Digital Transformation and Manufacturing Firm Performance: Evidence from China

Abstract

Based on the digital transformation practice of Chinese manufacturing enterprises, this paper sorts out the intrinsic mechanism of digital transformation affecting the performance of manufacturing enterprises systematically, based on the perspective of digital investment, and empirically tests the theoretical hypothesis using an unbalanced panel of China’s A-share listed manufacturing companies in Shanghai and Shenzhen from 2007 to 2020. The findings show that digital transformation enhances manufacturing firm performance significantly, and the conclusions still hold after using dynamic panel models, the instrumental variables approach, and a series of robustness tests; further analysis reveals that this effect is more significant in state-owned enterprises and manufacturing enterprises in regions with a higher degree of marketization. The results of the mediating effect model show that low-cost empowerment and innovation empowerment are important channels to improve the performance of manufacturing enterprises. Accordingly, this paper argues that accelerating the deep integration of digital technology and enterprise development is the key to achieving high-quality development and improving enterprise performance in the context of an uncertain business environment.

1. Introduction

In the post-epidemic era of the booming digital economy, digital transformation, which has become a major breakthrough in corporate transformation and upgrading, is an inescapable core mission for companies [1,2,3]. The potential of top-line growth and bottom-line savings from digital transformation projects exceeds that of other types of transformation initiatives [4,5,6]. In recent years, the rapid development of digital technologies, represented by 5G, cloud computing, artificial intelligence, etc., has continued to give rise to new forms and models that meet market needs and drive the digital transformation of real enterprises. According to the White Paper on China’s Digital Economy (2020) (https://baijiahao.baidu.com/s?id=1698072368884815563&wfr=spider&for=pc, accessed on 8 June 2022), digital technology is the core driver of the digital economy, and the integration of digital technology and the real economy has not only improved the digitalization, networking, and intelligence of enterprises but also accelerated the reconfiguration of economic development and governance models [5]. Currently, digital technologies have invaded all business sectors such as retail, advertising, healthcare, finance, manufacturing, and education [7]. Undoubtedly, companies are reshaping business models and industry boundaries through digital transformation [8]. Presently, against the background of the rapid development of technologies such as big data, artificial intelligence, and blockchain, the digital transformation of Chinese fintech companies is in full swing. However, whether the digital transformation of enterprises generates positive feedback on the value of manufacturing enterprises themselves is a question worth studying as well as the paths by which digital investment in enterprises affects corporate performance and whether there is a difference in the impact of different components of digital transformation on enterprise performance.

Digital transformation is the process by which an organization responds to environmental changes to transform its value creation process using digital technologies such as mobile computing, artificial intelligence, cloud computing, and the Internet of Things [9], and digital transformation has become a means for entrepreneurs to internalize external pressures as a motivation for change in the face of a competitive and unpredictable external environment [10]. A review of the existing literature reveals that the studies related to the topic of this paper can be divided into two main categories, the first is about the connotation and measurement of digital transformation in enterprises. From the perspective of capabilities, digital transformation facilitates the process of change and innovation in business models, customer systems, and utilization patterns [11,12]. A representative definition is the “structural change in which digital technology is integrated with business processes”, where the existing organizational, operational, and business models are changed or innovated by introducing digital technology [13]. The second is that some studies define a digital transformation from a process perspective. It is emphasized that digital technology facilitates the process of change and innovation in business models, customer systems, and utilization models, and makes effective changes to the organization to achieve access to organizational performance. Finally, some scholars define a digital transformation from the perspective of capabilities. For example, Berghaus and Back [14] consider digital transformation as digital innovation focused on improving the digital capabilities of the firm. Another category is the research on the impact of digital transformation on firms. Regarding the impact of digital transformation, scholars agree that the companies that have implemented digital transformation, taking advantage of the information, communication, and computing of digital technologies such as big data and cloud computing, can strengthen the inter-collaborative relationship between different participants inside and outside the companies to reduce production costs and, thus, improve the efficiency of business operations [15]. Domestic scholars have now explored the impact of digital transformation in depth and concluded that digital transformation is conducive to promoting enterprise effectiveness [16]. In addition, organizational learning theory points out that enterprises with a strong learning ability can learn external digital technology and knowledge faster and more effectively, and the improvement of learning ability needs management guidance and cultivation, such as holding regular industry sharing sessions and inviting research institutes and other institutions to organize regular exchanges on enterprise digital transformation. What is more, the establishment of cooperative relationships between enterprises and their peers, as well as industry chain enterprises, are beneficial to reducing the cost of carrying out innovation activities and obtaining precious information such as market conditions and technological changes promptly [17,18]. Moreover, digital transformation enhances the ability to predict the financial risks of enterprises, which is one of the effective ways to warn the financial risks of enterprises [19,20]. Meanwhile, both aspects of big data resources and integrated utilization of big data resources have significant positive effects on market performance and the financial performance of enterprises [21].

In summary, it can be seen that domestic and foreign scholars have made some valuable achievements in these studies through case studies, logical reasoning in economics and management, and empirical analysis, and now there are some shortcomings in the research on digital transformation in enterprise performance, including the following aspects: first, the research on the performance of manufacturing enterprises. Due to the special nature of the manufacturing industry, the current research on the performance of manufacturing enterprises is still inadequate. Second, there is still a relative lack of research on the impact of digital transformation on the performance of manufacturing enterprises. More studies regard the improvement of information technology level as an important feature of digital transformation, but, from the profound connotation of digital transformation, it is still insufficient to evaluate the digital transformation of manufacturing enterprises by only applying the information technology or investment level. Third, the impact of how digital transformation affects the performance of manufacturing enterprises is still relatively small, and the existing research on this is still relatively limited and needs to be systematically developed further.

Given this knowledge, this paper empirically investigates the impact and mechanism of digital transformation on the performance of manufacturing enterprises from the perspective of enterprise digital investment based on the data of a sample of manufacturing enterprises listed in Shanghai and Shenzhen A-shares in China from 2007 to 2020. Compared with the previous studies, the incremental work and marginal contributions of this paper are: first, based on the existing studies, this paper further clarifies the behavioral nature of enterprise digital transformation, combining the theories related to digital transformation with the digital transformation strategy practice of manufacturing enterprises, and helps to deeply understand the status and effectiveness of digital transformation of manufacturing enterprises in China. Second, distinguishing itself from the existing studies [3], this paper reconstructs enterprise digital transformation indicators from the perspective of digital investment, based on the existing methods, and combines the digital transformation practices of manufacturing enterprises to reveal the impact of digital transformation on the performance of manufacturing enterprises, which is an effective supplement to the previous studies, enriching the previous studies on the impact effects of enterprise digital transformation. Third, this paper explores the mechanism and heterogeneity of the effect of digital transformation on the performance of manufacturing firms. The heterogeneous effects of digital transformation on the performance of manufacturing firms are explored in terms of the nature of property rights and the degree of marketability, and the mechanism of the effect of digital transformation on the performance of manufacturing firms is revealed in terms of low-cost empowerment and innovation empowerment.

2. Theoretical and Hypothesis

2.1. Digital Transformation and Manufacturing Enterprise Performance

The difference between digital transformation and the term “digitalization” is that digitalization covers the digitization of data from paper-based carriers to electronic information, which is also referred to as e-transformation. Digitization is needed to optimize the digital transformation context, however, digitization is not the same as the term digital transformation [22]. Companies are increasingly aware of the need to adapt their activities, strategies, and routines to the challenges of the “new normal” [23], and they have started to optimize and improve the efficiency of their processes through digital tools to ensure business continuity [24]. Currently, the digital transformation of enterprises in China has made some achievements, digital management is popular in half of the enterprises, with manufacturing companies far ahead in development, and intelligent production and network collaboration are growing steadily every year, leading to intelligent exploration. Moreover, digital investment has a positive contribution to the dynamic capabilities of enterprises, and individual forgetfulness and entrepreneurial orientation play a positive moderating role between the digital investment and dynamic capabilities of enterprises as well as innovation performance [25]. On the one hand, digital transformation helps to enhance the competitive ability of the firm. Digital transformation facilitates the enhancement of the firm’s understanding of the resources owned by the firm and the degree of utilization of these resources, and the integration of the firm as a heterogeneous resource, in turn, enhances the firm’s competitiveness. This is because digital resources are scarce, valuable, difficult to be replaced, and difficult to be imitated; digital assets have become an important source for enterprises to continuously gain competitiveness, and the digital resources owned are not only a direct source of competitiveness for enterprises but also can be transformed into knowledge through data to gain continuous competitiveness. On the other hand, enterprises empower the business process through new technologies such as big data, artificial intelligence, and blockchain, which in turn improve business performance. In the real business process, enterprises use the existing measurement of new technologies to digitally transform existing products and services or develop new digital products and services which accelerates enterprise innovation, explores new market opportunities, and explores new business models, which can improve the added value of products and thus enhance the competitiveness of products in the industry. In addition, data-driven processes and smart manufacturing can improve the efficiency of project management and professional work, increase capacity utilization, and achieve cost reduction and efficiency gains [13]. It can be seen that the company’s digital transformation business strategy choices reflect its behaviors and efforts to achieve corporate excellence and its desire to increase the value of the company by integrating different digital technologies into its operations and processes. The company’s digital transformation strategy elevates its culture to maximize the value of the company. At the same time, digital transformation improves communication within the company [23]. Through digital transformation, management communicates more effectively with shareholders, employees, and other stakeholders. The improved communication provides cost savings to the company. Moreover, digital transformation can help companies develop new networks and improve their international competitiveness. In essence, digital transformation reduces organizational barriers [26]. Thus, digital transformation-enhanced companies support their access to new forms of knowledge and relationships, resulting in better resources for innovation and future development, and market internationalization. In short, digital transformation strategies improve the operational efficiency, cost savings, and innovation of companies [7]. As a result, companies that underwent digital transformation performed better than those that did not.

However, the historical experience of ebb and flow is always a constant reminder that several flaws of today’s digital transformation have become apparent. Companies that actively embrace digital transformation should both actively join the flood and think calmly, trying to avoid falling into various pitfalls [27]. One is the strategic trap that companies are rushing to provide digital solutions as a new growth point. In the initial stage of digital transformation, for most companies, the first thing to do is to try to realize the business process of their own business, increase the added value of their products, and then improve the service capability and market competitiveness of their products. Additionally, now more and more companies are keen to provide digital solutions as an important business and new growth point in the future. The second is the organizational trap, that is, the hope that the information technology department will assume the heavy responsibility of digital transformation, ignoring at least two important risks: on the one hand, the existing enterprise information technology department lacks the business capabilities to encourage what the digital transformation should be, and on the other hand, the existing information technology department may reduce the importance of the digital transformation work. The third trap is the tool trap, that is the thinking that the introduction of digital transformation tools will be all right, or blindly introducing a variety of the latest tools with the lack of knowledge of their core technological innovation, business model innovation, and other aspects of sustained new breakthroughs. The fourth, is the governance trap, that is, thinking that seizing the data has an advantage. The ultimate value of data lies in its analysis to obtain results that can be used for forensic and predictive purposes, as well as to promote business improvement, and the value of data needs to be realized in the transmission and analysis of applications. The fifth is the performance trap, which is to assume that digital transformation will yield immediate results. Digital transformation is a strategic action that requires long-term practice, long-term investment, and a focus on the long-term performance improvement of the business [28]. Moreover, business managers may not fully implement the company’s digital transformation strategy. In the presence of agency conflicts, managers may not act in the best interests of shareholders. Instead, they will seek their private interests [29]. Therefore, digital transformation strategies may not imply better corporate performance. Accordingly, we propose the following competing research Hypotheses H1a and H1b.

Hypothesis 1a (H1a).

Digital transformation enhances manufacturing firm performance and there is a positive relationship between the two.

Hypothesis 1b (H1b).

Digital transformation decreases the performance of manufacturing firms, and they are negatively correlated.

2.2. Digital Transformation, Heterogeneity, and Manufacturing Firm Performance

The ownership structure of enterprises affects the difficulty of accessing resources for digital transformation. Currently, in the context of deepening the reform of state-owned enterprises, the state has clear policy instructions for state-owned enterprises to become bigger and stronger. SOEs often have significant comparative advantages in physical resources such as resources and technology, as well as soft factor resources such as policy information due to their special status [30], and through digital transformation, SOEs are more likely to improve their corporate performance [31]. Meanwhile, according to the corresponding view of the scale effect, when technology is at a certain level, the expansion of scale also helps to enhance advantages and efficiencies, which in turn promotes higher firm performance. Enterprises with more resources can better utilize the advantageous resources of information networks, give full play to digital advantages, and form a strong effect. Compared with non-SOEs, SOEs usually have strong advantages in capital supply and resources. With the implementation of modern enterprise management system reform in recent years, SOEs can more fully run their accumulated organizational management and business experience to maximize the advantages brought by digital transformation to improve enterprise performance. In addition, in the process of digital transformation, SOEs can make full use of their state-owned capital background and government backing to quickly concentrate the required digital transformation resources and establish a perfect digital operation system and structure, which, in turn, can promote the rapid improvement of enterprise performance [32]. Therefore, we propose the research Hypothesis H2.

Hypothesis 2 (H2).

Qualifying other conditions constant, state-owned manufacturing firms are more likely to significantly amplify the positive impact of digital transformation on firm performance.

Digital transformation is a new service industry and model, and the development of this new industry and model faces obstacles in terms of institutional mechanisms. A high level of marketization helps business managers to devote more attention to productive activities [33]. The level of economic development is more uneven across China, and in regions with a high level of marketization, the rules in business interactions are clearer, which can also provide favorable conditions for the corresponding digital transformation as a new service industry and a new role to play. Firstly, regions with a high level of marketization tend to be more receptive to new things, and the digital transformation function can be well played. Second, a high level of marketization can help to avoid the adverse effects of the early digital economy stage faced with greater obstacles and possible efficiency losses, providing good conditions and support for the digital transformation of manufacturing enterprises. Moreover, a high level of marketization can help reduce the institutional transaction costs of manufacturing enterprises in the process of digital transformation. Finally, a high level of marketization also helps to circumvent the negative effects caused by the uncertainty and complexity of the internal and external environments of enterprises [34]. Therefore, digital transformation may have a more significant effect on improving the performance of manufacturing firms in regions with a higher degree of marketization, leading us to propose the research Hypothesis H3.

Hypothesis 3 (H3).

Qualifying other conditions constant, manufacturing firms in regions with higher levels of marketization are more likely to significantly amplify the positive impact of digital transformation on firm performance.

2.3. Digital Transformation, Channel Mechanism, and Manufacturing Enterprise Performance

Through digital transformation, manufacturing enterprises can optimize different aspects of their logistics, production, and back-end services to significantly reduce their costs and achieve low-cost empowerment. Digital transformation can help facilitate the creation, analysis, and utilization of large amounts of data, build an intelligent logistics supply chain by using modern technology, innovate the means of communication, and empower the overall operational efficiency of the entire industrial chain with industry chain integration and deep user involvement [35]. In addition, digital transformation helps to improve traditional work processes and methods and reshape organizational structures by using modern digital technologies such as advanced and effective big data, artificial intelligence, cloud technology, and the Internet of Things. Companies can effectively reduce the redundant resources of the enterprise by optimizing the existing business processes, improving the efficiency of the entire industrial chain by vigorously developing intelligent logistics, and laying out the development of digital warehousing logistics and intelligent suppliers and warehouse management using intelligent logistics and smart procurement systems [36]. Especially important, digital transformation greatly reduces the human capital cost of enterprises, which can effectively enhance the scale of operation and strengthen management to reduce costs and increase efficiency. In the current context of China’s aging population and slowing urbanization, enterprises are generally facing rising labor costs, and modern intelligent industrial robots, an efficient labor alternative, can further reduce the human capital costs of enterprises based on optimizing their business patterns and business models, effectively reducing their operating costs. Therefore, we propose the research Hypothesis H4.

Hypothesis 4 (H4).

Digital transformation enhances the performance of manufacturing enterprises through low-cost empowerment.

Digital transformation also helps facilitate the transformation of firms’ innovation potential into reality, and these innovations can significantly change firms’ existing business models and value propositions [37]. Digital transformation drags business model reconfiguration and innovation by reshaping industries, building alternative products, and creating new business service and value payment models [38]. First, digital transformation makes knowledge resources richer, innovation comes from the recombination of existing knowledge, and enterprises can better perceive industry or competitors’ information and services promptly through digital transformation as well as data platforms, artificial intelligence, and other data analysis methods to obtain more information about enterprise production and operation and product technology, thus providing support for enterprise innovation. Second, digital transformation can change the innovation process of enterprises. Enterprises can effectively accelerate their innovation process by replacing the original model with microservices [39,40] and the DEVOPS development model [41] through digital transformation. Third, digital transformation promotes open innovation in enterprises and enhances the ability of enterprises to utilize and develop external innovation resources, which can achieve a better collection of high-quality ideas and accelerate innovative product development. Therefore, we propose the research Hypothesis H5.

Hypothesis 5 (H5).

Digital transformation enhances the performance of corporate manufacturing firms through innovation empowerment.

3. Materials and Methods

3.1. Sample and Data Collection

This paper selects the data of Chinese manufacturing companies listed in Shanghai and Shenzhen A-shares from 2007 to 2020 as the initial research sample to empirically examine the impact of digital transformation on the performance of manufacturing companies. This paper treats the initial research sample as follows: considering the special characteristics and possible effects of market trading behavior of delisted ST companies, ST companies are excluded from the sample period; financial companies and companies with serious data deficiencies are excluded, and the unbalanced panel data consisting of 5569 company-annual observations of 911 listed companies are finally obtained after processing. In addition, to exclude the influence of extreme outliers on the empirical results, all continuous variables in this paper are shrunken at the 1% and 99% quartiles. The financial data used in this paper is obtained from the CSMAR database (https://www.gtarsc.com/, accessed on 8 June 2022), and the corporate digital transformation data are obtained from the CSMAR (https://www.gtarsc.com/, accessed on 8 June 2022) and Wind (https://www.wind.com.cn/, accessed on 8 June 2022) databases.

3.2. Variable Definition

3.2.1. Dependent Variables

Corporate performance (ROA). The main measures of corporate performance in the existing literature are the return on assets (ROA), return on net assets (ROE), and TobinQ value (TobinQ). Since the Tobin Q (TobinQ) value is a corporate market indicator and the return on assets (ROA) and return on net assets (ROE) are corporate financial indicators, this paper, based on combining the financial indicator system of the DuPont analysis and referring to related studies [42,43,44], selects the return on assets (ROA) as an evaluation indicator of manufacturing enterprise performance and selects the return on net assets (ROE) as an alternative indicator of enterprise performance for robustness testing.

3.2.2. Independent Variable

Digital transformation (DT). According to the portrayal pattern of previous studies, there are four main ways to portray the digital transformation of enterprises: First, the digital transformation of enterprises is considered as a dummy variable, and a 0–1 assignment is made based on the interim or periodic reports of listed companies to manually collect whether the enterprises have conducted digital transformation in the current year. Second, the amount of technology investment is applied to measure the digital transformation of enterprises [45]. Third, the ratio of digital intangible assets to the percentage of total intangible assets is applied to measure the digital transformation of the firm [46]. In particular, when the intangible assets disclosed in the notes to the financial statements of listed companies contain keywords such as “software”, “network”, and “digital”, which are equal to the keywords of related digital transformation, they are considered as digital intangible assets to the total intangible assets. The digital intangible assets are considered as digital intangible asset pairs and summed up; finally, the ratio of total digital intangible assets to total intangible assets is used to measure the digital transformation of enterprises. Fourthly, the digital transformation indicators of enterprises are constructed by the text analysis method [47,48]. The Python crawler function was used to gather and organize the annual reports of all A-share listed enterprises in Shanghai and Shenzhen exchanges, and all text contents were extracted through the Java PDFbox library and filtered in the text contents of annual reports according to the enterprise digital transformation characteristic words, and, finally, the word frequencies of key technology directions were summed up to measure the enterprise digital transformation by the total word frequencies obtained at last.

Through comparative analysis and combined with this paper’s research [43,47,48], the input of enterprise digital transformation is further taken into account based on whether the enterprise has carried out digital transformation, which can more accurately and objectively measure whether the enterprise has carried out substantial digital transformation. On the basis of the keywords related to digital transformation [46], such as “network” and “digital”, this paper further adds “digitalization”, “artificial intelligence”, “Artificial intelligence”, “intelligent logistics”, “intelligent storage”, “intelligent manufacturing”, “intelligent platform”, etc., the keywords related to the digital transformation of manufacturing enterprises [44], such as “intelligent platform”, are defined as “digital technology intangible assets”, and the proportion of intangible assets in this year is calculated to measure the digital transformation of enterprises. In addition, to further ensure the reliability of the screening results, a manual review was conducted. In the robustness check section, the variables of enterprise digital transformation were constructed based on the text analysis method.

3.2.3. Control Variables

Since corporate performance is influenced by a variety of factors, to exclude the influence of possible factors and concerning relevant studies [43,44], this paper selects variables from the corporate characteristic level and corporate governance level to be constrained. The variables selected are as follows: firm size (Size), nature of ownership (Soe), financial leverage (LEV), enterprise value (TobinQ), book-to-market ratio (MB), inventory turnover (Turnover), earnings volatility (Volatility), controlling shareholder ownership (Sharehold), equity checks and balances (Balance), the Dual, Board size, Perprofit, Intensity, Freecash, Cash, Turnover, and Mfree. The definitions of the above variables are detailed in

Table 1. Variable description table.

Variable Type	Variables	Variable Symbols	Variable Definition
Dependent Variables	Corporate Performance	ROA	The ratio of the company’s net income to total assets
Independent Variable	Digital Transformation	DT	The ratio of digital intangible assets to total intangible assets
Control Variables	Company Size	Asset	Total company assets are taken as the logarithm
	Nature of ownership	Soe	State-owned enterprises take 1 and non-state-owned enterprises take 0
	Financial leverage	LEV	The ratio of the company’s year-end liabilities to total assets
	Enterprise Value	TobinQ	The ratio of market capitalization to total assets
	Market capitalization book-to-bill ratio	MB	The ratio of company market capitalization to total assets
	Inventory turnover rate	Turnover	The ratio of the company’s operating costs to the ending balance of inventories
	Earnings Volatility	Volatility	Three-year volatility of the ratio of EBIT to total assets
	Controlling shareholder’s shareholding	Sharehold	The shareholding ratio of the company’s controlling shareholders
	Shareholding Checks and Balances	Balance	The ratio of the shareholding of the 2nd–5th largest shareholder to the shareholding of the first largest shareholder
	Two jobs in one	Dual	Take 1 if the chairman and general manager are the same people, otherwise, take 0
	Board Size	Board	Number of board directors plus one to take the logarithm
	Per capita profit generation	Perprofit	The ratio of net profit to the number of employees
	Employee Intensity	Intensity	The ratio of the number of employees at the end of the year to the operating revenue for the year
	Free cash flow	Freecash	EBITDA + depreciation and amortization-working capital additions-capital expenditures
	Net cash flow from operations	Cashflow	Net cash flows from operating activities
	Overhead rate	Mfree	The ratio of administrative expenses to operating income

.

3.3. Model Setting

To test Hypotheses H1a and H1b, i.e., the impact of digital transformation on the performance of manufacturing firms [44], the following regression models are set up.

$$RO{A}_{i,t}={\alpha }_0+{\alpha }_{1}D{T}_{i,t}+\gamma X_{i,t}+\sum YEAR+\sum IND+{\epsilon }_{i,t}$$

(1)

where $RO{A}_{i,t}$ is the firm performance of firm i in year t, $D{T}_{i,t}$ is the digital transformation of firm i in year t, $X_{i,t}$ is the set of all control variables, $YEAR$ is the year dummy, $IND$ is the industry dummy, and ${\epsilon }_{i,t}$ is the random error term. If the sign direction of the estimated coefficient ${\alpha }_1$ is significantly positive, it indicates that digital transformation will significantly improve the performance of manufacturing firms, and Hypothesis H1a holds. This paper predicts that the estimated coefficients of $RO{A}_{i,t}$ are significantly positive in the baseline regression.

4. Results

4.1. Descriptive Statistical Analysis

In this paper, descriptive statistics were conducted for the main variables and the results are shown in

Table 2. Descriptive statistics of variables.

Variables	Obs	Mean	Std. Dev.	Min	p25	p50	p75	Max
ROA	5569	0.0426	0.104	−0.556	0.0261	0.0532	0.0845	0.247
DT	5569	0.180	0.252	0	0.00734	0.0637	0.239	1.000
Asset	5569	21.66	1.115	19.49	20.85	21.52	22.31	25.01
Soe	5569	0.246	0.431	−0.333	0	0	0.221	1.492
Cashflow	5569	2.800 × 108	7.770 × 108	−7.090 × 108	6.964 × 106	7.750 × 107	2.370 × 108	5.430 × 109
LEV	5569	0.362	0.214	0.0397	0.196	0.332	0.492	1.236
MB	5569	0.318	0.159	−0.0471	0.202	0.298	0.418	0.744
Mfree	5569	0.133	0.115	0.0158	0.0630	0.102	0.161	0.737
Turnover	5569	110.4	613.9	−171.3	2.274	3.842	6.894	5065
Volatility	5569	0.0511	0.0764	0.00127	0.0113	0.0239	0.0577	0.495
Sharehold	5569	33.63	14.33	7.500	22.46	31.45	43.28	70.77
Balance	5569	0.842	0.652	0.0367	0.330	0.675	1.188	2.983
Dual	5569	0.378	0.483	−0.0563	0	0	1	1
Board	5569	2.088	0.213	1.386	1.946	2.197	2.197	2.833
Perprofit	5569	36,750	378,378	−2.53 × 106	19,031	55,785	118,075	1.018 × 106
Intensity	5569	1.855	1.389	0.113	0.888	1.509	2.447	8.157
Freecash	5569	−3.980 × 107	1.370 × 109	−5.950 × 109	−3.000 × 108	5.360 × 107	2.540 × 108	6.240 × 109
TobinQ	5569	2.495	1.587	0.837	1.514	2.007	2.892	10.17

. It can be seen that the mean value of corporate performance (ROA) is 0.0426, the minimum value is −0.556, the maximum value is 0.247, and the standard deviation is 0.104, which is a large difference, indicating that there is a large difference in the level of corporate performance among different manufacturing companies. Digital transformation (DT) has a mean value of 0.180, a minimum value of 0, a maximum value of 1, and a standard deviation of 0.252, with a large difference page, indicating that there is a large gap between the digital transformation of different manufacturing companies. In addition, the results of descriptive statistics of other variables are within a reasonable range and the influence of empirical results of outliers can be excluded.

4.2. Analysis of Multiple Regression Results

Hausman’s test suggests that the firm fixed-effect model should be adopted. The test results of model (1) are presented in

Table 3. Baseline regression results.

Variables	(1)	(2)	(3)	(4)	(5)
	ROA	ROA	ROA	ROA	ROA
DT	4.7076 ***	5.9388 ***	10.9207 ***	10.8432 ***	11.8057 ***
	(3.184)	(3.901)	(4.477)	(4.465)	(5.261)
Asset		−0.0146 ***	0.0329 ***	0.0449 ***	0.0203 ***
		(−3.730)	(5.436)	(7.073)	(3.408)
Soe		−0.0235	0.0077	0.0024	−0.0049
		(−1.511)	(0.361)	(0.110)	(−0.249)
Cashflow		0.0000	0.0000	0.0000	0.0000
		(0.792)	(0.748)	(0.559)	(0.521)
LEV			−0.5119 ***	−0.5096 ***	−0.3751 ***
			(−33.734)	(−33.757)	(−24.949)
MB			0.0969 ***	0.0706 **	−0.0174
			(3.263)	(2.337)	(−0.576)
Mfree			−0.0510 **	−0.0431 *	−0.0457 **
			(−2.197)	(−1.860)	(−2.121)
Turnover			−0.0000	−0.0000	−0.0000
			(−0.419)	(−0.408)	(−0.893)
Volatility			1.8486 ***	1.8483 ***	1.8159 ***
			(496.958)	(499.923)	(497.814)
Sharehold				0.0031 ***	0.0021 ***
				(5.935)	(4.445)
Balance				−0.0185	−0.0083
				(−1.631)	(−0.794)
Dual				0.0058	0.0050
				(0.486)	(0.459)
Board				0.0102 ***	0.0087 **
				(2.621)	(2.450)
Perprofit					0.0000 ***
					(29.175)
Intensity					−0.0046 ***
					(−2.992)
Freecash					0.0000 ***
					(2.907)
TobinQ					−0.0151 ***
					(−9.163)
Constant	1.0290	1.3451	−0.6259 ***	−1.0510 ***	−0.4553 ***
	(1.038)	(1.390)	(−4.788)	(−7.090)	(−3.236)
Controls	NO	YES	YES	YES	YES
YEAR/IND	YES	YES	YES	YES	YES
Observations	5569	5569	5569	5569	5569
Number of companies	911	911	911	911	911
R-squared	0.409	0.412	0.559	0.583	0.579

Notes: * p < 0.1, ** p < 0.05, *** p < 0.01.

. The regression results in column (1) show that before adding the control variables, the regression coefficient of digital transformation (DT) is 4.7076 and the t-value is 3.184, which is significantly positive at the 1% level. The estimated coefficients of the regression results are more significant after adding the control variables. Additionally, from column (5), it can be seen that after adding the control variables, the regression coefficient of digital transformation (DT) is 11.8057 with a t-value of 5.261, which is significantly positive and more significant at the 1% level, indicating that for every 1 unit increase in digital transformation, the performance of manufacturing firms increases by 1180.57%. The above results indicate that digital transformation helps to improve the performance of manufacturing firms and supports the content of Hypothesis H1a. In addition, the regression results for the control variables are also largely consistent with existing studies and are not repeated here.

4.3. Treatment of Endogeneity Issues

4.3.1. Adopting Dynamic Panel Data Econometric Model

Considering that the performance of manufacturing enterprises may have a high degree of autocorrelation, the performance of current enterprises may be affected by the previous value and show the characteristics of inertia. However, the lag term of enterprise performance is not controlled in the static panel regression model and this omission of variables will lead to an endogeneity problem. It should not be overlooked that business risk and individual firm characteristic variables (e.g., firm asset size, financial leverage, etc.) may also be causally linked and endogenous. To this end, with reference to existing studies [37], this paper introduces the first-order lagged term of the explanatory variable firm performance in the econometric model (1) and establishes a dynamic panel data econometric model for robustness testing. The details are as follows

$$RO{A}_{i,t}={\alpha }_0+{\alpha }_{1}RO{A}_{i,t-1}+{\alpha }_{2}D{T}_{i,t}+\gamma X_{i,t}+\sum YEAR+\sum IND+{\epsilon }_{i,t}$$

(2)

Due to the presence of lagged terms in the explanatory variables, the generalized moment estimation method (GMM) is used in this paper to overcome the endogeneity problem in the model (1). To overcome the drawbacks of both differential GMM (DIF-GMM) and systematic GMM (SYS-GMM) estimation methods in the dynamic panel data econometric model, this paper uses both the differential GMM method and systematic GMM method for regression.

Columns (1) to (4) in

Table 4. Results of the endogeneity as question treatment test.

Variables	DIF-GMM		SYS-GMM		IV-2SLS
	(1)	(2)	(3)	(4)	(5)	(6)
	ROA	ROA	ROA	ROA	ROA	ROA
DT	4.6261 ***	4.4969 ***	4.2664 ***	4.5858 ***	2.4850 **	2.4037 ***
	(15.104)	(2.650)	(10.064)	(5.684)	(2.143)	(2.978)
Asset	−0.0371 ***	−0.0356 ***	−0.2223 ***	−0.2248	−0.0275 ***	−0.0291 ***
	(−23.986)	(−3.668)	(−16.431)	(−1.324)	(−7.723)	(−3.753)
Cashflow	0.0000	0.0000	0.0000 ***	0.0000	0.0000	0.0000
	(0.454)	(0.414)	(9.253)	(1.137)	(0.425)	(0.296)
LEV	0.0970 ***	0.0856	0.7017 ***	0.7833	0.0701 ***	0.0631
	(9.806)	(0.808)	(10.111)	(0.837)	(7.933)	(0.898)
MB	0.1898 ***	0.1625 ***	−0.6223 ***	−0.5513	0.1018 ***	0.0801 *
	(27.698)	(2.833)	(−10.392)	(−0.758)	(5.741)	(1.915)
Mfree	−0.1529 ***	−0.1488 ***	−1.1034 ***	−1.1293	−0.1048 ***	−0.1016 ***
	(−20.438)	(−3.494)	(−15.214)	(−1.266)	(−8.107)	(−3.287)
Turnover	−0.0000	−0.0000	−0.0000	0.0000	0.0000	0.0000
	(−0.774)	(−0.935)	(−0.034)	(0.052)	(0.450)	(1.021)
Volatility	−1.1074 ***	−1.0888 ***	−0.1357 **	−0.1579	−0.9202 ***	−0.9519 ***
	(−57.749)	(−3.469)	(−2.147)	(−0.244)	(−44.747)	(−3.437)
Perprofit	0.0000 ***	0.0000 ***	0.0000 ***	0.0000 **	0.0000 ***	0.0000 ***
	(23.204)	(2.684)	(20.624)	(2.543)	(30.901)	(3.600)
Intensity	−0.0033 ***	−0.0029	−0.0398 ***	−0.0387 ***	−0.0014 *	−0.0015
	(−9.491)	(−1.514)	(−13.587)	(−3.329)	(−1.727)	(−1.099)
Freecash	0.0000	0.0000	−0.0000 ***	−0.0000	0.0000	0.0000 **
	(1.111)	(0.955)	(−7.370)	(−0.784)	(0.884)	(1.981)
TobinQ	0.0078 ***	0.0068 **	−0.0140 ***	−0.0144	0.0057 ***	0.0059 **
	(13.302)	(2.442)	(−10.877)	(−0.602)	(4.037)	(2.464)
Soe		−0.0012		0.0812	0.0156	0.0098
		(−0.057)		(0.615)	(1.335)	(0.702)
Sharehold		0.0003		0.0005	0.0002	0.0004
		(0.695)		(0.198)	(0.788)	(0.753)
Balance		0.0172 *		−0.0605	0.0190 ***	0.0167 **
		(1.760)		(−0.932)	(3.240)	(2.295)
Dual		0.0198 **		0.0251	0.0217 ***	0.0181 **
		(2.340)		(0.429)	(3.710)	(2.398)
lnBoard		0.0099		−0.1806	0.0013	0.0053
		(0.385)		(−1.068)	(0.083)	(0.280)
L.ROA	0.0216 ***	0.0281 ***	0.2027 ***	0.1962 ***
	(3.174)	(3.235)	(3.872)	(3.384)
Observations	3657	3655	3657	3655	4374	4374
Number of companies	643	643	643	643	655	655
R-squared					0.589	0.588
Centered-R2					0.1663	0.1705
AR(1)—p-value	0.000	0.000	0.000	0.000
AR(2)—p-value	0.573	0.592	0.586	0.475
Hansen—p-value	0.115	0.174	0.192	0.159
Anderson canon. corr. LM					0.000	0.000
Statistical p-value
Cragg−Donald Wald F					313.77	313.77
Statistical quantities					<16.38>	<16.38>

Note: L denotes lagged one period; *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively. The null hypothesis for the AR(1) and AR(2) tests is that there is no first-order and second-order autocorrelation in the model residual terms, respectively. Since heteroskedasticity exists in the model data results, the Sargan test is valid only under homoskedasticity and is invalid under heteroskedasticity, so the results of the Hansen test are reported in this paper, and the null hypothesis of the Hansen test is that the instrumental variables are valid. The identifiable test for instrumental variables was used with the Anderson canon. corr. LM statistic and the null hypothesis is that the instrumental variables are not correlated with endogenous variables and there is under-identification. Weak instrumental variables were tested using the Cragg−Donald Wald F statistic with Stock-Yogo’s critical value in pointed brackets, with the null hypothesis that the instrumental variables are weakly instrumental.

show the dynamic panel model regression results. Among them, columns (1) to (2) are the test results of differential moment estimation, and columns (3) to (4) are the test results of systematic moment estimation. The p-values of the AR(1) test are less than 0.1, which rejects the original hypothesis and indicates the existence of first-order autocorrelation in the residuals; the p-values of the AR(2) test are greater than 0.1, which indicates that there is no second-order autocorrelation in the residuals; and the p-values of the Hansen test are greater than 0.1, which indicates that the selection of instrumental variables is reasonable. The above test results verify the reasonableness of the model set. The regression coefficients of the test results show that the regression coefficients of the lagged terms of corporate performance (ROA) are all significantly positive at the 1% level, indicating that corporate performance has inertia characteristics. The estimated coefficient of the core explanatory variable digital transformation (DT) is still significantly positive at the 1% level, indicating that digital transformation will improve the performance of manufacturing firms and that the findings of this paper are reliable.

4.3.2. Adopting Instrumental Variables Approach

To overcome the model endogeneity problem more effectively, this paper uses external instrumental variables to test the endogeneity of the benchmark regression model. In this paper, the lagged period of numerical transformation (L.DT) is used as an instrumental variable to perform the endogeneity test [44]. Columns (5) and (6) in Table 4 show the regression results of the first and second stages of the instrumental variables, respectively. The regression results in column (6) show that the regression coefficient of digital transformation (DT) is 2.4037, which is significantly positive at the 1% level after controlling for the endogeneity of the test using instrumental variables. Further, the p-values of the under-identification test Anderson canon. corr. LM statistics are less than 0.1, rejecting the original hypothesis that L.DT and DT do not want to be statistically related. The statistical value of the weak instrumental variable test Cragg−Donald Wald F is 313.77, which is greater than the corresponding Stock−Yogo critical value of 16.38 under 10% bias, which means that the original hypothesis of the weak instrumental variable is rejected. Thus, this result indicates that the results of the main test remain unchanged after the two-stage regression analysis using instrumental variables to control for the endogeneity problem.

4.4. Robustness Tests

To further verify the reliability of the above results, this paper also conducted robustness tests in terms of replacing the explanatory variables and replacing the core explanatory variables, respectively. The test results are shown in

Table 5. Robustness test results.

Variables	(1)	(2)	(3)	(4)
	ROE	ROE	ROA	ROA
DT	0.0174 ***	4.8586 ***
	(3.150)	(5.679)
DT1			0.7107 ***	1.7770 ***
			(2.740)	(5.573)
Asset		−0.0788 *		−0.0276 ***
		(−1.714)		(−4.091)
Soe		−0.2364 *		0.0140
		(−1.927)		(1.073)
Cashflow		0.0000		0.0000
		(0.633)		(0.668)
LEV		0.4441 ***		0.0268
		(3.500)		(0.412)
MB		0.6258 **		0.1170 **
		(2.027)		(2.309)
Mfree		−0.8078 ***		−0.0752 ***
		(−3.753)		(−3.291)
Turnover		−0.0000		−0.0000
		(−0.385)		(−0.291)
Volatility		−0.1006 ***		−0.7835 ***
		(−3.307)		(−3.282)
Sharehold		0.0041		0.0008 **
		(1.199)		(2.217)
Balance		−0.0499		0.0240 ***
		(−0.656)		(3.050)
Dual		0.0435		0.0169 ***
		(0.459)		(2.820)
Board		0.0110		0.0016
		(0.404)		(0.857)
Perprofit		0.0000 ***		0.0000 ***
		(4.729)		(3.823)
Intensity		−0.0011		−0.0030 *
		(−0.076)		(−1.910)
Freecash		−0.0000		0.0000
		(−0.376)		(1.035)
TobinQ		0.0284		0.0100 **
		(1.642)		(2.411)
Constant	−0.0122	1.2075 ***	0.1946 ***	0.7847 ***
	(−0.214)	(3.157)	(7.146)	(5.046)
Controls	NO	YES	NO	YES
YEAR/IND	YES	YES	YES	YES
Observations	5569	5569	5569	5569
Number of companies	911	911	911	911
R-squared	0.583	0.607	0.652	0.545

Notes: * p < 0.1, ** p < 0.05, *** p < 0.01.

.

In the previous analysis, this paper used the return on assets (ROA) to measure the performance of manufacturing firms, and here this paper uses the return on net assets (ROE) to measure the performance of firms [44]. As can be seen from columns (1) and (2) in

Table 6. Heterogeneity test results.

Variables	(1)	(2)	(3)	(4)
	State-Owned-Enterprises	Non-State-Owned Enterprises	High Marketability	Low Market Degree
DT	4.4143 ***	0.1256	9.7151 ***	−0.9435
	(4.994)	(0.953)	(3.975)	(−0.960)
Asset	−0.0003	0.0223 ***	−0.0153 ***	0.0140 **
	(−0.142)	(2.742)	(−5.673)	(2.190)
Soe			0.0293 *	−0.0085
			(1.844)	(−0.430)
Cashflow	−0.0000	0.0000 ***	0.0000	0.0000
	(−0.390)	(3.069)	(1.303)	(0.937)
LEV	−0.1264 ***	−0.4832 ***	0.0801 ***	0.0922 ***
	(−9.142)	(−26.107)	(6.733)	(5.300)
MB	−0.0189	−0.0329	0.1381 ***	0.0779 **
	(−1.198)	(−0.871)	(6.612)	(2.560)
Mfree	−0.1547 ***	0.0094	−0.0700 ***	−0.0638 ***
	(−8.516)	(0.382)	(−3.253)	(−3.469)
Turnover	0.0000	−0.0000	0.0000 ***	−0.0000
	(0.759)	(−0.709)	(3.355)	(−0.943)
Volatility	−0.1794 ***	1.8415 ***	−1.0866 ***	1.7056 ***
	(−6.781)	(412.486)	(−42.428)	(405.767)
Sharehold	0.0004 *	0.0028 ***	0.0011 ***	0.0021 ***
	(1.670)	(4.437)	(6.664)	(3.682)
Balance	0.0025	−0.0113	0.0151 ***	−0.0035
	(0.424)	(−0.765)	(4.594)	(−0.282)
Dual	0.0068	−0.0039	0.0072	0.0150
	(1.172)	(−0.284)	(1.639)	(1.227)
Board	0.0019	0.0136 ***	−0.0003	0.0066 *
	(1.310)	(2.755)	(−0.222)	(1.776)
Perprofit	0.0000 ***	0.0000 ***	0.0000 ***	0.0000 ***
	(13.761)	(23.793)	(24.617)	(31.099)
Intensity	−0.0008	−0.0086 ***	−0.0010	0.0003
	(−1.584)	(−3.780)	(−0.597)	(0.270)
Freecash	0.0000	0.0000 ***	−0.0000	0.0000 *
	(0.747)	(5.769)	(−1.421)	(1.926)
TobinQ	0.0008	−0.0183 ***	0.0181 ***	−0.0117 ***
	(0.658)	(−9.695)	(9.393)	(−8.664)
Constant	0.0867	−0.5138 ***	0.2733 ***	−0.5539 ***
	(1.590)	(−2.657)	(4.516)	(−3.675)
Controls	YES	YES	YES	YES
YEAR/IND	YES	YES	YES	YES
Observations	1308	4034	3226	2308
Number of companies	195	747	593	507
R-squared	0.602	0.587	0.596	0.573

Notes: * p < 0.1, ** p < 0.05, *** p < 0.01.

, the regression coefficients are still significantly positive at the 1% level after replacing the explanatory variables, regardless of whether control variables are included or not, which again verifies the reliability of the findings.

In addition, referring to the studies of Wu et al. [48], Ni and Liu [47], and Yuan et al. [49], a machine learning-based text analysis method was used to construct corporate digital transformation indicators to measure the degree of Measuring Digital Transformation in Manufacturing Companies ([48]). Specifically, by first utilizing the semantic representation of national policies related to the digital economy, and combined with Python word separation processing and manual recognition, a lexicon of enterprise digital transformation was constructed from the application layer and technical layer; secondly, all the annual reports of A-share listed enterprises in Shanghai and Shenzhen exchanges were collected and organized with the help of the Python crawler function, and all the text contents were extracted through the Java PDFbox library and according to the enterprise digital transformation feature words in the annual report text content screening; finally, with the help of Python software for word root identification counting, while thoroughly screening manually to ensure the accuracy of the results, the word frequency of key technology directions are summed up to the final total word frequency obtained to measure the digital transformation of enterprises (DT1). Columns (3) and (4) in Table 6 show the regression results after replacing the core explanatory variables with DT1. It can be seen that the regression coefficients for digital transformation (DT1) are all significantly positive at the 1% level, indicating that digital transformation enhances the performance of manufacturing firms, again demonstrating the robustness of the paper’s findings.

5. Further Analysis

5.1. Heterogeneity Analysis

To examine the impact of digital transformation on the performance of manufacturing enterprises, this paper groups them according to the nature of property rights and the degree of marketization, respectively. First, this paper divides the sample enterprises into state-owned enterprises and non-state-owned enterprises according to the nature of property rights and then performs group regressions. The regression results are shown in columns (1) and (2) in Table 6. It can be seen that, for state-owned enterprises, the regression coefficient is 4.4133, which is significantly positive at the 1% level, while the regression coefficient of the non-state-owned enterprise group is not significant, indicating that the impact of digital transformation on the performance of manufacturing enterprises of state-owned nature is more significant which verifies the content of Hypothesis H2. Second, this paper uses the Fan Gang Marketability Index [50] to measure the degree of marketability (https://cmi.ssap.com.cn/, accessed on 8 June 2022) and divides the sample into a high marketability group and a low marketability group based on the median of the sample. The test results are shown in columns (3) and (4) in Table 6. It can be seen that the regression coefficient is 9.7151 in the higher marketization group, which is significantly positive at the 1% level, which means that digital transformation has a more significant impact on the performance of manufacturing firms in the higher marketization areas and verifies the content of Hypothesis H3.

5.2. Mechanism Test

According to the previous theoretical analysis, digital transformation can improve the performance of manufacturing companies through the channels of low-cost and innovation empowerment. To verify the existence of the above channel, regarding existing studies [51], the following mediating effect model is set up in this paper.

$$RO{A}_{i,t}={\alpha }_0+{\alpha }_{1}D{T}_{i,t}+\gamma X_{i,t}+\sum YEAR+\sum IND+{\epsilon }_{i,t}$$

(3)

$$Mediato{r}_{i,t}={\vartheta }_0+{\vartheta }_{1}D{T}_{i,t}+\gamma X_{i,t}+\sum YEAR+\sum IND+{\epsilon }_{i,t}$$

(4)

$$RO{A}_{i,t}={\theta }_0+{\theta }_{1}D{T}_{i,t}+{\theta }_{2}Mediato{r}_{i,t}+\gamma X_{i,t}+\sum YEAR+\sum IND+{\epsilon }_{i,t}$$

(5)

where ${\mathrm{Mediator}}_{\mathrm{i},\mathrm{t}}$ is the mediating variable, which represents the cost margin (CPM) and innovation investment (R&D) variables, ${\alpha }_1$ reflects the total effect of bank competition on the business risk of the entity, ${\theta }_1$ indicates the direct effect of bank competition on the business risk of the entity, and the magnitude of the intermediation effect is measured by ${\vartheta }_1\times {\theta }_2={\alpha }_1-{\theta }_1$. According to the mediation effect test procedure, if the coefficients ${\alpha }_1, {\vartheta }_1,$ and ${\theta }_2$ are significant and the coefficient ${\theta }_1$ becomes smaller or decreases in significance compared to ${\alpha }_1$, it indicates the existence of a mediation effect.

Table 7. Results of intermediate effect test.

Variables	(1)	(2)	(3)	(4)
	CPM	ROA	R&D	ROA
DT	−10.6742 ***	0.3021 ***	7.9536 ***	0.3921 ***
	(−5.158)	(3.153)	(3.609)	(6.108)
CPM		−1.0777 ***
		(−5.882)
R&D				1.4351 ***
				(7.946)
Asset	0.0181 ***	0.0005	0.0180 ***	0.0005
	(3.324)	(1.347)	(3.302)	(1.302)
Soe	−0.0018	−0.0014	−0.0009	−0.0013
	(−0.101)	(−1.317)	(−0.053)	(−1.233)
Cashflow	0.0000	0.0000 *	0.0000	0.0000 *
	(0.312)	(1.902)	(0.340)	(1.924)
LEV	−0.3668 ***	0.0264 ***	−0.3669 ***	0.0263 ***
	(−26.658)	(27.342)	(−26.654)	(27.366)
MB	0.0156	−0.0340 ***	0.0156	−0.0341 ***
	(0.561)	(−17.280)	(0.558)	(−17.354)
Mfree	−0.0376 *	−0.0038 ***	−0.0376 *	−0.0039 ***
	(−1.890)	(−2.708)	(−1.887)	(−2.780)
Turnover	−0.0000	−0.0000	−0.0000	−0.0000
	(−0.939)	(−0.263)	(−0.937)	(−0.277)
Volatility	1.5401 ***	0.1548 ***	1.5402 ***	0.1549 ***
	(462.677)	(104.804)	(462.435)	(105.088)
Sharehold	0.0019 ***	0.0000 *	0.0018 ***	0.0000 *
	(4.339)	(1.790)	(4.315)	(1.720)
Balance	−0.0066	−0.0013 **	−0.0066	−0.0013 **
	(−0.697)	(−2.160)	(−0.697)	(−2.156)
Dual	0.0041	−0.0002	0.0039	−0.0003
	(0.410)	(−0.366)	(0.392)	(−0.466)
Board	0.0079 **	0.0002	0.0584 **	0.0004
	(2.439)	(0.858)	(2.204)	(0.212)
Perprofit	0.0000 ***	−0.0000 ***	0.0000 ***	−0.0000 ***
	(29.830)	(−5.268)	(29.824)	(−5.261)
Intensity	−0.0042 ***	0.0004 ***	−0.0041 ***	0.0004 ***
	(−2.960)	(3.949)	(−2.950)	(4.067)
Freecash	0.0000 ***	−0.0000 *	0.0000 ***	−0.0000 *
	(2.974)	(−1.735)	(2.981)	(−1.729)
TobinQ	−0.0116 ***	−0.0024 ***	−0.0117 ***	−0.0024 ***
	(−7.617)	(−22.401)	(−7.635)	(−22.444)
Constant	−0.4135 ***	−0.0038	−0.4675 ***	−0.0025
	(−3.228)	(−0.457)	(−3.423)	(−0.284)
Controls	YES	YES	YES	YES
YEAR/IND	YES	YES	YES	YES
Observations	5569	5569	5569	5569
Number of companies	911	911	911	911
R-squared	0.602	0.715	0.596	0.651

Notes: * p < 0.1, ** p < 0.05, *** p < 0.01.

shows the results of the mechanical test. Columns (1)–(2) are the results of the test for low-cost empowerment. The core explanatory variable in column (1) is corporate cost margin (CPM), and it can be seen that the regression coefficient of digital transformation (DT) is −10.6742, which is significantly negative at the 1% level, indicating that digital transformation reduces the operating costs of manufacturing firms. From column (2), it can be seen that digital transformation has a significant positive effect on firm performance after adding the mediating factor cost margin (CPM), and the coefficient ${\theta }_1$ becomes smaller and more significant compared to ${\alpha }_1$. According to the principle of mediating effect, it can be seen that low sunk cost empowerment plays a part in mediating effect in the process of digital transformation on the performance of manufacturing firms, which verifies Hypothesis H4. Similarly, column (3)–column (4) shows the test results of innovation input. From column (3), it can be seen that digital transformation has a significant positive effect on innovation investment after adding the mediator innovation investment (R&D), indicating that digital transformation stimulates firms’ innovation investment initiative. From column (4), it can be seen that after adding the mediating factor of innovation investment (R&D), digital transformation has a significant negative effect on manufacturing firm performance and the coefficient ${\theta }_1$ becomes smaller and more significant compared to ${\alpha }_1$. According to the principle of mediating effect, it is known that innovation investment plays a part in mediating the effect in the process of digital transformation on the manufacturing firm performance, which verifies Hypothesis H5.

In summary, it can be seen that digital transformation has improved the performance of manufacturing firms by reducing operating costs and increasing innovation investment, which also indicates that low-cost empowerment and innovation empowerment are important mechanisms of digital transformation affecting the performance of manufacturing enterprises.

To further ensure the significance of the mediating effect, this paper uses the coefficient product test to test the significance of the mediating effect [52]. The absolute values of the statistics Z are 4.96 and 5.62, which are greater than the critical value of 0.97 at the 5% significance level, and the original hypothesis is rejected. Research hypotheses H4 and H5 are further verified.

6. Conclusions

Based on the background of digital transformation and the practice status of digital transformation in manufacturing enterprises in China, this paper systematically analyses the intrinsic mechanism of digital transformation affecting the performance of manufacturing enterprises and empirically tests the theoretical hypothesis using an unbalanced panel of A stock market manufacturing communication listed companies in China from 2007 to 2020. The findings show that (1) digital transformation improves manufacturing firm performance. (2) The findings still hold after using dynamic panel regression econometric models, the instrumental variables approach to consider potential endogeneity issues, and a series of robustness tests. (3) The results of the heterogeneity analysis show that the positive effect of digital transformation on the performance of manufacturing firms in state-owned enterprises and more market-oriented regions is more pronounced compared to other firms. (4) The effect channel test based on the mediating effect model finds that there is a partial mediating effect of low-cost empowerment and increased innovation investment between digital transformation and the performance of manufacturing firms, and there is a transmission channel from digital transformation to improving the performance of manufacturing firms in the process of digital transformation by exerting influence on the performance of manufacturing firms, that is, from digital transformation to low-cost empowerment/innovation empowerment and, finally, to manufacturing enterprise performance improvement.

Nevertheless, this paper has the following limitations. First, measuring the digital transformation of companies based on digital investment suffers from serious data deficiency and measurement bias, which can be supplemented by using digital transformation keyword text mining based on combining machine learning and text analysis in the future; second, this study is based on manufacturing data in China, and to make the study more extensive, it can be conducted based on samples from more countries and industries in the future with more empirical tests; finally, the geographical distribution of these companies, etc., are not discussed in this study based on global data, which can be explored in depth in combination with spatial econometric models in the future.