Seoul Center for Finance and Innovation Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea © 2023 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW Washington DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org This work is a product of the staff of The World Bank with external contributions. The find- ings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy, completeness, or currency of the data included in this work and does not assume responsibility for any errors, omissions, or discrepancies in the information, or liability with respect to the use of or failure to use the information, methods, processes, or conclusions set forth. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Nothing herein shall constitute or be construed or considered to be a limitation upon or waiver of the privileges and immunities of The World Bank, all of which are specifically reserved. Rights and Permissions The material in this work is subject to copyright. Because The World Bank encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Any queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522- 2625; e-mail: pubrights@worldbank.org. World Bank Group Finance, Competitiveness and Innovation Global Practice Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea Seoul Center for Finance and Innovation june 2023 Acknowledgment The note was prepared by a team led by Anwar Aridi (Senior Private Sector Specialist, World Bank’s Finance Competitiveness and Innovation Global Practice), which included Bogang Jun (Inha Uni- versity) as lead author, Kibum Kim (Private Sector Specialist, WB FCI GP), and Kyeyoung Shin (Con- sultant, WB FCI GP). The note benefited from the guidance of the World Bank Management, Zafer Mustafaoglu (Practice Manager) and Jason Allford (Country Manager), and from feedback and comments provided by Alvaro Gonzalez (Lead Economist, WB FCI GP), Marcio Cruz (Principal Economist, IFC), Francesca de Nicola (Senior Economist, IFC), Jonathan David Timmis (Economist, WB EAP Chief Economist Unit), and Pierre-Alexandre Balland (Utrecht University). Ryosun Jang (Consultant, WB) and Wil- liam Shaw (Consultant, WB) edited the report. This knowledge note was made possible by a grant from the Korean Ministry of Economy and Fi- nance, provided through the Seoul Center for Finance and Innovation and the WBG Korea Office. 4 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea Contents Acknowledgment............................................................................................................4 Abstract.........................................................................................................................6 01 Introduction.............................................................................................................7 02 Literature review..................................................................................................... 10 2.1 Technological diversification of firms...................................................................... 11 2.2. Industry 4.0 technologies ....................................................................................13 03 Data........................................................................................................................15 3.1 Financial data.................................................................................................... 16 3.2 Patent data........................................................................................................ 16 3.3 NTIS data............................................................................................................17 04 Methodology........................................................................................................... 18 4.1 Measuring technological relatedness..................................................................... 19 4.2 Measuring technology complexity ........................................................................20 4.3 The empirical model............................................................................................21 05 Results................................................................................................................... 23 5.1 Technology space of Korean manufacturing firms.................................................... 24 5.2 Regression results .............................................................................................. 29 06 Conclusion and discussion....................................................................................... 33 References................................................................................................................... 36 Appendices.................................................................................................................. 41 Appendix 1............................................................................................................... 42 Appendix 2.............................................................................................................. 43 5 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea Abstract This research identifies factors that affect the technological transition of firms toward industry 4.0 technologies, focusing on capabilities and policy impacts using relatedness and complexity measures. A unique dataset of Korean manufacturing firms’ patents and their financial and market information was used for the analysis. The research follows the Principle of Relatedness, which is a recently shaped empirical principle in the field of economic complexity, economic geography, and regional studies. A technology space is built, and each firm’s footprint in the space is traced. Firms’ technology space can help identify those firms that successfully develop new I4 technol- ogies and can examine whether their accumulated capabilities in their previous technology do- mains positively affect their technological diversification and which factors play a critical role in their transition towards I4. In addition, the research analyzes the role of government policy in supporting firms’ knowledge activities in new I4 technologies by combining data on whether the firms received government support for R&D activities. The analysis finds that government support is relatively more effective for firms that possess low technological relatedness and less so for firms that already have accumulated related technologies. This research is expected to inform policy design related to supporting firm diversification towards I4 technologies. 6 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea 01 Introduction 7 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea 1. Introduction In the last half-century, rapid structural change has transformed the Republic of Korea’s economy from an agricultural society to an industrialized society. Achieving industrialization is not an easy task (Wade, 2018). Only a few countries outside Europe and among European offshoots (e.g., Aus- tralia, New Zealand, Canada, and the United States) have achieved industrialization by escaping the Malthusian trap. Excluding city-states such as Singapore and Hong Kong, very few countries in the world (e.g., Japan, Taiwan, Korea) have achieved sustainable economic development capa- ble of transforming their economies from a backward state to an advanced industrialized society (Gerybadze, 2018). This study focuses on Korea, especially its recent experience with the Fourth Industrial Revolution. Scholars have argued that we are living in the era of the Fourth Industrial Revolution (Liao et al., 2017; Lu, 2017), although debate exists over its discontinuity with the previous industrial revo- lution (Rifkin, 2011; Nuvolari, 2019). Living in the Fourth Industrial Revolution implies that the windows of opportunity for economic development are now open for developing countries (Perez, 2003). According to Perez (2003), a technological revolution would provide the best opportunities for catching up with technical changes sufficient for initiating and advancing the development process. This is because every country is a beginner in the early stage of a new techno-economic paradigm, and the probability of success by leap-frogging increases at that stage. Perez (2003) argues that each technological revolution is a cluster of technological systems. For example, during the Second Industrial Revolution, mass production and its successive systems allowed economies to achieve structural change, while the Third Industrial Revolution in around the 1970s was associated with information technology. The most recent industrial revolution is industry 4.0 (I4) (Liao et al., 2017; Lu, 2017; Popkova et al., 2019).1 I4 technologies are expected to provide the domain knowledge that affects a wide range of economic sectors and can therefore be regarded as the core technologies in the Fourth Industrial Revolution. Firms were the main actors creating new technologies and sectors during earlier technological revolutions. Lundvall (1992) argues that innovation at the country level depends on its National In- 1 The term industry 4.0 was introduced as part of a German industrial policy that aimed to improve the production system by combining the IT system with the manufacturing system. However, industry 4.0 is often used interchangeably with the Fourth Industrial Revolution. 8 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea novation System (NIS), consisting of not only enterprises but also research institutes, universities, and government. From a broader perspective, innovation includes accumulated human capital, dynamics of the labor market, learning among enterprises, and sets of government policies. Kim et al. (2000) also describes how human capital formation, the inflow of foreign technologies, and various government policies provide the environment for firms’ innovation activities. Within this environment, firms develop their strategies at the microeconomic level to sustain growth by acquiring technological capabilities. In the long run, these firms’ innovation activities work as the main engine of economic development and lead to the emergence of new sectors (Saviotti and Pyka, 2013; Saviotti et al., 2016). This study, therefore, focuses on the main actor in innovation: the firm. What are the characteristics of I4 technologies, and how do firms diversify their capabilities into I4 technologies? Furthermore, how do policies support firms in the process of acquiring emerg- ing technologies? This paper tries to identify the factors that affect the technological transition of firms toward industry 4.0 technologies by looking at the technological trajectories of Korean firms. The core questions are as follows. (i) What are the characteristics of I4 technologies, and how do these change in terms of complexity? (ii) Among the three factors, that are technological relatedness, complexity, and government policy, which ones play a critical role in firms’ techno- logical diversification, especially when firms enter the I4 technology space? (iii) Which type of firm is more likely to succeed in entering a new technology that is associated with I4 technologies? The target audiences of this research are firms that aim to diversify their technological capabilities towards I4 technologies, in addition to researchers and policymakers who aim at designing and implementing innovation policies that support firms’ capability upgrading. To explore these questions, the research adopts a methodology from the Principle of Relatedness. The principle indicates that firms, cities, regions, and countries are more likely to undertake new economic activities, such as those related to new technologies, new products, and new industries when they already conduct related activities (Hidalgo et al., 2018; Hidalgo, 2021). In their sem- inal work, Hidalgo et al. (2007) construct a product space using world trade data and track the trajectories of industrial diversification by countries in the space. Following Hidalgo et al. (2007), in this paper, the technology space of Korean manufacturing firms is built and the footprints of technological diversification by the firms are traced to examine firms’ capability transition towards I4 technologies. 9 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea 02 Literature review 10 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea 2. Literature review 2.1 Technological diversification of firms Diversification is one of the characteristics of modern enterprises (Chandler, 1990; Chandler, 1993; Chandler et al., 2009). Scholars have found various factors that affect the patterns and probability of success in diversification. For example, the firm’s size (Freeman, 2013) or age (Huergo and Jau- mandreu, 2004) can affect technological diversification. Furthermore, the amount of R&D invest- ment (Griliches, 1998) and qualitative aspects of R&D investment, such as the persistence of the investment, also affect a firm’s technological expansion. Factors associated with human capital, such as the CEO’s expertise in technology and the collaborative environment among R&D work- ers, also affect technological diversification. Finally, firms’ technological strategies (for example, exploitation or exploration) can also affect the pattern of diversification. Firms expand their tech- nological boundaries resulting from the interactions among these factors, rather than a stochastic matter of a single factor (Weaver, 1991; Hidalgo, 2021). Teece (1980, 1982) explains firms’ technological diversification by building a theory of multi-prod- uct firms (i.e., firms with a diversified portfolio of related products). He argues that given the fungible and tacit characteristics of organizational knowledge, profit-seeking firms diversify in a way that avoids the high transaction costs associated with trading services and specialized assets in various markets. The direction of diversification, however, is not random but shows a path-de- pendent pattern. Analyzing US data from 1987, Teece et al. (1994) shows that the most common way for a firm to diversify is by adding related activities. Focusing on technological diversification in firms, Jaffe (1986) introduces a measure of a firm’s technological distance by examining its patents. To characterize the technological position, Jaffe (1986) uses the distribution of firms’ patents by patent class and defines a cosine similarity in- dex that represents the change in distribution over time. He finds evidence that firms’ patents, together with their profits and market value, are systematically related to the technological posi- tion of their research programs, and that movement in the technology space follows the pattern of contemporaneous profits at different technological positions. Breschi et al. (2003) also stud- ies technological diversification in a firm by introducing technological relatedness, focusing on the development of its core technology. They calculate the cosine similarity index by examining the co-occurrence of International Patent Classification (IPC) codes in every patent and find that knowledge relatedness, measured using the cosine similarity index, is a critical factor in firms’ technological diversification. However, the methodology used in previous studies (e.g., using cosine similarity and defining relatedness centering on core activities) has room for improvement. First, the process of aggre- 11 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea gating all technologies to obtain the technological relatedness of a whole industry and calculating firms’ technological relatedness based on their core technology, does not reflect firms’ hetero- geneous characteristics. Similarly, applying industry-level technological relatedness to measure within-firm technological relatedness is rather vague because the level does not take into account the firm’s accumulated technological infrastructure or path-dependent characteristics. Moreover, defining a firm’s core technology as the highest share of IPC code could be an artificial interpre- tation of the analysis. Because the result combines several core technologies, it is hard to differ- entiate the unique core technology. Hence, it is necessary to develop a measure of relatedness involving a firm’s idiosyncratically accumulated technological infrastructure. To better capture firms’ technological trajectories based on what technologies they already own, the methodology from the Principle of Relatedness is used. This empirical principle indicates that firms, cities, regions, and countries are more likely to enter new activities, such as new technolo- gies, new products, and new industries, when they already have related activities in them (Hidalgo et al., 2018; Hidalgo, 2021). For example, by analyzing world trade data, Hidalgo et al. (2007) cal- culate the density of every country’s product and find that countries are more likely to diversify ex- port products toward products with higher density. Similarly, Jun et al. (2020) expand the density measure into three relatedness measures and find that countries are more likely to diversify their exporting products even in bilateral trade when they have more product relatedness, importer relatedness, and exporter relatedness. Similarly, Neffke et al. (2011) use Swedish data on product portfolios in manufacturing plants to show that the density of related industries in a region affects the probability of success for entering a new industry. The Principle of Relatedness holds in regions entering a new technology (Kogler et al., 2013). Us- ing US patent data, Kogler et al. (2013) find that cities are more likely to enter a new technology when the city has a higher density of the new technology. They also find that cities with a higher technology density tend to exhibit faster technological development. The findings of Kogler et al. (2013) and Rigby (2015) show that technological relatedness determines the path of knowledge accumulation at the city level. Although Kogler et al. (2013) and Rigby (2015) analyze patent data, their unit of interest is not firms but geographic regions. Kim et al. (2021) examine the role of technological relatedness in firms’ technological diversifi- cation at the firm level using Korean data and find that a firm is more likely to diversify into new technology when it already has related technologies. However, this research covers the entire range of technology owned by a manufacturing firm, instead of focusing on a certain type of tech- nology such as I4 technologies. Moreover, Kim et al. (2021) do not examine the role of government support in a firm’s technological diversification. This research explores the effect of technological relatedness and government support on a firm’s pattern of technological diversification at a firm level, focusing on I4 technologies. Along with the relatedness measure, a complexity measure is also used to capture the complex characteristics of technologies. The complexity measure represents the competencies of the eco- nomic agents or the sophistication of their economic activity, conserving each characteristic of the activity or agent and by considering their interactions as well (Hidalgo and Hausmann, 2009). In their seminal work, Hidalgo and Hausmann (2009) measure two types of complexity, that of economic agents (countries, regions, or firms) and that of activities (production of goods, pro- vision of services, knowledge creation, or patenting), by looking at a country’s export products. 12 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea The complexity of a country represents that country’s degree of diversification in export products, reflecting information about the complexity of each product. The more a country engages in di- verse and complex activities, the more capable the country is. The more an activity is engaged in by a large number of capable countries, the more sophisticated the activity is. Until now, studies of complexity have demonstrated its effects on the future economic growth of countries (Hidalgo and Hausmann, 2009; Hausmann et al., 2014) and on various types of economic activities, such as services (Stojkoski et al., 2016), employment (Fritz and Manduca, 2021; Wohl, 2020), technol- ogy (Petralia et al., 2017; Balland and Rigby, 2017; Balland et al., 2019), and products (Hidalgo and Hausmann, 2009; Hausmann et al., 2014; Domini, 2022). This study explores the factors that affect the technological diversification of firms focusing on the effect of technological relatedness, together with the effect from the complexity of technology. In addition, the research explores the role of government policy in support of diversification. Gov- ernment support for the R&D agenda is justified because private sector R&D investments are sub- optimal compared to the desired societal level. Since the output of firms’ R&D activities shows the characteristics of public goods, the firms are likely to under-produce technological knowledge because of appropriability issues (Stephan, 1996). Furthermore, considering the uncertainty and indivisibilities of R&D activities, firms are reluctant to engage in R&D that requires high costs and substantial resources (Arrow, 1962). The role of government in achieving the optimal social level of R&D for technological knowledge has been emphasized (Stephan, 1996), and institutions such as patent or government subsidies that help achieve the optimal level of knowledge creation are often regarded as the prime engines of technological progress (Cohen et al., 2000; Rockett, 2010) that further the economic development of countries (North, 1990; Acemoglu et al., 2005). The role of government becomes more critical during the early stages of technological revolutions because the uncertainty of R&D activities tends to be higher than at other stages (Perez, 2003), and the demand for new products from new technologies often lags behind the speed of tech- nological change (Dosi, 1982). Various technologies that initiated the new technological regimes have been seeded with government support. For example, technology related to packet switch- ing funded by the Defense Advanced Research Project Agency (DARPA) resulted in Transmission Control Protocol/Internet Protocol (TCP/IP), which is the cornerstone of information and commu- nication technologies (ICT) (Greenstein, 2010). Mazzucato (2015) discusses various examples of innovation and invention that were spearheaded by the state’s visible hand. Upon the emergence of a new technological paradigm during the Fourth Industrial Revolution, the marginal effect of government support becomes bigger as such a paradigm shift accompanies high uncertainty in R&D. For this reason, the effect of government support, especially its direct support of firms’ I4 technologies development, becomes an important area of concern for re- searchers and policymakers. 2.2. Industry 4.0 technologies There are various definitions of I4 technologies, and there is no formal classification of them (Bal- land and Boschma, 2021; Balland et al., 2019). For example, World Bank (2020) defines digital 13 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea industry 4.0 technologies as technologies that belong to the following three categories based on the underlying efficiency improvement caused by the technology group: (1) informational tech- nologies, which leverage big data and analytics (e.g., cloud computing, big data analytics, and machine learning); (2) operational technologies, which replace labor by combining data with automation (e.g., Internet of Things (IoT), 3D printing, and smart drones); and (3) transactional technologies, which match supply and demand such as in digital platforms and distributed ledger technologies. Ciffolilli and Muscio (2018) classifies I4 technology into eight categories based on expert peer reviews that focus on the input of the R&D process. The categories include (1) advanced man- ufacturing solutions, (2) additive manufacturing, (3) augmented reality, (4) simulation between interconnected machines that optimize processes, (5) horizontal and vertical integration tech- nologies that integrate information within the value chain, (6) the industrial internet and cloud that support multidirectional communication between production processes and products, (7) cyber-security that secures network operations, and (8) big data and analytics that optimize prod- ucts and processes. Using patent data, Balland and Boschma (2021) focus on the output of R&D activities and catego- rize I4 technology into 10 categories based on the Cooperative Patent Classification (CPC) code of the OECD-REGPAT database: (1) additive manufacturing; (2) artificial intelligence; (3) augmented reality; (4) autonomous robots; (5) autonomous vehicles; (6) cloud computing; (7) cybersecuri- ty; (8) quantum computers; (9) machine tools; and (10) system integration. CPC provides one of the most precise technological classifications broken down into around 250,000 categories. To identify the patents of I4 technologies, they check the CPC code of patents and reconstruct cate- gories indirectly by combining sub-categories. Furthermore, they analyze the abstracts of patent data in case these categories do not allow them to identify the I4 technology. Considering that this research is interested in the output of firms’ R&D activities, the definition and classification of Balland and Boschma (2021) is followed. 14 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea 03 Data 15 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea 3. Data 3.1 Financial data This study focuses on Korean firms in the manufacturing sector2 listed in three stock markets; (i) the Korea Composite Stock Price Index (KOSPI), a major stock market that lists large companies, (ii) Korea Securities Dealers Automated Quotations (KOSDAQ), which lists promising small and medium-sized enterprises (SMEs) or venture companies, and (iii) the Korea New Exchange (KON- EX), a stock market for SMEs with lower listing thresholds. Considering that derivatives are the only excluded segment from the Korean capital market, most of the listed companies in Korea are included in this research. KisValue, a dataset from National Information & Credit Evaluation Inc. (NICE), is used to collect firms’ financial information. It provides a variety of information, including external audited firms’ general information (e.g., founding year, number of employees, listing or delisting dates) and fi- nancial information (e.g., income statement, cash flow statement, valuations). Key financial ratios such as debt and profit ratios can be calculated using the KisValue dataset. 3.2 Patent data The European Patent Office (EPO) Worldwide Patent Statistical Database (PATSTATS) was used to obtain the patent dataset. PATSTATS was created by the EPO at the request of the OECD, and it is updated twice a year (Kang and Tarasconi, 2016). This study uses the Spring 2021 edition, which covers years from 1984 to 2021. PATSTATS covers more than 90% of the world’s patent authorities, including the Korean Intellectual Property Office (KIPO), the United States Patent and Trademark Office (USPTO), and the European Patent Office (EPO). The data contain comprehensive informa- tion on each patent, including applicants, inventors, publications, citations, filing country, filing date, registration status, and CPC codes. In this paper, the CPC code is used as a proxy for tech- nology. Merging the patent with financial data was challenging as the patent applicant was not a unique identifier. Some companies used non-unified company names, and some changed their compa- 2 According to the World Bank, value added by the manufacturing sector as percent of total GDP was 25 percent in 2021. 16 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea ny names, and for this reasons, one applicant had a variation of written names. For this reason, matching a patent with its company owner is not always accurate. Accordingly, Hall et al. (2001), Thoma et al. (2007), Julius and De Rassenfosse (2014), and He et al. (2018) try to standardize the names of applicants, with Kim et al. (2016), Lee et al. (2019), and Kang et al. (2019) focusing on Korean firms. This study uses the OECD Harmonised Applicant Names (HAN) database, which is based on text-matching algorithms such as the one presented by Kang et al. (2019). The HAN database provides a unique firm identifier called the HAN-ID that harmonizes the names of ap- plicants in different countries. However, there are still mismatches and errors in the OECD HAN, and no changes in the company name are considered (Kang et al., 2019). This study, therefore, unifies both mismatches in the HAN ID and errors in applicants’ names to obtain the name of a representative applicant. Next, this representative applicant’s name is matched with KisValue data that provide financial information. The result is a unique, unbalanced panel dataset of listed Korean manufacturing companies. A total of 1,196 manufacturing firms were matched with various applicant names within PATSTATS, and therewith, 388,454 patent applications from 2005 to 2018. 3.3 NTIS data Government can support firm’s R&D activities either through direct support, such as grants or matching grants to target firms, or through indirect support, such as R&D tax incentives. Informa- tion on whether firms that conducted R&D activities in I4 technologies received direct government support are used for the analysis. The National Science Technology Information Service (NTIS) database, managed by the Korean Ministry of Science and ICT (MSIT), allows us to assess the im- pact of public support on firms' knowledge activities in industry 4.0 technologies. This study only considers whether direct government support was allocated to firms, without considering other information, such as the amount of support or the characteristics of government-funded projects. Among various proxies that may show the outcome of a project (success or failure of the project, journal publications, patents filed, etc.), information on applied or granted patents is chosen as the outcome of government-supported projects. Among the 696,293 applied or granted patents from 99,482 government-funded projects between 2007 and 2018, 568,447 patents were selected for the analysis.3 The year a patent was granted is considered the starting year of the project. A total of 34,947 cases were found where firms developed new technologies based on the CPC code from 2007 to 2018. 3 When there are multiple contributors to the outcome of a government project, each contributor claims and reports their share of the contribution. An organization was regarded as owning the patent when the report on the government-funded project claimed that the contribution of the organization was above 50%. 17 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea 04 Methodology 18 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea 4. Methodology 4.1 Measuring technological relatedness A measure of relatedness is introduced to estimate the proportion of related technologies already existing in a firm (Hidalgo et al., 2007; Kogler et al., 2013; Boschma et al., 2014; Gao et al., 2021; Jun et al., 2019; Kim et al., 2021). First, the CPC codes of each patent and firm are connected by building a CPC code–firm bipartite network in which the weight of the link is the number of CPC codes possessed by the firm. Every patent requires one or more CPC codes to classify its technol- ogy. All the CPC codes in one patent are examined, instead of using the representative CPC code. For example, if firm i obtains a patent at time t, where the CPC codes are A, B, and C according to the KIPO, it is regarded as meaning the company developed all three technologies at time t. In a family of patents, although patents are identical, a new CPC code (e.g., technology D) could be added following the request of an examiner when filing the patent at another patent office (e.g., USPTO). After considering a patent granted by (for example) the KIPO or USPTO as one family pat- ent, all non-overlapping technology areas (CPC codes A–D) are synthesized and regarded as the technology of the corresponding patent. Next examined are co-occurrence of CPC codes within patents of the same firm, and then, the proximity between technologies α and β is estimated by following the method of Hidalgo et al. (2007). Proximity (φα,β,t ) indicates the minimum value of the pairwise conditional probability that two technologies have a comparative advantage together within the same firm: φα,β = min {Pr(RTAα|RTAβ),Pr(RTAβ|RTAα)} (1) where RTA stands for “revealed technological advantage”: Pi,α,t (2) ∑α Pi,α,t RTAi,α,t = ∑i Pi,α,t ∑i ∑α Pi,α,t in which Pi,α,t is the number of patents related to technology α possessed by firm i at time t (Bal- assa, 1965). RTA indicates the comparative advantage of firm i in technology α by measuring whether the share of technology α in its total technologies is greater than that of the average firm. We say firm i de- velops a comparative advantage in technology α at time t when its RTAi,α,t transitions from RTAi,α,t <1 to RTAi,α,t ≥1. Considering that previously developed technologies require some time (often more 19 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea than three years) to affect the next generation of new related technologies, the three years before the development of a new technology also are examined. When defining a firm’s development of technology α at time t, that technology’s RTA value is below 1 at time t − 1, over 1 at time t, and it holds its value above 1 at t + 1 and t + 2 (Bahar et al., 2014). Lastly, the proximity among technologies is used to aggregate the related technologies of a firm— termed the density (ωi,α,t ) of the related technology of firm i. Density measures how a firm’s exist- ing technology portfolio is related to technology α from among all its technologies. Formally, the density of the related technology for technology α of firm i at time t is given by ∑β φα,β,t Ui,β,t ωi,α,t = (3) ∑β φα,β,t where φα,β,t is the proximity between technologies α and β, and Ui,β,t is 1 if firm i has an RTA in tech- nology β in year t (RTAi,α,t ≥1), but is 0 otherwise. 4.2 Measuring technology complexity Along with the relatedness measure, a complexity measure is also used. Based on their obser- vation that some countries export various kinds of products, and others export only a few kinds of products, Hidalgo and Hausmann (2009) ask, “What are the distinguishing characteristics of these two different kinds of countries and their exported products” (Hidalgo and Hausmann, 2009). To answer this, they introduce a methodology called the Method of Reflection (MOR). From a bipartite network, this methodology can reduce the information of one dimension (for example, producing a sophisticated product) while preserving the rest of the information of the opposite di- mension (for example, products that are produced by a country whose economy is sophisticated). As a result, the MOR independently provides two types of symmetric information: (1) about actors (in the country, region, city, and firm, among others), and (2) about activities (industry, product, technology, and occupation, among others) that constitute a bipartite network. First, the Location Complexity Index (LCI) measures how much comparative advantage a location has in various economic activities, reflecting information on the complexity of economic activities. According to the work of Hidalgo and Hausmann (2009), countries with diversified export prod- uct portfolios also tend to have a high LCI level. That is because countries that are more likely to export several kinds of products can produce more complicated products that cannot easily be manufactured by many countries. Countries with few kinds of export products tend to have a low LCI because they produce a few products that are ubiquitous and that can be manufactured by many other countries. Second, the Economic Complexity Index (ECI) measures how frequently a certain activity is participated in at various locations while preserving the location’s diversification information. We can intuitively understand that economic activities are more sophisticated when only a small number of countries can develop and possess them. Contrarily, if every country can participate in, and practice, a certain economic activity, that economic activity would have a low level of difficulty; in other words, it has little complexity. 20 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea By using MOR, we can reduce two different (but connected) pieces of information (location or economic activity) from the bipartite network to information of only one dimension by recursively calculating the average of diversification and ubiquity. The following equations are the general forms of LCI and ECI. Because we are interested in technology as an economic activity and in firms for their locations, the technology expression is unified with subscript t and the expression for firms takes subscript f: 1 Complexity of firm: Kf,N = ∑ M K (4) Kf,0 t f,t t,N-1 1 Complexity of technology: Kt,N = TCIt,N = ∑ M K (5) Kt,0 f f,t f,N-1 where Mf,t is a matrix composed of firms and technologies with an RCA above 1, which is calculated from equation (2). By using MOR, we can calculate the Complexity of a firm and the Complexity of a technology by averaging out previous characteristics levels in neighboring nodes that are itera- tively positioned in the opposite dimension. When we set two different nodes as a starting point and a destination in the technology dimension, there exist numerous routes stopped by nodes of an opposite dimension (a space composed of firms). The value of Complexity can have different meanings based on the iteration number, N(≥0). N indi- cates how many times it iterates through nodes of different dimensions to reach a destination. The initial condition of complexity, starting with N = 0, simply means the degree of a node in a network, and the number of links connecting it with other nodes within the opposite dimension. Diversification of firm: Kf,0 = ∑t Mf,t (6) Ubiquity of firm: Kt,0 = ∑f Mf,t (7) Kf,0 and Kt,0, respectively, denote the technological diversification of a firm (the number of technol- ogies developed by the firm) and the ubiquity of a technology (the number of firms that develop a certain technology). As N increases, we can average them out so that the values for Complexity of a firm and for Complexity of a technology converge. Because it iterates in increments of N until we cannot get additional information, interactions were stopped at the 20th run following the rule of thumb of this methodology. 4.3 The empirical model To examine the factors that affect a firm’s diversification to new technologies, the following multi- variate logit model was constructed: Ui,α,t+2 = β0 + β1ωi,α,t + β2TCIα,t + β3Govi,α,t + β4Firmi,t + β5Technologyα,t + θt + εi,α,t (8) Ui,α,t+2 is 1 when firm i successfully enters new technology α at time t+2; otherwise, it is 0. The main explanatory variables, ωi,α,t and TCIα,t represent the density of related technologies and the tech- 21 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea nological complexity at time t, respectively. Govi,α,t indicates whether firm i received government R&D support for developing technology α. The second half of the equation (8) consists of control variables: Firmi,t includes (i) Agei,t (the ten- ure of firm i since its inception); (ii) num_employeei,t, which controls for the size effect of the firm; (iii) Profit_ratioi,t and Debt_ratioi,t, which are profit-to-sales and total liabilities to total assets in year t, respectively. Profit_ratioi,t and Debt_ratioi,t control for the quantitative aspects of the firm (representing a capital structure that can capture its value and expected growth); and Number of RTAi,t represents all technologies that offer an advantage within firm i (this value reflects the quan- titative aspect of a technological capability, rather than the qualitative aspect). The other vector, Technologyα,t, includes the number of competitors (num_competitorα,t) in order to examine the firm’s technological environment. This variable is the number of firms that have a comparative advantage in technology α in the industry to which the firm belongs. This variable captures two as- pects: how many learning opportunities exist in the industry, and how many competitors for that technology exist in the industry. Finally, year-fixed effects (θt) are added to control for nationwide time trends. εi,α,t is the error term. Table 1 below summarizes the variables used for the empirical model. Table 1. Description of the variables Variables Description Ui,α,t+2 Binary variable depending on whether firm i successively entered new technology α at time t+2 ωi,α,t Density of related technologies TCIα,t Technological complexity Govi,α,t Binary variable depending on whether firm i received government R&D support for developing technology α Agei,t Tenure of firm i since its inception num_employeei,t Number of employees Profit_ratioi,t Profit ratio Debt_ratioi,t Debt ratio Number of RTAi,t Number of technologies that offer revealed technological advantage within firm i num_competitorα,t Number of competitors in technology α θt Year-fixed effects 22 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea 05 Results 23 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea 5. Results 5.1 Technology space of Korean manufacturing firms To provide context for the empirical analysis that examines the factors that affect Korean firms’ knowledge activities in I4 technologies, it is important to visualize the technology space of Korea’s manufacturing firms. Figure 1 (A) depicts the position of 10 different I4 technologies of Korean manufacturing firms in the technology space. The ten I4 technologies are highlighted in green, and their related technologies in gray. As seen in the figure, all ten I4 technologies are located at the core or in an adjacent cluster at the upper-left corner of the space. The size of the node is pro- portional to the number of patents belonging to each CPC code. Among the ten I4 technologies, quantum computers, cloud computing, and cybersecurity are the major patented technological fields by the Korean manufacturing firms. (See the Appendix for the technology space covering all CPC codes.) Figures 1 (B) and (C) explore the relationships among the ten I4 technologies. Proximity between technologies is represented by the thickness of the edge in Figure 1 (B) and by colors of the heat map in Figure 1 (C). We can see there is a high level of proximity between cybersecurity and aug- mented reality. The second-highest level can be found between autonomous robots and system integration. In addition, the relationship among cybersecurity, augmented reality, and cloud com- puting exhibits high proximities. This indicates that technologies in cybersecurity and augmented reality are likely to share common technological capabilities of Korean firms, resulting in patenting in tandem. Likewise, high proximity among technologies in cybersecurity, augmented reality, and cloud computing implies they require similar technological capabilities. 24 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea Figure 1: Technology space of Korean manufacturing firms: Industry 4.0 technologies A. Position of I4Ts in Technology Space Cloud computing Quantum computers Autonomous robots System integration Artificial intelligence Cybersecurity Augmented reality Quantum computers Machine tools Machine tools Additive manufacturing Autonomous vehicles b. relatedness between i4ts c. i4ts hierarchical matrix Autonomous vehicles 1.0 Additive manufacturing 0.9 Artificial intelligence Quantum computers 0.8 Machine Augmented reality System tools Φ ≥ 0.7 0.7 Autonomous robots integration 0.6 ≥ 0.6 Autonomous vehicles 0.5 Cloud computing Artificial Cloud ≥ 0.5 0.4 Cybersecurity intelligence computing 0.3 Machine tools ≥ 0.4 0.2 Quantum computers 0.1 Autonomous robots Φ ≤ 0.4 System integration 0.0 Additive manufacturing Artificial intelligence Augmented reality Autonomous robots Autonomous vehicles Cloud computing Cybersecurity Machine tools Quantum computers System integration Link weight Cybersecurity Augmented reality (Proximity) Additive manufacturing Source: Author’s Note: (A) highlights the position of the ten I4 technologies in the technology space (2010-2019). The thickness of the links in (B) indicates the proximity between pairs of technologies. The colors in (C) indicates the level of relatedness between pairs of technologies. 25 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea Next, to examine the level of complexity of I4 technologies, patent data covering the 12 years from 2007 to 2018 are aggregated, and the ranking of complexity is calculated using equation (5). Fig- ure 2 shows the complexity ranking of I4 technologies and how they evolved over the 12 years. The aggregated number of patents for each I4 technology is plotted on the x-axis after log transforma- tion. On the y-axis, the average complexity ranking for each I4 technology is plotted. How com- plexity ranking and the number of the patents evolved over the period is shown over three phases: period 1 (2007-2010), period 2 (2011-2014), and period 3 (2015-2018).4 The ten I4 technologies are in different colors and the shapes of nodes represent different time periods. A large value on the y-axis implies that the complexity of the technology is low, or in other words more ubiquitous with many other firms already having developed it. For example, from period 1 to period 3, artificial intelligence became more ubiquitous, although the total number of patents remains low. Total applied patents for eight I4 technologies increased from period 1 to period 3, except for quantum computers and machine tools. In the first period (2007-2010), the I4 technology most commonly possessed by Korean manufacturing firms was quantum computers, and the least com- mon was artificial intelligence. A small number of patents in artificial intelligence were owned by a relatively small number of firms, while relatively more patents in quantum computing were de- veloped by a larger number of firms. The least ubiquitous technology in period 3 was autonomous vehicles because this field was nascent at the time, with only a few specialized firms developing technologies related to autonomous vehicles. 4 The global financial crisis happened in period 1, and Period 2 was the recovery period from the crisis. In period 3, a seminal paper in AI was introduced (LeCun et al., 2015), and the phenomena associated with the Fourth Industrial Revolution started to be vividly observed. 26 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea Figure 2: Complexity ranking and number of patents of I4 technologies 500 Average ranking of complexity 400 300 200 100 0 4 6 8 10 log (number of patents) Time I4 technology Period 1 (2007-2010) Additive manufacturing Period 2 (2011-2014) Artificial intelligence Period 3 (2015-2018) Augmented reality Autonomous robots Autonomous vehicles Cloud computing Cybersecurity Machine tools Quantum computers System integration Source: Author’s Note: The x-axis is the total number of patents (in log form) and the y-axis is average rankings for complexity of CPCs within each I4 technology. Table 2 shows the names of the top three companies for each of the I4 technologies based on the number of patents. Well-known Korean conglomerates such as Samsung, LG, SK, and POSCO were found to be the key players in the I4 technologies space. 27 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea Table 2: Top three companies in terms of numbers of patents in I4 technologies Technology Artificial Addictive Autonomous Cloud rank intelligence manufacturing robots Cybersecurity computing 1 Samsung SDI POSCO LG Electronics Samsung Electronics Samsung Electronics 2 Samsung Electronics Hyundai Steel Samsung Electronics Samsung SDI Samsung SDI 3 LG Electronics POSCO Chemtech Samsung SDI LG Electronics LG Electronics Technology Quantum Autonomous Augmented System rank competers vehicles reality Machine tools integration 1 Samsung SDI LG Electronics Samsung SDI Doosan Infracore LG Electronics 2 Samsung Electronics Samsung Electronics Samsung Electronics Samsung Electronics LS Industrial Systems 3 SK hynix Samsung SDI LG Electronics Samsung SDI Samsung Electronics Source: Author’s calculation using the European Patent Office (EPO) Worldwide Patent Statistical Database (PATSTATS) 28 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea Firms that developed I4 technologies tended to be slightly younger than other firms. From 2008 to 2014, there were 879 firms with at least one patent in I4 technologies from among all 1,180 firms. When only looking at 2014, 683 firms had at least one patent in I4 technologies from among all 1,059 firms. In 2014, the average age of I4 technology firms was 25.56 years, while that of all firms that owned any kind of patent was 26.80 years. These differences are significant at the 95% confidence interval.5 I4 firms were not significantly larger than other firms. From 2008 to 2014, the average number of employees in firms that owned I4 technologies was 740, whereas that of all firms was 630 employ- ees. In 2014, the average number of employees in I4 technology firms was 1,175, while that of all firms in the sample was 906. These differences in size are not significant at the 95% confidence interval.6 5.2 Regression results This section examines factors that affect Korean manufacturing firms’ technological diversification and their transition into industry 4.0 technologies. The main empirical results are summarized in Table 3, where all control and fixed variables are included (Appendix 2 provides summary statistics and detailed regression results). The key findings are: (i) Technological relatedness has a positive and significant impact on firms’ diversification. An increase in technological relatedness, ωi,α,t , by 1 unit results in a 47.6% increase in the odds of developing technology α and a 96.9% increase for the I4 technologies, provided other variables remain fixed. This finding aligns with the previous literatures that argue that firms are more likely to develop technologies that are related to the technologies they already have. Result from Table 3 Column (2) similarly shows that firms are likely to enter a new I4 technology when they have related technology base. (ii) Result from Table 3 column (1) shows that technology complexity has no significant impact on firms’ diversification. However, a closer examination through split samples by the level of related- ness reveals that the relationship between technological complexity and diversification is stron- ger for firms with a high level of technological relatedness. Table 4 column (3) to (6) shows that technological complexity has a positive and significant impact on diversification only when firms have high relatedness. Controlling all variables, an increase of 1 unit in technological complexi- ty (TCIα,t ) enhances the odds of technological diversification by 50.0% when firms possess high technological relatedness. This result implies that firms need to accumulate related technologies in order to develop more complex and non-ubiquitous technologies. Balland et al. (2019) explain 5 The p-value from a t-test that asked whether the average age of firms with I4 technologies was less than that of all firms was 0.017 at the 95% confidence interval. 6 The p-value of the t-test that asked if the average size of firms with I4 technologies was greater than that of all firms was 0.1645 at the 95% confidence interval. 29 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea it as a diversification dilemma, meaning that a firm cannot develop a more difficult technology even though it is more attractive. (iii) Government support has a positive and significant impact on diversification for both samples in Table 3. An increase of 1 unit in government support, Govi,α,t enhances the odds of technological development by 64.4% for all technologies and 79.8% for industry 4.0 technologies. However, as seen in Table 4, this relationship is significant only for firms with a low level of relatedness, per- haps because firms with high level of relatedness can diversify independently even without public support. This finding held true when we examined the regression result covering I4 technologies in Table 5 column (3) and (4). Government support has a significant impact on diversification only for firms that possessed low technology relatedness. This suggests that public programs are likely more effective when targeted at firms with low levels of technological relatedness. Table 3: Summary of the regression results (1) All technology (2) I4 technologies ωi,α,t 0.476*** 0.969*** (0.014) (0.340) TCIα,t 0.351*** (0.061) Govi,α,t 0.644*** 0.798*** (0.051) (0.238) Agei,t 0.050*** -0.038 (0.008) (0.084) num_employeei,t 0.086*** 0.367 (0.015) (0.248) num_competitorα,t 0.277*** 0.522*** (0.005) (0.120) num_RTAi,t 0.009 -1.014*** (0.038) (0.386) Profit_ratioi,t -0.001 0.025 (0.003) (0.031) Debt_ratioi,t 9.652*** -4.164 (1.666) (14.838) Time fixed effect Yes Yes Observations 5,320,088 17,937 Pseudo R2 0.1274 0.1193 *p < 0.1; **p < 0.05; ***p < 0.01 Note: Technological complexity (TCIα,t ) is not included in the independent variables, as the sample consists of only I4 technolo- gies where the level of complexity is similar. 30 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea Table 4: Regression result covering all technologies broken down by the level of relatedness (1) High (2) Low (3) High (4) Low (5) High (6) Low relatedness relatedness relatedness relatedness relatedness relatedness TCIα,t 0.362*** -0.347 0.594*** -0.216 0.500*** -0.213 (0.076) (0.330) (0.089) (0.337) (0.093) (0.358) Govi,α,t 0.253*** 3.817*** -0.020 2.691*** -0.022 2.685*** (0.068) (0.457) (0.072) (0.720) (0.072) (0.721) Agei,t 0.140*** -0.182*** 0.139*** -0.191*** (0.013) (0.067) (0.013) (0.068) num_employeei,t -0.028 0.844*** -0.042** 0.844*** (0.020) (0.190) (0.021) (0.191) num_ 0.227*** 0.480*** 0.234*** 0.475*** competitorα,t (0.006) (0.041) (0.006) (0.042) num_RTAi,t 0.850*** 0.820*** 1.150*** 0.704*** (0.053) (0.224) (0.062) (0.246) Profit_ratioi,t -0.007 0.024 -0.010** 0.034 (0.005) (0.018) (0.004) (0.021) Debt_ratioi,t 4.845** -36.148** 7.841**** -35.156** (2.399) (15.153) (2.455) (15.327) Time fixed effect No No No No Yes Yes Costant -5.964*** -6.755*** -8.437*** -6.059*** (0.341) (1.481) (0.410) (1.505) Observations 617,932 617,929 532,008 532,006 532,008 532,006 Pseudo R2 0.0005 0.0100 0.0231 0.0604 0.0285 0.0654 *p < 0.1; **p < 0.05; ***p < 0.01 31 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea Table 5: Regression result covering I4 technologies broken down by the level of relatedness (1) All firms (2) All firms (3) High density (4) Low density with I4T with I4T ωi,α,t 0.286*** 0.969*** (0.054) (0.340) Govi,α,t 0.544** 0.798*** 0.814*** 1.514*** (0.221) (0.238) (0.263) (0.541) Agei,t -0.038 -0.101 -0.025 (0.084) (0.095) (0.163) num_employeei,t 0.367 0.315 2.692*** (0.248) (0.268) (0.829) num_competitorα,t 0.522*** 0.519*** 1.435*** (0.120) (0.106) (0.319) num_RTAi,t -1.014*** -0.189 -0.147 (0.386) (0.158) (0.331) Profit_ratioi,t 0.025 0.017 0.150* (0.031) (0.033) (0.083) Debt_ratioi,t -4.164 -3.724 -32.556 (14.838) (17.497) (31.216) Time fixed effect Yes Yes Yes Yes Observations 20,244 17,937 8,968 8,968 Pseudo R2 0.0567 0.1193 0.0786 0.2416 32 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea 06 Conclusion and discussion 33 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea 6. Conclusion and discussion This research leverages patent data and relatedness analysis to analyze factors that have contrib- uted to Korean manufacturing firms’ technology diversification patterns, particularly diversifica- tion into the I4 space. First, it provides a visualization of the Korean manufacturing firms’ product space and the relative positioning of I4 technologies. In terms of the absolute number of patents owned by Korean manufacturing firms, quantum computers, cloud computing, and cybersecurity appear as dominant I4 technologies. From the analysis of Korean manufacturing firms’ technol- ogy space, we show that cybersecurity and augmented reality technologies have the closest re- lationship, implying those two technologies share common capabilities. Moreover, autonomous robots and system integration show high proximity to each other. Despite the increasing number of patents on autonomous vehicles, the technology turns out to be less ubiquitous than others. This implies that only a few companies have the capabilities to develop technologies related to autonomous vehicles. For the most part, the Korean firms that published patents in I4 technology space were mostly large and well-established firms. A few key findings could be emphasized. Both technological relatedness and technological com- plexity have a positive, significant impact on firms’ diversification. Technological complexity is more important when firms have a high level of related technologies. Similarly, firms with a high level of related technologies are more likely to engage in new I4 technologies. This result suggests that in order to transition into more complex and non-ubiquitous technologies, accumulating re- lated technologies is needed. Direct government support increases the likelihood that a firm diversifies into I4 technologies. Public support is more effective for firms that possess low technological relatedness, and less so for firms that already have enough accumulated related technologies. This may imply that public programs aiming to promote activities in the I4 space are better off targeting support towards firms with low levels of technological relatedness. This research has a few limitations. First, while this research focuses on the firm’s innovation activ- ity output, which is patenting, various R&D input factors such as human capital, financial resourc- es, local infrastructure, and macroeconomic conditions also affect firms’ technological diversi- fication. Second, while this research examined the effect of direct government support, indirect government supports, which include tax incentives for R&D, loan guarantees also play a significant role in technological innovation. For example, the Korean government has been raising the corpo- rate tax break for SMEs and large corporations through Enforcement Decree of the Restriction of 34 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea Special Taxation Act. In this Act, most of the industry 4.0 technologies are listed as national stra- tegic technologies, and firms that invest in selected national strategic technologies can receive tax benefits. Third, in this research the government support variable is treated a dummy variable, depending on whether the firm participated in the government-funded project. Disaggregating this variable through leveraging granular data on the amount and duration of the support and also distinguishing different types of direct government support (e.g., grants, matching grants, innovation vouchers, loans) could help identify the ideal type and level of government support. Fourth, firms may file patents strategically as a defensive tactic or to block competitors. This can inflate patenting activity, especially of manufacturing firms, and this research does not distin- guish between truly innovative and strategic patenting. Fifth, the micro-mechanism of relatedness among different technologies is not empirically examined. According to Boehm et al. (2022), input capabilities, which can stem from human capital, institutions, and tacit knowledge already within the firm, determine the evolution of the product space of a multi-product firm and consequently decide the pattern of technological diversification. Last, the paper is solely based on the findings from the Korean manufacturing sector, excluding firms in the service sector. 35 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea References 36 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea References Acemoglu, D., Johnson, S. and Robinson, J.A., Boschma, R., 2017. Relatedness as driver of 2005. Institutions as a fundamental cause regional diversification: A research agenda. of long-run growth. Handbook of economic Regional Studies 51, 351–364. growth, 1, pp.385-472 Boschma, R., Balland, P.A., Kogler, D.F., Arrow, K. (1962) “Economic Welfare and 2014. Relatedness and technological change the Allocation of Resources to Invention”. in cities: the rise and fall of technological In Nelson, R., Ed. The Rate and Direction knowledge in US metropolitan areas from 1981 of Inventive Activity: Economic and Social to 2010. Industrial and Corporate Change 24, Factors. National Bureau of Economic 223–250. Research: Cambridge, MA., 609-626. Boschma, R., Iammarino, S., 2009. Related Bahar, D., Hausmann, R., Hidalgo, C.A., variety, trade linkages, and regional growth in 2014. Neighbors and the evolution of the italy. Economic Geography 85, 289–311. comparative advantage of nations: Evidence of international knowledge diffusion? Journal of Boschma, R., Minondo, A., Navarro, M., International Economics 92, 111 –123. 2013. The emergence of new industries at the regional level in spain: Aproximity approach Balassa, B., 1965. Trade liberalisation and based on product relatedness. Economic “revealed” comparative advantage 1. The Geography 89, 29–51. manchester school 33, 99–123. Breschi, S., Lissoni, F. and Malerba, F., 2003. Balland, P.A. and Boschma, R., 2021. Mapping Knowledge-relatedness in firm technological the potentials of regions in Europe to diversification. Research policy, 32(1), pp.69- contribute to new knowledge production in 87. Industry 4.0 technologies. Regional Studies, 55(10-11), pp.1652-1666. Chandler, A.D., 1990. Integration and Diversification as Business Strategies--An Balland, P.A., Boschma, R., Crespo, J., Rigby, Historical Analysis. Business and Economic D.L., 2019. Smart specialization policy in the History, pp.65-73. european union: relatedness, knowledge complexity and regional diversification. Chandler, A.D., 1993. The visible hand. Harvard Regional Studies 53, 1252–1268. university press. Balland, P.A., Rigby, D., 2017. The geography Chandler, A.D., 2009. Scale and scope: The of complex knowledge. Economic Geography dynamics of industrial capitalism. Harvard 93, 1–23. University Press. Boehm, J., Dhingra, S. and Morrow, J., 2022. Ciffolilli, A., Muscio, A., 2018. Industry 4.0: The comparative advantage of firms. Journal of national and regional comparative advantages Political Economy, 130(12), pp.000-000. in key enabling technologies. European Planning Studies 26, 2323–2343. 37 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea Cohen, W. Nelson, R. Walsh (2000) Hall, B.H., Jaffe, A.B., Trajtenberg, M., 2001. “Protecting Their Intellectual Assets: The NBER Patent Citation Data File: Lessons, Appropriability Conditions and Why U.S. Insights and Methodological Tools. Working Manufacturing Firms Patent (or Not)”. National Paper 8498. National Bureau of Economic Bureau of Economic Research Working Paper Research. 7552. Hausmann, R., Hidalgo, C.A., Bustos, S., Domini, G., 2022. Patterns of specialization Coscia, M., Simoes, A., 2014. The atlas of and economic complexity through the lens of economic complexity: Mapping paths to universal exhibitions, 1855-1900. Explorations prosperity. Mit Press. in Economic History 83, 101421. He, Z.L., Tong, T.W., Zhang, Y., He, W., 2018. Dosi, G., 1982. Technological paradigms Constructing a chinese patent database of and technological trajectories: a suggested listed firms in china: Descriptions, lessons, interpretation of the determinants and and insights. Journal of Economics & directions of technical change. Research Management Strategy 27, 579–606. policy, 11(3), pp.147-162. Hidalgo, C.A., 2021. Economic complexity Freeman, C., 2013. Economics of industrial theory and applications. Nature Reviews innovation. Routledge. Physics Frenken, K., Van Oort, F., Verburg, T., 2007. Hidalgo, C.A., Balland, P.A., Boschma, R., Related variety, unrelated variety and regional Delgado, M., Feldman, M., Frenken, K., economic growth. Regional Studies 41, Glaeser, E., He, C., Kogler, D.F., Morrison, 685–697. A., Neffke, F., Rigby, D., Stern, S., Zheng, S., Zhu, S., 2018. The principle of relatedness, Fritz, B.S., Manduca, R.A., 2021. The economic in: Morales, A.J., Gershenson, C., Braha, complexity of us metropolitan areas. Regional D., Minai, A.A., Bar-Yam, Y. (Eds.), Unifying Studies 55, 1299–1310. Themes in Complex Systems IX, Springer International Publishing, Cham. pp. 451–457. Gao, J., Jun, B., Pentland, A. Zhou, T., Hidalgo, C.A., 2021. Spillovers across industries Hidalgo, C.A., Hausmann, R., 2009. The and regions in china’s regional economic building blocks of economic complexity. diversification. Regional Studies, 1–16. Proceedings of the national academy of sciences 106, 10570–10575. Gerybadze, A., 2018. Industrial development strategies in asia: The influence of friedrich Hidalgo, C.A., Klinger, B., Barabási, A.L., list on industrial evolution in Japan, South Hausmann, R., 2007. The product space Korea and China, in: The Economic Thought of conditions the development of nations. Friedrich List. Routledge, pp. 223–244. Science 317, 482–487. Griliches, Z., 1998. Patent statistics as Huergo, E., Jaumandreu, J., 2004. How does economic indicators: a survey, in: R&D and probability of innovation change with firm age? productivity: the econometric evidence. Small Business Economics 22, 193–207. University of Chicago Press, pp. 287–343. Jaffe, A.B., 1986. Technological opportunity and spillovers of R&D: evidence from firms’ patents, profits and market value. 38 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea Julius, T., De Rassenfosse, G., 2014. Liao, Y., Deschamps, F., Loures, E.d.F.R., Harmonising and matching ipr holders at ip Ramos, L.F.P., 2017. Past, present and future australia. of industry 4.0-a systematic literature review and research agenda proposal. International Jun, B., Alshamsi, A., Gao, J., Hidalgo, C.A., journal of production research 55, 3609–3629. 2020. Bilateral relatedness: knowledge diffusion and the evolution of bilateral trade. Lu, Y., 2017. Industry 4.0: A survey on Journal of Evolutionary Economics, 1–31. technologies, applications and open research issues. Journal of industrial information Kang, B., Tarasconi, G., 2016. Patstat revisited: integration 6, 1–10. Suggestions for better usage. World patent information 46, 56–63. Lundvall, B.A., 1992. National innovation systems: towards a theory of innovation and Kang, T., Baek, C., Lee, J.D., 2019. Effects of interactive learning. Londres: Pinter. knowledge accumulation strategies through experience and experimentation on firm Mazzucato, M., 2015. Building the growth. Technological Forecasting and Social entrepreneurial state: A new framework for Change 144, 169–181. envisioning and evaluating a mission-oriented public sector. Levy Economics Institute of Bard Kim, J., Lee, C.Y., Cho, Y., 2016. Technological College Working Paper, (824). diversification, core-technology competence, and firm growth. Research Policy 45, 113–124. Neffke, F., Henning, M., Boschma, R., 2011. How do regions diversify over time? industry Kim, L., Nelson, R.R., Nelson, R.R., 2000. relatedness and the development of new Technology, learning, and innovation: growth paths in regions. Economic Geography Experiences of newly industrializing 87, 237–265. economies. Cambridge University Press. Nuvolari, A., 2019. Understanding successive Kim, S.H., Jun, B. and Lee, J.D., 2021. industrial revolutions: A “development block” Technological relatedness: how do firms approach. Environmental Innovation and diversify their technology? Center for Open Societal Transitions 32, 33–44. Science. Perez, C., 2003. Technological revolutions and Kogler, D.F., Rigby, D.L., Tucker, I., 2013. financial capital. Edward Elgar Publishing. Mapping knowledge space and technological relatedness in us cities. European Planning Petralia, S., Balland, P.A., Morrison, A., Studies 21, 1374–1391. 2017. Climbing the ladder of technological development. Research Policy 46, 956–969. LeCun, Yann, Yoshua Bengio, and Geoffrey Hinton. “Deep learning.” nature 521, no. 7553 Popkova, E.G., Ragulina, Y.V., Bogoviz, A.V., (2015): 436-444. 2019. Industry 4.0: Industrial revolution of the 21st century. Springer. Lee, J., Lim, H., Kim, S., Song, K., Jung, J.Y., 2019. Korea patent data project (KoPDP): Rifkin, J., 2011. The third industrial revolution: Contents and methods. Available at SSRN how lateral power is transforming energy, the 3404698. economy, and the world. Macmillan. 39 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea Rigby, D.L., 2015. Technological relatedness Wade, R., 2018. Governing the market. and knowledge space: Entry and exit of us Princeton University Press. cities from patent classes. Regional Studies 49, 1922–1937. Weaver, W., 1991. Science and complexity, in: Facets of systems science. Springer, pp. Rockett, K., 2010. Property rights and 449–456. invention. In Handbook of the Economics of Innovation (Vol. 1, pp. 315-380). North- Wohl, I., 2020. The Method of Reflections Holland. and US Occupational Employment. Office of Industries, US International Trade Saviotti, P.P., Pyka, A., 2013. From necessities Commission. to imaginary worlds: Structural change, product quality and economic development. World Bank, 2020. Europe 4.0: Addressing the Technological Forecasting and Social Change digital dilemma. 80, 1499–1512. Saviotti, P.P., Pyka, A., Jun, B., 2016. Education, structural change and economic development. Structural Change and Economic Dynamics 38, 55–68. Stephan, P. E. (1996). The economics of science. Journal of Economic literature, 34(3), 1199-1235. Stojkoski, V., Utkovski, Z., Kocarev, L., 2016. The impact of services on economic complexity: Service sophistication as route for economic growth. PloS one 11, e0161633. Teece, D.J., 1980. Economies of scope and the scope of the enterprise. Journal of economic behavior & organization, 1(3), pp.223-247. Teece, D.J., 1982. Towards an economic theory of the multiproduct firm. Journal of Economic Behavior & Organization, 3(1), pp.39-63. Teece, D. J., Rumelt, R., Dosi, G., & Winter, S., 1994. Understanding corporate coherence: Theory and evidence. Journal of economic behavior & organization, 23(1), 1-30. Thoma, G., Torrisi, S., et al., 2007. Creating powerful indicators for innovation studies with approximate matching algorithms: a test based on PATSTAT and Amadeus databases. Università commerciale Luigi Bocconi. 40 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea Appendices 41 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea Appendix 1 Figure 3 is the technology space of Korea, based on patent data from 2010 to 2019. Each node represents a technological category expressed with a CPC code, while links represent the prox- imity between technologies. The radius of a node can be viewed as broadly indicative of the total number of patents in that CPC code. The technology space for Korea exhibits a core/periphery structure such that the subcategories of physics and mechanical engineering, lighting, heating, and weapons are likely located at the center, while those of human necessities, textiles, and paper are at the periphery. Simultaneously, we see a coherent cluster in the upper left-hand corner that consists of subcategories for electricity, physics, mechanical engineering, lighting, heating, and weapons. This cluster is from Samsung, which owns 31% of all Korean patents and specializes as an electronics and chip maker. Figure 3: Technology space of Korean manufacturing firms from 2010 to 2019 A. Technology Space Node color (cpc category) A: Human necessities B: Performing operations; transporting C: Chemistry; metallurgy D: Textiles; paper E: Fixed constructions F:  Mechanical engineering; lighting; heating; weapons; blasting engines or pumps b. hierarchical matrix G: Physics H: Electricity 1.0 Y:  General tagging of new technological developments; 0.9 general tagging of cross-sectional technologies spanning over several sections of the IPC; 0.8 technical subjects covered by former USPC cross- reference art collections (XRACs) and digests 0.7 0.6 Link weight (Proximity) 0.5 Φ ≥ 0.7 ≥ 0.6 ≥ 0.5 ≥ 0.4 Φ ≤ 0.4 0.4 0.3 0.2 Source: Kim et al. (2021) Note: The color of a node classifies it by one-digit CPC code, 0.1 and its radius is proportional to the number of patents belong- 0.0 ing to each CPC code. 42 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea Appendix 2 This appendix provides detailed information on the relationships between the main independent variables (technological complexity, technological relatedness and government support) and technological diversification. To estimate the effect of technological relatedness and complexity described in equation (8), Box-Cox transformation was applied to all variables (except for the binary variables) and to TCIα,t based on the year. TCIα,t was rescaled to 0-1 and log transformed. Table 6 shows the summary statistics of the variables. Table 6: Summary statistics Statistic N Mean St.Dev. Min Max Ui,α,t+2 6,179,317 0.002 0.050 0 1 ωi,α,t 6,179,317 0.007 2.176 -8.712 6.802 TCIα,t 6,179,317 4.510 0.157 2.145 4.615 Govi,α,t 6,179,317 0.004 0.062 0 1 Agei,t 5,320,088 -0.00004 1.336 -5.036 4.113 num_employeei,t 5,320,088 0.000 0.636 -4.145 2.729 num_competitorα,t 5,320,088 0.007 2.064 -5.087 5.395 num_RTAi,t 5,320,088 0.00001 0.791 -1.552 2.193 Profit_ratioi,t 5,320,088 0.201 3.026 -47.713 43.374 Debt_ratioi,t 5,320,088 0.0000 0.005 -0.013 0.038 Source: Author’s Note: all variables underwent Box Cox transformation Table 7 is the correlation table for all the variables. We can see that ωi,α,t is highly correlated with the number of other kinds of a firm’s technologies. Interestingly, ωi,α,t is not highly correlated with a firm’s basic characteristics, such as age (Agei,t), size (num_employeei,t), or financial structure (Profit_ratioi,t and Debt_ratioi,t ), but is highly correlated with the number of revealed technologi- cal advantage (Num_RTAi,t). To avoid the multicollinearity problem among independent variables, the variance inflation factor (VIF) was measured, and the value was not very high (1.005). Another test was on the VIF for Agei,t and num_employeei,t , and the result was 1.063, which implies that all the variables can be simultaneously considered in the regression model. 43 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea Table 7: Correlation table num_ num_ Debt_ Ui,α,t+2 ωi,α,t TCIα,t Govi,α,t Agei,t employeei,t competitorα,t num_RTAi,t Profit_ratioi,t ratioi,t Ui,α,t+2 1 ωi,α,t 0.064 1 TCIα,t -0.001 0.004 1 Govi,α,t 0.022 0.085 0.0005 1 Agei,t 0.022 0.207 -0.00001 0.028 1 num_employeei,t 0.036 0.364 0.0002 0.055 0.465 1 num_competitorα,t 0.035 0.255 -0.046 0.010 0.0000 0.000 1 num_RTAi,t 0.054 0.865 0.000 0.079 0.223 0.372 0.000 1 Profit_ratioi,t -0.001 -0.019 0.028 -0.001 -0.071 -0.048 -0.00004 -0.002 1 Debt_ratioi,t 0.004 0.033 -0.0002 0.003 -0.045 0.006 0.0000 0.033 -0.128 1 Source: Author’s Tables 8 provides the regression results covering all technologies with and without fixed effects. Table 8: Regression results covering all technologies (1) (2) (3) (4) (5) ωi,α,t 0.525*** 0.518*** 0.454*** 0.476*** (0.003) (0.004) (0.011) (0.014) TCIα,t 0.003 0.316*** 0.351*** (0.048) (0.056) (0.061) Govi,α,t 0.694*** 0.645*** 0.644*** (0.048) (0.050) (0.051) Agei,t 0.056*** 0.054*** 0.050*** (0.008) (0.008) (0.008) num_employeei,t 0.155*** 0.078*** 0.086*** (0.014) (0.014) (0.015) num_competitorα,t 0.352*** 0.283*** 0.277*** (0.005) (0.005) (0.005) num_RTAi,t 1.261*** 0.067** 0.009 (0.014) (0.032) (0.038) Profit_ratioi,t 0.0001 0.0003 -0.001 (0.003) (0.003) (0.003) Debt_ratioi,t 7.116*** 9.813*** 9.652*** (1.654) (1.656) (1.666) Time fixed effect No No No No Yes Costant -6.681*** -6.694*** -6.756*** -8.220*** (0.012) (0.217) (0.014) (0.253) Observations 6,179,317 6,179,317 5,320,088 5,320,088 5,320,088 Pseudo R2 0.1044 0.1053 0.1156 0.1247 0.1274 *p < 0.1; **p < 0.05; ***p < 0.01 44 Manufacturing Firm Diversification into Industry 4.0 Technologies: Evidence from Korea Seoul Center for Finance and Innovation Website: https://worldbank.org/seoulcenter Seoul Center for Finance and Innovation