The role of business visits in fostering R&D investment

Vivarelli, Marco; Piva, Mariacristina; Tani, Massimiliano Working Paper The role of business visits in fostering R&D investment UNU-MERIT Working Papers, No. 2025-010 Provided in Cooperation with: Maastricht Economic and Social Research Institute on Innovation and Technology (UNU-MERIT), United Nations University (UNU) Suggested Citation: Vivarelli, Marco; Piva, Mariacristina; Tani, Massimiliano (2025) : The role of business visits in fostering R&D investment, UNU-MERIT Working Papers, No. 2025-010, United Nations University (UNU), Maastricht Economic and Social Research Institute on Innovation and Technology (UNU-MERIT), Maastricht, https://doi.org/10.53330/DTMT2556 This Version is available at: https://hdl.handle.net/10419/326939 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-sa/4.0/ #2025-010 The role of business visits in fostering R&D investment Marco Vivarelli, Mariacristina Piva and Massimiliano Tani Published 3 April 2025 DOI: https://www.doi.org/10.53330/DTMT2556 Maastricht Economic and social Research institute on Innovation and Technology (UNU-MERIT) email: [email protected] | website: http://www.merit.unu.edu Boschstraat 24, 6211 AX Maastricht, The Netherlands Tel: (31) (43) 388 44 00 UNU-MERIT Working Papers ISSN 1871-9872 Maastricht Economic and social Research Institute on Innovation and Technology UNU-MERIT UNU-MERIT Working Papers intend to disseminate preliminary results of research carried out at UNU-MERIT to stimulate discussion on the issues raised. https://creativecommons.org/licenses/by-nc-sa/4.0/ 1 The role of business visits in fostering R&D investment Marco Vivarelli 1 Department of Economic Policy, Università Cattolica del Sacro Cuore, Milano, Italy Maastricht Economic and Social Research Institute on Innovation and Technology, UNUMERIT, Maastricht, The Netherlands IZA, Bonn, Germany Email: marco.vivare[email protected] Mariacristina Piva Department of Economic Policy, Università Cattolica del Sacro Cuore, Piacenza, Italy Email: [email protected] Massimiliano Tani School of Business, The University of New South Wales, Canberra, Australia IZA, Bonn, Germany Email: [email protected] Abstract Labor mobility is considered a powerful channel to acquire external knowledge and trigger complementarities in the innovation and R&D investment strategies; however, the extant literature has focused on either scientists’ mobility or migration of high-skilled workers, while virtually no attention has been devoted to the possible role of short-term business visits. Using a unique and novel database originating a country/sector unbalanced panel over the period 1998-2019 (for a total of 8,316 longitudinal observations), this paper aims to fill this gap by testing the impact of BVs on R&D investment. Results from GMM-SYS estimates show that short-term mobility positively and significantly affects R&D investments; moreover, our findings indicate - as expected - that the beneficial impact of BVs is particularly significant in less innovative countries and in less innovative industries. These outcomes justify some form of support for BVs within the portfolio of the effective innovation policies, both at the national and local level. Keywords: Business visits; labor mobility; knowledge transfer; R&D investments Acknowledgements Marco Vivarelli and Mariacristina Piva acknowledge the support by the Italian Ministero dell’Università e della Ricerca (PRIN-2022, project 2022P499ZB: “Innovation and labor market dynamics”; principal investigator: Marco Vivarelli; funded by the European Union – Next Generation EU, Mission 4, Component 2, CUP J53D23004830008). The views and opinions expressed are only those of the authors and do not necessarily reflect those of the European Union or the European Commission. Neither the European Union nor the European Commission can be held responsible for them. 1 Corresponding author 2 1 Introduction Innovation policy can be summarised as the set of strategies and measures implemented by governments to promote and support innovation for the purpose of enhancing economies’ productivity and growth through technological advancement (Schot and Steinmuller, 2018). Its relevance has grown over time, as the notion that scientific and technical knowledge has moved away from its original characterization as a global public good whose availability and transferability across the globe makes its spatial location less relevant for accessing its benefits. Under such approach (often adopted by the mainstream international economics, based on the Heckscher-Ohlin-Stolper-Samuelson framework), low-income countries were predicted not only to access knowledge developed in high-income countries without hurdles, but also to be able to use it, grow faster and ‘catch up’ with their high-income counterparts. Historically, this has often not occurred, as experience has shown that scientific knowledge is typically localised and ‘sticky’ (i.e. geographically contained to where investments in research are made, see Heimeriks and Boschma, 2014), tacit (i.e. not coded, but embodied in individuals, see Polanyi, 1966), path-dependent (David, 1975, Arthur, 1994) and its transfer is conditional on the presence of ‘absorptive capacity’ (Cohen and Levinthal, 1989 and 1990) within knowledge recipients, i.e. prior experience in research activities and adequate social capabilities in the knowledge transfer’s destination place (see Abramowitz, 1986; Lee, 2016, 2019 and 2024). Of course, as new challenges emerge alongside the limitations of the prevailing reference model - as is currently the case with the United Nations’ SDGs 1 - the theoretical frames underpinning innovation policy naturally evolve. Yet, adapting the reference model may at 1 In more recent times, innovation policy has been asked to take a more proactive and experimental stance to shed light on the transformative changes required to address the social and environmental challenges contained in the United Nations’ Sustainable Development Goals. 3 times underplay elements that appear to have little consequence until an unexpected shock may reveal otherwise. One such example is the Covid-19 pandemic, which, for a time, has reduced to almost nil physical interactions between people in several countries. As well-known sources of innovation like relations between producers and suppliers, and producers and final customers, also occur through people’s interactions, will lower innovation activity follow fewer interactions due to the pandemic? It is probably too early to observe such so recent dynamics in official statistics, but existing empirical evidence supports the hypothesis that work-related mobility contributes to generate and transmit productive knowledge, even when it occurs over short periods of time, as in the case of business visits (from now on: BVs). The relevant literature (see next section) notes that short-term business visits are neither side effects of international trade and investment flows nor ‘consumption’ items, as accounted for in company books. Can they instead be viewed, at least in part, as a trigger factor in fostering investment in knowledge production activities, such as R&D investments? Or as a strategic choice to access fundamental external knowledge able to increase the expected profitability of R&D expenditures, especially by organisations, regions and countries constrained by geography and resources availability? If so, could some incentives for BVs be included in innovation policies? Indeed, on the one hand, the extant (rare) literature on BVs has investigated their (indirect) impact on productivity and economic growth, but not (with one exception, see next section) their direct effect on innovation activities (such as R&D investments). On the other hand, previous literature has clearly underlined that innovation is characterized by complementarities and super-additive effects (Milgrom and Roberts, 1990 and 1995) which can be considered to be the inner rationale of formal cooperative R&D (Veugelers, 1997; Cassiman and Veugelers, 2002; Piga and Vivarelli, 2003 and 2004) and ‘open innovation’ (Chesbrough, 2003; 4 Chesbrough et al. 2006). Under the same approach, labor mobility can be considered as an investment in accessing information and competences as it is a powerful channel to acquire external knowledge and trigger complementarities in the innovation and R&D investment strategies (Braunerhjelm et al., 2020). Nevertheless, previous research has devoted much attention to long-term labor mobility - both in terms of scientists’ mobility (Geuna, 2015; Verginer and Riccaboni, 2021; Lissoni and Miguelez, 2024) and migration of high-skilled workers (Breschi and Lissoni, 2009; Bosetti et al., 2015; Lissoni, 2018; Fassio et al., 2019) - but virtually no attention has been paid to the possible role of short-term BVs in enhancing knowledge and productivity. We aim to fill these gaps with the present study, whose purpose is to empirically test - in a cross-country sectoral econometric framework - the effect of BVs on R&D investments by combining R&D data from the OECD with novel information from a proprietary database collecting international business visits expenditures worldwide. As better qualified in Section 3, in this study BVs are defined as work-related labour movements lasting less than 3 months, involving no change of residence and hence generally not capped by immigration authorities. Furthermore, we will try to assess whether BVs are particularly crucial in countries that are not within the club of the R&D worldwide champions (the hypothesis being that this channel of knowledge transfer might be particularly important for countries that - although characterized by an adequate absorptive capacity – most need outside knowledge). As a preview of the findings (see Section 4 for a detailed discussion), our results highlight that BVs do raise R&D investments, with an elasticity of about 4-5% and that this effect is particularly significant for countries that are not R&D-leaders. However, in the current policy debate BVs are still considered as consumable expenditures rather than an activity more akin to an investment in accessing and generating productive knowledge. Therefore, we argue for a 5 reconsideration of short-term labor mobility as an explicit target of innovation policy under the prior that its recognition will prompt the development of suitable national and regional incentives to promote its occurrence (see Section 5). The rest of the paper is organised as follows. Section 2 provides a review of the relevant literature, ahead of the summary of data and methodology (Section 3). The results are discussed in Section 4, while Section 5 offers key concluding remarks and policy considerations. 2 The literature The extant literature within the domain of the economics of innovation extensively highlights the role of external knowledge (Section 2.1), but is extremely limited as far as the role of BVs is concerned (Section 2.2). 2.1 The importance of external knowledge Innovation literature and innovation policy have evolved considerably over time. Initially, they were largely focused on promoting internal research and development (R&D) within national boundaries, with an emphasis on supporting specific industries, companies (“national champions”) or technologies deemed critical for national competitiveness. This approach was rooted in the belief that governments could play a crucial role in driving innovation by funding basic research, providing subsidies for R&D activities, and protecting intellectual property rights (Nelson, 1959). However, as the understanding of innovation processes evolved, so did the approaches to their design and implementation. The emergence of the National Innovation Systems (NIS) framework in the late 20th century marked a significant shift in the conceptualisation of 6 innovation policy, as it emphasised the importance of interactions between the actors of an innovation system (firms, universities, research institutions, and government agencies) in generating new products and processes. In particular, it recognised that innovation is not a linear process but rather a complex, interactive, and systemic phenomenon 2 that requires coordination and collaboration across different agents. Moreover, internal and external knowledge generate complementarities that in turn originate super-additive effects in terms of innovative performance (Freeman, 1987; Lundvall, 1992; Nelson, 1993). The development of innovation systems at the national, regional (Cooke et al., 1997), and sectoral (Malerba, 2002) level encapsulates the tenet that scientific and technological knowledge is cumulative and path-dependent (David, 1975; Arthur, 1988), contains important tacit elements, and does not freely or automatically travel over geographical and cultural distances. Instead, it is ‘sticky’ (von Hippel, 1994), and typically exists both outside and within the successful innovator (e.g. March and Simon, 1958; Mansfield, 1968; Rosenberg and Steinmuller, 1988). However, an organisation’s ability to recognise and absorb this external knowledge, and gain an edge over competitors as a consequence, depends on its ‘absorptive capacity’ (Cohen and Levinthal, 1989 and 1990) and on its ‘dynamic’ capabilities (Teece et al., 1997): a set of skills, knowledge, and competencies that organisations develop and accumulate over time. By interacting, each organisation learns new information, problems and solutions, which can be linked to its existing knowledge stock. In turn, these novel linkages expand problem-solving capabilities and skills within individuals and organisations, raising their efficient absorption of new information (Cohen and Levinthal, 1989; Teece et al., 1997), and their creativity (Shalley 2 The interdisciplinary and multi-faceted nature of innovation policy has continued to date, encompassing a wide range of policy instruments and strategies that cover both supply-side (e.g. R&D funding, tax incentives, and support for education and skills development), and demand-side incentives (e.g. public procurement, standards, and regulations targeting innovative products and services) (e.g. Edler & Georghiou, 2007; Mazzucato, 2018). 13 and not the number of passengers or trips 10 - hence it represents the aggregate expenditure associated with BVs for a given sector/country/year and its evolution over the period covered. Second, they refer to overall short-term expenditures, national and international, where the international component, due to collecting procedures, is typically dominant. The database, which reports BV expenditures and the value of output is, unsurprisingly, commercially sensitive as it is used by airlines to forecast and decide their load capacity in each country or regional market. The access to this novel and unique database was made possible through GBTA's agreement to share their information at a discount and financial support from the University of New South Wales awarded to one of the authors. Our empirical strategy - in a cross-country sectoral framework - is based on the merger of three datasets: the NBTA/GBTA database for BVs described above and the publicly available OECD-ANBERD and OECD-STAN databases for economic variables, including R&D investments. The industry classification for both datasets is ISIC Rev.4. The final sample is an unbalanced (due to OECD missing values) panel covering 30 industries (manufacturing and services) for 25 countries in the 1998-2019 timespan, with a total of 8,316 longitudinal observations. All the monetary series have been corrected for purchasing power parities, expressing, at the end, values in constant prices and PPP 2010 US dollars. In order to consider the well-known dynamic and path-dependent dimension of R&D investments 11 , we set up a specification in a Dynamic Panel Data (DPD) framework (1): 10 This means that a longer travel by a top manager is valued more than a shorter travel by a middle-level manager. 11 See the seminal contributions of Arthur (1988 and 1994) and David (1975 and 1985) and what discussed in the previous Sections 1 and 2. The path-dependent nature of R&D investments calls for the inclusion of the lagged dependent variable as in eq.1. 14 𝑙𝑛𝑅𝐷𝑖𝑗𝑡 = 𝛼𝑙𝑛𝑅𝐷𝑖𝑗𝑡−1 + 𝛽1𝑙𝑛𝐵𝑉𝐶𝑖𝑗𝑡 + 𝛽2𝑙𝑛𝐸𝑖𝑗𝑡 + 𝛽3𝑇𝑅𝐴𝐷𝐸𝑗𝑡 + 𝐷𝑢𝑚𝑚𝑖𝑒𝑠 + 𝜈𝑖𝑗𝑡 (1) with: i (sector) = 1,…, 30; j (country) = 1,…, 25; t (time) = 1998,…, 2019; ln = natural logarithm The annual R&D investments at the sectoral level is, therefore, framed within a dynamic specification. The measure of our key impact variable is the entire business visits capital (BVC) obtained from the original BVs flows through the Perpetual Inventory Method (PIM) 12 . Controls are E (the total number of employees per sector) which accounts for the size of the industry, and TRADE intensity ( (𝐼𝑚𝑝𝑜𝑟𝑡+𝐸𝑥𝑝𝑜𝑟𝑡) 𝐺𝐷𝑃 ∗100 ) 13 , to account for international exposure 14 (indeed, this alternative channel of technology transfer might drive innovative investments potentially replacing the role of short-term mobility of workers; see Acharya and Keller, 2009; Brancati et al., 2024). BVs are measured as stocks rather than flows since it is the cumulated acquired knowledge which may affect the current R&D decisions, rather than the sole contemporaneous flow. Indeed, both tangible and intangible capitals are generally measured in terms of stocks in the relevant innovation literature (see Hall and Mairesse, 1995; Parisi, et al., 2006; Ortega et al., 2014). Moreover, using BVs stock allows emphasizing learning and cumulative patterns related to short-term mobility and so considering the mobility effort a longer and more effective source of innovation. Finally, since stocks incorporate the accumulated investments in the past, 12 𝐵𝑉𝐶0= 𝐵𝑉𝑠0 (𝑔+ 𝛿) ; 𝐵𝑉𝐶𝑡= 𝐵𝑉𝐶𝑡−1(1 − 𝛿)+ 𝐵𝑉𝑠𝑡 where BVC is the capital (stock); BVs measure the investment flow; δ is the depreciation rate of 15% (as BVs - seen as a channel of knowledge acquisition - should have a fast degree of obsolescence, similar in principle to the standard discount rate for R&D proposed by Hall, 2007 and Hall et al., 2009); finally, g is computed as an “ex post” 3-year compound growth rate (see Piva et al., 2023). 13 Obtained from the World Bank open data source. 14 This data is only available at the country-year level and so it is repeated for all industries in a given country in a given year. 15 the risk of endogeneity is significantly mitigated. The sample composition by countries is presented in Table 1. Table 1: Sample composition by countries Country Observations Australia 283 Austria 477 Belgium 493 Canada 431 Chile 63 Czechia 516 Denmark 232 Finland 412 France 332 Germany 472 Greece 164 Hungary 324 Ireland 205 Italy 510 Japan 319 Korea 494 Mexico 432 Netherlands 129 Norway 353 Portugal 399 Slovakia 147 Spain 440 Sweden 210 United Kingdom 105 United States 374 Total 8,316 In order to explore possible heterogeneity across nations, we identify two groups of countries according to their innovation intensity (𝑅&𝐷 𝐺𝐷𝑃) at the macro level: the top 5 leader innovators 16 (Korea, Sweden, Japan, US and Germany 15 ) and non-leader innovators (the remaining 25 countries). The sectoral composition of the sample is presented in Table 2. Table 2: Sample composition by industries Industries High (H) Low (L) ISIC Rev. 4 Observations Agriculture, forestry and fishing L 01-03 375 Food products; beverages L 10-11 79 Tobacco products L 12 69 Textiles L 13 217 Wearing apparel L 14 195 Leather and related products, footwear L 15 210 Wood and products of wood and cork, except furniture; articles of straw and plaiting materials L 16 344 Paper and paper products L 17 341 Printing and reproduction of recorded media L 18 331 Coke and refined petroleum products L 19 279 Chemicals and chemical products; pharmaceuticals, medicinal chemical and botanical products H 20-21 323 Rubber and plastics products; other non-metallic mineral products H 22-23 353 Basic metals H 24 370 Fabricated metal products, except machinery and equipment L 25 373 Computer, electronic and optical products H 26 346 Electrical equipment H 27 347 Machinery and equipment n.e.c. H 28 360 Motor vehicles, trailers and semi-trailers; other transport equipment H 29-30 369 Furniture; other manufacturing; repair and installation of machinery and equipment H 31-33 362 Construction L 41-43 384 Accommodation and food service activities L 55-56 266 Publishing activities L 58 172 Motion picture, video and television programme production, sound recording and music publishing activities; programming and broadcasting activities L 59-60 149 Telecommunications L 61 239 Computer programming, consultancy and related activities; information service activities H 62-63 215 15 The top 5 R&D performers of our sample were identified based on the 2019 R&D intensity value, considering that 2019 is the most recent year of the time-span of our analysis. 17 Financial and insurance activities L 64-66 334 Real estate activities L 68 258 Professional, scientific and technical activities H 69-75 256 Administrative and support service activities L 77-82 233 Arts, entertainment and recreation L 90-93 167 Total 8,316 Note: ‘High’ R&D intensity industries (H) include manufacturing and non-manufacturing industries in (High R&D intensity + Medium-high R&D intensity + Medium R&D) intensity groups based on OECD taxonomy, while ‘Low’ (L) include manufacturing and non-manufacturing industries in (Medium-low R&D intensity + Low R&D intensity) groups (see Galindo-Rueda and Verger, 2016). The large number of industries, albeit with an unbalanced dimension, provides a comprehensive picture of the economic structure of the countries analysed. This allows us to take into consideration another possible source of heterogeneity, namely the different innovation propensity across the different industries (Arbelo, et al., 2024). Therefore, following the OECD taxonomy (Galindo-Rueda and Verger, 2016), we cluster industries labelling them ‘High R&D intensity industries’ if they belong to the High R&D intensity + Medium-high R&D intensity + Medium R&D intensity groups, while ‘Low R&D intensity industries’ belong to the Medium-low R&D intensity and Low R&D intensity groups. Table 3 presents descriptive statistics and correlation matrix for the whole sample. As can be seen, a positive, statistically significant and relatively high in magnitude correlation between BVs and R&D emerges from this very preliminary test. 18 Table 3: Descriptive statistics and correlation matrix Mean (St.Deviation) ln(RD) ln(BVC) ln(E) ln(RD) 4.30 (2.40) ln(BVC) 6.42 (1.76) 0.532* ln(E) 4.73 (1.75) 0.341* 0.627* TRADE 0.79 (0.43) -0.204* -0.202* 0.347* Notes: - Employees are expressed in thousands of persons engaged, monetary variables are expressed in millions of constant PPP 2010 US dollars. - * Significant at 95% 4 Results As far as the econometric methodology is concerned, the DPD specification requires GMMfamily estimators to generate unbiased estimates. In particular, given the very high AR(1) correlation of our dependent variable (R&D, equal to 0.97), we opted for a GMM-SYS as the best unbiased estimator (see Blundell and Bond 1998; Pellegrino et al., 2019; Damioli et al., 2021). In Tables 4 and 5 our attention will focus on GMM-SYS estimated coefficients, where the lagged R&D is treated as endogenous 16 , although POLS and FE estimates are also reported 17 as controls. 16 A number of Hansen tests were run to assess the potential endogeneity of other regressors. Results provided evidence of their exogeneity. Indeed, the BVC variable is a stock already considering - by construction - BVs flows in previous years. 17 POLS is affected by upward bias in estimating the lagged dependent variable, meanwhile a downward bias is characterizing the case of the FE estimator. As it can be seen in Tables 4 and 5, the GMM-SYS estimator of the lagged dependent variable is always within these upper and lower bounds, as required. 19 In Table 4 (column 3) the dependent variable is confirmed to be strongly persistent and autocorrelated with a highly significant coefficient of about 0.9. Our variable of interest, BVC, turns out to have a positive and very significant impact on R&D, with an elasticity equal to 4.4%. Employment, as size control variable, has an expected positive and significant effect on R&D, while TRADE does not seem to affect in a significant way the innovative investments at the sectoral level 18 . With regard to the standard GMM diagnoses, the AR(1) and AR(2) tests and the non-significant Hansen test reassure us on the proper choice of the instruments matrix (see Bond 2002). Our key result - using the whole available sample - supports our hypothesis that short-term mobility (i.e. ideas circulation and face-to-face interactions), positively affects innovative investments. In addition, digging into the sample composition and considering the two groups of leader and non-leader R&D countries, we test the same relationship for the sole non-leader innovative countries (column 6) to evaluate if this channel might play a stronger role in countries that are further away from the innovation frontier. Here the hypothesis is that weaker countries in terms of domestic knowledge generation (those with lower R&D/GDP ratios) may benefit more from the knowledge transfer associated with BVs in comparison with innovative leaders . The results indicate that this is the case, as the beneficial impact of BVC is primarily due to its effect in less innovative countries (elasticity increasing to 4.7%). This implies that the free exchange of ideas and people could be essential for their innovative performance. 18 This outcome may be due to the imperfect measure we have at disposal, that is the national figure repeated at the industry level (see above). 20 Table 4: Dependent variable: ln(RD) (1) WHOLE SAMPLE OLS (2) WHOLE SAMPLE FE (3) WHOLE SAMPLE GMMSYS (4) NON-LEADER INNOVATORS POLS (5) NON-LEADER INNOVATORS FE (6) NON-LEADER INNOVATORS GMM-SYS Lagged ln(RD) 0.979*** (0.002) 0.687*** (0.008) 0.904*** (0.029) 0.976*** (0.002) 0.680*** (0.009) 0.874*** (0.003) ln(BVC) 0.009** (0.004) 0.019* (0.012) 0.044*** (0.015) 0.010*** (0.005) 0.015 (0.021) 0.047*** (0.017) ln(E) 0.011** (0.004) 0.043*** (0.014) 0.016*** (0.009) 0.015* (0.005) 0.035** (0.015) 0.027* (0.014) TRADE 0.045 (0.055) 0.029 (0.055) 0.002 (0.001) 0.041 (0.060) 0.004 (0.065) 0.037 (0.068) Constant -0.034 (0.121) 1.221*** (0.136) -0.280 (0.279) 0.093 (0.054) 1.216*** (0.147) 0.052 (0.172) Timedummies Yes Yes Yes Yes Yes Yes Countrydummies Yes - Yes Yes - Yes Timedummies Wald test (p-value) 3.93*** 9.30*** 4.79*** 3.39*** 10.10*** 5.24*** Countrydummies Wald test (p value) 2.42*** - 0.96 2.10*** - 1.13 Hausman test 1426.95*** 1152.94*** Adj. R2 0.97 0.96 R2 within 0.60 0.61 AR(1) AR(2) Hansen test -7.84*** 0.64 222.18 -7.13*** -0.49 205.25 Number groups 604 488 Number obs. 8,316 6,447 Notes: - In columns (4), (5), (6), the top RD/GDP performers in 2019 were excluded (Korea, Sweden, Japan, US, Germany) - * Significant at 90%; ** Significant at 95%; *** Significant at 99% 21 As a complementary exercise, we run the same estimation focusing on the top 5 R&D investors (Table 5). As obvious (column 3), the persistence of R&D is higher in leader innovative countries (96%), implying that in these countries industries are keener to invest in R&D in a stable manner. Turning our attention to our key impact variable, the BVC, although positive, is no longer statistically significant. Our interpretation is that BVs as a channel of knowledge acquisition is not so important in those countries that can rely on an excellent, established, and continuous production of domestic knowledge (while it turns out to be essential for all the other countries that are not within the club of the R&D champions worldwide, see Table 4, column 6). As the diverse industrial structure of economies could influence the observed differences and shape the outcomes, we classify industries as either ‘High’ or ‘Low’ innovative (based on the OECD classification - see Table 2) to discover and qualify potential differences in our results. We present the results in Table 6. The GMM-SYS estimates reveal that BVC has the highest and most significant impact on R&D investments in ‘High’ R&D industries in non-leader countries (column 2). This suggests that external knowledge transfer can be crucial in boosting R&D in industries that are more inclined towards innovation but are situated in non-leader countries. In addition, ‘Low’ industries, in both leader and non-leader innovator countries (columns 3 and 4, respectively), benefit from the mobility of people, with an elasticity of about 3-4%. To summarize, high-tech industries in leader countries do not seem to need BVs as a source of viable knowledge; BVs are instead crucial in non-leader countries and low-tech industries. However, while weaker situations benefit more from BVs in general, the most significant and larger coefficient is detected in the high-tech industries in the non-leader countries, reminding us of the key role of the absorptive capacity (see previous sections). 22 Table 5: Dependent variable: ln(RD) in Leader Innovators (Korea, Sweden, Japan, US, Germany) (1) LEADER INNOVATORS POLS (2) LEADER INNOVATORS FE (3) LEADER INNOVATORS GMM-SYS Lagged ln(RD) 0.987*** (0.004) 0.700*** (0.017) 0.955*** (0.018) ln(BVC) 0.007 (0.007) -0.012 (0.039) 0.023 (0.014) ln(E) -0.001 (0.010) 0.404*** (0.082) 0.001 (0.010) TRADE -0.081 (0.156) 0.033 (0.135) -0.428 (0.458) Constant 0.112 (0.128) -0.318 (0.489) 0.047 (0.094) Time-dummies Yes Yes Yes Country-dummies Yes - Yes Time-dummies Wald test (p-value) 1.32 3.96*** 1.78** Country-dummies Wald test (p-value) 3.05** - 2.18* Hausman test 310.56*** Adj. R2 0.98 R2 within 0.64 AR(1) AR(2) Hansen test -3.34*** 1.59 101.18 Number groups 116 Number obs. 1,869 Notes: - To avoid over-identification (see Roodman, 2009), number of instruments is 115 in columns (3) - * Significant at 90%; ** Significant at 95%; *** Significant at 99% 29 Ortega-Argilés, R., Piva, M. & Vivarelli, M. (2014). The transatlantic productivity gap: is R&D the main culprit? Canadian Journal of Economics, 47(4), 1342–1371. Parisi, M.L., Schiantarelli, F. &. Sembenelli, A. (2006). Productivity, innovation creation and absorption, and R&D. Microevidence for Italy, European Economic Review, 50(8), 2037– 61. Pellegrino, G., Piva, M. & Vivarelli, M. (2019. Beyond R&D: the role of embodied technological change in affecting employment. Journal of Evolutionary Economics, 29, 1151-1171. Piga, C. & Vivarelli, M. (2003). Sample selection in estimating the determinants of cooperative R&D. Applied Economics Letters, 10(4), 243-246. Piga, C. & Vivarelli , M. (2004). Internal and external R&D: A sample selection approach. Oxford Bulletin of Economics and Statistics, 66(4), 457-482. Piva, M., Tani, M. & Vivarelli, M. (2018). Business visits, knowledge diffusion and productivity. Journal of Population Economics, 31(4), 1321-1338. Piva, M., Tani, M. & Vivarelli, M. (2023). The productivity impact of short‑term labor mobility across industries. Small Business Economics, 60(2), 691-705. Polanyi, M. (1966). The Tacit Dimension. Peter Smith. Rogers, M. (1995). International knowledge spill-overs: A cross-country study. In S. Dowrick (Ed.), Economic approaches to innovation. Edward Elgar Publications. Romer, P.M. (1994). The origins of endogenous growth. Journal of Economic Perspectives, 8(1), 3-22. Roodman, D. (2009) A note on the theme of too many instruments. Oxford Bulletin of Economics and Statistics, 71(1), 135–158. Rosenberg, N. & Steinmueller, W.E. (1988). Why are Americans such poor imitators?. American Economic Review, 78(2), 229-234. Salt, J., Singleton, A., & Hogarth, J. (1994). Europe’s international migrants: Data sources, patterns and trends. HMSO. Schot, J., & Steinmueller, W. E. (2018). Three frames for innovation policy: R&D, systems of innovation and transformative change. Research Policy, 47(9), 1554-1567. Shalley, C.E., Zhou, J. & Oldham, G.R. (2004). The effects of personal and contextual characteristics on creativity: Where should we go from here?. Journal of Management, 30(6), 933-958. Tani, M. (2014). Business visits and the quest for external knowledge. Research in Labor Economics, 40, 293–324. Teece, D.J., Pisano, G.P. & Shuen, A. (1997). Dynamic capabilities and strategic management. Strategic Management Journal, 18(7), 509-533. Torre, A. & Rallet, A. (2005). Proximity and localization. Regional Studies, 39(1), 47–59. United Nations - UN (1998). Recommendations on statistics of international migration -United Nations, ST/ESA/STAT/SER.M/58/Rev.1. Verginer, L. & Riccaboni, M. (2021). Talent goes to global cities: the world network of scientists’ mobility. Research Policy, 50(1), 104127. Veugelers , R. (1997). Internal R&D expenditures and external technology sourcing. Research Policy, 26(3): 303-316. von Hippel, E. (1987). Cooperation between rivals: Informal know-how trading. Research Policy, 15(6), 285–305. 30 von Hippel, E. (1994). “Sticky information” and the locus of problem solving: Implications for innovation. Management Science, 40(4), 429-439. Zellner, C. (2003). The economic effects of basic research: evidence for embodied knowledge transfer via scientists’ migration. Research Policy, 32(10),1881–1895. The UNU-MERIT WORKING Paper Series 2025-01 Development strategies for the green hydrogen economy in emerging economies by Fabianna Bacil, Anthony Black, Marina Domingues, Jun Jin, Rasmus Lema, Glen Robbins and Sören Scholvin 2025-02 Do global value chains and local capabilities matter for economic complexity in EU regions? by R. Boschma, E. Hernández-Rodríguez, A. Morrison and C. Pietrobelli 2025-03 Innovation and productivity: the recent empirical literature and the state of the art by Jacques Mairesse, Pierre Mohnen and Ad Notten 2025-04 Digital Technologies, Hiring, Training, and Firm Outcomes: Evidence from AI and ICT Adoption in Indian firmss by Sneha Marydas, Nanditha Mathew, Giordano De Marzo and Carlo Pietrobelli 2025-05 New technologies and employment: the state of the art by Marco Vivarelli and Guillermo Arenas Díaz 2025-06 Formal-Informal Supply Chain Linkages and Firm Productivity in Sub-Saharan Africa: The Role of Human Capital by Gbenoukpo Robert Djidonou, Neil Foster-McGregor and Nanditha Mathew 2025-07 Internet Access and Youth’s Mental Health and Well-being: Evidence from Ethiopia by Trang Thi Pham and Pui-Hang Wong 2025-08 A Twin Transition or a policy flagship? Emergent constellations and dominant blocks in green and digital technologies by Linnea Nelli, Maria Enrica Virgillito and Marco Vivarelli 2025-09 Good for business, not so much for the environment? Entry into importing and the energy intensity of Indian plants by Michele Delera, Nanditha Mathew and Tania Treibich 2025-10 The role of business visits in fostering R&D investment by Marco Vivarelli, Mariacristina Piva and Massimiliano Tani