Inter-Organizational Connectivity, Digital Transformation, and Firm Ambidextrous Innovation: A Coupled Perspective on Innovation Ecosystems and Digitalization

Abstract

In the context of the explosive growth of the digital economy, how inter-organizational connectivity affects corporate ambidextrous innovation has emerged as a pressing issue in the current digital economy. Based on the perspectives of the innovation ecosystem and digital coupling, this paper explores the inner mechanism of this issue through structural modeling by using the data of China’s high-tech enterprise alliance cooperation from 2015 to 2022. It is found in the empirical study that the local efficiency and reach rate of the digital innovation ecosystem have an inverted U-shaped relationship with exploratory innovation, and the local efficiency and reach rate of the digital innovation ecosystem have a negative effect on firm exploitative innovation. In addition, the level of firms’ digital transformation mediates the relationship between the local efficiency, reach rate, and ambidextrous innovation. The level of market development plays a moderating role in the relationship between the local efficiency, reach rate, and ambidextrous innovation. The findings provide a theoretical basis for the digital innovation ecosystem to realize the role of a “resource pool” through structural connections, which in turn provides important guidance for the digital transformation and innovation development of high-tech enterprises.

1. Introduction

In the current society, some scholars hold that the digital economy and the real economy are in a state of deep integration [1,2]. The complex and ambiguous relationship between the digital competitive environment, firms’ digital transformation, and corporate innovation has left companies confused about whether to integrate into the digital ecological environment and whether digital transformation is beneficial to their development [2,3]. This is mainly due to the fact that digital technologies, such as AI, are not yet well integrated into the production processes of companies. At the same time, some companies have found digital transformation to be an important prerequisite for innovative development [1,2]. The approach of firm digital transformation mainly is to use technologies such as artificial intelligence technology, blockchain technology, cloud computing technology, big data technology, and digital application technology. Studies have shown that there is a “substitution elasticity” between AI and labor, and the majority of innovation managers believe that AI has great potential to improve the efficiency of innovation activities [4,5,6,7]. Therefore, unlocking the “black box” of the relationship among the digitally embedded ecological environment, firm digital transformation, and firm innovation has important theoretical value for solving how the digital environment affects firm innovation and how to enhance the dynamics of firm digital transformation.

Ecosystem is used to characterize the competitive environment [8,9,10]. In the context of the digital economy, new types of business entities such as digital firms and digital platforms have emerged, and they are integrated into the ecosystem to form a new structure, which is defined as a “digital innovation ecosystem” by some scholars [11]. A digital innovation ecosystem is a dynamic collective of interdependent actors and the resources they draw on to innovate with digital technology [12]. How digital technologies affect the integration of systems [13,14] has rarely been studied directly. How do participants in digital innovation ecosystems strive to integrate to achieve connectivity, and how do digital technologies work? Given that the interactions between a variety of entities are typical features of the ecosystem [15], a digital innovation ecosystem relies on the digital technologies that are used to connect relationships between organizations, acting as a “resource pool” [16,17,18]. Therefore, it is crucial for us to construct a structural characterization of digital firms’ engagement from a network perspective and to investigate how digital innovation ecosystems achieve connectivity and how they rely on structural connectivity to function as a “pool of resources” that can influence firms to achieve successful innovation.

Ambidextrous innovation refers to a firm’s capacity to balance exploitative and exploratory innovation activities, which requires the rational allocation of resources in line with strategic goals [19]. Specifically, exploration focuses on technological breakthroughs and knowledge creation, while exploitation emphasizes the application and refinement of existing knowledge and skills [20]. In practical terms, this means enterprises need to simultaneously invest in developing new technologies or markets (exploration) and optimizing existing products or processes (exploitation) to adapt to dynamic market changes. More studies have focused on exploring the innovation performance of firms [21]. However, there is less evidence on ambidextrous innovation in the context of digitalization. Digital innovation ecosystems are characterized using “resource pools” created by connectivity to meet the resource needs of firms for innovation. Therefore, this study highlights the significant implications of examining the mechanisms that influence ambidextrous innovation from a complex digital innovation ecosystem perspective for firm innovation practices.

The literature lacks a systematic overview of the relationship between digital innovation ecosystems, firm digital transformation, and exploratory and exploitative innovations, despite some studies focusing on knowledge transfer and collaborative competition [9]. However, the moderating effect of firm digital transformation still lacks strong theoretical support and empirical tests. Prior studies suggested that inter-organization connectivity relies on the firm digital technology, as digital technology promotes the motivation to collaborate and enhances the firm’s ability to link with the outside [22,23]. It has also been shown that digital technologies can help firms to collect resources to promote exploratory and exploitative innovation [24]. Therefore, we can see that firm digital transformation can act as a bridge, connecting the connectivity of a digital innovation ecosystem and firm innovation. In addition, there is heterogeneity in the mechanisms by which the embedding of different digital technologies, such as AI, affects firm innovation. However, previous studies have not explored the heterogeneity of different digital technologies in depth. Moreover, few studies explore moderating mechanisms for improving firm digital transformation and innovation. An enabling environment facilitates collaboration among participants within their ecosystem, and innovation ecosystems would benefit from a degree of market development that promotes innovation collaboration [25,26]. Therefore, the research framework should consider the impact of the degree of market development on the relationship between the connectivity of the organization, firm digital transformation, and firm ambidextrous innovation.

Against this backdrop, our study aims to uncover how digital innovation ecosystems act as a “resource pool” through structural connectivity (local efficiency, reach rate) to promote exploratory and exploitative innovation, validate the mediating role of firm digital transformation and the moderating role of market development, and reveal the heterogeneous effects of five digital technologies (AI, blockchain, etc.) on ambidextrous innovation.

This paper is structured as follows: Section 2 presents the literature review and theoretical hypotheses; Section 3 describes the research methodology (sample, variables, model); Section 4 presents the empirical results; Section 5 discusses the findings; Section 6 gives the conclusions and limitations of the paper; Section 7 describes the theoretical contributions and originality; and Section 8 presents the management implications.

2. Literature Review and Theoretical Hypotheses

2.1. The Connectivity of the Digital Innovation Ecosystem and Ambidextrous Innovation

Iansiti and Levien [27] define an ecosystem as a loose network. A network is a com-bination of loosely coupled relationships formed by self-organized cooperation among organizations [28,29]. We measure connectivity in terms of both local efficiency and reach rate. Local efficiency refers to the non-redundant connections between firms and partners [30]. Specifically, the structure of the network composed of the firm and other nodes is not affected after removing a given neighboring node, which quantifies the resistance of the firm to small-scale failures. If a firm is insensitive to small-scale failures, it indicates that the firm has a strong alternative ability and can continuously maintain local connectivity. The reach rate is the set of the number of firms and the path length that a firm can reach [31]. The more nodes a firm can reach, the stronger the connectivity it has and the more knowledge information it can obtain from the other node. Additionally, the completeness of the transferred knowledge is directly related to the path length between organizations [31]. Generally, both local connectivity is maintained and the number and path length of connections is achieved. Therefore, we choose both local efficiency and the reach rate to measure connectivity.

Exploratory innovation is the innovative ability of a firm to explore new possibilities for producing new products or services, while exploitative innovation is the ability of a firm to use old resource to improve products or services progressively [20]. On the one hand, exploratory innovation focuses on pursuing new areas of opportunity for the organization’s development [32,33]. As local efficiency and the reach rate increase, firms are more resistant to failure. Additionally, the connectivity between the firm and digital entities such as digital firms and digital platforms is high. Digitalization promotes inter-organizational cooperation and thereby accelerates knowledge acquisition [34]. External resources stimulate exploratory innovation. Currently, firms are motivated to focus more on exploratory innovation through obtaining heterogeneous resources, instead of reducing exploitative innovation. On the other hand, with the further improvement in local efficiency, the higher stability of local cooperation, and the greater number of nodes that firms can reach leading to path dependence, the enthusiasm of firm’s exploratory innovation will decrease. Considering that exploitative innovation focuses on the use of existing knowledge in search processes [33,35,36], due to the path-dependence at this stage, firms are likely to pursue exploitative innovation by integrating and utilizing old resources.

However, research on traditional innovation ecosystems has mainly focused on research issues related to traditional environments [16,17,37]. For example, some studies have explored the mechanisms of measuring and influencing the ecological niche fitness of innovation ecosystems, the resilience of innovation ecosystems and their evolutionary mechanisms [37], and the synergistic mechanisms of key technologies of core firms in innovation ecosystems [16]. There are relatively few empirical studies on the characteristics of innovation ecosystems in the context of digitization, and even fewer studies from a network perspective. Therefore, combining the digitalization and innovation system network perspectives, this study proposes the following two hypotheses for the impact mechanism of innovation ecosystem connectivity on firms’ ambidextrous innovation in the context of digitalization.

H1a. 

The local efficiency, reach rate of the digital innovation ecosystem, and firm exploratory innovation show an inverted U-shaped relationship.

H1b. 

The local efficiency, reach rate of the digital innovation ecosystem, and firm exploitative innovation show a U-shaped relationship.

2.2. The Mediating Role of the Firm Digital Transformation Level

Firms use digital technologies to transform their production operation systems and management models and core business processes, resulting in disruptive changes, which is the process of digital transformation [38]. When the local efficiency and reach rate of firms in the digital innovation ecosystem are high, it means that there are more connections between traditional organizations and other digital organizations and that the structure is relatively stable, which makes it easier to facilitate the firm’s digital transformation. This process facilitates the digital transformation of firms. Second, digitalization deepens collaboration within the ecosystem [39,40]. Firms with a high local efficiency and reach rate often occupy an important position, which helps firms to drive the digital transformation process and further optimize and solidify their core positions.

Digital technology is not only a production tool and management means for firm managers, but it can also empower firms. First, advances in information technology make it easier for firms to access external heterogeneous resources [41], and digital capabilities promote inter-organization linkages to achieve the effective matching of innovation supply and demand [40]. Second, in the framework of established task models [42], while digitization has an alternative role to routine tasks, digital technologies can liberate non-routine tasks to form the data assets of an organization [43], thus enhancing the efficiency and quantity of access to external heterogeneous resources and then favoring innovation [42]. In short, firms’ high connectivity in digital innovation ecosystems provides a resource space for their digital transformation [44]. At the same time, by collaborating with other participants in the innovation ecosystem, firms can leverage partnerships to increase the path to acquire knowledge [27,45,46], thereby obtaining more external resources for exploratory innovation and exploitative innovation [47].

Notably, the literature lacks a systematic overview of the relationship between digital innovation ecosystems, the digital transformation of firms, and exploratory and exploitative innovation, although some studies focus on knowledge transfer and collaborative competition [9]. More studies have explored the mechanisms by which digital transformation affects firm performance (e.g., [48]), labor income distribution (e.g., [49]), and so on. It has also been shown that digital technologies can help firms collect resources and promote exploratory and exploitative innovation [24]. In addition, there is heterogeneity in the impact mechanisms of the digital transformation of different digital technologies, such as artificial intelligence, on enterprise innovation, and previous studies have not explored the heterogeneity of different digital technologies in depth. Therefore, in order to fill the research gap on the role of digital transformation in bridging between innovation ecosystems and corporate innovation, a four-point mediating role hypothesis is proposed.

H2a. 

The firm digital transformation level mediates the relationship between the local efficiency of the digital innovation ecosystem and its exploratory innovation.

H2b. 

The firm digital transformation level mediates the relationship between the reach rate of the digital innovation ecosystem and its exploratory innovation.

H2c. 

The firm digital transformation level mediates the relationship between the local efficiency of the digital innovation ecosystem and its exploitative innovation.

H2d. 

The firm digital transformation level mediates the relationship between the reach rate of the digital innovation ecosystem and its exploitative innovation.

2.3. The Moderation Effect of Market Development

Markets have always been an important informal institutional environment that influences firms’ behaviors. Marketization represents the degree of market activity [50]. Some authors argued that informal institutions play an important role in innovation ecosystems formed by the free association of firms [51,52], and it can influence the internal operation of ecosystems as participants work together to create value [44,46]. When market development level is high, market competition destabilizes the local co-operation among firms, which leads to the reduction of the local efficiency of the innovation ecosystem, thus discouraging exploratory innovation and exploitative innovation. Second, market activity destabilizes local cooperation while increasing new cooperation [53], and the connections among organizations increase. In addition, the reach rate of inter-organization is enhanced.

The existing literature suggests that an enabling environment favors collaboration among participants within the ecosystem and that innovation ecosystems will benefit from the degree of market development that fosters innovation collaboration [25,26]. Few studies have delved into the moderating mechanisms that improve digital transformation and innovation in firms. Therefore, this paper incorporates the degree of market development into the research framework with the following two hypotheses.

H3a. 

The market development level negatively moderates the relationship between local efficiency and exploratory and exploitative innovation.

H3b. 

The market development level positively moderates the relationship between the reach rate and exploratory and exploitative innovation.

Based on the above ideas, we develop a comprehensive theoretical model of moderating mediators. The independent variables are the local efficiency and reach rate among organizations within a digital innovation ecosystem. The mediating variable is firm digital transformation. In addition, the moderating variable is market development. The conceptual model is shown in Figure 1.

3. Methodology

3.1. Samples and Data

According to the Classification of High-Tech Industries (Manufacturing) (2017), the six categories of pharmaceutical manufacturing; aviation, spacecraft, and equipment manufacturing; electronics and communications equipment manufacturing; computer and office equipment manufacturing; medical instrumentation and apparatus manufacturing; and information chemicals manufacturing together constitute the high-tech manufacturing industry. We chose China’s high-tech manufacturing industry as the object of study for three reasons. First, China had 53,345 patents in 2018, ranking second in the world. However, the innovation performance of China’s high-tech manufacturing industry is not high, with only 42.9% of high-tech manufacturing enterprises [35]. The innovation research of the high-tech manufacturing industry needs to be further strengthened. Second, the innovation of the high-tech manufacturing industry requires continuous R&D investment, which is difficult to be supported by the strength and resources of a single enterprise, and more alliance cooperation will be generated among enterprises; plus, the data of enterprise cooperation in the innovation ecosystem are relatively complete and available. Third, the digital transformation of the high-tech manufacturing industry is more difficult than other industries, and the research can provide insights for the transformation and development of the high-tech manufacturing industry.

To test the hypotheses, this paper collects two types of data: (1) The alliance cooperation data of six high-tech manufacturing industries in China from 2015 to 2022. During the alliance cooperation data collection process, we distinguish between digital enterprises, digital platform enterprises, and high-tech manufacturing enterprises. Based on the existing literature, we define digital enterprises as enterprises that can provide technological tools and are integrators of digital technologies such as intelligent management systems, software, algorithms, and data. Platform enterprise is a new type of economic form in which the platform operator provides differentiated products and services to different subjects on all sides of the platform, thus integrating the resources and relationships between multiple subjects and maximizing the benefits of multiple subjects. Taking pharmaceutical manufacturing as an example, with reference to the alliance cooperation data collection method, the advanced search in Baidu was set to “contain any keywords for cooperation, platform, digitalization, enterprise” and “contain complete keywords for pharmaceutical manufacturing”. With the help of crawler software, the cooperation data from 1 January 2015 to 31 December 2022 were searched year by year. The types of cooperation in network sales and life services are eliminated, and the types of R&D cooperation are retained. Finally, the alliance cooperation network is constructed. According to the existing research method in academia [54], the storage time of a cooperation relationship is generally 3–5 years, and the average survival time of an optional alliance cooperation network is 3 years, so 2015–2022 is divided into six time windows, and a total of 835 enterprises in six industries are collected in 2015–2022, for a total of unbalanced panel data of 5010 sample points. Figure 2 shows the firms’ alliance cooperation networks under the 2015–2017 time window. (2) Considering the availability of basic enterprise data and market development environment data, listed enterprises in the first type of enterprise alliance cooperation network sample are selected as research objects, and enterprise-related data are collected through CSMAR and other databases.

We extracted three types of digital partnerships by identifying inter-organizational relationships in digitally embedded networks, including digital platform-leading partnerships, digital firm-utilizing partnerships, and horizontal cooperation between digital platforms, digital firms, and high-tech manufacturing firms (shown in Figure 3).

Figure 3a shows the digital platform-leading partnership. It is mainly manifested in that the platform dominates the establishment of cooperative relationships with external organizations for regulating united standards and achieves inter-organization synergy by regulating the digital standards. It makes it easier for different players to connect and coordinate through the platform [55]. Therefore, a digital platform firm is defined as a technology platform firm that is built to meet a certain technology standard or to help realize a partner’s technology needs.

Figure 3b is a digital firm-utilizing partnership. It is mainly manifested that high-tech manufacturing firms need to utilize the digital functions of digital firms in order to realize the purpose of digital function embedding. It will provide traditional firms with IT tools, such as intelligent management systems, software, and other IT services [56,57].

Figure 3c is a horizontal cooperative relationship between digital platforms, digital firms, and high-tech manufacturing firms. The platform and the participants establish a horizontal and equal cooperative relationship to jointly improve the platform technology level and for common development purposes. Based on the above digital embedding structure, we construct the cooperation network. Additionally, we use this as a standard to calculate various types of indicators.

3.2. Measures

The dependent variable is ambidextrous innovation. Firm innovation activities can be categorized into exploratory innovation and exploitative innovation. Exploratory innovation represents the innovation of a drastic, fundamental type of activity [58]. Exploitative innovation is the use of existing technology for product improvement. Currently, there is no standardized measure of ambidextrous innovation. Some studies have used patent data to measure ambidextrous innovation by comparing the first four numbers of the IPC classification number in patents. Benner and Tushman [58] used firms’ technological innovation behaviors as a criterion for categorizing the development of a new product or service as an exploratory innovation, whereas product process improvement, quality control improvement, adding product features, and reducing production costs are categorized as exploitative innovation. There was also a study that used firms’ R&D data to measure ambidextrous innovation [59]. According to the document “Firm Accounting Standard No. 6—Intangible Assets” (2006), the investment in R&D projects within firms in China is differentiated into research-state investment and development-state investment. Research-state investment is more inclined toward exploratory expenditures, which have greater risks and outcome uncertainty. Based on the division of this document and referring to Bi [59], we measure the exploratory innovation investment by using the expensed expenditures of the firm’s R&D activities, while the capitalized expenditures of the firm’s R&D activities are used to measure exploitative innovation. Among them, the expenditure on R&D activities of an enterprise refers to the expenditure incurred by the enterprise to carry out R&D activities on new products, new technologies, new processes, and new materials. If an enterprise chooses to expense R&D expenditures, then in the year of R&D, the enterprise can recognize the actual R&D inputs as costs and expenses in the current period’s profit and loss, which has great risks and belongs to exploratory innovation. If it is capitalized, the enterprise is required to recognize the R&D expenditure as an intangible asset when certain conditions are met and to amortize it to profit or loss over the future use period of the intangible asset, which belongs to relatively stable investment and is in line with the low-risk characteristics of exploitative innovation.

The independent variable is the inter-organization connectivity in the digital innovation ecosystem. It has been measured using the ratio of the number of maximum connected subgraphs [31,60]. Several studies have used structural indicators of the network, reach rate [31], and local efficiency [28] to measure network connectivity. The reach rate is the number of firms and path lengths that can be reached by the alliance firms. Local efficiency is the structural connectivity characteristics of the network that remain unchanged after removing a given node and quantifies the network’s resistance to small-scale failures [41]. If the local efficiency of the network is high, it means that it is insensitive to small-scale faults, and the network is resilient. If the local efficiency is low, it indicates sensitivity to small-scale faults, implying that some nodes are too important. Based on existing studies, we believe that connectivity needs to consider not only the local connectivity efficiency but also the number of connections. Therefore, we measure network connectivity in terms of the reach rate and local efficiency.

Local efficiency: Studies have shown that non-redundant linkages between firms and partners will increase their innovation capacity [61]. The formula is as follows [57]:

$${Local Efficiency}_{it}=[\sum _j(1-\sum _q{p}_{iq}{m}_{iq})]/N_i, j\ne q$$

(1)

where j is all the points connected to i, and q is every third party except i or j; is the proportion of the relationship that i puts into q; is the marginal strength of the relationship from j to q. The maximum value is 1 in binary networks. When j is connected to q, =1. Conversely, =0 represents the redundancy between i and j. Ni is the number of coalitional partners connected to i. Its value range is [0, 1].

Reach rate: To obtain the reach rate of the individual network for each time window, both the number of nodes and the path length that a firm can reach are considered [31]. The formula is as follows:

$$Reach=\left[\sum _{i=1}^n\sum _{j=1}^{m}1/d_{ij}\right]/n$$

(2)

is the distance (shortest path length) from firm i to firm j, m is the number of nodes that firm i can reach, i ≠ j, and n is the number of nodes in the network in the range [0, n].

The mediating variable is the digital transformation level of high-tech manufacturing firms. Previous research suggests that describing digital transformation in annual reports can be achieved by analyzing digital-related word frequencies [62,63]. Some studies use “keywords + expert ratings” to judge the digital transformation level [64]. This provides useful insights for this study. Therefore, according to the definition criteria of digital transformation in the CSMAR database, the classification of digital technology includes five categories: artificial intelligence technology, blockchain technology, cloud computing technology, big data technology, and digital application technology. We obtain the frequency measure of five digital technologies adopted by listed firms from the CSMAR database.

The moderating variable is market development. Earlier studies used European bank for reconstruction and development (EBRD) transition indicators and have found that market-oriented reforms have significant explanatory power for economic growth [65]. The NERI index compiled in 2016 is used to measure the development of the market in the provinces [66,67]. Management scholars have utilized the National Economic Research Index (NERI) to assess market environment quality in China, and their studies have been published in top management journals [68,69]. Referring to Wei [70], we utilize China’s provincial market index (hereinafter referred to as the NERI) compiled by the National Economic Research Institute of the China Reform Foundation, by quantitatively scoring five secondary sub-indices in the areas. It systematically measures the progress of marketization in 31 provinces in China. The China Marketisation Index by Province (the ‘Marketisation Index’) is an index system that measures the relative progress of marketization in provinces, autonomous regions, and municipalities directly under the central government by means of data standardization.

In order to better measure and analyze the main variables in this study, it is necessary to control for other variables that may affect firm ambidextrous innovation. Therefore, this study adds the control variables of network level, firm attributes level, and regional level.

Control variables at the level of the network structure: It has been shown that some structural attributes of the network affect firm innovation [60]. Therefore, in order to identify the factors that influence firm innovation, we control two structural attributes, including network density and degree centrality. They are calculated by using Ucinet 16.0 software.

Network density: Network density is the ratio of the number of edges actually in the network to an upper limit on the number of edges that can be accommodated [71]. The formula is as follows:

$$Density=l/[{\displaystyle \frac{n\left(n-1\right)}{2}}]$$

(3)

l is the actual number of connections in the network; n denotes the number of nodes in the network, taking the range of [0, 1]. A larger value indicates a higher network density.

Degree centrality is one of the most important structural indicators to describe the location of the network [72]. The formula is as follows:

$${DC}_{deg }\left(v\right)={\displaystyle \frac{d_v}{\left|N\right|-1}}$$

(4)

where DCdeg (v) is the degree centrality of node v, and N is the number of participants in the individual network of v. dv is the number of partners that the organization has directly connected in its network.

Control variables at the firm attributes level: Drawing on existing research, this study controls for the following variables. (1) Firm age—The longer a firm survives, the more knowledge is accumulated, which may affect firm innovation [73]. Measure the age of the firm by subtracting the year of the study from the year of the firm’s founding and adding 1 to take the natural logarithm. (2) Number of R&D personnel—The larger the number of employees, the larger the firm is and the more likely it is to have a “scale effect” that may affect firm innovation [74]. (3) Registered capital—Registered capital represents the size of resources at the firm’s disposal, and controlling for this variable can exclude the interference of firms’ innovative activities due to differences in initial resources [75]. (4) Net profit—Net profit is an important manifestation of the enterprise’s operating results. Controlling this variable can control the influence of enterprise profitability on ambidextrous innovation [76]. (5) Return on assets—Return on assets reflects the operational efficiency of the firm, and controlling this variable helps to isolate the impact of the firm’s operational efficiency on ambidextrous innovation [76]. (6) Asset liability ratio—The asset liability ratio measures the long-term solvency of the enterprise, and controlling this variable can eliminate the interference of financial risk factors on binary innovation [77]. (7) Year dummy variable—Controlling for the possible effects of different years.

Control variables at the regional level: The firm digital transformation is affected by the development level of the local digital economy, so we consider the digital development level of the region. To promote national information consumption, the Ministry of Industry and Information Technology (MIIT) identified 25 national information consumption demonstration cities across the country in December 2015. Furthermore, it released the document “The 2016 National Information Consumption Demonstration Cities Construction Guide”. The guide suggests that these cities should accelerate the upgrading of information infrastructure and accelerate technological innovation. Therefore, we use as a dummy variable whether the province where the city is located belongs to the National Information Consumption Demonstration Cities, with 1 belonging to the National Information Consumption Demonstration Cities and 0 not belonging to the National Information Consumption Demonstration Cities. The measure of all variables is shown in

Table 1. Variables and measures.

Variable Category	Variable Label	Variable Measure
Dependent variable	Exploratory innovation (EI-1)	Expensed expenses for corporate R&D activities
	Exploitative innovation (EI-2)	Capitalized expenditures for corporate R&D activities
Independent variable	Local efficiency (LE)	The global efficiency of the network consisting of its neighboring nodes after removing a given node
	Reach rate (RR)	The inverse of the length of the shortest path the firm can reach
Mediating variable	Digital transformation (DT)	The sum of the word frequency of five categories: artificial intelligence technology, blockchain technology, cloud computing technology, big data technology, and digital application technology
Moderating variable	Market development (MT)	Market indices constructed by the NERI
Control variable	Firm age (FA)	Year of study minus year of business establishment plus 1 to take the natural logarithm
	Number of R&D personnel (RDP)	Number of R&D personnel
	Registered capital (RC)	Registered capital
	Net profit (NP)	Net profit
	Return on assets (ROA)	Return on assets
	Asset liability ratio (AIR)	Asset liability ratio
	Network density (ND)	Ratio of the number of edges actually present in the network to the upper limit of the number of edges
	Degree centrality (DC)	Number of direct links owned by the node

.

3.3. Research Model

To test the U-shaped relationship between the local efficiency and reach rate and firms’ exploratory innovation and exploitative innovation, these models are constructed as follows:

$$E{I}_{\_} 1=a_o+a_{1}LE+a_2{LE}^2+a_3\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}+{\mathsf{\epsilon }}_1$$

(5)

$$E{I}_{\_} 1={\beta }_o+{\beta }_{1}RR+{\beta }_2{RR}^2+{\beta }_3\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}+{\mathsf{\epsilon }}_2$$

(6)

$$E{I}_{\_} 2={\sigma }_o+{\beta }_{1}LE+{\beta }_2{LE}^2+{\beta }_3\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}+{\mathsf{\epsilon }}_3$$

(7)

$$E{I}_{\_} 2={\rho }_1+{\rho }_{1}RR+{\rho }_1{RR}^2+{\rho }_1\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}+{\mathsf{\epsilon }}_4$$

(8)

In order to test the mediating role of firm digital transformation between the local efficiency and exploratory innovation, the reach rate and exploratory innovation, the local efficiency and exploitative innovation, and the reach rate and exploitative innovation, the model is constructed as follows:

$$E{I}_{\_} 1={\pi }_o+{\pi }_{1}LE+{\pi }_2{LE}^2+{{\pi }_{3}DT+\pi }_4\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}+{\mathsf{\epsilon }}_5$$

(9)

$$E{I}_{\_} 1={\partial }_o+{\partial }_{1}RR+{\partial }_2{RR}^2+{\partial }_{3}DT+{\partial }_4\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}+{\mathsf{\epsilon }}_6$$

(10)

$$E{I}_{\_} 2={\mu }_o+{\mu }_{1}LE+{\mu }_2{LE}^2+{\mu }_{3}DT+{\mu }_4\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}+{\mathsf{\epsilon }}_7$$

(11)

$$E{I}_{\_} 2={\gamma }_o+{\gamma }_{1}LE+{\gamma }_2{LE}^2+{\gamma }_{3}DT+{\gamma }_4\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}+{\mathsf{\epsilon }}_8$$

(12)

In order to test the moderating role of the market development between the local efficiency and exploratory innovation, the reach rate and exploratory innovation, the local efficiency and exploitative innovation, and the reach rate and exploitative innovation, the model is constructed as follows:

$$E{I}_{\_} 1={\eta }_0+{\eta }_{1}LE+{\eta }_2{LE}^2+{\eta }_{3}MI+{\eta }_{4}LE\times \mathrm{M}\mathrm{I}+{\eta }_5{LE}^2\times \mathrm{M}\mathrm{I}+{\eta }_{6}Control+{\mathsf{\epsilon }}_9$$

(13)

$$E{I}_{\_} 1={\delta }_0+{\delta }_{1}RR+{\delta }_2{RR}^2+{\delta }_{3}MI+{\delta }_{4}RR\times \mathrm{M}\mathrm{I}+{\delta }_5{RR}^2\times \mathrm{M}\mathrm{I}+{\delta }_{6}Control+{\mathsf{\epsilon }}_{10 }$$

(14)

$$E{I}_{\_} 2={ϵ}_0+{ϵ}_{1}LE+{ϵ}_3{LE}^2+{ϵ}_{4}MI+{ϵ}_{5}LE\times \mathrm{M}\mathrm{I}+{ϵ}_6{LE}^2\times \mathrm{M}\mathrm{I}+{ϵ}_{7}Control+{\mathsf{\epsilon }}_{11}$$

(15)

$$E{I}_{\_} 2={\theta }_0+{\theta }_{1}RR+{\theta }_2{RR}^2+{\theta }_{3}MI+{\theta }_{4}RR\times \mathrm{M}\mathrm{I}+{\theta }_5{RR}^2\times \mathrm{M}\mathrm{I}+{\theta }_{6}Control+{\mathsf{\epsilon }}_{12}$$

(16)

In addition, a, β, σ, ρ, π, ∂, μ, γ, η, δ, ϵ, and θ in the model are the regression coefficients of the model intercept term, the independent variables, and the control variables, respectively.

4. Analyses and Results

4.1. Basic Data Analysis

This study used Stata 15.0 to perform variable descriptive statistics and correlation tests on the data, and the results are shown in

Table 2. Descriptive statistics and correlation analyses.

Number	Variable	Mean	SD	1	2	3	4
1	LE	4.446 × 107	7.997 × 109	1.000
2	RR	0.506	0.0229	0.0001 (0.9768)	1.000
3	MD	10.02	1.544	−0.009 (0.5049)	−0.040 *** (0.0050)	1.000
4	DT	20	44.60	−0.005 (0.7182)	−0.020 (0.1493)	0.089 *** (0.0001)	1.000

Note: z-statistics in parentheses, *** p < 0.001.

. Since there are non-equilibrium panel data, a fixed-effect threshold regression model was used for hypothesis testing [78]. Furthermore, the absolute values of the correlation coefficients between all variables in the table are less than 0.7, so the possibility of multicollinearity between variables is excluded.

4.2. Hypotheses Testing

The main effects test assesses the relationship between network connectivity (local efficiency, reach rate) and ambidextrous innovation (exploratory innovation, exploitative innovation). Model 1 and Model 5 in

Table 3. The main effect test.

Variable	EI-1				EI-2
	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6	Model 7
RDP	0.029 * (1.41)	0.028 (1.43)	0.085 *** (5.31)	0.041 * (2.37)	0.021 * (0.61)	−0.012 * (−0.33)	−0.012 * (−0.34)
FA	0.0008 *** (5.24)	0.0008 *** (5.25)	0.0004 *** (−4.58)	0.0002 (0.19)	0.0003 (1.18)	0.0004 (1.28)	0.0004 (1.38)
RC	−0.032 * (−1.17)	−0.032 (−1.18)	0.037 (0.11)	0.035 (0.56)	−0.015 (−0.72)	−0.013 (−0.25)	−0.013 (−0.25)
NP	−0.067 (−0.12)	−0.072 (−0.96)	−0.028 *** (−2.84)	−0.099 (−1.38)	0.081 (0.01)	0.087 (0.06)	0.072 (0.05)
ROA	−0.090 *** (−5.15)	−0.089 *** (−5.21)	−0.018 *** (−6.57)	−0.098 *** (−5.77)	0.023 (0.77)	0.026 (0.82)	0.026 (0.82)
AIR	−0.016 *** (−1.05)	−0.016 *** (−1.05)	−0.013 *** (−6.72)	−0.028 * (−1.85)	0.022 * (0.79)	0.024 * (0.85)	0.024 * (0.84)
ND	−0.041 (−0.13)	−0.041 (−0.16)	−0.002 (−0.94)	−0.055 (−0.43)	−0.041 (−0.17)	−0.017 (−0.07)	−0.019 (−0.08)
DC	−0.014 (−1.3)	−0.014 (−1.3)	−0.0001 (−0.11)	−0.014 (−1.24)	0.023 (0.12)	0.041 (0.2)	0.043 (0.21)
LE		0.033 * (2.10)				−0.019 * (−2.30)
LE2		−0.045 * (2.13)
RR			0.014 * (1.12)	0.338 *** (3.25)			−0.018 * (−2.76)
RR2				−0.648 *** (−3.21)
Year	Control	Control	Control	Control	Control	Control	Control
Province	Control	Control	Control	Control	Control	Control	Control
Constant	0.0168 *** (5.63)	0.0168 *** (5.63)	0.00921 * (1.24)	0.0267 *** (3.78)	0.0283 *** (4.59)	0.0269 *** (4.97)	0.0174 (1.3)
F	6.24	6.87	7.00	6.47	6.00	6.00	6.00
R2	0.3162	0.3162	0.3165	0.3134	0.3134	0.3134	0.3134

Note: z-statistics in parentheses, *** p < 0.001, * p < 0.05.

only include the control variables. The findings indicate that it is crucial to choose these control variables. Main effect test: Model 2 adds local efficiency and the squared term of local efficiency based on Model 1, and the coefficient of the square of local efficiency is −0.045 (p < 0.1). Therefore, the local efficiency has an inverted U-shaped relationship with exploratory innovation. Model 3 and Model 4 add the reach rate and the squared term of the reach rate based on Model 1, separately, and the coefficient of the squared term of the reach rate is −0.648 (p < 0.01). Therefore, H1a is verified. Model 6 adds local efficiency to Model 5, and the coefficient of local efficiency is −0.019 (p < 0.1), indicating that local efficiency negatively affects exploitative innovation. Model 7 adds the reach rate to Model 5, and the coefficient of the reach rate is −0.018 (p < 0.1), indicating that the reach rate negatively affects exploitative innovation. There is a different result with H1b.

Table 4. Mediating effect test.

Variable	DT			EI-1	EI-2
	Model 8	Model 9	Model 10	Model 11	Model 12
RDP	0.000707 *** (2.95)	0.000707 *** (2.95)	0.0007 *** (2.92)	0.0273 *** (1.37)	−0.0996 (−0.28)
FA	2.819 *** (15.13)	2.817 *** (15.12)	2.962 *** (12.24)	0.00771 *** (4.91)	0.00437 (1.53)
RC	0.00189 (0.57)	0.00188 (0.57)	0.00196 (0.59)	−0.00323 (−1.19)	−0.00120 (−0.24)
NP	0.00176 * (1.92)	0.00176 * (1.92)	0.00173 * (1.89)	−0.0740 (−0.98)	0.0132 (0.10)
ROA	−0.0449 (−0.02)	−0.0290 (−0.01)	−0.0709 (−0.03)	−0.0895 *** (−5.21)	0.00255 (0.82)
AIR	−1.794 (−0.95)	−1.786 (−0.94)	−1.867 (−0.99)	−0.00162 (−1.04)	0.00233 (0.83)
ND	0.799 (0.51)	0.798 (0.51)	0.685 (0.43)	−0.00414 (−0.32)	−0.00147 (−0.06)
DC	−0.0487 (−0.36)	−0.0488 (−0.36)	−0.05 (−0.37)	−0.00143 (−1.30)	0.00396 (0.20)
LE		−0.00214 * (−1.50)
LE2		0.001 * (−1.45)
RR			−0.677 * (−1.46)
RR2			0.878 * (1.65)
DT				0.00109 * (1.86)	−0.028 * (−1.21)
Year	Control	Control	Control	Control	Control
Province	Control	Control	Control	Control	Control
Constant	−0.340 *** (−9.37)	−0.339 *** (−9.36)	−0.506 *** (−5.37)	0.0172 *** (5.70)	0.0259 *** (4.74)
F	30.62	27.57	25.39	6.94	6.00
R2	0.0620	0.0621	0.0629	0.0164	0.0164

Note: z-statistics in parentheses, *** p < 0.001, * p < 0.05.

reports the mediating effect of the firm digital transformation level. We use hierarchical regression analysis to examine the mediating effect of the firm digital transformation level on the relationship between connectivity (local efficiency, reach rate) and ambidextrous innovation. Model 8 verifies the relationship between the control variables and digital transformation. Model 9 adds the independent variable local efficiency and the squared term of local efficiency based on Model 8 to verify the relationship between local efficiency and digital transformation, and the coefficient of local efficiency is −0.021 (p < 0.1). The coefficient of digital transformation is 0.011 (p < 0.1) in Model 11; thus, H2a is validated by the test results of Model 8 and Model 11. It indicates that the degree of digital transformation has a mediating role in the relationship between local efficiency and exploratory innovation. The coefficient of the squared term of the reach rate is 87.788 (p < 0.1) in Model 10; thus, H2b is validated by Model 11 and Model 10. The coefficient of digital transformation is −0.028 (p < 0.1) in Model 12, and H2c and H2d are validated through Model 9 and Model 12, and Model 10 and Model 12, respectively.

According to Preacher and Hayes [79], Bootstrap is used to further verify the mediating effect of digital transformation. It can be seen from

Table 5. The mediating effect of digital transformation in the relationship between local efficiency and ambidextrous innovation.

Independent Variable	Indicator	Value	Boot	Boot CI	Boot CI	z
			Standard Deviation	Lower	Upper
Exploratory innovation	indirect effect	−0.000194	0.0004	−0.000979	−0.00059	−0.49 *
	Direct effect	0.000636	0.000726	−0.000786	−0.000786	0.88 ***
Exploitative innovation	Indirect effect	0.00228	0.00011	−0.00193	−0.00139	0.21 *
	Direct effect	−0.00256	0.00166	−0.00352	−0.003	−0.15 ***

Note: |z|> 0.9115, p < 5%, and * in the table indicates significance at the 5% level; *** in the table indicates significance at the 0.1% level.

that the upper and lower bounds of the 95% confidence interval for the mediating effect of digital transformation do not include 0, and the upper and lower bounds for the direct effect of local efficiency on exploratory innovation do not include 0, indicating that digital transformation has a mediating role in the relationship between local efficiency and exploratory innovation. H2a is further validated. Similarly, H2c is further validated. Moreover, H2b and H2d are further validated.

Table 6. The moderating effect test.

Variable	EI-1			EI-2
	Model 13	Model 14	Model 15	Model 16	Model 17	Model 18
RDP	0.081 ***	0.082 ***	0.061 ***	0.08 ***	0.08 ***	0.06 ***
	(5.09)	(5.15)	(3.74)	(5.09)	(5.15)	(3.74)
FA	−0.004 ***	−0.004 ***	−0.004 ***	−0.004 ***	−0.003 ***	−0.003 ***
	(−5.14)	(−4.97	(−5.10)	(−5.14)	(−4.97)	(−5.10)
RC	−0.012	−0.013	−0.018	0.012	0.013	0.018
	(0.36)	(0.39)	(0.53)	(0.36)	(0.39)	(0.53)
NP	−0.030 ***	−0.029 ***	−0.028 **	−0.003 ***	−0.003 ***	−0.028 ***
	(−3.03)	(−2.99)	(−2.83)	(−3.03)	(−2.99)	(−2.83)
ROA	−0.018 ***	−0.018 ***	−0.018 ***	−0.018 ***	−0.018 ***	−0.018 ***
	(−6.49)	(−6.48)	(−6.34)	(−6.49)	(−6.48)	(−6.34)
AIR	−0.013 ***	−0.014 ***	−0.016 ***	−0.013 ***	−0.013 ***	−0.013 ***
	(−6.63)	(−6.60)	(−6.45)	(−6.63)	(−6.60)	(−6.45)
ND	−0.020	−0.0019	−0.019	−0.002	−0.002	−0.002
	(−0.89)	(−0.88)	(−0.85)	(−0.89)	(−0.88)	(−0.85)
DC	0.026	0.0025 *	−0.023	0.0003	−0.0002	−0.0002
	(0.14)	(−1.74)	(−0.13)	(0.14)	(−0.10)	(−0.13)
LE	0.082			0.082
	(0.99)			(0.99)
LE2	0.081
	(0.99)
RR		−0.253 *			−0.253 *
		(−1.74)			(−1.74)
RR2		0.463 *
		(1.67)
MD	0.021 ***	0.021 ***	0.018 ***	0.002 ***	0.002 ***	0.002 ***
	(6.36)	(6.52)	(5.37)	(6.36)	(6.52)	(5.37)
LE × MD	0.035 *			−0.028
	(1.36)			(−0.30)
LE2 × MD	−0.028 *
	(−1.30)
RR × MD		0.021 *			0.049 *
		(1.87)			(1.49)
RR2 × MD		−0.047 *
		(1.49)
Year	Control	Control	Control	Control	Control	Control
Province	Control	Control	Control	Control	Control	Control
Constant	0.0231 ***	0.0315 **	0.0244 ***	0.023 ***	0.032 ***	0.024 ***
	(6.67)	(2.69)	(6.85)	(6.67)	(2.69)	(6.85)
F	14.07	13.25	17.45	14.07	13.25	17.45
R2	0.0327	0.0333	0.0402	0.033	0.033	0.040

Note: z-statistics in parentheses, *** p < 0.001, ** p < 0.01, * p < 0.05.

reports the moderating effect of the market development. To test the moderating role of market development in the inverted U-shaped relationship between local efficiency, the reach rate, and exploratory innovation, separately, we establish the regression model y = β0 + β1 x + β2 x2 + β3 xz+β4 x2 z + β5 z, where z is the moderating variable. According to Haans, Pieters, and He [80], if β4 is significantly positive, the U-shaped curve will flatten; if β4 is significantly negative, the U-shaped curve will become steeper. According to Aiken and West [81], the moderating effect of the U-shaped curve can be tested if the coefficients of the independent variable × moderating variable and the quadratic term of the independent variable × moderating variable are both significant, indicating that the moderating variable alters the curve’s shape and slope. The slopes of the regression lines for high and low moderating effects k were obtained by grouping the mean values of the moderating variables plus or minus one standard deviation, as follows:

$$k={\beta }_1+\left(2{\beta }_2+{2\beta }_{4}z\right)x+{\beta }_{3}z$$

(17)

Model 13 shows that the coefficient of the interaction term between the squared term of local efficiency and the market development is −0.028 (p < 0.1), indicating that the market development negatively moderates the inverted U-shaped relationship between local efficiency and exploratory innovation and that the curve is steeper. Model 14 shows that the coefficient of the interaction term between market development and the quadratic term of the reach rate is 0.047 (p < 0.1), indicating that market development positively moderates the inverted U-shaped relationship between the reach rate and that the curve is flatter. According to Model 13, k = (0.162 − 0.056z) x + 0.035z + 0.082. In the low local efficiency interval, the simple slope of the high market development is higher than that of the low market development (0.117 > 0.047), indicating that the high level of market development enhances the positive effect of local efficiency on exploratory innovation within the low local efficiency interval. In the high local efficiency interval, the simple slope of the high market development level is lower than the slope of the low market development (0.223 < 0.265). It indicates that higher market development weakens the negative effect of local efficiency on exploratory innovation in the high local efficiency interval. The coefficient on the primary term of market development and the reach rate is not significant, indicating that it does not change the overall slope of the curve.

Model 16 shows that the coefficient of the interaction term between local efficiency and market development is −0.028, which is not significant, indicating that market development does not moderate the negative impact of local efficiency on exploitative innovation. Model 17 shows that the coefficient of the interaction term between the reach rate and market development is 0.049 (p < 0.1). Therefore, H3a is partially validated, and H3b is validated (Figure 4).

4.3. Robustness Test

4.3.1. Instrumental Variable Method

Economic efficiency metrics may have reverse cause-and-effect. Firms with better economic performance are in turn conducive to the digital transformation of the firm. Referring to Bartik’s instrumental variables approach [82], the instrumental variable for digital transformation is the product of the mean value of other firms’ digital transformation levels in the previous year and the growth rate of internet access in the country (excluding the province in which the firm is located). We enhance the exclusion by using data from the year prior to the sample rather than the sample period and multiply the instrumental variable by the national growth rate of the number of people accessing the Internet, which excludes data from the province.

Table 7. Instrumental variable regression.

Variable	EI-1	EI-2
	Model 1	Model 2
RDP	0.009 ***	−0.001 **
	(2.72)	(−2.17)
FA	−0.001 ***	−0.00001
	(−3.65)	(−0.04)
RC	0.057	−0.069
	(0.78)	(−0.59)
NP	−0.052 **	−0.032
	(−2.21)	(−0.84)
ROA	−0.019 ***	−0.001
	(−3.36)	(−0.10)
AIR	−0.015 ***	−0.0003
	(−3.44)	(−0.05)
ND	−0.005	−0.009
	(−1.04)	(−1.26)
DC	0.001	−0.001
	(1.27)	(−1.14)
DT_Tool Variables	0.007 ***	−0.001 *
	(2.97)	(−1.29)
Constant	0.034 ***	0.040 ***
	(7.78)	(5.66)
F	9.51	8.94
R2	0.556	0.538

Note: z-statistics in parentheses, *** p < 0.001, ** p < 0.01, * p < 0.05.

reports the two-stage least squares (2SLS) regression results. The correlation coefficient between the tool variable of digital transformation level and exploratory innovation is 0.007 (p < 0.1), and the correlation coefficient between the tool variable of digital transformation level and exploitative innovation is −0.001 (p < 0.1). It indicates that after mitigating possible endogeneity problems, digital transformation still significantly improves firm exploratory and exploitative innovation.

4.3.2. Substitution of Variables

According to Guo and Zhu [83], the asset proportions of software investment and digital hardware investment projects are used to measure the level of enterprise digital transformation. In particular, the software investment ratio is the net value of software assets in intangible assets divided by total net assets, and the digital hardware investment ratio is the net value of office equipment, electronic equipment, etc. in fixed assets divided by total net assets. The clustering coefficients and number of node connections are used instead of local efficiency and reach, and the NEDI is used instead of the market index for validation. The robust-type test results in coefficients that remain significant.

4.4. Heterogeneity Test for Digital Technology

Different digital technologies have different impacts. The previous literature tends to ignore these differences. Referring to the study of Füller et al. [84], the digital technologies used by enterprises are classified into five types: artificial intelligence technology, blockchain technology, cloud computing technology, big data technology, and digital application technology based on the classification of digital transformation keywords. Their heterogeneity is also analyzed.

The results in

Table 8. Heterogeneity test 1.

Variable	EI-1					EI-2
	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6	Model 7	Model 8	Model 9	Model 10
RDP	0.078 ***	0.084 ***	0.081 ***	0.074 ***	0.067 ***	−0.012 **	−0.012 **	−0.019 **	−0.012 **	−0.013 **
	(4.82)	(5.24)	(5.01)	(4.62)	(4.05)	(−2.28)	(−2.35)	(−2.26)	(−2.36)	(−2.35)
FA	−0.0004 ***	−0.0004 ***	−0.0003 ***	−0.0004 ***	−0.0003 ***	0.0008	0.042	−0.079	0.03	0.091
	(−4.64)	(−4.64)	(−4.45)	(−4.85)	(−4.42)	(0.03)	(0.02)	(−0.00)	(0.01)	(0.04)
RC	0.032	0.04	0.053	0.068	0.095	−0.066	−0.065	−0.066	−0.065	−0.064
	(0.87)	(0.12)	(0.16)	(0.20)	(0.28)	(−0.60)	(−0.59)	(−0.60)	(−0.59)	(−0.58)
NP	0.027	−0.028 ***	−0.028 ***	−0.027 ***	−0.029 ***	−0.027	−0.027	−0.027	−0.026	−0.027
	(0.38)	(−2.82)	(−2.77)	(−2.67)	(−2.89)	(−0.82)	(−0.81)	(−0.83)	(−0.80)	(−0.82)
ROA	−0.018	−0.018 ***	−0.018 ***	−0.018 ***	−0.018 ***	−0.002	−0.002	−0.002	−0.002	−0.002
	(−0.53)	(−6.53)	(−6.53)	(−6.52)	(−6.54)	(−0.21)	(−0.19)	(−0.21)	(−0.19)	(−0.19)
AIR	−0.013	−0.013 ***	−0.013 ***	−0.013 ***	−0.013 ***	−0.001	−0.001	−0.001	−0.001	−0.001
	(−1.06)	(−6.62)	(−6.69)	(−6.52)	(−6.73)	(−0.15)	(−0.12)	(−0.14)	(−0.12)	(−0.13)
ND	−0.018 ***	−0.002	−0.002	−0.002	−0.002	−0.010	−0.010	−0.010	−0.010	−0.010
	(−6.58)	(−0.91)	(−0.89)	(−0.99)	(−0.89)	(−1.36)	(−1.36)	(−1.37)	(−1.36)	(−1.36)
DC	−0.00013 ***	−0.00003	−0.000033	−0.00006	−0.00003	−0.001	−0.001	−0.001	−0.001	−0.001
	(−0.54)	(−0.14)	(−0.18)	(−0.34)	(−0.15)	(−1.10)	(−1.12)	(−1.10)	(−1.12)	(−1.12)
Artificial intelligence	0.0017 *** (5.28)					−0.005 (−0.47)
Blockchain technology		0.002 *** (2.98)					0.001 (0.31)
Cloud computing			0.0008 ***					−0.00007
			(3.01)					(−0.77)
Big data				0.0002 ***					0.002 *
				(5.73)					(1.25)
Digital technology applications					0.0001 *** (4.45)					0.003 * (1.35)
Constant	0.042 ***	0.043 ***	0.042 ***	0.042 ***	0.042 ***	0.038 ***	0.038 ***	0.038 ***	0.038 ***	0.038 ***
	(25.23)	(25.58)	(25.09)	(25.38)	(25.07)	(6.95)	(6.93)	(6.98)	(6.92)	(6.86)
F	13.26	13.54	13.55	15.99	14.65	1.567	1.554	1.605	1.551	1.556
R2	0.031	0.026	0.026	0.031	0.028	0.003	0.003	0.003	0.003	0.003

Note: z-statistics in parentheses, *** p < 0.001, ** p < 0.01, * p < 0.05.

show that the correlation coefficients of AI technology, blockchain technology, cloud computing technology, big data technology, and digital application technology with exploratory innovation are all significant; the correlation coefficients are 0.0017 (p < 0.01), 0.002 (p < 0.01), 0.0008 (p < 0.01), 0.0002 (p < 0.01), and 0.0001 (p < 0.01), respectively. For exploitative innovation, only big data technology and digital application technology showed significant correlation coefficients with results of 0.0002 (p < 0.1) and 0.003 (p < 0.1), respectively. In terms of “process-oriented” and “goal-outcome-oriented” categorization [85,86], artificial intelligence technologies, blockchain technologies, cloud computing technologies, and digital application technologies belong to the digital transformation of “process-oriented” and “goal-outcome-oriented” categories, which mainly realize the digital transformation of firms through intelligent firms’ production processes. Big data technology and digital application technology belong to the digital transformation of “target result-oriented”, which mainly realizes the digital transformation of firms through intelligent embedded products or intelligent access to end-user data.

5. Discussion

First, local efficiency and the reach rate have an inverted U-shaped relationship with exploratory innovation. Local efficiency and the reach rate negatively affect exploitative innovation. The results are the same as H1a and different from H2a. We find that moderate local efficiency is more conducive to increasing firm exploratory and exploitative innovation than lower and higher local efficiency. We believe that the main reason is that as inter-organizational local efficiency increases, the stability of inter-organizational linkages increases, which facilitates the acquisition and integration of heterogeneous resources and thus enhances exploratory innovation. Local efficiency and the reach rate have a negative impact on firm exploitative innovation. The reasons are as follows: first, in the policy encouragement background of exploratory innovation in China, the region gives policy encouragement and subsidies to high-tech firms’ cooperation based on technology R&D, which leads high-tech firms to pay attention to exploratory innovation based on technology cooperation while ignoring exploitative innovation. In addition, China, as an emerging economy, has been carrying out exploitative innovation for a long period, and the more mature development process requires less capital investment to achieve.

Second, firm digital transformation mediates the relationship between local efficiency and exploratory innovation, local efficiency and exploitative innovation, the reach rate and exploratory innovation, and the reach rate and exploitative innovation. High connectivity in digitalized innovation ecosystems provides a resource space of digital technologies for their digital transformation [18,47]. Moreover, by collaborating with digital firms and digital platforms, firms can reserve a digital technological base for digital transformation. By employing the higher information dissemination and sharing of digital technologies and the automation of production and management they bring, firms can use digital technologies to increase the path to acquire knowledge in interdependent inter-organizational relationships and thus obtain more external knowledge to realize ambidextrous innovation.

Third, the market development level plays a negative moderating role between local efficiency and exploratory innovation and between local efficiency and exploitative innovation. Market development plays a positive moderating role between the reach rate and exploratory innovation, and the reach rate and exploitative innovation. When the degree of market development is high, fierce market competition destabilizes the local cooperation among firms, which leads to a reduced local efficiency of the digital innovation ecosystem. Therefore, it is not conducive to firm exploratory innovation and exploitative innovation. In addition, market activity destabilizes local cooperation while increasing new cooperation among more firms, thus leading to a rapid increase in the number of connections among firms, which enhances the reach rate of the innovation ecosystem.

Fourth, the local efficiency of the digital innovation ecosystem is conducive to AI technologies embedded in firms and is not conducive to big data technologies and cloud computing embedded in firms. The reach rate is conducive to the realization of five digital technologies, including AI technologies, blockchain technologies, big data technologies, cloud computing technologies, and digital application technologies. Second, all five digital technologies are conducive to promoting firm exploratory innovation, and only big data technology and digital technology promote firm exploitative innovation. Furthermore, according to the classification of “process-oriented” and “goal-result-oriented”, artificial intelligence technology, blockchain technology, and cloud computing technology belong to the “process-oriented” category embedded in the digital transformation, which mainly realizes the digital transformation of firms through intelligent firm production processes. Big data technology and digital application technology belong to the “goal-result-oriented” of digital transformation, which mainly realizes the digital transformation of firms by embedding intelligent products or intelligently acquiring end-user data. Exploratory innovation requires a long-term R&D process and the setting of R&D goals, so both types of digital technologies have a facilitating effect on exploratory innovation. In contrast, for result-oriented exploitative innovation, only result-oriented digital technologies play a positive role.

6. Conclusions

This study explores how inter-organizational connectivity affects the ambidextrous innovation of enterprises and answers the questions of whether enterprises “need to integrate into the digital ecosystem” and “whether digital transformation is beneficial to the development of enterprises”. From the perspective of coupling innovation ecosystems and digitalization, this study establishes a structural model of digital innovation ecosystems by using the data of China’s high-tech enterprise alliance cooperation from 2015 to 2022.

It is found in the empirical study that the local efficiency and reach rate of digital innovation ecosystems have an inverted U-shaped relationship with exploratory innovation, and the local efficiency and reach rate of digital innovation ecosystems have a negative effect on firm exploitative innovation. Second, the firm digital transformation level plays a mediating role in the relationship between the local efficiency and exploratory innovation, the reach rate and exploratory innovation, the local efficiency and exploitative innovation, and the reach rate and exploitative innovation. Finally, the market development level negatively moderates the inverted U-shaped relationship between the local efficiency and exploratory innovation and positively moderates the relationship between the reach rate and exploratory innovation, and the reach rate and exploitative innovation. In addition, we empirically analyze the different impacts and paths of a total of five types of digital transformation technologies, namely, artificial intelligence, blockchain, cloud computing, big data technology, and digital application technology, on firms’ ambidextrous innovation.

The findings suggest that inter-organizational connectivity has an inverted “U”-shaped relationship with firms’ exploratory innovation in digital innovation ecosystems, which has a negative impact on firms’ exploitative innovation. Structural connectivity in digital innovation ecosystems drives firms’ ambidextrous innovation capabilities by facilitating their digital transformation, with different digital transformation technologies playing different roles. Second, a dynamic market environment does not always play a fully positive moderating role.

Nonetheless, this study is not without limitations. First, this work does not cover informal institutional environments, which include factors such as social, internal organizational, ecological, economic, and technological environments [87]. Therefore, future research should further integrate the impact of different informal institutional environments. Second, the concept of digital application technology is taken into account in this research due to the availability of data for measuring digital transformation variables. In the future, the concept of digital application technology, as well as other measurements for digital transformation, should be further explored. Third, there are limitations in the measurement of digital transformation. In future research, more methods will be used to measure this variable and compare and analyze it.

7. Theoretical Contributions and Originality

7.1. Theoretical Contributions

First, this study responds to the call by Walrave et al. for more empirical research on innovation ecosystems [88]. We clarify how digital ecosystems function as “resource pools” through structural attributes such as local efficiency and the reach rate, thereby expanding the application scenarios of innovation ecosystem theory in digital contexts.

Second, from the perspective of the resource-based view, this study reveals how the connectivity of digital innovation ecosystems affects ambidextrous innovation by influencing firms’ digital transformation. It fills the research gap in the resource-based view regarding the “conversion of network structural resources into innovation capabilities”, extends the application of the resource-based view from traditional resources to network structural resources in digital ecosystems, and clarifies the path of resource conversion.

Third, through a mediating effect model, this study verifies that digital transformation is an important bridge connecting digital innovation ecosystems and ambidextrous innovation and clarifies the mechanism among the three.

Fourth, this study incorporates the level of market development, which characterizes the market environment, into the research model; explores its moderating role in the relationships between organizational connectivity, digital transformation, and ambidextrous innovation; reveals the boundary conditions for market environments to drive corporate innovation; and improves the theoretical understanding of how institutional environments affect corporate innovation within innovation ecosystems.

7.2. Originality

The originality of this study lies in filling the existing literature gaps and enriching existing empirical work. Previous studies hypothesized that innovation ecosystems improve firm performance through evolution and internal knowledge transfer (e.g., [89,90]) but paid little attention to the structural characteristics of digital innovation ecosystems and their intrinsic mechanisms with firms’ ambidextrous innovation. This study is the first to reveal the non-linear relationship between inter-organizational connectivity and exploratory innovation, filling this gap. Meanwhile, this study enriches the literature on “digital innovation ecosystem–firm digital transformation–firm ambidextrous innovation”, breaking through previous discussions on the relationship between innovation ecosystems and firm innovation, as well as between digital transformation and knowledge transfer efficiency [48,89].

In addition, the previous literature analyzed the impact of digital transformation on small and medium-sized enterprises but lacked research on specific industries [91]. This paper expands the research scope of digital transformation from the perspectives of industry characteristics and technology classification. We selected the high-tech manufacturing industry for analysis and, through empirical research, distinguished the heterogeneous impacts of technologies such as artificial intelligence and blockchain on ambidextrous innovation, clarifying the unique paths of different technologies in driving innovation.

Furthermore, we have also conducted in-depth research on the role of the market environment in innovation ecosystems. In the innovation ecosystem formed by the free association of enterprises, the market environment, as the most direct living environment for social entities, plays a crucial role. Although there is a broad consensus on the value of innovation ecosystems (e.g., [51,52,92,93]), how the market environment affects corporate innovation has not been fully elaborated in existing studies.

8. Management Implications

Our study provides some important managerial insights for high-tech manufacturing firms.

First, higher local efficiency and reach are not more conducive to exploratory innovation. For example, Huawei maintains the threshold of inter-organizational connectivity through a “partner rotation system”, which validates the model’s conclusion that inter-organizational connectivity has an inverted “U” relationship with firms’ exploratory innovation. Based on this, enterprise managers need to dynamically monitor the local efficiency and reach rate of the innovation ecosystem, and when the local efficiency of the cooperation network exceeds the threshold, they should break the path dependence by introducing new cooperation subjects (digital platforms or startups) and developing cross-field cooperation paths (e.g., setting up a special fund for cross-industry technology docking) and optimizing the inter-organizational cooperation structure on a regular basis, in order to continue to unleash the resource integration advantages of the innovation ecosystem. The government should encourage cooperation between digital entities and traditional enterprises, build a platform for inter-enterprise cooperation, and provide policy concessions.

Second, facing exploration may lead to an infinite cycle of “failure traps” [94]. Over-reliance on old resources may form a “capability rigidity trap” [95,96]. Therefore, enterprises need to build an ambidextrous innovation resource allocation mechanism, allocate the R&D budget reasonably according to exploratory innovation and utilization innovation, and enhance the risk-resistant capability with diversified product portfolio and technological innovation capability, so as to cope with market competition and uncertainty risk.

Third, digital transformation plays an important role in connecting the digital innovation ecosystem with enterprise innovation. For example, according to Haier’s annual report 2021, Haier uses digital platforms to connect the resources of many enterprises, realizes digital technology transformation through AI industrial quality inspection system upgrading and production line intelligent transformation, and significantly improves enterprise innovation capability. The hypothesis that digital transformation plays a mediating effect is indirectly verified. Based on this, enterprises need to promote the upgrading of managers’ digital awareness, such as requiring corporate executives to regularly participate in hands-on training such as AI big model and digital governance and leading digitalization pilot projects. In addition, according to our conclusion, AI technology, blockchain technology, and cloud computing technology mainly realize the digital transformation of enterprises through the intelligence of their production processes. Big data technology and digital application technology mainly realize the digital transformation of enterprises through embedding smart products or intelligently acquiring end-user data. Enterprises need to match the corresponding transformation programs according to their own strategic needs and choose the digital technology suitable for enterprise development.

Finally, in response to the double-edged sword effect of the market environment, enterprises should build a “double-cycle cooperation structure” and sign long-term technology-sharing agreements with core local enterprises in their fields. At the same time, they should dynamically dock with emerging partners through innovation platforms and adopt a short-term project-based cooperation approach, so as to establish a dynamic balance between short-term cooperation returns and long-term technology collaboration.