Assessing the Impact of Digital Transformation on Manufacturing Enterprises’ Performances: An Efficiency Perspective

Abstract

In recent years, the impacts of the new scientific and technological revolution on the industrial system and production modes have begun to emerge. Digital transformation is gradually being integrated into the production behaviors of manufacturing enterprises and has become a component of the micro-economy. We aim to find better methods for measuring digital transformation and to analyze its impact on both market performance and innovation performance within manufacturing enterprises. To achieve our goals, we employ an empirical study to examine the influence of digital transformation on market and innovation performance using panel data for Chinese listed manufacturing enterprises from 2012 to 2021. We emphasize digital transformation efficiency and affirm its role in relieving financing constraints. Our study shows that digital transformation effectively improves both the market and innovation performance of manufacturing enterprises. Moreover, it mitigates the financing constraint dilemma, resulting in performance enhancement. Heterogeneity analysis indicates that digital transformation has a more significant promotional effect on the market and innovation performance of large-scale and mature enterprises. Our research offers fresh perspectives for better understanding digital transformation, enriching the body of work on the impact of digital transformation in manufacturing enterprises and its underlying mechanisms.

1. Introduction

Digital technologies, such as AI, big data, the Internet of Things, etc., have emerged as the primary drivers of the latest technological revolution. Enterprises have begun to strategically and comprehensively apply digital technologies and engage in continuous digital transformation to obtain better performance and competitive advantages. Currently, scholars’ research on digital transformation primarily centers on its fundamental characteristics (Van Veldhoven & Vanthienen, 2022; Singh & Hess, 2020), precursors (Verhoef et al., 2021; Jose et al., 2024), outcomes (Fitzgerald et al., 2014; Kretschmer & Khashabi, 2020; Gao et al., 2025), and mechanisms for value creation (Frank et al., 2019; Matt et al., 2015; Wang et al., 2023). Viewed at the enterprise level, digital transformation is an endeavor that enables innovation and organizational change, aiming to enhance enterprise performance.

Digital transformation changes the processes and/or organizational structure of a manufacturing enterprise, and thus managers should focus on its efficiency to balance the input and output. Efficiency is usually understood as the output produced by consuming a certain amount of input, and this process requires not only the pursuit of higher economic efficiency, but also the rational allocation of human, capital, financial, and other resources (Camanho et al., 2024). The efficiency of enterprise digital transformation can comprehensively reflect the current transformation capability of the enterprise, indicating the utilization and allocation of resources in the transformation process. Managers who possess a deeper understanding of the resource combination of inputs and outputs in the transformation process can enhance their competitive advantage. This, in turn, stimulates enterprises to expedite the transformation process (Sirmon et al., 2011).

Academic and industrial discourse on digital transformation’s business impact remains inconclusive. Scholars debate if increased digital investments lead to a “digital paradox” or improved performance. Short-term benefits may exist, but long-term impacts are uncertain and it is unclear if digital transformation’s potential is overstated (Zhu et al., 2021). Additionally, it is argued that existing theoretical frameworks are inadequate at fully underpinning the ongoing depth of research in this area (Smith & Beretta, 2021). Chinese manufacturing’s digital transformation is just starting, and systematic research on it is scarce. There is no agreement on its impact on performance; therefore, how digital transformation affects performance and its variability among enterprises is still unclear.

In the process of digital transformation, enterprises need to adopt digital technology to redesign their workflows, which will generate more management investment in the short term (Trantopoulos et al., 2017). If digital transformation investment is not effectively utilized and does not match the degree of transformation, leading to a low efficiency, the overall performance may be unsatisfactory. Therefore, it is necessary to explore the impact of digital transformation on performance from the perspective of digital transformation efficiency.

In this paper, we focus on the efficiency of digital transformation, which reflects the ability of enterprises to utilize investment during the process of digital transformation. This measurement can better describe the degree of digital transformation while considering the costs incurred during the transformation process. As our research sample, we use China’s A-share listed manufacturing enterprises from 2012 to 2021, during which Chinese manufacturing enterprises carried out large-scale digital transformation and the impact was the most significant. We employ a DEA to assess the digital transformation efficiencies of the enterprises, empirically analyze the impact of digital transformation on enterprise performance, explore variations in impact across different contexts through a heterogeneity analysis, and examine the potential mechanism of financing constraints between digital transformation and enterprise performance.

We find that the efficiency of digital transformation has a significant promotion effect on the market and innovation performances of the enterprises, and this promotion effect is more obvious in large enterprises and mature enterprises. In addition, digital transformation can improve the enterprises’ performances by alleviating their financing constraints.

Our study contributes to the literature by (1) supplementing a digital transformation measurement method using a DEA to evaluate enterprise efficiency and (2) using the new measurement to empirically investigate the effect of digital transformation efficiency on Chinese manufacturing enterprise performance, complementing existing research on digital transformation outcomes at the enterprise level. Our findings provide insights into manufacturing enterprises’ digital transformations, providing more targeted theoretical guidance for enterprise practice.

The rest of the paper is organized as follows: Section 2 contains a literature review; Section 3 contains the theoretical analysis and research hypotheses; Section 4 outlines the research design; Section 5 presents the analysis of the empirical results; and Section 6 comprises the discussion and conclusions.

2. Related Works

2.1. Measurement of Digital Transformation

Currently, the primary issue requiring attention in the field is how to assess and measure digital transformation. This constitutes a complex and challenging task, with numerous methodologies employed in existing studies. Some scholars employ the adoption rate of enterprise technology and digital tools as a metric (Miuller et al., 2018). This approach has limitations, as it fails to encompass information regarding an enterprise’s strategy, processes, profitability, and innovation capabilities. Others utilize text analysis to quantitatively assess the degree of digital transformation (Fang & Liu, 2024). While this method provides relatively comprehensive information, there remains the potential for improvement in terms of precision.

2.2. Data Envelopment Analysis (DEA)

Data envelopment analysis (DEA) is a highly effective method for measuring efficiency, enabling organizations to assess how effectively they convert multiple inputs into multiple outputs. To achieve digital transformation, enterprises require corresponding shifts across all internal levels by examining cross-departmental inputs and outputs. This encompasses clear strategic direction and leadership (Korachi & Bounabat, 2020), the possession of resources and capabilities (Chanias et al., 2019), the transformation of business models (F. Li, 2020), and the adaptation of core operations (Rodríguez-Abitia & Bribiesca-Correa, 2021). DEA can encompass multifaceted information such as enterprises’ resource allocation, costs, and performance, striving to cover the aforementioned dimensions of digital transformation. This enables the measurement of digital transformation from an efficiency perspective.

2.3. The Impact of Digital Transformation on Enterprise Performance

Regarding the impact of digital transformation on enterprise performance, existing studies have achieved some results. On the one hand, enterprises implement digital transformation through the utilization of digital technology, which accelerates the pace of innovation and can effectively enhance enterprise performance. Enterprises can introduce new products and services more easily due to the homogeneity and re-programmability of digital technology (Yoo et al., 2011). New business models based on digitalization expand the user base and increase access to new customer information and market opportunities (Kraus et al., 2021; Usai et al., 2021). Moretti and Biancardi (2020) stated that the higher the quality of digital transformation, the higher the productivity of the organization and the greater the positive impact on the performance of the organization.

On the other hand, although digital transformation has the potential to increase business revenues, some enterprises have difficulty in taking the financial risk and may not obtain the expected results. Tortorella and Fettermann (2018) noted that investment in enterprise informatization might result in a decline in business performance, highlighting the existence of an “IT productivity paradox” phenomenon. Nolan’s model states that enterprises must invest heavily in IT early during an information transformation, which will not boost their performance immediately; instead, effective improvement will occur in the integration and data management stages (King & Kreamer, 1984). Between the early stages of digital transformation and effective performance improvement, enterprises face more financial risks, which are difficult for some enterprises to bear and may prevent them from achieving the expected performance (Nolan, 2012).

In addition, some empirical results have indicated that enterprises’ investments in digitalization yield little immediate results, with neither positive nor negative effects. For example, Usai et al. (2021) found that digital technologies have a minimal impact on enterprises’ innovation performance in the short term, but digital transformation mitigates enterprise info asymmetry and enhances market visibility. However, increased external attention intensifies pressure, which may lead to more conservative investment choices, deterring risky yet lucrative projects. Bayo-Moriones et al. (2013) explored how the adoption of information and communication technology (ICT) in SMEs positively correlates with organizational performance. However, the findings indicate that this correlation exhibits a delayed impact contingent upon the specific type of ICT.

To sum up, in current research, there remains no consensus among scholars regarding the impact of digital transformation on enterprise performance and further research is required.

3. Theoretical Analysis and Research Hypotheses

China’s manufacturing industry has expanded quickly. Using its cost advantages, it has taken on labor-intensive tasks for multinational enterprises. China’s manufacturing system is the world’s largest (L. Li, 2013), but it is unstable due to outdated practices and a lack of advanced technology. Big data and IoT are pushing the sector towards smarter operations, reducing China’s cost advantage (Zhong & Xu, 2017). Porter’s diamond theory shows that basic production factors are declining in importance, while advanced factors are crucial for a competitive advantage (Pol, 2020). To stay competitive, Chinese manufacturers need to adopt digital technologies to enhance production, R&D, and adaptability, promoting high-quality growth and market competitiveness (L. Li, 2018).

Digital technology provides key benefits such as data sharing, replication, and reuse, which enhance market operations and business growth. Referring to the definition of efficiency, we use “digital transformation efficiency” to measure the digital transformation of enterprises. Digital transformation is a process involving multiple inputs and outputs, in which enterprises need to make reasonable use and allocations of invested resources to achieve the highest possible output (Camanho et al., 2024). The efficiency of enterprise digital transformation can reflect the proportion between input and output in the process of digital transformation and reflect the utilization and allocation of resources by the enterprise, so as to comprehensively reflect the transformation ability and level of the enterprise (K. Chen & Guan, 2012). Digital transformation allows for iterative improvements, boosting productivity and enterprise market performance (Y. Chen & Wang, 2019). In this way, the market performance of the enterprise can be improved. Additionally, it facilitates efficient knowledge sharing and tech trend tracking and guides R&D efforts. Platforms utilizing open-source tech or standards can further enhance R&D and innovation performance (Jacobides et al., 2018). Digital transformation impacts enterprises by improving both market and innovation aspects.

Digital transformation can enhance enterprise market performance for two reasons. Firstly, digital transformation boosts market performance by enabling firms to harness unique resources (Nambisan et al., 2017). Helfat emphasizes the importance of reconfiguring, integrating, and transforming these resources for sustained competitiveness (Helfat, 2007). Efficiency management theory states that a higher enterprise transformation efficiency means easier access to valuable resources. Enterprises can also control digital and other resources precisely and more effectively, enhancing risk prediction and decision-making accuracy (Camanho et al., 2024). Secondly, digital technology is a key factor in transformation, shaping competitive advantages and boosting performance. It is a vital tool for enterprises to increase revenue (Wen et al., 2022; Demirkan et al., 2016). Efficient transformation with digital technology helps enterprises meet evolving market demands and customer needs, reducing time costs and enhancing customer satisfaction and loyalty through optimized products and services.

Consequently, enterprises can attain a competitive advantage in the marketplace.

Therefore, we propose the following hypothesis:

Hypothesis 1.

The digital transformation of Chinese manufacturing enterprises positively influences the enterprises’ market performance.

Digital transformation drives enterprise innovation and the need for change continues beyond initial success. Therefore, companies must develop an awareness of and ability for ongoing innovation (Sousa-Zomer et al., 2020). On the one hand, digital technology reallocates resources to generate business value, enabling enterprises to modify product structures and production processes, thus improving new product development efficiency (Nambisan, 2017). On the other hand, enterprises explore new distribution channels and enhance customer value through digital transformation. Cenamor states that new business models, such as subscription services, online marketplaces, and platform-based ecosystems, have increased companies’ ability to innovate and use digital technologies to transform operations and deliver value to customers (Cenamor et al., 2019).

Therefore, we propose the following hypothesis:

Hypothesis 2.

The digital transformation of Chinese manufacturing enterprises positively influences the enterprises’ innovation performance.

Digital transformation is a lengthy process that involves integrating digital technology into business operations and introducing costs and risks. Digital transformation has the potential to enhance the transparency of enterprise information (Luo, 2022). This transformation encourages engagement with financial institutions, facilitating financing activities and deeper collaboration. By leveraging advanced technology and credit systems, businesses can create secure processes, improve information disclosure, and lower cooperation costs.

Overcoming financing challenges provides secure funding, boosting market expansion and flexibility. This will help companies improve financing flexibility and strengthen information collection in the financial market, thus reducing information and transaction costs (C. Li et al., 2023). These improvements will enhance the financial and market performance of enterprises. Therefore, we propose the following hypothesis:

Hypothesis 3.

The digital transformation of Chinese manufacturing enterprises can improve the enterprises’ market performance by easing enterprise finance constraints.

Digital transformation reformulates an enterprise’s business and financing operations through digital technology integration, enhancing its technological capabilities and innovation dynamics. Financing institutions gain better insights into enterprises’ operations and profitability, allowing for improved credit assessments and financing options. Efficiently transformed enterprises can leverage external resources and use digital technology to identify innovative projects, injecting funds into R&D to enhance knowledge (Yu & Dou, 2020; Lange et al., 2020). Enterprises can then attract innovative enterprises to participate in collaborative innovation, thus improving performance at the innovation level.

Therefore, we propose the following hypothesis:

Hypothesis 4.

The digital transformation of Chinese manufacturing enterprises can improve the enterprises’ innovation performance by easing enterprise financial constraints.

4. Empirical Design

4.1. Sample and Data

Given the time required for data collection, organization, and analysis, there is an inherent lag in the research data. Moreover, external factors such as global pandemics and economic fluctuations may introduce anomalies or fluctuations in the data, thereby impacting the study’s accuracy.

We take Chinese A-share listed enterprises in the manufacturing industry as the research sample and select the years 2012–2021 as the time window. During that period, Chinese manufacturing enterprises undertook large-scale digital transformation initiatives, with the impact of such transformation being most pronounced and demonstrably significant during this timeframe, thereby serving as a highly representative example. We process the sample as follows: delete the enterprises with missing data about key variables; delete the enterprises with ST, *ST, and PT; and perform a two-sided shrinkage treatment at the 1% level on all the continuous variables to prevent the extreme values and outliers in the data from interfering with the empirical results. After the above processing steps, the balanced panel data which include 800 observations of 80 enterprises are finally obtained. The relevant patent data used in this paper was sourced from the CNRDS database; the data concerning digital transformation and other attributes of enterprises was sourced from the CSMAR database.

4.2. Variables

4.2.1. Dependent Variables

This study examines two dependent variables: market performance, measured using Tobin’s Q (TobinQ), and innovation performance (EFFI). Market performance is assessed by economic outcomes such as price, output, cost, and technological progress within a market structure, influenced by specific market behaviors (Bharadwaj et al., 1999). Tobin’s Q is calculated as the ratio of a firm’s market value to the replacement cost of its assets. For robustness, return on total assets (ROA) is also used as an indicator of market performance.

Innovation performance is assessed by output and efficiency. Output is typically gauged by patent data, with a focus on applications due to grant delays. Efficiency is the output per R&D dollar spent (Camanho et al., 2024). Our study employs invention patent applications and their ratio to R&D spending as innovation indicators. We use patent applications as a baseline measure of performance and innovation efficiency across robust tests.

4.2.2. Independent Variable

This study focuses on digital transformation efficiency (DT), a measure of how well an enterprise manages resources during its transformation. We utilize data envelopment analysis (DEA) to measure this variable. DEA is a non-parametric test method in which the organizations under evaluation are termed decision-making units (DMUs). DEA can measure how effectively each DMU converts multiple inputs into multiple outputs, and determine which units achieve optimal performance under given resource constraints. Its core principle involves assessing relative efficiency by comparing the ratio of inputs to outputs across units, focusing solely on the relationship between inputs and outputs without requiring prior assumptions about production or utility functions. By calculating efficiency scores for each unit, DEA identifies those performing more efficiently relative to others and pinpoints units capable of increasing output without additional inputs, thereby enhancing efficiency.

We use DEA to evaluate DT for two main reasons. Firstly, the DEA method can simultaneously consider multiple input and output indicators. Digital transformation efficiency encompasses information regarding an enterprise’s strategic management, digital technology infrastructure, business process optimization, organizational culture, and other aspects. The multiple input and output indicators within DEA can cover all facets of an enterprise, offering greater comprehensiveness and representativeness. Secondly, DEA constitutes a non-parametric testing method. It requires no assumptions regarding the distribution of an enterprise’s digital transformation data or financial information, nor does it necessitate assigning weights to input and output indicators, which lends it to a degree of objectivity. In light of the above, this paper employs DEA to measure the efficiency values of digital transformation.

In our experiment, we select the Banker–Charnes–Cooper model (BCC model) from the family of data envelopment models. This model serves as a method for evaluating the performance of relatively efficient decision-making units. The BCC model permits the existence of imperfect variability between inputs and outputs, enabling it to better accommodate data from real-world scenarios. The following presents a simplified derivation of the BCC model. Taking the efficiency index of the kth decision unit as the objective and subjecting it to the constraint of the overall efficiency of all decision units, the following model is derived:

$$E_k=Max{\displaystyle \frac{{{\displaystyle \sum }}_{r=1}^s{u}_r{Y}_{rk}}{{{\displaystyle \sum }}_{i=1}^m{v}_i{X}_{ik}}}$$

(1)

$$s.t.{\displaystyle \frac{{{\displaystyle \sum }}_{r=1}^s{u}_r{Y}_{rk}}{{{\displaystyle \sum }}_{i=1}^m{v}_i{X}_{rk}}}\le 1,u_r, v_i\ge \epsilon >o, j=1, \dots , n$$

$X_{ik}$ denotes the total input of the $k$th factor by the $i$th decision unit, $Y_{rk}$ denotes the total output of the $k$th product by the $r$th decision unit, while $v_i$ and $u_r$ denote the weighting coefficients for the $i$th type of input and the $r$th type of output, respectively.

Through Charnes–Cooper transformation, this can be converted into the following linear programming model:

$$E_k=Max{{\displaystyle \sum }}_{r=1}^s{u}_r{Y}_{rk}$$

(2)

$$s.t.{{\displaystyle \sum }}_{i=1}^m{v}_i{X}_{rk}=1,$$

$${{\displaystyle \sum }}_{r=1}^s{u}_r{Y}_{rk}-{{\displaystyle \sum }}_{i=1}^m{v}_i{X}_{rk}\le 0u_r, v_i\ge \epsilon >o, j=1, \dots , n$$

Model (1) is a fractional linear programming problem, which not only poses computational difficulties but also carries the risk of infinite solutions. It must be converted into a linear equation by setting the denominator to 1, yielding Model (2).

$$Min{h}_k={\theta }_k-\epsilon ({{\displaystyle \sum }}_{i=1}^m{s}_i^{-}+{{\displaystyle \sum }}_{r=1}^s{s}_r^{+})$$

(3)

$$s.t.{{\displaystyle \sum }}_{j=1}^n{\lambda }_j=1$$

$${{\displaystyle \sum }}_{j=1}^n{\lambda }_j{X}_{ij}-{\theta }_k{X}_{ik}+s_i^{-}=0$$

$${{\displaystyle \sum }}_{j=1}^n{\lambda }_j{X}_{ij}-s_r^{+}={\gamma }_{ik}, {\lambda }_j, s_i^{-}, s_r^{+}\ge 0$$

Converting the linear problem of Model (2) into its dual form by introducing slack variables $s_r^{+}$ and residual variables $s_i^{-}$ yields Model (3). The inequality constraints are transformed into equality constraints, thereby simplifying the model.

Initial input and output indicators were selected based on the literature, considering indicator comparability, hierarchy, and data accessibility (Yang et al., 2023; Kao et al., 2022). The selection aims to encompass the human resources, funding, and other resources needed for digital transformation, as well as strategic, process, and revenue information to ensure indicator completeness and model accuracy.

Selected indicators are shown in

Table 1. Primary input–output variable indicators.

Variable Indicators	Type	Explanation
Digital Technology	Input	Frequency of featured words of digital technology in annual reports.
Total amount of investment	Input	Total amount of project investment with digital keywords in the name.
Number of R&D personnel	Input	Total number of R&D personnel.
Number of R&D projects	Input	Total number of R&D projects.
R&D funding	Input	Total amount of R&D funding.
Number of infrastructures built	Input	Cumulative number of participations in the construction of national-level science and technology innovation bases as of that year.
Digital strategy looking forward	Output	Years when digitized keywords first appeared in the annual report.
Digital strategy continuity	Output	Total years digital keywords appear in annual report.
Digital breadth	Output	Number of digitized keyword types in each annual report.
Digital intensity	Output	Ratio of digitized keywords to total words in each year’s annual report.
Technology innovation	Output	Frequency of related feature words.
Process innovation	Output	Frequency of related feature words.
Business innovation	Output	Frequency of related feature words.
Digital innovation qualification	Output	Number receiving the same recognized project that year.
Scientific knowledge generation	Output	Number of science and technology papers published in journals.
The number of industry standards developed	Output	Number of participations in the formulation of industry standards in that year.

.

We conducted factor analysis on the DEA model’s input and output data due to potential non-linear relationships. The KMO and Bartlett’s tests, as shown in

Table 2. KMO and Bartlett’s test on input index.

KMO Sampling Suitability Quantity		0.734
Bartlett’s Test of Sphericity	chi-square	793.732
	free degree	21
	significance	0.000

and

Table 3. KMO and Bartlett’s test on output index.

KMO Sampling Suitability Quantity		0.704
Bartlett’s Test of Sphericity	chi-square	434.211
	free degree	55
	significance	0.000

, confirm good correlations among the indicators, making them suitable for factor analysis. Data was standardized during preprocessing to avoid errors from unit and range differences. Principal component analysis and Kaiser’s method identified the input and output indicators, as detailed in

Table 4. List of input and output indicators.

Input Indicators	Output Indicators
Digital Technology Total amount of investment Number of R&D personnel Number of R&D projects R&D funding Number of infrastructures built	Digital strategy looking forward Digital strategy continuity Digital breadth Digital intensity Technology innovation Process innovation Business innovation Digital innovation qualification Scientific knowledge generation The number of industry standards developed

. Among them, the input indicators are digital technology, the total amount of investment, the number of R&D personnel, the number of R&D projects, the R&D funding, and the number of infrastructures built; the output indicators are the digital strategy looking forward, the digital strategy continuity, the digitization breadth, the digital intensity, technology innovation, process innovation, business innovation, digital innovation qualification, scientific knowledge generation, and the number of industry standards developed. Then we use DEA to analyze and obtain the efficiency value of digital transformation.

4.2.3. Mediating Variable

In this paper, we use financing constraints (FCs) as the mediating variable, the difficulties firms face in accessing funds due to issues like information asymmetry and high costs. We use the enterprise finance cost ratio as a proxy for FCs, where a higher ratio signifies greater financing constraints and a lower ratio indicates fewer constraints.

4.2.4. Control Variables

To control the influence of other factors on the enterprises’ performances, in this paper we draw upon existing studies and select the following variables as control variables (Chang et al., 2015): enterprise gearing ratio (Lnleverage), the ratio of total liabilities to total assets of an enterprise, which controls the impact of capital structure on performance; enterprise age (Lnage), which controls the strategic decision-making behavior of enterprises at different stages of development; enterprise size (Lnsize), the total assets of the enterprise at the end of the period; cash ratio (Lncashradio) = cash and cash equivalents closing balance/current liabilities; asset structure (Lntang) = (net Fixed Assets + net Inventories)/total assets; and book-to-market ratio (Lnmbratio), controlling the enterprises’ ability to grow. Additionally, we control for industry fixed effects and year fixed effects. The definition and measurement of each variable are provided in

Table 5. Variable definition table.

Variable Name	Variable Identification	Measurement
Market Performance	TobinQ	Market value/replacement cost of assets
	ROA	Net profit/average total assets
Innovation Performance	EFFI	Ln (number of invention patent applications for the year)
	Patent_RD	Total number of invention patent applications/Total R&D investment
Digital Transformation Efficiency	DT	DEA
Financing Constraints	FC	Financial expenses/operating income
Gearing Ratio	Lnleverage	Ln (total liabilities/total assets)
Age	Lnage	Ln (enterprise age)
Cash Ratio	Lncashradio	Ln (closing balance of cash and cash equivalents/current liabilities)
Enterprise Size	Lnsize	Ln (total assets at the end of the specific year)
Asset Structure	Lntang	Ln ((net fixed assets + net inventories)/total assets)
Book-to-Market Ratio	Lnmbratio	Ln (total assets/market capitalization)

.

4.3. Models

To verify Hypotheses 1 and 2, the following panel regression model applied in this study can be expressed in the following form:

$$Tobin{Q}_{it}={\alpha }_0+{\alpha }_{1}D{T}_{it}+{\alpha }_{2}Control{s}_{it}+\sum Industr{y}_i+\sum Yea{r}_t+{\epsilon }_{it}$$

(4)

$$EFF{I}_{it}={\beta }_0+{\beta }_{1}D{T}_{it}+{\beta }_{2}Control{s}_{it}+\sum Industr{y}_i+\sum Yea{r}_t+{\epsilon }_{it}$$

(5)

Subscripts i and t, respectively, denote enterprises and years. $Tobin{Q}_{it}$ refers to the proxy variable of manufacturing enterprises’ market performance, $EFF{I}_{it}$ refers to the proxy variable of manufacturing enterprises’ innovation performance, and $D{T}_{it}$ represents an indicator of digital transformation. If the coefficient ${\alpha }_1$ is significant and positive, Hypothesis 1 is verified. If the coefficient ${\beta }_1$ is significant and positive, Hypothesis 2 is verified. $Control{s}_{it}$ represents the control variables. Models (4) and (5) also include industry fixed effects $\sum Industr{y}_i$ and time fixed effects $\sum Yea{r}_{it}$ to control for industry-invariant and time-invariant characteristics.

In order to accurately measure the mediating effect of financing constraints in the path of digital transformation’s effect on enterprise performance, in this paper we establish the following mediation regression model:

$$F{C}_{it}={\delta }_0+{\delta }_{1}D{T}_{it}+{\delta }_{2}Control{s}_{it}+\sum Industr{y}_i+\sum Yea{r}_t+{\epsilon }_{it}$$

(6)

$$Tobin{Q}_{it}={\mu }_0+{\mu }_{1}D{T}_{it}+{\mu }_{2}F{C}_{it}+{\mu }_{3}Control{s}_{it}+\sum Industr{y}_i+\sum Yea{r}_t+{\epsilon }_{it}$$

(7)

$$EFF{I}_{it}={\omega }_0+{\omega }_{1}D{T}_{it}+{\omega }_{2}F{C}_{it}+{\omega }_{3}Control{s}_{it}+\sum Industr{y}_i+\sum Yea{r}_t+{\epsilon }_{it}$$

(8)

where the mediating variable $F{C}_{it}$ is the financing constraint; ${\delta }_1$ is the regression coefficient of digital transformation on financing constraints; ${\mu }_1$ is the direct effect of digital transformation on market performance; ${\mu }_2$ is the regression coefficient of financing constraints on market performance; ${\omega }_1$ is the direct effect of digital transformation on innovation performance; and ${\omega }_2$ is the regression coefficient of financing constraints on innovation performance. If the coefficients ${\delta }_1$ and ${\mu }_2$ are significant and negative, and the coefficient ${\mu }_1$ is positively significant and lower than the coefficient ${\alpha }_1$ in baseline regression, Hypothesis 3 is verified. If the coefficients ${\delta }_1$ and ${\omega }_2$ are significant and negative, and the coefficient ${\omega }_1$ is positively significant and lower than the coefficient ${\beta }_1$ in baseline regression, Hypothesis 4 is verified.

5. Results

5.1. Results of Baseline Regression

In this paper, we assess the impact of digital transformation on enterprises’ market performance and innovation performance. In

Table 6. The impact of digital transformation on enterprise performance.

Variable	Market Performance		Innovation Performance
	(1)	(2)	(3)	(4)
	TobinQ	TobinQ	EFFI	EFFI
DT	0.248 *** (0.088)	0.278 *** (0.084)	0.718 *** (0.235)	0.840 *** (0.218)
_cons	−0.881 *** (0.098)	−0.320 (0.618)	2.633 *** (0.238)	−2.712 (2.028)
N	800	800	800	800
R-squared	0.498	0.532	0.540	0.591
Controls	No	Yes	No	Yes
Industry	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes

Note: Numbers in parentheses are standard errors; *** (1%), ** (5%), and * (10%) indicate significance at the corresponding level.

, we display the empirical results of the baseline regression. The results of the baseline regression show that the regression coefficients of the main independent variable (DT) are significantly positive at the 1% level, which implies that digital transformation significantly improves both the market performance and innovation performance of enterprises, confirming hypotheses H1 and H2 in this study.

5.2. Robustness Tests

To check the robustness of the results, we perform robustness tests, the results of which are detailed in

Table 7. Results of the robustness analysis.

Variable	(1)	(2)	(3)	(4)	(5)	(6)
	ROA	Patent_RD	TobinQ	EFFI	TobinQ	EFFI
DT	0.023 *** (0.005)	0.017 *** (0.003)
L.DT			0.221 *** (0.070)	0.156 *** (0.050)
L2.DT					0.133 * (0.069)	0.166 *** (0.055)
_cons	0.083 *** (0.014)	0.692 *** (0.023)	−0.498 (0.563)	0.716 (0.477)	−0.737 (0.573)	0.586 (0.490)
N	800	800	720	720	640	640
R-squared	0.586	0.735	0.544	0.622	0.525	0.620
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Industry	Yes	Yes	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes	Yes	Yes

Note: Numbers in parentheses are standard errors; *** (1%), ** (5%), and * (10%) indicate significance at the corresponding level.

. Initially, we use ROA, the average total assets, to substitute TobinQ as a market performance metric, with the outcomes in column (1). Next, we use Patent_RD, the number of licensed patents per R&D dollar, to substitute the EFFI as an innovation performance metric (Hirshleifer et al., 2013), with the outcomes in column (2).

Lastly, to account for time lags, we lag the digital transformation efficiency variable (DT) by one and two periods. The results of the regression are displayed in columns (3)–(6). The robustness test outcomes are consistent with the baseline model, suggesting that the study’s findings are robust.

5.3. Endogeneity Analysis

To address endogeneity from sample selectivity bias, we use propensity score matching (PSM) with k-nearest-neighbor matching (k = 4). Before matching, there is a significant standardization bias, but post matching, all covariate standardization biases are under 10%. Columns (1) and (2) in

Table 8. Results of endogeneity analysis.

Variable	PSM		OLS
	(1)	(2)	(3)	(4)	(5)
	TobinQ	EFFI	DT	TobinQ	EFFI
DT	0.166 *** (0.027)	0.124 *** (0.038)		1.613 *** (0.617)	1.705 *** (0.615)
instrumental variable (L2.DT)			0.169 *** (0.031)
_cons	1.180 *** (0.027)	3.651 *** (0.019)	1.005 ** (0.432)	−1.018 (0.921)	0.658 (0.844)
N	750	750	640	640	640
Controls	Yes	Yes	Yes	Yes	Yes
Industry	Yes	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes	Yes
Kleibergen–Paap rk LM statistic			28.073 *** (p value 0.000)
Kleibergen–Paap Wald rk F statistic			28.993 [16.38]

Note: Numbers in parentheses are standard errors; *** (1%), ** (5%), and * (10%) indicate significance at the corresponding level. Critical values of the F-test for weak instrumental variable identification at the 10% level are in square brackets.

show the PSM regression results. The DT regression coefficient remains significantly positive at the 1% level, aligning with the baseline results, confirming that digital transformation boosts firm performance robustly.

High-performing enterprises often invest more in digital transformation, potentially enhancing efficiency. To address endogeneity, we use an instrumental variable method with a two-stage least squares regression. We draw upon existing research and select the lagged variable from two periods prior as an instrumental variable (Zhang et al., 2023). The lagged form of the independent variable is an instrumental variable favored by numerous scholars, as it effectively resolves endogeneity issues. The results show a significant positive correlation between the instrumental variable and digital transformation efficiency. The second-stage regression also indicates a significant positive effect of digital transformation efficiency on performance. The instrumental variables are validated with a low LM statistic p-value and a high F-statistic, confirming the robustness of the baseline regression results after accounting for endogeneity.

5.4. The Mediation Effect of Financing Constraints

Table 9. The mediating mechanism of financing constraints.

Variable	(1)	(2)	(3)	(4)	(5)
	TobinQ	FC	TobinQ	EFFI	EFFI
DT	0.278 *** (0.084)	−0.001 ** (0.000)	0.248 *** (0.082)	0.840 *** (0.218)	0.776 *** (0.215)
FC			−34.869 *** (12.288)		−74.046 ** (31.346)
_cons	−0.320 (0.618)	0.012 *** (0.004)	0.101 (0.618)	−2.712 (2.028)	−1.817 (2.003)
Boot 95%	[0.0078387, 0.1289372]			[0.0084812, 0.0781988]
N	800	800	800	800	800
R-squared	0.532	0.186	0.543	0.591	0.597
Controls	Yes	Yes	Yes	Yes	Yes
Industry	Yes	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes	Yes

Note: Numbers in parentheses are standard errors; *** (1%), ** (5%), and * (10%) indicate significance at the corresponding level. Confidence intervals for mediating effects are in square brackets.

shows that the financing constraints (FCs) mediate the link between digital transformation (DT) and enterprise performance. Columns (1) and (4) show the overall impact of digital transformation on the market performance and innovation performance of enterprises, column (2) shows the impact of digital transformation on corporate financing constraints, and columns (3) and (5) show the regression results after taking both the financing constraint variable and the digital transformation variable as explanatory variables. Columns (3) and (5) show that the negative coefficients for FCs and positive ones for DT indicate FCs’ mediating role, suggesting that DT improves enterprises’ market and innovation performance by easing financing issues. Bootstrap tests support this, with values outside the 95% confidence interval confirming hypotheses H3 and H4. Therefore, DT enhances market and innovation performance by reducing financing constraints.

5.5. Heterogeneity Analyses

Digital transformation depends on long-term investments in human capital and finances. Large enterprises possess financial resources, large scales, and targeted talent reserves, enabling efficient resource utilization during transformation. They can also withstand uncertainties encountered during the transformation process and adjust strategies, resources, and plans for long-term development. SMEs, however, may face budget, cost, and talent shortages, hindering progress and efficiency. Their lower tier in the industrial chain also challenges transforming inputs into productive outputs, impacting economic outcomes.

We investigate how digital transformation affects enterprises’ performances among sizes by categorizing enterprises based on revenue. Referring to the Regulations on the Classification Standards for Small and Medium-sized Enterprises issued by the Ministry of Finance, we divide the sample enterprises into large enterprises and small and medium-sized enterprises (SMEs) according to their operating income. Columns (1) and (3) in

Table 10. Heterogeneity test results.

Variable	Market Performance		Innovation Performance		Market Performance		Innovation Performance
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Large Enterprises	SMEs	Large Enterprises	SMEs	Mature Enterprises	Growth-Stage Enterprises	Mature Enterprises	Growth-Stage Enterprises
DT	0.455 *** (0.109)	0.095 (0.108)	0.455 *** (0.109)	0.095 (0.108)	0.304 *** (0.104)	0.103 (0.116)	0.340 *** (0.059)	0.121 (0.116)
_cons	2.599 * (1.482)	−4.548 *** (1.670)	2.599 * (1.482)	−4.548 *** (1.670)	−0.198 (0.731)	−0.214 (1.286)	0.925 * (0.506)	−1.071 (0.983)
N	580	220	580	220	653	147	653	147
R-squared	0.614	0.507	0.614	0.507	0.533	0.696	0.631	0.774
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Industry	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

Note: Numbers in parentheses are standard errors; *** (1%), ** (5%), and * (10%) indicate significance at the corresponding level.

show the regression results for the large enterprises, while columns (2) and (4) show the results for the SMEs. The results show that digital transformation significantly improves the performance in large enterprises, while its impact on SMEs is positive but not significant. This indicates that digital transformation benefits large enterprises more.

Enterprises operate at various life cycle stages with unique objectives. Growth-stage enterprises aim for survival, facing challenges in digital transformation due to limited experience, restricted access to market information, and immature profit models. Moreover, the scarcity of talent contributes to a relatively sluggish pace in the digital transformation process. Conversely, mature enterprises prioritize stable development and expansion, with managers recognizing digital transformation as a business model. They actively seek development momentum through resource integration, with defined objectives and systematic strategic planning for higher input-to-output ratios.

To study the effects of digital transformation throughout the business cycle, we divide the enterprises into mature and growth stages based on their age. Mature enterprises, over ten years old, have overcome initial challenges and are more established (Jefferson et al., 2008). Columns (5) and (7) in Table 10 show the regression results for the mature enterprises, while columns (6) and (8) show the regression results for the growth-stage enterprises. The results show that digital transformation significantly improves the performance of mature enterprises, while its impact on growth-stage enterprises is not statistically significant. This indicates that digital transformation has a greater impact on the performance of mature enterprises.

6. Discussion and Conclusions

We use data from 2012 to 2021 to assess the digital transformation efficiency of Chinese A-share manufacturing enterprises with DEA. We then examine how the efficiency affects enterprises’ performances. The results show that digital transformation positively affects market and innovation performance and the findings hold up under robustness and endogeneity tests.

The mediation analysis reveals that digital transformation contributes to both market and innovation performance improvements by alleviating financing constraints. This mechanism shows that enterprises that use digital technologies more efficiently improve information transparency, enhance credibility with financial institutions, and gain better access to capital.

Heterogeneity analysis indicates that the benefits of digital transformation efficiency differ in different types of enterprises. Large and mature enterprises gain more substantial advantages compared with SMEs and enterprises in the growth stage. This variation shows that digital transformation is not only a technological process but also a managerial consideration that depends on resource endowment, organizational maturity, and strategic alignment.

Our study enriches the existing literature by shifting attention from the extent of digitalization to its efficiency, thereby offering a more nuanced understanding of how digital transformation creates value. By highlighting efficiency as the bridge between the input and output of digital transformation, our study expands both the definition and measurement of digital transformation and contributes new empirical evidence to the field.

For policymakers, we suggest that governments should create supportive environments to enable enterprises, especially SMEs, to overcome financing barriers during digital transformation. Governments should strengthen policy guidance, optimize financial resource allocation, and promote digital financing infrastructure through preferential loans, guarantees, and fintech-driven platforms. This will help balance digital transformation opportunities across enterprise types and sustain the competitiveness of the manufacturing sector.

For enterprises, the results of our study provide actionable insights. Firstly, enterprises should view digital transformation as a process of capability reconfiguration rather than a technology upgrade. By improving data integration, financial transparency, and coordination among departments, enterprises can better convert digital investments into performance gains. Growth-stage enterprises should focus on building core digital capabilities early and optimizing resource use, while mature enterprises should leverage their financial and technological advantages to lead comprehensive digitalization. SMEs, in particular, should enhance their financial management, strengthen reputation mechanisms, and engage in collaborative financing models to ease constraints and improve resilience.

Despite the robustness of our results, several limitations should be acknowledged. We use 80 manufacturing enterprises from 2012 to 2021 in our sample, which is relatively small and industry-specific. Future research could broaden the scope to include other sectors and extend the observation period to capture the effects of more recent digital policies and technologies. Additionally, while efforts were made to reduce interference, other factors could still have influenced the results. Future research could use alternative empirical methods to test the findings’ robustness.

Future work could also explore how emerging digital technologies reshape efficiency and financing dynamics across industries. Comparative studies among different economies may reveal institutional factors that affect the relationship between digital transformation efficiency and enterprises’ performances. Furthermore, building upon our research findings, more extensions to the study may be pursued. We assess enterprises’ digital transformation from the perspective of digital transformation efficiency, whilst the efficiency derived from DEA represents efficiency among decision-making units. This also indicates that the scope of sample coverage could be further narrowed. For instance, by selecting enterprises possessing certain characteristics (such as high-tech enterprises or state-owned enterprises) as samples for more in-depth analysis, we may potentially discover additional distinct influencing mechanisms.