Digital Transformation, Ambidextrous Innovation and Enterprise Value: Empirical Analysis Based on Listed Chinese Manufacturing Companies

: Text analysis was used in this study to construct the index of digital transformation degree in manufacturing enterprises. On this basis, an empirical test was conducted on the relationship between digital transformation, ambidextrous innovation, and enterprise value. The results show that: digital transformation has a signiﬁcant positive impact on the enterprise value of manufacturing enterprises. Digital transformation can promote the rise of enterprise value through technological innovation, business model innovation, and their combination, where ambidextrous innovation plays a mediating role. For manufacturing enterprises, the impact of digital transformation on enterprise value has a single-threshold effect and a double-threshold effect, respectively, with technological innovation and business model innovation being the threshold variables. The synergy of technological innovation and business model innovation plays a stronger mediating role compared with a single innovation model. The conclusions of this study have certain theoretical and practical signiﬁcance for systematically revealing the intermediate mechanism of digital transformation affecting enterprise value and then promoting manufacturing enterprises to make remarkable accomplishments in digital transformation.


Introduction
The manufacturing industry has been the "ballast stone" of China's economy for many years. Since 2010, China's manufacturing added value has accounted for 19.8% of the global manufacturing added value, enabling China to surpass the United States in one fell swoop and continuously maintain its identity as the world's largest manufacturing country [1]. However, China's manufacturing is also faced with two difficulties: first, the contribution of manufacturing to China's GDP has been decreasing in nine consecutive years, down 5.88% from 32.06% in 2011 to 26.18% in 2020; second, China's manufacturing is not strong despite its large scale, with its low-end manufacturing having excess capacity and high-end manufacturing relying heavily on Europe and the United States. This is still the basic situation of China's manufacturing industry. Noticeably, digital transformation is the key to solve this dilemma. Particular to manufacturing enterprises, digital transformation is the adoption of information technology in the production process [2], whereby all links of the enterprise value chain are digitized to realize transformational changes in the business operation flow [3]. In this process, digital transformation works through the coordination or interaction of ambidextrous innovation, namely between technological innovation and business model innovation. Specifically, business model innovation can provide insights into the unmet needs of consumers [4] and point out a clear direction for technological innovation, while technological innovation can change the internal elements required for Sustainability 2022, 14,9482 2 of 20 business model innovation and the way it is implemented [5]. Their mutual complementarity facilitates the integration of internal and external innovation resources [6], gives impetus to the connection among marketing, production, management, and R&D, reduces the cost of data collection, storage, transmission, and analysis, and helps enterprises stay afloat and sustain success in a turbulent environment [7], thus achieving the transformation of the manufacturing industry from low-end to high-end and ensuring the continuous growth of enterprise value. China also regards the digital economy as an important strategy to revitalize the real economy, cultivate new economic momentum, and take the vantage place globally. In the Outline of the 14th Five-year Plan, it is clearly proposed to "create new advantages of digital economy" [8]. In the Guiding Opinions on Accelerating the Cultivation and Development of High-quality Manufacturing Enterprises issued in July 2021 jointly by six departments including the Ministry of Industry and Information Technology and the Ministry of Science and Technology, it is further highlighted that the development scheme of digital transformation in the manufacturing industry should be guided through the implementation of intelligent manufacturing engineering, the actual practice of digital transformation in the manufacturing industry, and the application of 5G in innovation actions [9]. Therefore, considering that the digital economy is a momentous national strategy, digital transformation is no longer a "choice" for manufacturing enterprises, but a "required option" that determines their survival and long-term development.
Under the background of China's focus on the development of the digital economy, relevant data were collected in this study from listed manufacturing companies from 2013 to 2020 to analyze the impact of digital transformation on manufacturing companies' enterprise value and examine the mediating role of ambidextrous innovation, that is technological innovation and business model innovation, in the relationship between digital transformation and enterprise value. The conceptual framework and research idea of this study are shown in Figure 1. This study aims to systematically reveal the intermediate mechanism of digital transformation affecting enterprise value and provide a decision-making reference for promoting digital transformation in the manufacturing industry and boosting the deep integration of the digital economy and the real economy.
elements required for business model innovation and the way it is implemented [5]. Their mutual complementarity facilitates the integration of internal and external innovation resources [6], gives impetus to the connection among marketing, production, management, and R&D, reduces the cost of data collection, storage, transmission, and analysis, and helps enterprises stay afloat and sustain success in a turbulent environment [7], thus achieving the transformation of the manufacturing industry from low-end to high-end and ensuring the continuous growth of enterprise value. China also regards the digital economy as an important strategy to revitalize the real economy, cultivate new economic momentum, and take the vantage place globally. In the Outline of the 14th Five-year Plan, it is clearly proposed to "create new advantages of digital economy" [8]. In the Guiding Opinions on Accelerating the Cultivation and Development of High-quality Manufacturing Enterprises issued in July 2021 jointly by six departments including the Ministry of Industry and Information Technology and the Ministry of Science and Technology, it is further highlighted that the development scheme of digital transformation in the manufacturing industry should be guided through the implementation of intelligent manufacturing engineering, the actual practice of digital transformation in the manufacturing industry, and the application of 5G in innovation actions [9]. Therefore, considering that the digital economy is a momentous national strategy, digital transformation is no longer a "choice" for manufacturing enterprises, but a "required option" that determines their survival and long-term development.
Under the background of China's focus on the development of the digital economy, relevant data were collected in this study from listed manufacturing companies from 2013 to 2020 to analyze the impact of digital transformation on manufacturing companies' enterprise value and examine the mediating role of ambidextrous innovation, that is technological innovation and business model innovation, in the relationship between digital transformation and enterprise value. The conceptual framework and research idea of this study are shown in Figure 1. This study aims to systematically reveal the intermediate mechanism of digital transformation affecting enterprise value and provide a decisionmaking reference for promoting digital transformation in the manufacturing industry and boosting the deep integration of the digital economy and the real economy.
The remainder of this article is organized as follows. The theoretical analysis and research hypotheses are presented in Section 2. Section 3 illustrates the research design and model construction. Section 4 focuses on the empirical analysis of the data and the discussion of the results and findings. Finally, the conclusions, implications, limitations, and future research suggestions are presented, respectively, in the last section.  The remainder of this article is organized as follows. The theoretical analysis and research hypotheses are presented in Section 2. Section 3 illustrates the research design and model construction. Section 4 focuses on the empirical analysis of the data and the discussion of the results and findings. Finally, the conclusions, implications, limitations, and future research suggestions are presented, respectively, in the last section.

Theoretical Analysis and Research Hypotheses
The influencing factors of enterprise value have been more fully explored in the existing articles. At the macro level, scholars have analyzed the factors influencing enterprise value from various perspectives including the uncertainty of economic policy [10] and intellectual property protection [11]. At the micro level, some scholars have analyzed the impact of green innovation [12], organizational capital [13], and social responsibility information disclosure [14] on enterprise value. These studies provide a well-established theoretical basis and rich inspirations for understanding the enhancement and mechanism of enterprise value from a "macro-micro" perspective.
Under the digital economy, digital transformation is an important stage for manufacturing enterprises to realize the shift from low-end manufacturing to smart manufacturing, and this transformation shall be reflected in enterprise value. The existing academic articles mostly focus on digital transformation and enterprise performance, but there is still no consensus. Carlos et al. argued that digital transformation enables manufacturers to obtain greater profits, productivity, and competitiveness [15]. However, the digitalization paradox suggests that although companies may invest in digitalization, they often fail to achieve the expected revenue enhancement [16]. Some scholars have also analyzed the relationship between digital transformation and productivity [17,18] or between innovation performance [19] and business model innovation [20], while there are rare direct studies on digital transformation and enterprise value. Similarly, for ambidextrous innovation, the existing studies are more concentrated on exploratory innovation and exploitative innovation [21], while the matching of technological innovation and business model innovation has not been included in the framework of research on the relationship between digital transformation and enterprise value. In this view, this study intends to theoretically analyze the acting and channel mechanism of "digital transformation-enterprise value among manufacturing enterprises".

Enterprise Digital Transformation and Its Impact on Enterprise Value
In the context of the digital economy, digital transformation has a positive impact on enterprise value. According to the 2021 Digital Transformation Index of Chinese Enterprises released by Accenture, enterprises' digital advantages are doubled and transformed into financial advantages, and the gap with other enterprises is as much as 3.7-times. First, digital transformation improves enterprise value via a data-driven effect. On the one hand, manufacturing enterprises exploit digital technology to build a big data platform, actualize data mining and utilization [22], optimize resource allocation in the production process, and reduce the waste of resources in their operation. On the other hand, through the data-driven effect, manufacturing enterprises can accurately predict and quickly respond to customer needs, perceive digital business opportunities [23], and then increase their enterprise value. Secondly, manufacturing enterprises improve their enterprise value through the connection function enabled by digital transformation. Digital technology has universal connectivity, which can strengthen internal communication within the enterprise, realize the coordination and sharing of information resources among various departments, enhance the cooperation and integration of partners in the business ecosystem, build a multi-party synergistic ecosystem [20], drive content providers to change from selfsearchers in the market to "adherents" in the social market [24], and heighten the efficiency of communication between the enterprise and its external environment, thereby raising the enterprise's value-added. Finally, digital transformation contributes to a higher level of enterprise value by reducing costs. Those manufacturing enterprises with higher sales costs and operating costs are more willing to implement "Internet+" transformation [25]. Reducing transaction costs is the important purpose of digital transformation. For instance, Goldfarb and Tucker (2019) argued that digital technology such as the Internet can cut down the costs of searching, replication, transportation, tracking, and verification [26]. At the same time, digital technology can also help manufacturing enterprises follow the mode of mass production, optimize their production process, augment production efficiency, and Sustainability 2022, 14, 9482 4 of 20 minimize production costs [27]. In addition, the management and decision-making system that relies on data-enabling platforms can effectively obtain useful information, alleviate information asymmetry, increase the accuracy and timeliness of decision-making, decrease operating costs, improve financial performance [28], and drive the increment of enterprise value. Therefore, the following research hypothesis H1 is proposed in this study: Hypothesis 1 (H1). Digital transformation by manufacturing enterprises can improve enterprise value.

Mediating Role of Technological Innovation
The Theory of Endogenous Growth holds that technological innovation is the source of economic growth, providing strategic guidance for enterprises to transform from being factor-driven to being innovation-driven. As the main body of technological innovation, enterprises can not only enhance their production efficiency, but also improve the quality of their products through R&D investment. Obviously, technological innovation plays a significant role in improving enterprise value. On the one hand, digital transformation realizes the digitization of manufacturing enterprises in information, process, production, marketing, and management, offers basic support for the interpretation and output of information, and breaks through the boundary constraints of traditional elements with the help of technological innovation [29], eventually enhancing the production efficiency of manufacturing enterprises for other factors and bringing enterprises more value [30]. For example, blockchain allows for the creation of a publicly accessible and collaborative data flow system that turns data into mobile assets. Therefore, it can be utilized to construct a variety of open and integrated business models that enhance collaboration with other organizations, promote collaborative technological innovation, and also, facilitate the access of businesses to innovation resources. Similarly, digital twin solutions empower companies to combine machine learning, artificial intelligence, and software analytics with the data collected from production plants to create digital simulation models that can reduce costs, minimize investment risks, and increase business productivity [31]. On the other hand, the intelligent production system built through digitization enables employees of medium and large companies to participate in innovation, fosters the flow of information, ideas, and views among employees, produces a significant impact on the initiation of innovation [32], gives play to the spillover effect of knowledge sharing, cultivates employees' ability, stimulates their potential, and empowers them, leading to higher productive value in employees [23]. It also attains and drives the connection of R&D, production, marketing, and other value chain links, advances enterprises' competence in technological innovation, and enables enterprises to maintain their core competitiveness. From these effects, it can be seen that digital transformation plays a decisive role in achieving the maximization of enterprise value [33]. However, the positive effect of digital transformation on enterprise value may be characterized by the existence of a "threshold", depending on the level of technological innovation. A lower level of technological innovation generally prevents digital transformation from exerting a network effect and a spillover effect, thus undermining the promoting effect of digital transformation on the enhancement of enterprise value; when the level of technological innovation is higher than the threshold, its own spillover effect and the network effect of digital transformation come into play. The knowledge sharing spillover effect of technological innovation is geometrically amplified by digital transformation, forming a positive driving system in which digital transformation has an accelerating effect on the enhancement of enterprise value. Therefore, the following research hypotheses H2 and H3 are put forward herein: Hypothesis 2 (H2). Digital transformation by manufacturing enterprises can improve enterprise value by promoting technological innovation.

Hypothesis 3 (H3).
Digital transformation by manufacturing enterprises has a non-linear spillover effect on enterprise value as the level of technological innovation increases.

Mediating Role of Business Model Innovation
Business model innovation is conducive to reshaping the value chain of manufacturing enterprises, upgrading the original trading network of the business ecosystem, boosting the reallocation of resources, and consequently, enhancing enterprise value [34]. Digital transformation is related to the use of digital technology to achieve significant business improvement [35], triggering changes in management structure, decision-making procedures and systems, and giving rise to business model innovation [36]. It has an important impact on organizational change, thus contributing to the growth of enterprise income and the elevation of profits. Meanwhile, digital transformation alters the business operation mode and renders a closer relationship between enterprises and between enterprises and consumers. Consumers and external enterprises' active participation in the process of value creation can abate information asymmetry, weaken the irrationality of managers' decision-making behavior, and thereby, exalt managers' level of corporate governance [37], eventually actualizing the improvement of enterprise value. In addition, digital transformation and business model innovation are continuously blurring the boundaries among transaction subjects in economic activities, resulting in the declining marginal cost of linkage among transaction subjects in the value network. Consequently, each transaction subject is allowed to effectively improve efficiency. Moreover, this effect becomes increasingly obvious as the level of business model innovation increases. In this case, the network effect is established upon enterprises' digital transformation. Accordingly, the following research hypotheses H4 and H5 are proposed in this study:

Hypothesis 4 (H4). Digital transformation by manufacturing enterprises can improve enterprise value by promoting business model innovation.
Hypothesis 5 (H5). Digital transformation by manufacturing enterprises has a non-linear spillover effect on enterprise value as the level of business model innovation increases.

Mediating Role of Technological Innovation and Business Model Innovation
In the research of innovation-driven enterprises, technological innovation and business model innovation constitute an ambidextrous innovation model [38]. Technological innovation ameliorates the production process by developing and introducing new technologies and creates subversive products to reinforce enterprise competitiveness. Business model innovation updates the original transaction structure and transforms technology input into economic value [39]. Evidently, technological innovation and business model innovation are complementary to each other and form a two-wheel driving effect. They work together to bring long-term performance to the business [40]. According to the research by Tong et al. (2021) [41], the matching between technological innovation and business model innovation can be measured by the degree of their combination: combination indicates whether an enterprise has resources to improve the level of one type of innovation while maintaining the level of the other type or whether it has resources to simultaneously make these two types of innovation reach a high level. When the enterprise pushes these two types of innovation jointly to a higher level or improves the level of one type of innovation with the other type remaining unchanged, it is considered to have a high combination. Otherwise, it is believed to have a low combination. A high combination has a positive impact on enterprise performance.
Under the circumstances of digitalization, collaborative innovation driven by "technological innovation + business model innovation" is a powerful engine for enterprise transformation, upgrading, and high-quality development [42]. Manufacturing enterprises can rely on digital technology, especially big data, to attain the informatization of products, businesses, production, marketing, and other processes, make full use of data platforms to collect and analyze information, and even supervise the whole process. Hence, manufactur- ing enterprises can take advantage of digital technology to support the dynamic coordination between technological innovation and business model innovation, so as to raise enterprise value; besides, digital transformation empowers manufacturing enterprises to alter the occurrence of their original business model, shorten the distance with consumers, and grasp the direction for technological innovation through the rapid acquisition and effective utilization of consumption data. In the meantime, the continuous incremental level of technological innovation can further provide strong support for business model innovation. Therefore, digital transformation enables enterprises to foster the connection between technological innovation and business model innovation and carry out value promotion through their combined effect. Therefore, the following research hypotheses H6 is put forward herein: Hypothesis 6 (H6). Digital transformation by manufacturing enterprises can improve enterprise value by promoting the combination between technological innovation and business model innovation.

Data Sources
Given that 2013 is an important time point for the combination of digital reality and virtual reality, the listed companies in Shanghai and Shenzhen A-share manufacturing sector from 2013 to 2020 were selected as the research sample and screened as follows: (1) excluding the companies with missing data and (2) excluding ST and the companies delisting during the period. A research sample of 7312 companies including 914 manufacturing listed companies was finally obtained. The data above are from CSMAR Database, Wingo Textual Analytics Database and CNRDS China Research Data Services Platform.

Explained Variable
Enterprise value: Enterprise value is mainly measured by economic value-added (EVA), ROA, Tobin's Q, and other related indicators of financial performance. In line with the practice by Liu Lifu and Du Jinmin [43], economic value-added (EVA) was used in this study to measure enterprise value. Economic value-added (EVA) can make up for the defect of conventional performance indicators in accurately reflecting the value created by the company for shareholders. It is a financial indicator that reflects the maximization of shareholders' wealth. Economic value-added (EVA) is expressed by net operating profit after tax after deducting all capital inputs including equity and debt. In addition, in order to avoid the possible impact of extreme values on the reliability of empirical results, the method of Winsorize was used for tail reduction of variables at 1% and 99%.

Explanatory Variable
Digital transformation: Some scholars have made beneficial attempts at the quantitative measurement of digital transformation. For example, Qi et al. [37] measured digital transformation by the proportion of the detailed intangible assets related to the digital economy in the total intangible assets at the end of the year disclosed in the notes to listed companies' financial reports. Meanwhile, Zhao et al. [44] adopted "scoring by keyword + expert" to judge the degree of digital transformation in each company. Besides, in the field of text analysis, Hu et al. [45] followed the method of keyword + word2vec similar word expansion to study variables. Word frequency, as a common way to measure concepts, reflects the extent to which the text attaches importance to each word. It provides a useful enlightenment for this study. Therefore, the approach of word frequency analysis, namely keyword + word2vec similar word expansion, was employed in this study to measure the behavior of digital transformation.
Specifically, the existing classic articles that focus on the theme of digitalization, the existing key words that have been used to measure in digital transformation, and key policy documents were drawn in this study to summarize some specific keywords regarding digital transformation such as "digital technology", "digitalization", "intelligence", and "Internet+". Because the same concept is generally expressed by diverse words with similar semantics, it is necessary to conduct similar word expansion on each keyword. In this study, word2vec was applied to obtain the words that have the highest similarity with the existing keywords. The repeated words in the set yielded were deleted, and 188 keywords were finally obtained. Eventually, the proportion of the total word frequency of "digital transformation" in the total word frequency of the MD&A text was multiplied by 100 to obtain the index of digital transformation. Moreover, the method of Winsorize was used for tail reduction at 1% and 99%. When this index is larger, it means that the company has a higher degree of digital transformation.

Mediating Variables
Technological innovation: The number of patents is an important indicator to measure an enterprise's innovation level. The enterprise with a larger number of patents has a higher level of technological innovation. In accordance with the measurement method adopted by Zhao et al. [44], the natural logarithm of the number of patents plus 1 was used for measurement.
Business model innovation: A business model is a system composed of a series of value activities aiming at pursuing business profits, in which value creation, value transmission, and value realization are the three core elements. According to the measurement method adopted by Liu et al. [46], a measurement system was established in this study to cover a total of 9 indicators in three dimensions including value creation, value transmission, and value realization. Business model innovation was measured after determining the index weight through the entropy weight method [47]. Details are shown in the variables' notes in Table 1. Combination of technological innovation and business model innovation: As for the two-wheel driven combination of technological innovation and business model innovation within the enterprise, according to the treatment by Cao [48] and Tong [41], combination was measured by the product of technological innovation and business model innovation.

Controlled Variables
The comprehensive analysis of the impact of digital transformation by listed manufacturing companies on enterprise value requires setting the following controlled variables that may affect enterprise value: (1) enterprise size, presented by the natural logarithm of total assets; (2) enterprise age, presented by subtracting the year of establishment from the current year; (3) current ratio, presented by dividing current assets by total assets; (4) ownership concentration, presented by the sum of the shareholding ratios of the top five major shareholders; (5) capital intensity, represented by dividing total assets by operating income; and (6) CEO duality, being 1 for the CEO duality of the chairman and general manager and 0 otherwise The specific definition of each variable is shown in Table 1.

Model Construction
Total effect model: The following benchmark regression model was set in this study: where the explained variable EVA represents the economic added value of the enterprise; the core explanatory variable DIGI refers to enterprise digital transformation; Controls means the controlled variable described above; µ Y is the year-fixed effect; δ I is the industryfixed effect. Mediating effect model: The same explanatory variable and controlled variables in model (1) were used to construct the model (2)~(3): MEDIATOR in Models (2) and (3) is the mediating variable, referring to technological innovation, business model innovation, and the combination of technological innovation and business model innovation, respectively. Threshold effect model: The following threshold effect model was set in this study: where Adj it is the threshold variables, referring to technological innovation and business model innovation; I(·) is an exponential function that takes the value of 1 in the case where the conditions are met or 0 otherwise. Model (4) is a single-threshold model, which can be transformed to a multi-threshold model, depending on the number of thresholds. Table 2 shows the results of the descriptive statistical analysis of each variable. From the means of the whole sample, it can be seen that the index of digital transformation has the following characteristics: the average word frequency of digital transformation is 0.333, which means that every 100 words in MD&A contain 0.333 digital transformation words; the minimum is 0, and the maximum is 1.941, indicating that there are significant differences in digital transformation among manufacturing enterprises. Additionally, the maximum of technological innovation is 9.549, and the minimum is 0, indicating that there are significant differences in technological innovation among manufacturing enterprises. According to the description of the variables related to enterprise economic added value, the maximum of economic added value is 45.58 and the minimum is −19.51, indicating that there are great differences in enterprise value among manufacturing enterprises.

Correlation Analysis
Before the regression analysis of the model, the method of the Pearson correlation coefficient was used to test the correlation between different variables. The specific results are shown in Table 3.  It can be seen from the results in the table that all the correlation coefficients are lower than 0.5, indicating that there is no multicollinearity among the variables of the model. Meanwhile, the correlation coefficient matrix shows that enterprise value has a significantly positive correlation with both technological innovation and business model innovation. To some extent, this suggests that technological innovation and business model innovation may have an impact on enterprise value. There is a significantly positive correlation between the explanatory variable digital transformation and the explained variable enterprise value at the level of 1%. H1 could be preliminarily verified, thus laying a foundation for further verification of the acting mechanism between variables by regression analysis.

Regression Analysis
The STATA software was used to carry out multiple linear regression on the relationship between "enterprise digital transformation and enterprise value" in two steps. Yearand industry-fixed effects were only controlled in Model (1), while Model (2) had controlled variables added (enterprise size, enterprise age, current ratio, ownership concentration, capital intensity, and CEO duality) on the original basis. Table 4 reports the core results of testing the relationship between "enterprise digital transformation and enterprise value". In the benchmark regression, the regression coefficient of DIGI in Model (1) is 1.740, significantly positive at the level of 1%; the regression coefficient of DIGI in Model (2) is 0.809, smaller than that in Model (1), probably due to the absorption of some factors affecting enterprise value after adding the controlled variables. However, the value of R2 increases and passes the statistical significance test at 1%. This shows that the enhancement of digital transformation in manufacturing enterprises would significantly improve enterprise value, and H1 can be verified.

Mediating Effect Analysis
The mediating effect model proposed by Wen and Ye (2014) [49] was applied in this study to assist the discussion on the mechanism path of enterprise digital transformation affecting enterprise value. The corresponding results are shown in Table 5. Note: ***, **, * represent significance levels of 1%, 5% and 10% respectively. Table 5 show the results of testing H2. Accordingly, technological innovation increases significantly with the rise of enterprise digital transformation and thereby promotes the improvement of enterprise value. Column (2) shows the degree to which digital transformation affects technological innovation. The regression results show that the regression coefficient between digital transformation and technological innovation is 0.637 and significant at the level of 1%. This indicates that the degree of technological innovation is higher when the degree of digital transformation is higher. Column (3) reveals the joint impact of technological innovation and digital transformation on enterprise value. These two variables both pass the significance test. The regression coefficient of digital transformation is 0.537, which is less than 0.809 in the total effect regression. It can be known that technological innovation exerts a partial mediating effect. The results above suggest that H2, that digital transformation affects enterprise value through technological innovation, can be verified.

Columns (1)-(3) in
Columns (4) and (5) in Table 5 show the results of testing H4. In Column (4), the coefficient of the index of digital transformation is significantly positive (significant at the confidence level of 1% with the coefficient being 0.002), indicating that there is more sufficient business model innovation when the degree of enterprise digital transformation is higher. In Column (5), the coefficient of digital transformation and business model innovation is significantly positive, indicating that business model innovation can promote the improvement of enterprise value. In this sense, business model innovation also exerts a partial mediating effect. Therefore, business model innovation plays a mediating role in the relationship between enterprise digital transformation and enterprise value, suggesting that H4 is established. Table 5 also shows the results of identifying the mechanism concerning the "combination of technological innovation and business model innovation". It can be found that the regression coefficient of enterprise digital transformation on the combination of technological innovation and business model innovation is positive and highly significant. This means that digital transformation can promote the matching of technological innovation and business model innovation and provide a strong force to drive the promotion of enterprise value. Specifically, in Column (6) of Table 5, the coefficient of digital transformation is significantly positive (significant at the confidence level of 1% with the coefficient being 0.132), suggesting that digital transformation can improve the combination of technological innovation and business model innovation. The coefficient of digital transformation in Column (7) is 0.435, passing the statistical significance test at the confidence level of 10%, and decreases to some extent compared to its value when technological innovation and business model innovation are the mediating variables. Meanwhile, the combination of technological innovation and business model innovation passes the significance test at the confidence level of 1%, with the coefficient being 2.841. Accordingly, the combination of technological innovation and business model innovation has a stronger mediating effect than a single innovation model. Therefore, H6 can be verified, that is digital transformation affects enterprise value through the combination of technological innovation and business model innovation.

Threshold Effect Analysis
Digital transformation drives manufacturing enterprises to obtain higher enterprise value, but this driving effect may not be significant before technological innovation and business model innovation have reached a certain level, suggesting that digital transformation probably has a "threshold effect" on the enhancement of enterprise value. The threshold effect analysis proposed by Hansen [50] was applied in this study with technological innovation and business model innovation being the threshold variables. The results of 300 iterations of the bootstrap method, where the xthreg command [51] was used to process relevant data, show that the threshold variable of technological innovation significantly passes the single threshold other than the double and triple thresholds. In contrast, business model innovation significantly passes the double threshold, as shown in Table 6. Meanwhile, the likelihood ratio function Figures 2 and 3 shows that the single threshold of technological innovation is 5.2470 and the double thresholds of business model innovation are 0.1768 and 0.1858, respectively. The regression model with the corresponding number of thresholds was set up in the test above to validate the existence of a threshold. The regression results are shown in Table 7. The results in Column (1) obtained from using technological innovation as the threshold variable show that the impact of digital transformation on enterprise value has a significant interval effect as technological innovation changes from a low to a high level. Besides, the marginal interval effect coefficients are 0.579 and 3.045. As the level of technological innovation increases, the impact of digital transformation on enterprise value changes from a non-significant to a significant level. In other words, the degree to which digital transformation by manufacturing enterprises affects enterprise value differs significantly at different technological innovation degree intervals. When manufacturing enterprises have a high level of technological innovation (RES > 5.2470), the marginal effect coefficient is 3.045, passing the significance test at the confidence level of 10%, five-times that of enterprises with a low level of technological innovation. This shows that the promoting effect of digital transformation on enterprise value has a non-linear spillover among manufacturing enterprises with the increase of technological innovation level. The possible reason is manufacturing enterprises with a higher level of technological innovation are more able to support their digital transformation. Their innovation awareness and innovation concepts are more aligned with digital transformation, which in turn improve the rate of achieving success through technological innovation, thus accelerating the increase of their own enterprise value.
significantly passes the single threshold other than the double and triple thresholds. In contrast, business model innovation significantly passes the double threshold, as shown in Table 6. Meanwhile, the likelihood ratio function Figures 2 and 3 shows that the singl threshold of technological innovation is 5.2470 and the double thresholds of busines model innovation are 0.1768 and 0.1858, respectively.    significantly passes the single threshold other than the double and triple thresholds. In contrast, business model innovation significantly passes the double threshold, as shown in Table 6. Meanwhile, the likelihood ratio function Figures 2 and 3 shows that the single threshold of technological innovation is 5.2470 and the double thresholds of busines model innovation are 0.1768 and 0.1858, respectively.     Note: ***, **, * represent significance levels of 1%, 5% and 10% respectively.
The regression estimation results in Column (2) obtained from using business model innovation as the threshold variable show that there is a strong threshold relationship between digital transformation and enterprises value. Noticeably, the effect of digital transformation on enterprise value passes the 1% statistical significance test in all cases. When manufacturing enterprises' business model innovation is less than 0.1858, there is a negative correlation between digital transformation and enterprise value. However, the degree to which digital transformation affects enterprise value decreases from −7.895 to −1.953 as the level of business model innovation increases. To be specific, when business model innovation is less than 0.1768, digital transformation increases by 1% and enterprise value decreases by 7.895%. With business model innovation between 0.1768 and 0.1858, digital transformation increases by 1% and enterprise value decreases by only 1.953%. This indicates that at the early stage of digital transformation, the mismatch between the business model and digital transformation due to the low level of business model innovation and the failure to give full play to the role of digital technology in enhancing enterprise value drive digital transformation to have a negative impact on enterprise value. Nevertheless, this negative impact decreases as the level of business model innovation increases. When the level of business model innovation is greater than 0.1858, the impact coefficient is 1.392, suggesting the generation of a significant positive promoting effect. Accordingly, when manufacturing enterprises' business model innovation is in a high-level range, it has a synergistic effect with digital transformation and contributes to the positive promoting effect of digital transformation on enterprise value. The possible reason is that after the level of business model innovation is improved considerably and reaches the second threshold, it brings about the upgrade of the organizational structure and value network. Under these circumstances, enterprises can perceive information regarding products, customers, and markets in time and rapidly iterate on innovation. Along with the Metcalfe effect of digitalization, the spillover effect of digitalization can fully reduce the marginal cost and the transaction cost accumulated by users. At the same time, digital transformation and the business model match and coordinate with each other more effectively, leading to a great increase in manufacturing enterprises' productivity and a subsequent rise in their enterprise value.

Heterogeneity Analysis
Due to the differences among manufacturing enterprises in attribute characteristics, it is likely that digital transformation varies in its impact on enterprise value under the background of the digital economy. In this regard, it is possible to start from the micro characteristics of manufacturing enterprises. On the one hand, the enterprises in the sample were divided into state-owned enterprises and non-state-owned enterprises for the grouping test according to their ownership nature; on the other hand, they were divided into large-scale enterprises and small-scale enterprises for the grouping test according to their enterprise size. The division by enterprise size followed the practice by SunHui [47], where the mean of enterprise size (22.30) is the critical value. A company is a large-scale enterprise when its size is greater than the mean or a small-scale enterprise when its size is lower than the mean.
It can be seen from Columns (1) and (2) of Table 8 that enterprise digital transformation significantly improves the enterprise value of non-state-owned enterprises and state-owned enterprises, but it has a greater role in promoting the enterprise value of state-owned enterprises. The possible reason lies in the closer ties between state-owned enterprises and the government. Under the background where the digital economy has become a national strategy, state-owned enterprises are more willing to cater to the government's policy guidance and have a strong desire to foster digital transformation. In addition, their capital, scale, and policy advantages make it easier for them to integrate digital transformation with their own strengths and enhance their enterprise value. Compared with state-owned enterprises, non-state-owned enterprises often do not have capital advantages to invest sufficiently in the long, complex project of digital transformation and generally fail to meet the capital requirement of long-term investment. As a direct result, they "dare not transfer" or "cannot transfer". In addition, since they are constrained by the impact from the external environment and their own weak ability to resist market risks, they lag behind state-owned enterprises in the effect of digital transformation.
Columns (3) and (4) in Table 8 list the results of the test regarding the difference in the attribute characteristic of enterprise size. Compared with small-scale enterprises, large-scale enterprises receive a more obvious effect of enterprise digital transformation in improving enterprise value. It is therefore argued here that large-scale enterprises have easier access to the scale effect, stronger anti-risk ability, and considerable financial support for digital transformation and are more likely to give full play to the effect of digital transformation; on the contrary, it is difficult for small-scale enterprises to break through the boundary of resource constraints. Since the capital guarantee required for digital transformation cannot be met, digital transformation produces a smaller effect in enhancing enterprise value. Note: ***, **, * represent significance levels of 1%, 5% and 10% respectively.

Robustness Test
The robustness of the model was tested in this study to further enhance the persuasiveness of the core conclusions. Firstly, it was tested by replacing the explanatory variable. In this study, economic value-added (EVA) was replaced with the values of Tobin's Q, ROA, and ROE to test the regression relationship between digital transformation and enterprise value and the corresponding mediating effect. The results show no substantive differences, suggesting that the empirical results are still robust. Secondly, it was tested by examining the mediating effects. The Sobel test and bootstrap were used in this study to verify the robustness of the mediating results, as shown in Table 9. According to the Sobel test results, the Z values of technological innovation, business model innovation, and the combination of technological innovation and business model innovation are 6.78, 3.783, and 8.31, respectively, all passing the statistical significance test. This indicates that the three mediating effects are established. Among them, the proportions of these three mediating effects are 33.6%, 26.7%, and 46.2%, respectively, indicating that the effect of the matching of technological innovation and business model innovation is better than that of a single innovation model. As shown by the results of bootstrapping by 1000 times, the confidence intervals of the three mediating paths do not contain zero, indicating that the results are robust. Therefore, the integration of the two methods shows that the conclusions are still robust.

Research Conclusions and Implications
The 914 A-share manufacturing listed companies in Shanghai and Shenzhen from 2013 to 2020 were taken as the research sample in this study to explore the degree of enterprise digital transformation through text analysis. At the same time, the fixed effect model was applied to conduct empirical analysis and economic explanation of the relationship and mechanism of "digital transformation-enterprise value". The following conclusions were drawn: (1) Digital transformation has a significant positive impact on enterprise value. (2) Technological innovation and business model innovation play a mediating role in the process of digital transformation affecting enterprise value. The enhancement of ambidextrous innovation can effectively boost enterprise value. (3) With the improvement of technological innovation and business model innovation, digital transformation by manufacturing enterprises has a non-linear spillover effect on the improvement of enterprise value, indicating that technological innovation, business model innovation, and digital transformation can jointly form a driving force for the improvement of enterprise value among manufacturing enterprises. (4) Compared with a single innovation model such as technological innovation and business model innovation, the combination of technological innovation and business model innovation has a more significant impact on enterprise value. The matching relationship between technological innovation and business model innovation has an enhanced mediating effect on the increase of enterprise value.
There are the following policy implications from this study. First, the government and relevant departments should continue to promote the construction of the digital economy infrastructure such as big data and artificial intelligence, thereby laying a solid foundation for enterprises in various fields to engage in digital transformation. At the same time, it is necessary to build a digital supply chain development system for the manufacturing industry, give public access to digital resources and capabilities, and implement the inclusive service of "cloud-supported acquisition of digital intelligence", so as to create a favorable external service environment for manufacturing enterprises to pursue digital transformation. Furthermore, the government should support and participate in construction of innovation platforms such as manufacturing innovation centers and knowledge sharing innovation centers, build the connection between the government, schools, enterprises, and scientific research institutions, shape a strategic alliance of ambidextrous innovation, carry out collaborative innovation, increase the research and demonstrative applications in the core fields of digital economy such as 5G application and network security, and make vigorous efforts to protect the intellectual property of innovation achievements. Second, the government should combine the local characteristics and resource advantages of each region and manufacturing enterprises themselves and formulate targeted policies and measures to support the digital transformation of manufacturing enterprises. In particular, small-and medium-sized private manufacturing companies are often constrained by the cost of investment in digitalization and external factors, while digital transformation can promote the transformation of labor-intensive enterprises to intelligent manufacturing enterprises and reduce the pressure caused by rising labor costs. Therefore, those policies that benefit enterprises should be effectively implemented to foster the digital transformation of manufacturing enterprises. Third, manufacturing enterprises should seek digital transformation methods that optimize business information management through the deep incorporation of digital technology into sales, production, and management, and thus realize network collaboration and intelligent manufacturing. Moreover, they should also establish a digital management team, build a data management platform, and use massive data to construct a more comprehensive business evaluation system to improve their operation efficiency and accelerate the increase of enterprise value. Fourth, enterprises should continue to upgrade their innovation ability. On the one hand, technological innovation should be exploited to improve their product competitiveness, maintain their competitive advantage, and create more market value; on the other hand, business model innovation is a key way to mobilize the innovation enthusiasm of all players, attain the process innovation of sales, management, and production activities, and form a new pattern of digital manufacturing. In the meantime, they should attach importance to the synergy between technological innovation and business model innovation and improve the level of comprehensive innovation through engaging in technological innovation under the assistance of business model innovation, so as to ensure the synergy between technological innovation and business model innovation, give full play to the maximum utility, and raise enterprise value to the largest extent.

Research Limitations and Prospects
As the research on enterprise digital transformation is still in its infancy, the mechanism of digital transformation affecting enterprise value was explored in this study merely from the perspective of innovation. It is suggested to further study this black-box mechanism from different perspectives in the future. In addition, it is still necessary to discuss how the matching relationship between technological innovation and business model innovation affects enterprise value. Including manager behavior in the framework of their matching relationship can be a further research direction.