Should It Always Be Central? Substitution Effects of Multi-Network Embeddedness on Absorptive Capacity

Abstract

In the knowledge economy, organizations embedded within large network systems enhance their capability to identify, acquire, assimilate, and effectively utilize external knowledge. Given the diversity and complexity of these networks, examining the factors that influence organizational absorptive capacity from a multi-network perspective is both essential and timely. However, how organizations can strategically allocate positions across different networks to enhance absorptive capacity remains unclear. Drawing on social network theory and the knowledge-based view, this study proposes that the relational embeddedness of organizations across networks produces an interactive substitution effect on absorptive capacity. Using a composite database of patent, financial, and organizational data, we annually constructed multiple networks of listed companies in the global automobile and components industry and empirically tested the model through multiple stepwise regression analysis. The results indicate that relational embeddedness in both cooperation and knowledge networks positively affects absorptive capacity; nevertheless, relational embeddedness in knowledge networks can limit the positive effect of cooperation network embeddedness on organizational absorptive capacity. This interaction highlights the substitution effect between different network roles in shaping absorptive capacity.

1. Introduction

With the increasing complexity of the environment and the development of economic globalization, organizations can no longer adapt to the rapid iteration of technology and gain sustainable competitive advantages by relying solely on their own resources and technologies [1]. Therefore, organizations need to continuously absorb external resources and technologies and integrate them with their own knowledge to maintain their competitive advantage [2]. When the internal development needs of an organization do not match its existing resources and structure, the organization needs to obtain resources from the environment or seize opportunities. This absorption process represents the behavior of organizations to obtain resources and information from external systems such as cooperation networks or knowledge networks, which can help organizations cross boundaries, make up for deficiencies and leverage their strengths.

The concept of absorptive capacity was first proposed by Cohen and Levinthal [3], and refers to the ability of an organization to identify, acquire, assimilate and effectively utilize external knowledge. These three dimensions include not only the ability to imitate the products or processes of other organizations, but also the ability to utilize less commercialized knowledge. Zahra and George [4] expanded absorptive capacity to the dynamic ability of companies to innovate and apply knowledge in order to gain and maintain competitive advantages. Organizational absorptive capacity is affected by many factors. By combing through existing research results, it is found that organizational absorptive capacity is affected by prior knowledge base [5], R&D investment level [3], internal organizational structure [6], etc. Cohen and Levinthal [3] first focused on the cognitive basis of individual absorptive capacity based on prior knowledge background, and then described the factors that influence absorptive capacity at the organizational level, how the absorptive capacity of an organization differs from that of individual members, and the role of diversity of expertise within an organization. The development of absorptive capacity is path dependent [7], and the lack of early investment in a particular area of expertise may hinder the organization’s future development of technology in that area.

It is necessary and important to study the factors affecting organizational absorptive capacity from the perspective of multiple networks, because organizations are often embedded in multiple complex social networks [8,9]. Network embeddedness refers to the deep participation of an enterprise in a certain social network through its relationships with other organizations or individuals to obtain resources, information or support. Network embeddedness can be studied from two dimensions: relational embeddedness and structural embeddedness [10], and the former can better reflect the central position of the node in the network [11]. The position of organizations in different social networks has a great influence on their use of various knowledge and resources [12]. However, the mechanism of how organizations reasonably allocate their positions in multiple networks to improve their absorptive capacity is still unclear. The flow of knowledge resources in cooperation networks is crucial for enhancing absorptive capacity, while the characteristics of the organizational knowledge system, represented by knowledge networks, also directly influence the knowledge absorption and assimilation processes. However, the limited knowledge resources in competitive networks have less impact on absorptive capacity.

Therefore, our research is dedicated to investigating the impact of cooperation and knowledge network embeddedness on company absorptive capacity, with a particular focus on examining the potential interactions between these two forms of embeddedness. Through theoretical analysis and literature review, we construct a relationship model linking multiple network embeddedness to organizational absorptive capacity. We then empirically test this model using data from the global automobile and components industry, analyzing the results of regression and robustness tests to draw conclusions. Although previous studies have examined the impact of network embeddedness on absorptive capacity, our research offers several key innovations. First, unlike existing studies that primarily focus on a single type of network, we expand the analysis to a multi-network perspective. Second, we focus specifically on the global automobile and components, which has been undergoing significant transformation over the past decade. In this context, absorptive capacity is especially crucial for firms navigating such a rapidly changing environment. Third, this study also highlights the potential substitution effect between cooperation network and knowledge network embeddedness. Specifically, we show that the embeddedness in knowledge networks may, in some cases, limit the positive impact of cooperation networks on absorptive capacity.

2. Hypotheses

2.1. Relational Embeddedness Within the Cooperation Network and Absorptive Capacity

Developing and maintaining absorptive capacity is crucial to the long-term survival and development of a company because absorptive capacity can strengthen, supplement or reposition the organization’s knowledge portfolio [13]. From the perspective of social network theory, all R&D economic activities of an organization are embedded in the social relationship network to which the company belongs [14], and the external relationship network is an important carrier for organizations to obtain scarce resources such as knowledge and engage in innovation activities [15,16]. From the perspective of the resource-based view, different organizations have different absorptive capacities, and the level of resource acquisition and development can help organizations enhance their competitive advantage [17].

Research cooperation between organizations can take many forms, such as technology sharing, joint ventures, or joint R&D. The general characteristic of a collaborative relationship is that organizations work on different research projects with non-overlapping partners, and this collaborative relationship can be represented by a network [18]. In a collaborative network, the organization with higher degree centrality is connected to more organizations, so it has a positional advantage in obtaining corresponding knowledge and information resources [19]. In the knowledge identification and acquisition phase, the organization with higher degree centrality has direct connections with multiple organizations, so it has more opportunities to acquire knowledge [20]. In the assimilation and transformation stage, due to the convenience of direct contact, the organization with higher degree centrality can more quickly transform external knowledge into its own knowledge [21]. In the effective utilization phase, the organization with higher degree centrality can obtain information that can be skillfully applied in the process of direct communication with other organizations, which makes it easier for the organization to directly utilize the corresponding knowledge [22]. The more connections an organization has with other organizations, the higher its centrality. This advantage can enhance the organization’s knowledge recognition, absorption and transformation capabilities, and effective utilization capabilities, that is, it can enhance the organization’s absorptive capacity. Based on this, this study proposes the following hypothesis:

H1: 

The relational embeddedness within the cooperation network can positively affect organizational absorptive capacity.

2.2. Relational Embeddedness Within the Knowledge Network and Absorptive Capacity

A cooperation network usually refers to a network of connections established between enterprises, organizations or individuals through common goals, resource sharing, and collaboration mechanisms. The knowledge network focuses on the flow of knowledge, and we can analyze both from the perspective of knowledge elements and their associations, as well as from the perspective of the flow of knowledge and information between companies. The nodes in the cooperation network represent companies, and we consider a tie between two companies if they collaborate with each other (e.g., jointly apply for a patent). But when we talk about knowledge networks, we usually regard knowledge elements as nodes of knowledge networks, and if two knowledge elements can be integrated into one achievement (such as two IPCs included in one patent), we consider that there is a tie between them. Since we are studying the embeddedness of company in knowledge networks, we need to attribute the average network characteristics of knowledge elements in the company’s knowledge element portfolio to the organizational level, so as to characterize the company’s embeddedness in the knowledge network.

Cohen and Levinthal [3] emphasize that prior related knowledge enhances a firm’s absorptive capacity, improving its ability to assimilate and apply new knowledge. An organization’s knowledge network is distinct from its cooperation network [23], exhibiting varying structural characteristics and impact paths on the organization’s absorptive capacity. Based on the knowledge-based view, the knowledge network comprises elements and connections from the organizational knowledge portfolio [24]. Knowledge elements typically encompass the initial findings of the scientific or technological research community regarding a specific subject, comprising facts, theories, methods, and procedures [23]. Broadly, knowledge elements are interdependent and collectively form a larger knowledge system, representing distributed knowledge repositories and agents that seek, relay, and generate knowledge. Ties in knowledge networks both facilitate and limit the acquisition, transfer, and creation of knowledge [25].

The organization’s relational embeddedness in the knowledge network describes the strength of the combination between its knowledge elements and others. The high centrality of organizational knowledge indicates that it forms the core of the knowledge portfolio, with extensive connections to other organizational knowledge elements, reflecting the organization’s substantial reserves and widespread application [25]. The more central an organization is in the knowledge network, the broader its internal element combination experience and the greater the visibility in its technical field, increasing potential for combining organizational knowledge with other elements [26]. Organizations with higher relational embeddedness in the knowledge network can more effectively recognize, assimilate and apply external knowledge due to their rich knowledge base. Based on this, this study proposes the following hypothesis:

H2: 

The relational embeddedness within the knowledge network can positively affect organizational absorptive capacity.

2.3. Interaction Effect of Relational Embeddedness Across Multiple Networks

The relational embeddedness within the knowledge network signals the combinatory potential of organizational knowledge. The potential for knowledge combination relies on the natural connection level between the subject matter of a knowledge element and other knowledge elements [27]. Delving into the experiences of knowledge elements and their combinations tends to heighten their perceived combinatorial potential within the organizational research community. By investigating familiar knowledge elements for combinatorial opportunities, researchers can cultivate routines, standards, and modules that boost their innovation efficiency [28]. Simultaneously, the social construction of the combinatorial potential of knowledge elements is shaped by researchers’ beliefs regarding the feasibility and desirability of combining them with other elements [29], and robust beliefs typically steer resources towards exploring such opportunities.

The mechanism becomes more complex when we take collaborative networks into account. Ties in knowledge networks both facilitate and limit the acquisition, transfer, and creation of knowledge [25]. Effective cooperation networks frequently incorporate both formal and informal mechanisms, including joint ventures, alliances, and social networks, to facilitate the acquisition and integration of external knowledge by organizations [30]. Organizations often tend to favor the familiar and mature, as well as to search for solutions near to existing ones [31], which in some form creates a knowledge trap hindering the absorption of external knowledge. The influence of multiple networks on organizational absorptive capacity is interactive [32,33,34]. Knowledge networks and cooperation networks exhibit some overlap in their effects; thus, a higher centrality in the knowledge network can diminish the positive impact of cooperation network embeddedness on the organization’s absorptive capacity. Based on this, this study proposes the following hypothesis:

H3: 

The positive impact of the relational embeddedness within the cooperation network on absorptive capacity will be weakened by the relational embeddedness within the knowledge network.

3. Methods

3.1. Data Collection

Our analysis hinges on a secondary dataset containing basic information, financial data and patent data of listed companies in the automobile and components industry from the BVD database. We sourced the list and codes of listed companies in the automobile and components industry from the OSIRIS database, a total of 879 companies. Then, we imported these codes into the ORBIS database to gather basic information and financial data from 2011 to 2020, and into the Orbis Intellectual Property database to obtain the 216,840 patent data from 2008 to 2017.

3.2. Data Preprocessing

We used the 216,840 patent data to build the cooperation network, knowledge network and competition network of the global automobile and components industry from 2008 to 2017 year by year. Patent documents contain a wealth of information, such as the application year, applicants, patent owner, and IPCs. Patent information can be used to build various networks embedding companies [35]. First, we constructed cooperation networks based on the co-occurrence information of applicants in each year’s patents. The nodes in the cooperation network represent companies, and we consider a tie between two companies if both are applicants for a patent. Node characteristics, such as degree centrality and the constraint value of each node, were calculated using Pajek. Second, we constructed knowledge networks based on the co-occurrence information of IPCs in each year’s patents, and we calculated node characteristics using Pajek. The nodes in the knowledge network represent IPCs, and we consider a tie between two IPCs if both appear in the same patent. However, it is important to note that our research operates at the organizational level. As such, we need to attribute the network characteristics of IPC nodes in the knowledge network to the network level. By analyzing the dataset, we derived companies’ IPC portfolio and, ultimately, transferred the IPC characteristics in the knowledge network to the companies’ knowledge network embeddedness by calculating the average IPC characteristics within the portfolio. Third, we constructed competition networks based on the overlap information in IPC portfolios between companies, and we calculated node characteristics using Pajek. The nodes in the competition network represent companies, and we consider a tie between two companies if their IPC portfolios include the same IPCs.

Finally, these organizational-level characteristics of listed companies in multiple networks for each year were matched with their relevant basic information and financial data, resulting in an unbalanced panel dataset of 879 listed companies in the global automobile and components industry from 2008 to 2020. It is important to note that the final number of effective samples for empirical testing depends on data matching, influenced by two main factors: first, a company’s presence in a specific network in a given year is uncertain; second, this study applied a three-period lag to the network-related variables when testing causal relationships, which impacted the final sample size.

3.3. Measurement

3.3.1. Dependent Variable

The dependent variable is absorptive capacity (abbreviated as Abs). Existing studies offer various measures of absorptive capacity, including the number of patents [31], the number of developers [36], R&D expenditure [37], and R&D intensity [38]. To more effectively examine the impact of a firm’s embeddedness in the external network environment on its absorptive capacity, we use R&D intensity as a proxy for absorptive capacity. We measured organizational absorptive capacity using standardized R&D intensity, defined as the ratio of R&D expenditure to revenue [38]. This measure is commonly used to assess the efficiency of a company’s R&D activities and reflects its ability to acquire, assimilate, and transform external knowledge. While R&D expenditure is typically used as a direct measure of innovation output, we note that there is a significant difference between R&D expenditure and R&D intensity. Cohen and Levinthal mention that while R&D obviously generates innovations, it also develops the company’s ability to identify, assimilate, and exploit knowledge from the environment-what they call a company’s ’learning’ or ’absorptive’ capacity [39]. R&D intensity, though closely related to R&D investment, focuses on how efficiently a company utilizes its R&D resources, which is a key aspect of its absorptive capacity. Absorptive capacity involves the company’s ability to acquire and integrate external knowledge, whereas innovation capability focuses on the company’s ability to translate that knowledge into innovative outputs. Absorptive capacity is conceptually distinct from innovation capability. We believe that R&D intensity can serve as a reasonable proxy for absorptive capacity, particularly in the context of examining how external network embeddedness can enhance a company’s capacity to absorb knowledge. This metric reflects the company’s learning efforts and accumulated knowledge, while accounting for the impact of company size on R&D expenditure. A higher R&D intensity indicates a stronger absorptive capacity.

3.3.2. Independent Variables

Our two independent variables are relational embeddedness in the cooperation network and in the knowledge network. The existing literature provides varying interpretations and measurement methods for network embeddedness. Some scholars, for example, combine degree centrality, betweenness centrality, eigenvector centrality, and other forms of centrality as structural measures of network embeddedness [40,41]. Other studies differentiate network embeddedness into relational embeddedness and structural embeddedness [10,42]. In this framework, relational embeddedness is typically measured by degree centrality, reflecting the central position of companies within the network [34,43], while structural embeddedness is often represented by structural holes, indicating a company’s brokerage role within the network [43,44].

In this study, we use degree centrality to represent the relational embeddedness of cooperation networks and knowledge networks, based on two main considerations. First, we argue that degree centrality effectively captures a company’s position and resource acquisition potential within a network. Degree centrality, a classic concept in social network analysis [45], is commonly used to measure relational embeddedness. In most social networks, some nodes have higher degree centrality, positioning them at the network’s core, where information flows more easily. For companies embedded in the network, occupying a central position enables more convenient access to information [46,47]. In both cooperation and knowledge networks, degree centrality not only reflects a company’s connections to other nodes but also its ability to control resources and occupy a central role in information flow. Thus, degree centrality serves as a useful indicator of relational embeddedness between companies and their partners or knowledge sources. Second, one of the core objectives of this study is to explore the interaction between cooperation networks and knowledge networks. We believe that degree centrality can capture both structural and relational embeddedness, offering an intuitive measure of a company’s ability to occupy a central position in multiple networks. This dual role of degree centrality allows us to simultaneously assess its advantages in relational embeddedness across different network types. To better test causal relationships, we used a three-period lag for the independent variables when estimating the model.

(1) The first independent variable is relational embeddedness in the cooperation network (abbreviated as CPRre). Network centrality determines the dominant position of network nodes [48]. Existing studies usually use degree centrality, closeness centrality or betweenness centrality to measure the centrality of network nodes [49,50]. We use degree centrality as a proxy variable for relational embeddedness in a cooperation network, based on the following considerations. First, degree centrality reflects the cohesion of a node within the network, measured by the number of direct connections (edges) linking the node to others. Through these interactions, a node can access and leverage the resources embedded in the relationships. Second, in a cooperation network, the degree centrality of a node represents the extent of an organization’s cooperation, defined by the number of other organizations directly connected to it. A higher degree centrality indicates greater cohesion for the organization within the network and access to more abundant collaborative resources. Additionally, to avoid issues with reverse coefficients and suppression effects, the combined use of multiple centrality indicators in the estimation model should be minimized [51]. Therefore, we did not use the average of degree centrality and other centrality indicators. Instead, we directly adopted degree centrality as a proxy variable for relational embeddedness in a cooperation network.

(2) The first independent variable is Relational Embeddedness in the Knowledge Network (abbreviated as KLDre). The relational embeddedness of a company’s knowledge network refers to the extent to which the knowledge elements it possesses are central within the network. The degree centrality of nodes in the knowledge network quantifies the compatibility of knowledge elements. A higher degree centrality indicates a greater likelihood that the knowledge element it represents can be integrated and innovated with other elements. In this paper, we focus on listed companies within the automobile and components industry as our primary subject of investigation. The knowledge element portfolio associated with these companies typically comprises a vast array of knowledge components, with significant variation in the quantity of these elements among different firms. This diversity highlights the complex and dynamic nature of knowledge within the industry. Consequently, we employed the average degree centrality of knowledge elements within the company’s knowledge elements set as a metric to assess the company’s centrality within its embedded knowledge network. This approach enables a nuanced understanding of the company’s influence in the broader knowledge network. The calculation formula is as follows:

$$KLDr{e}_i={\displaystyle \frac{{\displaystyle {\sum }_{k=1}^{n_i}degre{e}_k}}{n_i}}$$

(1)

where KLDre_i represents relational embeddedness of company i within the knowledge network, degree_k denotes the degree centrality of knowledge element k in the knowledge network, and n_i refers to the total amount of knowledge elements in the knowledge element portfolio of company i. We utilized standardized KLDre to validate the regression model.

3.3.3. Moderating Variable

We also use knowledge network relational embeddedness as a moderating variable to examine its effect on the relationship between relational embeddedness in the cooperation network and absorptive capacity. The characteristics of a company within its knowledge network, particularly the nature of the knowledge it possesses (such as compatibility and cross-correlation), influence how information is transmitted and received. Thus, relational embeddedness in the knowledge network as a moderating variable is measured by the standardized average degree of knowledge elements in the company knowledge element portfolio.

3.3.4. Control Variables

To minimize the marginal effects of possible alternative explanatory variables, we included seven control variables related to the company’s organizational characteristics when estimating the impact of the its embeddedness in multiple networks on its absorptive capacity. These control variables pertain to the formation of absorptive capacity and include basic factors such as the company’s market attractiveness, patent stock, and age, as well as network characteristics such as relational embeddedness in the competition network, structural embeddedness in the cooperation network, structural embeddedness in the competition network, structural embeddedness in the knowledge network.

(1) Market Attractiveness (abbreviated as Mat). A company’s absorptive capacity is influenced by the external environment, particularly market attractiveness. When market attractiveness is higher, competitive pressure from the external environment compels companies to absorb more external technology and knowledge [3]. We utilized a comprehensive evaluation of market attractiveness for companies, provided annually by the ORBIS database, to assess their market attractiveness. The data are compiled by IPBI, a partner of BVD, using advanced data mining techniques and indicator-based valuation methods. The evaluation incorporates 25 key indicators, such as forward and backward citations, family size, geographical coverage, patent age, and legal status. By referencing historical patent transaction databases and manually reviewing patent mergers, acquisitions, valuation projects, and auctions, it assesses the level of competition in a company’s technological domain from an intellectual property perspective. The market attractiveness is quantified on a scale of 0 (minimum) to 100 (maximum) within a specific market segment. This study utilized the standardized results of these data to gauge the market attractiveness.

(2) Patent Stock (abbreviated as Pst). A company’s patent stock typically reflects its technological accumulation. Companies with extensive patent stock are often better positioned to absorb and transform external knowledge [3]. We measured patent stocks by the annual accumulated patents of listed companies in the automobile and components industry. Different from previous studies that used the number of patents of the focal organization before it became a sample to measure patent stock [35,52], we calculated the total number of patents of focal company in each year based on existing patent information to reflect the changing characteristics of patent stocks over time, thereby measuring the patent stock that takes into account dynamic changes.

(3) Age. As a company continues to exist, it gains a lot of resources, technology and experience, and its absorptive capacity also grows accordingly [53]. We measured the age of a company by calculating the number of years since its registration up to the observation period. Specifically, we used the difference between the sample year and the registration year of listed companies in the automobile and components industry as a proxy for company age. In estimating the model, this study standardized this variable.

(4) Relational Embeddedness in the Competition Network (abbreviated as CPTre). The relational embeddedness of a company in a competition network reflects its position and influence within the industry. By controlling for the relational embeddedness in the competition network, we can effectively isolate the effects of external factors such as resource acquisition, technology diffusion, and competitive strategy [19] on the enterprise’s absorptive capacity. We measured the relational embeddedness of a company in the competition network using its degree centrality within the competitive network. A higher degree centrality indicates greater competitive pressure or more diversified business activities for the company. In our model estimation, we utilize standardized degree centrality to assess it.

(5) Structural Embeddedness in the Cooperation Network (abbreviated as CPRse). Companies with high structural embeddedness act as a bridge for knowledge transfer, enabling them to communicate between different information groups, which may make it easier for them to acquire external knowledge and improve their absorptive capacity [19]. Structural holes quantify the degree of disconnection between network nodes and other related network nodes, which is also a very common network feature measurement method [54]. Burt measured the structural holes of nodes using network redundancy [55]. This study referred to the calculation method of structural holes in previous studies [23], using 2 minus the total constraint (calculated using a three-step method) to indicate the control advantage created by the company by spanning the structural hole. Additionally, our model estimation employed the standardized structural hole indicator to assess the structural embeddedness of the company in cooperation network structure.

(6) Structural Embeddedness in the Competition Network (abbreviated as CPTse). Companies with high structural embeddedness in competition networks can leverage their unique positional advantages to access more resources and technological information [55], potentially enhancing their absorptive capacity. We used standardized structural hole indicator of company in the competitive network in which they are embedded to measure the structural embeddedness in the competition network.

(7) Structural Embeddedness in the Knowledge Network (abbreviated as KLDse). The structural characteristics of a company’s knowledge element portfolio in the knowledge network exhibit interdisciplinary traits, which influence the efficiency of cooperation and communication, thereby impacting the company’s absorptive capacity [56]. We used standardized average structural holes of knowledge elements within the company’s knowledge element portfolio as a metric to assess the company’s structural hole indicator within its embedded knowledge network. The standardized structural hole index of company in the competitive network in which they are embedded to measure the structural embeddedness in the competition network.

3.4. The Analytical Model

3.4.1. Model Settings

We used a linear regression model to examine the relationship between multiple network embeddedness and organizational absorptive capacity. Based on the variable settings, the following regression model is established to test the three research hypotheses of this study. Model 0 is the regression model of control variables and explained variables; Model 1 is used to test hypothesis H1, Model 2 is used to test hypothesis H2, and Model 3 is used to test hypothesis H3. At the same time, Model 3 is also the full variable model of this study.

$$Ab{s}_t=f(Ma{t}_t,Ps{t}_t,Ag{e}_t,CPTr{e}_t,CPRs{h}_t,CPTs{h}_t,KLDs{h}_t)$$

(2)

$$Ab{s}_t=f(Ma{t}_t,Ps{t}_t,Ag{e}_t,CPTr{e}_t,CPRs{h}_t,CPTs{h}_t,KLDs{h}_t,CPRr{e}_{t-3})$$

(3)

$$Ab{s}_t=f(Ma{t}_t,Ps{t}_t,Ag{e}_t,CPTr{e}_t,CPRs{h}_t,CPTs{h}_t,KLDs{h}_t,KLDr{e}_{t-3})$$

(4)

$$\begin{array}{l}Ab{s}_t=f(Ma{t}_t,Ps{t}_t,Ag{e}_t,CPTr{e}_t,CPRs{h}_t,CPTs{h}_t,KLDs{h}_t,\\ CPRr{e}_{t-3},KLDr{e}_{t-3},CPRr{e}_{t-3}\ast KLDr{e}_{t-3})\end{array}$$

(5)

3.4.2. Model Checking

The data for this study comprise unbalanced panel data. Prior to conducting regression analysis, the Hausman test is employed to assess individual fixed effects between multiple network embeddedness and organizational absorptive capacity, thereby unifying the fixed effects and random-effects models. The significance of the difference in parameter estimators is tested, with results presented in

Table 1. Hausman test results of regression analysis of multiple network embeddedness and organizational absorptive capacity.

Variable	Model 0		Model 1		Model 2		Model 3
	FE	RE	FE	RE	FE	RE	FE	RE
Relational Embeddedness in the Cooperation Network			0.044 **	0.034 *			0.072 ***	0.064 ***
			(0.022)	(0.020)			(0.025)	(0.023)
Relational Embeddedness in the Knowledge Network					0.039 **	0.035 **	0.041 **	0.042 **
					(0.017)	(0.016)	(0.017)	(0.016)
Relational Embeddedness in the Cooperation Network × Relational Embeddedness in the Knowledge Network							−0.034 **	−0.031 **
							(0.016)	(0.015)
Market Attractiveness	0.011 *	0.013 **	0.010 *	0.012 *	0.014 **	0.014 **	0.010	0.012 *
	(0.006)	(0.006)	(0.006)	(0.006)	(0.006)	(0.006)	(0.006)	(0.006)
Patent Stock	0.000	0.021	0.001	0.023	0.001	0.019	0.005	0.026
	(0.022)	(0.022)	(0.022)	(0.022)	(0.022)	(0.022)	(0.022)	(0.022)
Age	0.117 **	0.099 ***	0.148 ***	0.108 ***	0.071	0.087 ***	0.137 **	0.106 ***
	(0.054)	(0.031)	(0.056)	(0.032)	(0.057)	(0.032)	(0.061)	(0.032)
Relational Embeddedness in the Competition Network	0.019	0.041 ***	0.007	0.030 *	0.010	0.031 *	−0.016	0.007
	(0.017)	(0.015)	(0.018)	(0.017)	(0.018)	(0.016)	(0.019)	(0.018)
Structural Embeddedness in the Cooperation Network	−0.009	−0.007	−0.011	−0.009	−0.009	−0.006	−0.012	−0.010
	(0.008)	(0.008)	(0.008)	(0.008)	(0.008)	(0.008)	(0.008)	(0.008)
Structural Embeddedness in the Competition Network	0.012	0.011	0.014	0.012	0.014	0.014	0.016	0.015
	(0.018)	(0.018)	(0.017)	(0.018)	(0.017)	(0.017)	(0.017)	(0.017)
Structural Embeddedness in the Knowledge Network	−0.015	−0.014	−0.015	−0.014	−0.023	−0.023	−0.026	−0.024
	(0.016)	(0.015)	(0.015)	(0.015)	(0.016)	(0.016)	(0.016)	(0.016)
Constant	−0.224 ***	−0.199 ***	−0.265 ***	−0.208 ***	−0.153 **	−0.179 ***	−0.229 ***	−0.188 ***
	(0.071)	(0.053)	(0.073)	(0.053)	(0.077)	(0.054)	(0.080)	(0.054)
Chi2	27.55		31.85		28.92		33
Prob > chi2	0.00		0.00		0.00		0.00

The standard error is shown in brackets. *** p < 0.01, ** p < 0.05, and * p < 0.1.

. The null hypothesis of the Hausman test is that there is no significant difference between the parameters obtained by the two methods [57]. Prob > chi2 in the test results indicates the probability of making a true rejection (Type I) error by rejecting the null hypothesis. In other words, the smaller the probability, the more likely you are to reject the null hypothesis. The p-values in the Hausman test results for Models 1, 2, and 3 are all less than 0.1, leading to the rejection of the null hypothesis. This indicates a significant difference in coefficient estimates between the fixed-effects model and the random-effects model, suggesting that the fixed-effects model should be used for the subsequent regression analysis. 

4. Analysis and Results

4.1. Descriptive Statistics and Correlation Analysis

Table 2. Descriptive statistics and correlation coefficient.

Variable	Mean	SD	1	2	3	4	5	6	7	8	9	10
1. Absorptive Capacity	0.0285	0.0592	1
2. Relational Embeddedness in the Cooperation Network	15.57	43.06	0.170 ***	1
3. Relational Embeddedness in the Knowledge Network	40.36	38.26	0.067 ***	0.271 ***	1
4. Market Attractiveness	51.68	26.80	0.071 ***	0.0250	0.042 **	1
5. Patent Stock	7535	34,989	−0.002	0.005	0.021	0.107 ***	1
6. Age	34.11	27.06	−0.080 ***	0.134 ***	0.065 ***	0.023 **	0.015	1
7. Relational Embeddedness in the Competition Network	434.4	596.3	0.100 ***	0.712 ***	0.302 ***	0.028	0.070 ***	0.354 ***	1
8. Structural Embeddedness in the Cooperation Network	1.484	0.349	0.263 ***	0.239 ***	−0.020	0.037	0.057 **	0.398 ***	0.370 ***	1
9. Structural Embeddedness in the Competition Network	1.944	0.107	0.029	0.05 1 **	0.270 ***	0.025	0.012	0.118 ***	0.193 ***	0.099 ***	1
10. Structural Embeddedness in the Knowledge Network	1.585	0.358	0.076 ***	0.192 ***	0.440 ***	0.031 *	−0.015	0.131 ***	0.282 ***	0.024	0.544 ***	1

*** p < 0.01, ** p < 0.05, and * p < 0.1.

presents the descriptive statistics, including the mean, standard deviation, maximum, and minimum values of key variables, as well as the bivariate Pearson correlation coefficients. The correlation coefficient matrix reveals a significant positive correlation between CPRre and Absorptive Capacity, as well as between KLDre and Absorptive Capacity. Additionally, all correlation coefficients among the variables are below 0.8 [58], indicating no significant multicollinearity issues, thus allowing inclusion in the regression model for further analysis.

4.2. Empirical Results

To explore the impact of multiple network embeddedness characteristics on organizational absorptive capacity, this study employed a fixed effects linear regression method to empirically test three hypotheses using unbalanced panel data from 897 listed companies in the automobile and components industry between 2008 and 2017. First, control variables such as Mat, Ps, Age, CPTre, CPRse, and KLDse are included in the model. Next, CPRre and KLDre are sequentially added to test their direct effects on Abs. Finally, the interaction term of CPRre and KLDre is included to examine the moderating effect of KLDre on the relationship between CPRre and Abs. Based on the aforementioned steps, we established four models: Model 0, Model 1, Model 2, and Model 3.

The regression results are displayed in

Table 3. Regression analysis results of multiple network embeddedness and absorptive capacity.

Variable	Absorptive Capacity
	Model 0	Model 1	Model 2	Model 3
Market Attractiveness	0.011 *	0.010 *	0.014 **	0.010
	(0.006)	(0.006)	(0.006)	(0.006)
Patent Stock	0.000	0.001	0.001	0.005
	(0.022)	(0.022)	(0.022)	(0.022)
Age	0.117 **	0.148 ***	0.071	0.137 **
	(0.054)	(0.056)	(0.057)	(0.061)
Relational Embeddedness in the Competition Network	0.019	0.007	0.010	−0.016
	(0.017)	(0.018)	(0.018)	(0.019)
Structural Embeddedness in the Cooperation Network	−0.009	−0.011	−0.009	−0.012
	(0.008)	(0.008)	(0.008)	(0.008)
Structural Embeddedness in the Competition Network	0.012	0.014	0.014	0.016
	(0.018)	(0.017)	(0.017)	(0.017)
Structural Embeddedness in the Knowledge Network	−0.015	−0.015	−0.023	−0.026
	(0.016)	(0.015)	(0.016)	(0.016)
Relational Embeddedness in the Cooperation Network		0.044 **		0.072 ***
		(0.022)		(0.025)
Relational Embeddedness in the Knowledge Network			0.039 **	0.041 **
			(0.017)	(0.017)
Relational Embeddedness in the Cooperation Network × Relational Embeddedness in the Knowledge Network				−0.034 **
				(0.016)
Constant	−0.224 ***	−0.265 ***	−0.153 **	−0.229 ***
	(0.071)	(0.073)	(0.077)	(0.080)
Number of observations	675	675	675	675
Number of groups	148	148	148	148
R-sq	0.027	0.034	0.037	0.054
p	0.046	0.019	0.012	0.001

The standard error is shown in brackets. *** p < 0.01, ** p < 0.05, and * p < 0.1.

, where β denotes the regression coefficient. Model 0 estimates the effect of control variables on tissue absorptive capacity. Subsequently, CPRre and KLDre were added to Model 1 and Model 2, respectively, to test hypotheses H1 and H2. The results indicate a significant increase in the models’ R-squared values (from 0.027 to 0.034 and 0.037, respectively) after adding these variables. CPRre positively impacts the organization’s absorptive capacity (β = 0.044, p < 0.1), as does KLDre (β = 0.039, p < 0.1), thus verifying hypotheses H1 and H2. Additionally, Model 3 incorporated the interaction term of CPRre and KLDre to test H3. The results show a significant improvement in the R-squared value (to 0.054), with a negative and significant impact coefficient for the interaction term (β = −0.034, p < 0.1), confirming hypothesis H3.

To clarify the moderating effect of KLDre on the relationship between CPRre and Abs, it is illustrated in a diagram. Figure 1 depicts this interaction: the solid line labeled “Low KLDre” represents the relationship when KLDre is one standard deviation below its mean, while the dotted line labeled “High KLDre” represents the relationship when set to one standard deviation above its mean. The slope of the solid line is steeper than that of the dotted line, indicating that the relational embeddedness within the knowledge network negatively moderates the relationship between the relational embeddedness within the cooperation network and organizational absorptive capacity. This supports hypothesis H3, which states that higher relational embeddedness within the knowledge network weakens the positive effect of the relational embeddedness within the cooperation network on organizational absorptive capacity.

5. Robustness Testing

To mitigate the potential interference of reverse causality, the network characteristic-related variables in both the independent and control variables were lagged by three periods. The results indicate that the embeddedness of cooperation network relationships has a significant positive effect on absorptive capacity. To address the endogeneity issues arising from omitted variables, this study employed a fixed-effects model. Specifically, in addition to control variables such as market attractiveness, patent stock, and age, we also included other network embeddedness characteristics as control variables. The results continue to support the positive impact of cooperative network relational embeddedness on absorptive capacity. Considering that different model settings and potential endogeneity issues arising from sample selection may affect empirical results, this study employs various approaches to conduct robustness checks on the relationship between multiple network embeddedness and organizational absorptive capacity.

5.1. The Heckman Two-Stage Model

Considering the strategic differences and distinctive characteristics of companies, the decision to join a cooperative network is not random but rather a comprehensive one that involves factors such as market expansion and organizational practices. To address the endogeneity problem caused by sample selection bias, this study employs the Heckman two-stage model. In the first stage, we generate a dummy variable indicating whether a company joins a cooperation network and use it as the dependent variable in the selection equation. We then add number of companies in the group, patent stock and age to the selection equation, and estimate the probability of a company joining the cooperation network using a Probit model. The Inverse Mills Ratio (IMR) is also calculated at this stage. In the second stage, the IMR is included in the regression model excluding the control variables related to network embeddedness. The estimation results are shown in

Table 4. Estimation results of the Heckman two-stage model.

Variable	Absorptive Capacity
Relational Embeddedness in the Cooperation Network	0.152 ***
	(0.018)
Market Attractiveness	0.032 **
	(0.015)
Patent Stock	0.026
	(0.032)
Age	0.223 ***
	(0.058)
IMR	0.270 ***
	(0.104)
Constant	−0.655 ***
	(0.191)
Number of observations	5714
Selected	736
Non-selected	4978
p	0.000

The standard error is shown in brackets. *** p < 0.01, ** p < 0.05.

. The coefficients for cooperative network relational embeddedness and the IMR in the regression results are 0.15 (p < 0.01) and 0.27 (p < 0.01), respectively. These results indicate that, after controlling for sample selection bias, the embeddedness of a company’s cooperation network relationship continues to significantly promote its absorptive capacity.

5.2. Propensity Score Matching

To further verify the impact of companies joining the cooperation network on absorptive capacity, we employ the propensity score matching (PSM) method. Using firm characteristics such as market attractiveness and firm age, we calculate the propensity score for each firm using a Logit model. The companies are then matched 1:1 using the nearest neighbor matching method. After matching, only the samples with non-zero weights are used for regression analysis. The estimation results are shown in

Table 5. Estimation results of propensity score matching.

Variable	Absorptive Capacity
Relational Embeddedness in the Cooperation Network	0.100 **
	(0.048)
Market Attractiveness	0.010
	(0.012)
Patent Stock	0.008
	(0.045)
Age	0.221 **
	(0.100)
Relational Embeddedness in the Competition Network	0.067 *
	(0.040)
Structural Embeddedness in the Cooperation Network	−0.005
	(0.011)
Structural Embeddedness in the Competition Network	0.020
	(0.041)
Structural Embeddedness in the Knowledge Network	−0.021
	(0.021)
Constant	−0.334 ***
	(0.082)
Number of observations	250
Number of groups	112
p	0.019

The standard error is shown in brackets. *** p < 0.01, ** p < 0.05, and * p < 0.1.

. The regression coefficient for the embeddedness of cooperative network relationships is 0.10 (p < 0.05), indicating that companies joining the cooperation network have a significant positive impact on their absorptive capacity.

5.3. Alternative Model Testing

The third robustness testing method employed in this study involves a replacement econometric analysis using a random-effects model to empirically examine the relationship between multiple network embeddedness and organizational absorptive capacity, as shown in

Table 6. Robustness test results using random-effects model.

Variable	Absorptive Capacity
	Model 0	Model 1	Model 2	Model 3
Market Attractiveness	0.013 **	0.012 *	0.014 **	0.012 *
	(0.006)	(0.006)	(0.006)	(0.006)
Patent Stock	0.021	0.023	0.019	0.026
	(0.022)	(0.022)	(0.022)	(0.022)
Age	0.099 ***	0.108 ***	0.087 ***	0.106 ***
	(0.031)	(0.032)	(0.032)	(0.032)
Relational Embeddedness in the Competition Network	0.041***	0.030 *	0.031 *	0.007
	(0.015)	(0.017)	(0.016)	(0.018)
Structural Embeddedness in the Cooperation Network	−0.007	−0.009	−0.006	−0.010
	(0.008)	(0.008)	(0.008)	(0.008)
Structural Embeddedness in the Competition Network	0.011	0.012	0.014	0.015
	(0.018)	(0.018)	(0.017)	(0.017)
Structural Embeddedness in the Knowledge Network	−0.014	−0.014	−0.023	−0.024
	(0.015)	(0.015)	(0.016)	(0.016)
Relational Embeddedness in the Cooperation Network		0.034*		0.064 ***
		(0.020)		(0.023)
Relational Embeddedness in the Knowledge Network			0.035 **	0.042 **
			(0.016)	(0.016)
Relational Embeddedness in the Cooperation Network × Relational Embeddedness in the Knowledge Network				−0.031 **
				(0.015)
Constant	−0.199 ***	−0.208 ***	−0.179 ***	−0.188 ***
	(0.053)	(0.053)	(0.054)	(0.054)
Number of observations	675	675	675	675
Number of groups	148	148	148	148
p	0.000	0.000	0.000	0.000

The standard error is shown in brackets. *** p < 0.01, ** p < 0.05, and * p < 0.1.

. Model 0 represents the regression between control variables and Abs. Models 2 and 3 add CPRre and KLDre, respectively, while Model 4 incorporates both variables along with their interaction term. The results indicate that in the random-effects model, the regression coefficient for CPRre in Model 1 is significant and greater than zero (β = 0.034, p < 0.1), supporting hypothesis H1. Similarly, the coefficient for KLDre in Model 2 is significant (β = 0.035, p < 0.1), reinforcing hypothesis H2. In Model 3, the coefficient for the interaction term of CPRre and KLDre is negative and significant (β = −0.031, p < 0.1), while both individual coefficients remain significant, thus supporting hypothesis H3.

5.4. Region Fixed Effects

The fourth robustness testing method employed in this study supplements regional variables by using the countries and regions of listed companies in the automobile and components industry as dummy variables, replacing individual fixed effects. The results are presented in

Table 7. Robustness test results for the supplementary variable of region fixed effects.

Variable	Absorptive Capacity
	Model 0	Model 1	Model 2	Model 3
Market Attractiveness	0.013 **	0.013 **	0.015 **	0.013 **
	(0.006)	(0.006)	(0.006)	(0.006)
Patent Stock	0.017	0.019	0.015	0.021
	(0.021)	(0.021)	(0.021)	(0.021)
Age	0.050	0.059 *	0.030	0.050
	(0.034)	(0.034)	(0.035)	(0.035)
Relational Embeddedness in the Competition Network	0.039 **	0.027 *	0.026 *	0.004
	(0.015)	(0.017)	(0.016)	(0.018)
Structural Embeddedness in the Cooperation Network	−0.008	−0.010	−0.008	−0.011
	(0.008)	(0.008)	(0.008)	(0.008)
Structural Embeddedness in the Competition Network	0.010	0.011	0.013	0.014
	(0.017)	(0.017)	(0.017)	(0.017)
Structural Embeddedness in the Knowledge Network	−0.010	−0.010	−0.021	−0.022
	(0.015)	(0.015)	(0.016)	(0.015)
Region Fixed	Yes	Yes	Yes	Yes
Relational Embeddedness in the Cooperation Network		0.032 *		0.060 ***
		(0.019)		(0.023)
Relational Embeddedness in the Knowledge Network			0.043 ***	0.048 ***
			(0.016)	(0.016)
Relational Embeddedness in the Cooperation Network × Relational Embeddedness in the Knowledge Network				−0.028 *
				(0.015)
Constant	0.003	−0.005	0.045	0.037
	(0.436)	(0.437)	(0.435)	(0.437)
Number of observations	675	675	675	675
Number of groups	148	148	148	148
p	0.000	0.000	0.000	0.000

The standard error is shown in brackets. *** p < 0.01, ** p < 0.05, and * p < 0.1.

. Model 0 incorporates the original control variables and the region dummy variable (region fixed) to regress the organization’s absorptive capacity. Models 1 and 2 add CPRre and KLDre, respectively. Model 3 is a full variable model that includes interaction terms for both types of embeddedness. The regression coefficient for CPRre in Model 1 is significant and positive (β = 0.032, p < 0.1), supporting hypothesis H1. In Model 2, the coefficient for KLDre is also significant and positive (β = 0.043, p < 0.1), reinforcing hypothesis H2. Model 3 shows a significant and negative coefficient for the interaction term of both types of embeddedness (β = −0.028, p < 0.1), while both individual coefficients remain significant, thereby supporting hypothesis H3.

5.5. Legal Form Fixed Effects

The fifth robustness testing method employed in this study supplements the legal form variable by using the registered legal form of listed companies in the automobile and components industry as a dummy variable, replacing individual fixed effects to control the model. The results are presented in

Table 8. Robustness test results for the supplementary variable of legal form fixed effects.

Variable	Absorptive Capacity
	Model 0	Model 1	Model 2	Model 3
Market Attractiveness	0.013 **	0.012 **	0.015 **	0.012 **
	(0.006)	(0.006)	(0.006)	(0.006)
Patent Stock	0.015	0.017	0.013	0.019
	(0.021)	(0.021)	(0.021)	(0.021)
Age	0.078 **	0.086 ***	0.065 **	0.083 ***
	(0.030)	(0.031)	(0.031)	(0.031)
Relational Embeddedness in the Competition Network	0.036 **	0.025	0.025	0.002
	(0.015)	(0.017)	(0.016)	(0.018)
Structural Embeddedness in the Cooperation Network	−0.007	−0.009	−0.007	−0.010
	(0.008)	(0.008)	(0.008)	(0.008)
Structural Embeddedness in the Competition Network	0.011	0.012	0.013	0.014
	(0.017)	(0.017)	(0.017)	(0.017)
Structural Embeddedness in the Knowledge Network	−0.012	−0.011	−0.021	−0.023
	(0.015)	(0.015)	(0.016)	(0.016)
Legal Form Fixed	Yes	Yes	Yes	Yes
Relational Embeddedness in the Cooperation Network		0.033 *		0.063 ***
		(0.019)		(0.023)
Relational Embeddedness in the Knowledge Network			0.038 **	0.044 ***
			(0.016)	(0.016)
Relational Embeddedness in the Cooperation Network × Relational Embeddedness in the Knowledge Network				−0.030 **
				(0.015)
Constant	−0.225 ***	−0.233 ***	−0.204 ***	−0.212 ***
	(0.051)	(0.051)	(0.051)	(0.052)
Number of observations	675	675	675	675
Number of groups	148	148	148	148
p	0.000	0.000	0.000	0.000

The standard error is shown in brackets. *** p < 0.01, ** p < 0.05, and * p < 0.1.

. Model 0 includes the original control variables and the legal form dummy variable (legal form fixed) to assess Abs. Models 1 and 2 add CPRre and KLDre, respectively. Model 3 is a comprehensive model that incorporates CPRre and KLDre along with their interaction. The results indicate that the coefficient for CPRre in Model 1 is significant and positive (β = 0.033, p < 0.1), supporting hypothesis H1. In Model 2, the coefficient for KLDre is also significant and positive (β = 0.038, p < 0.1), reinforcing hypothesis H2. Additionally, the coefficient for the interaction term in Model 3 is significant and negative (β = −0.030, p < 0.1), while both CPRre and KLDre coefficients remain significant, further supporting hypothesis H3.

5.6. Heterogeneity Analysis

To further examine how the impact of explanatory variables on absorptive capacity varies across different internal and external characteristics, we conducted two subgroup regressions based on firm size (measured by employee count) and whether the company is located in an emerging economy. The estimation results are shown in

Table 9. Estimation results of heterogeneity analysis.

Variable	Internal		External
	Number of Employees (≤10,000)	Number of Employees (>10,000)	Developed Economy	Emerging Economy
Relational Embeddedness in the Cooperation Network	0.0757	0.0585 *	0.0573 **	0.0439
	(1.27)	(1.67)	(2.19)	(1.07)
Relational Embeddedness in the Knowledge Network	0.0519 ***	0.0478	0.0427 **	0.0196
	(2.64)	(0.88)	(2.18)	(0.72)
Market Attractiveness	0.00187	0.0151	0.0105	0.0142
	(0.22)	(1.21)	(1.49)	(1.34)
Patent Stock	0.0153	−0.0229	0.00432	0.0175
	(0.69)	(−0.41)	(0.19)	(0.05)
Age	0.0805	0.0624	0.0836	0.295 ***
	(1.08)	(0.43)	(1.23)	(2.75)
Relational Embeddedness in the Competition Network	−0.059	−0.0259	−0.0197	0.0973 ***
	(−1.42)	(−0.78)	(−0.91)	(2.68)
Structural Embeddedness in the Cooperation Network	−0.00592	−0.0236	−0.0157	−0.00382
	(−0.50)	(−1.13)	(−1.46)	(−0.52)
Structural Embeddedness in the Competition Network	0.0157	0.221	0.0168	0.0516
	(1.00)	(0.68)	(0.87)	(0.79)
Structural Embeddedness in the Knowledge Network	−0.0266	0.0123	−0.0244	−0.026
	(−1.62)	(0.16)	(−1.25)	(−1.31)
Constant	−0.217 **	−0.0499	−0.131	−0.365 ***
	(−2.21)	(−0.19)	(−1.25)	(−6.09)
R-sq	0.0524	0.0451	0.0362	0.407
Number of observations	295	276	570	105

The standard error is shown in brackets. *** p < 0.01, ** p < 0.05, and * p < 0.1.

.

Heterogeneity of Internal Characteristics of Companies: To examine the heterogeneity in internal characteristics, we created a dummy variable based on company size, setting companies with more than 10,000 employees count to 1 and the others to 0. We then performed inter-group comparisons in the regression model. The results show that the relational embeddedness of the cooperation network has a significant effect only in companies with more than 10,000 employees (β = 0.0585 *, p < 0.1), whereas it is not significant in companies with fewer than 10,000 employees. This may be because large companies typically have more stable resources and more sophisticated management systems, allowing them to better utilize information and resources from cooperation networks to enhance their absorptive capacity. In contrast, small companies may face resource and management constraints, which limit the effectiveness of network embeddedness on their absorptive capacity. On the other hand, the relational embeddedness of the knowledge network is significant only in firms with fewer than 10,000 employees (β = 0.0519 ***, p < 0.01), while it is not significant in firms with more than 10,000 employees. This could be because larger companies tend to rely on complex knowledge systems for knowledge integration, whereas smaller companies focus on core knowledge elements to improve their competitive advantage. In this context, knowledge elements with high centrality in the knowledge portfolio are particularly valuable for smaller firms, helping them absorb, transform, and utilize knowledge more effectively.

Heterogeneity of External Environmental Characteristics of Enterprises: We categorize the regions of the sample companies into emerging economies and developed economies, and then conduct inter-group comparisons in the regression models. The results show that both the relational embeddedness of the cooperation network and the relational embeddedness of the knowledge network are significant in companies from developed economies (β = 0.0573 ** and β = 0.0427 **, respectively), but not in firms from emerging economies. We can try to explore the reasons from two aspects. First, cultural and institutional differences play a key role. Companies in developed economies benefit from a more transparent and stable institutional environment, where strong institutional support for cooperation promotes absorptive capacity by enhancing the effectiveness of collaborative efforts. In contrast, the less developed institutional environment in emerging economies can restrict the flow of knowledge through company cooperation, thereby diminishing the impact of cooperation network embeddedness on absorptive capacity. Second, differences in the maturity of knowledge networks also contribute to this disparity. Developed economies typically have well-established knowledge networks, facilitating efficient knowledge transfer between companies and supporting the absorption and integration of external knowledge. On the other hand, knowledge networks in emerging economies tend to be less mature, with companies often lacking effective mechanisms for knowledge integration. As a result, companies in emerging economies struggle to fully absorb and utilize external knowledge, making the impact of knowledge network embeddedness on their absorptive capacity less significant.

6. Discussion

This study empirically examines the relationship between company embeddedness characteristics in multiple networks and organizational absorptive capacity, along with the interactive effects of knowledge network and cooperation network embeddedness on absorptive capacity. Specifically, based on social network theory, resource-based view, and knowledge-based view, the research investigates: (1) the positive impact of cooperation network embeddedness on organizational absorptive capacity, (2) the positive impact of knowledge network embeddedness on absorptive capacity, and (3) the boundary conditions affecting the relationship between cooperation network embeddedness and absorptive capacity within the context of multiple network embeddedness. This comprehensive analysis aims to deepen the understanding of how multiple network embeddedness characteristics influence organizational absorptive capacity.

The finding that the relational embeddedness within the cooperation network enhances their absorptive capacity aligns with previous research suggesting that cooperative R&D models can improve technology and knowledge absorptive capacity. This study further investigates the catalytic mechanism of a company’s central position within the cooperation network on absorptive capacity. Companies with high cooperation network embeddedness can leverage their positional advantages to gather substantial information, facilitating the identification of potential knowledge and opportunities in the external environment, thereby enhancing their absorptive capacity. Additionally, a strong node position in the network fosters significant knowledge flow, directly promoting the internalization of external knowledge, increasing the company’s knowledge stock, and transforming it into proprietary knowledge. Thus, through cooperation networks, the influx of external knowledge and technology effectively enhances a company’s ability to identify and internalize both internal and external knowledge, catalyzing their absorptive capacity.

The finding that the relational embeddedness within the knowledge network enhances their absorptive capacity aligns with the view that companies foster knowledge transformation and improve innovation performance through external and local searches, from the perspectives of knowledge portfolio and organizational learning. This study quantifies the relationship between company knowledge characteristics and organizational absorptive capacity through a network lens. High relational embeddedness in the knowledge network suggests that a company’s knowledge portfolio comprises more fundamental knowledge elements, which are often linked to key technologies within a specific field and drive the internalization and transformation of knowledge. Furthermore, greater knowledge network relational embeddedness signifies higher compatibility among knowledge elements in the company’s knowledge portfolio. These compatible knowledge elements can be easily recombined with others, facilitating the integration of externally acquired knowledge with the internal knowledge stock. By maintaining a robust collection of basic and compatible knowledge elements, companies can more effectively internalize and transform knowledge, thereby enhancing their absorptive capacity.

The finding that a company’s relational embeddedness in the knowledge network inhibits the positive effect of its cooperation network relational embeddedness on absorptive capacity. Specifically, while the relational embeddedness within the knowledge network positively impacts organizational absorptive capacity, its interaction with the relational embeddedness within the cooperation network has a negative effect. This suggests that the two forms of embeddedness overlap in their influence on improving absorptive capacity, indicating a substitution effect. The potential for combining knowledge elements within the knowledge network is socially constructed through the collaborative behaviors of R&D personnel, who actively mine, combine, and transform knowledge elements in the company’s knowledge portfolio based on their insights. This process effectively substitutes for the informational advantages derived from the company’s central position in the cooperation network. Consequently, the relational embeddedness within the knowledge network exerts an inhibitory moderating effect on the relationship between the relational embeddedness within the cooperation network and absorptive capacity, representing the substitution mechanism of the relational embeddedness within the knowledge network over the relational embeddedness within the cooperation network in enhancing absorptive capacity.

This study makes several important theoretical contributions to the literature on network embeddedness and absorptive capacity. First, while previous studies have focused on the independent effects of a single type of network embeddedness on absorptive capacity, our study extends the existing research perspective by examining the interactive effects between these two types of network embeddedness. Specifically, our results highlight that while both types of network relational embeddedness have a positive impact on absorptive capacity, relational embeddedness in knowledge networks may, in some cases, limit the positive effects of the relational embeddedness within the cooperation network on absorptive capacity. This finding challenges the traditional view that all forms of network embeddedness work together in a complementary manner [59,60]. Second, previous studies have typically focused on the effects of either cooperation networks or knowledge networks [61,62,63]. The multi-network perspective introduced in our study aligns with recent calls for more nuanced models of network dynamics in organizational research [64,65]. Third, our study builds on the view that absorptive capacity is not only influenced by a company’s internal characteristics but is also significantly affected by its relationship with the external environment [3], thereby enriching our understanding of how firms leverage external network relationships to enhance their dynamic capabilities [53].

The implications of these findings for managers are that they should strategically capitalize on the organization’s unique position within multiple networks to enhance absorptive capacity. Companies should effectively leverage the benefits of internal and external learning. When knowledge elements within a company’s knowledge portfolio exhibit strong relational embeddedness, they typically represent foundational knowledge that integrates well with other elements, potentially enhancing the development of key technologies. However, in such cases, knowledge and resources acquired from cooperation networks may be challenging to incorporate and absorb. By strategically managing the relational embeddedness levels in both networks, companies can optimize their learning advantages, achieving greater improvements in absorptive capacity with lower investment. Specifically, company should strengthen their relational embeddedness in cooperation networks to better acquire and integrate external knowledge. However, they must also recognize that excessive embeddedness in knowledge networks may restrict information and knowledge flows, thereby hindering absorptive capacity. Therefore, managers should strike a balance in network relationships and adapt their strategies to maintain a favorable position in collaborative networks while avoiding the potential drawbacks of over-reliance on specific networks. To achieve this balance, companies could engage in cross-sector collaborations, participate in interdisciplinary platforms, and adopt flexible strategies, which facilitate dynamically adjusting their central position in both networks. By doing so, managers can ensure that their company remains capable of absorbing new knowledge while avoiding the risks associated with excessive network embeddedness.

7. Conclusions

This study developed a relationship model linking multiple network embeddedness to organizational absorptive capacity, grounded in social network theory and the knowledge-based view. After analyzing the existing literature, an empirical test of the model was conducted. The results were further discussed, highlighting the catalytic mechanisms of the relational embeddedness within the cooperation network and the impact of the relational embeddedness within the knowledge network on organizational absorptive capacity, as well as the alternative mechanisms by which both types of embeddedness enhance absorptive capacity.

First, this study highlights the dual sources of technology, knowledge, and information acquisition for companies: the cooperation network and the knowledge network. It examines how the relational embeddedness within these networks influences organizations’ absorptive capacity, elucidating the process by which companies enhance their identification, acquisition, absorption, transformation, and effective utilization of knowledge through multiple network embeddedness. Second, it distinguishes the mechanisms by which the relational embeddedness within the cooperation and knowledge networks affects absorptive capacity. Specifically, embeddedness in cooperative networks enhances organizations’ absorptive capacity by improving their ability to recognize and acquire external knowledge, while also facilitating the transformation and absorption of internal and external knowledge through the compatibility of knowledge elements within their knowledge portfolio. Finally, this study posits that relational embeddedness in the knowledge network increases companies’ understanding of the characteristics and compatibility of their knowledge elements, prompting them to explore combinations of existing knowledge. This exploration overlaps with the extensive knowledge search in cooperation networks, suggesting that knowledge network embeddedness may inhibit the enhancement of absorptive capacity fostered by cooperation network embeddedness.

While this research has achieved some results and offers some implications for management theory and practice, there are still some limitations that require improvement in future research. First, to more effectively demonstrate the impact of network embeddedness on a company’s absorptive capacity, this study focuses on the automobile and components industry, a sector in the process of moving from traditional fields to emerging fields such as intelligence and cleanliness, which may limit the external validity of our findings. Future research should consider a broader range of sample sources, including both traditional industries and emerging industries, to enhance the universality and generalizability of the results. Second, this study collected data for empirical testing of the theoretical model by matching and integrating multiple database resources. Although this approach has improved data reliability, it has also introduced challenges regarding data applicability, potentially hindering our comprehensive understanding of the impact process especially in terms of the measurement and formation mechanism of absorptive capacity. Future research could conduct more qualitative explorations of specific cases to provide deeper insights. Additionally, we primarily examined the linear relationship between network embeddedness and absorptive capacity. However, the actual situation is likely more complex, as network relationships may exhibit multiple embeddedness or interactive effects. Future research should investigate how different types of networks influence each other, thereby providing a more comprehensive understanding of their impact mechanisms on absorptive capacity.