key: cord-0877933-6u18n4pm authors: Cheng, Long; Wang, Meng; Lou, Xuming; Chen, Zifeng; Yang, Yang title: Divisive Faultlines and Knowledge Search in Technological Innovation Network: An Empirical Study of Global Biopharmaceutical Firms date: 2021-05-24 journal: Int J Environ Res Public Health DOI: 10.3390/ijerph18115614 sha: c343a17bb381db48e97a665387756f14218baae5 doc_id: 877933 cord_uid: 6u18n4pm Divisive faultlines caused by the uneven distribution of relationship strength play an essential role in knowledge search in the technological innovation network, which serves as an important requirement for the technological innovation network’s macro level to expand to the meso-subgroup level and promote its healthy development. Given that the biopharmaceutical industry, as a high-tech industry, plays a vital role in promoting healthy development, this paper uses the joint patent applications of global biopharmaceutical firms from 2003 to 2018 as a sample to construct a technological innovation network, to explore the relationship between divisive faultlines and knowledge search in the technological innovation network. We also study the moderating effect of structural holes in this relationship. The empirical results show that divisive faultlines significantly affect the depth of knowledge search in the technological innovation network. Divisive faultlines have an inverted U-shaped effect on the breadth of knowledge search in the technological innovation network. Structural holes positively moderate the relationship between divisive faultlines and depth of knowledge search but negatively moderate the inverted U-shaped relationship between divisive faultlines and breadth of knowledge search. This research reveals the relationship between divisive faultlines and the knowledge search in the technological innovation network. The research results provide a theoretical basis and management enlightenment to improve biopharmaceutical firms’ knowledge search ability and promote healthy and sustainable development. In recent years, with the frequent occurrence of global public health events, health issues have become the focus of global attention. Biopharmaceuticals use modern bioengineering technology to create drugs with special curative effects, which play an essential role in treating significant diseases that seriously threaten human health [1, 2] . Many countries in the world unanimously cultivate the biopharmaceutical industry as a new economic growth point, accelerating to seize the commanding heights of "biological economy" [3] . China regards biopharmaceuticals as one of the strategic emerging industries, an essential support for Healthy China. Lalor et al. [4] confirmed that biopharmaceuticals are closely related to developing the national economy and improving people's healthy life quality. A statistical report by the US market research company IMS Health in 2017 pointed out that the global biopharmaceutical industry's market size is developing rapidly. Its sales growth rate exceeds the growth rate of the gross world product (GWP). With the accelerating economic globalization process and the increasingly complex market environment, the biopharmaceutical industry's sustainable development is inseparable from rich knowledge, technology, and other resources [5] . As an effective way for Int. J. Environ. Res. Public Health 2021, 18 , 5614 2 of 20 firms to acquire diversified knowledge, a knowledge search helps enterprises enrich their knowledge base, improving their competitive advantages [6] . As an emerging technology industry, the biopharmaceutical industry usually faces problems such as high risk, high investments, and long returns on investments in the research and development of biopharmaceuticals [7] . It is difficult for biopharmaceutical firms to rely on their limited resources and ability to acquire diversified knowledge to ensure the successful research and development of new drugs. Therefore, firms will shift from closed innovation to open innovation and achieve success in technological innovation activities through cooperative innovation with other innovation subjects [8] . The technological innovation network composed of multiple firms or organizations through cooperative innovation relationships is an important carrier and organizational form for cooperative innovation activities between enterprises [9, 10] . Su and Vanhaverbeke [11] found that technological innovation networks can improve corporate knowledge search efficiency and have a significant effect on integrating innovation resources and enhancing innovation capabilities. Balachandran et al. [12] confirmed that firms could acquire heterogeneous knowledge and resources in the innovation network to improve innovation performances. In practice, the United States has formed industrial clusters with the five major biotechnology industrial areas of San Francisco, Boston, Washington, North Carolina, and San Diego to improve the country's pharmaceutical technology capabilities and promote the development of the biopharmaceutical industry. Research has proven that the technological innovation network is conducive to the firms' knowledge search, such as in the early drug discovery phase [13] . However, with the diversification of subjects and inter-firm relationships in the technology innovation network, firms have a preference in choosing innovation partners. Firms tend to maintain strong relationships with partners who have a historical basis for cooperation or high trust, while maintaining weak relationships with other partners, resulting in the uneven distribution of the strength of the relationships between firms in the technological innovation network [14] . The research further finds that, with the diversification of subjects and inter-firm relationships in the technology innovation network, the uneven distribution of the inter-firm relationship strength will lead to obvious agglomeration or alliances among members in the network, and this small group formed by the close connections among members is defined as a subgroup [15] . Subgroups are a common phenomenon in technological innovation networks. Jojo et al. [16] built a technological innovation network based on the alliance of 203 biopharmaceutical firms around the world and found that there are a large number of subgroups in the network, and cohesive subgroups will affect the knowledge search among firms. Heidl et al. [17] proceeded from the perspective of the modular structure at the meso level, such as alliances or factions and found that the uneven distribution of the strength of the relationship between firms will trigger the group discontinuity mechanism, which makes the technological innovation network divide into multiple subgroups and affects the knowledge flow and knowledge search in the technological innovation network. Yi et al. [18] , from the perspective of similarity selection and interactions between enterprises, found that divisive faultlines have a negative impact on knowledge sharing in the technological innovation network. Although existing studies have proven that there are divisive faultlines in the technological innovation network and impact on the knowledge search and performance of firms, few empirical studies use a knowledge search as a dependent variable at the meso subgroup level. Whether and how divisive faultlines affect knowledge searches in technological innovation networks has not been thoroughly elucidated to date. This paper introduces the theory of divisive faultlines from the level of the meso subgroup to explore the mechanism of the impact of divisive faultlines on a knowledge search in the technological innovation network. In addition, because firms are embedded in technological innovation networks, the opportunities for knowledge exposure to firms in different network locations are various. In particular, firms occupying structural holes have information advantages and control advantages in the innovation network, which will affect the knowledge search of enterprises to a large extent [19, 20] . Therefore, it is necessary to explore the moderating effect of structural holes on the relationship between divisive faultlines and knowledge search. By using a sample of global biopharmaceutical firms, this research incorporates divisive faultlines, knowledge searches, and structural holes into the same framework, which is helpful to understand the impact of divisive faultlines on a knowledge search in the technological innovation network, and the moderating effects of structural holes on them, and to provide a theoretical basis and a new perspective for improving the efficiency of a knowledge search in biopharmaceutical firms. The contributions of this paper are mainly reflected in the following aspects: First, the existing knowledge search research primarily focuses on the micro self-centered network level and the macro integrated network level. However, there are few studies from the meso subgroup level. This paper investigates the impact of divisive faultlines on a knowledge search in the technological innovation network, which enriches the literature in the research field of the influencing factors of the knowledge search. Second, based on the existing research, this paper introduces divisive faultlines to the network level. Combined with the characteristics of the technology innovation network, we analyze divisive faultline connotations from the perspective of relationship embeddedness and study the manifestation of divisive faultlines in the innovation network. It enriches the related research on divisive faultlines and makes up for the lack of attention paid to divisive faultlines at the firm and network levels. Finally, based on the structural holes theory, we explore the structural hole moderating effect on the relationship between divisive faultlines and a knowledge search. As structural factors, structural holes can affect the role of factors at the level of relationships between members, which provides a new perspective for firms on how to use the positive effect of divisive faultlines and avoid the negative effect of divisive faultlines to improve the ability of a knowledge search and promote sustainable development. The paper is organized as follows. Section 2 introduces the literature review. Section 3 proposes the hypotheses and theoretical models. Section 4 shows the research design. Section 5 shows our empirical results. Conclusions are given in the last section. To solve the standing problems of uncertainty, resource scarcity, and limited innovative ability inside the firm in a modern innovation environment, the technology innovation network, a form of network organization based on common innovation goals is established by collaborators between firms and others or organizations [10, 21] . The concept of an innovation network was first proposed by Freeman [22] , who believed that an innovation network is the basis for systemic innovation, and a network is mainly constructed by innovation cooperation between firms. Jianbo et al. [23] found that firms' research and development (R&D) cooperation in the technological innovation network can overcome innovation impediments, share R&D costs, gather resources, exchange technology, and share benefits. For example, Shanghai Pudong Zhangjiang Hi-Tech Park's Pharmaceutical Research Institute once cooperated with the Green Valley Holding (Group) in the park on drug research and development. The Green Valley Holding (Group) provided funds, and the Pharmaceutical Research Institute invested in technological achievements; the cooperation greatly improved the R&D progress and completed phase I, II, and III clinical trials in just over one year. A knowledge search refers to the specific search strategy that firms choose to realize innovation. Petruzzelli et al. [24] pointed out in the research that the technological development of the biopharmaceutical industry needs to rely on the search and integration of a large amount of diversified knowledge for a long time. At present, domestic and foreign scholars have studied the connotation of a knowledge search from different perspectives, such as open innovation theory, transaction cost theory, and social capital theory, and the dimensions of a knowledge search are divided from a variety of different perspectives [25, 26] . According to different search strategies, the existing research divides a knowledge search into two dimensions: depth of the knowledge search and breadth of the knowledge search. Since these two dimensions can more accurately reflect the scope and degree of a knowledge search, they are adopted by more and more scholars [27] . Therefore, this paper follows this division method and divides the knowledge search into two dimensions: depth of the knowledge search and breadth of the knowledge search. Among them, the depth of knowledge search refers to the frequency of repeated visits and existing knowledge by firms, emphasizing the focus of knowledge [28] . The breadth of knowledge search refers to the extent of the fields and channels involved in a firm's search for external knowledge, emphasizing the diversification of knowledge [29] . The related research of the technological innovation network believes that [30] the close connection between firms encourages members to excavate and revisit existing knowledge continuously and deepen the depth of the knowledge search. Simultaneously, the sparse connections among firms provide channels for firms to search for external knowledge, which is conducive to firms acquiring diversified knowledge [31] . However, some scholars believe that [32] too close or too sparse relationships between firms will limit firms' opportunities to acquire novel knowledge and hinder firms' innovation performances. This study believes that the uneven distribution of the strength of inter-firm relationships will affect the depth and breadth of a knowledge search, and the mechanism of uneven distribution of the strength of interorganizational relationships on the depth and breadth of a knowledge search is not clear. Therefore, it is necessary to clarify the influence mechanism of the uneven distribution of the strength of inter-firm relationships on the depth and breadth of a knowledge search. Divisive faultlines refer to the internal differentiation tendency of the whole network caused by the difference in the degree of shared experience among members in the process of firms' interactive innovations, which is an important factor in forming and developing subgroups in the technological innovation network [33] . Take China's COVID-19 vaccine development as an example; in order to maximize the success rate and speed of COVID-19 vaccine research and development, the scientific research team divided the development of a COVID-19 vaccine into five main technical routes: live vaccines, adenovirus vector vaccines, attenuated influenza virus vector vaccines, recombinant protein vaccines, and nucleic acid vaccines, which will also mean that the research and development of the new crown vaccine is mainly divided into five subgroups. There are differences in the shared experiences between members of the various subgroups; members in the same subgroup will form close cooperation, and the relationships between different subgroups will be sparse, so that the technological innovation network formed with the goal of COVID-19 vaccine research and development will cause potential divisive faultlines. In the literature, related scholars have conducted research on divisive faultlines. The concept of divisive faultlines was first proposed by Lau and Murnighan [34] in the study of team diversity. Those team members with multiple identical attributes (e.g., gender, race, age, and other demographic attributes) may have strong cohesion, so that the team is divided into two or more different subgroups, and the members of each subgroup have high similarities. On this basis, Heidl et al. [17] believed that there are also divisive faultlines in the technological innovation network, because the partner selection tendency of "establishing strong relationships based on historical cooperation" will cause the cohesion of firms on a local scale, resulting in experience between partners. The degree of sharing is different, resulting in the challenges of subgroups of "within the group" and "outside the group". With deepening research, Zhang et al. [35] and Zhang and Guler [36] found that divisive faultlines have a significant influence on the formation of cooperation between firms and the changes of network members. Xinghua et al. [33] and Long et al., [37] extended divisive faultlines between firms to the technological innovation network, believing that the pre-relationship and the embeddedness of various relationships would cause divisive faultlines, making the members of the subgroup more willing to establish cooperative relations with familiar firms and more inclined to search for knowledge within the known range. This study believes that the uneven distribution of the strength of relationships between firms will cause potential divisive faultlines, making the technological innovation network present the characteristics of local close connections and sparse global connections. Firms will choose different knowledge search strategies in the process of innovation. Therefore, it is necessary to explore divisive faultlines' influential mechanisms on the depth and breadth of a knowledge search from the meso subgroup level. Structural holes refer to the voids between unconnected actors in a technological innovation network. If an actor connects two actors who are not directly connected, the actor is considered to occupy structural holes [38] . Structural holes depict the nonredundant connection between the two actors. Shi et al. [39] believed that firms occupying structural holes can quickly acquire nonredundant knowledge, accelerate the accumulation of coding knowledge, and deepen the understanding of existing knowledge. Long and Xinghua [40] believed that the widespread structural holes in the network are conducive to forming weak relationships between firms and encouraging firms to search for knowledge from a more extensive scope. However, some scholars believe that firms occupying structure holes do not establish a direct close connection with other connected firms, which will reduce the partner's trust, aggravating the relationship breakdown between firms and, thus, hinder firms from acquiring knowledge from the outside [41] . Based on the above analysis, this study believes that structural holes, as a structural factor, have advantages in the network that can affect the depth and breadth of a knowledge search. More importantly, when structural holes directly or indirectly act on the level of interfirm relationships, they will also affect the depth and breadth of a knowledge search. Therefore, it is necessary to explore the moderating effect of structural holes further. In the technological innovation network, cooperation between firms often relies on the norms and conventions formed by historical cooperation experiences. Partners with solid relationships maintain intense and frequent contacts, forming cohesive subgroups, while partners with weak relationships are queued outside the cohesive subgroups [33] . A cohesive subgroup has the characteristics of a close and substantial connection and knowledge focus, which is conducive to firms' in-depth searches. First, under the influence of inter-organization conventions, firms may not be willing to spend time and cost to establish new relationships but prefer to establish a deeper connection with historical partners, deepening the understanding of the existing domain knowledge, and to improve the efficiency of the knowledge reorganization and utilization in the process of innovation [42] . Second, under the effect of solid relationships, the close ties between firms establish a bridge for the flow of knowledge in the innovation network. The network members can search for knowledge quickly and accurately, reducing knowledge search costs [43] . Finally, as time goes by, the stronger the innovation network is, the higher the cohesion within the subgroup will be. Due to the existence of a high degree of trust mechanisms, members in the same subgroup are more willing to carry out knowledge reciprocity with local members, so that the knowledge can be focused on a specific field. The ability of members to absorb knowledge is enhanced, thus improving the depth of the knowledge search [44, 45] . Take the filed patents of biologics-based drugs from the new York-based Pfizer in 2017 and 2018, for example. A strong, cohesive subgroup, the formation of divisive faultlines, could be observed between Pfizer and its collaboration partners. Among the 16 patents filed in 2017, 13 came either from the old partners (four in total) such as Merck, ABBVIE STEMCENTRX LLC, and BRISTOL-MYERS SQUIBB CO or from itself (nine in total). Similar cooperation subgroups could be found in the filed patents in 2018, with nine patents out of 10 coming from subgroups (five in total) or itself (four in total). A closer look into the knowledge of filed patents in these two years revealed a deepened depth of the knowledge search. Twenty-five out of 26 patents focused on the development of antibody or antibody-drug conjugates, which could be used to treat cancer. Among them, six patents, in collaboration with Merck, disclosed antibody therapies targeting a specific target, PD-L1. These patents described deep knowledge regarding the clinical treatments, ranging from antibody formulation to diagnostic antibody, as well as combinational treatment, of an anti-PD-L1 antibody and a target specific inhibitor. Based on the above analysis, we proposed the following hypothesis: Hypothesis 1. Divisive faultlines have a positive impact on the depth of a knowledge search in the technological innovation network. Divisive faultlines promote the depth of a knowledge search by forming highly cohesive subgroups, but the impact on the breadth of the knowledge search has two sides. On the one hand, when divisive faultlines are low, the network relationships are evenly distributed, and firms are either unfamiliar or very familiar, which is not conducive to the breadth of a knowledge search. Specifically, when firms are not familiar with each other, due to a lack of trust, the knowledge exchange between members of the subgroups is not deep, and the willingness to share knowledge is low, which hinders members in subgroups from searching for knowledge from the outside [18] . However, when firms are very familiar with each other, due to the cognitive constraints of homogenized knowledge, members of the subgroups rely excessively on the inherent innovation mode, making the subgroup internal information redundant and difficult to search for diversified knowledge [46] . On the other hand, with the strengthening of divisive faultlines, the distribution of relationships in the network become uneven. The differences in the degree of experience-sharing between members becomes more prominent, which quickly causes close connections between members of the subgroups and sparse connections between the subgroups. Halevy et al. [20] believed that the combination of close internal connections and appropriate bridging relationships can effectively increase the brokerage value and help network members search for diverse knowledge. The close contact within the subgroup provides members with stable resources and channels for information communication, reducing the knowledge search risk. The bridging relationship outside the subgroup establishes a communication channel for different subgroups, bringing heterogeneous knowledge to the subgroup members [47] . Especially, firms occupying an important position in the network can easily attract members outside the group to actively introduce their innovative resources and diversified knowledge to bridge the relationships built between groups [48] . In other words, strong relationships within subgroups and bridging relationships between subgroups increase the interactions between subgroups, and members can continuously search for new knowledge elements from the outside, expand the existing knowledge bases, and promote technological innovation in biopharmaceutical firms. However, excessively high divisive faultlines have aggravated the factional gatherings among members, caused faction disputes between subgroups, and hindered general social exchanges among members in the technological innovation network. First, the interactions between members are limited to the group, and other resources such as knowledge and information are difficult to flow freely between the subgroups. As a result, the subgroup members can only obtain partial knowledge but cannot search for knowledge from a prominent scope [36] . Second, the enhancement of divisive faultlines will aggravate the degree of polarization of the subgroups in the technological innovation network, leading to limited cohesion between the subgroups. Further, amplifying the conflicts between members within and outside the subgroups, causing members inside and outside the subgroup to protect their knowledge, members of the network cannot obtain diverse knowledge. Third, the higher the divisive faultlines in the technological innovation network, the more significant the heterogeneity of knowledge between the subgroups, and the lower the transfer rate of knowledge between the subgroups, which will increase the difficulty of a knowledge search and is not conducive to the breadth of a knowledge search [37] . The "lightspeed" success of the Pfizer-BioNTech COVID-19 vaccine could be used to explain the impact of divisive faultlines and breadth of knowledge in the biopharmaceutical industry. An analysis of the collaboration partners of both companies before 2018 exhibited a rather confined innovation network. Besides, filed patents showed that Pfizer focused on the development of antibody treatments for cancer and BioNTech concentrated on RNA biotechnology, indicating a relatively narrow knowledge search within two companies resulting from a high level of divisive faultlines. However, the incentive to develop an effective influenza vaccine in 2018 drove Pfizer to find new collaborators and broaden the breadth of the knowledge search for vaccines outside its innovation networks connecting with old partners. Even though the first collaboration between Pfizer and BioNTech did not yield promising vaccines for the flu, an elementary subgroup free of conflicts and competition was formed between them, with BioNTech responsible for the vaccine discovery and Pfizer for the clinical trials and vaccine manufacturing. This kind of medium divisive faultline accelerated the development and wide distribution of the COVID-19 vaccine. Based on the above analysis, we proposed the following hypothesis: Hypothesis 2. Divisive faultlines have an inverted U-shaped effect on the breadth of a knowledge search in the technological innovation network. Structural holes occupying the technological innovation network can strengthen the relationship between divisive faultlines and depth of a knowledge search. First, the occupants of structural holes can accurately and timely obtain a high innovation value from numerous knowledge streams and effectively control the direct contact between partners, saving maintenance partners. The time and effort spent on redundant relationships reduces the cost of a knowledge search [49] . Second, structural holes have a higher visibility in the network, and the degree of integration of the resources in the network is also higher [32] . Firms occupying higher structural holes are more likely to gain recognition from the subgroup members. Such recognition enables firms to identify with existing knowledge, which is conducive to firms focusing on refining and reorganizing the existing knowledge elements in innovation, improving their ability to absorb knowledge more. It has a positive effect on the depth of the knowledge search [39, 44] . Finally, firms occupying structural holes have a higher impact on group activities. If members of the group are already familiar with this knowledge, they will be able to use more knowledge through reorganization to achieve innovation [50] . Therefore, the higher the divisive faultlines, the easier the firms occupying the structural holes may influence the members in the subgroups and spread the innovative knowledge, conducive to the in-depth mining and utilization of their knowledge. However, the diverse knowledge brought by divisive faultlines has two sides for firms occupying higher structural holes. On the one hand, structural holes negatively regulate the relationship between divisive faultlines and the breadth of knowledge search. When a divisive faultline's strength is low, high structural holes will make the network's diversified structural characteristics more obvious. The members of the network will face the risk of information overload. They will easily fall into cognitive inertia, which weakens the firm's willingness to acquire new knowledge [40] . Simultaneously, high structural holes reduce the trust between members, increase the cooperation costs and potential risks between firms, and make it more difficult for firms to obtain external resources, making enterprises more inclined to dig and use existing knowledge [41] . On the other hand, structural holes can break through the boundaries between subgroups, act as boundary crossers, and slow down the negative relationship between divisive faultlines and the breadth of the knowledge search. First, with the advantages of crossfirm boundaries and technical fields, firms occupying structural holes can cooperate with partners in other subgroups, conducive to forming bridging relationships between network subgroups to a certain extent, alleviating strong divisive faultlines. The structural isolation between groups reduces the difficulty of strong divisive faultlines for the knowledge flow and knowledge sharing between subgroups [51] . Second, structural holes establish a cooperative relationship between members inside and outside the subgroup, so that disconnected members have a common third party to monitor its partner behavior, curb opportunistic behavior, and prevent further destructive divisions [20] . Finally, with the help of the information advantages of structural holes, the occupants of structural holes can gain greater power in acting as an "information bridge", which is conducive to establishing extensive cooperative relations between firms and other partners. Diversified knowledge, information, and other resources can be smoothly circulated in the network, prompting firms to search for diversified knowledge [49] . Based on the above analysis, we proposed the following hypotheses: In summary, based on theoretical analysis and research hypotheses, the theoretical model constructed in this paper is shown in Figure 1 . diversified structural characteristics more obvious. The members of the network will face the risk of information overload. They will easily fall into cognitive inertia, which weakens the firm's willingness to acquire new knowledge [40] . Simultaneously, high structural holes reduce the trust between members, increase the cooperation costs and potential risks between firms, and make it more difficult for firms to obtain external resources, making enterprises more inclined to dig and use existing knowledge [41] . On the other hand, structural holes can break through the boundaries between subgroups, act as boundary crossers, and slow down the negative relationship between divisive faultlines and the breadth of the knowledge search. First, with the advantages of cross-firm boundaries and technical fields, firms occupying structural holes can cooperate with partners in other subgroups, conducive to forming bridging relationships between network subgroups to a certain extent, alleviating strong divisive faultlines. The structural isolation between groups reduces the difficulty of strong divisive faultlines for the knowledge flow and knowledge sharing between subgroups [51] . Second, structural holes establish a cooperative relationship between members inside and outside the subgroup, so that disconnected members have a common third party to monitor its partner behavior, curb opportunistic behavior, and prevent further destructive divisions [20] . Finally, with the help of the information advantages of structural holes, the occupants of structural holes can gain greater power in acting as an "information bridge", which is conducive to establishing extensive cooperative relations between firms and other partners. Diversified knowledge, information, and other resources can be smoothly circulated in the network, prompting firms to search for diversified knowledge [49] . Based on the above analysis, we proposed the following hypotheses: In summary, based on theoretical analysis and research hypotheses, the theoretical model constructed in this paper is shown in Figure 1 . As a typical knowledge-intensive industry, the biopharmaceutical industry integrates knowledge in different fields, such as biology, pharmacy, medicine, and As a typical knowledge-intensive industry, the biopharmaceutical industry integrates knowledge in different fields, such as biology, pharmacy, medicine, and biochemistry. Different from the traditional pharmaceutical industry, which mainly develops synthetic small molecular drugs, the biopharmaceutical industry concentrates on the development of biologics-based drugs, including peptides, antibody-drug conjugates, enzymes, and nucleic acid-based compounds. Compared with traditional drugs, biologics-based drugs outcompete them, owing to the advantages of high specificity and obtained immunogenicity, which have made the biopharmaceutical industry stand out in recent years. Especially after the outbreak of COVID-19 in 2020, many countries have successively listed COVID-19 as a strategic industry, which has promoted the development of the biopharmaceutical industry. Therefore, we used global biopharmaceutical firms as the research object. A patent is the primary carrier of cooperation and innovation among firms. There are a large number of structured data fields in patent text, such as the patent number, patentee name, patentee code, application date, international patent classification (IPC), etc. The patentee and the patent right code provide data support for the construction of cooperation networks between firms, and the IPC classification number offers a feasible solution for the measurement of a knowledge search. Take the patent US2018296691-A1 that Pfizer and AbbVie jointly applied in 2017 as an example; the two firms have their own unique codes in the patents, namely PFIZ-C and ABBI-C, and there are three IPC classes in the patent, including A61K-031/704, A61K-047/68, and C07K-016/30. When measuring a knowledge search, we used the first four digits of the IPC (such as A61K or C07K). Therefore, this paper used patent data in the biopharmaceutical firms to verify the hypothesis. The process of collecting and cleaning patent data in biopharmaceutical firms includes the following steps: First, the Derwent Innovations Index database contains the patent information of 41 patent institutions around the world, since 1963, covering more than 100 countries, which is the authoritative data source for global scientific and technological intelligence agencies. Therefore, we searched for biopharmaceutical patents from January 1, 2003 to December 31, 2018 through the Derwent Patent Database. Second, we cleaned the downloaded patent data. Since the patentee had many inconsistencies in the same firm's names, such as name changes, abbreviated names, parent and subsidiary companies, and missing letters, this paper used the patentee code in the Derwent patent database to unify the names of the patentees. Simultaneously, we selected two or more patents with the patentee code, because the patents contained two or more different patent rights codes to represent the achievements of innovations between firms, so the code of the patentee was removed as the cooperative patent of the individual. In view of the fact that some firms' temporary entry or exits will have a particular impact on the technological innovation network, this paper only retains firms that have participated in at least three collaborations in different years. Finally, we referred to previous studies [52] , based on the cooperation between firms to apply for patents, we established a five-year mobile time window to divide the data from 2003 to 2018 into 11 time windows (i.e., 2003-2007, 2004-2008, . . . 2013-2017) and the corresponding technological innovation networks. Restricted by the lack of variable information and other problems, finally, we obtain the patent data of 509 biopharmaceutical firms. Therefore, this paper's empirical analysis was based on the unbalanced panel data of 509 biopharmaceutical firms from 2003 to 2018 (N = 1798). To more comprehensively measure the degree and scope of the knowledge search in a firm's technological innovation network, we drew on the related research of Katila and Ahuja [28] to divide the knowledge search into two dimensions: depth of the knowledge search and breadth of the knowledge search. Depth of the knowledge search represents the degree of focus or specialization of knowledge, and breadth of the knowledge search represents the range or diversification of knowledge. Based on the availability of the existing research and patent data, this paper measured the depth of the knowledge search by calculating the average number of times that companies reused the knowledge in patent applications in year t. At the same time, by referring to the relevant research of scholars such as Rongkang et al. [53] , the breadth of the knowledge search was measured by calculating the number of IPC categories (that is, taking the first four digits of the IPC patent number) contained in the patents filed by the firm in year t. In this paper, based on the relevant research of Heidl et al. [17] , we measured divisive faultlines mainly by the discreteness of the strength of the binary relation of the selfcentered network. First, the historical relationship strength was measured by the duration of the relationship in years. According to the 5-year time window, the duration of the relationship between each self-centered network member pair in the past five years (t-5 to t-1) was calculated, with a value range of 0-5. For example, the historical relationship strength value was 1 if the member pair lasted for 1 year in the past 5 years. The strength value was 2 if the pair lasted for 2 years, and so on. Second, when we measured the uneven distribution of relationships among firms, it was measured by the standard deviation of the strength of the relationship between each pair of members in the self-centered network within a 5-year time window, because the standard deviation meant the discrete type of binary relationship distribution. The standard deviation was 0, which indicated that the relationship strength distribution between the member pairs in the self-centered network was uniform, which meant that the network's divisive faultlines were low. The larger the standard deviation, it indicated that the distribution of the relationship strength between the pair of self-centered network members was exceptionally uneven, which meant that the network's divisive faultlines were high. In this paper, based on the research of Jia and Yue [49] and Zhang and Luo [54] , we used the constraint index in the Burt [38] indicator to calculate the firm's structural holes in the technological innovation network from t-5 to t-1s year. Its calculation formula is as follows: where p iq denotes the investment proportion of firm i in the relationship of firm j in the technological innovation network, and ∑ q p iq p qj is the total amount of indirect relationship of firm i in the network. Based on the research on the divisive faultlines and knowledge search in the technological innovation network, this paper argued that the factors at the technological innovation network level and the firm level will have an impact on the knowledge search, including network size, network density, betweenness centrality, knowledge base, and technological R&D capability. Network size: the larger the network's size, the larger the number of firms in the technology innovation network, and the more complex the interaction process between members will be. Therefore, this paper controlled the network size by calculating the number of members directly related to the firm from t-5 to t-1 years. Network density: the greater the density of the technological innovation network, the closer the connection between members, and the greater the impact on the behavior of the firms in the network. The calculation formula is as follows: Network density = 2L/(N × (N − 1)), where L is the actual relationship number in the technological innovation network, and N is the number of firms in the technological innovation network. This formula expresses the ratio of the actual relationship number to all possible relationship numbers in the five years from t-5 to t-1 in the technological innovation network. Betweenness centrality: the higher the centrality of a firm in the technological innovation network, the more important its position in the network, and the stronger the control over the surrounding firms. The calculation formula is as follows: Betweenness Centrality = ∑ b