Unleashing the Power of Digital Transformation: Boosting Green Total Factor Productivity in China’s Energy Enterprises

Abstract

This study examines the influence of digital transformation (DT) on green total factor productivity (GTFP) for A-share-listed energy enterprises from 2013 to 2022. The study findings demonstrate that DT can be advantageous in improving GTFP, and this conclusion remains valid even after performing the endogeneity treatment and the robustness test. The mechanism test suggests that improving green technology innovation and alleviating financing constraints are significant transmission paths. The heterogeneity analysis’ findings show that state-owned enterprises, large-scale enterprises, and new energy enterprises benefit more significantly from DT’s favorable efficacy in improving GTFP. By focusing on energy enterprises, this study expands upon the prior research on digital transformation in the micro field. Moreover, this study reveals the critical path of digital transformation in enhancing GTFP, thus enriching its theoretical connection. GTFP will be enhanced by the joint development of digital facilities by enterprises and the government, as well as using distinct digital strategies.

1. Introduction

Resource shortages and environmental pollution issues are exacerbated by energy enterprises’ manufacturing processes, which frequently include excessive energy use and emissions [1]. There has been a historical turning point in the green transformation of energy companies due to the implementation of the “30 × 60 dual-carbon” target [2]. Furthermore, the goal of raising renewable energy’s percentage in total energy consumption from 15% to 85% by 2050 highlights the importance of renewable energy even more [3]. However, energy enterprises are still facing great difficulties in building a resource-saving and low-carbon energy system and achieving green production [4]. In striving to adapt to a fast-changing environment, energy enterprises are aggressively adopting digital transformation (DT) by decreasing pollutants while enhancing production efficiency and economic value added [5]. Artificial intelligence, big data, blockchain, and cloud computing (ABCD) constitute the core underlying technology architecture for DT [6]. Through these technologies, energy enterprises can optimize the capacity of resource allocation and the safety and security of energy production, transmission, trading, and consumption, considerably advancing the pace of energy revolution [3]. According to the China Digital Economy Development Report (2022) released by the China Academy of Information and Communications Technology (CAICT), investment in the digitalization of energy producers in China has increased, from CNY 654.42 billion in 2015 to CNY 1051.5 billion in 2022 [3].

The DT of energy enterprises is worthy of attention due to the following reasons. First, energy enterprises are vital players in the achievement of “dual-carbon” goals. The total energy consumption of China in 2022 was 159.39 EJ, with a year-on-year growth of 0.9% and representing 26.4% of the total global energy consumption [7]. Moreover, China’s energy consumption has expanded at an average annual pace of 3.1% during the previous decade, maintaining its lead as the greatest worldwide energy consumer for 14 years [2]. However, in this process, the development mode of energy enterprises due to their high consumption and pollution has produced massive amounts of carbon emissions [5]. By 2030, carbon emissions will amount to 12.1 billion tons, with those of the energy sector exceeding 1.8 billion tons [8]. Thus, one key strategy for meeting the dual-carbon goal on time is for energy enterprises to cut their carbon emissions [9,10]. Second, applying DT is seen to be a vital step towards assisting energy enterprises to transition to low-carbon operations. Digital technology could enable energy enterprises to conduct feasibility studies for energy conservation and emission reduction and to precisely track the dynamics of carbon emissions [11]. For example, smart power grids based on digital technology can precisely predict the path of carbon emissions during energy production, which is beneficial for controlling pollutant emissions from the source [12]. In the production process, digital technology enables the efficient coordination of power generation, transmission, and distribution and solves problems such as low efficiency and long industrial chains [13]. Therefore, energy quality and efficiency are improved and carbon emissions reduced. Finally, digitization exhibits great potential in improving the sustainable productivity of energy enterprises [14]. Digitalization has intrinsically low marginal costs and the minimal emissions of pollutants as components of production [2]. Thus, the increase in the degree of DT is itself an optimization of the factor structure that can facilitate the shift in energy enterprises towards green production [15]. Additionally, digital technology has broken the boundaries and barriers of regional markets, allowing green elements to flow freely and fully in a larger space [16]. This is conducive to the precise matching of various energy resources in the process of production and transportation, thereby improving green productivity.

This study’s primary goal is to evaluate the relationships and mechanisms of transmission between energy enterprises’ green total factor productivity (GTFP) and digital transformation (DT). The following represents this study’s anticipated contributions. First, this study reveals a relationship between DT and GTFP from a micro viewpoint. Most previous researchers in this field have examined the relationship between DT and green production from a macro standpoint, with few studies focusing on microenterprises. Therefore, based on the micro view, with the emphasis on energy enterprises, this study investigates the efficacy of DT in fostering the green transformation of energy by performing the empirical and theoretical analyses of the process. Second, this study broadens the research related to digital transformation. Existing studies have investigated the economic benefits of DT more from the aspect of productivity, but environmental factors have been largely neglected. By considering production efficiency and green sustainability, this study bridges this gap, thus providing a better theoretical basis for DT to promote environmental protection. Finally, this study provides practical insights from different perspectives for energy enterprises undergoing digital transformation. More specifically, the empirical models control for ownership, size, and nature of business, and the results demonstrate that the implementation of digital strategy exhibits heterogeneity, according to an enterprise’s status as being State-Owned, the scale of its operations, and whether it uses new or traditional energy. This important theoretical foundation provides an essential insight for differential policy governance and guidance.

The following is the format for the remaining portion of this article: The related literature is included in Part 2. The theoretical analysis and hypothesis are covered in Part 3. The approach is introduced in Part 4. The data sources and descriptive statistics are presented in Part 5. The empirical outcome is analyzed in Part 6, and Part 7 provides the policy implications.

2. Literature Review

2.1. The Application of Digital Transformation in Energy Enterprises

The application of DT in energy enterprises mainly focuses on energy efficiency and energy security. Big data and artificial intelligence technology has been effectively applied in oil and gas enterprises in areas such as health and safety environments, refining, and transportation, resulting in an increased energy efficiency [17,18]. Furthermore, DT improves energy technology innovation [6] and reduces energy use [19], which makes energy systems safer and more efficient and enables enterprises to maximize the efficiency of resource allocation [1]. DT is also beneficial for upgrading the operational efficiency of China’s power enterprises [20]. The DT of power plants and organizations that sell renewable energy will result in the advantages of collaboration, encouraging the improvement in energy efficiency [3,13].

2.2. Digital Transformation and Green Development

There is no consensus in a previous study on the connection between DT and green development, but it primarily presents three viewpoints: positive influence, negative impact, and nonlinear effect. First, a majority of studies demonstrate that DT stimulates green development. Digital transformation effectively replaces conventional components and may reduce pollution to the environment and enhance green growth by boosting energy efficiency [21] and optimizing industrial structure [22]. DT can successfully persuade heavily polluting or established businesses to uphold their environmental obligations and engage in green growth [23]. Specifically, DT can better promote the energy conservation and emission reduction of enterprises, further meeting the inherent requirements of green development [24,25]. The degree of green growth of mineral energy enterprises has increased with digitalization [4,26].

On the contrary, some scholars have pointed out that digital transformation has an environmental damage effect; thus, it is not conducive to green development [27,28]. There is increasing evidence that digital technology is mostly energy-intensive; thus, the deepening of digitization will result in a higher power consumption [29]. Furthermore, the rapid adoption of the Internet and the large-scale development of digital infrastructure have resulted in resource and energy consumption, which have increased environmental degradation [30].

In addition, a few scholars have studied the nonlinear relationship between DT and green development. The degree of digitization and the level of green development display an inverted U-shaped curve [31]. It means that, in the early stage of digital development, it has a significant role in promoting green development, and, when it develops to a certain stage, the promotional effect gradually weakens [32]. Another study has innovatively suggested a theoretical framework for analyzing and explaining the nonlinear link that exists between digitalization and green development [33], which provides a theorical basis for future research. Several studies have focused on the regional spillover effects, pointing out that digital technology has an inverted U-shaped regional spillover effect on China’s carbon emissions [34]; that is, the carbon emissions in surrounding areas may increase in the early stage of digital technology development, but these carbon emissions will be mitigated after the further development of digital technology. These studies provide rich research ideas for the academic community and promote the sustainable development of this field.

2.3. Digital Transformation and Green Total Factor Productivity

The majority of previous studies on the connection between DT and GTFP are concentrated on the macro level, which includes nations and cities. In terms of cities, DT involving information and communication technology will positively promote the GTFP of Chinese cities [35], but its impact varies with the degree of resource misallocation, and it has a positive and nonlinear impact on the GTFP growth based on digital development [36]. For nations, digital technology has the potential to lower factor flow transaction costs, eliminate time and space restrictions, increase production efficiency, and advance GTFP [37]. Moreover, input digitization can increase the economic generation of per unit energy, reduce energy intensity [38] and energy consumption [39], and, ultimately, achieve the improvement in GTFP. However, its nonlinear impact is likely to remain “U-shaped” in the short term [40]. In addition, a few scholars concentrate on the micro level, mainly on enterprises. Digitization has a major beneficial impact on manufacturing enterprises’ green productivity [2], but its promotional degree will vary with the features of enterprises, industry characteristics, and provincial environmental factors [25].

In summary, scholars have studied DT’s effects on GTFP in great detail at the macro level, but there have not been many studies conducted at the micro level. This is especially true for energy enterprises, which have been largely disregarded as a crucial area of economic growth. Therefore, this study focuses on energy enterprises to explore the direct impact of DT on GTFP. In addition, an analysis of the mediating effect between the two is made, which extends the existing literature.

3. Research Hypotheses

With the deepening of digitization, digital technology continues to empower energy enterprises in manufacturing and innovation activities and constantly improves GTFP. In addition, the information asymmetry between enterprises and investors has weakened due to the use of digital technologies, thus enabling enterprises to obtain more financial support to engage in green production. More specifically, enterprises have their own green production project planning, operation data, profit expectations, and other information, while investors know little about this core information. A low level of information hinders investment decisions, resulting in obstacles for corporate green production project financing. However, digital technology can achieve real-time information sharing, and investors can easily obtain the dynamic data of enterprises and quickly judge the investment value, thus assisting enterprises to obtain more funds for green production. Therefore, this study proposes the first hypotheses:

Hypothesis 1:

DT improves the GTFP of energy enterprises.

This study mainly discusses the influence of DT on GTFP via two mechanisms of green technology innovation and financing constraints. First, through encouraging the development of green technologies, digital transformation contributes to an increase in GTFP. DT can completely substitute traditional manufacturing technology that causes a high level of pollution with green technology. Digital technologies facilitate the continuous progress and enhancement in environmentally friendly technologies through horizontal information sharing, technical modularization, and both the vertical and “bottom-up” reconstruction of industrial systems [41]. With the penetration of green technology into various production stages of enterprises, it may instantly decrease energy consumption and improve productivity in the manufacturing processes of energy enterprises and stimulate the enhancement of GTFP [42,43]. Furthermore, DT is advantageous for enhancing green technology innovation’s efficacy. Energy enterprises use the communication platform of big data, artificial intelligence, and the cloud computing architecture industry chain to share data of products from production to sales and to optimize and improve innovative activities in real time [23,44]. Furthermore, DT drives the efficient docking of production factors and capacity sharing between enterprises, so that energy enterprises can extend the innovation output boundary with limited resources and greatly strengthen the innovation efficiency and economic benefits of enterprises, thus enhancing GTFP.

Hypothesis 2:

DT contributes to GTFP by enhancing green technology innovation.

Second, DT improves GTFP by easing the financing constraints of energy enterprises. Moreover, digital technology can alleviate the information asymmetry between energy enterprises and financial institutions. Under the traditional financing mode, there is a significant information gap between energy enterprises and financial institutions: it is difficult for energy enterprises to fully understand the details of financing products of financial institutions, and financial institutions also lack a sufficient understanding of the operation status and project prospects of energy enterprises, which restricts the decision making of both parties. However, with the characteristics of data integration and real-time interaction, digital technology can ensure the two-way synchronous update of information, thereby effectively breaking the information barrier. Energy enterprises can have direct, effective, and timely access to the amount, interest rate, loan speed, and other corporate financing data through digital platforms, thus selecting the most appropriate financing model according to their actual situation. Furthermore, financial institutions can also provide a more appropriate loan amount according to the business dynamics of enterprises [45]. With strong financial support, energy enterprises have more resources and capabilities to cater to the market for green development and enhance GTFP. Furthermore, DT can bring together various types of factor subjects through open and shared circulation platforms. In addition to financial institutions, social capital can aid in performing investment activities that broaden financing channels [6]. The introduction of new investors increases GTFP, enhances financing speed and efficiency, and lessens the limitations imposed by supply and demand for capital. Therefore, this study proposes the following hypotheses based on the above analysis:

Hypothesis 3:

DT contributes to GTFP by alleviating financial constraint.

4. Methodology and Data

4.1. Basic Model

The primary goals of the benchmark regression model are to confirm the first theoretical hypothesis and investigate the consequences of DT on the GTFP of energy enterprises. Furthermore, the estimation results of the fixed-effects (FE) model are impartial and uniform; thus, reliable regression results can be obtained [25]. Therefore, this study constructs the following panel regression model for empirical estimation:

$${GTFP}_{it}={\alpha }_0+{\alpha }_1{DT}_{it}+{\alpha }_2{Controls}_{it}+{\mu }_i+{\epsilon }_{it}$$

(1)

where the explanatory variable ${GTFP}_{it}$ represents the green total factor productivity of i enterprise in t year, and the core explanatory variable ${DT}_{it}$ is the degree of digital transformation of i enterprise in t year. The set of control variables is denoted by ${Controls}_{it}$, the fixed effects of firm and time are expressed as ${\mu }_i$, and the residual is presented by ${\epsilon }_{it}$. The significance and influence direction of the estimated coefficient ${\alpha }_1$ of DT is the focus of this study. If ${\alpha }_1$ is greater than 0 and passes the 10% level of statistical significance test, it indicates that DT contributes to enhancing energy enterprises’ GTFP; contrarily, it suggests that DT inhibits GTFP.

4.2. Measurement for GTFP

The specific methodology for measuring the GTFP is as follows:

An environmental technology model is firstly defined. It is assumed that there are n decision-making units in the model (DMUj, j = 1, 2, …, n) and each decision-making unit has three input–output indicators, including “input x”, “expected output ye”, and “unexpected output yu”, which is shown as Equation (1).

$${\displaystyle \frac{P}{x_0,y_0}}=\left\{\left(\overline{x},{\overline{y}}^e,{\overline{y}}^u\right)|\overline{x}\ge {\sum }_{k=1}^n{\lambda }_k{x}_k,{\overline{y}}^e\le {\sum }_{k=1}^n{\lambda }_k{\overline{y}}^e,{\overline{y}}^u\le {\sum }_{k=1}^n{\lambda }_k{\overline{y}}^u,\lambda \ge 0\right\}$$

(2)

Under the assumption of m “inputs”, s1 “expected output”, and s2 “unexpected output”, the Super-SBM model which considers non-expected outputs is shown as follows:

$$min\theta ={\displaystyle \frac{\frac{1}{m}\sum _{t=1}^m\left\{\frac{\overline{x_i}}{x_{ik}}\right\}}{\frac{1}{s_1+s_2}\left\{\frac{\sum _{r=1}^{s_1}{\overline{y}}^e}{y_{ik}^e}+\frac{\sum _{r=1}^{s_1}{\overline{y}}^u}{y_{ik}^u}\right\}}}$$

(3)

$$\overline{x}\ge \sum _{j=1,j\ne k}^n{x}_{ij}{\lambda }_j,{\overline{y}}^u\le \sum _{j=1,j\ne k}^n{y}_j{\lambda }_j$$

$$\overline{x}\ge x_0,0\le y_k^e,{\overline{y}}^u\ge y_k^u,\sum _{j=1,j\ne k}^n{\lambda }_j=1,\lambda \ge 0$$

where the desired efficiency value is represented by θ, the weight vector by λ, and the measured decision unit by k. Under the set production possibility and appropriate time period, the Super-SBM model can be described as folows:

$$\overline{D_0^G}(x^t,y^t,b^t;y^t,{-b}^t)$$

(4)

However, considering that the SBM model can only measure GTFP at a certain time cross-section and is not sensitive to the process of dynamic change, in view of this, Oh [46] proposed GML index, which can observe the change in productivity among decision-making units over time. Therefore, this paper further derives the GML index based on the directional distance function obtained by the super-efficient SBM.

$${GML}_t^{t+1}={\displaystyle \frac{1+\overline{D_0^G}(x^t,y^t,b^t;y^t,{-b}^t)}{1+\overline{D_0^G}(x^{t+1},y^{t+1},b^{t+1};y^{t+1},{-b}^{t+1})}}$$

(5)

The GML index is broken down into GTC and GEC under constant returns to scale. GEC measures the intertemporal global technical efficiency change. GEC > 1 implies that efficiency has improved significantly compared to the previous period, while GEC < 1 indicates the opposite. GTC assesses the worldwide shift in technical advancement throughout time, including the changes brought about by the advances in production technology and process innovation. GTC > 1 suggests that the level of technology is advancing, with an expected increase in desirable output, while GTC < 1 indicates that technology is degrading, potentially leading to an increase in undesired output.

In terms of specific indicator selection, this paper combines the “economic–resource–energy–environmental” factors to construct the GTFP measuring framework. As indicated in

Table 1. Input–output indicators of energy enterprise GTFP.

Indicators	Index	Measure Method	Data Source
Input	Labor	Number of employees in the enterprise	CSMAR
	Capital	Total fixed assets of the enterprise	CSMAR
	Energy	Referring to Wang et al. [25]	SSCY
Desired output	Industrial value added	Total business income of the enterprise in the current year	CSMAR
Non-desired output	SO2, Smoke and dust, Wastewater	Referring to Wang et al. [37]	SSCY

, the input indicators are labor, capital, and energy, while the output indicators are desired and non-desired output. Since the absence of specific data at the enterprise level on electricity, energy consumption, and industrial waste emissions, referring to Wang et al. [25], this article employees the enterprise’s industrial power usage in the city where it is located and the percentage of enterprise employees in the employment of urban personnel in the city as a proxy variable.

The following is the calculation for non-desired outputs:

First, the adjustment coefficient of each pollution indicator for each prefecture-level city is calculated:

$$W_j={\displaystyle \frac{P_{ij}/\sum P_{ij}}{O_i/\sum O_i}}$$

(6)

where P_i represents the emissions of pollutant j (j = 1, 2, 3) of city i, and ∑ P_i denotes the sum of all cities’ pollutant emission. O_i denotes the total amount of the industrial output value of city i, and ∑ O_i is the sum of all city.

Second, the emission of pollutant j in city i after weighted adjustment is calculated:

$${em}_{ij}=W_i\times Y_{ij}$$

(7)

where Y_ij is a representation of the pollutant j’s initially emission in city i.

Finally, the emission of pollutant j of enterprise k in city i is obtained:

$${em}_{kj}={em}_{ij}\times \left({\displaystyle \frac{O_k}{\sum O_k}}\right)$$

(8)

where O_k is the total industrial output value of enterprises k, and ∑ O_k is the total industrial output of the prefecture-level city where enterprises k is located, respectively.

4.3. Mechanism Testing

A further exploration of the transmission mechanism of DT affecting GTFP is beneficial for better understanding the relationship between the two. Mechanism testing is a fitting method for investigating transmission channels, which visualizes the mediated effects [47]. By referring to the research of Wu et al. [48] and Cheng et al. [49], this study integrates financial constraints (FCs) and green technological innovation (GTI) as influence mechanisms into the model. The specific regression equations are shown in Equations (9) and (10).

$${GTI}_{it}={\beta }_0+{\beta }_1{DT}_{it}+{\beta }_2{Controls}_{it}+{\mu }_i+{\epsilon }_{it}$$

(9)

$${FC}_{it}={\xi }_0+{\xi }_1{DT}_{it}+{\xi }_2{Controls}_{it}+{\mu }_i+{\epsilon }_{it}$$

(10)

The judgment criteria of GTI are presented as follows: If ${\beta }_1$ > 0, it implies that the effect of DT on GTFP can be optimistically transmitted through GTI; that is, when DT exerts a positive (or negative) influence on GTI, GTI also positively (or negatively) impacts GTFP. Conversely, if ${\beta }_1$ < 0, it suggests that the effect of DT on GTFP is negatively transmitted through GTI; that is, when DT exerts a positive (or negative) influence on GTI, GTI negatively (or positively) impacts GTFP. In a similar manner, the decision-making procedure of the mechanism variable FC may be conducted.

4.4. Data Selection

In this study, 207 energy enterprises are eventually obtained based on the A-share-listed enterprises in China’s Shanghai and Shenzhen stock exchanges and the national economic industry classification criteria, and samples labeled as Special Treatment (ST) and ST*, as well as incomplete information and unknown transactions, are deleted. The sample period of data for each variable is from 2013 to 2022. Additionally, data of the explanatory variables, control variables, and mediating variables (financing constraints) are derived from the annual report of the enterprise; the explained variable is obtained from the China Stock Market Accounting Research Database (CSMAR) and the Statistical Year book of Chinese Cities (SSCY); and another mediating variable (green technology innovation) is obtained from World Intellectual Property Organization (WIPO). Noteworthily, samples are based on the company-year as the observation unit.

4.4.1. Explanatory Variables

Leng and Zhang [50] regarded digital transformation (DT) as the explanatory variable. This study first determines the words related to digital transformation, mainly including artificial intelligence, blockchain, cloud computing, and big data (ABCD). Then, Python 3.9 software is used to search, match, and score the frequency of related words in listed enterprises’ annual reports, and negative words with prefixes such as “no”, “not”, and “none” with the root word are excluded. The more times the related words appear, the higher the degree of enterprise digital transformation. Finally, the number of words obtained are added and logarithmically processed.

4.4.2. Explained Variable

Green total factor productivity (GTFP) serves as the explained variable. The study by Tone [51] first employs the Slack-Based Measure (SBM) model to construct the measurement system of GTFP from three aspects: inputs, desired outputs, and undesired outputs. Among them, labor, capital, and energy are considered as input indicators, industrial value added is regarded as a desired output, and three industrial wastes (smoke and dust, SO₂, and wastewater) are taken as undesired outputs. The SBM model is insensitive to the process of dynamic change and can only quantify GTFP at a certain time. In view of this, the study by Oh [46] splits GTFP into green technological progress change (GTC) and green efficiency change (GEC) using the Global Malmquist–Luenberger (GML) index.

4.4.3. Mechanism Variable

Green technological innovation is the first mechanism variable (GTI). The GTI of enterprises is generally measured by the number of green patents. Based on Wu et al. [48], to obtain the GTI, the State Intellectual Property Office (SIPO) matches the green patent application data of listed enterprises with the International Green Patent Classification List, which was launched by the World Intellectual Property Organization (WIPO) in 2010. The resulting data are then logarithmically processed. Notably, the data are processed by adding 1 before taking the logarithm to avoid nulls as much as possible.

In terms of financing constraints (FCs), the existing research on the measurement of FCs mainly includes KZ, WW, and SA indices. However, the first two indices involve the financial indicators of the enterprise in the calculation process, which may pose endogenous risks. Therefore, based on Hadlock and Pierce [52], the SA index, which is more exogenous than other variables and fluctuates less over time, is used in this study to quantify the extent of financial limitations faced by energy enterprises. The specific calculation formula is as follows:

SA = −0.737Size + 0.043(Size)² − 0.040Age

(11)

where Size is obtained by logarithmicizing the total assets of the enterprise at the end of the year, and the units of total assets are converted to millions of CNY to better reflect their economic meaning. Age is represented by the establishment year of the listed enterprise.

4.4.4. Control Variables

To limit the influence of external variables on energy enterprises’ GTFP, this study uses four control variables for the regression equation: the enterprise’s size (Size), debt-to-asset ratio (Debt), Tobin’s Q value (Tobin Q), and the size of board of directors (Board). Among them, Size leads to differences in capital acquisition and resource allocation, thus affecting environmental performance and reflecting in GTFP [25]. Debt measures the enterprise’s operating capacity, and the increase in debt hinders its environmental investment and green innovation, thereby reducing GTFP [53]. Tobin’s Q value indicates enterprises’ investment incentives, which affect enterprises’ energy input and consumption [25]. The size of the Board can not only promote the GTFP of enterprises by optimizing decision making but also may play a reverse restrictive role due to communication and coordination problems [2].

Detailed variable definitions are shown in

Table 2. Variable definitions.

	Variables	Symbol	Definition
Explained variable	Green total factor productivity	GTFP	Log (measured using super-efficiency SBM-GML + 1)
	Technical efficiency changes	GEC	Measured using super-efficiency SBM-GML
	Technological progress changes	GTC	Measured using super-efficiency SBM-GML
Explanatory variables	Digital transformation	DT	Log (the total number of keywords related to digitalization + 1)
Mechanism variable	Green technology innovation	GTI	Log (the number of green patent applications + 1)
	Financing constraints	FC	SA = −0.737Size + 0.043(Size)2 − 0.040Age
Control variable	The scale of the enterprise	Size	Log (Total assets for the year)
	Debt-to-asset ratio	Debt	Year-end total liabilities divided by year-end total assets
	Tobin Q	Tobin Q	Market value divided by replacement cost
	Board size	Board	The number of board of directors

.

4.5. Descriptive Statistics

Table 3. Descriptive statistics of the variables.

Variables	Mean	Std. Dev	Min	Median	Max
GTFP	1.026	0.079	0.880	1.023	1.176
DT	0.158	0.414	0.000	0.030	4.340
GTI	1.370	1.409	0.000	1.099	6.879
FC	5.339	1.887	0.615	5.167	13.637
Size	4.684	21.127	0.017	0.969	273.319
Debt	0.509	0.209	0.021	0.510	3.262
Tobin Q	1.691	1.199	0.227	1.337	17.653
Board	8.881	1.840	3.000	9.000	18.000

presents the descriptive statistics. The mean and median values of DT are 1.158 and 0.030, respectively. The mean is larger than the median, indicating that the degree of DT of the sample energy enterprises exhibits a right-skewed distribution. The maximum (4.340), minimum (0.000), and standard deviation (0.414) values suggest that the digitization levels of energy enterprises vary widely and are unevenly distributed. In terms of GTFP, the mean, median, standard deviation, maximum, and minimum values are 1.026, 1.023, 0.079, 1.176, and 0.880, respectively. The minimum value of GTI is 0, demonstrating some companies have not engaged in green innovation activities. The maximum value of FC is 13.637, indicating some companies face financing constraints, and affect their investment choices. The standard deviation value of Size 21.127, which is the largest among variables, demonstrating energy enterprises’ scale vary widely. The mean and median values of Debt are 0.509 and 0.510, showing energy enterprises own reasonable debt level, and this is beneficial for their development. The median of Tobin Q = 1.34; hence, the majority of sample companies are growth companies. The median of Board is 9, which meets the legal requirements, and can implement effective corporate governance.

Table 4. The results of correlation.

Variables	GTFP	DT	GTI	FC	Size	Debt	Tobin Q	Board
GTFP	1.000
DT	0.277 ***	1.000
GTI	0.244 ***	0.163 ***	1.000
FC	0.149	−0.012	0.469 ***	1.000
Size	0.018	0.142 ***	0.042	−0.355 ***	1.000
Debt	0.025	−0.037 *	0.134 ***	0.257 ***	−0.128 ***	1.000
Tobin Q	0.087 ***	−0.018	−0.173 ***	−0.498 ***	0.267 ***	−0.114 ***	1.000
Board	0.043 ***	−0.026	0.403 ***	0.523 ***	−0.158 ***	0.017	−0.114 ***	1.000

* p < 0.1, *** p < 0.01.

further shows the results of correlation. We find that the absolute values of pairwise correlations among variables are less than 0.5, and even less than 0.2, demonstrating multicollinearity does not exist, and we can implement further analysis. Some major economic interpretations are shown as follows: (1) For GTFP, DT and GTI are positive and significantly correlated, which aligns with the theory of “digital transformation enabling green development”, and highlights technology upgrades as a critical engine for low-carbon transitions. (2) For GTI, DT is positive and significantly correlated, demonstrating digital technologies provide foundational tools for green innovation, such as simulation modeling and data-driven R&D optimization. (3) For Size, DT is positive and significantly correlated, but not significantly related to GTFP, showing large enterprises can better absorb fixed costs of digital transformation, but scale advantages may not directly translate to GTFP gains. (4) For Tobin Q, FC is negative and significantly correlated, indicating high financial constraints may limit investments in high-return projects, depressing market valuations. (5) For Debt, Size is negative and significantly correlated, showing smaller enterprises rely more on debt financing, while larger firms diversify risks through equity markets, leading to lower leverage ratios at scale. (6) For Board, GTI is positive and significantly correlated, revealing larger boards may foster diversified decision making for green innovation and resource acquisition.

5. Results and Discussion

5.1. Benchmark Regression

Table 5. The results of benchmark regression.

	(1) GTFP	(2) GTFP	(3) GTC	(4) GEC
DT	0.0063 ** (0.0021)	0.0060 ** (0.0021)	0.0093 ** (0.0030)	0.0041 * (0.0022)
Size		0.0006 * (0.0003)	0.0008 * (0.0004)	0.0003 (0.003)
Debt		0.0011 (0.0025)	0.0021 (0.0035)	0.0011 (0.0026)
Tobin Q		0.0016 * (0.0008)	0.0026 * (0.0011)	0.0012 (0.0008)
Board		0.0009 * (0.0004)	0.0012 * (0.0006)	0.0004 (0.0004)
year	yes	yes	yes	yes
firm	yes	yes	yes	yes
N	2070	2070	2070	2070
R2	0.9692	0.9693	0.515	0.9670

Standard errors in parentheses * p < 0.1, ** p < 0.05.

displays the results of the baseline regression. Only the explained variable and the primary explanatory variable are present in Column (1), whereas control variables are added in Column (2), and both columns include fixed effects in terms of enterprise and time. Overall, the two regression findings’ coefficients are strikingly positive at the 1% level, suggesting that, the more intense the DT, the more it contributes to the increase in GTFP. From an economic point of view, for every 1% increase in the degree of digital transformation, there will be a corresponding 0.006% growth in enterprise GTFP, which supports the positive effect of Hypothesis 1. In terms of the environmental impact, DT can optimize production processes, facilitate the efficient allocation of energy resources, and achieve the full potential of environmental sustainability. The adoption of digital transformation aids energy enterprises in resource management and energy input monitoring, which lowers production-side energy consumption and pollutant emissions. In addition, energy enterprises may monitor environmental pollutants on the output side and set up and execute environmental management systems with the assistance of DT. In terms of economic benefits, DT has a cost-saving effect, enabling energy enterprises to reduce costs and improve efficiencies, thereby increasing revenues. Customers’ fulfillment can be enhanced by digital transformation, which, in turn, improves business performance [54]. In brief, DT enhances GTFP by encouraging economic benefits and environmental advantages for energy enterprises.

Referring to Wang et al. [25], this paper further examines the effect of DT on the GTC and GEC from dynamic dimensions, and the outcomes are displayed in Columns (3) and (4), respectively. The regression findings show that the coefficients of GTC and GEC are significant at the 5% and 10% levels, respectively, indicating that DT has a dual role in promoting them. However, the enhancement in GTC is the primary source of DT’s contribution to GTFP. The reason for this may be that the DT of energy enterprises has strengthened the degree of connection between internal resources and technology and information and has resulted in the advancement in technological level. In addition, technological progress impacts the entire production process, and the “multiplier effect” drives the increase in expected output and the decrease in undesired output, thus improving GTFP [25]. The weaker contribution of technical efficiency may be because digitization is a long-term transformation activity, which makes the differentiated input factors in the initial stage insensitive to the performance of digital technologies. As digital change proceeds, the integration of elements gradually rises, and the gradual increase in the integration of various factors leads to the improvement in production efficiency.

5.2. Robustness Tests

This study conducts robustness tests from the following three aspects, and all the test results are provided in

Table 6. The valid robustness results of replacing DT, removing specific year and zero-digitization enterprises.

	(1) GTFP	(2) GTFP	(3) GTFP
DT	0.006 ** (0.002)	0.005 ** (0.003)	0.004 ** (0.002)
Controls	yes	yes	yes
year	yes	yes	yes
firm	yes	yes	yes
N	2070	1863	2040
Adj. R2	0.71	0.968	0.968

Standard errors in parentheses ** p < 0.05.

. First, the independent variable of DT is replaced. Based on Cheng et al. [49], the digital transformation of enterprises can also be measured by the proportion of the digital-technology-related assets to the total intangible assets in the financial report. Therefore, to further confirm the reliability of the findings, this study uses the percentage of digital technology assets as the new alternative independent variable, and the results are displayed in Column (1). The prior findings of this work are robust, as evidenced by the positivity and significance of the coefficient estimates, which agree with the outcomes of the benchmark regression. Second, financial crisis factors are considered. After experiencing a severe financial shock, enterprises may prefer to increase the factor inputs of the production process to improve economic efficiency, while the pollution emissions may be ignored and may lower GTFP. Additionally, digital technology is characterized by unpredictability and expensive conversion costs; thus, financial shocks further weaken the risk tolerance of enterprises, thus stagnating digital transformation. Therefore, this study excludes the data from 2015 (the stock market crash of China) to account for endogenous disturbances brought on by significant financial crises, and the result is shown in Column (2). At the 1% level, the DT coefficient is positive, which means the positive effect in Hypothesis 1 still holds true. Finally, the strategic behavior of the enterprises is also considered. The degree of digitization disclosed in the enterprises’ annual reports is likely to be higher than the actual progress [55]. Thus, this study removes samples with a digitization level of 0 to avoid the influence of enterprises’ strategic behavior, and the result is shown in Column (3). The reliability of the study conclusion is further supported by the stability of the direction and significance of the coefficient of DT.

5.3. Mechanism Test

In this section, we utilize Sobel’s test [56] to check the mechanisms of GTI and FC.

Table 7. The results of mechanism test.

	(1)	(2)	(3)	(4)
	DT→GTI	GTI→GTFP	DT→FC	FC→GTFP
DT	0.144 ** (0.069)		−0.210 ** (0.106)
GTI		0.007 *** (0.001)
FC				−0.038 *** (0.003)
Year	yes	yes	yes	yes
Firm	yes	yes	yes	yes
N	2070	2070	2070	2070
Adj. R2	0.760	0.968	0.760	0.968

Standard errors in parentheses ** p < 0.05, *** p < 0.01.

shows the corresponding results. Column (1) shows the regression coefficient of DT on GTI is 0.144, passing the 5% significance level, and the standard error (SE) is 0.069. Column (2) shows the regression coefficient of GTI on GTFP is 0.007, passing the 1% significance level, and the SE is 0.001. With two different coefficients and their SE, we obtain the statistics of Sobel test as 1.997, and the p-value as 0.045, demonstrating the mechanism of GTI exists. As mentioned above, DT increases the benchmark for innovative energy-efficient green technologies in the energy sector, providing effective technical support and corresponding resource conditions for GTFP. Digital technology facilitates the integration, sharing, and application of knowledge and technology across sectors, maximizing the effectiveness of green innovation resources and empowering the development of green technology [57]. Through information transmission and innovation investment, GTI creates economic value while reducing the resource consumption and environmental costs, thereby providing sufficient incentive for the promotion of GTFP [58,59,60]. Thus, hypothesis H2 is verified.

Column (3) indicates the regression coefficient of DT on FC is −0.210, which is significant at the 5% level, and SE is 0.106. Column (4) shows the regression coefficient of FC on GTFP is −0.038, passing the 1% significance level, and the SE is 0.003. With two different coefficients and their SE, we obtain the statistics of the Sobel test as 1.957, and the p-value as 0.050, demonstrating the mechanism of FC exists. Digitization reduces the degree of financing limitations by lowering the financing barrier for investors and addressing the issue of information asymmetry both inside and outside of energy enterprises [4]. Easier financing constraints have made it more accessible for energy enterprises to obtain investment from financial institutions and social investors. Sufficient funds may enhance an energy enterprise’s ability to take on financial risks and allow them to invest more in modernizing technology, machinery, production methods, and managerial expertise, all of which have a positive effect on GTFP [61]. Thus, the H3 proposed in this study is verified.

5.4. Heterogeneity Analyses

5.4.1. Nature of Property Rights

The capital market enterprises with varying ownership structures exhibit notable distinctions in their resource endowments and governance processes, and these variations are reflected in the extent of the impact of digitization on GTFP. According to Liao et al. [62],

Table 8. The results of heterogeneity test.

	(1) State-Owned	(2) Non-State-Owned	(3) Large-Scale	(4) Small-Scale	(5) New Energy	(6) Traditional Energy
DT	0.005 ** (0.002)	0.000 (0.002)	0.010 ** (0.005)	0.000 (0.002)	0.005 ** (0.002)	0.001 (0.004)
Controls	yes	yes	yes	yes	yes	yes
year	yes	yes	yes	yes	yes	yes
firm	yes	yes	yes	yes	yes	yes
N	1090	980	810	1260	1290	980
Adj. R2	0.968	0.968	0.966	0.968	0.967	0.970

Standard errors in parentheses ** p < 0.05.

’s Columns (1) and (2) display the results of the division of the sample enterprises in this study into state-owned and non-state-owned categories according to the nature of their property rights. The explanation for the limited significance of the promotional impact of DT on GTFP in state-owned enterprises is as follows: First, the state-owned enterprises have the advantage of being well-embedded in the national credibility chain. They are more likely to obtain resource favoritism and additional privileges to effectively alleviate the financial pressure in the process of technology research and development and infrastructure construction. In the context of the digital age, this is conducive to the state-owned enterprises seizing market dominance and enabling their digital change behavior to form an effective feedback loop in the process of enhancing GTFP [63]. In addition, state-owned enterprises have a strong policy orientation. They must take the lead and set an example in the process of green development, and they are subjected to increased oversight from the public and government. Therefore, when the country advocates digital strategies, state-owned enterprises are more willing to efficiently utilize digital transformation to reap environmental benefits, thus improving GTFP [64]. In contrast, non-state-owned businesses like to contend with fiercer market competition, and this provides them with a stronger motivation and willingness to improve their internal operational processes and capabilities. However, due to the greater financial pressure and imperfect management system, their intrinsic motivation for digital transformation is limited, and the promotion of GTFP may ultimately be only marginally impacted.

5.4.2. Enterprise Scale

The digital transformation resources owned by enterprises of different scales may vary, resulting in inconsistent green productivity. Based on Cheng et al. [49], the results are displayed in Columns (3) and (4) of Table 8 after the sample is split into large-scale and small-scale businesses based on the average size of the business. The findings demonstrate that, while the small-scale enterprises’ DT-to-GTFP regression coefficient is positive, it does not pass the significance test. In contrast, DT contributes better to the GTFP of large-scale enterprises, which passes the significance test at the 5% level. The potential explanations for this are as follows: Large-scale enterprises generally have a clear leading position in the industry and have sufficient capital, strong risk resistance, and high bargaining power relative to upstream and downstream enterprises [65]. Therefore, such enterprises are more conducive to the implementation of digital transformation practices. In contrast, small-scale enterprises are easily marginalized by financial institutions and face more severe financing difficulties due to their small size, short establishment period, lack of collateral assets, and difficulties in disclosing their financial status. Under the conditions of external financial constraints, small-scale enterprises have insufficient incentives to undergo digital transformation, and the quantity and quality of transformation are far inferior to those of large-scale enterprises. Furthermore, due to factors such as geographic location, development prospects, and level of innovation, small-scale enterprises are not sufficiently attractive to external investors and digital technology talents, and they do not meet sufficient conditions for digital transformation, resulting in a relatively slow digitalization process. Therefore, the impact of DT on GTFP is more evident for large-scale enterprises than for small-scale enterprises.

5.4.3. Energy Type

Since the effect of digitalization on distinct types of energy enterprises is different, it is also necessary to divide the sample into new energy and traditional energy enterprises for performing the heterogeneity analysis. China’s energy industry categorization standard divides energy production into two categories: conventional energy, which comprises thermal power, coal mining and processing, and natural gas and oil extraction, and the wind, solar, and biomass subsectors [4]. Table 8’s Columns (5) and (6) present the results. It is evident from these results that, although DT has a minor positive influence on GTFP in traditional energy enterprises, it has a significant positive impact in new energy enterprises. The underlying explanation is that new energy enterprises are more concerned with environmental benefits. New energy sources are characterized by the fact that they do not produce additional pollution or waste like fossil fuels, which is key in solving environmental problems. Digital technology optimizes the efficient production, dispatch, and consumption of new energy, fosters the intelligence of the energy industry chain, and assists in maximizing the effectiveness of green output [66]. Conversely, traditional energy enterprises primarily apply digital technologies to meet the energy supply and create economic benefits, exacerbating energy consumption or only increasing the total factor productivity rather than GTFP [67].

6. Conclusions and Policy Implications

This study empirically examines the relationship between DT and GTFP based on the data of A-share-listed energy enterprises from 2013 to 2022. The path to improving the effectiveness of digital transformation is deciphered in this study. In addition, by analyzing the heterogeneity, this study provides practical insights from different perspectives for enterprises to implement digital transformation strategies. Contrary to the previous works of literature, the fresh results of our paper are reflected in the following three aspects. First, some studies demonstrate DT hinders the improvement of GTFP [19]. Our benchmark regression shows that the enhancement in GTFP is facilitated by energy enterprises implementing digital transformation strategies. Second, some works of literature show that GTI and FC are not important mechanisms [49,68]. Third, some cases do not find significant heterogeneity in enterprise size and other fields [69,70]. Our results demonstrate that state-owned enterprises, large-scale enterprises, and new energy enterprises are more capable of taking full advantage of digitization to improve their GTFP.

Drawing on the empirical analysis, three conclusions are presented. First, the DT of energy enterprises has a major beneficial impact on GTFP. This finding has been validated using a variety of robustness tests. Additionally, the decomposition of GTFP reveals that DT is the main driving force of GTC, implying that the implementation of digital strategies is an advantageous approach for resolving the disagreement between environmental preservation and economic growth. Second, DT can improve GTFC by upgrading GTI or relieving FC. Finally, the effect of DT on GTFP is heterogeneous. The powerful resource capacity of state-owned enterprises, the sufficient capital of large-scale enterprises, and the greater emphasis on environmental benefits by new energy enterprises are possible explanations for the differentiation. Based on the results, this study put forward several suggestions. First, governments should strongly support the development of digital technologies. Second, distinct strategies for digital transformation should be implemented by energy enterprises according to their actual growth circumstances. Finally, investors should enhance communication with energy enterprises to reduce information asymmetries.

In light of the aforementioned findings, several policy suggestions are made. First, governments should strongly support the development of digital technologies. Relevant departments should strengthen the policy and institutional framework to encourage energy enterprises to carry out digital technology innovation and achieve digital transformation as soon as possible. Furthermore, to fully achieve the potential of data elements, the government should also invest more resources in digital technology research and development. Accelerating the development of digital infrastructure, including 5G mobile networks and blockchain, is a preferable starting point for transforming data into a new key production factor, thus providing a strong impetus for green development. Second, distinct strategies for digital transformation should be implemented by energy enterprises according to their actual growth circumstances. For example, non-state-owned and traditional energy enterprises could use digital transformation to achieve the redevelopment of basic elements to compensate for the lack of green resources. Small-scale enterprises can focus more on maximizing the green value of digitization by leveraging digital technologies to add value to their existing products and technologies. Briefly, enterprises should consider their own nature and development direction to formulate targeted transfer paths. Finally, the behavior of investors cannot be ignored. Investors are encouraged to take an active interest in and enhance communication with energy enterprises to reduce information asymmetries between them and identify enterprises with investment potential. Meanwhile, investors can increase their investments moderately, which allows enterprises to reduce their operating costs and have sufficient funds for digitization and green development.

Although some preliminary conclusions can be drawn from this study, several limitations remain that present opportunities for future research. These limitations include the fact that some newly established enterprises are not incorporated in this study and that enterprise-level energy consumption data are difficult to collect. Future research can add newly established enterprises to the sample to increase the sample size. Furthermore, new measurement indicators can be constructed to measure energy consumption at the enterprise level.