikanth, 2016; Zhao, 2006; Zhao & Islam, 2017). This organization of R&D characterized by strong internal linkages and cross-unit interdependence in conjunction with a segmentation of R&D activities across locations makes it more difficult for collocated (rival) firms to access (the full range of) knowledge assets and set of technologies required to develop and commercialize innovations. Alcácer and Zhao (2012) confirm that knowledge flows between firms with collocated R&D activities are reduced if firms make use of such internal linkages. Zhao and Islam (2017) exam globally operating pharmaceutical firms and observed smaller local spillovers from R&D units of large firms, perhaps due to stronger IP control systems and internal linkage opportunities. Belderbos and Somers (2015) find that internal linkages of regional R&D leaders in Europe reduce the incentives of rival firms to collocate. Zhao (2006) shows that in countries with weak IPR protection internal linkage strategies are more frequently observed, while Nandkumar and Srikanth (2016) show that such strategies are associated with a reduced involvement of local inventors.22
In addition, Berry (2017) finds that strengthening control over foreign subsidiary operations in weak IPR countries by employing expatriates increases parents’ knowledge transfer to the subsidiaries -which is likely to be related to lower (perceived) IP risks.
Zhao (2006) argues that only firms that have developed the capabilities to embed internal linkage strategies in their R&D organization and to organize for effective complementarity knowledge development activities across locations can benefit from R&D investments in weak IP environments.
Although prior research has reported positive returns to foreign R&D activities in general for MNEs in terms of productivity growth and innovation output (e.g., Belderbos, Lokshin, & Sadowski, 2015; Griffith, Harrison, & Van Reenen, 2006; Kafouros, Wang, Mavroudi, Hong, & Katsikeas, 2018), the challenges and performance consequences of conducting R&D in countries with weak IPR protection have not been investigated. Similarly, the literature on internal linkage strategies has not examined whether and under which circumstances such strategies, that are arguably costly to implement (Alcácer & Zhao, 2012; Zhao, 2006), indeed have an expected positive financial return. This paper contributes by examining the firm performance effects of foreign R&D investments in weak IPR countries in conjunction with the internal linkage strategy developed by the MNE. As we are interested in the misappropriation hazard hampering future profitability, we examine firms' market valuation as a forward looking measure of performance. We relate this to announcements on new cross-border R&D investments and IP risks in the countries of those investments in conjunction with the internal linkage strategy characterizing the firm's existing R&D organization. We derive the specification for estimation from a dynamic market valuation model, allowing for a gradual convergence in firm performance and the identification of the effects of new R&D investments on firm value. The analysis controls for other influences on market value such as R&D expenditures and patent applications, and corrects for potential endogeneity bias by employing generalized method of moments estimation (system-GMM).
We draw inferences from panel data (2003–2014), both on a yearly and monthly basis, on 117 leading MNEs across manufacturing and knowledge intensive service industries, investing in 1,763 foreign R&D projects in 68 countries. Our empirical results show that the generally positive effects of new cross-border R&D investments on market valuation can turn negative if the new R&D investments are located in countries with weak IP protection. These negative consequences of new R&D investments in weak IPR countries are reduced and can be fully neutralized if firms have developed a pronounced internal linkage strategy in the context of weak IP protection environments. In contrast, if such linkage strategies are developed in strong IP protection environments, they do not reduce the harmful consequences of IP risks associated with new R&D investments and have a negative association with market value. We also show that the new R&D investments in weak IPR environments by firms that have developed a pronounced internal linkage strategy exhibit significantly smaller local knowledge outflows in such environments.
Our study contributes to the literatures on international R&D, knowledge spillovers and IP protection strategies, and the performance effects of R&D. Our findings help reconciling the twin phenomena of rising R&D investments in weak IPR environments and MNEs' concerns about the potential misappropriation of knowledge and technologies in such environments.
2 THEORY AND HYPOTHESES
Our study of the market valuation consequences of R&D investments abroad has antecedents in prior work regarding the relationship between R&D and firm market valuation (Blundell, Griffith, & Van Reenen, 1999; Hall, Jaffe, & Trajtenberg, 2005; Kanwar & Hall, 2017; Sandner & Block, 2011). This literature has argued and observed a positive influence of R&D assets on market value, which is ascribed to the future revenue generating potential of such assets.
Foreign R&D activities provide firms with a greater opportunity to access knowledge assets and R&D resources that are not easily acquired in their home countries, and facilitate local adaptation of existing products in order to increase sales in foreign markets (Belderbos, 2003; Kuemmerle, 1999; Lewin et al., 2009; Liu & Chen, 2012; Song, Asakawa, & Chu, 2011; Zhao, 2006). Prior studies have suggested that foreign R&D can increase MNEs' productivity and innovation performance (Belderbos et al., 2015; Griffith et al., 2006; Kafouros et al., 2018; Lahiri, 2010). We may thus expect that R&D investments abroad are positively related to firms' market valuation.
However, if these R&D investments are located in countries with weak IP protection, MNEs may be confronted with infringement and misappropriation of their knowledge assets by local rivals. This can seriously threaten the MNE's market position if rivals can benefit from infringement by bringing products (more rapidly) to the market, hampering the MNE from reaping the (full) benefits of its R&D assets. Such unintended knowledge outflows and associated misappropriation can occur because collocation and spatial proximity enhance knowledge spillovers (e.g., Belenzon & Schankerman, 2013; Peri, 2005). An important channel of knowledge spillovers is inter-firm mobility of scientists and engineers employed by the firm (Agarwal, Ganco, & Ziedonis, 2009; Kim & Marschke, 2005; Schmiele, 2013). Firms have fewer (legal) measures under weak IPR regimes to prevent misappropriation stemming from mobility of their R&D workers (Ganco, Ziedonis, & Agarwal, 2015; Garmaise, 2009).
Unintended knowledge spillovers may stimulate the birth of imitators that steal market share of the investing firm, and such “business stealing effects” are likely to reduce expectations of future profits and hence market value (Bloom, Schankerman, & Van Reenen, 2013). Prior work on IP protection, in the specific context of differences in trade secret protection laws between US states, has shown that differences in IP protection influence investor expectations and affect firm market value (Castellaneta, Conti, & Kacperczyk, 2017).
The above arguments imply that, while new foreign R&D investments abroad are expected to increase market value, weak IPR protection in the host countries of these investments is likely to reduce it. This leads to the following baseline hypotheses:
Hypothesis 1.New R&D investments abroad by an MNE are associated with higher firm market value.
Hypothesis 2.Weaker IPR protection in the countries in which an MNE invests in R&D is associated with lower firm market value.
In order to prevent knowledge leakages from foreign R&D units to local rival firms in countries with weak IPR protection, MNEs can and do consider concentrating R&D in their home country instead (Belderbos, Leten, & Suzuki, 2013; US OTA, 1994). However, this is not always an option for firms aiming to expand sales in foreign markets through successful local adaptation or to leverage cost effective (or unique) R&D resources in foreign countries. Prior studies have suggested that one way to reduce the risk of substantive knowledge leakage to rivals is to organize geographically dispersed R&D relying on internal linkages across R&D units (Alcácer & Zhao, 2012; Belderbos & Somers, 2015; Nandkumar & Srikanth, 2016; Zhao, 2006; Zhao & Islam, 2017). Formal models of modular organizational design with cross-unit interdependence have also shown the potential positive impact on intellectual property protection of this design, as agents (R&D workers) only have access to partial knowledge and have reduced bargaining power and incentives to leave the focal firm (Baldwin & Henkel, 2015).
Cross-unit and in particular cross-country linkages in R&D projects, related to intra-firm collaboration among researchers of the firm based in different locations, restrict the formation of local specialized capabilities and hence collocated rival firms' potential access to the full set of knowledge elements needed to develop and commercialize a technology. This organization keeps complementary knowledge resources residing in other units well protected, while the co-specialized nature of knowledge assets reduces the potential for imitation (Teece, 1986). The R&D organization with internal linkages reduces the scope and value of knowledge that R&D personnel may bring to collocated firms if R&D workers would leave the MNE and change employment. Internal linkage strategies furthermore allow for better monitoring and control of local R&D activities and a swifter response to misappropriation threats. Alcácer and Zhao (2012) show that such internal linkages are associated with lower knowledge spillovers to collocated rival firms, but did not examine the influence of heterogeneous local IPR protection conditions on the performance effects of linkage strategies. Zhao (2006) suggests that an internal R&D organization involving the use of internal linkages should be considered as a capability that MNEs develop to effectively transfer, integrate, utilize and protect their global knowledge in the context of important disparities in local IPR regimes. Hence, only those MNEs that have built up such capabilities to cope with weak IPR environments are expected to be able to apply an internal linkage strategy effectively to new R&D investments.
Internal linkage strategies will have the greatest potential to reduce knowledge spillovers and misappropriation by local firms if IPR policies provide little protection against such misappropriation. Hence, the negative consequences of foreign R&D investment in weak IPR environments, in terms of undesired knowledge outflows and reduced ability of the MNE to appropriate the returns to R&D, are likely to be reduced if the MNE has developed an R&D organization featuring an internal linkage strategy to cope with weak IP environments. This suggests that the negative influence of new R&D investment in weak IPR countries is attenuated for MNEs that have developed such an internal linkage strategy:
Hypothesis 3.The negative association between weak IPR protection in the countries in which the MNE invest in R&D and firm market value is attenuated if the MNE has developed an internal linkages strategy in weak IP environments.
There are also potential drawbacks and costs of an internal linkage strategy (Alcácer & Zhao, 2012). The distribution of tasks for an R&D project across various R&D units requires well designed protocols and interfaces, which need to be developed (Zhao, 2006) and requires allocation of resources to coordination efforts. Exchanging complex knowledge and collaborating at large geographic distance will either involve costly and time consuming travel or more limited face-to-face contact that are of prime importance for tacit knowledge exchange (e.g., Castellani, Jimenez, & Zanfei, 2013; Lane & Lubatkin, 1998). Second, collaboration and partitioning of knowledge development to enhance control and limit substantial knowledge outflows may not avoid a degree of redundancy in R&D activities across locations and may not necessarily be compatible with an allocation of R&D tasks corresponding with relative capabilities across units. Third, an internal linkage strategy to reduce knowledge spillovers implies less autonomy for local units. Extant research has suggested that a lack of autonomy of, and delegation to, local units and their researchers may stifle creativity, motivation, and project efficiency (e.g., Gambardella, Kashabi, & Panico, 2020), with as potential consequence reduced effective new product development (Beugelsdijk & Jindra, 2018).
Because an internal linkage strategy involves costs and potential inefficiencies, such a strategy is only expected to have positive effects on firm market value if it serves its purpose in terms of mitigating risk of knowledge spillovers and misappropriation. If such strategies are not applied to address R&D knowledge spillover concerns in weak IP environments, the costs may outweigh the benefits. Hence, in the absence of new R&D investments in high IP risk locations, an internal linkage strategy is likely to have a negative association with market value:
Hypothesis 4.In the absence of R&D investments in locations with weak IPR protection, there is a negative association between market value and internal linkage strategy.
3 DATA, MEASURES AND EMPIRICAL MODEL
3.1 Sample and data
We employ panel data on 117 leading multinational firms headquartered in 20 OECD countries, 2003–2014, to test our hypotheses. The sample firms are among the top 10 largest players in the European market in their respective industries (in terms of sales). The focus on European market leaders stems from the use of secondary data gathered to examine technology and market leadership in Europe (Commission of the European Communities, 2010). From this exercise, 162 publicly listed firms are identified with technology development (patent) activities. Among these 162 firms, 117 invested in new cross-border foreign R&D projects in at least 1 year during 2003–2014 (the period for which we have foreign R&D investment data) and are included in the analysis. We focus on firms with R&D investments as this allows to examine the differences in performance consequences for these firms depending on the IP environment of the countries in which the firms invest and the firms' internal linkage strategy—the focus of our research.33
The non-investing firms often have not applied for patents or applied only for a handful of patents, which prohibits the construction of a reliable internal linkage variable. Including observations of firms that never invested furthermore would add no variance to the focal variables and effects of interest. We discuss possible selection issues in the supplementary analysis section.
Advantages of the empirical setting are frequent observations on foreign R&D projects in a variety of countries, and the variation over home countries and industries. Despite the focus on European market leadership, the firms are also headquartered in the U.S. (27) and Japan (13) in addition to European countries, such as France (18), Germany (14), the United Kingdom (13), Italy (4), the Netherlands (4), and Switzerland (4). The firms are active across a range of industries, including technology intensive sectors such as electronics and electrical equipment (30), pharmaceuticals & chemicals (20), and automobiles & transportation equipment (16), but also encompassing industries such as food, beverages and tobacco (15).44
In a separate appendix, we list the sample firms and their country of origin, industry, and number of foreign R&D investments.
The panel is slightly unbalanced due to delistings and mergers, which results in a panel of 1,350 firm-year observations. We also estimate monthly market valuation models to enable analysis of the consequences of individual R&D investments and to evaluate effects closer to the moment of the announcement. This analysis includes 16,046 firm-month observations.
Data on foreign R&D investments are obtained from the Financial Times' fDi Markets database. The fDi Markets database draws on media information, firm reports, and various other sources regarding foreign direct investments (FDI). Information is available on investing firms, types of FDI (among which R&D) and destination countries. The fDi Markets database claims to be representative for global greenfield FDI flows and has been widely used by international organizations such as the OECD, World Bank, and UNCTAD, and by scholars (e.g., Castellani & Lavoratori, 2020; Castellani et al., 2013; Castellani & Pieri, 2013; Crescenzi, Pietrobelli, & Rabellotti, 2014; see OECD, 2016 for more details). The sample firms invested in 1,763 foreign R&D projects during the observation period 2003–2014. The database provides the year and month of the investment announcement but not the specific date of each investment. We use the information on the month of the R&D investments to estimate a version of the dynamic market valuation model on a monthly panel.
Information on patents applied by the focal firms is drawn from the PATSTAT database, 2016 spring edition. We use patent family IDs to track the firms' global patenting activities (e.g., Berry, 2017). Patent information is retrieved at the consolidated level and includes patent applications of consolidated subsidiaries identified through ORBIS data on firms' global affiliates, annual reports, and data on M&As from the Zephyr database. The MNEs on average applied for 871 patents yearly. Finally, financial information was extracted from Worldscope.
3.2 Measures
3.2.1 Tobin's Q and R&D investments
The dependent variable is Tobin's Q, in the definition of Chung and Pruitt (1994), which has been widely used in the strategy, market value, and R&D literatures (Belderbos, Cassiman, Faems, Leten, & Van Looy, 2014; Blundell et al., 1999; Czarnitzki, Hall, & Oriani, 2006; Czarnitzki, Hussinger, & Leten, 2020; Griliches, 1981; Hall et al., 2005; Hall, Thoma, & Torrisi, 2007; Jaffe, 1986; Jayachandran, Kalaignanam, & Eilert, 2013; Kanwar & Hall, 2017; Li & Tallman, 2011; Petrenko, Aime, Recendes, & Chandler, 2019 and Sandner & Block, 2011). In our case, the dependent variable is derived from a market value equation in which market value is determined as the expected discounted future profits of the firm both due to tangible assets and intangible (knowledge) assets (Peters & Taylor, 2017). The core explanatory variables are Foreign R&D, IP Risk of Foreign R&D, and Linkage Strategy. Foreign R&D (Hypothesis 1) is the count of new cross-border R&D investments in a year, or month, by the focal firm. Linkage Strategy is the observable use of internal linkage strategy in weak IP environments in the existing R&D operations of the MNE (see below); we will use the abbreviated term for notational convenience in the remainder of the paper.
3.3 IP risk of Foreign R&D
and a country's score on Impartial Courts (IC) in the Economic Freedom of the World (EFW) report by the Fraser Institute. The GP and IC scores are normalized by the annual maximum value of each index, such that the composite index ranges from 0 to 1. The GP index is widely used in the literature and is based on statutory information on patent laws (e.g., Athukorala & Kohpaiboon, 2010; Belderbos, Lykogianni, & Veugelers, 2008; Branstetter, Fisman, & Foley, 2006; Nandkumar & Srikanth, 2016). An often voiced criticism of the index is that a statutory indicator will not effectively capture the actual enforcement level of IPR laws. We follow recent work (Hu & Png, 2013; Maskus & Yang, 2013) by multiplying the GP index by IC as a measure of enforcement of (patent) laws, with IC referring to the perceived impartiality and quality of the legal system for private firms. The measure of IP risk of Foreign R&D is 1 minus the index of IPR protection, summed over all destination countries of the R&D investments.66 In case of multiple R&D investments in a country, its IP risk enters this sum multiple times. We adopt this specification separating the overall “value” of IP Risk from the volume of Foreign R&D, such that we can identify the moderating effect of Linkage strategy on IP risk of Foreign R&D specifically.
We note that, since a firm can have multiple R&D investments in a year, the IP risk measure in the yearly panel models aggregate IP risk over the countries of investment. In the monthly panels, in contrast, we are able to assess IP risk of a single country. Hypothesis 2 suggests a negative sign.
3.4 (Internal) linkage strategy
Following prior studies (e.g., Alcácer & Zhao, 2012; Belderbos & Somers, 2015), (Internal) Linkage Strategy is measured as the degree to which inventors based in different countries contribute to firms' patents (cross-country inventor collaboration), as identified from inventors' addresses. First, we calculate the country dispersion of inventors listed on each patent, indicated by a Blau index (1 minus the Herfindahl index of the shares of inventors' resident countries listed on a patent). This dispersion index increases if inventors of a patent are based in multiple countries. Hence, it captures the extent to which the knowledge creation process is segmented and distributed over multiple (country) locations.77
While prior studies (e.g., Alcácer & Zhao, 2012) only identified whether or not a patent was the product of a cross-country inventor team, our measure also considers the degree of dispersion of inventors across countries.
Second, the effectiveness of an internal linkage strategy will depend on the country context in which it is used. In particular, the IP protection capabilities related to linkage strategy are likely to be salient if the strategy is developed in high IP risk contexts (Alcácer & Zhao, 2012). Hence, to take into account this “fit” of firms' internal linkage strategies and IP risk consistent with Hypothesis 3, we multiply the inventor dispersion score of each patent with the highest IP risk score among the countries of the inventors. Subsequently, to arrive at a firm-level variable for internal linkage, we average this weighted dispersion across all foreign invented patents of the focal firm.
The linkage variable is based on the MNEs' patents in the year t − 1 and hence reflects MNEs' strategy and capabilities in their existing R&D organization. Measuring this in t − 1 ensures that investors can observe this strategy and infer from this observation what the likely financial performance benefits are of new R&D investments by the firm in risky IP environments. Firms that have proven capable in their prior R&D activities to utilize internal linkages to increase control over, and limit the extent of, knowledge outflows in high risk locations, are expected to be able to apply such linkage strategy also to their new R&D investments. In supplementary analysis, we test and confirm this assumption. Measuring internal linkages in t − 1 in addition ensures that the focal R&D project investments in year t cannot influence it, avoiding potential simultaneity. Hypothesis 3 suggests a positive sign for the interaction of IP Risk of Foreign R&D and Linkage Strategy and Hypothesis 4 suggests a negative sign for Linkage Strategy if it is not used in conjunction with IP risks of Foreign R&D.
3.5 Control variables
We include a range of control variables that might influence the expectations on future performance of firms. Positive future profit expectations related to foreign R&D investments may be heterogeneous and depending on the market potential and technological capabilities of destination countries, reflecting market access and technology sourcing motivations, respectively (Kafouros et al., 2018; Kuemmerle, 1999; von Zedtwitz & Gassmann, 2002). The analysis therefore controls for GDP and the ratio of the number of relevant patents over GDP in the destination countries of firms' R&D investments (averaged over the countries of investments in the year). We match the detailed sector of the focal firms' investment projects listed in the fDi Markets database with (NAICS) industry codes, using a published concordance table. These industry codes are subsequently matched with a technology field classification (Lybbert & Zolas, 2014). We count those patent families invented in a potential host country that have technology fields (IPCs) that correspond to the sector of the investment.
We also control for a potential confounding alternative effect of internal linkage strategies. Prior studies on the performance effects of foreign R&D have suggested that innovation performance is enhanced by intra-firm collaboration, due to its potential to enhance knowledge integration and cross-fertilization (Frost & Zhou, 2005; Lahiri, 2010; Singh, 2008). In order to isolate the IP protection benefits of linkage strategies from this other potential influence, we construct and include the variable Linkage for knowledge sourcing. The variable is constructed as the weighted average of host country's patent stock in sector of the firm, using the share of the firm's foreign inventors involved in cross-country internal linkages with the host country as weights.
The analysis also includes conventional control variables in the literature relating to information that becomes available to stock market investors and that may change the expected future cash flows of the firm. We follow prior studies in including firm R&D expenditures, patents over R&D, forward citations to patents, and the importance of self-citations among these citations (e.g., Belenzon, 2011; Czarnitzki et al., 2020; Hall et al., 2005). Citation lags are restricted to a 4-year window including the application year. Since patent applications are not always directly visible to investors upon filing, we take a 1 year lag for the patent-related control variables (Belderbos et al., 2014). Finally, the equation for estimation includes the lagged value of Tobin's Q and sets of dummies for the industries of investment (24 sectors), home country, and time (years or months). We discuss the model for estimation in the next paragraph.
3.6 Empirical model
:
(1)The (proportional) growth in Tobin's Q is a function of the past level of Tobin's Q, the growth in physical assets (Δa), R&D expenditures (RD), the number of new Foreign R&D investments (Inv) of the firm, the IP risk of Foreign R&D (Risk) in the countries of these R&D investments, and the interaction between IP risk of Foreign R&D and Linkage Strategy (LS). The main effect of Linkage Strategy (LS) measures potential cost effects. The equation is augmented with the control variable Linkage for knowledge sourcing (LS_KS) and a set of variables ∑rμrΡrit representing R&D host country characteristics (GDP and technological capabilities) and patent and citation related variables. In addition, sets of sector, year, and home country dummies are included (the vector 
) and there is a firm-specific component added to the growth equation (τi),while εit are serially uncorrelated error terms.
3.7 Estimation strategy
We estimate Equation (1) both on yearly and on monthly data. The advantage of exploiting monthly variation is that it allows focusing on the market valuation consequences of single R&D investments in one particular country, whereas in the yearly panel firms often have multiple R&D investments in a year. The finer granularity of monthly data avoids aggregation of the R&D investment and IP risk variables and allows analysis of the consequences of individual R&D investments closer to the moment of the announcement. To identify effects of single R&D investments, we leave out 240 firm-month observations with multiple investments, and arrive at a panel of 16,046 firm-month observations on 116 firms.99
One firm (Borealis) lacked monthly market value information and had to be omitted from this analysis.
Employing monthly data, however, also has its limitations and drawbacks in our setting. Since no monthly information is available on accounting data such as R&D expenditures and capital, we have to allow imprecision in the measurement of other key drivers of market value in the dynamic model. We could measure market value as an end of month variable, and we could turn patent variables into monthly data by looking at publication dates. We used quarterly accounting reports data for assets with interpolation to arrive at a monthly measure. R&D expenditures are yearly data and cannot be assigned to months, since the distribution of R&D investments is unknown to investors and probably even to the firm; hence we divide R&D by 12. Internal linkage strategy continues to be measured in the year prior to the investment.
We estimate Equation (1) in level form—bringing the past level of Q to the right hand side—with the System General Method of Moments (GMM-SYS) method due to Blundell and Bond (1998). This approach is suitable for panel data models that include lagged values of the dependent variable, in which case conventional regression analysis and fixed and random effects models are inappropriate (Arellano, 2003). The latter methods are known to provide biased estimates, in particular in short panels such as our yearly panel, or when the autoregressive parameter is high (Bun & Carree, 2005; Hsiao, 2014; Nickell, 1981). GMM-SYS delivers unbiased and consistent parameter estimates and can correct for potential endogeneity of the regressors by employing lagged values of the regressors as instruments (Roodman, 2009b). This allows for the introduction of predetermined, but not strictly exogenous variables, such as the past realizations of the dependent variable (Kripfganz, 2016). The method also accommodates unobserved heterogeneity through a firm-specific component in the level equations and enhances estimates in the presence of autocorrelation and heteroscedasticity. Because of these properties, strategy scholars have increasingly adopted this estimation method when estimating dynamic panel models (e.g., Alessandri & Seth, 2014; Chizema, Liu, Lu, & Gao, 2015, Girod and Withington, Girod & Whittington, 2017, Gómez & Maícas, 2011, and Milanov & Shepherd, 2013).
GMM-SYS estimates both a level equation and an equation in first differences to improve identification and efficiency. In the level equation we include sector, year, and home country dummies as exogenous variables. To avoid instrument proliferation, we limit the number of lags for the instruments to three (Roodman, 2009a) and impose equal moment conditions, restricting the number of instruments to one for each variable and lag. Following Alessandri and Seth (2014), we further strengthen the instrument set by including two additional variables as exogenous instruments for the key focal variables Foreign R&D investment and IP risk (of Foreign R&D) exposure, to enhance identification. We use the number of R&D entries in the year and country by non-focal firms from the same home country in other industries as an external instrument for R&D investment, and the weighted IP risk of R&D investments by such other firms as an external instrument for the IP risk variable. R&D investment behavior by loosely connected firms is an indicator of shocks across years and countries that also affect R&D decisions of focal firms, and information on investments by same home-country firms is also more likely to reach the focal firm. By limiting R&D investments to firms outside the focal industry, we avoid that the instrument is correlated with the dependent variable.1010
The specification relying on R&D investments outside the focal industry follows the logic of the R&D investment and spillover literature (Zacchia, 2020). In contrast, same industry investments may impact on the focal firms' market valuation upon entry through industry agglomeration effects. We note that results are robust to the exclusion of the external instruments.
3.8 Basic statistics
Table 1 shows the main host countries receiving R&D investments from the 117 MNEs and these host countries' (weak) IP scores. The most prominent destinations of foreign R&D investments by the MNEs are China (271 projects), India (207), the U.S. (146), the U.K. (98), and Singapore (92). We observe important differences in the (weak) IP scores. Germany has the lowest IP risk whereas Brazil, India, and China have four to five times larger (weak) IP scores compared to Germany.
| Country | # projects | Share (%) | Weak IP score |
|---|---|---|---|
| China | 271 | 15.37 | 0.55 |
| India | 207 | 11.74 | 0.54 |
| USA | 146 | 8.28 | 0.23 |
| UK | 98 | 5.56 | 0.17 |
| Singapore | 92 | 5.22 | 0.21 |
| Spain | 82 | 4.65 | 0.49 |
| Brazil | 56 | 3.18 | 0.66 |
| Germany | 55 | 3.12 | 0.14 |
| Canada | 45 | 2.55 | 0.22 |
| France | 44 | 2.50 | 0.29 |
| Other countries | 667 | 37.8 | 0.55 |
| Total | 1763 | 100 | — |
- Note: Average scores, 2003–2014.
Descriptive statistics of the (transformed) variables and their correlations are shown in Table 2. The coefficient of correlation between IP risk of Foreign R&D and Foreign R&D is high (0.96) in the yearly data because IP risk of Foreign R&D by design increases in the number of R&D entries, and since IP risk of Foreign R&D is necessarily zero for firm-year observations with zero foreign R&D. In the monthly panel the correlation is reduced but still high (0.88), again due to the fact that IP risk is only observed for firm-month observations with a foreign R&D investment. It is the variation in IP risk among the subset of observations with investments that allows identification of the influence of IP risk. We also note that for a supplementary analysis (shown in a separate appendix), we instrument for both variables, which importantly reduces effective correlation: instrumented versions of the variables have much lower correlation coefficients: 0.25 (monthly) and 0.27 (yearly). We return to this issue and discuss alternative specifications of the estimated models in the supplementary analysis section.
| Yearly panel | Variable | Mean | SD | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Log(Q)t | 0.37 | 0.36 | ||||||||||||||
| 2 | Foreign R&D | 0.22 | 3.04 | 0.11 | |||||||||||||
| 3 | IP risk (of Foreign R&D) | 0.07 | 0.60 | 0.07 | 0.96 | ||||||||||||
| 4 | Linkage strategy | 0.05 | 0.05 | 0.08 | 0.00 | −0.00 | |||||||||||
| 5 | Linkage strategy—high risk | 0.08 | 0.11 | 0.08 | −0.03 | −0.04 | 0.82 | ||||||||||
| 6 | Linkage strategy—low risk | 0.05 | 0.08 | 0.05 | 0.10 | 0.09 | −0.03 | −0.47 | |||||||||
| 7 | Linkage for knowledge sourcing | 0.25 | 0.55 | 0.09 | −0.01 | 0.00 | −0.07 | −0.11 | 0.13 | ||||||||
| 8 | Log(Q)t − 1 | 0.37 | 0.37 | 0.88 | 0.11 | 0.07 | 0.07 | 0.08 | 0.06 | 0.10 | |||||||
| 9 | dlog(Assets) | 0.03 | 0.18 | 0.09 | 0.02 | 0.01 | −0.01 | 0.01 | −0.00 | 0.03 | 0.23 | ||||||
| 10 | R&D expenditure | 0.08 | 0.12 | −0.27 | 0.01 | 0.05 | −0.12 | −0.09 | −0.07 | −0.01 | −0.33 | −0.13 | |||||
| 11 | Patents/R&D | 0.01 | 0.06 | 0.20 | 0.43 | 0.39 | 0.04 | −0.04 | 0.16 | −0.03 | 0.20 | 0.00 | −0.05 | ||||
| 12 | Citations/patents | 1.87 | 1.63 | 0.31 | 0.03 | 0.03 | 0.05 | 0.11 | −0.01 | 0.02 | 0.33 | −0.04 | 0.01 | 0.00 | |||
| 13 | Self-citations/citations | 0.27 | 0.19 | 0.13 | 0.08 | 0.07 | 0.16 | 0.13 | 0.03 | −0.01 | 0.11 | 0.05 | −0.10 | 0.15 | −0.12 | ||
| 14 | Host countries' GDP | 1.40 | 1.43 | 0.04 | 0.07 | 0.11 | 0.01 | 0.05 | −0.06 | 0.01 | 0.05 | −0.04 | 0.11 | −0.01 | 0.16 | 0.02 | |
| 15 | Host technological capability | 0.02 | 0.06 | 0.08 | 0.01 | 0.02 | −0.01 | 0.02 | −0.03 | 0.03 | 0.10 | 0.03 | 0.06 | −0.00 | 0.17 | −0.03 | 0.29 |
| Monthly panel | Variable | Mean | SD | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Log(Q)t | −0.05 | 0.54 | ||||||||||||||
| 2 | Foreign R&D | 0.06 | 0.60 | −0.08 | |||||||||||||
| 3 | IP risk (of Foreign R&D) | 0.02 | 0.24 | −0.09 | 0.88 | ||||||||||||
| 4 | Linkage strategy | 0.05 | 0.05 | 0.08 | −0.01 | −0.01 | |||||||||||
| 5 | Linkage strategy—high risk | 0.08 | 0.11 | 0.08 | −0.02 | −0.02 | 0.82 | ||||||||||
| 6 | Linkage strategy—low risk | 0.05 | 0.08 | 0.06 | 0.01 | 0.01 | −0.03 | −0.47 | |||||||||
| 7 | Linkage for knowledge sourcing | 0.25 | 0.55 | 0.06 | 0.03 | 0.04 | −0.07 | −0.12 | 0.14 | ||||||||
| 8 | Log(Q)t − 1 | −0.05 | 0.54 | 0.99 | −0.08 | −0.09 | 0.08 | 0.08 | 0.07 | 0.05 | |||||||
| 9 | dlog(Assets) | 0.00 | 0.04 | 0.04 | −0.00 | −0.00 | −0.01 | 0.00 | 0.00 | 0.02 | 0.09 | ||||||
| 10 | R&D expenditure | 0.01 | 0.02 | −0.27 | 0.22 | 0.32 | −0.06 | −0.06 | −0.02 | −0.02 | −0.28 | −0.02 | |||||
| 11 | Patents/R&D | 0.00 | 0.01 | −0.07 | 0.00 | 0.00 | 0.02 | 0.03 | −0.03 | −0.01 | −0.07 | −0.01 | −0.02 | ||||
| 12 | Citations/patents | 1.24 | 1.62 | 0.18 | 0.01 | 0.02 | 0.00 | 0.06 | −0.02 | 0.03 | 0.18 | 0.00 | 0.03 | −0.03 | |||
| 13 | Self-citations/citations | 0.22 | 0.25 | 0.02 | 0.00 | 0.00 | 0.06 | 0.09 | −0.03 | 0.01 | 0.02 | −0.01 | 0.02 | 0.01 | 0.13 | ||
| 14 | Host countries' GDP | 2.07 | 7.37 | 0.02 | 0.33 | 0.33 | 0.02 | 0.04 | −0.03 | 0.01 | 0.01 | −0.00 | 0.04 | 0.03 | 0.08 | 0.03 | |
| 15 | Host technological capability | 0.09 | 1.17 | 0.03 | 0.06 | 0.06 | 0.00 | 0.03 | −0.02 | 0.02 | 0.03 | 0.01 | 0.01 | 0.00 | 0.06 | 0.01 | 0.30 |
- Note: Significant correlations (p < .05) are in bold. IP Risk of Foreign R&D and Foreign R&D are actual values.
It can be calculated that the mean of IP risk of Foreign R&D is 0.70 for firm-year observations above the median of Linkage Strategy, and 0.41 for firms below the median; this difference is significant (p = .000). This observation confirms the expectation that MNEs with strong IP strategies are more often investing in countries with weaker IPR protection where misappropriation risks are high, in line with the patterns observed in Zhao (2006), and of Alcácer and Zhao (2012) in the particular context of the semiconductor industry.
Other prima facie evidence can be obtained by comparing cases of market value responses to R&D announcements. For instance, we observe that Bayer, characterized by relatively high internal linkage capabilities, saw its market value increase in the months it announces R&D investments in China or Russia, while Astra Zeneca, with substantially lower internal linkage capabilities, saw its market value decline in the months of R&D investments around the same time in these countries. In the empirical analysis we examine if these varied patterns of market valuation consequences are systematic.
4 EMPIRICAL RESULTS
Models 1–7 in Table 3 present the coefficient estimates of the model for yearly panels. For all models, the GMM diagnostics confirm the validity of the sets of instruments (the Hansen J test is far from significant) and the Wald statistics confirm model significance. Model 1 shows results with control variables only, model 2 adds the two linkage strategy variables, and models 3–5 in turn add the focal variables to test the hypotheses. The incremental Wald tests confirm a significant increase in model fit due to the addition of the focal variables in models 3–5.
| (1) | (2) | (3) | (4) | (5) | (6) | (7) | |
|---|---|---|---|---|---|---|---|
| Dependent variable = Log(Q)t | Controls only | Linkage | H1 baseline | H1 + H2 | H1 + H2 H3 + H4 | Linkage low vs. high risk | Linkage low vs. high risk |
| Foreign R&D | 0.002 | 0.028 | 0.033 | 0.030 | 0.028 | ||
| (0.000) | (0.000) | (0.000) | (0.000) | (0.000) | |||
| IP risk of Foreign R&D | −0.138 | −0.636 | −0.526 | −0.398 | |||
| (0.000) | (0.002) | (0.008) | (0.024) | ||||
| IP risk*linkage strategy | 7.729 | 6.125 | |||||
| (0.014) | (0.040) | ||||||
| Linkage strategy | 0.806 | 1.079 | 0.973 | 0.794 | |||
| (0.221) | (0.153) | (0.190) | (0.256) | ||||
| IP risk*linkage strategy—high risk | 3.471 | ||||||
| (0.006) | |||||||
| IP risk*linkage strategy—low risk | 1.001 | ||||||
| (0.131) | |||||||
| Linkage strategy—high risk | 0.119 | −0.053 | |||||
| (0.549) | (0.856) | ||||||
| Linkage strategy—low risk | −0.537 | −0.595 | |||||
| (0.036) | (0.048) | ||||||
| Linkage for knowledge sourcing | 0.149 | 0.106 | 0.147 | 0.150 | 0.130 | 0.140 | |
| (0.082) | (0.171) | (0.088) | (0.082) | (0.086) | (0.000) | ||
| Log(Q)t-1 | 0.760 | 0.683 | 0.751 | 0.609 | 0.604 | 0.635 | 0.625 |
| (0.000) | (0.000) | (0.000) | (0.000) | (0.000) | (0.000) | (0.000) | |
| dlog(Assets) | −0.342 | −0.320 | −0.589 | −0.452 | −0.481 | −0.505 | −0.491 |
| (0.344) | (0.216) | (0.114) | (0.097) | (0.034) | (0.029) | (0.032) | |
| R&D expenditures | 0.314 | 0.181 | 0.043 | 0.183 | 0.372 | 0.319 | 0.058 |
| (0.145) | (0.197) | (0.773) | (0.180) | (0.016) | (0.060) | (0.677) | |
| Patents/R&D | 0.451 | 0.504 | 0.494 | 0.437 | 0.479 | 0.456 | 0.423 |
| (0.001) | (0.000) | (0.003) | (0.005) | (0.000) | (0.000) | (0.007) | |
| Citations/patents | 0.037 | 0.034 | 0.021 | 0.031 | 0.039 | 0.040 | 0.032 |
| (0.002) | (0.007) | (0.110) | (0.009) | (0.001) | (0.001) | (0.009) | |
| Self-citations/citations | −0.032 | 0.038 | −0.035 | 0.010 | 0.016 | −0.011 | −0.104 |
| (0.836) | (0.768) | (0.811) | (0.942) | (0.910) | (0.923) | (0.589) | |
| Host countries' GDP | −0.000 | −0.009 | −0.003 | −0.001 | 0.003 | 0.001 | −0.002 |
| (0.951) | (0.153) | (0.708) | (0.832) | (0.725) | (0.860) | (0.842) | |
| Host technological capability | −0.273 | 0.095 | −0.214 | −0.123 | −0.227 | −0.201 | −0.075 |
| (0.288) | (0.709) | (0.639) | (0.618) | (0.393) | (0.421) | (0.761) | |
| Constant | −0.056 | −0.066 | −0.116 | 0.084 | 0.040 | 0.191 | 0.316 |
| (0.487) | (0.556) | (0.254) | (0.458) | (0.705) | (0.081) | (0.003) | |
| Time/country/industry FE | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| # instruments | 107 | 113 | 116 | 119 | 122 | 125 | 128 |
| Hansen J p-value | 0.503 | 0.624 | 0.753 | 0.730 | 0.554 | 0.808 | 0.909 |
| Incremental Wald chi2 (p-value) | 5.3 (0.07) | 14.9 (0.00) | 38.9 (0.00) | 6.1 (0.01) | |||
| Model fit Wald chi2 (p-value) | 1,179.4 (0.00) | 1,216.8 (0.00) | 1,292.6 (0.00) | 4,107.6 (0.00) | 4,834.1 (0.00) | 5,477.3 (0.00) | 6,042.8 (0.00) |
| Observations | 1,350 | 1,350 | 1,350 | 1,350 | 1,350 | 1,350 | 1,350 |
| Number of firms | 117 | 117 | 117 | 117 | 117 | 117 | 117 |
- Note: Standard errors are clustered by firm. p-values are shown in parentheses. Incremental Wald tests show the significance of the improvement in model fit compared to the prior model. Variables are defined in Equation (1). The non-nested specification in models 6 and 7 precludes the calculation of incremental Wald tests. Linkage low vs. high risk is measured for the existing R&D organization of the firm in t − 1.
The control variables in model 1 and the more elaborate models display expected patterns in general, in particular in the fully specified model 5. The coefficient of the lagged Tobin's Q is positive and significant (β = 0.604,p < .000), with the estimate suggesting a convergence of 24 to 40% across the models 1–5. The coefficient on physical asset growth of −0.48 in model 5 (p = .034) suggests that the elasticity of market value with respect to physical assets is 0.52. The coefficient on R&D (β = 0.372,p = .016) suggests a marginal rate of return of 0.37, which is comparable to earlier findings in the literature (e.g., Czarnitzki et al., 2006; Griliches, 1981). The patent and citation indicators, except the self-citation rate, have positive and statistically significant coefficients. The coefficients of GDP and patent intensity of R&D host countries do not reach conventional statistical significance levels.
In model 2, the main effect of Linkage strategy is not significant, while Linkage for sourcing has a positive sign and is marginally significant (β = 0.149, p = .082). Model 3 shows that the coefficient of Foreign R&D is positive and highly significant (β = 0.002,p < .000) in support of the baseline hypothesis. The coefficient on Foreign R&D is also positive and significant in Model 4 and increases substantially in size (β = 0.028, p < .000), while IP risk of Foreign R&D has a negative and significant coefficient (β = − 0.138, p < .000), in support of Hypothesis 2. In Model 5, the coefficient on the interaction term of Linkage strategy and IP risk of Foreign R&D is positive and significant (β = 7.729,p = .014), consistent with Hypothesis 3. The coefficient on Linkage Strategy itself, indicating the effect in the absence of new R&D investments, is insignificant (β = 0.794, p = .256), such that Hypothesis 4 is rejected.
We assess the economic significance of the focal variables by calculating the proportional effect of a standard deviation increase in Foreign R&D and IP risk of Foreign R&D on Tobin's Q. The estimated effect of Foreign R&D in Model 5 suggests that a standard deviation increase in foreign R&D investment is associated with a 5.4% increase in Tobin's Q (p < .000). As the effect of IP risk of Foreign R&D is conditional on firms' Linkage Strategy in Model 5, we estimate proportional changes in Tobin's Q due to a standard deviation increase in IP risk of Foreign R&D at different percentiles of Linkage Strategy (Figure 1). A standard deviation increase in IP Risk of Foreign R&D is associated with a significant and substantial drop in market value of more than 10% if Linkage Strategy is in 5th percentile. This decline is reduced to 6.2% (p < .000) and 2.6% (p < .000) if Linkage Strategy is in the median and in the 75th percentile quartile, respectively. The effect of IP Risk of Foreign R&D is close to zero and becomes insignificant at the 85th percentile of Linkage Strategy (p = .794). Hence, the negative consequences of foreign R&D investments in high IP risk countries are mitigated and even neutralized for firms that have adopted a pronounced internal linkage strategy.1111
For the extreme case of the highest linkage strategy (percentile 100), the confidence interval around a positive effect of IP risk just suggests significance.
We did not find support for Hypothesis 4 suggesting that internal linkage strategies are likely to be associated with higher cost and potential lower efficiency of the R&D organization. One possible reason for this may be that we did not sufficiently consider if firms use the strategy in the most appropriate, high IPR risk, contexts and develop capabilities in such environments. We further examined this by differentiating between two Linkage Strategy variables, depending on whether the strategy developed for the firm's existing R&D activities is used in relatively high or relatively low IP risk environments. We take the second part of the Linkage Strategy construction—the mean of the highest IP risk score among the countries of the inventors. We distinguish whether this mean score for a firm-year observation is below or above the sample median. Linkage Strategy—high risk is the Blau index of inventor dispersion but takes only positive values if this mean of the risk score is above the sample median, else zero. Linkage Strategy—low risk in contrast is the Blau index of inventor dispersion taking only takes positive values if the mean is below the sample median. We adopt this specification both in the yearly and in the monthly models. Results are shown in models 6 and 7 of Table 3 and model 2 in Table 4. In the yearly model (model 6) the main effect of Linkage Strategy—high risk is insignificant (p = .549), while in contrast, Linkage Strategy—low risk is negative and significant (β = − 0.537,p = .036). This finding suggests that for the intensive use of cross-unit collaboration in relatively strong, rather than weak, IP environments the costs can indeed outweigh the benefits, providing qualified support for Hypothesis 4. In addition, we find in model 7, that when the two Linkage Strategy variables are interacted with IP Risk of Foreign R&D, that only the interaction of Linkage Strategy—high risk with IP Risk of Foreign R&D is significant (β = 3.471,p = .006). This provides further support for Hypothesis 3.
| (1) | (2) | |
|---|---|---|
| Dependent variable = Log(Q)t | H1 + H2 + H3 | Linkage low vs. high IP risk |
| New Foreign R&D Investment | 0.138 | 0.125 |
| (0.025) | (0.011) | |
| IP risk of Foreign R&D | −0.464 | −0.300 |
| (0.004) | (0.021) | |
| IP risk*linkage strategy | 7.161 | |
| (0.023) | ||
| Linkage strategy | −0.136 | |
| (0.125) | ||
| IP risk*linkage strategy—high risk existing R&D locations | 0.999 | |
| (0.084) | ||
| IP risk*linkage strategy—low risk existing R&D locations | 0.339 | |
| (0.556) | ||
| Linkage strategy—high risk existing R&D locations | −0.035 | |
| (0.411) | ||
| Linkage strategy—low risk existing R&D locations | −0.086 | |
| (0.070) | ||
| Internal linkage for knowledge sourcing | 0.050 | 0.047 |
| (0.060) | (0.060) | |
| Log(Q)t − 1 | 0.969 | 0.960 |
| (0.000) | (0.000) | |
| dlog(Assets) | −0.637 | −0.640 |
| (0.000) | (0.000) | |
| R&D expenditures | 0.017 | 0.228 |
| (0.943) | (0.228) | |
| Patents/R&D | −0.445 | −0.273 |
| (0.369) | (0.555) | |
| Citations/patents | −0.003 | 0.003 |
| (0.830) | (0.827) | |
| Self-citations/citations | −0.098 | −0.023 |
| (0.245) | (0.743) | |
| Host countries' GDP | −0.001 | −0.000 |
| (0.197) | (0.482) | |
| Host technological capability | −0.007 | −0.007 |
| (0.198) | (0.221) | |
| Constant | −0.074 | 0.033 |
| (0.009) | (0.227) | |
| Time/country/industry FE | Yes | Yes |
| # instruments | 133 | 139 |
| Hansen J p-value | 0.334 | 0.701 |
| Model fit Wald chi2 (p-value) | 4,802.4 (0.00) | 6,597.3 (0.00) |
| Observations | 16,046 | 16,046 |
| Number of firms | 116 | 116 |
- Note: Standard errors are clustered by firm. p-values are shown in parentheses. Variables are defined in Equation (1). Linkage low vs. high risk is measured for the existing R&D organization of the firm in t − 1.
The economic effects of Linkage Strategy developed in strong IPR environments rather than in risky IP setting are illustrated in Figure 2, which shows the proportional change in Tobin's Q due to a standard increase in Linkage Strategy—low risk based on the results of model 7, Table 3. The effects of Linkage Strategy depend on its moderation of the effect of the IP risk of new foreign R&D investments, hence Figure 2 shows the calculated proportional increase in Tobin's Q contingent on IP Risk of Foreign R&D. We observe an overall significantly negative effect of internal linkage strategy if there are no new R&D investments with IP risk (IP Risk of Foreign R&D = 0), which pertains to about half of the observations. A 5.9% decline (p = .048) in Tobin's Q is estimated in these circumstances. At the same time, we should bear in mind that internal linkage strategy if employed in knowledge rich environments has an additional positive influence on market value. In model 7, linkage for sourcing is highly significant, with the coefficient suggesting that a standard deviation increase is associated with a 8% increase in Tobin's Q.
4.1 Monthly panel results and supplementary analysis
We estimated the dynamic market valuation model also on monthly data, results of which are presented in Table 4. We observe a similar pattern for the focal variables, with a positive and significant coefficient on Foreign R&D (β = 0.138,p = .025), a negative and significant coefficient on IP Risk of Foreign R&D (β = − 0.464,p = .004) and a positive and significant coefficient on the interaction term (β = 7.161,p = .023). The coefficient on the interaction term suggests relatively strong mitigating effects of a Linkage Strategy on IP risk of Foreign R&D, and calculations confirm similar patterns as those described in Figure 1. As to the additional effects of differentiated internal Linkage Strategy, only a qualitatively similar pattern is observed in the monthly panel (Model 2 in Table 4) with significance levels for the negative main effect of Linkage Strategy—low risk and the positive interaction of Linkage Strategy—high risk with IP Risk of Foreign R&D reducing to below the 5% level (p-values of .070 and .084, respectively).
We conducted a number of supplementary analyses to examine the robustness of the findings reported in Tables 3 and 4, of which we relegate the key results to a separate appendix. We obtain qualitatively and quantitatively similar results in both yearly and monthly panels if (a) we omit the two external instruments from the GMM estimation; (b) we add new R&D entries due to M&As with a technology motivation; (c) we utilize (limited) information on the size of R&D investments abroad to weigh foreign R&D investments and IP risk; (d) we use an alternative measure of IP Risk of Foreign R&D (an average rather than a sum) that only has a limited correlation with foreign R&D investment; (e) we use an alternative linkage strategy indicator without weighting by IP risk; (f) we control for potential sample selection bias by estimating a Heckman's selection model; and (g) if we exclude the control variables measuring patent activity from the models. Finally, an alternative—more restrictive—estimation method (quasi-maximum likelihood estimation) and a less appropriate—biased—alternative (ordinary least squares) estimation method delivered qualitatively similar results.
4.2 Post-investment analysis: local internal linkages and local spillovers
Our analysis of the relationship between internal linkage strategy, R&D in weak IPR countries, and market valuation makes two important assumptions: (a) firms that have developed capabilities in their global R&D organization to prevent knowledge spillovers through an internal linkage strategy are better able to apply this strategy to their new R&D investments in comparison with firms that have not previously developed such capabilities; (b) Firms that are better able to apply an internal linkage strategy to new R&D investments in weak IPR locations are also able to reduce local knowledge spillovers. We examine the available evidence related to those 901 new R&D investment projects by the sample firms for which we can derive sufficient information on post-investment patent applications. We exclude investments between 2012–2014 to ensure that 5 years of subsequent patent applications are observed, and we require 10 or more local patents in a 5 year period to make reliable inferences.
Table 5 shows that firms with high or low internal linkage capabilities both exhibit higher internal linkages for their new R&D investments (in the 5 years after the investment) in high IP risk environments than in low IP risk environments (distinguished at the sample median). However, the average degree of internal linkage in high IP risk environments, as identified by patents invented at the new R&D locations, is substantially and significantly higher for the firms with prior high (above the sample median) internal linkage capabilities. Consistent patterns are also identified if one examines the citations by local firms to these locally invented patents of the MNCs in the second panel of Table 5. The knowledge spillover rate, measured by such citations (e.g., Alcácer & Zhao, 2012; Zhao & Islam, 2017), of high internal linkage capability firms is less than half the rate of low internal linkage capability firms, and this difference is significant.
| High ex-ante internal linkage | Low ex-ante internal linkage | Row gap (p-value) | ||
|---|---|---|---|---|
| Ex-post internal linkage | .125 | .078 | .000 | |
| Of which: | In weak IPR countries | .153 | .109 | .000 |
| In strong IPR countries | .098 | .053 | .000 | |
| Column gap (p-value) | .000 | .000 | ||
| China | .145 | .104 | .000 | |
| USA | .048 | .036 | .023 | |
| Column gap (p-value) | .000 | .000 | ||
| High ex-ante internal linkage | Low ex-ante internal linkage | Row gap (p-value) | ||
|---|---|---|---|---|
| Local ex-post R&D spillovers | .039 | .098 | .000 | |
| Of which: | In weak IPR countries | .012 | .029 | .015 |
| In strong IPR countries | .064 | .153 | .001 | |
| Column gap (p-value) | .004 | .000 | ||
| China | .018 | .041 | .012 | |
| USA | .331 | .445 | .309 | |
| Column gap (p-value) | .000 | .000 | ||
- Note: High or low internal linkage and weak or strong IPR categorizations are based on the median splits across the R&D investments in the sample. Both ex-ante and ex-post internal linkage scores are weighted by IP risk of recipient host countries. Calculations are restricted to 901 R&D investments, 2003–2011, for which at least 10 local patents in 5 years subsequent to the investment are observed. Local spillovers are the ratio of the number of citations by local firms to the focal firm's local patents in the R&D host country over the number of local patents. Local citations are measured up to 5 years from the investment year. The numbers for the largest recipient countries cover 218 investments in China and 106 investments in the USA.
We observe on average lower spillover intensities (but given the weak IP protection environment, arguably with more serious consequences) in weak IP environments, which will also be driven by the on average relatively smaller pool of local firms with relevant capabilities to absorb such knowledge. A more precise comparison can be made looking at investments in particular countries. Table 5 also shows the internal linkage strategy and local knowledge spillover rates specifically for R&D investments in China and the United States, respectively. China and USA are the top countries receiving R&D investments by the sample firms representing a prominent high IP risk and a prominent low IP risk environment, respectively. We observe clear patterns again in differences in internal linkage intensities between the two R&D investment environments, and between high and low internal linkage capability firms. Interestingly, spillovers rates are significantly different between the two groups of firms in China, but not in the United States. Overall, the patterns illustrated in Table 5 confirm the assumptions we made are consistent with the empirical results and theoretical arguments.
5 DISCUSSION AND CONCLUSION
We find that foreign R&D investments are associated with economically sizeable increases in market value, but that IP risks in R&D host countries with a weak IPR regime reduce market value if the MNE has not developed capabilities to embed such R&D in their global R&D operations through an internal linkage strategy. Our findings contribute to the literature on internal linkages and IP (e.g., Alcácer & Zhao, 2012; Belderbos & Somers, 2015; Nandkumar & Srikanth, 2016; Zhao, 2006; Zhao & Islam, 2017) by empirically investigating the interplay between R&D investments, internal linkage strategy, and weak IPR environments in influencing the expected future returns to foreign R&D—as indicated by the investing firm's market value. Our study is the first to demonstrate that such internal linkage strategies—intra-firm cross-country inventor collaboration—can have substantial financial performance effects, by reducing (investors' concerns regarding) misappropriation of a focal MNE's technologies in weak IPR environments. We also show that firms with experience in utilizing internal linkage strategies are indeed more capable of implementing internal linkage strategies in high risk R&D environments and are more effective in keeping knowledge outflows to local firms limited in such environments. On the other hand, we also find—qualified—evidence that an R&D organization relying on distributed R&D tasks and internal linkages across dispersed R&D units can increase costs and hamper efficiency of global R&D, with negative net market valuation effects if an internal linkage strategy is not applied predominantly in weak IPR environments.
Our study also contributes to an expanding stream of literature on the determinants and performance effects of R&D globalization (Athukorala & Kohpaiboon, 2010; Belderbos et al., 2013, 2015; Berry, 2014; Griffith et al., 2006; Kafouros et al., 2018; Kumar, 2001; Penner-Hahn & Shaver, 2005). While previous studies on the role of IPR protection in host countries have shown a negative (average) association between cross-border R&D investments and a lack of host country IPR protection (e.g., Belderbos et al., 2013; Branstetter et al., 2006; Smith, 2001), our study suggests that the effects of IPR protection on a MNE's preferences for R&D investment locations will be heterogeneous and depend on the IP protection capabilities of the MNE. Only firms with the capability to integrate and build on technologies developed in different R&D units through an R&D organization embedding internal linkages are able to benefit from conducting R&D and benefit from the available human capital and market opportunities in weak IP environments. Our study highlights that MNEs' “home bias” in R&D (the disproportional concentration of R&D in MNEs' home countries) may be mitigated if a suitable internal linkages strategy and R&D organization is developed—as indicated by a positive correlation between internal linkage strategy and foreign R&D investments in countries with high IP risks.
Our results show that internal linkage strategies to protect proprietary knowledge and technologies abroad are another moderator of the relationship between foreign R&D and firm performance. Prior studies on the (innovation) performance effects of foreign R&D concluded that performance is enhanced by intra-firm collaboration, and attributed this to its knowledge creation potential through knowledge integration and cross-fertilization (Frost & Zhou, 2005; Lahiri, 2010; Singh, 2008). In the context of a dynamic market valuation model, we confirmed a positive association between market value and the adoption of linkage strategies in countries with sample opportunities for knowledge sourcing. Our analysis suggests that the knowledge creation effects of internal linkage strategies can only be properly examined if the simultaneous influence on performance through knowledge protection is taken into account.
At the same time, we find first preliminary evidence of the costs of internal linkage strategies. There may be drawbacks of internal linkage strategies in terms of increased coordination costs, redundancy in knowledge linkages, and reduced autonomy and creativity in local R&D units (e.g., Alcácer & Zhao, 2012; Beugelsdijk & Jindra, 2018; Gambardella et al., 2020). Our study suggests that the intensive use of cross-unit collaboration in relatively strong IP environments is negatively associated with firm performance. This further attests to the salient role that internal linkages strategies of MNEs play—primarily in addressing knowledge spillovers and misappropriation risks in weak IPR environments.
Finally, our study contributes to the literature on firms' knowledge assets, appropriation strategies, and market valuation (e.g., Castellaneta et al., 2017; Czarnitzki et al., 2020; Hall et al., 2005, 2007) by distinguishing the influence of foreign R&D investments from overall R&D expenditures, and by considering that environmental influences that increase the risks of R&D (Bromiley, Rau, & Zhang, 2017) and reduce market value, can be counteracted by appropriate strategies to prevent outward knowledge spillovers.
The managerial implications of our study are relatively straightforward. Foreign R&D in risky IP environments can increase performance if firms align their R&D organization to these environments by allowing for strong control and interdependence across R&D units. Internal linkages may also aid knowledge sourcing and knowledge integration within the MNE, but advantages have to be weighed against the generally higher costs of operating a collaborative R&D network. Our findings suggest that if internal linkage strategies are used in high IP risk environments, they create value and mitigate the expected hazards of investing in high IP risk countries. Yet internal linkage strategies used in relatively low IP risk countries do not serve such purpose and reduce market value, due to their higher costs. The patterns we observed for internal linkages in newly established R&D units showed that firms that have developed an R&D organization embedding an internal linkage strategy tend to apply this in new R&D activities in weak as well as strong IP environments.
An interesting implication for management and research on global innovation and R&D management (e.g., Lahiri, 2010; Belderbos et al., 2015; Kafouros et al., 2018; Von Zedtwitz & Gassmann, 2002) is that firms face a challenge to differentiate R&D practice and organization across countries, while still maintaining a global R&D organization focused on global knowledge integration and effective appropriation. This challenge increases the complex tradeoffs in delegation and autonomy decisions revolving about local resources, motivation, and incentives (Gambardella et al., 2020). Further disentangling the positive consequences of internal linkage strategies in terms of IP protection and knowledge creation, investigating the circumstances under which these strategies may increase the costs of global R&D, and considering other means to control knowledge outflows abroad such as the use of expatriates (Berry, 2017) are topics worthy of attention in future studies.
Our research has several limitations, of which we discuss the most salient ones. First, our sample, while displaying heterogeneity in foreign R&D investment patterns and the adoption of internal linkages strategies, focused on leading MNEs with headquarters in Europe, the US, and Japan. It is clear that results are difficult to generalize to smaller MNEs with fewer opportunities to engage in internal linkage strategies. MNEs based in emerging economies with fewer knowledge assets may be more eager to benefit from knowledge spillovers in weak IPR environments than worried about protecting their own technologies (cf. Alcácer & Chung, 2007; Shaver & Flyer, 2000).
Second, although a more precise identification of the effects of foreign R&D on market value may be obtained by examining stock price reactions at the time decisions are publicly announced, we could not employ such an event analysis as our data source on R&D does not include specific dates of the investments. Instead we had to rely on new information reaching the market in a given year or month. Although the analysis of monthly data in a more granular approach confirmed findings in the yearly panel, future research can improve identification if specific dates of investment announcements can be investigated. Third, our analysis focused on patent protection and enforcement, while the protection of trade secrets and enforcement of non-compete clauses are also likely to play an important role (e.g., Castellaneta et al., 2017; Garmaise, 2009).
Fourth, while the creation of internal linkage capabilities presents itself in our analysis as a tradeoff between its costs on the one hand, and protection against knowledge outflows in weak IP environments as well as potential knowledge sourcing and integration benefits in knowledge rich environments on the other hand, our analysis did not allow to delineate this complex tradeoff precisely. Such a tradeoff is likely to exhibit heterogeneity across host countries and could in addition also depend on the market potential of the country for the investing firm. A limitation of our study was that we could not include information on industry-specific market opportunities. Future studies could investigate the role of local market opportunities in greater detail.
Finally, IP risks in countries with weak IP protection will be most salient if the R&D investment is collocated with local rival firms in the same sub-national cluster (Alcácer & Zhao, 2012), but we only differentiated foreign R&D at the country level and moreover did not specifically identify rivals and their related absorptive capacity creating misappropriation risks. Identifying (the location of) potential IP infringers is beyond the scope of this research but is likely to provide more accurate estimates of IP risks and their consequences for firm performance. An assessment of spillover and misappropriation risks at the cluster level in the context of R&D investments should also take into account that even if a MNE chooses a location outside the cluster, rivals may (re)locate to this location in the future. These issues offer fruitful avenues for further research and exploration.
ACKNOWLEDGMENTS
The authors wish to thank the editor Minyuan Zhao, two anonymous reviewers, Pierre Mohnen, John Hagedoorn, Katrin Hussinger, Keun Lee, Boris Lokshin, and participants at the ISS conference in Seoul, the Strategy Science Conference Doctoral Workshop in Philadelphia, and seminars at UNU-MERIT, KU Leuven and NEOMA BS for comments on earlier drafts. The authors acknowledge financial support from the Research Foundation Flanders (FWO) under grant No. G075520N. Jinhyuck (Joseph) Park acknowledges support from UNU-MERIT.
Open Research
DATA AVAILABILITY STATEMENT
Measures constructed by the authors will be made available upon request except for data with commercial restrictions set by the data providers.
