Exploring science-and-technology-led innovation: A cross-country study

Raghupathi, Viju; Raghupathi, Wullianallur Article Exploring science-and-technology-led innovation: A crosscountry study Journal of Innovation and Entrepreneurship Provided in Cooperation with: Springer Nature Suggested Citation: Raghupathi, Viju; Raghupathi, Wullianallur (2019) : Exploring science-andtechnology-led innovation: A cross-country study, Journal of Innovation and Entrepreneurship, ISSN 2192-5372, Springer, Heidelberg, Vol. 8, Iss. 1, pp. 1-45, https://doi.org/10.1186/s13731-018-0097-0 This Version is available at: https://hdl.handle.net/10419/259578 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/4.0/ RESEARCH Open Access Exploring science-and-technology-led innovation: a cross-country study Viju Raghupathi 1* and Wullianallur Raghupathi 2 * Correspondence: [email protected] 1 Koppelman School of Business, Brooklyn College of the City University of New York, 2900 Bedford Ave, Brooklyn, NY 11210, USA Full list of author information is available at the end of the article Abstract Countries can enhance endogenous innovation using multifaceted incentives for science and technology indicators. We explore country-level innovation using OECD data for research and development (R&D), patents, and exports. We deploy a dual methodology of descriptive visualization and panel regression analysis. Our results highlight industry variances in R&D spending. As a nation develops, governmental expenditure on R&D decreases, and businesses take on an increasing role to fill the gap, increasing local innovation. Our portfolio of local versus foreign resident ownership of patents highlights implications for taxation/innovation policies. Countries with high foreign ownership of patents have low tax revenues due to the lack of associated costs with, and mobility of income from, patents. We call on these countries to devise targeted policies encouraging local patent ownership. Policy makers should also recognize factors influencing high-technology exports for innovation. Lastly, we call on countries to reinstate horizontal and vertical policies, and design national innovation ecosystems that integrate disparate policies in an effort to drive economic growth through innovation. Keywords: Innovation, Patents, Science and technology, Research and development, Exports, Panel analysis Introduction Innovation is a key driver of economic growth and a prime source of competition in the global marketplace (Organization for Economic Cooperation and Development (OECD) 2005); at least 50% of growth is attributable to it (Kayal 2008; Organization for Economic Cooperation and Development (OECD) 2005). Notable in this regard are the levels of adoption and creation of technological innovation (Grupp and Mogeec 2004;Niosi2010) and technological learning (Koh and Wong 2005; Organization for Economic Cooperation and Development (OECD) 2005)increating this expansion. Acountry’seconomicgrowthprogressesthroughthreestagesoftechnological change and productivity, namely factor-driven growth, investment-driven growth, and innovation-driven growth (Koh and Wong 2005; Rostow 1959;World Economic Forum (WEF) 2012). Factor-driven economies produce goods based on natural endowments and low labor cost; investment-driven economies accumulate capital (technological, physical and human) and offer investment incentives; and innovation-driven economies emphasize research and development (R&D), Journal of Innovation and Entrepreneurshi p © The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 https://doi.org/10.1186/s13731-018-0097-0 entrepreneurship, and innovation. Progressing from one growth stage to another involves transitioning from a technology-importing economy that relies on endowments, capital accumulation, infrastructure, and technology imitation, to a technology-generating economy that focuses on creating new products or knowledge using state of the art technology (Akcali and Sismanoglu 2015). In the creation of new products or knowledge, economies use a systemic approach to represent the interaction of public and private institutions with policies, incentives, and initiatives (Organization for Economic Cooperation and Development (OECD) 1997). Institutions include research entities, public laboratories, innovative enterprises, venture capital firms, and organizations that finance, regulate, and enable the production of science and technology (Malerba 2004; Mazzoleni and Nelson 2007;Niosi2010). The government also plays a key role in supporting these institutions through policies, targeted incentives, R&D collaboration, and coordinated infrastructure. According to the Oslo manual (Organization for Economic Cooperation and Development (OECD) 2005), innovation has been defined as the implementation of a new or significantly improved product/service, process, marketing method, or organizational method in business practices, workplace organization, or external relations. Though many innovation studies use variations of this definition, the common thread is Schumpeter (2008). Schumpeter proposes that the crux of capitalism lies in production for the mass market through creative destruction, that is, the continuous process of generating new products, processes, markets, and organizational forms that make existing ones obsolete (Lee 2015). In today’s digital era, technology is integral to advancing such creative destruction. It is no surprise that innovation vis-à-vis technological innovation drives economic growth. Technological innovation is measured using science and technology (S&T) indicators. These indicators include resources devoted to R&D, patents, technology balance of payments, and international trade in R&D-intensive industries. The importance given to S&T indicators increased with the call for a comprehensive analysis of the economy that not only incorporates economic indicators, but also those that represent knowledge creation (Lepori et al. 2008). However, for S&T to translate to improved economic development (represented by improved quality of life, as well as wealth and employment creation), it needs to be geared towards bringing new products/processes into the marketplace—that is, towards innovation (Siyanbola et al. 2016). Attaining national development goals requires evidence-based and informed policy-making. Incorporating S&T indicators offers the scientific evidence needed to effectively design, formulate, and implement national innovation policies that contributetoeconomicdevelopment.Webaseourstudyonthispremiseandutilizethe S&T indicators of R&D, patents, and exports to explore country-level technological innovation for policy analysis. We offer research-based suggestions for governments to compare innovation policy initiatives, seek insights into solving national development problems, identify outstanding best practices, and work collaboratively. The rest of the paper is organized as follows: section 2 describes the research background; section 3 covers methodology; section 4 discusses the analyses results and discussion; section 5 offers scope and limitations; section 6 covers contributions and implications for future research; and section 7 presents the conclusions of our research. Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 2 of 45 Research background There are several theoretical models that offer rationales for technological innovation. The technology gap model developed by Posner (1961) explains how countries that are technologically advanced introduce new products into a market and enjoy the innovative advantage (Krugman 1986). However, this comparative advantage is transient and shifts over time to other countries that show sustained innovative activities due to technological improvements. The product life cycle hypothesis shows that industrialized countries with a high degree of human capital and R&D investment produce more technical innovations and new products (Maksimovic and Phillips 2008). These countries enjoy the comparative advantage early in the product’s life cycle. But as the country exports and the product becomes more standardized, it allows other countries to reproduce at a lower cost (with advanced technology) and gain market share. The endogenous growth model shows that technology, knowledge, and human capital are endogenous and primary contributors to innovation and economic growth (Gocer et al. 2016). It is clear that technological innovation offers a country a competitive edge that contributes to economic development and growth (Gocer et al. 2016). Therefore, measuring this phenomenon takes on increasing significance at national and global levels. We now describe our research framework and conceptualization of the innovation phenomenon. Research framework for innovation In this research, we adapt the comprehensive framework presented at the OECD workshop for national innovation capability using S&T indicators (Qiquan et al. 2006). According to Miles and Huberman (1994), a conceptual framework explains either graphically or in narrative form the main things to be studied—the key factors, concepts, or variables—and the presumed relationships among them. Using this definition, we have laid out the key concepts in our research and how they relate to the overall phenomenon of innovation. We conceptualize innovation using the three components of inputs, knowledge creation and absorption, and outputs. Inputs to innovation are represented by efforts at research and development, including the expenditure and personnel hired for R&D. Knowledge creation and absorption represents national efforts at motivating and rewarding the innovative process. The outputs of innovation represent exports of products and services. Figure 1shows the research framework. Inputs—R&D expenditure and R&D personnel R&D is an important input to national innovation. It includes creative work undertaken systematically to increase the stock and the use of knowledge to devise new applications—both of which have the potential to influence innovation. R&D expenditure is often used to encourage innovation and provide a stimulus to national competitiveness. Research has used R&D expenditure as a percentage of GDP (referred to as R&D intensity) to explain the relationship between firm size and innovative effort (Cohen 2010) and as an input to innovation (Kemp and Pearson 2007). In general, developed countries have higher R&D intensity than developing countries. For a country, the gross domestic expenditure on R&D (GERD), which represents the expenditure on scientific research and experimental development, offers an indication Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 3 of 45 of the allocation of financial resources to R&D in terms of the share in the GDP. Sufficient R&D funding is essential for innovation, economic growth, and sustainable development. Changes in R&D expenditure suggest evolving long-term strategies and policies related to innovation for economic development. GERD can be broken down among the performance sectors of business enterprise, government, higher education, and private not-for-profit institutions serving households. Business enterprise expenditure on R&D (BERD) is an important indicator of business commitment to innovation. Although not all business investments yield positive results, efforts towards R&D signal a commitment to the generation and application of new ideas that lead to new or improved products/services for innovation. Research suggests that R&D spending is associated with productivity and GDP growth. An increase of 0.1% point in a nation’s BERD to GDP ratio could eventually translate to a 1.2% increase in GDP per capita (Expert Panel on Business Innovation 2009). Also, over the last few years, R&D intensity in the business sector has varied considerably between countries (Falk 2006). It is, therefore, useful to analyze this at a global level. The government intramural expenditure on R&D (GOVERD) represents efforts by the government to invest in R&D. Several motivations have been proposed for the same. Endogenous theories present such investment to be the foundation for economic growth (Griliches 1980). Governments respond to market failures in which firms under-invest due to the risk of externalities and information issues (Arrow 1962). Additionally, government funding can stimulate corporate R&D activities (Audretsch et al. 2002; Görg and Strobl 2007). The average government-funded R&D expenditure in 24 OECD countries doubled in three decades, from $6.04 billion in 1981 to $12.3 billion in 2008 (in US dollars, constant prices) (Kim 2014). Higher education expenditure on R&D (HERD) has been the focus of much research since the 1980s. Research has studied the transfer of knowledge and technology Fig. 1 Framework for national innovation Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 4 of 45 between the university and industry, using firmor industry-level data (Adams et al. 2001; Collins and Wakoh 2000; Furman and MacGarvie 2007; Grady and Pratt 2000; Siegel et al. 2001). Since the 1990s, higher education institutions have played an increasingly important role in regional and national development in OECD countries, owing to the growth of strategic alliances across industry, research institutions, and knowledge intensive business services (Eid 2012). There is growing recognition that R&D in higher education institutions is an important stimulus to economic growth and improved social outcomes. Funding for these institutions comes mainly from the government, but businesses, too, fund some activities (Eid 2012). Private, not-for-profit organizations are those that do not generate income, profits, or other financial gain. These include voluntary health organizations and private philanthropic foundations. Expenditure on R&D represents the component of GERD incurred by units belonging to this sector. In addition to expenditure, R&D personnel are an input to the innovation phenomenon in a country. R&D personnel refer to all human capital including direct service personnel, such as managers, administrators, and clerical staff, who are involved in the creation of new knowledge, products, processes, and methods, and can be employed in the sectors of public, private, or academia. Knowledge creation and absorption—patents In the innovation framework, knowledge creation is the process of coming up with new ideas through formal R&D (Organization for Economic Cooperation and Development (OECD) 2005). Knowledge absorption is the process of acquiring and utilizing knowledge from entities such as universities, public research organizations, or domestic and international firms (Organization for Economic Cooperation and Development (OECD) 2005). Factors influencing knowledge absorption include human capital, R&D, and linkages with external knowledge sources. On a national level, the creation and absorption of knowledge is manifested through the evolution of intellectual property (IP). Countries institute regulatory frameworks in the form of patents and copyrights, to protect intellectual property and innovation (Blind et al. 2004). The rationale for protection arises from the fact that innovation amounts to knowledge production, which is inherently non-rival and non-excludable. Non-rival refers to the notion that the amount of knowledge does not decrease when used by others, and non-excludable refers to the unlimited ability of others to use and benefit from the knowledge once it is produced. Countries, therefore, institute legal systems to protect the rights of inventors and patent holders. An example is the Bayl-Dohl Act in the USA. By calibrating the strength of patent protection rights, policymakers can influence national innovation systems (Raghupathi and Raghupathi 2017). Legal protection and exclusivity of the use of knowledge allows investment in R&D and leads to the production of knowledge and innovation. Research in the area of patents has often centered on whether stronger IP rights lead to more innovation (Hall 2007;HuandJaffe2007; Jaffe 2000) and on the endogeneity of patent rights on industries (Chen 2008;Moser2005;Qian2007; Sakakibara and Branstetter 2001). As patent rights change at a national level, industries within a country may react differently according to the importance of such rights to the respective industries (Rajan and Zingales 1998). Exploring patent applications or distribution by industry Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 5 of 45 is therefore a key estimator of the extent of national innovation (Qian 2007). Patent laws have a significant effect on the direction of technological innovation in terms of which industries have more innovations (Moser 2005). In addition to using individual patent applications as a measure of patent activity, national innovation research also uses patent families, which are sets of patents/applications covering the same invention in one or more countries. These applications relate to each other by way of one or several common priority filings (Organisation for Economic Cooperation and Development OECD 2009). Patent families reflect the international diffusion of technology and represent an excellent measure of national innovativeness (Dechezlepreˆtre et al. 2017). In the current research, in contrast to studies that look at domestic patent families in an American (Hegde et al. 2009) or European context (Gambardella et al. 2008; Harhoff 2009; van Zeebroeck and van Pottelsberghe 2011), we adopt a global approach and consider patent families in all three major patent systems: the United States Patent and Trademark Office, the European Patent Office, and the Japan Patent Office. This examination allows us to identify possible international patent-based indicators that enable rigorous cross-country comparisons of innovation performances at national and sectoral levels. Outputs—exports Exports represent an output of the innovative activity of a country. Endogenous growth models suggest that firms must innovate to meet stronger competition in foreign markets (Aghion and Howitt 1998; Grossman and Helpman 1991;Hobday 1995). Firms enhance their productivity prior to exporting in order to reach an optimum level that qualifies them to compete in foreign markets (Grossman and Helpman 1995). Upon entry into the export market, continued exposure to foreign technology and knowledge endows a “learning-by-exporting”effect that offers economies of scale that further enable covering the cost of R&D (Harris and Moffat 2012). In the context of our research framework, we explore the following research questions using a descriptive visualization and an econometric panel regression approach: How do the S&T indicators (R&D expenditure, patents, and exports) influence countrylevel innovation? How do countries around the world differ in terms of innovation with S&T indicators? Research methodology We now discuss our methodology in studying science and technology indicators for national innovation. Table 1summarizes the research methodology. Data collection and variable selection We downloaded innovation data from the Master Science and Technology Indicator (MSTI) database of the Organization for Economic Cooperation and Development (OECD), for the years 2000 to 2016 (https://stats.oecd.org). The MSTI database contains indicators that reflect efforts towards science and technology of Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 6 of 45 OECD member countries and non-member countries. The data covers provisional or final results, as well as forecasts from public authorities. MSTI indicators include, among others, resources dedicated to research and development, patent (including patent families), and international trade in R&D intensive industries. Table 2shows the variables in the research. Research and development variables include gross domestic expenditure on R&D (GERD) for the sectors of business enterprise, higher education, government, and private/non-profit households; funding sources include industry, government, abroad, and other national sources. For business enterprise, we included the industries of aerospace, computer/electronic/optical, and pharmaceutical. We also look at R&D personnel as a percentage of the national total in the abovementioned business sectors. In patents, we consider indicators such as the number of triadic patent families, which includes patents filed at the offices of the European Patent Office, the United States Patent and Trademark Office, and the Japan Patent Office, for the same invention by the same applicant or inventor. Patent variables also include those representing the international flow of patents and cross-border ownerships in the inventive process. Patents are considered for the technology sectors of biotechnology, information and communication technology (ICT), environmental technology, and pharmaceutical. As exports represent theoutputsoftheinnovativeprocess,weconsider total exports in the three industries of aerospace, computer/electronic/optical, and pharmaceutical. We use panel data that includes variables for multiple indicators spanning multiple countries and time period. There are several ways to group country-level data. We use the income-level classification of the World Bank of high, upper-middle, lower-middle, and low income. However, due to the lack of availability of data, we only focus on upper-middle and high-income categories. In addition, we use the region classification of East Asia and Pacific, Europe and Central Asia, Latin America and the Caribbean, Middle East and North Africa, North America, South Asia, and Sub-Saharan Africa. Table 1 Research methodology Data collection Data source: https://stats.oecd.org/ Years: 2000–2016 Variable selection Science and technology indicators (MSTI database) Research & development—R&D expenditure; R&D personnel Patents—international cooperation in technology; patent filing; patent by technology Exports—exports by industry Analytics platform/tools selection Tableau and R Analytics implementation Visualization—Tableau Panel analysis—R Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 7 of 45 Table 2 Variables in the research Category Variable Description Research and development R&D expenditure BERD (%) Business enterprise expenditure on R&D (BERD) as a % of GDP BERD by industry (current PPP dollars $) Business enterprise expenditure on R&D (BERD) performed in the industries of service, aerospace, computer, electronic and optical, pharmaceutical, GERD (%) Gross domestic expenditure on R&D as a % of GDP GERD by sector of performance GERD in the sectors of: business enterprise (GERD_BUS_PERFORM), government (GERD_GOV_PERFORM), higher education (GERD_HIGH_PERFORM), and private sector (GERD_PRIVATE_PERFORM) GERD by source of funds GERD financed by other national sources, government (GERD_GOV_FINANCE), industry (GERD_IND_FINANCE), and abroad (GERD_ABROAD_FINANCE) GOVERD Government intramural expenditure on R&D (GOVERD) as a % of GDP HERD Higher education expenditure on R&D as a % of GDP R&D personnel GOVER_RESEARCHER Government researchers as a % of national total BUS_RESEARCHER Business enterprise researchers as a % of national total HIGH_RESEARCHER Higher education researchers as a % of national total Patents Patent filing PATENT_TRIADIC Number of “triadic”patent families (priority year), based on inventor(s)'s countries of residence PATENT_PCT Number of patent applications filed under the PCT(priority year), based on inventor(s)’s countries of residence PATENT_EPO Number of patent applications filed under the European Patent Office (priority year), based on inventor(s)'s countries of residence International cooperation in patents COOP_COINVENTOR% The total percentage of patents invented with foreign co-inventors under EPO where reference date is prior date. COOP_OWN% The total percentage of patents owned by foreign residents in EPO where reference date is prior date. COOP_ABROAD% The total percentage of patents invented abroad in EPO where reference date is Prior Date. Patents by technology PATENT_BIO Number of patents in the biotechnology sector - applications filed under the PCT (priority year), based on inventor(s)’s countries of residence PATENT_ICT Number of patents in the ICT sector—applications filed under the PCT, based on inventor(s)’s countries of residence (priority year) PATENT_ ENV Number of patents in the ENV_TECH sector—applications filed under the PCT, based on inventor(s)’s countries of residence (priority year) PATENT_PHARMA Number of patents in the PHARMA sector—applications filed under the PCT, based on inventor(s)’s countries of residence (priority year) Exports Export of particular industries EXPORT_AERO Total exports: aerospace industry (current prices) EXPORT_IT Total exports: computer, electronic, and optical industry (current prices) EXPORT_PHAR Total exports: pharmaceutical industry (current prices) Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 8 of 45 America, the service and computer industries dominate throughout the time span, with the aerospace industry taking the smallest share—less than 20%. In the sub-Saharan region, the service industry plays a dominant role. Note that our analysis for this region is limited because data is missing for the years 2010 to 2015, but we chose to include in our analysis the data that is available in order to provide as holistic an outlook as possible. In general, it appears that developed regions focus more on the technology sectors while developing regions focus more on service sectors. R&D personnel We looked at the number of researchers engaged in each performance sector as a percentage of the national total. Regions and countries differ in the allocation of personnel in the different sectors of business, government, and higher education. In developing regions since the government invests more in R&D expenditure, we expected the same trend to reflect in personnel. Figure 10 shows the sectoral distribution of R&D personnel as a percentage of national totals for each region. Figure 10 clearly displays structural variations in the regions for R&D personnel. The regions of East Asia and Pacific, Europe and Central Asia, and North America reveal a steady business sector pattern, engaging the highest share of personnel; the government sector engages the lowest. In other regions, such as Latin America and the Caribbean, the education sector leads in personnel. The business sector is significant as it shows a steady annual increase. It appears that, in developed regions the business sector engages a higher share of R&D personnel than in developing regions. We now move on to analyze patents. Fig. 10 R&D Personnel by region and sector Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 15 of 45 Patents The average number of patent applications filed under the Patent Cooperation Treaty (PCT) reflects the extent of technological innovation in a country (Fig. 11). Figure 11 shows the patent applications filed under PCT for the years 2000 to 2016. In the figure, green denotes a high number and red a low number of applications. The USA has the highest number of patent applications, followed by Japan and China. By comparison, Russia, Argentina, and other countries see a very low number of patent applications, signaling a need for innovative focus. As for the number of patent applications filed under European Patent Office (EPO) (Fig. 12), the USA leads Japan, Germany, and France. We now turn to applications for triadic patent families (Fig. 13). Fig. 11 Average number of patent applications filed under PCT Fig. 12 Average number of patent applications filed under EPO Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 16 of 45 A triadic patent family is defined as a set of patents registered in patent offices of various countries to protect the same invention. The European Patent Office (EPO), the Japan Patent Office (JPO), and the United States Patent and Trademark Office (USPTO) are the three major patent offices worldwide. Counting triadic patent families begins with each inventor’s country of residence and the initial date of registration of the patent. Indicators based on patent families normally enhance the international comparability and the quality of patent indicators. The greatest number of triadic patent families originated in the USA, followed by Japan, Germany, and France. In general, countries like the USA, Japan, Germany, and France are high in the number of patent filings under EPO and PCT. Fig. 13 Applications for triadic patent families Fig. 14 Percentage of patents owned by foreign residents under EPO Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 17 of 45 Co-inventions reflect international collaboration and represent the flow of information and knowledge across borders. They also indicate the flow of funds from multinational companies for R&D. Figure 14 shows the analysis for the percentage of patents owned by foreign residents under EPO. In Fig. 14, countries represented in green indicate a high percentage of ownership of patents with foreign residents; the darker the green, the higher the percentage. Countries in red indicate a low percentage of ownership. The US, Japan, and other EU economies have a relatively low share of patents owned by foreign residents, and most of the patent ownership in these countries is local. By contrast, countries like Argentina, Russia, and Mexico show a high percentage of patents owned by foreign residents. These countries rely on foreign collaboration to strengthen their resources and facilities for innovation (Raghupathi and Raghupathi 2017). This signals a need to strengthen local innovation by targeting education systems to offer relevant skills and knowledge that foster growth. The portfolio of ownership of patents between local and foreign residents is an interesting revelation that offers implications for national policies on taxation and innovation, and will be discussed further in our conclusions. In the analysis of the percentage of patents invented abroad (Fig. 15), the difference globally is not as varied as it is for patents owned by foreign residents (as shown earlier, in Fig. 14). As Fig. 15 illustrates, the majority of countries in our study show less than 1% of patents are invented abroad. Only Switzerland (1.16%) and Ireland (1.28%) show relatively high levels. We looked next at the differences in distribution of patents by sector. We considered the various technology domains of environmental technology, pharmaceutical, ICT, and biotechnology. Figure 16 shows the comparison of the number of patents in each technology domain to the benchmark of the total number of patent applications under PCT. Environmental technology is the application of environmental science, green chemistry, environmental monitoring, and electronic technology to monitor and conserve the natural environment and resources and to mitigate the negative effects of humans on the environment. Sustainable development is the primary objective of environmental technology. Figure 16 shows the number of patents in the environmental technology sector and indicates a steady increase from 2000 to 2011 and a sudden decrease from Fig. 15 Percentage of patents invented abroad under EPO Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 18 of 45 2011 to 2013. That said, the total number of PCT patents over the years shows a trend of increase. The biotechnology sector includes technological applications that use biological systems or living organisms to develop or make products for specific use. The number of patents in the pharmaceutical and biotechnology sectors has been increasing over the years. Among all the sectors and throughout the time span we studied, ICT saw a high number of patents and a consistently positive trend. The number of patents in this sector equals the total number under PCT, highlighting the overall dominance of the ICT sector in the patent industry. We then looked for significant associations among sets of innovation indicators. We started with expenditure on R&D and the R&D personnel in the sectors of business, government, and higher education. Association between R&D expenditure and R&D personnel Governments use R&D statistics collected by countries from businesses to make and monitor policy related to national science and technology. These stats also feed into national economic statistics, such as GDP and net worth. Different performance sectors may have different kinds of associations between R&D expenditure and personnel. Figure 17 shows the associations between BERD and R&D personnel in the business sector. Figure 17 shows a significant positive association (p< 0.0001) in the business sector between expenditure on R&D (BERD) and R&D personnel. Interestingly, the implication is that we should also see an increase in R&D personnel in this sector. But this is not the case when we break down the analysis by region. In Latin America and Caribbean, while the percentage of researchers in the business sector is high, the Fig. 16 Distribution of patents under PCT for various technology domains Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 19 of 45 expenditure is consistently low for all countries in the region. The explanation is that in developing regions such as these, while there is recognition of the need to focus on R&D by employing more personnel, the cost of deployment is relatively low. By contrast, in North America, both the R&D expenditure and R&D personnel are high, likely because of the high cost of deployment of personnel. In East Asia and Pacific, while some countries are high in both business expenditure and personnel, others are low in both. Across all regions, some countries show a large business expenditure on R&D with no associated increase in personnel. Examples include Israel in Latin America and Caribbean, Japan in East Asia and Pacific, and Slovenia in Europe and Central Asia. On the flip side, there are countries, Romania and Ireland (in the region of Europe and Central Asia) among them, that show a large fluctuation in personnel with little change in expenditure. In general, there is a positive association between R&D expenditure and R&D personnel in the business sector. Figure 18 shows the analysis of R&D personnel and intramural expenditure on R&D (GOVERD) for the government sector. Figure 18 shows a significant positive association (p< 0.0001) between GOVERD and R&D personnel. Though the region of East Asia and Pacific is similar to Latin America and Caribbean in government expenditure on R&D, it has a lower percentage of R&D personnel. This can be attributed to a relatively high cost of labor in East Asia & Pacific compared to Latin America & Caribbean. In North America, Fig. 17 Association between BERD and R&D personnel in business enterprise sector Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 20 of 45 Fig. 18 Association between GOVERD and R&D personnel in government sector Fig. 19 Association between HERD and R&D personnel in higher education Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 21 of 45 both expenditure and personnel are lower than those of East Asia & Pacific. This is due to the increased emphasis on R&D by the business sector over the government. Figure 19 shows the relationship between expenditure and personnel in the higher education sector (HERD). In the case of higher education (Fig. 19), we did not find a significant association between the expenditure on R&D and the percentage of researchers (p> 0.05). It is important to analyze gross domestic expenditure on R&D (GERD) because it represents an aggregate of the sectors of business, government, and higher education and because it is considered the preferred method for international comparisons of overall R&D expenditure. Figure 20 shows the relationship between expenditure (GERD) and R&D personnel in all the sectors. The relationship between GERD and the percentage of researchers is significant and positive (p< 0.0001) for the business sector, but significant and negative (p< 0.001) for the government and higher education sectors. This means that in the business sector, an increase in expenditure is associated with an increase in the R&D personnel percentage of researchers). In the government and higher education sectors, an increase in expenditure is associated with a decrease in personnel. This highlights the fact that, in general, most of the R&D expenditure and personnel come from the business sector and not from the government or Fig. 20 Association between GERD and R&D personnel in the sectors Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 22 of 45 higher education sectors. We searched for associations between national exports and business expenditure on R&D (BERD) by industry to see if there were dominant patterns. Association between exports and BERD by industry We analyze business expenditure on R&D and exports for different industries. Figure 21 shows the analysis for the aerospace industry. Figure 21 depicts the exports and BERD for the aerospace industry for each country. The intensity of the color indicates the quantity of exports, while grid size denotes expenditure. Only countries for which data can be adequately mapped are shown in the Fig. 22 Exports and BERD in computer, electronic, and optical industry Fig. 21 Exports and BERD in the aerospace industry Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 23 of 45 diagram. The USA is the leader in exports and expenditure in this industry. France, Germany, the UK, and others show high exports and relatively low expenditure. Japan and China are low in both expenditure and exports in the aerospace industry. Figure 22 shows the exports and business expenditure on R&D for the computer/electronic/optical industry. Fig. 24 Association between patents with foreign co-inventors and R&D expenditure Fig. 23 Exports and BERD in pharmaceutical industry Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 24 of 45 increases, the percentage of patents invented abroad decreases. This finding has important policy implications for governments of developing countries, which should direct more resources to R&D with a view to improving local innovation and its contribution to GDP. This is discussed in detail in the conclusions section. We now discuss our second approach of econometric panel analysis. Econometric panel analysis Our panel analysis follows a threefold structure: first, we perform a regression analysis on exports for each industry; we then analyze the influence of R&D expenditure on patents; and lastly, we explore the international ownership of and investment in patents and the influence on exports. We deploy the PLM package of R for all the analyses. There are certain steps that need to be taken in order to prepare the data for panel analysis. As a first step to ensuring integrity, we inspected the dataset for missing data (Table 3). As the results show, some variables had more than 20% missing values. We deleted these and used the Random Forest algorithm to fill in values for the remaining variables. The descriptive statistics for the complete dataset are depicted in Table 4. The next step was to ensure that the data is stationary and usable for panel analysis. For this, we did unit root testing with Augmented Dickey Fuller (ADF) values (Table 5). As seen in Table 5, the ADF values are all significant (p< 0.01), confirming the appropriateness of data for panel analysis. The next test was to check for multicollinearity among variables. Since the preliminary correlation analysis revealed high correlation between certain variables, we did the variance inflation factor (VIF) for the variables within each industry (Table 6). The results showed some VIFs above 10, confirming multicollinearity. We therefore deleted these variables and reran the test for the remaining. The results were now satisfactory, with all VIFs below 10. Table 5shows the results before and after multicollinearity analysis for each industry. The variables are now ready to be deployed into a regression model for each industry. In panel analysis, the commonly used approaches include independently pooled model, fixed-effect model (also known as first differenced model), and random effectmodel.Theequationfortheindependently pooled model is shown below: Table 8 Test results of random effects for computer industry Table 9 Comparison of fixed effects and random effects model for computer industry Comparison results of fixed effects and random effects model Hausman test chisq = 14.363 df = 17 pvalue = 0.6413 Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 31 of 45 yit ¼aþX K K¼2 βitxkit þεit In this equation, arepresents the intercept, β it represents the coefficient for each attribute, x kit represents the attributes, and ε it represents the residual of this model. The equation for the fixed-effect model is as follows: yit ¼λiþX K K¼2 βitxkit þεit In the equation above, λ i represents the intercept for each individual in the panel dataset, β it represents the coefficient for each attribute, x kit represents the attributes, and ε it represents the residual of this model. Table 10 Random effects model for the computer industry Random effects model Coefficients: Estimate Std. error tvalue Pr(>|t|) INTERCEPT 5.708 1.0029 5.6913 0.00000001906*** BERD% 0.04068 0.25145 0.1618 0.871529 GOVERD% 2.8649 1.0004 2.8636 0.004323** GOVER_RESEACHER% −0.015821 0.01784 −0.8868 0.37549 HERD% −1.0879 0.67842 −1.6036 0.109294 HIGH_RESEARCHER% −0.0028366 0.0082533 −0.3437 0.731188 PATENT_ICT 0.00069838 4.823e−05 14.4809 < 2.20e−16*** PATENT_TRIADIC −0.000060034 8.828e−05 −0.68 0.496731 COOP_COINVENTOR% −0.00075916 0.0050732 −0.1496 0.881094 COOP_OWN% −0.0096889 0.0060417 −1.6037 0.10927 COOP_ABROAD% −0.013263 0.0042571 −3.1156 0.001916** PATENT_ ENV −0.0024575 0.0001768 −13.904 < 2.20e−16*** INCOME_LEVEL 0.14524 1.1914 0.1219 0.903013 OECD_MEM −3.0259 1.2481 −2.4243 0.015607* GERD_GOV_PERFORM −0.036709 0.018841 −1.9484 0.051797. GERD_HIGH_PERFORM 0.01069 0.013953 0.7661 0.443882 GERD_OTHER_FINANCE −0.018753 0.019972 −0.939 0.3481 GERD_GOV_FINANCE −0.0096609 0.010862 −0.8894 0.374113 GERD_ABROAD_FINANCE −0.016215 0.0089331 −1.8151 0.069961. PATENT_EPO 0.000079401 3.614e−05 2.1974 0.028345* Adj. R-Squared: 0.40962 F-statistic: 722.097 p-value: < 2.22e−16 Table 11 Comparison of pooling and fixed effects models for aerospace industry Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 32 of 45 The equation for the random-effect model is shown below: yit ¼aþλiþX K k¼2 βitxkit þεit In the equation, arepresents the intercept of the model, λ i represents the intercept for each individual in the panel dataset, β it represents the coefficient for each attribute, x kit represents the attributes, and ε it represents the residual of this model. However, as seen from the results of the ADF test (Table 5), all the variables satisfied the stationary assumption, thereby eliminating the need for first-differenced models. We therefore decide on the final model through a two-step comparison between the pooling (mixed effects) and fixed effects model, and the fixed effects and random effects model. We first use the analysis of variance (ANOVA) Ftest for both the time and country dimensions. Then, we use the L-M test to check for random effects, and the Hausman test to compare the leading influence of fixed or random effects models. We now discuss the regression analysis by industry. Regression analysis for the computer industry We first did a regression analysis on exports for the computer industry. Table 7shows the results for the comparison of the pooling (mixed) and fixed effects models for the industry. As shown in Table 7, the individual fixed effects model is better than the mixed effects model (p< 0.0001) and the time fixed effects model (p< 0.0001). We used the LM test (Table 8) to check for random effects, and the Hausman test (Table 9) to analyze the significance of the effects. The LM test (Table 8) confirms the significance of the random effects (p< 0.0001). The results of the Hausman test (Table 9) indicate that the random effects model is better than the fixed effects model. Accordingly, we ran the random effects model for the computer industry (Table 10). As Table 10 shows, five variables are significant in influencing exports: business expenditure (BERD), government expenditure (GOVERD), income level, gross expenditure on high education performance (GERD_HIGH_PERFORM), and patent Table 12 Test results of random effects for aerospace Table 13 Comparison of fixed effect and random effects model for aerospace Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 33 of 45 Table 14 Aerospace individual fixed effects model result Table 15 Aerospace random effects model result Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 34 of 45 filings under EPO (PATENT_EPO). Of these, government expenditure is the most significant, since we see that a unit increase in government expenditure is associated with an almost 3-unit (2.8649) increase in exports. Time and country are significant influencers in this industry, along with government and business expenditure, income level, and educational performance. Even though patents have a positive influence on the final exports, the low coefficient (0.0000794) indicates that the effect is not as significant as for other variables. Regression analysis for the aerospace industry Table 11 shows the comparison between the pooling (mixed) and fixed effects models for the aerospace industry. As shown in Table 11, the individual fixed effects model is better than mixed effects model (p< 0.0001). In comparing the mixed and time fixed effects models, we see that that time is not a significant influencer of exports in the industry. The results of the comparison between fixed effects models also support the conclusion that individual fixed effects model is better (p< 0.0001). Next, we performed the LM test to check for random effects (Table 12) and the Hausman test to compare the random and the fixed effects models (Table 13). TheresultsoftheLMtest(Table12) show the random effects to be significant (p< 0.0001) in the analysis for the industry. The results of the Hausman test (Table 13)showthefixedeffects model to be better than the random effects model (p< 0.0001). We therefore ran a fixed effects model analysis for the aerospace industry (Table 14). However, the results in Table 14 show that the model is not significant (p> 0.05) in explaining the relationships in the panel dataset for the industry. In light of this, we performed a random effects model analysis for the aerospace industry (Table 15). In the random effects model shown in Table 15, the variables that have a positive influence on exports include business expenditure (BERD), government expenditure (GOVERD), income level, gross expenditure in the performance sector of higher education (GERD_HIGH_PERFORM), patent filings under EPO (PATENT_EPO), under ICT (PATENT_ICT), triadic patents (PATENT_TRIADIC), patents in environmental technology (PATENT_ENV), and OECD membership (OECD_MEM). In terms of patents, Table 16 Comparison of pooling and fixed effects models for pharmaceutical Table 17 Test results of random effects for pharmaceutical Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 35 of 45 we see that all four types of patents have a positive influence on exports and the influence is much higher here than in the computer industry. Income-level and OECD status (being a member) show a positive influence on exports in this industry. Exports analysis for the pharmaceutical industry The comparison between the fixed and mixed effects for the pharma industry (Table 16) shows that the individual fixed effects model is better (p< 0.0001). Time does not seem to be a significant factor in this industry (p> 0.05). In the comparison between the fixed effects models, the results also support the conclusion that individual fixed effects model is better (p< 0.0001). We did the LM test to see if random effects are significant (Table 17) and the Hausman test to compare the random and fixed effects models (Table 18). From the LM test results (Table 17), we see that the random effects are better (p< 0.0001), and from the Hausman test (Table 18), we see that the random effects model better explains the relationships in this industry. Accordingly, we did the random effects model for the pharma industry (Table 19). The results in Table 19 show that patents, in terms of the number of applications, as well as ownership and funding from abroad, play a more prominent role in exports of this industry than in the others. Table 18 Comparison of fixed effects and random effects models for pharmaceuticals Comparison results of fixed effects and random effects model Hausman test chisq = 24.539 df = 16 pvalue = 0.07838 Table 19 Pharmaceutical random effects model result Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 36 of 45 Patents and government-related variables In this section, we analyze the relationship between the number of patent applications and the variables relating to government, such as expenditure and personnel. The dependent variable is the number of patents under PCT (PATENT_PCT). The six independent variables include GOVERD, GOVER_RESEARCHER, INCOME_LEVEL, OECD_MEM, GERD_GOV_PERFORM, and GERD_GOV_FINANCE. Using a correlation matrix, we found that most of the coefficients are less than 0.5 for the variables (Table 20). We also calculated the variance inflation factor (VIF) to test for multicollinearity (also shown in Table 20). The results of the table show all variables having a VIF of less than 5, thereby indicating no multicollinearity. Table 21 shows the comparison between the pooling (mixed) effects and the fixed effects models. The results indicate that the individual fixed effects model is better than both the mixed effects (p< 0.0001) and the time-fixed effects models (p< 0.0001). We did the LM test (Table 22) and confirmed that the random effects model is significant (p< 0.0001) in this analysis. From the Hausman test (Table 23), we confirmed that the individual fixed effects model is better than the random effects model (p< 0.0001). Accordingly, we ran the individual fixed effects model for patents (Table 24). The results of the individual fixed effects model (Table 24) show government expenditure on R&D (GOVERD) and gross expenditure on R&D in the government sector (GERD_GOV_PERFORM) as having a significant positive effect on the number of patent applications (PATENT_PCT). GOVERD, in particular, has a very large influence—a unit increase brings about 7040 unit increase in patents. This reflects Table 20 Correlation coefficients and VIF test results Table 21 Comparison of pooling and fixed effects models for pharma Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 37 of 45 the importance of government investment in R&D as a driving factor for patents and for innovation. International patents and gross expenditure on R&D We now explore the influence of variables relating to gross expenditure on R&D (GERD, GERD_BUS_PERFORM, GERD_GOV_PERFORM, GERD_OTHER_FINANCE, GERD_IND_FINANCE, GERD_ABROAD_FINANCE, and GERD_HIGH_PERFORM), on the percentage of patents owned by foreign residents (COOP_OWN). We checked for multicollinearity with the VIF test and resolved it by deleting the affected variables, and redoing the test. Table 25 shows the results before and after multicollinearity analysis. Table 26 shows the comparison between pooling (mixed) and fixed effects models. The results in Table 26 show that the fixed effects model is better (p< 0.0001) than the mixed effects model. Also, between the fixed effects model, the individual effects model is better (p< 0.0001). This means the influence of R&D expenditure on patents varies among countries. On the contrary, the fixed effects model has no effect by time (p> 0.05). Table 27 shows the LM test, and Table 28 shows the Hausman test. From the results of the LM test (Table 27), we see that the random effects model is significant (p< 0.0001). The Hausman test (Table 28) shows that the individual fixed effects model is good for this analysis (p< 0.0001). Therefore, we constructed the individual fixed effects model for the variables (Table 29). According to the model results in Table 29, GERD financing from abroad (GERD_ABROAD_FINANCE) and gross expenditure on R&D by higher education sector (GERD-HIGH-PERFORM) have negative coefficients. This is the case with business (BERD) and government expenditures (GOVERD) as well. Of the variables, government expenditure has the largest influence on patent ownership by foreign residents since a unit increase in government expenditure on R&D is associated with a decrease of 27.94 units in patents owned by foreign residents. This depicts a large impact. In summary, a few things stand out from the panel analysis. First, it follows that the random effects model is better at explaining the relationship between exports and the independent variables in all the three industries. Time and country factors are significant in the model. Government expenditure on R&D has a positive influence on exports in all three industries (coefficients = 2.865, 0.085, and 0.563 respectively), implying that increased government expenditure will positively influence innovation through exports. Business expenditure, government expenditure, and gross expenditure in higher education are all key drivers for country-level innovation. Table 22 Comparison of pooling and random effects model Table 23 Comparison of fixed and random effects model Comparison results of fixed effects and random effects model Hausman test chisq = 14.363 df = 17 pvalue = 0.6413 Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 38 of 45 Second, the computer industry has the largest R-square value among the three industries (0.41 in computer, 0.02 in Aerospace, and 0.005 in Pharmaceutical), indicating that the independent variables will have a stronger influence on exports in the computer industry than in the others. The results also point out that, of the three, investment of R&D in the computer industry will have a stronger influence on national innovation. Third, for all three industries, in terms of patents, government expenditure and government financing have a positive correlation with the number of patent applications. For instance, one unit increase in government expenditure (GOVERD) shows a 7040-unit increase in number of patent applications. This reflects the potential for government investment in R&D to influence innovative efforts. Fourth, there is a negative correlation between R&D expenditure being financed from abroad and patents owned by foreign residents. This reflects the necessity for countries to ramp up internal financing for R&D so as to encourage local innovation. Scope and limitations There are some limitations to our study. First, although we cover a period of 16 years, future studies can look at a longer span, facilitating the prospect of uncovering more trends and patterns in the data. Second, we explore associations but not causality in the relationships among S&T indicators for innovation. Third, we consider a small segment of innovation indicators relating to S&T, whereas there is a gamut of variables that can be Table 24 Individual fixed effects model result Individual fixed effects model Coefficients: Estimate Std. error tvalue Pr(>|t|) GOVERD% 7040.2247 2420.4771 2.9086 0.003759*** GOVER_RESEACHER% −5.0068 44.4386 −0.1127 0.910331 GERD_GOV_PERFORM −119.8164 45.9508 −2.6075 0.009338*** GERD_GOV_FINANCE 6.0481 23.246 0.2602 0.794812 Adj. R-Squared: 0.022922 F-statistic: 2.14624 p-value: 0.005758 Table 25 VIF tests before and after resolving multicollinearity Before multicollinearity analysis After multicollinearity analysis Variables VIF Variables VIF COOP_OWN% 2.342497 COOP_OWN% 2.338598 GERD% 2610.814611 GOVERD% 2.388626 GOVERD% 39.041936 HERD% 3.02646 HERD% 124.939324 BERD% 4.469952 BERD% 1860.667698 GERD_GOV_PERFORM 4.156039 GERD_BUS_PERFORM 56.633482 GERD_OTHER_FINANCE 1.183356 GERD_GOV_PERFORM 22.763118 GERD_GOV_FINANCE 2.755481 GERD_OTHER_FINANCE 1102.518588 GERD_ABROAD_FINANCE 1.420166 GERD_GOV_FINANCE 6401.936929 GERD_HIGH_PERFORM 3.677083 GERD_IND_FINANCE 7809.030778 GERD_ABROAD_FINANCE 2758.600606 GERD_HIGH_PERFORM 36.270409 Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 39 of 45 incorporated in future research. Fourth, the data are extracted from a secondary data source (OECD); the aggregated data from multiple sources/models inherently poses some limitations. For instance, data is missing for some years or regions—either it was not collected, or it was collected but not reported. We use patents as an indicator of innovation, but patents are subject to certain drawbacks. Many inventions are not patented, or inventors utilize alternative methods of protection such as secrecy and lead-time. Furthermore, the propensity for patenting varies across countries and industries. Differences in patent regulations and laws pose a challenge for analyzing trends and patterns across countries and over time. Contributions and future research Despite the limitations, our study contributes in many ways to the literature on innovation and policy making. While most studies adopt a firm or enterprise level of innovation analysis, we deploy a country-level analysis using a large and comprehensive dataset from the OECD. The breadth of the indicators in the dataset allows for in-depth multi-dimensional analysis. We utilize a dual methodology of visualization and panel analysis, each of which offers a suite of benefits in terms of research insights and knowledge on a phenomenon. Visualization is an assumption-free and data-driven approach, allowing the data to speak for itself. With no pre-conceived notions, the methodology allows for previously undetected patterns and relationships to emerge from the data. Panel analysis provides the researcher with a large number of data points and reduces the issue of multicollinearity among the explanatory research variables. It therefore improves the efficiency of econometric estimates and allows for multidimensional investigation of a phenomenon. In addition to the contributions in terms of methodologies, the research adds to the literature on empirical innovation studies that deploy an analytic approach. By comparing innovation indicators at a national level, this study calls on policy makers to design appropriate horizontal or vertical S&T policies. The analysis of R&D expenditure by sector and by industry, along with R&D personnel, allows for effective and optimum resource allocation and talent distribution. Patent analysis is done incorporating individual applications as well as triadic families, thereby offering an individual and holistic perspective. The study presents insights on the phenomenon of international Table 26 Comparison of pooling and fixed effects models Table 27 Comparison of pooling and random effects model Raghupathi and Raghupathi Journal of Innovation and Entrepreneurship (2019) 8:5 Page 40 of 45