Foreign knowledge sources and innovation: Differences across large and small and medium-size multinational enterprises (MNEs)

International Review of Economics and Finance 92 (2024) 741–757 Available online 18 February 2024 1059-0560/© 2024 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Foreign knowledge sources and innovation: Differences across large and small and medium-size multinational enterprises (MNEs) María Jesús Rodríguez-Gulías, Sara Fern´ andez-L´ opez, David Rodeiro-Pazos * Department of Finance and Accounting, Universidade de Santiago de Compostela, Avda. do Burgo, s/n., 15782 Santiago de Compostela, Galicia, Spain ARTICLE INFO Keywords: Cross-border knowledge sourcing Internationalisation spillovers Foreign knowledge sources R&D-based innovation Multilevel analysis Multinational enterprises ABSTRACT Multinational enterprises (MNEs) can take advantage of cross-border knowledge sourcing and internationalisation spillovers. Nevertheless, the influence of these two foreign knowledge sources on innovation has been scarcely analysed in the MNEs. The purpose of this paper is three-fold. First, to apply multilevel modelling in the study of MNEs’ innovation. Second, to analyse how the region’s foreign knowledge sources affect innovation based on R&D in MNEs. Third, to explore whether this influence differs between large and small and medium-size MNEs (SM-MNEs). We used multilevel panel data analysis in a sample of 1471 Spanish MNEs over the period 2009 to 2016. The results indicate that the firm-level factors have a stronger relationship with the propensity to innovate in MNEs than the region’s foreign knowledge sources. Nevertheless, the effect of these region’s knowledge sources in explaining the differences across MNEs in innovation is more substantive in large MNEs than in SM-MNEs. 1. Introduction There is an extensive theoretical and empirical literature on the drivers of firm innovation. These innovation drivers have been grouped into external and internal factors by the literature on the geography of innovation (L´ opez-Bazo & Motell´ on, 2018; Sternberg & Arndt, 2001). The former are usually linked to a set of location-bound advantages mainly stemming from the external knowledge sources (Raspe & Van Oort, 2009), while the later are mostly associated with the firm’s ability to use and exploit these external resources to innovate (Acs et al., 2009; Jantunen et al., 2008; Srholec, 2010). Despite the vast research on firm innovation, it was only in the last decade that both types of factors began to be analysed jointly by considering the hierarchical structure of the data, which implies to acknowledge the existence of heterogeneous firms within a given location (Rodríguez-Gulías et al., 2020). Compared to firms which sell on the national market only, internationally-active firms have developed knowledge-based capabilities that help them to compete in international markets. In a recent study, Alguacil et al. (2023) obtained that, for some regions and sectors, European firms investing in third markets are more intensive in R&D and innovation activities. Additionally, they are exposed to a wider range of external knowledge sources than domestic firms (Criscuolo et al., 2010; Wagner, 2006). In other words, from the geography of innovation perspective, the multinational enterprises (MNEs) are in better conditions to * Corresponding author. E-mail addresses: [email protected] (M.J. Rodríguez-Gulías), [email protected] (S. Fern´ andez-L´ opez), david. [email protected] (D. Rodeiro-Pazos). Contents lists available at ScienceDirect International Review of Economics and Finance journal homepage: www.elsevier.com/locate/iref https://doi.org/10.1016/j.iref.2024.02.036 Received 25 March 2021; Received in revised form 11 December 2022; Accepted 13 February 2024 International Review of Economics and Finance 92 (2024) 741–757 742 internally exploit a broader range of external knowledge resources than firms only selling in domestic markets (Du & Temouri, 2011; Fleury, Fleury, & Borini, 2013). Furthermore, the MNEs ranked top in the total factor productivity pecking order likely dominated in innovation or management skills (Tang, 2017). However, to the best of our knowledge, there is no studies exploring how external knowledge and internal resources jointly affect the MNEs’ innovation by taking into account the hierarchical structure of the both level of analysis (first research gap). In addition, within the external resources available to innovate, the literature on innovation has overlooked the effects of foreign knowledge sources on firm innovation. Good examples of foreign knowledge sources are cross-border knowledge sourcing and export spillovers. The former is important to MNEs (Choung & Hwang, 2000), however it has hardly been studied empirically. Similarly, although the literature on international business (IB) acknowledges the role of export spillovers on firm internationalisation (Fletcher & Harris, 2012; García-Cabrera et al., 2017; Requena-Silvente & Castillo-Gim´ enez, 2007), it has ignored their effects on the MNEs’ innovation (second research gap). Both types of foreign knowledge sources can influence the innovative outputs of firms, especially in the case of MNEs as they often interact with other internationally-active firms in a certain location. Also, in the IB literature, Zaheer and Nachum (2011) point out that not all MNEs are equally good at making the most of the local environment, which concurs with the claim of the literature of geography of innovation to put the emphasis on the individual firm (Wang & Ling, 2013). Particularly, a relevant firm characteristic whose relationship with innovation has been subject of debate is firm size. In this respect, Garg and Zhao (2018) also claim for further investigation on foreign knowledge sources referred to small MNEs (third research gap). This paper aims to fill the abovementioned three research gaps in the literature on the MNEs’ innovation. More specifically, we explore whether the cross-border knowledge sourcing and export spillovers, together with other location-bound advantages have an effect on the MNEs’ innovation, and whether such effects depend on the firm size. To this end, we apply multilevel panel data analysis in a sample of 1471 Spanish internationally-active manufacturing firms over the period 2009 to 2016. In light of the pressing nature of the topic, this paper makes the following contributions. First, the use of the multilevel models with cross-interactions allows detaching the influence of the macro-level external knowledge sources and micro-level internal resources on the MNEs’ innovation. In this respect, this paper extends the recent stream of the literature of the geography of innovation that calls for the accommodation of the hierarchical structure of the data through multilevel models when studying firm innovation (Autio et al., 2014; Beugelsdijk, 2007; L´ opez-Bazo & Motell´ on, 2018; Rodríguez-Gulías et al., 2016). Second, this paper explores two new types of foreign knowledge sources available at a regional level. In so doing, new evidence on how “explicit” foreign knowledge (i.e. cross-border knowledge sourcing) and “tacit” foreign knowledge (i.e. export spillovers) impact on MNEs’ innovation is provided. Moreover, to this end a rich set of firm and locational characteristics is considered in order to reduce the perils of confounding the effect of the region’s foreign knowledge sources with that of the omitted variables (L´ opez-Bazo & Motell´ on, 2018). Third, this paper responds to the claim of gaining insights into the differences in foreign knowledge sourcing across large and small and medium-size MNEs (SM-MNEs) (Garg & Zhao, 2018). Finally, the Spanish case during the sample period (2009–2016) offers an interesting environment for such an analysis. Until 2014, Spain was affected by a severe economic downturn. To face the crisis, the government encourages firms to increase exports, which are highly concentrated in a few large companies (Dur´ an, 2014), while the small ones are less productive than their European counterparts (Mora Sanguinetti & Fuentes, 2012). These data reinforce the idea that Spanish SM-MNEs remain weak compared to their European counterparts. Similarly, Spanish firms show low levels of innovative activities. In 2014, 36.4% of the Spanish firms innovated, compared to 49.1% of the European ones (EUROSTAT, 2018). In 2016, the R&D financial effort of the Spanish firms was half the European average effort (COTEC, 2017). In such resource-constrained environment, gaining insights into how the region’s foreign knowledge sources affect the innovation of MNEs will allow for better design policies aimed at supporting them. The remainder of the paper is organised as follows: Section 2 introduces the positioning of the paper. Section 3 describes the data and the specification of the models used in the empirical analysis. Section 4 discusses the econometric results, and, finally, Section 5 concludes and provides the implications for literature and practice. 2. Foreign knowledge sources, firm size and MNES’ innovation: a firm-centered approach 2.1. The need of multilevel analysis of the MNEs’ innovation The geography of innovation literature agrees that firm innovation is determined by internal and external factors (Jaffe et al., 1993). From this approach, external resources are available to all firms in a particular location and the most efficient firms at exploiting them would obtain competitive advantages. The literature on knowledge management also insists on this idea by highlighting that effective knowledge management allows combining internal and external sources of knowledge enhancing firm innovation (Andersson et al., 2016). Accordingly, a burgeoning number of studies have examined the effects of both sources of resources on firm innovation (Vega-Jurado et al., 2008). The balance of findings indicates that locational characteristics matter for innovative performance, but the firm heterogeneity plays the major role (L´ opez-Bazo & Motell´ on, 2018; D’Agostino & Moreno, 2019). Compared to domestic enterprises, internationally-active firms are exposed to a wider range of external knowledge sources (Criscuolo et al., 2010; Wagner, 2006). MNEs interact with customers, suppliers, financial institutions or research centers (Perri et al., 2017) other than domestic ones, which serve as sources of tacit knowledge (Forero-Pineda et al., 2010). In the same line of reasoning, the learning-by-exporting hypothesis establishes that exporting enables firms’ domestic innovation after entry; that is, the competence with their international rivals (Bellmann et al., 2013) leads exporters to attain innovative capabilities to exploit new knowledge (Du & Temouri, 2011; Fleury et al., 2013). In this respect, several studies show that firms that are active on international markets learn more M.J. Rodríguez-Gulías et al. International Review of Economics and Finance 92 (2024) 741–757 743 from external resources and generate more innovative outputs (Criscuolo et al., 2010; Wagner, 2006). To sum up, from the geography of innovation perspective, MNEs are not only exposed to a wider set of external knowledge sources than domestic firms, but they also have developed ‘incremental’ innovative capabilities to exploit them. Accordingly, an extensive stream of the literature on IB has explored MNEs’ innovation and knowledge management, mainly at a micro-level of analysis (see Andersson et al. (2016) for a recent survey). However, there exists scarce literature combining both internal resources (at a firm-level) and external knowledge spillovers (at a macro-level) as determinants of the innovative outputs of the MNEs (Andersson et al., 2016). More specifically, even when literature on the geography of innovation has embraced the view that local factors are generic resources available to all firms in a certain location, and the literature on IB indicates that not all MNEs are equally good at exploiting the location-bound advantages (Zaheer & Nachum, 2011), the question of how external knowledge and internal resources jointly affect the MNEs’ innovation by taking into account the hierarchical structure of the both levels of analysis remains opened. Drawing on previous streams of the literature, this paper aims to fill this first (methodological) research gap by applying multilevel models. 2.2. The effects of the region’s cross-border knowledge sourcing and internationalisation spillovers on MNEs’ innovation Regarding the external resources available to innovate, the emphasis has been placed on the location-bound advantages stemming from knowledge (Raspe & Van Oort, 2009). In this field, the most intensively researched locational characteristics have been those referred to technological resources and production spillovers available in a certain location, usually in a region (Feldman & Audretsch, 1999). The former have been traditionally associated with the R&D inputs, such as R&D infrastructure (Beugelsdijk, 2007), whereas the latter have been linked to innovative outputs, such as the stock of patents. The underlying argument is that the region’s R&D inputs and outputs can, in turn, work as an input for innovative activities of the firms in that region (Corsi & Prencipe, 2016; Czarnitzki & Hottenrott, 2009; Powers & McDougall, 2005; Raspe & Van Oort, 2009). This reasoning could also be applied to the two foreign knowledge sources explored in this paper: cross-border knowledge sourcing and internationalisation spillovers. As mentioned, a branch of IB literature has studied the links between innovation and knowledge management at a micro-level (Andersson et al., 2016). This research generally acknowledges that MNEs can tap into diverse pockets of knowledge around the world (Criscuolo et al., 2010; Perri et al., 2017; Wagner, 2006), enhancing their competitive advantages (Andersson et al., 2016). More specifically, within the four knowledge sub-processes distinguished by Andersson et al. (2016) for MNEs, sourcing is the process by which managers identify and gain access to knowledge created outside the boundaries of the firm (Eisenhardt & Santos, 2001). When this external sourcing of knowledge is combined with the internal innovative capabilities, it contributes to the creation of new competences for the MNEs (Andersson et al., 2016). Thus, Choung and Hwang (2000) find that MNEs from the emerging markets can build their innovative capabilities and internal resources by learning from their international rivals through approaches such as licensing. Stemming from previous arguments, we argue that MNEs can benefit from the cross-border knowledge sourcing available in a certain region. Particularly, in this paper cross-border knowledge sourcing is measured as the sum of payments for foreign licenses or technical assistance by firms in the region as a percentage of regional Gross Domestic Product (GDP). The acquisition of foreign knowledge may provide actors with the novelty and variety that are needed to enrich local innovation processes (Perri et al., 2017). The higher the volume of payments for foreign licenses or technical assistant, the higher the number of the firms exploiting foreign knowledge. Thus, the interactions with these geographically proximate firms can act as a catalyst for the innovative activities of the MNEs. Based on previous arguments, we propose the following hypothesis. H1. The innovation of MNEs is positively influenced by the region’s cross-border knowledge sourcing. The IB literature agrees that external factors as export spillovers can help firms to internationalize (Fletcher & Harris, 2012; García-Cabrera et al., 2017). From this approach, when local firms are exposed to other internationally-active firms, they acquire information about foreign markets due to a spillover effect (Requena-Silvente & Castillo-Gim´ enez, 2007). This effect is more relevant for those firms lacking internal export knowledge or experience (Acs et al., 1994). In this paper we argue that the export spillovers can also affect innovative activities. More specifically, interactions with other internationally-active firms located in the region can work as a channel to acquire information about new products and processes from foreign markets. According to De Clercq et al. (2008), exports in the region could be used for measuring the sources of export spillovers. Accordingly, in this study we use the natural logarithm of the region’s exports to proxy export spillovers. Although the success of the firm to acquire and exploit external knowledge through innovation will depend on its innovative capabilities (Jantunen et al., 2008), firms located in territories with a high presence of internationally-active actors will be more exposed to export spillovers compared to firms in territories with a low presence of exporters, giving rise to an export spillover effect on firms’ innovation. This reasoning, which could apply to all firms in a given location, is especially relevant for MNEs that interact daily with other internationally-active firms for ‘international reasons’ rather than for ‘innovative reasons’. Stemming from previous arguments, the next hypothesis is proposed. H2. The innovation of MNEs is positively influenced by the region’s export spillovers. In short, we argue that, together with the technological and production spillovers in a certain region, MNEs can benefit from the cross-border knowledge sourcing and export spillovers. By exploring these new types of foreign knowledge sources, the paper aims to fill the aforementioned second research gap. In so doing, we provide new evidence on how “explicit” foreign knowledge (cross-border knowledge sourcing) and “tacit” foreign knowledge (export spillovers) affect MNEs’ innovation. To the best of our knowledge, no study has addressed these issues. M.J. Rodríguez-Gulías et al. International Review of Economics and Finance 92 (2024) 741–757 744 2.3. The role of firm size in MNEs’ innovation In the IB literature, Zaheer and Nachum (2011) highlight that MNEs differ in their ability to exploit the location-bound advantages. Similarly, the geography of innovation literature insists on bringing the firm heterogeneity to the centre of the debate to understand technological innovation from a multilevel approach (Wang & Lin, 2013). In this respect, the firm size has emerged as an important firm characteristic to explain differences in exploiting external knowledge spillovers. Indeed, the need of further research on foreign knowledge sourcing in small MNEs (i.e. the third research gap) has been recognized by Garg and Zhao (2018). Different authors have been analized the relationship between firm characteristics as size and the market scope of firms’ internationalisation decisions (Alguacil et al., 2017; Helpman et al., 2004; Yeaple, 2009), but the relationship between firm size and innovation still remains a subject of debate. Overall, the authors concur with the argument behind this potential relationship; that is, the firms that rely more on external knowledge sources develop their internal innovative capabilities to a greater extent, pushing them to innovate. Nevertheless, the authors differ in the type of firms that most often gain access to external knowledge sources. Drawing on the resource-based view of the firm, some studies argue that the fewer internal resources of small firms lead them to rely more on external knowledge sources than larger firms. Conversely, other studies state that large firms may be at advantage in comparison with their smaller counterparts because their higher internal resources give them access to external knowledge sourcing (Fontana et al., 2006; Laursen & Salter, 2004). For example, the professional profile of the employees in large firms usually includes fields of study such as science and engineering (Laursen & Salter, 2004), which encourage them to engage in R&D activities with external actors (Barbosa & Eiriz, 2009; Chen et al., 2012). To sum up, this paper aims to tackle the abovementioned issues: (1) whether regional knowledge spillovers and internal resources jointly affect the MNEs’ innovation by taking into account the hierarchical structure of the levels of analysis (region or macro-level and firm or micro-level); (2) whether the new proposed types of foreign knowledge sources, namely cross-border knowledge sourcing and internationalisation spillovers, shape the effect of MNEs’ internal resources associated with innovative activities, and (3), whether such effects depend on the MNE’s size. In so doing, this paper fits with the recent branch of the literature on geography of innovation that claims the need of jointly exploring the influence of external and internal drivers of firms’ innovation and the literature on IB that claims the need of analysis cross-border knowledge sourcing in MNEs in general, and in small MNEs in particular. 3. Methodology 3.1. The data and sample At the firm level, the micro-data were drawn from the ESEE (Encuesta sobre Estrategias Empresariales, or Survey on Business Strategies), a panel of Spanish manufacturing firms. Data are annually collected by the Ministry of Industry and the SEPI Foundation. ESEE database covers firms with 10 or more employees; while all firms with more than 200 employees are included, a representative sample for small companies employing between 10 and 200 employees is annually surveyed. ESEE contains information on a wide range of topics, including innovative activities (R&D activities, patents or innovation) and internationalisation (exports and imports). From the initial sample consisted of 3004 firms over the period 2009–2016 some adjustments were made. Firms were eliminated from the sample for the following reasons: (1) the firm’s location information was unclear, (2) the firm presented two or fewer observations, which does not fulfil the requirements of the multilevel analysis (Snijders & Bosker, 1999), and (3) the firm had not exported in any of the year of the study period. To the purpose of this work, firms with exports sales in at least one year of observation were considered internationally-active firms or multinational enterprises (MNEs). Thereby, the final sample used in this study is formed by 1471 MNEs observed during the period 2009–2016. At the regional level, data for Spanish regions were compiled from two sources: the database of the Spanish National Statistics Institute (INE) and the database of the Spanish Patent and Trademark Office (OEPM), both official and public databases. Fig. 1. Firms with R&D-based innovation by year: number and percentage. M.J. Rodríguez-Gulías et al. International Review of Economics and Finance 92 (2024) 741–757 745 3.2. Definition and measurements of the variables The Oslo Manual (OECD, 2018) defines a business innovation as “a new or improved product or business process (or combination thereof) that differs significantly from the firm’s previous products or business processes and that has been introduced on the market or brought into use by the firm” and the research and experimental development (R&D) as “a creative and systematic work undertaken in order to increase the stock of knowledge and to devise new applications of available knowledge” (OECD, 2018). Following these definitions, and similarly to Spescha and Woerter (2019), it was constructed a dependent variable (INNO_RD) that takes the value 1 if the firm had product or process innovation in the year and developed R&D activities in the previous year, and 0 otherwise. The purpose of the dependent variable is to measure innovations emerging from R&D activities; therefore, R&D activities in the year preceding the innovation are considered, as it takes time for R&D activities to be converted into innovation results (Hall et al., 2010). Close to 44% of the sample firms presented innovation based on R&D (INNO_RD) in at least one year of de study (2009–2016). By year, the percentage of firms with R&D based innovation fluctuated between 31% and 41% (Fig. 1). By regions, the average percentage of firms with R&D-based innovation is drastically different. Hence, there were regions with average values higher than 40% (Cantabria) or 30% (Basque Country, Catalonia or Aragon) or even 0% (Balearic Island) (Fig. 2). In order to test the proposed hypotheses, two key independent variables were defined at a regional level (Table 1). Thus, the crossborder knowledge sourcing was measured as the sum of payments for foreign licenses or technical assistance by firms in the region as a percentage of regional Gross Domestic Product (GDP) (IMPTECH_GDP). Following De Clerq et al. (2008), the export spillovers were proxied by the natural logarithm of the region’s annual exports in an attempt to reflect the international orientation of each region (LNREGEXPORT). Two more variables were considered at a regional level. To collect the region’s technological resources, the percentage of Gross Domestic Product (GDP) dedicated to R&D (RD_GDP) was used. Also, the natural logarithm of the number of patent applications per 100 thousand inhabitants (LNPAT_100 TH) was included as proxy of the region production spillovers. At a firm level, 8 explanatory variables were included as potential drivers of MNEs’ innovation. In order to reflect the use of crossborder knowledge sourcing inside the firm, a dummy variable that takes the value 1 for companies that pay for foreign licenses or technical assistance in the current year (IMPTECH) was considered. As proxies of the firm R&D intensity two variables were added measuring the internal R&D effort (INTRD_SALES) and the external R&D effort (EXTRD_SALES), calculated as the percentage of sales dedicated to internal R&D and external R&D, respectively. A dummy variable that takes the value 1 if the firm had registered at least one patent or utility model in the current or in the previous years (CUMPAT) was also considered. Firm size was measured by the natural logarithm of the number of employees (LNEMP). Similarly, in an attempt to capture the internationalisation level of the sample firms, the natural logarithm of the sales from export (LNEXPORT) was included. The remaining two variables are the natural logarithm of the firm age (LNAGE) and a dummy that take the value 1 for firms in high-medium technology sectors according to the Eurostat classification 1 (HIGHTECH). The descriptive statistics of the independent variables are displayed in Table 2. Concerning the regional level, in average terms, 0.05% and 1.07% of GDP are dedicated to acquiring cross-border knowledge and to conduct R&D activities, respectively. The average number of patent applications is 6.62 per 100 thousand inhabitants and the average value of the regional exports is 13,000 million euros. Figures A1 to A.4 of Annex A report a deeper description of the regional variables under analysis showing their annual values in each region over the period 2009–2016. At the firm level, 6.84% of firms pay for foreign licenses or technical assistance, which involves the use of cross-border knowledge sourcing. The average percentage of sales dedicated to internal and external R&D expenditures are about 0.70% and 0.30%, respectively. Near 17% of firms report cumulative patent activity and the average annual export sales are about 41 million euros. The average number of employees is close to 229, the average age of firms is near 35 years, and 36.35% of them operate in high-medium technology sector. Finally, the correlation matrix of the independent continuous variables is shown in Table 3. In view of the correlation matrix data, possible multicollinearity problems were noted and then the variance inflation factor (VIF) was calculated and showed in Annex B. There were no VIF values higher than 3 and the average VIF was 1.63, which could be considered an acceptable level (Greene, 2012). 3.3. Strategy of estimation and model specification Given that the dataset involves observations (level 1) for MNEs (level 2) nested in regions (level 3) a three-level multilevel model was used to analyse the effect of external (regional) and internal factors on the probability of having R&D-based innovation. Furthermore, since the dependent variable (INNO_RD) is a dummy variable, a logistic model was assumed. In particular, we estimated three-level logistic random intercept models, which can be written as a latent response, and assumed fixed slopes. The Model 0 was the empty or null model, which only included the firms’ random intercept and without explanatory variables. 3.3.1. Model 0 (null model) INNO_RD* ijk =β 0 +v k +u j +e* ijk 1 Eurostat uses the aggregation of the manufacturing industry according to technological intensity and based on the NACE Rev.2 at the two-digit level. M.J. Rodríguez-Gulías et al. International Review of Economics and Finance 92 (2024) 741–757 746 Where INNO_RD* ijk is the observed R&D-based innovation for occasion i in the company j in the region k, β 0 is the overall mean of INNO_RD* ijk (across all groups), v k is the level 3 random effect (regional random effect), u j is the level 2 random effect (firm random effect), and e* ijk is the level 1 residual (occasion residual) with mean zero and variance σ 2 e* (assuming a logistic distribution of σ 2 2 ) Fig. 2. Average percentage of firms with R&D-based innovation by region. Table 1 Definitions of dependent and independent variables. Group Factor Variable Measure Source Dependent Innovation based on R&D INNO_RD 1 for the firms with innovation in the current year and R&D activities in the lagged year, and 0 otherwise. ESEE Regional level factors Foreign knowledge sources Cross-border knowledge sourcing IMPTECH_GDP Firms’ expenditures for foreign licenses or technical assistance/GDP (%) ESEE & INE Internationalisation spillovers LNREGEXPORT Natural logarithm of the region’s annual exports. INE Domestic knowledge sources Technological resources RD_GDP R&D expenditures/GDP (%) INE Production spillovers LNPAT_100 TH Natural logarithm of the number of patent applications per 100 thousand inhabitants. OEPM Firm level factors Innovative activities Internal use of cross-border knowledge sourcing IMPTECH 1 for companies that payed for foreign licenses or technical assistance in the current year, and 0 otherwise. ESEE R&D intensity INTRD_SALES Internal R&D expenditures/Sales (%) ESEE EXTRD_SALES External R&D expenditures/Sales (%) ESEE Patent activity CUMPAT 1 for the firms that had registered at least one patent or utility model in the current or in the previous years. ESEE Internationalisation LNEXPORT Natural logarithm of the annual firm exports. ESEE Size LNEMP Natural logarithm of the number of employees. ESEE Age LNAGE Natural logarithm of the firm’s age. ESEE Industry HIGHTECH 1 for the firms in high-medium technology sectors, and 0 otherwise. ESEE 2 Since the standard logistic distribution has a variance of π 2 /3 =3.29 this can be taken as the occasion or level 1 variance (Snijders & Bosker, 1999). M.J. Rodríguez-Gulías et al. International Review of Economics and Finance 92 (2024) 741–757 747 (Bellmann et al., 2013; Leckie, 2013; Steele, 2009). The Model 1 included the firm’s characteristics related to their internal innovative abilities. 3 3.3.2. Model 1 (with internal factors) INNO_RD* ijk =β 0 +β 1 IMPTECH ijk +β 2 L.INTRD_SALES ijk +β 3 L.EXTRD_SALES ijk +β 4 L.CUMPAT ijk +β 5 LNEXPORT ijk +β 6 LNEMP ijk +β 7 LNAGE jk +β 8 HIGHTECH jk +v k +u j +e* ijk The Models 2 incorporated the region’s knowledge sources. To avoid potential multicollinearity problems, the regional variables were added one by one resulting in four alternative specifications (Model 2.1 to 2.4). 3.3.3. Model 2.1 to 2.4 (with internal and regional factors) INNO_RD* ijk =β 0 +β 1 IMPTECH ijk +β 2 L.INTRD_SALES ijk +β 3 L.EXTRD_SALES ijk +β 4 L.CUMPAT ijk +β 5 LNEXPORT ijk +β 6 LNEMP ijk +β 7 LNAGE jk +β 8 HIGHTECH ijk +β 9 REGION ik +v k +u j +e* ijk Where REGION ik refers to the 4 regional level variables (IMPTECH_GDP ik , LNREGEXPORT ik, RD_GDP ik and LNPAT_100TH ik ). In order to investigate the differences in the determinants of R&D-based innovation in MNEs with different size, the sample was divided into two groups (large firm subsample and small and medium-size enterprise (SM-MNEs) subsample) and the models were reestimated. The large firm subsample comprises the MNEs with 200 or more employees, while the SM-MNEs subsample contains those with less than 200 employees. In order to know the relative magnitude of the variance components, intraclass correlation coefficients (ICCs) are reported in all estimated models for both, regional and firm level. The regional level ICC is calculated as the correlation between two observations i and i´within the same region k, but different firm j and j´: Table 2 Descriptive statistics of the independent variables. Variable Obs. Mean Std. Dev. Min. Max. INTERNAL FACTORS IMPTECH 9742 0.0684 0.2524 0 1 INTRD_SALES a 9693 0.6964 2.1419 0.000 43.792 EXTRD_SALES a 9725 0.2969 1.1964 0.000 33.957 CUMPAT 9742 0.1723 0.3777 0 1 EXPORT b 9731 41,000,000 259,000,000 0 6,740,000,000 EMP b 9742 228.7016 750.4114 1 13,091 AGE b 9742 35.1621 20.6070 1 177 HIGHTECH 9742 0.3635 0.4810 0 1 EXTERNAL FACTORS IMPTECH_GDP a 136 0.0543 0.1182 0.000 0.679 REGEXPORT b 136 13,000,000,000 14,100,000,000 834,000,000 65,100,000,000 RD_GDP a 136 1.0738 0.4790 0.320 2.230 PAT_100TH b 136 6.6188 3.7921 1.461 19.528 Notes: a Variables are in percentage terms; b Variables are not in logs. Table 3 Correlation matrix. (1) (2) (3) (4) (5) (6) (7) (8) (9) (1) INTRD_SALES 1 (2) EXTRD_SALES 0.3208* 1 (3) LNEXPORT 0.1083* 0.1082* 1 (4) LNEMP 0.1144* 0.1283* 0.7436* 1 (5) LNAGE 0.0161 0.0307* 0.1699* 0.2023* 1 (6) IMPTECH_GDP 0.0220* 0.0122 −0.0053 0.0232* −0.0728* 1 (7) RD_GDP 0.1025* 0.0689* 0.1203* 0.0781* 0.0790* −0.1727* 1 (8) LNPAT_100 TH 0.0818* 0.0337* 0.0510* 0.0172 0.018 0.1834* 0.6396* 1 (9) LNREGEXPORT 0.0366* 0.0282* 0.0611* 0.0071 0.1223* −0.0808* 0.5005* 0.3388* 1 Notes: Table shows the Pearson correlation coefficients for the continuous variables considered in the empirical analysis. *, **, *** denote significance at the 5%, 1%, and 0.1% levels. 3 Note that variables referred to R&D intensity and patent activity are delayed one year to be consistent with the definition of the dependent variable, which takes the value 1 if the firm has innovated in products or processes in the year and had carried out R&D activities in the previous year. M.J. Rodríguez-Gulías et al. International Review of Economics and Finance 92 (2024) 741–757 748 ρ v=corr(inno rdijk ,inno rd´ i ´ jk)= α 2 v α 2 v+ α 2 u+ α 2 e∗ The firm level ICC is calculated as the correlation between two observations i and i´within the same firm j and therefore the same region k: ρ vu =corr(inno rdijk,inno rd´ ijk)= α 2 v+ α 2 u α 2 v+ α 2 u+ α 2 e∗ 4. Empirical results 4.1. The effect of the region’s foreign knowledge sources: a multilevel panel data analysis Table 4 shows the estimates of the three-level logistic random intercept models for the effect of regional and firm level variables on R&D-based innovation (INNO_RD) of the global sample. As mentioned, the Model 0 is the null model, whereas the Model 1 includes the firms’ internal factors and the Models 2 add regional factors one by one. Table 4 Effect of regional and firm level variables on R&D-based innovation: global sample. Model 0 Model 1 Model 2.1 Model 2.2 Model 2.3 Model 2.4 CONSTANT −3.969*** −11.551*** −11.760*** −10.632*** −12.460*** −12.529*** (0.232) (0.814) (0.828) (2.473) (0.865) −0.931 FIXED EFFECTS Internal factors IMPTECH 0.241 0.218 0.233 0.234 0.219 (0.244) (0.245) (0.243) (0.244) (0.244) L.INTRD_SALES 0.682*** 0.681*** 0.682*** 0.675*** 0.676*** (0.050) (0.050) (0.050) (0.050) (0.050) L.EXTRD_SALES 0.403*** 0.402*** 0.406*** 0.400*** 0.402*** (0.065) (0.065) (0.065) (0.065) (0.065) L.CUMPAT 0.048 0.069 0.051 0.090 0.070 (0.199) (0.200) (0.199) (0.200) (0.199) LNEXPORT 0.120** 0.117* 0.128** 0.120** 0.124** (0.046) (0.047) (0.046) (0.046) (0.046) LNEMP 1.371*** 1.381*** 1.361*** 1.376*** 1.370*** (0.117) (0.118) (0.117) (0.117) (0.117) LNAGE 0.173 0.200 0.205 0.192 0.225 (0.169) (0.171) (0.170) (0.169) (0.169) HIGHTECH 0.182 0.189 0.197 0.123 0.177 (0.198) (0.199) (0.198) (0.199) (0.198) External factors IMPTECH_GDP 1.532* (0.711) LNREGEXPORT −0.045 (0.104) RD_GDP 0.668** (0.216) LNPAT_100 TH 0.403+ (0.206) RANDOM EFFECTS σ 2 v 0.355 5.05E-02 0.097 2.37E-18 1.98E-30 7.08E-21 σ 2 u 24.868 10.294 10.397 10.365 10.611 10.464 Nº observations 8994 7420 7420 7420 7420 7420 Nº superclusters (Regions) 17 17 17 17 17 17 Nº clusters (Firms) 1471 1448 1448 1448 1448 1448 Log likelihood −3219.29 −2554.487 −2552.051 −2555.008 −2550.909 −2553.326 Wald Х 2 . 492.03 490.89 497.22 493.63 496.62 Prob >Х 2 . 0.000 0.000 0.000 0.000 0.000 LR test Х 2 (vs. logistic model) 3847.66 1628.91 1632.19 1627.86 1636.06 1630.3 Prob ≥Х 2 0.000 0.000 0.000 0.000 0.000 0.000 LR test Х 2 (vs. 2-level model) 14.25 Prob ≥Х 2 0.0002 Prob ≥Х 2 (divide by 2) 0.0001 ICC REGION 1.24% 0.37% 0.71% 0.00% 0.00% 0.00% ICC FIRM|REGION 88.46% 75.87% 76.13% 75.91% 76.33% 76.08% Notes: Standard errors in brackets; +, *, **, *** denote significance at the 10, 5, 1, and 0.1% levels. M.J. Rodríguez-Gulías et al. International Review of Economics and Finance 92 (2024) 741–757 749 The LR test comparing the Model 0 with the simplest two-level firm-occasion model rejects the null hypothesis that there are no regional supercluster effects (H 0 : σ 2 v =0). This result indicates that the regional dimension partly determines the R&D-based innovation of the MNEs. Hence, MNEs located in the same region are more homogenous with each other in terms of R&D-based innovation than with other MNEs located in different regions. Located in different regions. It is worth to note that the intraclass correlation coefficient (ICC), calculated as 0.355/(0.355 +24.868+3.29) =0.0125, shows that 1.25% of the residual variation in the propensity to have R&D-based innovation is attributable to the region’s unobserved characteristics. The estimates indicate that R&D-based innovation of the MNEs is positively affected by the region’s cross-border knowledge sourcing (IMPTECH_GDP), supporting the H1. This finding suggests that MNEs in regions with a high volume of payments for foreign licenses or technical assistant are more exposed to such foreign knowledge that allows them to enrich their innovation activities (Perri et al., 2017). Innovation by MNEs not acquiring foreign licences or technical assistance may also result through interactions with geographically proximate MNEs acquiring foreign licences or technical assistance. In contrast, the export spillovers available in a region (LNREGEXPORT) fail to be significant (H2). While the effect of export spillovers on the propensity to internationalize is clearly contrasted in the literature (Requena-Silvente & Castillo-Gim´ enez, 2007), it does not seem to exist on MNEs’ innovation. Additionally, at the regional level, the region’s R&D financial effort (RD_GDP) and the production spillovers (LNPAT_100 TH) positively affect R&D-based innovation. At the firm level, the estimates show that four out the eight internal factors positively affect R&D-based innovation in MNEs. In particular, the one period lagged internal (INTRD_SALES) and external R&D intensity (EXTRD_SALES) significantly rise the probability of having R&D-based innovation. Similarly, the probability of innovating increases with both the firm international dimension (LNEXPORT) and the overall firm size (LNEMP). However, firm age (LNAGE), paying for foreign licenses or technical assistance (IMPTECH), patenting (CUMPAT) and operating in high-medium technology (HIGHTECH) fail to be significant for all estimated models. Previous findings underline that the internal innovative resources of MNEs play a more substantive role in explaining R&D-based innovation than the region’s external knowledge sources, although the latter also play a relevant role. More in detail, the statistics of the Model 1 clearly show that there is regional variability (or a “regional effect”). Thus, applying multilevel models seems the most appropriate empirical approach to study the innovation performance also for MNEs (first research gap) and, in addition, the region’s cross-border knowledge sourcing emerges as an important source of foreign knowledge whose role in firms’ innovative performance deserves to be analysed in depth (second research gap). 4.2. The effect of foreign knowledge sources: differences by firm size To address the third research gap, we split the overall sample into two sub-samples according to firm size. The percentage of observations with R&D-based innovation is 51.01% for large MNEs and 17.94% for SM-MNEs over the study period. By year, these percentages change lightly as it is shown in Fig. 3. In order to test whether the effects of the region’s foreign knowledge sources on R&D-based innovation can differ between large firms and SM-MNEs firms, the models in Table 4 are re-estimated by subsamples (see Table 5 and Table 6). The distinction between large and SM-MNEs reveals some interesting insights. First, neither cross-border knowledge sourcing (H1) nor export spillovers (H2) are significant in any of the subsamples analysed. Therefore, the evidence found in the global sample in favour of the effect of cross-border knowledge sourcing on R&D-based innovation in MNEs could be considered weak. Second, while the region’s technological resources (RD_GDP) are positively related to R&D-based innovation of MNEs regardless firm size, the region’s production spillovers (LNPAT_100 TH) remains significantly associated with the innovative performance of the large MNEs only (Table 5). These results suggest that large MNEs have a greater ability to exploit external sources of knowledge available in the region where they are located than SM-SMEs. The greater internal resources of the former would allow them to tap regional sources of knowledge more successfully than the latter, increasing their innovation performance. Third, concerning the firm’s internal factors, several differences are observed. Thus, the previous patent activity (CUMPAT) and the international dimension of the firm (LNEXPORT) remain significant and positively related to innovation in the subsample of SM-MNEs, but these relationships are not significant in the subsample of large MNEs. In contrast, firm age (LNAGE) is positively associated with Fig. 3. Firms with R&D-based innovation by subsample and year: number and percentage. M.J. Rodríguez-Gulías et al. International Review of Economics and Finance 92 (2024) 741–757 756 ANNEX B. Table B.1 Collinearity Diagnostics Variable VIF SQRT VIF Tolerance R-Squared INTRD_SALES 1.13 1.06 0.8869 0.1131 EXTRD_SALES 1.12 1.06 0.8893 0.1107 LNEXPORT 2.28 1.51 0.4395 0.5605 LNEMP 2.30 1.52 0.4343 0.5657 LNAGE 1.06 1.03 0.9454 0.0546 IMPTECH_GDP 1.22 1.10 0.8203 0.1797 RD_GDP 2.31 1.52 0.4336 0.5664 LNPAT_100 TH 1.96 1.40 0.5112 0.4888 LNREGEXPORT 1.34 1.16 0.7473 0.2527 Mean VIF 1.63 References Acs, Z. J., Audretsch, D. B., & Feldman, M. (1994). R&D spillovers and recipient firm size. The Review of Economics and Statistics, 76(2), 336–340. https://doi.org/ 10.2307/2109888 Acs, Z. J., Braunerhjelm, P., Audretsch, D. B., & Carlsson, B. (2009). The knowledge spillover theory of entrepreneurship. Small Business Economics, 32(1), 15–30. https://doi.org/10.1007/s11187-008-9157-3 Alguacil, M., Martí, J., & Orts, V. (2017). Firm heterogeneity and the market scope of European multinational activity. International Review of Economics & Finance, 51, 645–659. https://doi.org/10.1016/j.iref.2022.08.017 Alguacil, M., Martí, J., & Orts, V. (2023). Firms’ characteristics and their international location strategy: Micro-level evidence from European countries. International Review of Economics & Finance, 83, 97–113. https://doi.org/10.1016/j.iref.2017.07.031 Andersson, U., Dasi, A., Mudambi, R., & Pedersen, T. (2016). Technology, innovation and knowledge: The importance of ideas and international connectivity. Journal of World Business, 51(1), 153–162. https://doi.org/10.1016/j.jwb.2015.08.017 Autio, E., Kenney, M., Mustar, P., Siegel, D., & Wright, M. (2014). Entrepreneurial innovation: The importance of context. Research Policy, 43(7), 1097–1108. https:// doi.org/10.1016/j.respol.2014.01.015 Barbosa, N., & Eiriz, V. (2009). Linking corporate productivity to foreign direct investment: An empirical assessment. International Business Review, 18(1), 1–13. https://doi.org/10.1016/j.ibusrev.2008.10.003 Bellmann, L., Crimmann, A., Evers, K., & Hujer, R. (2013). Regional determinants of establishments’ innovation activities: A multi-level approach. Bonn, Germany: Institute for the Study of Labor. working paper, IZA Discussion Papers, No 7572. Beugelsdijk, S. (2007). The regional environment and a firm’s innovative performance. A plea for a multilevel interactionist approach. Economic Geography, 83(2), 181–199. https://doi.org/10.1111/j.1944-8287.2007.tb00342.x Chen, Y. S., Chang, K. C., & Chang, C. H. (2012). Nonlinear influence on R&D project performance. Technological Forecasting and Social Change, 79, 1537–1547. https://doi.org/10.1016/j.techfore.2012.04.007 Choung, J., & Hwang, H. (2000). Transition of latecomer firms from technology users to technology generators: Korean semiconductor firms. World Development, 28 (5), 969–983. https://doi.org/10.1016/S0305-750X(99)00161-8 Corsi, C., & Prencipe, A. (2016). Improving innovation in university spin-offs: The fostering role of university and region. Journal of Technology Management and Innovation, 11(2), 13–21. https://doi.org/10.4067/S0718-27242016000200002 COTEC. (2017). Informe COTEC. Madrid: Fundaci´ on COTEC para la innovaci´ on tecnol´ ogica. Criscuolo, C., Haskel, J. E., & Slaughter, M. J. (2010). Global engagement and the innovation activities of firms. International Journal of Industrial Organization, 28(2), 191–202. https://doi.org/10.1016/j.ijindorg.2009.07.012 Czarnitzki, D., & Hottenrott, H. (2009). Are local milieus the key to innovation performance? Journal of Regional Science, 49, 81–112. https://doi.org/10.1111/j.1467- 9787.2008.00584.x D’Agostino, L. M., & Moreno, R. (2019). Green regions and local firms’ innovation. Papers in Regional Science, 98(4), 1585–1608. https://doi.org/10.1111/pirs.12427 De Clercq, D., Hessels, J., & van Stel, A. (2008). Knowledge spillovers and new ventures’ export orientation. Small Business Economics, 31, 283–303. https://doi.org/ 10.1007/s11187-008-9132-z Du, J., & Temouri, Y. (2011). Internationalisation and high growth firms: The OECD experience. In International conference on information management, technical management and intelligent technology (ICITI), 26–28 March 2011, Hong Kong. Eisenhardt, K. M., & Santos, F. M. (2001). Knowledge-based view: A new theory of strategy? In A. Pettigrew, H. Thomas, & R. Whittington (Eds.), Handbook of strategy and management. Sage Publications. EUROSTAT. (2018). Community Innovation Survey. Science and technology, Enterprises by main types of innovation, NACE Rev. 2 activity and size class. Retrieved from http://ec.europa.eu/eurostat/en/web/products-datasets/-/INN_CIS9_TYPE. Feldman, M. P., & Audretsch, D. B. (1999). Innovation in cities: Science-based diversity, specialization and localized competition. European Economic Review, 43, 409–429. Fletcher, D., & Harris, S. (2012). Knowledge acquisition for the internationalisation of the smaller firm: Content and sources. International Business Review, 21, 631–647. https://doi.org/10.1016/j.ibusrev.2011.07.008 Fleury, A., Fleury, M. L., & Borini, F. M. (2013). The Brazilian multinationals’ approaches to innovation. Journal of International Management, 19(3), 260–275. https:// doi.org/10.1016/j.intman.2013.03.003 Fontana, R., Geuna, A., & Matt, M. (2006). Factors affecting university-industry R&D projects: The importance of searching, screening and signaling. Research Policy, 35, 309–323. https://doi.org/10.1016/j.respol.2005.12.001 Forero-Pineda, C., Corredor, S., & Forero, N. (2010). Business networks and innovation in SMEs and large firms of a developing country. In The 5th IEEE international conference on management of innovation and technology. June 2010, Sinresearch gapore. García-Cabrera, A. M., García-Soto, M. G., & Su´ arez-Ortega, S. M. (2017). Macro-level spillovers and micro-level capabilities as antecedents of young SMEs’ propensity to export and to become a born global. The International Entrepreneurship and Management Journal, 13(4), 1199–1220. https://doi.org/10.1007/s11365-017-0451- x Garg, P., & Zhao, M. (2018). Knowledge sourcing by multidivisional firms. Strategic Management Journal, 39(13), 3326–3354. https://doi.org/10.1002/smj.2956 M.J. Rodríguez-Gulías et al. International Review of Economics and Finance 92 (2024) 741–757 757 Greene, W. H. (2012). Econometric analysis (7th ed.). New York: Pearson. Hall, B. H., Mairesse, J., & Mohnen, P. (2010). Measuring the returns to R&D. Handbook of the Economics of Innovation, 2, 1033–1082. https://doi.org/10.1016/S0169- 7218(10)02008-3 Helpman, E., Melitz, M. J., & Yeaple, S. R. (2004). Export versus FDI with heterogeneous firms. The American Economic Review, 94(1), 300–316. https://doi.org/ 10.1257/000282804322970814 Jaffe, A. B., Trajtenberg, M., & Handerson, R. (1993). Geographic localization of knowledge spillovers as evidenced by patentcitations. Quarterly Journal of Economics, 108(3), 577–598. https://doi.org/10.2307/2118401 Jantunen, A., Nummela, N., Puumalainen, K., & Saarenketo, S. (2008). Strategic orientation of born globals - do they really matter? Journal of World Business, 43, 158–170. https://doi.org/10.1016/j.jwb.2007.11.015 Laursen, K., & Salter, A. (2004). Searching high and low: What types of firms use universities as a source of innovation? Research Policy, 33, 1201–1215. https://doi. org/10.1016/j.respol.2004.07.004 Leckie, G. (2013). Module 11: Three-level multilevel models: Concepts, LEMMA (learning Environment for multilevel Methodology and applications). Centre for Multilevel Modelling, University of Bristol. Available in: http://www.bristol.ac.uk/cmm/learning/online-course/. March 2019. L´ opez-Bazo, E., & Motell´ on, E. (2018). Innovation, heterogeneous firms and the region: Evidence from Spain. Regional Studies, 52(5), 673–687. https://doi.org/ 10.1080/00343404.2017.1331296 Mora Sanguinetti, J., & Fuentes, A. (2012). An analysis of productivity performance in Spain before and during the crisis: Exploring the role of institutions. OECD Economics Department Working. Paper No. 973. OECD. (2018). OECD Oslo Manual 2018: Guidelines for collecting, Reporting and using Data on innovation (4th ed.). Paris, France; Eurostat: Luxembourg: OECD Publishing. https://doi.org/10.1787/9789264304604-en, 2018. Perri, A., Scalera, V. G., & Mudambi, R. (2017). What are the most promising conduits for foreign knowledge inflows? Innovation networks in the Chinese pharmaceutical industry. Industrial and Corporate Change, 26(2), 333–355. https://doi.org/10.1093/icc/dtx004 Powers, J., & McDougall, P. (2005). University start-up formation and technology licensing with firms that go public: A resource-based view of academic entrepreneurship. Journal of Business Venturing, 20(3), 291–311. https://doi.org/10.1016/j.jbusvent.2003.12.008 Raspe, O., & Van Oort, F. (2009). Localized knowledge externalities and firms: An intrinsic multilevel issue” in FIRB-RISC conference research and entrepreneurship in the knowledge-based economy. Milan, Italy: KITeS-Cespri, Bocconi University. Requena-Silvente, F., & Castillo-Gim´ enez, J. (2007). Information spillovers and the choice of export destination: A multinational logit analysis of Spanish young SMEs. Small Business Economics, 28, 69–86. https://doi.org/10.1007/s11187-005-7324-3 Rodríguez-Gulías, M. J., Rodeiro-Pazos, D., & Fern´ andez-L´ opez, S. (2016). The regional effect on the innovative performance of university spin-offs: A multilevel approach. Journal of the Knowledge Economy, 7(4), 869–889. https://doi.org/10.1007/s13132-015-0287-y Rodríguez-Gulías, M. J., Rodeiro-Pazos, D., Fern´ andez-L´ opez, S., & Nogueira-Moreiras, M. A. (2020). The effect of regional resources on innovation: A firm-centered approach. The Journal of Technology Transfer. https://doi.org/10.1007/s10961-020-09811-8 Snijders, T., & Bosker, R. (1999). Multilevel Analysis: An introduction to basic and advanced multilevel modeling. London/Thousand Oaks/New Delhi: Sage Publications Ltd. Spescha, A., & Woerter, M. (2019). Innovation and firm growth over the business cycle. Industry & Innovation, 26(3), 321–347. https://doi.org/10.1080/ 13662716.2018.1431523 Srholec, M. (2010). A multilevel approach to geography of innovation. Regional Studies, 44, 1207–1220. https://doi.org/10.1080/00343400903365094 Steele, F. (2009). Module 7: Multilevel models for binary responses: Concepts. In LEMMA (learning environment for multilevel methodology and applications). Centre for Multilevel Modelling, University of Bristol. Available in: http://www.bristol.ac.uk/cmm/learning/online-course/. March 2019. Sternberg, R., & Arndt, O. (2001). The firm or the region: What determines the innovation behavior of European firms? Economic Geography, 77, 364–382. https://doi. org/10.2307/3594106 Tang, M. C. (2017). Total factor productivity or labor productivity? Firm heterogeneity and location choice of multinationals. International Review of Economics & Finance, 49, 499–514. https://doi.org/10.1016/j.iref.2017.03.016 Vega-Jurado, J., Guti´ errez-Gracia, A., Fern´ andez-De-Lucio, I., & Manjarr´ es-Henríquez, L. (2008). The effect of external and internal factors on firms’ product innovation. Research Policy, 37, 616–632. https://doi.org/10.1016/j.respol.2008.01.001 Wagner, J. (2006). International firm activities and innovation: Evidence from knowledge production functions for German firms. In University of lüneburg working paper series in economics, 25. Lüneburg: Hamburgisches Welt-Wirtschafts-Archiv (HWWA). Wang, C. C., & Lin, G. C. (2013). Dynamics of innovation in a globalizing China: Regional environment, inter-firm relations and firm attributes. Journal of Economic Geography, 13(3), 397–418. https://doi.org/10.1093/jeg/lbs019 Yeaple, S. R. (2009). Firm heterogeneity and the structure of US multinational activity. Journal of International Economics, 78(2), 206–215. https://doi.org/10.1016/j. jinteco.2009.03.002 Zaheer, S., & Nachum, L. (2011). Sense of place: From location resources to MNE locational capital. Global Strategy Journal, 1, 96–108. https://doi.org/10.1002/gsj.2 M.J. Rodríguez-Gulías et al.