This study aims to explore the interactive mechanisms and transmission paths between digital transformation and corporate performance in Chinese manufacturing enterprises.
This study conducts theoretical research through a literature review and proposes research hypotheses. In the empirical study, statistical data from Chinese A-share manufacturing enterprises from 2014 to 2021 are selected as the sample, and a multiple regression model is constructed to test the hypotheses. Through descriptive analysis, regression analysis and correlation analysis, the relationship between enterprise digital transformation and enterprise performance is determined as well as the mediating role of enterprise innovation. Finally, robustness analysis is used to validate the research results.
Research has found that digital transformation can significantly improve corporate performance: innovation plays an intermediary role between digital transformation and corporate performance. It is worth noting that the mediating effect of innovation output has a lag period of about two years. In addition, property rights significantly moderate the effects of digital transformation. The Szfix regression coefficient of non-SOEs is much lower than that of SOEs, and digital transformation has a more significant impact on the performance of SOEs.
Compared with existing literature, the innovation of this study lies in the introduction of the concept of “enterprise innovation,” which not only expanded the scope of research on the impact of digital transformation on enterprise performance but also provides strong empirical evidence for the cyclical characteristics of innovation outcome conversion. In addition, this study also considers different types of enterprise property rights.
1. Introduction
The rapid development of the global digital economy has become a new engine for economic growth in countries and regions around the world (Xu and Zhang, 2020). On the one hand, under the wave of rapid development of the digital economy, China has undergone comprehensive and profound changes in all aspects of production, life, and ecology (Qi and Xiao, 2020). On the other hand, the vigorous development of the digital economy has also had a positive impact on the quality, efficiency, and momentum of China's economic development. As the basic unit of the macroeconomic system, the digital transformation and upgrading of enterprises is strategically significant in promoting the overall development of the digital economy and stimulating new drivers of economic growth.
A fundamental question is how enterprises can be motivated to embark on a digital transformation process. The central issue is whether enterprises' digitalization can contribute to enhancing their performance. The “IT paradox” endorsed by Hajli et al. (2015) and Mikalef and Pateli (2017) points out that digitisation does not necessarily improve corporate performance. Still, Qi and Cai (2020) point out that the mechanism of digitisation's impact on enterprise performance is multiple, and that the positive and negative effects of various impacts may be why performance is not significant, given the issues above. This research investigates the dynamic interplay between digital transformation and firm performance, examining how integrating digital technologies can drive enhanced operational efficiency and innovation, and ultimately boost overall business outcomes. Using data from listed companies, the study delves into whether digitalisation can promote firm performance and, if so, the mechanism behind this effect.
2. Research hypotheses
2.1 Impact of digital transformation on firm performance
Digital transformation has become an inevitable trend in business development. Scholars typically explore the relationship between enterprise digital (information technology) transformation and business performance from two main perspectives. One perspective is the information value-added theory, which is one of the productivity paradoxes. With the development of the digital economy, an increasing number of empirical studies support the “information value-added theory.” This perspective argues that digital transformation enhances business performance by improving information processing, utilisation, and overall value creation capabilities. Ji et al. (2025) propose that by adopting advanced information technology systems, enterprises can optimise operational processes, improve production efficiency, and gain competitive advantages, ultimately achieving significant improvements in overall performance. During digital transformation, enterprises treat data elements as core production means and objects of labour, establishing a data-driven production and operation system. Compared with traditional enterprises, data circulates more quickly and efficiently within the enterprise, and this new model can significantly improve decision-making efficiency, optimise resource allocation efficiency, and enhance market response flexibility. On the one hand, improved internal information transmission efficiency can help technical personnel improve their technical skills and optimise production methods. On the other hand, enterprises can further reduce costs and improve efficiency by continuously optimising production technology. From a comprehensive perspective, the application of digital technology and the additional costs required by enterprises exhibit a complementary relationship; that is, the stronger an enterprise's ability to adopt digital technology, the lower the additional costs required, and the higher the enterprise's performance level. Based on this, the thesis posits the following hypothesis:
Digital transformation by manufacturing firms can positively contribute to the improvement of firm performance.
2.2 The mediating role of corporate innovation on digital transformation and firm performance
Digital transformation primarily promotes enterprise innovation in the following three areas: First, as pointed out by Chen et al. (2019), the digital economy can incentivise manufacturing enterprises to engage in a series of innovations, including but not limited to product innovation, system innovation, and model innovation. Second, the efficiency, reliability, and high productivity of digital technologies can assist manufacturing enterprises in achieving innovative development. Third, the digital economic environment has a significant impact on consumer demand, prompting companies to increase innovation investment and drive new product development to meet consumer needs. Therefore, there is a significant intrinsic relationship between digital economic development and corporate innovation, which is undisputed. On the other hand, the relationship between corporate innovation and corporate performance has undergone significant changes due to digital economic development. With the expansion of the digital economy, innovation has become a key factor driving corporate performance growth (Fan, 2020). Enterprise digital transformation can significantly enhance innovation capabilities and indirectly improve overall enterprise performance through an intermediary role. Therefore, improvements in enterprise performance may stem from the promotional effect of digital transformation on innovation output. Drawing upon these findings, the thesis posits the following hypothesis:
Firm innovation output serves as a critical mediator linking digital transformation to firm performance.
Firm innovation investment functions as a significant mediator in the relationship between digital transformation and firm performance.
2.3 The impact of the nature of business ownership on performance
The emergence of digital technology has brought many benefits to manufacturing enterprises. However, due to differences in the nature of property rights, its impact will manifest itself in different ways. This article will analyse the impact on SOEs and non-SOEs.
On the one hand, SOEs have more prominent advantages than non-SOEs, which stem from two core strengths. Fang and Yi (2023) point out that, first, SOEs have unique advantages in innovation; second, SOEs shoulder specific responsibilities and missions. First, SOEs' innovation is mainly oriented towards long-term innovation with social value. This is similar to the concept of digital transformation. SOEs have a longer-term vision for digital innovation, rather than improving innovation efficiency in the short term. Si-ying and Bai (2023) point out that government policy support makes it easier for SOEs to innovate in high-end industries that are difficult for non-SOEs to enter. Second, SOEs undertake more important scientific and technological innovation tasks in areas such as national strategic security and work closely with the government, enabling them to obtain resources more quickly and achieve breakthroughs in high-precision technology fields. In contrast, non-SOEs are relatively lacking in relevant resources and government support and protection, so they tend to take a more conservative approach to digital investment. In addition, Huan et al. (2024) point out that compared with non-SOEs, SOEs have generated a significant “absorption effect.” This phenomenon is reflected in the higher density of R&D personnel with higher education backgrounds in SOEs and their significantly larger investment in R&D personnel than private enterprises. Existing data show that, compared with non-SOEs, the level of digital investment in SOEs has shown a sustained growth trend. Although the growth rate is relatively stable, the growth rate is still higher than that of non-SOEs, indicating that SOEs have more resources, greater advantages and greater responsibilities under China's digital strategy, and are therefore committed to digital infrastructure construction. This observation gives rise to the following hypothesis:
Digital transformation improves firm performance more significantly in SOEs than in non-SOEs.
In summary, this study conducts theoretical research through a literature review and proposes research hypotheses (see Table 1), aiming to explore the interactive mechanisms and transmission pathways between digital transformation, innovation, and performance in Chinese manufacturing enterprises. The research framework is illustrated in Figure 1.
Summary of research hypotheses
| Assumed number | Assumed content | Literature support |
|---|---|---|
| H1 | Digital transformation by manufacturing firms can positively contribute to the improvement of firm performance | Ji et al. (2025) |
| H2a | Firm innovation output serves as a critical mediator linking digital transformation to firm performance | Chen et al. (2019), Fan (2020) |
| H2b | Firm innovation investment functions as a significant mediator in the relationship between digital transformation and firm performance | |
| H3 | Digital transformation improves firm performance more significantly in SOEs than in non-SOEs | Fang and Yi (2023), Si-ying and Bai (2023), Huan et al. (2024) |
The figure shows three circular diagrams connected with blue arrows. The top circle has the text in the inner ring reading, “Firm innovation.” The upper half of the outer ring contains the text “Innovation Investment” with the number “01” on the center left, and the lower half of the outer ring contains the text “Innovation Output” with the number “02” on the center right. The bottom-left circle has the text in the inner ring reading, “Digital transformation by manufacturing firms.” The upper half of the outer ring contains the text “Non-S O Es” with the number “01” on the center left, and the lower half of the outer ring contains the text “S O Es” with the number “02” on the center right. The bottom-right circle has the text in the inner ring reading, “Firm performance.” The upper half of the outer ring contains the text “Non-S O Es” with the number “01” on the center left, and the lower half of the outer ring contains the text “SOEs” with the number “02” on the center right A rightward arrow points from the bottom-left circle to the bottom-right circle. A right arrow turns downward and points from the top circle downward to the bottom-right circle. An upward arrow turns rightward and points from the bottom-left circle to the top circle.Mechanism flow chart. Source: The authors
The figure shows three circular diagrams connected with blue arrows. The top circle has the text in the inner ring reading, “Firm innovation.” The upper half of the outer ring contains the text “Innovation Investment” with the number “01” on the center left, and the lower half of the outer ring contains the text “Innovation Output” with the number “02” on the center right. The bottom-left circle has the text in the inner ring reading, “Digital transformation by manufacturing firms.” The upper half of the outer ring contains the text “Non-S O Es” with the number “01” on the center left, and the lower half of the outer ring contains the text “S O Es” with the number “02” on the center right. The bottom-right circle has the text in the inner ring reading, “Firm performance.” The upper half of the outer ring contains the text “Non-S O Es” with the number “01” on the center left, and the lower half of the outer ring contains the text “SOEs” with the number “02” on the center right A rightward arrow points from the bottom-left circle to the bottom-right circle. A right arrow turns downward and points from the top circle downward to the bottom-right circle. An upward arrow turns rightward and points from the bottom-left circle to the top circle.Mechanism flow chart. Source: The authors
3. Research design
3.1 Sample selection and data sources
This study takes listed companies in the manufacturing industry in China's A-share market as the sample body, and conducts empirical analyses based on panel data for 2014–2021. To ensure the reliability of the research findings, strict screening criteria were applied during data processing. Only samples from companies listed continuously for eight consecutive years were included, ensuring consistency and robustness in the analysis.
Companies designated with ST status were excluded from the study to prevent potential distortions in the performance data;
Companies that have been delisted due to unusual financial circumstances have been excluded;
Firms listed for a shorter period, after 2014, were excluded from the study.
In this study, Winsor's deflation method is adopted to deal with the extreme values of continuous variables at the upper and lower 1%, effectively controlling the influence of extreme values on the research results. After screening, 7,210 balanced panel data observations from 1,030 listed companies are finally obtained. The pertinent financial indicators and patent data employed in this study were derived from the CSMAR database, CNRDS database and Stata 16.0 statistical software was used for subsequent empirical analyses.
3.2 Research variables
Table 2 provides a comprehensive overview of the key variables used in this study, including their definitions, measurement methods, and data sources. Firstly, regarding the dependent variables, enterprise performance is selected as the explanatory variable in this study. In contrast, following the study by Guo (2025), this research measures corporate financial performance using return on equity (Roe). This indicator can comprehensively reflect a firm's ability to generate profits using net assets, directly embodying the efficiency of capital appreciation and operating results, and is compatible with the capital operation characteristics of manufacturing enterprises. Meanwhile, as an indicator widely adopted in academic research, its data is derived from standardized databases, which can ensure the objectivity of measurement and the comparability of results, thus meeting the needs of this study. Second, regarding the independent variables, this study focuses on enterprise digital transformation. Given the absence of a unified standard for measuring digital transformation, this study employs a text analysis approach informed by Yuan et al. (2021). Data are extracted from the annual reports of A-share listed companies using Python, and a statistical model is built to quantify digital transformation through word frequency analysis with the Jieba thesaurus, whose keywords are primarily based on the research by Wu et al. (2021) and Ren and Guo (2017), with the key term database presented in Figure 2.The process for developing the digital transformation index unfolds in several key stages: initially, annual reports from Shanghai and Shenzhen A-share manufacturing firms (2014–2021) are gathered using Python-based web scraping techniques. These documents are then converted into TXT files through Xunjie PDF Converter for further analysis; second, extract all text content of the annual reports for text analysis and supplement the keywords with reference to relevant studies to build a keyword lexicon; third, perform word segmentation on the text of the enterprise annual reports based on the constructed keyword lexicon and count the disclosure frequency of keywords from the two dimensions of digital technology and digital business model; fourth, compute the corporate DX index: log(annual report DX keyword count+1). Finally, in terms of mediating variables, this study selected enterprise innovation as a mediating variable. Enterprise innovation is operationalized using two indicators: innovation input (the ratio of R&D expenditure to operating revenue) and innovation output (the number of patent applications). Finally, to mitigate the effects of potential confounders, this study controls for firm-level characteristics—including firm size, total asset growth rate, gearing ratio, major shareholder ownership, board independence, operating cash flow, cost of goods sold ratio, along with year and industry factors, as supported by existing literature.
Definition of variables
| Variable category | Variable name | Variable symbol | Calculation method |
|---|---|---|---|
| Explanatory variable | Corporate financial performance | Roe | Net profit/0. 5 (Net assets at the beginning of the year + Net assets at the end of the year) |
| Explanatory variable | Digital Transformation | Dig | Frequency obtained by text mining |
| Intermediary variable | Enterprise investment in innovation | Rd | R&D investment as a percentage of operating revenue |
| Enterprise innovation outputs | Innovation | Natural logarithm of the total number of patents granted to enterprises+ 1 | |
| Control variable | Enterprise size | Size | Natural logarithm of total assets |
| Total asset growth rate | Growth | Difference between assets at the end of the period and assets at the beginning of the period divided by assets at the beginning of the period | |
| Gearing | Leverage | Total liabilities divided by total assets at the end of the year | |
| Shareholding ratio of major shareholders | Shareholder | Number of shares held by the largest shareholder divided by total share capital | |
| Board independence | Indirecter | Number of independent directors divided by number of board members | |
| Operating cash flow | NC | Natural logarithm of net operating cash flow of the enterprise | |
| Cost of goods sold ratio | Cos | Business operating costs divided by operating revenue | |
| Vintages | Year | Virtual variable | |
| Sector | lnd | By industry classification standards issued by the Securities and Futures Commission in 2012 |
The diagram has two central boxes. The box on the left is labeled “Digital technology applications,” and the box on the right is labeled “Digital business models.” From the box “Digital technology applications,” 11 arrows point outward to the following labels, from the top in aa clockwise manner: “Office Automation System (O A),” “Enterprise Resource Planning (S A P),” “Industrial Robots,” “Product Lifecycle Management (P L M),” “Intelligent Manufacturing Innovation Platform (U 9),” “Electronic Asset Security System (E A S),” “Financial Information Management System (N C),” “Industrial Internet,” “Robotic Process Automation (R P A),” “Information Technology (I T),” and “Oracle Database Management System.” From the box “Digital business models,” 11 arrows point outward to the following labels, from the top in aa clockwise manner: “online,” “B 2 B,” “C 2 B,” “O 2 O,” “C 2 C,” “e-commerce,” “ecological coordination,” “(SaaS),” “online retail,” “offline,” and “B 2 C.” The two central boxes, “Digital technology applications” and “Digital business models,” are connected by a line.Dimensions and keywords of digital transformation of manufacturing enterprises, Source: Wu et al. (2021) and Ren and Guo (2017)
The diagram has two central boxes. The box on the left is labeled “Digital technology applications,” and the box on the right is labeled “Digital business models.” From the box “Digital technology applications,” 11 arrows point outward to the following labels, from the top in aa clockwise manner: “Office Automation System (O A),” “Enterprise Resource Planning (S A P),” “Industrial Robots,” “Product Lifecycle Management (P L M),” “Intelligent Manufacturing Innovation Platform (U 9),” “Electronic Asset Security System (E A S),” “Financial Information Management System (N C),” “Industrial Internet,” “Robotic Process Automation (R P A),” “Information Technology (I T),” and “Oracle Database Management System.” From the box “Digital business models,” 11 arrows point outward to the following labels, from the top in aa clockwise manner: “online,” “B 2 B,” “C 2 B,” “O 2 O,” “C 2 C,” “e-commerce,” “ecological coordination,” “(SaaS),” “online retail,” “offline,” and “B 2 C.” The two central boxes, “Digital technology applications” and “Digital business models,” are connected by a line.Dimensions and keywords of digital transformation of manufacturing enterprises, Source: Wu et al. (2021) and Ren and Guo (2017)
3.3 Modelling
This study constructs the subsequent econometric model, drawing upon the work of Wen and Ye (2014), to assess the influence of digital transformation on business performance and elucidate its underlying mechanisms.
To test Hypothesis 1, which examines the effect of digital transformation on business performance, a multiple regression model was developed as outlined below:
The model is adapted from Qi and Xiao (2020),which examines the multifaceted effects of digitalisation on manufacturing enterprise performance and its underlying mechanisms.
To evaluate the mediating role of corporate innovation in linking digital transformation to corporate performance (Hypothesis 2), we construct the following multiple regression model:
The above model is from Wen and Ye (2014), who discusses the impact of mediating effects.
To investigate how variations in enterprise property rights influence the relationship between digital transformation and firm performance, the following multiple regression models are established to test Hypothesis 3. Model (4) pertains to SOEs, while Model (5) applies to non-SOEs.
3.4 Descriptive statistical analyses
Table 3 summarizes the descriptive statistics derived from 7,210 observations of A-share listed companies. Enterprise performance ranges from −0.526 to 0.349, indicating notable variability among manufacturing firms. However, a standard deviation 0.117 suggests that most performance values are clustered toward the lower end. Values in the digital transformation dimension range from 0 to 3.363, with a standard deviation of 0.543, indicating moderate firm dispersion. The data suggest that only 33% of enterprises have initiated digital transformation, while more than half remain inactive. Additionally, among those that have begun the process, there are considerable differences in the extent of transformation. The relatively low overall level of digital transformation suggests that the industry remains in the nascent stage of digital development, which to some extent constrains enterprises' productivity enhancement, innovation synergy, and market responsiveness. Regarding innovation investment (Rd), the mean is 4.550, with values ranging from 0 to 20.01. This wide dispersion indicates substantial heterogeneity among firms and suggests significant potential for many enterprises to enhance their innovation investment. The standard deviation of 3.446 suggests that, despite some variation, the overall level of innovation investment across enterprises remains relatively low. Regarding innovation, the average is 2.097 with a standard deviation of 2.250, ranging from 0 to 7.192. This indicates substantial variability in patent applications among companies.
Descriptive statistics
| Variant | Sample size | Average value | Median | (Statistics) standard deviation | Minimum value | Maximum values |
|---|---|---|---|---|---|---|
| Roe | 7,161 | 0.060 | 0.061 | 0.117 | −0.526 | 0.349 |
| Szfix | 7,210 | 0.330 | 0.146 | 0.543 | 0 | 3.363 |
| L2.Innovation | 7,210 | 2.097 | 1.609 | 2.250 | 0 | 7.192 |
| Rd | 7,210 | 4.550 | 3.870 | 3.446 | 0 | 20.01 |
| Soe | 7,210 | 0.689 | 1 | 0.463 | 0 | 1 |
| Size | 7,210 | 22.27 | 22.12 | 1.118 | 20.23 | 25.56 |
| Growth | 7,170 | 0.144 | 0.0800 | 0.283 | −0.299 | 1.725 |
| Lev | 7,210 | 0.401 | 0.394 | 0.186 | 0.0570 | 0.865 |
| Shareholder | 7,210 | 0.321 | 0.301 | 0.134 | 0.0900 | 0.687 |
| Indirecter | 7,209 | 0.375 | 0.333 | 0.0530 | 0.333 | 0.571 |
| Nc | 7,210 | 5.718e+08 | 1.610e+08 | 1.402e+09 | 1.052e+09 | 9.522e+09 |
| Cos | 7,210 | 0.711 | 0.747 | 0.167 | 0.188 | 0.982 |
Note(s): In order to ensure the sample size and the “eight-year continuity” data, only companies with serious vacancies were excluded from the data processing process, so above the sample values in the table are slightly different
3.5 Correlation analysis
Table 4 reveals that the absolute correlation coefficients among the key variables are all below 0.6, suggesting that multicollinearity is not a significant concern. Additionally, the variance inflation factor (VIF) test results presented in Figure 3 further confirm that the VIF values for all variables are below 2, a series of statistical indicators that fully prove that the research model does not have a significant multicollinearity problem. The correlation coefficients suggest a statistically significant positive association between digital transformation and enhancements in enterprise performance. This finding not only supports Hypothesis H1 but also offers robust empirical evidence for the subsequent analysis. Implementing a digital transformation strategy by enterprises positively contributes to their performance improvement, laying a solid empirical foundation for future in-depth studies.
Correlation analysis
| Variables | (1) | (2) | (3) | (4) | (5) | (6) | (7) | (8) | (9) | (10) | (11) |
|---|---|---|---|---|---|---|---|---|---|---|---|
| (1) Roe | 1 | ||||||||||
| (2) Szfix | 0.0154 | 1 | |||||||||
| (3)L2.Innovation | 0.0567 | −0.0289 | 1 | ||||||||
| (4)Rd | −0.0326 | 0.275 | 0.0577 | 1 | |||||||
| (5) Soe | 0.0424 | 0.0412 | 0.0359 | −0.173 | 1 | ||||||
| (6) Size | 0.0692 | 0.0953 | 0.126 | −0.234 | 0.348 | 1 | |||||
| (7) Growth | 0.125 | 0.0612 | 0.00010 | 0.0334 | −0.0851 | 0.0187 | 1 | ||||
| (8) Leverage | −0.149 | −0.0982 | 0.0240 | −0.294 | 0.320 | 0.564 | −0.0244 | 1 | |||
| (9)Shareholder | 0.0607 | −0.0989 | 0.0639 | −0.162 | 0.236 | 0.237 | −0.0487 | 0.122 | 1 | ||
| (10)Indirecter | −0.0343 | 0.0195 | 0.0149 | 0.0676 | −0.0187 | 0.0318 | 0.00270 | −0.0003 | 0.0481 | 1 | |
| (11) Nc | 0.148 | −0.0578 | 0.0567 | −0.0586 | 0.0274 | 0.205 | −0.0338 | −0.0240 | 0.0770 | −0.0057 | 1 |
| (12) Cos | −0.259 | −0.163 | 0.00560 | −0.369 | 0.246 | 0.221 | −0.0987 | 0.473 | 0.0917 | −0.0239 | −0.138 |
The horizontal axis of the vertical bar chart lists 12 variables: “Szfix,” “Rd,” “L 2. Innovation,” “Lev,” “Size,” “Cos,” “Soe,” “Share,” “Nc,” “Growth,” “Indirecter,” and “Mean V I F.” Each variable has a single vertical bar. A legend at the bottom indicates that the bars represent “V I F.” The values shown above each bar are: Szfix: 1.14. Rd: 1.48. L 2. Innovation: 1.03. Lev: 1.88. Size: 1.75. Cos: 1.57. Soe: 1.24. Share: 1.11. Nc: 1.11. Growth: 1.06. Indirecter: 1.01. Mean VIF: 1.31.Variance inflation factor test, Source: Authors’ calculations using CSMAR data, CNRDS data and textual mining of annual reports (SSE/SZSE A-share manufacturers)
The horizontal axis of the vertical bar chart lists 12 variables: “Szfix,” “Rd,” “L 2. Innovation,” “Lev,” “Size,” “Cos,” “Soe,” “Share,” “Nc,” “Growth,” “Indirecter,” and “Mean V I F.” Each variable has a single vertical bar. A legend at the bottom indicates that the bars represent “V I F.” The values shown above each bar are: Szfix: 1.14. Rd: 1.48. L 2. Innovation: 1.03. Lev: 1.88. Size: 1.75. Cos: 1.57. Soe: 1.24. Share: 1.11. Nc: 1.11. Growth: 1.06. Indirecter: 1.01. Mean VIF: 1.31.Variance inflation factor test, Source: Authors’ calculations using CSMAR data, CNRDS data and textual mining of annual reports (SSE/SZSE A-share manufacturers)
3.6 Benchmark regression
The fixed-effects model regression results of Model 1 (see Table 5) show that for manufacturing enterprises, the coefficient of the digital transformation variable (Szfix) is 0.008 with a t-value of 2.27, which is significant at the 5% level. This result indicates that each unit increase in the degree of digital transformation of manufacturing enterprises leads to an average increase of 0.008 units in their return on equity (Roe), thus verifying Research Hypothesis 1. In terms of economic significance, considering that the standard deviation of digital transformation (Szfix) is 0.543 as shown in the descriptive statistics, it can be inferred that when the degree of digital transformation increases by one standard deviation, corporate performance (Roe) will rise by 0.008 × 0.543 = 0.0043. This is equivalent to 7.2% of the mean value of Roe (0.060) in the sample, indicating that the promoting effect of digital transformation on corporate performance has practical economic significance. These results suggest that digital transformation plays a significant role in enhancing the performance of manufacturing enterprises, thereby effectively driving the economic development of the manufacturing sector. Moreover, the regression outcomes for the control variables are in line with established research, further substantiating that digital transformation contributes positively to the performance of manufacturing enterprises. Therefore, with the arrival of the digital wave in enterprises, enterprises can seize digital technology platforms to create many opportunities for innovation and inject new vitality into the enterprise. In addition, through the promotion of digital transformation, enterprise resources can be effectively integrated. This dynamic improvement in capabilities enables enterprises to quickly coordinate their production and operation activities, thereby reducing enterprise costs, improving enterprise efficiency, and enabling enterprises to grow rapidly.
Baseline regression
| Variant | Roe | T-value | |
|---|---|---|---|
| Szfix | 0.0008** | 2.27 | |
| Size | 0.013*** | 3.42 | |
| Growth | 0.092*** | 21.62 | |
| Lev | −0.214*** | −15.61 | |
| Shareholder | 0.082*** | 3.65 | |
| Indirecter | −0.096*** | −2.83 | |
| Nc | 0.000*** | 9.25 | |
| Cos | −0.482*** | −25.17 | |
| Constant | 0.203** | 2.44 | |
| Company FE | control | ||
| Year FE | control | ||
| R-squared | 0.142 |
Note(s): ***p < 0.01,**p < 0.05,*p < 0.1, T-values in parentheses
3.7 Robustness tests
To enhance the robustness of the empirical analysis, this study employs the variable substitution method and the method of lagging core explanatory variables for robustness tests, as detailed below:
Variable Substitution Method
Drawing on the research of Guo (2025), this study uses return on assets (Roa) as an alternative indicator for robustness testing. The reasons for selecting this indicator are as follows: first, Roa and the benchmark indicator Roe are both core profitability metrics, which respectively reflect the efficiency of a firm's utilization of total assets and the profitability of net assets, enabling complementary verification from different dimensions; second, Roa has standardized data sources and is widely used in academic research. If the empirical results remain significant after substitution, it can further confirm the robustness of the conclusions. Column (1) of Table 6 reports the regression results after replacing the explained variable Roe with Roa. The regression coefficient of digital transformation is 0.003, which is significant at the 10% statistical level, indicating that enterprise digital transformation has a significant positive impact on return on assets. The result is robust, and Hypothesis H1 is verified.
Robustness tests
| Variables | Roa | L.Roe | L2.Roe |
|---|---|---|---|
| Szfix | 0.003* | 0.0031*** | 0.0030*** |
| (1.82) | (4.5131) | (4.1200) | |
| Controlled variable | control | control | control |
| Constant | −0.443*** | −0.4403*** | 0.159*** |
| (−26.41) | (−24.90) | (0.036) | |
| Company FE | control | control | control |
| Year FE | control | control | control |
| R-squared | 0.202 | 0.311 | 0.306 |
Note(s): ***p < 0.01,**p < 0.05,*p < 0.1. T-values in parentheses
Method of Lagging Core Explanatory Variables
Following the research approach of Li et al. (2022), considering the possible lagged effect of digital investment on enterprise performance and aiming to more accurately depict the dynamic relationship between them, this study conducts regression analyses by lagging the explained variable (ROE) by 1 and 2 periods respectively. Column (2) of Table 6 shows that the regression coefficient of enterprise digital transformation on enterprise performance lagged by 1 period is 0.0031, which passes the significance test at the 1% level; in Column (3), its regression coefficient on enterprise performance lagged by 2 periods is 0.0030, also significant at the 1% level. The above results indicate that even after incorporating lagged terms, enterprise digital transformation still has a significant positive effect on improving the performance of manufacturing enterprises, further strengthening the robustness of the previous conclusions. In addition, the regression coefficients of the 1-period and 2-period lags are not only significant but also close in value, implying that the impact of digital transformation on enterprise performance may be long-term—its effects not only manifest in the short term but also persist for a certain period. This long-term effect may arise because measures involved in digital transformation, such as technological upgrading, process optimization, and data-driven management, require a certain amount of time to gradually release their effectiveness in various business links of the enterprise.
3.8 Analysis of mechanisms
Column (1) of Table 7 presents the test results without including the mediating variable. The coefficient α1 of digital transformation (Szfix) is significantly positive at the 5% level, indicating that digital transformation has a direct impact on the explained variable. An analysis of Columns (2) and (3) of Table 7 shows that the impact of digital transformation on the current period and lagged one-period of corporate innovation is not significant; however, Column (3) reveals that the regression coefficient ω11 of digital transformation (Szfix) is significantly positive at the 5% level, and the coefficient ω12 of lagged two-period innovation (Innovation) is significantly positive at the 1% level. The above results satisfy the three conditions for mediating effect testing: α1, ω11, and ω12 are all significant, which indicates that there exists a partial mediating effect of innovation in the relationship between digital transformation and the explained variable when innovation is lagged by two periods; moreover, η4 and ω12 are significant, so the Sobel test is not required. In conclusion, Hypothesis H2a is supported by empirical evidence.
Tests for mediating effects of innovation output
| Variables | (1) | (2) | (3) | (4) | (5) |
|---|---|---|---|---|---|
| Roe | Innovation | Innovation_t−1 | Innovation_t−2 | Roe | |
| Szfix | 0.008** | 0.029 | 0.023 | 0.290*** | 0.008** |
| (2.27) | (0.29) | (0.24) | (5.07) | (2.25) | |
| Innovation | 0.003*** | ||||
| (4.51) | |||||
| Controlled variable | control | control | control | control | control |
| Constant | 0.203** | 4.124* | 2.81 | −5.603*** | 0.195** |
| (2.44) | (1.77) | (1.25) | (−5.77) | −2.36 | |
| CompanyFE | control | control | control | control | control |
| YearFE | control | control | control | control | control |
| R-squared | 0.265 | 0.006 | 0.005 | 0.029 | 0.267 |
Note(s): ***p < 0.01,**p < 0.05,*p < 0.1. T-values in parentheses
Table 7 presents the mediating effect test results of innovation output. In the first step, digital transformation's regression coefficient on corporate performance is 0.008 (significant at 5%). The second step shows its coefficients on current and one-period lagged innovation output are 0.029 and 0.023 (both insignificant), while that on two-period lagged innovation output is 0.290 (significant at 1%). In the third step, with both variables included, digital transformation's coefficient is 0.008 (significant at 5%) and innovation output's is 0.003 (significant at 1%). Results confirm innovation output partially mediates the relationship, with the mediating effect significant only with a two-period lag. Digital transformation directly boosts corporate performance and indirectly does so via increased two-period lagged innovation output, validating Hypothesis H2a.
Table 8 shows the mediating effect test results of innovation input. In the first step, digital transformation's regression coefficient on corporate performance is 0.008 (significant at 5%). The second step finds digital transformation's coefficient on innovation input is 0.242 (significant at 1%). In the third step, with both variables included, digital transformation's coefficient drops to 0.004 (significant at 10%), and innovation input's is 0.002 (significant at 1%).Results confirm innovation input partially mediates the relationship: digital transformation directly boosts corporate performance and indirectly does so via increased R&D input, validating Hypothesis H2b.
Mediation effect test for innovation inputs
| Variables | (1) | (2) | (3) |
|---|---|---|---|
| Roe | Rd | Roe | |
| Szfix | 0.008** | 0.242*** | 0.004** |
| (2.27) | (4.96) | (1.28) | |
| Rd | 0.002*** | ||
| (5.59) | |||
| Indirecter | −0.096*** | 0.391 | −0.109*** |
| (−2.83) | (0.77) | (−3.25) | |
| Shareholder | 0.082*** | 0.010*** | 0.001*** |
| (3.65) | (4.98) | (6.12) | |
| Growth | 0.092*** | 0.270*** | 0.033*** |
| (21.62) | (7.28) | (7.90) | |
| Lev | −0.214*** | 0.329*** | 0.026*** |
| (−15.61) | (3.33) | (4.07) | |
| Nc | 0.000*** | −0.679*** | −0.242*** |
| (9.25) | (−3.93) | (−6.62) | |
| Cos | −0.482*** | 0.000 | 0.000*** |
| (−25.17) | (0.54) | (3.60) | |
| Constant | 0.203** | −4.291*** | −0.459*** |
| (2.44) | (−5.23) | (−5.37) | |
| CompanyFE | control | control | control |
| YearFE | control | control | control |
| R-squared | 0.265 | 0.027 | 0.139 |
Note(s): ***p < 0.01, **p < 0.05, *p < 0.1 T-values in parentheses
3.9 Analysing the heterogeneity of business ownership
Based on the regression findings for the whole sample grouping in columns (1) and (2) in Table 9, the research sample is partitioned into two distinct groups, SOEs and non-SOEs, and models 4 and 5 are empirically tested, respectively. The research data for SOEs indicate that the regression coefficient for the Szfix variable is 0.027, accompanied by a T-value of 4.72, and is statistically significant at the 1% level. Additionally, the regression outcomes for the other control variables align with the conclusions of previous literature, further supporting the robustness of the findings. In contrast, the regression coefficient of Szfix for non-SOEs is only 0.003, significant at the 10% level—substantially lower than the 0.027 observed for SOEs, which is significant at the 1% level. This empirical evidence unequivocally corroborates Hypothesis 5, indicating that digital transformation exerts a significantly more pronounced positive influence on the performance of SOEs than non-SOEs.
Grouped regression analysis
| Variables | SOEs group | Non-SOEs group |
|---|---|---|
| Roe | Roe | |
| Szfix | 0.027*** | 0.003* |
| (4.72) | (1.65) | |
| Controlled variable | control | control |
| Constant | 0.469*** | 0.094 |
| (2.85) | (0.98) | |
| Company FE | control | control |
| Year FE | control | control |
| R-squared | 0.146 | 0.156 |
Note(s): ***p < 0.01,**p < 0.05,*p < 0.1 T-values in parentheses
Empirical regression analysis indicates that property rights' characteristics significantly influence the degree of digital transformation. Although digital transformation yields positive outcomes for SOEs and non-SOEs, its effect is markedly more pronounced within SOEs. Specifically, digital transformation exerts a more pronounced impact on the performance of SOEs. This phenomenon can be attributed to the unique advantages of SOEs regarding resource endowment, national strategic mission, and policy support. As the pillars of the national economy, SOEs bear the great responsibility of promoting the construction of “Digital China”, and their continuous investment in digital infrastructure and technological innovation gives them a competitive advantage over non-SOEs in high-end industrial innovation, thus taking the lead in the process of digital transformation.Furthermore, in terms of corporate governance structure, the institutional linkages between SOEs and government entities enable them to more efficiently access policy information and rapidly respond to national digitalization strategic initiatives.
4. Results and discussion
This study uses Chinese A-share listed manufacturing companies from 2014 to 2021 as its sample to systematically examine the impact mechanism and pathways of digital transformation on corporate performance. The findings reveal: First, digital transformation significantly enhances the performance of manufacturing companies. Second, corporate innovation plays a crucial mediating role between digital transformation and corporate performance. Specifically, digital transformation not only directly promotes performance improvement but also indirectly enhances performance through two pathways: increasing R&D investment (innovation input) and boosting patent output (innovation output). Notably, the mediating effect of innovation outputs exhibits a lag period of approximately two years, providing empirical evidence for the cyclical characteristics of innovation outcomes conversion. Third, the nature of property rights significantly moderates the effects of digital transformation. SOEs exhibit a notably stronger performance improvement effect from digital transformation compared to non-SOEs.
This study aligns with Qi and Cai (2020)'s perspective on the “multiplicity of digitalisation impact mechanisms,” indicating that when digitalisation is combined with process optimisation, it can overcome the “IT paradox” proposed by Hajli et al. (2015) and enhance corporate performance. Additionally, this study quantitatively validates Chen et al. (2019)'s concept of “digital economy-driven multidimensional innovation,” supplementing Fan (2020)'s theoretical framework through a path-specific analysis of innovation inputs and outputs. Unlike the conclusions of Fang and Yi (2023), which focus on the innovation advantages of SOEs, this study reveals that under the backdrop of digital transformation, the “resource diversion effect” (Huan et al., 2024) of SOEs further widens the performance gap between them and non-SOEs. Additionally, existing literature often limits its exploration of the mediating role of innovation to a single perspective. This study, however, examines the mediating effects from both innovation input and output dimensions, providing a more comprehensive understanding of the transmission mechanism between innovation, digital transformation, and corporate performance. Notably, existing research often overlooks the time lag inherent in the innovation conversion process. This study, however, employs dynamic analysis to find that innovation outputs require at least two years to exert a significant impact on corporate performance. This finding challenges the prevailing notion that innovation outcomes yield immediate results and poses a profound challenge to the currently dominant short-term performance evaluation paradigm.
5. Conclusion
This study makes a theoretical contribution by distinguishing the differentiated roles of innovation inputs and outputs, thereby deepening our understanding of the entire “input-output” process driven by digital transformation. Additionally, by introducing the perspective of property rights heterogeneity, it reveals the moderating role of institutional environments on digital transformation outcomes, providing a new analytical dimension for research on digital transformation in emerging economies.
This study has the following limitations: the sample scope is limited to A-share listed companies, which may affect the external validity of the conclusions; the measurement of digital transformation primarily relies on text analysis methods, and the characterization of the depth of digital technology application needs to be strengthened; the heterogeneous characteristics of manufacturing sub-industries have not been fully considered; meanwhile, regarding the time lag effect of innovation outcomes, although it has been found that innovation outputs require at least two years to exert a significant impact on corporate performance, the specific mechanism behind this time lag effect (such as differences in the time lag across industries and enterprise scales) has not been explored in depth.
Future research can be expanded in multiple aspects: first, expand the sample scope to non-listed companies to enhance the external validity of the conclusions; second, construct multi-dimensional digital transformation measurement indicators to strengthen the characterization of the depth of digital technology application; third, conduct comparative studies on industry segments to fully consider the heterogeneous characteristics of manufacturing sub-industries; fourth, expand the research perspective to test the role of different innovation measures (such as product launches and process improvements) in the relationship between digital transformation and corporate performance, so as to further enrich the understanding of the innovation mediating mechanism; fifth, deeply explore the role of the adoption of specific digital technologies such as artificial intelligence in improving corporate performance, and analyse the association and differences between such role and the overall effect of digital transformation, thereby providing more targeted theoretical guidance for the practice of corporate digital transformation.
Author contributions
All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by LiMing. Badrul Hisham Bin Kamaruddin reviewed and supervised the content of the manuscript and the research process. The first draft of the manuscript was written by LiMing and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.
