The Mutual Relationships Between ESG, Total Factor Productivity (TFP), and Energy Efficiency (EE) for Chinese Listed Firms

Abstract

This study examines the mutual relationships among ESG performance, total factor productivity (TFP), and energy efficiency (EE) in a sample of Chinese A-share listed firms from 2010 to 2022. This study shows that ESG has a significant promotional effect on TFP. Reducing financing constraints and inefficient investment are among the mediating mechanisms, and the latter plays a greater role. Heterogeneity analyses suggest that state-owned enterprises (SOEs) and heavy-polluting enterprises (HPEs) should be consistently committed to ESG responsibility fulfillment. Formal environmental regulation (FER) can be complementary to ESG, but informal environmental regulation (IER) has the opposite effect. TFP was instead suppressed by the triple combined effect of ESG with these two. The results of the threshold effects of ESG and EE indicate that the positive impact on EE becomes more pronounced as ESG performance improves. However, ESG performance varies across subdimensions. As green technology research and development efficiency (GRDE) and green technology transformation efficiency (GTTE) improve, stronger ESG promotes EE. This threshold effect also exhibits heterogeneity with respect to the ownership structure. Moreover, there is bidirectional causality between EE and TFP, and EE has a stronger positive effect on TFP. These findings reveal the optimal paths and potential risks for moving toward sustainability for firms.

1. Introduction

The Environment, Social, and Governance (ESG) concept emphasizes that firms should focus on environmental protection, social responsibility and improved governance. Unlike the traditional assessment of a company’s financial performance, ESG is an investment philosophy that focuses on a company’s environmental, social, and corporate governance performance, as well as indicators and frameworks for assessing a company’s sustainable development [1]. With such a development trend, it has become a mainstream trend for companies to integrate ESG performance into their core business in response to contemporary eco-economic development and environmental protection. On the one hand, firms that prioritize ESG practices often find themselves at an advantage, as reflected in their ability to attract significant investment and secure lower costs, while also mitigating potential environmental and social risks in their operations. On the other hand, firms engage in ESG practices with the fundamental aim of being sustainable in terms of economic and environmental benefits, and do not want the challenges of economic development such as additional costs attached to them [1]. Therefore, under the sustainable development goal of achieving both economic and environmental benefits, companies attempting to implement ESG strategies are inevitably subjected to the dual development pressures of improving productivity and environmental performance, which also emphasizes the mutually beneficial outcome of economic growth and environmental sustainability. Given this reality, the real growth drivers for businesses cannot bypass total factor productivity (TFP) and energy efficiency (EE), especially for emerging economies. The concept of TFP represents the contribution of all factors other than capital and labor factor growth to output growth or economic growth [2]. In addition, TFP also represents technological change at the productivity frontier of firms, and this technology-driven production activity cannot be separated from the low-carbon transition in energy, and the two are mutually reinforcing [3]. Increasing TFP for economic growth [3] and improving EE for cost efficiency [4] have become powerful strategies that companies have adopted to achieve sustainable development.

The core of energy policy discussions often revolves around energy efficiency [5]. For enterprises, focusing on EE as a major factor in sustainable development is more likely to achieve both emission reduction effects and economic benefits [6]. Moreover, the productive activities of enterprises, as a signal of the final effects of the implementation of various policies, are always challenging due to a variety of factors, including the promotion of a transition process that triggers changes in the field of energy technology in addition to other changes [7]. Whether the process of changes in the energy market leads to significant economic, environmental, and social benefits depends largely on three important factors: the environmental, social, and governance practices of enterprises [8]. A study demonstrated that there is a nonlinear relationship between the impact of ESG sub-dimensions on energy transition, and the perspective of this study includes Sweden, Switzerland, and Norway, as well as three emerging economies, India, South Africa, and Cambodia, with both economic and social factors emerging as key variables influencing the nonlinear relationship [8]. Such a discussion suggests that even though energy-related performance factors are included in the environmental performance sub-dimension of ESG, the relationship between them and ESG may not be linear, especially for emerging economies, and such a relationship is yet to be explored due to the different processes of energy transition.

Although scholars have conducted studies related to ESG, including market value [9,10,11], cost of debt [12,13,14,15], innovation capacity [16], CEO compensation [17,18], industry sector development [19], and others, the linkages among ESG, TFP, and EE are still relatively underexplored. Therefore, our study expands the understanding of the relevant components of ESG practices in an attempt to elucidate the relationship between ESG, TFP, and EE. Under the sustainable development goal of achieving win–win economic and environmental benefits, enterprises attempt to carry out ESG strategy implementation, bearing the dual development pressure of improving their TFP and EE. Therefore, our study attempts to clarify the relationship between them. First, the real relationship between ESG performance and TFP, which is an important driver of firm value realization and long-term development, and its heterogeneous characteristics, as well as the intrinsic mechanism of action between the two, need to be explored. In addition, we focus on external moderators such as heterogeneous environmental regulations, explore whether different environmental regulations have different effects on the relationship between the two, and try to explore the mechanism of the joint moderating effect. Second, although studies have demonstrated a nonlinear relationship between ESG performance and energy transition, they have not focused on the specific factor of EE. In addition, for China, one of the most competitive economies among emerging economies, the actual relationship between firm-level ESG performance and energy efficiency remains unclear, and the strength of the role of each sub-dimension on EE and the existence of significant heterogeneity characteristics need to be analyzed in depth. Third, the relationship between TFP and EE is the last element of our study, emphasizing exploring whether there is a bidirectional causality between the two in order to further broaden the research perspective and to complement the core of the full paper. Our study focuses on listed companies in China due to the fact that the path of ESG practices of Chinese companies is aligned with the country’s overall goal of high-quality economic development, and such alignment will introduce the attention of companies and a wider range of stakeholders. In addition, as the world’s largest developing and emerging economy, research around ESG, economic development, and energy will provide lessons for other developing countries on the path of ESG practices. From the perspective of data analysis, the large number of listed companies in China and the rapid development of ESG rating practices in recent years provide rich data resources for our empirical research. On the basis of clear theoretical relationships, we will provide practical guidance on the best path for Chinese companies to achieve future sustainable economic development goals in order to ultimately realize the goal of green growth.

The remainder of this paper is organized as follows: Section 2 presents a review and summary of the existing literature, along with the formulation of the research hypotheses. Section 3 offers an overview of the research methodology, describes the variables used, and provides details about the data employed. Section 4 provides the specific research results and accompanying discussion. Finally, Section 5 summarizes the paper and offers recommendations derived from the findings.

2. Literature Review and Theoretical Hypotheses

2.1. The Relationship Between ESG and TFP

TFP is the additional unpredictable productivity achieved by a firm under certain conditions of factor input [20]. The ESG framework represents an effective methodology for assessing the economic growth and environmental sustainability of firms and consists of three main dimensional components [21]. Among them, the environmental performance dimension focuses on whether firms mitigate climate change and improve resource efficiency through energy efficiency, environmental pollution reduction initiatives, etc. [22], and such specific energy-related practices will affect the improvement of TFP [23]. The contribution of social responsibility is manifested through improvements in employee satisfaction, including employee incentive mechanisms and attracting new talent; the optimization of costs and efficiency in supply chain management; risk control; and quality advantages and customer feedback in product management, all of which have positive effects on promoting TFP [24,25]. For corporate governance, ESG positively impacts TFP by increasing disclosure [26], loosening freedom of equity ownership, and foreign ownership [27]. As the first aspect of this study, we establish a hypothetical mechanism to lay the foundation for the following content of the study. Here, we expect the following:

H1. 

ESG performance positively affects TFP.

ESG performance has become an effective way to assess the long-term competitiveness of firms, which is reflected in influencing market reputation [25] as well as financial performance [28,29,30,31]. In terms of the long-term goal of sustainability, ESG performance can improve firms’ TFP by alleviating financing constraints [32,33]. The emergence of financing constraints depends on how well a firm is performing in the long run, and it can lead to negative impacts such as the inability of enterprises to cover the costs of infrastructure [34], the costs of technology to develop new products [35], and negative effects such as limiting the scale of production [36]. According to the information asymmetry theory, enterprises with good ESG performance can effectively reduce the information asymmetry between enterprises and investors by disclosing detailed ESG information, thus enabling investors to more comprehensively understand the enterprise’s operating conditions, risk level, and development prospects, enabling financial institutions and investors to have the ability to avoid risks, reducing investor compensation, and alleviating corporate financing constraints [37]. From the perspective of internal operations, companies with good ESG performance can significantly increase TFP by raising internal labor costs, attracting and retaining high-quality talent, optimizing the allocation of human resources, and reducing agency costs [38].

The capital market has gradually increased its focus on green finance and a low-carbon economy, and such investments tend to have ESG investment preferences [39]. Coupled with media attention and public scrutiny, ESG practices have also become a direct channel for companies to improve their reputation. According to reputation theory, excellent ESG performance will help to enhance the recognition and credibility of the firm in the competitive market, which has established a good image through positive actions in environmental protection, social responsibility fulfillment, product quality, and good corporate governance. As information asymmetry is reduced and market reputation is enhanced, investors in the market are willing to offer more favorable financing terms [25]. As a result, firms with good ESG performance can obtain capital at a lower cost and maintain a stable supply of capital, further improving the efficiency of capital utilization so that it can be invested in technological innovation, equipment upgrades, and so on, thereby increasing TFP.

Second, ESG performance can increase TFP by reducing inefficient investment behavior as a way to improve firms’ competitiveness. There are two competing views in the research on the impact of ESG on investment efficiency. First, enterprises engage in ESG strategies that lead to overinvestment [40]. This is due to the fact that the management of a firm may make inefficient investments for its own benefit, such as the emergence of behaviors such as gaining more control gains for the purpose of expanding the size of the firm profitably. In addition, the trade-off hypothesis states that CSR investment reduces the firm’s resources, affects its long-term competitiveness, and adversely affects its development [40]. However, as we mentioned in the previous section, with the role of mediator intermediaries, firms with good ESG performance tend to have a good reputation, which can cushion pressures from external sources and reduce agency costs [39]. Meanwhile, when firms are often affected by information asymmetry between internal and external markets, which makes them lack effective information judgment and be less sensitive to investment risks, it can lead to decision-making errors such as managerial shortsightedness, which can be mitigated by engaging in ESG practices [41,42,43]. This relationship can also be explained by the stakeholder theory; firms with good ESG performance are usually characterized by better social responsibility fulfillment and greater improvement in relationships with stakeholders and the consideration of their relevant interests and expectations [41,42]. Meanwhile, the conflict resolution hypothesis theory also clarifies that investment efficiency tends to improve when communication relationships between important stakeholders are effectively improved and the level of trust in each other increases, and ESG practices assume this role. When customers, employees, suppliers, and regulators all reasonably communicate performance in ESG practices, it limits or even prevents possible self-interested investment behavior by management, thus reducing inefficient investment behavior [43]. The fulfillment of ESG responsibilities enables corporate management to identify projects with long-term value and strategic significance, and therefore investment efficiency is improved, and the sensitivity of investment cash flows is reduced, thus increasing the efficiency of resource allocation and driving the increase in TFP. This positive relationship has been confirmed in recent years [41,42,43,44,45,46].

To explore the mechanism of ESG and TFP, the above three aspects cannot be ignored. Consistent with the views of previous studies, we propose Hypothesis 1a and Hypothesis 1b, which lay the foundation for the logic of the whole article, as follows:

H1a. 

ESG performance can promote TFP by alleviating financing constraints.

H1b. 

ESG performance can promote TFP by reducing inefficient investment.

In exploring the mechanism of ESG and TFP, it is often affected by many external factors; thus, the relationship between the two is adjusted to different degrees, especially as it relates to environmental regulation. From an internal point of view, environmental responsibility is a necessary condition for the future development of enterprises, and from an external point of view, such environmental regulation policies also become an external regulatory factor. In China, formal environmental regulation (FER) generally derives from the government’s mandatory and incentive-based environmental policies, and informal environmental regulation (IER) depends on the public’s environmental awareness [47]. However, the roles of the two types of ERs in the relationship between ESG performance and TFP are not always consistent. Scholars who follow the cost hypothesis of the crowding out effect hold the opposite view, which can lead to the inhibition of production factors due to high costs [48]. Neoclassical economists contend that the enforcement of stringent ER increases costs and diminishes accessible productive resources, ultimately impeding economic progress [49,50]. However, scholars who follow the innovation compensation effect in Porter’s hypothesis argue that in regions where environmental impact assessments are strictly implemented, listed companies will pay more attention to investments in pollution control and incentivize producers to make technological innovations and other modifications to production factors to compensate for environmental costs [51,52,53,54,55]. Moreover, the ESG ratings obtained annually from third-party organizations become an external environmental regulatory factor that companies must fulfill to be monitored by stakeholders, thus increasing their costs and fulfilling their ESG responsibilities. Pollution taxes, as a type of environmental regulation, have been found to be positively associated with environmental investments made by small and medium-sized enterprises (SMEs) [53]. FER echoes the ESG concept’s requirements for environmental performance, and together they push companies to fulfill their environmental protection responsibilities while promoting TFP. Therefore, ESG and this external monitoring mechanism as FER complement each other [52]. With the increasing popularity of the Internet, online searches and discussions about environmental assessment practices generated by media reporting channels are used by the public to monitor a variety of firms’ behaviors as another type of environmental regulation [56,57,58]. On the one hand, the public tends to be highly sensitive to negative news, and firms adopt short-term behaviors to maintain their image, leading to poor ESG practices, thus weakening the relationship between ESG performance and TFP. On the other hand, firms will bet more resources and funds on resolving negative public opinion, thus increasing the cost of non-production and having a dampening effect on TFP.

Nevertheless, firms are often subject to the dual effects of both FER and IER in the long-term process of fulfilling ESG practices. One study suggests that the joint effect of FER and IER will have a policy overlay effect on TFP improvement [59]. However, there is still a gap in research on whether the triple joint effect of ESG, FER, and IER can still promote TFP improvement. On the one hand, according to signaling theory, FER can be regarded as an environmental signal, and the strength of this signal is reflected through the specific requirements of local government documents [60]. At the same time, this environmental signal is transmitted to other stakeholders including the media and the public. It is also in the process of signaling that the effectiveness of IER’s media reputation mechanism is strengthened. As a result, when firms actively engage in ESG practices, they are constrained by the combined effect of these two types of environmental regulations, which makes them bear more governance costs to avoid environmental penalties and negative public opinion, and thus also makes them “greenwash” [52,61], which affects the increase in TFP. This affects the improvement of TFP. At the same time, according to the theory of firm behavior, firms in the same industry or the same region tend to follow the same behavior under the pressure of cost and regulation, and the continuation of low-carbon investment will lead to the effective expectation of the stakeholders to the firms higher than the actual range of the firms [62], and the increase in TFP will be inhibited. Therefore, it is urgent to explore whether the relationship between ESG practices and TFP changes under the combined influence of FER and IER. Based on the analysis of these theoretical mechanisms, we first consider FER and IER as external moderators from the market to examine whether they moderate the relationship between ESG performance and TFP and formulate Hypotheses 2a and 2b. Then, we try to integrate the joint moderating effect of FER and IER and formulate Hypothesis 2c. All the hypotheses are as follows:

H2a. 

FER positively modulates the promotional effect of ESG performance on TFP.

H2b. 

IER negatively modulates the promotional effect of ESG performance on TFP.

H2c. 

ESG, FER, and IER interact to consistently increase TFP, which means that when ESG, FER, and IER act in combination, they weaken the facilitating effect of ESG performance on TFP.

2.2. The Relationship Between ESG and EE

Environmental performance as measured by ESG performance has become an effective method to measure the efficiency of corporate energy use, and this rule has been recognized internationally [63,64]. Scholars have studied the relationship between green taxes, energy efficiency, and ESG performance in manufacturing industries around developing countries, such as Bangladesh [21]. They found that EE, as a key driver of sustainable practices, positively affects ESG performance and that EE has a distinctive role in influencing the three sub-dimensions of ESG performance. Although this facilitating effect is significant, because this research targets a specific industry and Bangladesh itself has unique geographical characteristics, many factors may limit the generalization of these findings. Another study centered around green businesses in Russia and Central Asia found that firms usually undertake the creation of projects related to the improvement of energy efficiency in the fulfillment of ESG responsibilities, and this behavior is also interconnected with the fulfillment of social responsibility, identifying that the environmental component of ESG performance contributes to EE [64]. The linkage of ESG practices in major developed economies with their economic strategy factors tends to be effective in improving EE [63]. It has still been shown that there is a threshold effect between ESG and energy transition [8], and that energy efficiency is an important driver of energy transition [65]. Although energy efficiency is one aspect of environmental performance in ESG performance, the formation of this threshold effect is often closely related to the three components of ESG, that is, the three components of ESG also influence the effect to varying degrees [8]. If the nature of the relationship between a particular two is nonlinear in nature, then the study of linear approaches may produce mixed and inclusive findings [66]. For Chinese listed companies that are currently facing the task of energy transition and at the same time actively fulfilling ESG practices, although sustainable economic development can effectively contribute to the improvement of EE [67], when companies face environmental pressures, the rules related to ESG will push them to prefer decoupling from the environment [68].

In fact, there are many references in the literature to the nonlinear relationship between ESG and other variables, such as ESG performance and firm efficiency [69], energy transition [8], economic uncertainty [70], debt structure [66], sustainability reporting [71], and outward foreign direct investment [72]. However, studies addressing the possible nonlinear relationship between ESG and EE are not very extensive.

In terms of environmental performance under the ESG framework, firms with better ESG performance may be incentivized to focus on end-of-pipe pollutant management, thereby neglecting front-end EE improvement, which may ultimately lead to the decoupling of environmental performance. According to resource base theory, in the early stage of ESG practice, firms commit to behaviors such as improving energy management systems and adopting energy-saving technologies to significantly improve EE, but as cost pressures increase, the marginal benefit of resources diminishes, and after a certain threshold is exceeded, the effect of additional resource inputs on EE improvement gradually diminishes. Meanwhile, the theory of ecological modernization suggests that as environmental performance increases to a higher level, enterprises will face technical bottlenecks, market constraints, and other problems. Under the influence of such factors, the relationship between ESG and EE is characterized by stages, thus leading to the emergence of a nonlinear relationship.

Different factors in social responsibility have different roles in influencing EE [73,74]. As shown in the study, the relationship between energy-adjusted EE and ESG performance is characterized by a nonlinear U-shape, which exists only in social responsibility performance and corporate governance performance [69]. According to stakeholder theory, different stakeholders have different concerns at each stage, which change with market conditions. When stakeholders are more concerned with short-term profits, firms may focus less on energy use efficiency improvement and devote themselves to profit-making behaviors, and even though ESG performance is still maintained, EE improvement may no longer be evident. As shown in the study, the relationship between energy-adjusted EE and ESG performance has a nonlinear U-shaped characteristic that exists only in social responsibility performance and corporate governance performance but not in environmental performance [69].

Good performance at the corporate governance level plays a key role in achieving EE and sustainable economic growth [63]. Relying on principal–agent theory, at the early stage of corporate governance improvement, rational board structure, effective monitoring mechanism, etc., can motivate the management to actively promote energy management improvement and enhance EE. However, when corporate governance reaches a high level, the marginal impact of further improvement of the governance structure on the improvement of energy efficiency may become smaller, because at this point, the management decisions have already been relatively scientific and reasonable, and the room for improvement is limited. Therefore, the ESG performance of firms has a nonlinear relationship with EE, and the tripartite performance score of ESG performance also has a nonlinear relationship with EE, which needs to be further explored. Now, we assume and verify it in this article with the following hypotheses:

H3. 

There is a threshold effect between ESG performance and EE, and enterprises can improve EE through ESG performance.

H3a. 

The effects of the three ESG components on EE vary and exhibit a nonlinear relationship.

If we consider the nonlinear threshold effects of ESG and EE, we need to consider what factors have this effect on them. Green technology innovation refers to the advanced technology of using clean energy and alternative fuels in the production process, which has a direct effect on energy efficiency [75]. Proponents of the ‘productivity paradox’ theory argue that a portion of green technological innovation may inhibit productivity gains because of the inability to harmonize the dynamic linkages of factors of production instead [76]. However, other scholars use the environmental Kuznets curve (EKC) as a benchmark, arguing that green technological innovation can always improve the efficiency of energy utilization by integrating resources through new technologies, processes, and materials, which ultimately has a positive effect on the economy [77]. A large body of literature confirms that the role of green technology innovation is particularly critical for environmental, social, and corporate governance performance [78,79,80]. However, most of the aforementioned measures of green technological innovation focus on the number of patented innovations, but a dramatic increase in technology patents does not directly represent improved technological innovation [81]. There is a lag effect between green technology R&D and enterprise value, as it is often characterized by long lead times and high costs [82], and there is complexity in implementing green energy technologies nationwide [83]. In 2022, the Chinese government issued a directive concerning the production process of green technological innovation, stating that green technological innovation consists of two phases in the production process, namely, patent research and development and the active transformation of results. However, there is often a time lag between the research and development stage and the transformation into productivity [84]. The issue of green technology research and development and the transformation of research results into real productivity at the microenterprise level in China often reflects the output-to-input ratio of technological innovation activities under resource and environmental constraints; therefore, to measure the efficiency of green innovation from the perspective of inputs and outputs [85,86], the stochastic frontier analysis (SFA) method [87] is a better approach. Moreover, the data envelopment analysis (DEA) method [88], slacks-based measure of efficiency–DEA (SBM-DEA) model [89,90,91], and CDM model [92] are also commonly used in studies measuring the efficiency of green technological innovation. In view of this, this paper employs the SBM-DEA model to explore the nonlinear relationship between ESG and EE using the green technology research and development efficiency (GRDE) and green technology transformation efficiency (GTTE) variables as critical variables [85,90]. Therefore, we continue with the following hypotheses:

H3b. 

The effect of the GRDE on ESG to EE shows a nonlinear relationship.

H3c. 

The effect of the GTTE on ESG to EE shows a nonlinear relationship.

2.3. The Relationship Between EE and TFP

For listed companies, identifying the sources of economic growth drivers is critical for examining the possible relationship between the environment and the economy. Because TFP is usually considered an explicit result of technological change, economic and technological change often have an inevitable connection with the efficiency of energy conversion. The national research literature surrounding this issue provides us with empirical experience [3,93,94]. Energy efficiency in Portugal is the unit driver of TFP growth, but there is a rebound effect between the two. Specifically, efficiency gains increase the output of the economy, whereas higher levels of economic output, in turn, require additional inputs of human labor as well as energy, thus increasing the efficiency of energy use [3]. This bidirectional relationship is reflected in the study of the energy–economy nexus at the national level and is applicable to the development strategies of other countries. First, energy efficiency curbs emissions of pollutants such as carbon dioxide [95]. If TFP is limited, then the energy factor will also be unstoppably limited, leading to the emergence of environmental pollution problems, which in turn affects the growth of TFP, even if this effect has a long-term relationship. Moreover, if the energy conversion efficiency is enhanced throughout the process of transportation to final use, the long-term growth effect on the economy will also be affected. Even an increase in energy efficiency is a key driver of an increase in TFP [94]. On the basis of the findings of the existing literature, exploring the relationship between the two at the level of Chinese micro firms as a way to enrich the research results in this area is not a novel perspective. Therefore, we propose the following hypothesis:

H4. 

EE has a bidirectional causal effect on TFP.

The research hypothesis framework of this study is briefly outlined in Figure 1, and this framework also basically summarizes the research idea of this paper. By analyzing the relationships among ESG and TFP, ESG and EE, and TFP and EE, we aim to provide evidence of the role of ESG performance in the sustainable development of Chinese firms from a multidimensional perspective and to offer robust theoretical support for the future practice of relevant government policymakers and business participants under the goals of firm sustainable development and a low-carbon transition.

3. Materials and Methods

3.1. Data Sources

We selected data from Chinese A-share listed companies between 2010 and 2022. The ESG practices of Chinese firms show a growing trend in the global ESG field [16,45]. In terms of time span, the sample covers the stages of China’s economic restructuring, industrial upgrading, and environmental policy strengthening. During this period, along with the continuous changes in China’s economic policies and financial markets, the introduction of a series of environmental laws and regulations, and the widespread use of the Internet [57], firms faced a variety of types of environmental regulations [47]. With the implementation of the new accounting standards in 2007, Chinese listed companies were required to prepare consolidated and parent company financial statements [96], and the active internal capital market provided rich financial data for studying the relationship between ESG and TFP. In addition, there are obvious differences in the nature of property rights of Chinese enterprises, with state-owned enterprises (SOEs) playing an important role in the piloting of major projects, and the financing environment is different from that of non-SOEs [96]. In terms of energy factors, China has a high share of coal consumption and faces challenges such as the clean utilization of coal, so companies are subject to both economic and environmental pressures in ESG practices. However, China’s relatively large market size, industrial agglomeration effect, scale effect and technology spillover effect are significant, which is also conducive to the study of the threshold and nonlinear relationship between ESG and EE. As many emerging economies are similar to China in terms of economic development, China’s experience in balancing development and ESG practices, and in improving TFP and EE, can help them formulate sustainable development strategies and provide new ideas and examples for the global exploration of the relationship between the three.

We applied the following criteria at the sample selection stage: First, we chose to exclude financial firms listed on stock exchanges due to the fact that they have different rules for financial reporting than firms in other industries, such as balance sheet characteristics that are significantly different from those of firms in other industries, as well as significant differences in terms of the regulatory environments to which they are subjected, and service efficiency standards that are conceptually different from those of firms [97], we chose to exclude these firms listed in the financial industry. Secondly, firms labeled as ST, *ST have anomalous nature in their observed operational data and hence are also excluded. In addition, to prevent the effect of outliers, we shrank the continuous variables at the 1% level. With these screening criteria, we ended up with a sample of 6095 valid observations. We used ESG rating data from the Bloomberg database [97,98]. The financial data of enterprises and the data related to pollution emissions and energy consumption were obtained from the Cathay Pacific China Stock Market & Accounting Research (CSMAR) database, the annual reports of each firm, and the annual ESG disclosure reports of each firm. The environmental regulation data are derived from the National Bureau of Statistics of China, and the IER data are derived from the Baidu Index platform. The green patent data of each firm come from the China Research Data Service Platform (CNRDS) database.

3.2. Variable Selection

3.2.1. Dependent Variables

For the relevance measure of TFP, we used the Olley–Pakes method (OP) to measure it and as the basis for the main regression analysis [20]. The OP method is based on the method of consistent semiparametric estimators. The main idea of the method is to find consistent control equations from firms’ input decisions. It is based on a dynamic panel data model, which fully considers the decision-making behavior of enterprises in different periods and the impact of changes in market dynamics on firms and is able to capture the heterogeneous characteristics of firms in different periods. In the measurement, we first used the perpetual inventory method to determine the capital stock of the firm as the capital input, the total number of employees of the firm in the year as the labor input variable, and the business revenue of the firm as the total output variable. At the same time, in order to ensure the accuracy and reliability of the measurement, we also used the Levinsohn–Petrin method (LP) to measure total factor productivity and applied it to the robustness test of the relevant empirical findings, which utilizes the intermediate inputs as the proxy variables and complements the OP method that utilizes the investment as the proxy variable [23,99]. In the measurement, the method involves intermediate goods inputs as cash paid to employees and the difference between operating costs and depreciation and labor compensation used to measure intermediate input variables.

Energy efficiency (EE) is an important aspect of sustainable development [21,75]. In all definitions, EE usually means using less energy to produce the same or more output, and energy productivity or energy intensity is usually considered as a proxy for energy efficiency in practice [73]. It has been found that China’s past reductions in energy intensity have been mainly attributed to improvements in energy efficiency [100]. With this background, in measuring EE, we used the proportion of total energy consumption to the operating revenue of the firm to measure the EE of the enterprises [75,95,100]. Here, the coal usage, natural gas usage, gasoline usage, and diesel usage of the listed companies were calculated as the energy consumption index of the listed companies on the basis of the energy conversion coefficients (uniformly converted to standard coal) and summed to obtain the total energy consumption. The conversion coefficients of standard coal for various fuels are from the Chinese Energy Statistical Yearbook.

3.2.2. Core Variable

We utilized ESG score data from Bloomberg to measure the ESG performance of Chinese listed enterprises. Bloomberg scores the ESG disclosures of Chinese listed companies, including the summed disclosure scores of environment, social responsibility, and corporate governance, to measure the ESG performance of Chinese listed enterprises; the higher the score is, the better the ESG performance of the enterprises, which has been widely noted by the capital market [21,44].

3.2.3. Mediator Variables

On the basis of the theoretical analysis, we selected the Kaplan and Zingales index (KZ) [101] to measure financing constraints [25,39]. The degree of inefficient investment (NEI) encompasses overinvestment and underinvestment [41], and its measurement is obtained through the following Equation (1). Specifically, the regression residuals ${\epsilon }_{i,t}$ obtained from estimation based on this equation are used to measure the deviation of firms’ investment efficiency [102]. That is, when the regression residuals are positive, the investment is considered to be overinvested; when the regression residuals are negative, the investment is considered to be underinvested. We used the absolute value of the regression residuals to measure the degree of inefficient investment [41,102], as below:

$$\begin{array}{l}{\mathrm{Invest}}_{\mathrm{i},\mathrm{t}}={\beta }_0+{\beta }_1{\mathrm{Growth}}_{\mathrm{i},\mathrm{t}-1}+{\beta }_2{\mathrm{Size}}_{\mathrm{i},\mathrm{t}-1}+{\beta }_3{\mathrm{Lev}}_{\mathrm{i},\mathrm{t}-1}+{\beta }_4{\mathrm{Cash}}_{\mathrm{i},\mathrm{t}-1}+{\beta }_5{\mathrm{Age}}_{\mathrm{i},\mathrm{t}-1}\\ +{\beta }_6{\mathrm{Ret}}_{\mathrm{i},\mathrm{t}-1}+{\beta }_7{\mathrm{Invest}}_{\mathrm{i},\mathrm{t}-1}+{\mathrm{Ind}}_{\mathrm{i}}+{\mathrm{Year}}_{\mathrm{t}}+{\epsilon }_{\mathrm{i},\mathrm{t}}\end{array}$$

(1)

where Invest is the amount of capital invested in the firm, which is the ratio of cash paid for the purchase of fixed assets, intangible assets, and other long-term assets minus net cash recovered from the disposal of fixed assets, intangible assets, and other long-term assets to total assets at the beginning of the period; Cash is the ratio of the total cash flow generated from operating activities to total assets to measure the cash level; and Ret is the annual individual stock return considering the reinvestment of cash dividends to measure the annual stock return. The industry factor (Ind) and time factor (Year) are also fixed.

3.2.4. Moderator Variables

Starting from the study of the external factors affecting the relationship between ESG and TFP, we selected the proportion of completed industrial pollution control investment in the secondary industry of the province where the listed firm’s registered address belongs to measure the formal environmental regulation intensity (FER) as the moderating variable [103]. In addition, we searched for “pollution” as a keyword in the Baidu search index to obtain the public environmental concern as informal environmental regulation (IER) by year and by region [56].

3.2.5. Threshold Variables

In this study, three pillar scores of ESG performance were selected: the E score, S score, and G score, which are considered drivers of EE. Technological innovation in the production process is divided into the technology development stage and the achievement transformation stage. We calculated the efficiency of both stages using the SBM-DEA model and obtained the green technology development efficiency (GRDE) and green technology achievement transformation efficiency (GTTE) as threshold variables [85,90]. Specifically, in the green technology R&D stage, the number of R&D personnel and R&D expenditures were selected as the initial inputs, and the intermediate output indicators were selected from the enterprise’s green patent data, including the number of green patent applications and the number of green patent authorizations. In the transformation and application stage, the intermediate outputs are the input indicators of this stage, whereas the business income, pollution emission index (the pollution data of firms include the calculation of chemical oxygen demand, ammonia nitrogen emissions, total nitrogen, total phosphorus, sulfur dioxide, nitrogen oxides, and soot, and the pollution emission index is calculated according to the entropy method after unification of units) and energy consumption index (the energy consumption data of firms includes the accounting of water consumption, electricity consumption, coal consumption, natural gas consumption, gasoline consumption, diesel consumption, central heating, and calculated as the energy consumption index of listed companies according to the energy conversion coefficient (uniformly converted into standard coal)) of the enterprise are the final output indicators.

3.2.6. Control Variables

To explore the relationship between ESG and TFP with more robust findings, we selected a series of control variables considering other factors that may affect enterprises’ TFP, including the capita asset size (Size), ownership concentration (OC), liability-to-asset ratio (Lev), return on assets (Roa), fixed asset ratio (Far), operating revenue (Rev), and growth potential (Growth). These control variables are widely recognized as having a nonnegligible effect on TFP [25,39]. In the subsequent robustness tests, we also add relevant control variables that may be neglected, including firm age (Age), total asset turnover (TAT), the proportion of independent directors (Indep), the level of enterprises’ wages (Salary), and Tobin’s Q (Tobin Q).

Table 1. Descriptive statistics of the major variables.

Variables	N	Mean	Median	Min	Max	SD
TFP	6095	1.940	1.939	1.638	2.220	0.122
EE	6095	0.005	0.002	0.000	0.042	0.007
ESG	6095	0.293	0.287	0.141	0.547	0.083
E score	6095	0.129	0.086	0.004	0.600	0.142
S score	6095	0.145	0.128	0.037	0.369	0.069
G score	6095	0.651	0.693	0.320	0.840	0.136
GRDE	6095	0.554	0.535	0.178	0.987	0.195
GTTE	6095	0.555	0.536	0.177	0.988	0.196
KZ	6095	−1.064	−1.066	−1.242	−0.906	0.065
NEI	6095	0.206	0.144	0.002	1.589	0.236
FER	6095	0.183	0.144	0.009	0.690	0.143
IER	6095	249.397	175.773	7.085	1273.994	264.447
Size	6095	23.230	23.180	20.576	26.652	1.210
OC	6095	0.371	0.357	0.082	0.764	0.159
Lev	6095	0.490	0.500	0.084	0.866	0.187
Roa	6095	0.046	0.039	−0.153	0.234	0.058
Far	6095	0.240	0.202	0.002	0.748	0.177
Rev	6095	22.603	22.570	19.642	26.233	1.356
Growth	6095	0.154	0.109	−0.466	1.833	0.322

shows the descriptive statistics of the variables. The sample includes observations from 6095 Chinese listed enterprises. The average TFP is 1.940, the standard deviation is 0.122, the maximum value is 2.220, and the minimum value is 1.638. EE has an average of 0.005, and the standard deviation is 0.007, with a small difference between the maximum and minimum values. For the ESG performance score, the average is 0.293, the standard deviation is 0.083, the maximum value is 0.547, and the minimum value is 0.141, which implies that the ESG performance of enterprises is mostly at the medium level. However, the maximum and minimum values of both the E score and the G score vary widely, which indicates that there are large differences in environmental protection responsibility fulfillment performance and corporate governance levels among most enterprises.

3.3. Model Construction

3.3.1. Model Construction for the Relationship Between ESG and TFP

To determine the relationship between TFP and ESG performance, we set the baseline model to affirm Hypothesis 1:

$$\mathrm{T}\mathrm{F}{\mathrm{P}}_{\mathrm{i},\mathrm{t}}={\alpha }_0+{\alpha }_1\mathrm{E}\mathrm{S}{\mathrm{G}}_{\mathrm{i},\mathrm{t}}+{\alpha }_2\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}{\mathrm{s}}_{\mathrm{i},\mathrm{t}}+\mathrm{I}\mathrm{n}{\mathrm{d}}_{\mathrm{i}}+\mathrm{Y}\mathrm{e}\mathrm{a}{\mathrm{r}}_{\mathrm{t}}+{\epsilon }_{\mathrm{i},\mathrm{t}}$$

(2)

where i denotes a listed firm and t denotes a certain year. TFP_i,t represents the TFP of firm i in year t and ESG_i,t represents the ESG rating of firm i in year t. α₁ and α₂ are the parameters to be estimated in the model. Controls refers to the set of control variables. Ind_i indicates the fixed effect of the industry. Year_t indicates the fixed effect of the year.

Based on the theoretical analysis in the previous section, in order to explore how ESG performance affects TFP by lowering financing constraints and reducing inefficient investment behavior, we first refer to the research of scholars to construct a mediation effect model [104], and we also use Sobel’s test to assess the mediation effect in order to quantify the relative importance of these mechanisms [105]. The following regression equation is constructed as follows:

$$\mathrm{M}\mathrm{e}{\mathrm{d}}_{\mathrm{i},\mathrm{t}}={\beta }_0+{\beta }_1\mathrm{E}\mathrm{S}{\mathrm{G}}_{\mathrm{i},\mathrm{t}}+{\beta }_2\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}{\mathrm{s}}_{\mathrm{i},\mathrm{t}}+\mathrm{I}\mathrm{n}{\mathrm{d}}_{\mathrm{i}}+\mathrm{Y}\mathrm{e}\mathrm{a}{\mathrm{r}}_{\mathrm{t}}+{\epsilon }_{\mathrm{i},\mathrm{t}}$$

(3)

$$\mathrm{T}\mathrm{F}{\mathrm{P}}_{\mathrm{i},\mathrm{t}}={\gamma }_0+{\gamma }_1\mathrm{E}\mathrm{S}{\mathrm{G}}_{\mathrm{i},\mathrm{t}}+{\gamma }_2\mathrm{M}\mathrm{e}{\mathrm{d}}_{\mathrm{i},\mathrm{t}}+{\gamma }_3\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}{\mathrm{s}}_{\mathrm{i},\mathrm{t}}+\mathrm{I}\mathrm{n}{\mathrm{d}}_{\mathrm{i}}+\mathrm{Y}\mathrm{e}\mathrm{a}{\mathrm{r}}_{\mathrm{t}}+{\mathrm{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(4)

where Med is the mediating variable, representing financing constraints (KZ) and inefficient investment (NEI), respectively. The other variables are defined as in Equation (2). In this case, determine whether the estimates of β₁ and ${\gamma }_1$ are significant by stepwise regression.

To explore whether FER and IER modulate the relationship between ESG and TFP, the following four models were constructed to synthesize this question:

$$\mathrm{T}\mathrm{F}{\mathrm{P}}_{\mathrm{i},\mathrm{t}}={\lambda }_0+{\lambda }_1\mathrm{E}\mathrm{S}{\mathrm{G}}_{\mathrm{i},\mathrm{t}}+{\lambda }_2\mathrm{E}\mathrm{S}{\mathrm{G}}_{\mathrm{i},\mathrm{t}}\times \mathrm{F}\mathrm{E}{\mathrm{R}}_{\mathrm{i},\mathrm{t}}+{\lambda }_3\mathrm{F}\mathrm{E}{\mathrm{R}}_{\mathrm{i},\mathrm{t}}+{\lambda }_4\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}{\mathrm{s}}_{\mathrm{i},\mathrm{t}}+\mathrm{I}\mathrm{n}{\mathrm{d}}_{\mathrm{i}}+\mathrm{Y}\mathrm{e}\mathrm{a}{\mathrm{r}}_{\mathrm{t}}+{\epsilon }_{\mathrm{i},\mathrm{t}}$$

(5)

where ${\lambda }_0$ is a constant term, and ${\lambda }_1,\dots ,{\lambda }_4$ is the coefficient to be estimated.

$$\mathrm{TF}{\mathrm{P}}_{\mathrm{i},\mathrm{t}}={\mathsf{\mu }}_0+{\mathsf{\mu }}_1\mathrm{ES}{\mathrm{G}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\mu }}_2\mathrm{ES}{\mathrm{G}}_{\mathrm{i},\mathrm{t}}\times \mathrm{IE}{\mathrm{R}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\mu }}_3\mathrm{IE}{\mathrm{R}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\mu }}_4\mathrm{Control}{\mathrm{s}}_{\mathrm{i},\mathrm{t}}+\mathrm{In}{\mathrm{d}}_{\mathrm{i}}+\mathrm{Yea}{\mathrm{r}}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(6)

where ${\mathrm{\mu }}_0$ is a constant term, and ${\mu }_1,\dots ,{\mu }_4$ is the coefficient to be estimated.

$$\begin{array}{l}\mathrm{T}\mathrm{F}{\mathrm{P}}_{\mathrm{i},\mathrm{t}}={\sigma }_0+{\sigma }_1\mathrm{E}\mathrm{S}{\mathrm{G}}_{\mathrm{i},\mathrm{t}}+{\sigma }_2\mathrm{F}\mathrm{E}{\mathrm{R}}_{\mathrm{i},\mathrm{t}}+{\sigma }_3\mathrm{I}\mathrm{E}{\mathrm{R}}_{\mathrm{i},\mathrm{t}}+{\sigma }_4\mathrm{E}\mathrm{S}{\mathrm{G}}_{\mathrm{i},\mathrm{t}}\times \mathrm{F}\mathrm{E}{\mathrm{R}}_{\mathrm{i},\mathrm{t}}+{\sigma }_5\mathrm{E}\mathrm{S}{\mathrm{G}}_{\mathrm{i},\mathrm{t}}\times \mathrm{I}\mathrm{E}{\mathrm{R}}_{\mathrm{i},\mathrm{t}}\\ +{\sigma }_6\mathrm{F}\mathrm{E}{\mathrm{R}}_{\mathrm{i},\mathrm{t}}\times \mathrm{I}\mathrm{E}{\mathrm{R}}_{\mathrm{i},\mathrm{t}}+{\sigma }_7\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}{\mathrm{s}}_{\mathrm{i},\mathrm{t}}+\mathrm{I}\mathrm{n}{\mathrm{d}}_{\mathrm{i}}+\mathrm{Y}\mathrm{e}\mathrm{a}{\mathrm{r}}_{\mathrm{t}}+{\epsilon }_{\mathrm{i},\mathrm{t}}\end{array}$$

(7)

where ${\sigma }_0$ is a constant term, and ${\sigma }_1,{\sigma }_2\dots ,{\sigma }_7$ is the coefficient to be estimated.

$$\begin{array}{l}\mathrm{T}\mathrm{F}{\mathrm{P}}_{\mathrm{i},\mathrm{t}}={\theta }_0+{\theta }_1\mathrm{E}\mathrm{S}{\mathrm{G}}_{\mathrm{i},\mathrm{t}}+{\theta }_2\mathrm{F}\mathrm{E}{\mathrm{R}}_{\mathrm{i},\mathrm{t}}+{\theta }_3\mathrm{I}\mathrm{E}{\mathrm{R}}_{\mathrm{i},\mathrm{t}}+{\theta }_4\mathrm{E}\mathrm{S}{\mathrm{G}}_{\mathrm{i},\mathrm{t}}\times \mathrm{F}\mathrm{E}{\mathrm{R}}_{\mathrm{i},\mathrm{t}}+{\theta }_5\mathrm{E}\mathrm{S}{\mathrm{G}}_{\mathrm{i},\mathrm{t}}\times \mathrm{I}\mathrm{E}{\mathrm{R}}_{\mathrm{i},\mathrm{t}}\\ +{\theta }_6\mathrm{F}\mathrm{E}{\mathrm{R}}_{\mathrm{i},\mathrm{t}}\times \mathrm{I}\mathrm{E}{\mathrm{R}}_{\mathrm{i},\mathrm{t}}+{\theta }_7\mathrm{E}\mathrm{S}{\mathrm{G}}_{\mathrm{i},\mathrm{t}}\times \mathrm{F}\mathrm{E}{\mathrm{R}}_{\mathrm{i},\mathrm{t}}\times \mathrm{I}\mathrm{E}{\mathrm{R}}_{\mathrm{i},\mathrm{t}}+{\theta }_8\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}{\mathrm{s}}_{\mathrm{i},\mathrm{t}}+\mathrm{I}\mathrm{n}{\mathrm{d}}_{\mathrm{i}}+\mathrm{Y}\mathrm{e}\mathrm{a}{\mathrm{r}}_{\mathrm{t}}+{\epsilon }_{\mathrm{i},\mathrm{t}}\end{array}$$

(8)

where ${\theta }_0$ is a constant term, and ${\theta }_1,{\theta }_2\dots ,{\theta }_8$ is the coefficient to be estimated.

3.3.2. Model Construction for the Nonlinear Relationship Between ESG and EE

This study examined the threshold effect between ESG performance and EE to verify Hypothesis 3, Hypothesis 3a, Hypothesis 3b, and Hypothesis 3c. On the basis of the previous analysis, we developed a panel threshold regression model as shown below [106]:

$$\begin{array}{l}\mathrm{E}{\mathrm{E}}_{\mathrm{i},\mathrm{t}}={\partial }_0+{\partial }_1\mathrm{E}\mathrm{S}{\mathrm{G}}_{\mathrm{i},\mathrm{t}}\cdot \mathrm{I}(\mathrm{t}{\mathrm{h}}_{\mathrm{i},\mathrm{t}}\le {\Gamma }_1)+{\partial }_2\mathrm{E}\mathrm{S}\mathrm{G}\cdot \mathrm{I}({\Gamma }_1<\mathrm{t}{\mathrm{h}}_{\mathrm{i},\mathrm{t}}<{\Gamma }_2)\cdots \cdots \\ +{\partial }_{\mathrm{n}}\mathrm{E}\mathrm{S}\mathrm{G}\cdot \mathrm{I}({\Gamma }_{\mathrm{n}-1}<\mathrm{t}{\mathrm{h}}_{\mathrm{i},\mathrm{t}}\le {\Gamma }_{\mathrm{n}})+{\partial }_{\mathrm{n}+1}\mathrm{E}\mathrm{S}\mathrm{G}\cdot \mathrm{I}(\mathrm{t}{\mathrm{h}}_{\mathrm{i},\mathrm{t}}>{\Gamma }_{\mathrm{n}})+\beta \mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}{\mathrm{s}}_{\mathrm{i},\mathrm{t}}+{\epsilon }_{\mathrm{i},\mathrm{t}}\end{array}$$

(9)

where i (i = 1, 2, ……, n) and t (t = 1, 2, ……, T) denote the number of firms and the number of observations, respectively. EE_i,t represents the energy efficiency of firm i in year t, which is the dependent variable. ESG_i,t represents the core variable as well as the threshold variable, denoting the ESG performance of firm i in year t. I(·) represents a conditional function, and ${\partial }_1,{\partial }_2,\dots \dots {\partial }_{\mathrm{n}-1},{\partial }_{\mathrm{n}}$ represents the coefficient to be estimated. th and $\Gamma$ constitute the threshold conditions in different situations, where th represents the threshold variables, including ESG, E score, S score, G score, GRDE, and GTTE, where $\Gamma$ represents the threshold value. Controls represents the control variables selected in this paper.

3.3.3. Model Construction for the Relationship Between EE and TFP

To test Hypothesis 4, we conducted Equation (10) to explore the impact of EE on TFP, which lags the TFP by one period and controls for the effect of current period TFP, and we also test the effect of TFP on EE in the lagged period and control for the effect of current period EE by conducting Equation (11). In view of the possible causal effects, we adopted the OLS regression method for the following models:

$$\mathrm{T}\mathrm{F}{\mathrm{P}}_{\mathrm{i},\mathrm{t}+1}={\delta }_0+{\delta }_1\mathrm{E}{\mathrm{E}}_{\mathrm{i},\mathrm{t}}+{\delta }_2\mathrm{T}\mathrm{F}{\mathrm{P}}_{\mathrm{i},\mathrm{t}}+{\delta }_3\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}+{\epsilon }_{\mathrm{i},\mathrm{t}}$$

(10)

$$\mathrm{E}{\mathrm{E}}_{\mathrm{i},\mathrm{t}+1}={\delta }_0+{\delta }_4\mathrm{T}\mathrm{F}{\mathrm{P}}_{\mathrm{i},\mathrm{t}}+{\delta }_5\mathrm{E}{\mathrm{E}}_{\mathrm{i},\mathrm{t}}+{\delta }_6\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}+{\epsilon }_{\mathrm{i},\mathrm{t}}$$

(11)

where TFP_i,t+1 represents the total factor productivity of the firm, which is specified with a lag of one period, and EE_i,t+1 represents the energy efficiency of the firm, which is specified with a lag of one period.

4. Empirical Results

4.1. Relationship Between ESG and TFP

4.1.1. Benchmark Regression and Robustness Check

Table 2. Regression and robustness check results.

Variables	TFP_OP	TFP_OP	TFP_LP	TFP_OP	TFP_OP	TFP_LP	TFP_OP	TFP_OP
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
ESG	0.128 ***	0.0457 ***	1.020 ***	0.0530 ***	0.136 ***	1.060 ***	0.0508 ***	0.0600 ***
	(6.79)	(3.18)	(8.20)	(4.07)	(6.67)	(7.91)	(3.26)	(4.27)
Size	0.0554 ***	0.0454 ***	0.571 ***	0.0963 ***	0.0550 ***	0.568 ***	0.0463 ***	0.0968 ***
	(48.67)	(25.60)	(76.00)	(75.07)	(45.97)	(72.22)	(24.71)	(71.36)
OC	0.0314 ***	−0.0378 ***	0.310 ***	0.0208 ***	0.0335 ***	0.322 ***	−0.0324 ***	0.0229 ***
	(4.60)	(−3.93)	(6.87)	(4.38)	(4.66)	(6.82)	(−3.19)	(4.61)
Lev	0.115 ***	0.0519 ***	0.923 ***	0.0263 ***	0.116 ***	0.930 ***	0.0484 ***	0.0249 ***
	(15.06)	(7.01)	(18.33)	(4.92)	(14.45)	(17.63)	(6.18)	(4.43)
Roa	0.279 ***	0.243 ***	2.407 ***	0.114 ***	0.296 ***	2.516 ***	0.254 ***	0.121 ***
	(12.88)	(16.53)	(16.86)	(7.06)	(12.96)	(16.80)	(16.31)	(7.13)
Far	−0.0897 ***	−0.0780 ***	−0.876 ***	−0.0671 ***	−0.0902 ***	−0.875 ***	−0.0827 ***	−0.0673 ***
	(−12.54)	(−9.49)	(−18.57)	(−13.61)	(−11.96)	(−17.67)	(−9.56)	(−12.94)
Growth	0.0282 ***	0.0292 ***	0.150 ***	0.00512 **	0.0264 ***	0.141 ***	0.0282 ***	0.00427 *
	(8.43)	(15.85)	(6.78)	(2.21)	(7.54)	(6.12)	(14.39)	(1.76)
Age				0.00733 ***				0.00793 ***
				(4.97)				(5.15)
TAT				0.147 ***				0.147 ***
				(80.12)				(75.85)
Indep				−0.0391 ***				−0.0353 ***
				(−3.10)				(−2.64)
Salary				−0.0476 ***				−0.0485 ***
				(−40.45)				(−39.00)
Tobin Q				−0.00117				−0.00137 *
				(−1.63)				(−1.75)
Constant	0.553 ***	0.865 ***	−4.970 ***	0.536 ***	0.556 ***	−4.922 ***	0.841 ***	0.537 ***
	(23.78)	(21.64)	(−32.42)	(29.91)	(22.80)	(−30.71)	(19.93)	(28.44)
Ind/Year	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Firm	No	Yes	No	No	No	No	Yes	No
N	6095	6095	6095	6095	5459	5459	5459	5459
Adj.R2	0.600	0.908	0.736	0.813	0.603	0.740	0.908	0.815

Notes: ESG represents the ESG value (×100); t statistics are reported in parentheses. *, **, and *** represent statistical significance at the 10%, 5%, and 1% levels, respectively.

provides the results of the regression and robustness tests. Column (1) shows the model without the addition of firm individual fixed effects, and it can be seen from the regression results that the regression coefficient of the explanatory variable ESG performance is a coefficient of 0.128, which is significant at the 1% level, which initially suggests that ESG performance has a positive impact on TFP. Column (2) adds individual firm fixed effects, and the results show that the estimated coefficient of ESG and TFP is 0.0457, which is also significant at the 1% level, indicating that for every unit increase in the level of firm’s ESG performance, TFP will increase by 0.0457 units. For the control variables, the regression coefficient for Far is consistently negative in both columns, while the regression coefficient for OC changes from positive to negative after controlling for individual firm fixed effects. In terms of economic significance, higher Far may imply that firms invest relatively large amounts in fixed assets, leading to the occupation of large amounts of capital, and that the untimely replacement of fixed assets may lead to low productivity, thus negatively affecting TFP. The addition of firms’ individual fixed effects can better control firms’ individual heterogeneity, making the real impact of OC on TFP visible. Therefore, too-high OC may be detrimental to the improvement of TFP, which is also confirmed in the robustness test.

We performed a series of robustness checks for such results: (1) The replacement of explanatory variables. We calculated TFP by the LP method [99] to replace TFP measured by the OP method as a robustness check to avoid endogeneity problems, and the robustness results are displayed in column (3) of Table 2; the coefficient is 1.020, significant at the 1% significance level. (2) We added control variables that may be missing in the benchmark regression: firm age (Age) inferred from the time of the firm’s listing; total asset turnover (TAT); the proportion of enterprises that are independently knowledgeable (Indep); and the firm’s wage level (Salary), which is a logarithmic value of the amount of money in cash paid to and for employees; and Tobin’s Q (Tobin’s Q). The robustness results are shown in Column (4), with a coefficient of 0.0530, which is significant at the 1% significance level. (3) To avoid possible disturbances during the period of the COVID-19 outbreak at the end of 2019, the sample data from 2020 were excluded from this robustness test, and the results are shown in Column (5); the coefficient is 0.136, which is significant at the 1% significance level. (4) On the basis of the exclusion of the 2020 sample data, several of the aforementioned robustness methods were tested, as shown in Column (6), Column (7), and Column (8). All the results are robust, thus proving Hypothesis 1 of this paper.

4.1.2. Mechanisms Analysis

The above regression analysis initially examines the impact of ESG performance on TFP. In the following, this paper further examines the mechanism of ESG performance on TFP in terms of the degree of financing constraints and the degree of inefficient investment.

Table 3. Mechanistic test results.

Variables	TFP	KZ	TFP	TFP	NEI	TFP
	(1)	(2)	(3)	(4)	(5)	(6)
ESG	0.128 ***	−0.0182 ***	0.126 ***	0.128 ***	−0.196 ***	0.129 ***
	(6.79)	(−2.98)	(6.70)	(6.79)	(−3.47)	(6.83)
KZ			−0.0867 **
			(−2.19)
NEI						−0.131 ***
						(3.93)
Size	0.0554 ***	−0.0474 ***	0.0513 ***	0.0554 ***	0.00123	0.0554 ***
	(48.67)	(−128.42)	(23.36)	(48.67)	(0.36)	(48.66)
OC	0.0314 ***	−0.00882 ***	0.0307 ***	0.0314 ***	−0.0136	0.0315 ***
	(4.60)	(−3.97)	(4.48)	(4.60)	(−0.67)	(4.61)
Lev	0.115 ***	0.0229 ***	0.117 ***	0.115 ***	0.0222	0.115 ***
	(15.06)	(9.26)	(15.22)	(15.06)	(0.97)	(15.05)
Roa	0.279 ***	−0.204 ***	0.261 ***	0.279 ***	−0.144 **	0.279 ***
	(12.88)	(−29.03)	(11.30)	(12.88)	(−2.23)	(12.90)
Far	−0.0897 ***	−0.00123	−0.0898 ***	−0.0897 ***	−0.0635 ***	−0.0894 ***
	(−12.54)	(−0.53)	(−12.56)	(−12.54)	(−2.97)	(−12.49)
Growth	0.0282 ***	−0.0406 ***	0.0247 ***	0.0282 ***	0.0551 ***	0.0279 ***
	(8.43)	(−37.41)	(6.65)	(8.43)	(5.52)	(8.33)
Constant	0.553 ***	0.0506 ***	0.557 ***	0.553 ***	0.243 ***	0.551 ***
	(23.78)	(6.71)	(23.89)	(23.78)	(3.50)	(23.71)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Ind/Year	Yes	Yes	Yes	Yes	Yes	Yes
N	6095	6095	6095	6095	6095	6095
Adj.R2	0.600	0.852	0.600	0.600	0.0362	0.600
Sobel	0.002 * (z = 1.765)			0.002 ** (z = 2.423)
Proportion of total effect that is mediated	0.012			0.201
Ratio of indirect to direct effect	0.013			0.199
Ratio of total to direct effect	1.013			1.008

Notes: In Soble’s test, when the Z-value is greater than 1.65 and the p-value is less than 0.05, it represents the rejection of the original hypothesis, and the mediating effect is established. t statistics in parentheses. *, **, and *** represent statistical significance at the 10%, 5%, and 1% levels, respectively.

shows the mediating mechanism results.

When the degree of financing constraints (KZ) is the mediating variable, the regression results for Equation (3) are shown in Column (1). The regression results for Equation (4) are shown in Columns (2) and (3). In Column (1), the regression coefficient of ESG is 0.128, which is significantly positive at the 1% level; the estimated coefficient of KZ in Column (2) is −0.0182, which is negative at the 1% level; and the estimated coefficient of KZ in Column (3) is −0.0867, which is also negative at the 1% level. At this point, Sobel’s estimate is 0.002 and is significant at the 10% level. These results suggest that firms’ ESG performance can improve TFP by alleviating financing constraints, as described in the previous theoretical analysis. Accompanied by more and more investors in the market paying more attention to ESG-related green investment projects [39], as well as based on the theory of information asymmetry, enterprises with good ESG performance increase the transparency of information by disclosing detailed ESG information. It not only effectively reduces the information asymmetry between enterprises and investors, but also enhances the ability of external stakeholders to supervise the company, as well as reducing the compensation effect of financial institutions and investors, thus alleviating the problem of corporate financing constraints and providing financial support for enterprises to carry out technological innovations, equipment upgrading, and other activities to improve TFP. Secondly, firms focusing on ESG practices tend to establish closer relationships with stakeholders, which helps firms to obtain more stable cash flow expectations, making financial institutions more willing to provide financing for firms, thus alleviating the financing constraints of firms, and further improving TFP. Hypothesis 1a is confirmed.

When the degree of inefficient investment (NEI) is used as a mediating variable, the regression results for Equation (3) are shown in Column (4), and the regression results for Equation (4) are shown in Column (5) and Column (6). In Column (4), the regression coefficient of ESG is 0.128, which is significantly positive at the 1% level; the estimated coefficient of NEI in Column (5) is −0.0182, which is negative at the 1% level; and the estimated coefficient of NEI in Column (6) is −0.0867, which is also negative at the 1% level. At this point, Sobel’s estimate is 0.002 and significant at the 5% level. In addition to this, we quantified the relative importance of these mechanism variables. When KZ is used as a mediating variable, the ratio of such a mediating effect to the total effect is 0.012, whereas when NEI is used as a mediating variable, the ratio of such a mediating effect to the total effect is 0.201. Thus, NEI exhibits a stronger mediating effect compared to KZ. Such results suggest that ESG performance can promote TFP by mitigating inefficient investment, and the promotion effect is more obvious compared to mitigating financing constraints. When firms engage in ESG practices, in addition to financial disclosure, firms are more transparent in disclosing ESG-related information, which makes the capital market pay more attention to firms’ investment management. Coupled with the combined effect of media and public scrutiny [39], corporate management’s improper behavior is suppressed. Second, firms that emphasize ESG practices pay more attention to optimizing strategic decisions, and the probability of phenomena such as information asymmetry will be relatively low [25]. Practicing ESG concepts will prompt enterprises to not only focus on short-term financial returns in the decision-making process, but also consider long-term development factors, such as the environment, society, and governance mechanisms, which makes firms more prudent and comprehensive in making investment decisions and avoids investment projects that only pursue short-term benefits but may be inefficient or even harmful in the long run [42,43]. Third, practicing ESG concepts will help firms identify and manage potential risks [37]. By responding to environmental risks such as climate change in advance, firms can avoid production disruptions, negative publicity, and other losses due to environmental issues, thus reducing the possibility of inefficient investments. On the social side, focusing on employee rights and community relations can avoid and reduce operational risks due to employee turnover, community opposition, and other issues, improve resource allocation efficiency, and promote TFP. These factors all contribute to the reduction of inefficient investment behavior, which ultimately helps to improve TFP. Hypothesis 1b of this study is confirmed.

4.1.3. Moderate Effect Analysis

The positive effects still need to be considered in the context of moderator effects, including the FER and the IER faced by enterprises. In this paper, we use the completed investment in industrial pollution control as a proportion of the secondary industry to measure the FER [107] and add the cross-multiplier term (ESG_FER) to examine the moderating influence of FER. This paper searches for the keyword “pollution” = in the Baidu search index to measure IER and adds the cross-multiplication term (ESG_IER) in the model to investigate the moderating effect of IER [56]. At the same time, we considered whether the relationship between ESG practices and TFP changed under the combined influence of FER and IER.

Columns (1) and (2) of

Table 4. Moderating effect results.

Variables	TFP	TFP	TFP	TFP	TFP	TFP
	(1)	(2)	(3)	(4)	(5)	(6)
ESG	0.531 ***	0.138 ***	0.515 ***	0.128 ***	0.133 ***	0.132 ***
	(22.10)	(7.23)	(21.67)	(6.79)	(6.95)	(6.93)
FER	0.0387 ***	0.0113			0.0181 **	0.0221 **
	(3.48)	(1.35)			(2.13)	(2.53)
ESG_FER	0.345 **	0.346 ***			0.230 **	0.200 *
	(2.56)	(3.43)			(2.21)	(1.91)
IER			0.0293 ***	0.00984 **	0.0110 **	0.0101 **
			(5.16)	(2.28)	(2.51)	(2.30)
ESG_IER			−0.131 *	−0.173 ***	−0.145 ***	−0.177 ***
			(−1.80)	(−3.17)	(−2.62)	(−3.09)
FER_IER					0.106 ***	0.111 ***
					(2.92)	(3.05)
ESG_FER_IER						−1.046 **
						(−2.14)
Size		0.0553 ***		0.0552 ***	0.0552 ***	0.0552 ***
		(48.60)		(48.38)	(48.35)	(48.37)
OC		0.0311 ***		0.0294 ***	0.0294 ***	0.0295 ***
		(4.54)		(4.27)	(4.28)	(4.28)
Lev		0.115 ***		0.116 ***	0.116 ***	0.116 ***
		(15.13)		(15.19)	(15.21)	(15.23)
Roa		0.279 ***		0.277 ***	0.277 ***	0.278 ***
		(12.91)		(12.80)	(12.79)	(12.83)
Far		−0.0907 ***		−0.0880 ***	−0.0883 ***	−0.0884 ***
		(−12.65)		(−12.23)	(−12.27)	(−12.28)
Growth		0.0284 ***		0.0281 ***	0.0282 ***	0.0281 ***
		(8.49)		(8.40)	(8.45)	(8.43)
Constant	1.779 ***	0.551 ***	1.782 ***	0.554 ***	0.550 ***	0.549 ***
	(237.89)	(23.72)	(248.76)	(23.83)	(23.67)	(23.66)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Ind/Year	Yes	Yes	Yes	Yes	Yes	Yes
N	6095	6095	6095	6095	6095	6095
Adj.R2	0.285	0.601	0.287	0.601	0.602	0.602

Notes: t statistics are reported in parentheses. *, **, and *** represent statistical significance at the 10%, 5%, and 1% levels, respectively.

show the moderating role of FER in the relationship between ESG and TFP, where the regression coefficients of ESG and the cross-multiplier term ESG_FER are both positive, suggesting that FER plays a positive moderating role. The stronger the degree of FER is, the more ESG can complement formal ER and play a more obvious role in the area. Columns (3) and (4) of Table 4 show the moderating effect of IER between ESG and TFP, and the regression coefficient of IER is positive at the 1% level of significance, indicating that increased public awareness of environmental issues can significantly improve the ESG performance of enterprises [56,57]. Nevertheless, the estimated coefficient of the interaction term between the two variables is significantly negative, which indicates that the higher the IER is, the smaller the promotion effect of ESG on TFP. It has been shown that IER leads to a widening of the pay gap among enterprises with high pollution emissions [108]. Moreover, IER also inhibits market access for polluting enterprises such that enterprises are pressured to reduce pollution emissions to reorganize their industrial layout [54]. In the production process, all kinds of behaviors under such pressure will have higher operating costs as well as pollution control costs, thus inhibiting the increase in TFP. Therefore, Hypothesis 2a and Hypothesis 2b are validated.

Column (5) of Table 4 reflects the results of Equation (7). The regression coefficient of the interaction term of FER and IER (FER_IER) is 0.106, which is significant at the 1% level, which suggests that FER and IER interactively affect firms’ TFP. When the triple interaction term is further added, Column (8) reflects the result of Equation (8) shows that the regression coefficient of the triple interaction term (ESG_FER_IER) is −1.046 and significant at the 5% level. This suggests that firms’ ESG practices, when influenced by the combined effects of FER and IER, will act together as a disincentive to TFP. We try to analyze such results. When the intensity of FER increases, this environmental signal will be transmitted to the public, and its sensitivity to negative environmental public opinion increases, and the intensity of IER increases at this time, thus when firms are still pursuing better ESG performance, it is more necessary for them to invest a large amount of resources in initiatives such as environmental protection technology improvement, pollution control cost improvement, social responsibility fulfillment, and internal system improvement. At this time, firms will probably focus more on meeting external requirements, so both production and non-production costs will increase significantly, and the phenomenon of “environmental over-compliance” will occur [96,109,110], which will hinder the improvement of TFP. Therefore, Hypothesis 2c is confirmed.

4.1.4. Heterogeneity Analysis

In this section, we analyze the heterogeneity of the impact of ESG performance on TFP for Chinese listed enterprises on the basis of the ownership structure, the nature of enterprises’ pollution emissions, and the attention that enterprises receive from capital market analysts, which is a necessary issue that must be considered in the development environment of Chinese enterprises. There are significant differences in the strategic integration of ESG concepts and the implementation of ESG disclosure policies among Chinese firms with different ownership structures. Therefore, we first divide the sample of enterprises into state-owned enterprises (SOEs) and non-state-owned enterprises (non-SOEs) (we use the CSMAR platform to find information provided on whether a particular shareholder holds shares in a State-owned legal entity and whether the shareholder type is the State or a State-owned legal entity. Unpublished information is then passed through other third-party online digital platforms (such as Aiqicha, which collects information from the State Enterprise Credit Information Publication System) to finalize the registration of the enterprise and to identify the shareholders or the ultimate beneficiaries of the enterprise).

As shown in columns (1) and (2) of

Table 5. Heterogeneity results within the impact of ESG performance on TFP.

Variables	SOEs	Non-SOEs	HPEs	Non-HPEs	High-AA	Low-AA
	(1)	(2)	(3)	(4)	(5)	(6)
ESG	0.165 ***	0.0873 ***	0.162 ***	0.0603 **	0.108 ***	0.175 ***
	(5.95)	(3.50)	(5.61)	(2.49)	(5.02)	(4.92)
Size	0.0512 ***	0.0583 ***	0.0535 ***	0.0547 ***	0.0527 ***	0.0600 ***
	(33.53)	(33.44)	(27.28)	(39.81)	(35.42)	(30.70)
OC	0.0242 **	0.0324 ***	0.0290 **	0.0179 **	0.0307 ***	0.0249 **
	(2.42)	(3.35)	(2.56)	(2.13)	(3.66)	(2.21)
Lev	0.114 ***	0.111 ***	0.0842 ***	0.122 ***	0.141 ***	0.0995 ***
	(10.89)	(9.96)	(6.59)	(13.00)	(13.70)	(8.68)
Roa	0.312 ***	0.284 ***	0.224 ***	0.303 ***	0.302 ***	0.308 ***
	(9.07)	(10.49)	(6.30)	(11.43)	(10.87)	(7.89)
Far	−0.0715 ***	−0.112 ***	−0.0612 ***	−0.166 ***	−0.0837 ***	−0.100 ***
	(−7.63)	(−10.05)	(−5.46)	(−16.67)	(−9.57)	(−8.44)
Growth	0.0409 ***	0.0169 ***	0.0350 ***	0.0234 ***	0.0128 ***	0.0425 ***
	(8.21)	(3.91)	(6.22)	(5.80)	(3.01)	(8.12)
Constant	0.638 ***	0.502 ***	0.601 ***	0.600 ***	0.607 ***	0.449 ***
	(20.60)	(13.91)	(15.09)	(21.14)	(19.74)	(11.02)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Ind/Year	Yes	Yes	Yes	Yes	Yes	Yes
N	3340	2755	2223	3871	3484	2608
Adj.R2	0.585	0.634	0.561	0.647	0.610	0.585
p value (χ2)	0.0295 **		0.0039 ***		0.0023 ***

Notes: t statistics are reported in parentheses. **, and *** represent statistical significance at the 5%, and 1% levels, respectively.

, the fulfillment of the ESG responsibilities of SOEs has a stronger facilitating effect on TFP, with a coefficient of 0.165, and non-SOEs also present a facilitating effect, but with a smaller coefficient of 0.0873. It has been shown that SOEs have more incentives to achieve sustainable corporate development than non-SOEs [38]. This is reflected in the ability to effectively monitor and constrain the behavior of controlling shareholders and discourage the pursuit of “short-term and quick” projects at the expense of ESG investments [96]. Compared with non-SOEs, SOEs have closer ties with government departments, and they have certain advantages in financing channels and tax incentives compared with non-SOEs, which provide a good guarantee for SOEs to practice ESG and improve TFP. From the perspective of corporate governance, SOEs’ directors, supervisors, and other senior managers are mostly appointed and removed by the government and are therefore subject to stricter supervision and better quality at the corporate governance level, thus SOEs are more inclined to actively fulfill their environmental protection and social responsibilities. However, in non-SOEs, the probability of risky issues such as agency problems and in-service consumption may be greater compared to SOEs.

In general, China’s high-emission, high-pollution listed enterprises often need to assume more environmental and social responsibilities and thus pay more attention to the accuracy and transparency of ESG information disclosure, which includes investing a lot of money into not only environmental protection and social responsibilities but also into avoiding negative public opinions on the market value of the impact of the cost of treatment [111]; thus, the financial performance of such enterprises is inhibited [17]. It has been shown that green finance [112], environmental regulation [113], and sustainable supply chains [114] significantly affect TFP in highly polluting firms, and it remains to be explored whether the effect of ESG performance on total factor productivity is heterogeneous among heavily polluting and nonpolluting firms. According to China’s “Guidelines for the Industry Classification of Listed Companies (2012 Revision)”and “List of Industry Classification Management for Listed Companies’ Environmental Protection Verification (2008)”, we categorize the sample into heavy-polluting enterprises (HPEs) and non-heavy-polluting enterprises (non-HPEs) (the industry classification code for heavily polluting listed companies is: B06, B07, B08, B09, C17, C19, C22, C25, C26, C28, C29, C30, C31, C32, D44. Enterprises belonging to these industry classifications are categorized as Highly Polluting Enterprises (HPEs), which include, for example, industries such as thermal power, steel, and cement, while enterprises in industries that do not fall within the scope of these classifications are regarded as non-Highly Polluting Enterprises (non-HPEs). The classification document reference is as follows: https://www.gov.cn/gzdt/2008-07/07/content_1038083.htm (accessed on 13 September 2024).

Columns (3) and (4) of Table 5 show that TFP can be significantly enhanced by ESG practices for both HPEs and non-HPEs, with coefficients of 0.162 and 0.0603, respectively, but the enhancement effect is more significant for HPEs. HPEs are often required to take on more environmental and social responsibility due to stricter government regulation [60]. In the process of fulfilling their social responsibilities, firms can obtain social capital through business–political relationships, and the external network of relationships formed during ESG practices is more conducive for HPEs to obtain knowledge and experience related to green innovation to be transformed into green innovation capabilities, thus promoting the improvement of TFP. On the other hand, it is also because the state has set stricter environmental restrictions on HPEs, and this pressure forces polluting enterprises to proactively transform and upgrade their environmental performance and reduce the risk of non-compliance, in order to minimize the environmental costs, thus promoting the improvement of TFP.

ESG performance is a significant investment reference standard for the capital market, and enterprises pay more attention to the performance of each part of the ESG strategy in the capital market. However, the pressure on enterprises to be noticed also makes managers unethical, and employees appear to bleach green behavior on ESG disclosure [115]. On the basis of previous research on the relationship between capital market attention and ESG performance, we take the level of analyst attention as a proxy variable of capital market attention, take the average value of analyst attention in the same industry as a truncated value, and categorize the sample into high analyst attention (High-AA) firms and low analyst attention (Low-AA) firms [52]. As shown in the results of Columns (7) and (8) of Table 5, we find that for enterprises characterized by greater capital market analyst attention, ESG contributes less to TFP. This is not surprising in practice, as intense capital market attention may indeed lead enterprises to engage in inappropriate behaviors such as “greenwashing” and emission falsification [52,61,99]. When too much money is invested in improper behaviors, the working capital cost is greater, and productivity is inhibited, thus preventing a positive increase in TFP.

4.2. Relationships Between ESG and EE

4.2.1. Threshold Effect Test

To examine whether threshold effects and nonlinear relationships occur, we adopted the methodology of [106] and performed the following tests.

Table 6. Threshold effect test.

Variables	Number of Thresholds	Value	F-Value	p-Value	RSS	MSE	Confidence Interval	10%	5%	1%
ESG	Single	26.245	13.040	0.007	3114.408	0.429	[25.224, 26.466]	8.859	10.549	12.560
E score	Double	15.464	16.300	0.000	2911.486	0.414	[15.373,15.796]	6.263	7.250	9.805
		14.286	26.780	0.000	2900.448	0.412	[13.863, 14.618]	9.372	12.679	19.532
S score	Single	11.548	25.190	0.000	2724.281	0.381	[11.397, 11.639]	13.124	14.671	17.548
G score	Double	41.071	27.250	0.000	3118.843	0.438	[38.049, 42.173]	6.979	8.704	11.552
		63.847	13.780	0.010	3112.813	0.438	[58.429, 64.419]	6.981	8.208	12.947
GRDE	Single	0.359	12.170	0.000	3114.781	0.429	[0.313,0.369]	7.033	8.244	10.152
GTTE	Double	0.283	8.710	0.037	3116.262	0.430	[0.246, 0.289]	7.006	7.979	10.964
		0.665	7.870	0.050	3112.884	0.429	[0.598,0.675]	6.679	7.859	11.138

Notes: The p value and the critical value were generated by repeated sampling 300 times with the Bootstrap. The fixed-effect model was used.

reports the results of the threshold tests for ESG, E score, S score, G score, GRDE, and GTTE. The results reveal that there is a single-threshold effect when ESG, the S score, and the GRDE are used as threshold variables. The threshold variables are ESG = 26.245, S score = 11.548, and GRDE = 0.359. When the E score, G score, and the GTTE are used as threshold variables, there is a double-threshold effect. The threshold variables are E score = 14.286 and E score = 15.464 for the E score, G score = 41.071 and G score = 63.847 for the G score, and GTTE = 0.283 and GTTE = 0.665 for the GTTE.

4.2.2. Threshold Regression Results

Model 1 in

Table 7. Threshold regression results of ESG for EE.

Model 1
Threshold Variable: ESG
ESG (ESG ≤ 26.245)	0.00410 ***
	(2.66)
ESG (ESG > 26.245)	0.00827 ***
	(8.00)
Constant	9.375 ***
	(30.37)
Control Variables	Yes
Controls	Yes
Adj.R2	0.0459

Notes: t statistics are reported in parentheses. *** represents statistical significance at the 1% level.

shows that the facilitating effect of ESG performance on EE gradually increases as the threshold for ESG increases. When the threshold is crossed, the coefficient increases from 0.00410 to 0.00827. The gradual increase in the marginal effect reflects the nonlinear character of the relationship between ESG and EE. The difference in the effects of different threshold intervals can be understood as the difference in the changes in costs and benefits at different stages of ESG investment. When ESG performance is relatively low, which means that it is in the early stages of development, the input cost of fulfilling environmental responsibilities is high, and the primary focus is on reducing emissions, without comprehensive attention being given to improving energy efficiency. However, when ESG performance crosses a certain threshold, better responsibility fulfillment pays more attention to energy adjustment, so EE is further promoted, which also reflects the scale effect.

The results of

Table 8. Threshold regression results of individual ESG pillars for EE.

Model 2		Model 3		Model 4
Threshold Variable: E Score		Threshold Variable: S Score		Threshold Variable: G Score
ESG (E ≤ 14.286)	0.0114 ***	ESG (S ≤ 11.548)	0.0123 ***	ESG (G ≤ 41.071)	0.00726 ***
	(10.05)		(10.28)		(3.01)
ESG (14.268 < E ≤ 15.464)	0.0136 ***	ESG (S > 11.548)	0.0107 ***	ESG (41.071 < G ≤ 63.847)	0.00582 ***
	(10.42)		(11.61)		(4.01)
ESG (E > 15.464)	0.0101 ***			ESG (G > 63.847)	0.00928 ***
	(11.15)				(9.01)
Constant	8.281 ***	Constant	8.515 ***	Constant	9.188 ***
	(28.64)		(28.39)		(29.83)
Controls	Yes	Controls	Yes	Controls	Yes
Adj.R2	0.0468	Adj.R2	0.0358	Adj.R2	0.0475

Notes: E represents the E score; S represents the S score; G represents the G score; and t statistics are reported in parentheses. *** represents statistical significance at the 1% level.

indicate that the threshold effects of the three pillars of ESG performance on EE are not exactly the same. According to the results of Model 2 in Table 8, the regression coefficient of ESG on EE changes as the E score increases. When the E score is ≤14.286, the coefficient is 0.0114. When the value of the E score is between 14.268 and 15.464, the coefficient is 0.0136, and through the comparative analysis of all the estimation results in Table 8, we find that this estimation is the largest among all the coefficient results. The environmental performance of enterprises directly reflects the extent to which enterprises cope with pollution emissions, the performance of renewable energy utilization, and the strength of environmental finance investment, and the improvement in environmental performance has a significant effect on ESG performance itself [116], which is consistent with our results showing a positive contribution in all three threshold regions. However, when the E score crosses the second threshold of 15.464, the marginal effect tends to decrease, at which point the regression coefficient of ESG on EE is 0.0101. Notably, overemphasizing environmental performance in ESG practices can also have the opposite effect [117].

When the E score is low, it means that the company is in the early stage of ESG practice. At this stage, the firm has just begun to implement ESG concepts and is more likely to focus on basic tasks such as pollution emissions or end-of-pipe pollution management, so it has not yet been able to comprehensively consider the improvement of EE at this stage. When ESG practices have reached a certain scale and synergies are generated among various inputs, firms are able to better integrate their resources, and the rewards of scale increase. However, when the E score crosses the second threshold variable, the marginal promotion effect tends to weaken. Firms with higher E scores may have excessive working capital invested in resolving issues such as environmental governance disputes, leading to an increase in multifaceted cost inputs [118]. Although environmental performance has positive externalities, such promotion is often unsustainable for enterprises’ long-term development strategy deployment without government policy subsidies or more market financing opportunities [119]. According to resource-based theory, at the initial stage of ESG practice, firms can significantly improve EE by improving energy management systems and adopting energy-saving technologies. However, as cost pressures increase, the marginal benefit of resource inputs diminishes, and once a specific threshold is exceeded, the effect of additional resource inputs on EE improvement will gradually diminish. Meanwhile, the theory of eco-modernization points out that when environmental performance reaches a high level, firms will face problems such as technological bottlenecks and market constraints.

Model 3 in Table 8 shows that the facilitating impact of ESG on EE decreases step by step as the value of the S score increases. The coefficient decreases from 0.0123 to 0.0107 after the S score crosses the threshold value of 11.548. Among mainstream international ESG rating agencies, Bloomberg, Thomson Reuters, FTSE Russell, the Dow Jones Sustainability indices (DJSI), Sustainalytics, the Carbon Disclosure Project (CDP), and MSCI all categorize stakeholder metrics in social responsibility performance to measure enterprises’ S scores for ESG ratings. This includes employees, investors, customers, and regulators, which are groups directly related to the company’s interests. Fulfilling ESG responsibilities will create long-term value for all stakeholders, including shareholders [120]. Whereas stakeholders play a role in ensuring consistent business benefits, each group has a role in ESG implementation [121]. It is from the perspective of stakeholder theory that the focus of different stakeholders at different stages changes with market conditions. When stakeholders are more focused on short-term profits, firms may reduce their investment in energy efficiency promotion in favor of profit-making activities. At this point, the promotion effect slows down. Therefore, enterprises with high S scores may pay more attention to investor sentiment, customer communication, and responses to regulators, thus neglecting the big picture of firm development and slowing the improvement in EE to some extent. The possibility remains that, with stakeholder communication being extremely important, unless such sustained high-density communication is able to influence the decisions and behaviors of the firm’s management and other stakeholders, this will only result in a scoring performance but will not add to its true value [122].

According to the results of Model 4 in Table 8, the regression coefficient of ESG on EE also varies as the G score increases. When the G score is ≤41.071, the regression coefficient of ESG on EE is 0.00726. The regression coefficient of ESG on EE decreases to 0.00582 when the value of the G score ranges from 41.071 to 63.847. As the G score increases further, the regression coefficient is 0.00928 when it exceeds 63.847, at which point ESG has the strongest positive impact on EE. Corporate governance mechanisms play a key role in achieving energy efficiency and sustainable economic growth [63]. In terms of executive characteristics and cognitive mechanisms, the real factors of corporate governance impact on corporate sustainability are usually directly related to the educational background, compensation and traits of the company’s executives, as well as to the level of executives’ cognitive awareness of environmental thinking and their concern for sustainability, which influences the company’s decision-making and actions in the process of corporate governance, and consequently, corporate sustainability [18]. On the basis of this mechanism, executive perceptions also trigger the mechanism of external effects. When a firm’s strategic decision makers are more sensitive to the risk awareness of environmental issues such as energy and are more concerned about the environment, it will attract the attention of green investors and market analysts, which in turn will contribute to the further fulfillment of environmental responsibility [114]. In addition, when the level of corporate governance is low, it reflects that the focus of the company’s strategy is biased toward short-term profitability and lacks clear planning and efficient execution in resource allocation. When the second threshold is exceeded, it implies that the firm integrates the ESG concept deeply into its strategy, clarifies the importance of EE enhancement in the long-term development of the firm, and cooperates with other resources to synergistically promote the related activities, at which time ESG has the strongest positive promotion effect on EE. The above results confirm Hypotheses 3 and 3a.

Model 5 and Model 6 in

Table 9. Threshold regression results of the GRDE and GTTE for EE.

Model 5		Model 6
Threshold Variable: GRDE		Threshold Variable: GTTE
ESG (GRDE ≤ 0.359)	0.00689 ***	ESG (GTTE ≤ 0.283)	0.00592 ***
	(6.03)		(4.40)
ESG (GRDE > 0.359)	0.0104 ***	ESG (0.283 < GTTE ≤ 0.665)	0.00987 ***
	(11.29)		(10.20)
		ESG (GTTE > 0.665)	0.0108 ***
			(11.79)
Constant	9.471 ***	Constant	9.450 ***
	(30.43)		(30.19)
Controls	Yes	Controls	Yes
Adj.R2	0.0465	Adj.R2	0.0453

Notes: t statistics are reported in parentheses. *** represents statistical significance at the 1% level.

show the regression results of the two technical variables as threshold variables. Model 5 shows that the facilitating effect of ESG on EE gradually increases as the value of ESG increases. The coefficient increases from 0.00689 to 0.0104 when the GRDE crosses the threshold value of 0.359. In Model 6, we find the same trend. When the GTTE ≤ 0.283, the coefficient is 0.00592. When the value of the GTTE is between 0.283 and 0.665, the coefficient increases to 0.00987. As the threshold increases, the regression coefficient is 0.0108 when it exceeds 0.665, at which point the facilitation is the strongest. We conclude that the GRDE better facilitates ESG and EE after a single threshold is crossed. The double-threshold effect presented by the GTTE also reflects that the facilitating effect is stronger when the threshold is constantly crossed. The results all suggest that the improvement in EE is directly related to technological factors and that green technology innovation can indeed promote EE [87].

In terms of the mechanism of action, a continuous increase in the threshold GRDE represents a continuous improvement in the efficiency of the relevant R&D, which means that the cost of developing and using renewable energy will be further reduced, and at the same time, the dependence on fossil energy extraction will be reduced [123]. Due to the scarcity of technological resources, as GRDE increases, it is gradually adapted to the productivity conditions of the society, which makes the promotion of ESG and EE increasing. From the numerical comparison, when GTTE further crosses the second threshold, its coefficient value (0.0108) is larger compared to that of GRDE (0.0104), and the promotion effect is more prominent compared to GRDE at the corresponding stage. According to the technology cycle theory, there are four stages in the life cycle of green technologies: emergence, development, diffusion, and maturity [124]. In the process of practicing the ESG concept, green technology development to the stage of conversion to productivity means that the enterprise goes from planning in advance for a more efficient way of energy utilization to integrating this result into the production process. The increasing threshold value of GTTE also represents the continuous development and maturity of the process of conversion to productivity, and the facilitating effect of ESG on EE also increases. Overall, in the analysis of the threshold effect, the overall performance of ESG is similar to the response trend of GRDE, both have a single-threshold effect, and both have a significant promotion effect successively with the increase in the threshold value. GTTE, although presenting a double-threshold effect, also has the trend of a continuously increasing promotion effect. From the perspective of sustainable development theory, green technology innovation remains a key factor in realizing corporate sustainable development [87,108,123], and ESG is a comprehensive measurement framework for corporate sustainable development [79]. The enhancement of GRDE and GTTE provides technical support for ESG practice, while ESG provides strategic guidance and institutional guarantees for the development of GRDE and GTTE. Continuous breakthroughs in R&D efficiency and transformation efficiency of green technologies will enable enterprises to have better performance in ESG to promote EE improvement and ultimately realize the sustainable development of enterprises. Therefore, Hypotheses 3b and 3c are validated. This study also provides a path for the development of green technology research and development strategies for listed enterprises in China.

4.2.3. Robustness Tests

To ensure the robustness of the results, we performed robustness checks. Since the COVID-19 epidemic broke out in China at the end of 2019 and Chinese listed enterprises were potentially affected both economically and financially, our robustness check excludes the data sample from 2020 to 2021.

Table 10. Excluding observations from 2020–2021: threshold effect test.

Variables	Number of Thresholds	Value	F-Value	p-Value	RSS	MSE	Confidence Interval	10%	5%	1%
ESG	Single	27.1502	17.36	0.003	2667.136	0.416	[26.2147,27.3011]	9.151	10.897	14.429
E score	Double	15.4636	17.16	0.003	2642.283	0.422	[15.2824,15.6448]	6.763	8.787	13.182
		13.9535	21.19	0.007	2633.366	0.421	[13.7119,14.3763]	7.938	12.928	18.755
S score	Double	12.0314	29.19	0.000	2692.986	0.425	[11.714,12.3337]	15.242	17.424	22.917
		7.3761	14.58	0.087	2686.803	0.424	[6.7412,7.6179]	13.494	16.451	19.907
G score	Double	39.2857	20.98	0.000	2421.369	0.392	[37.5,42.1734]	6.708	7.733	10.481
		77.6942	15.98	0.003	2415.118	0.391	[64.419,78.0554]	6.356	7.681	10.506
GRDE	Single	0.3661	13.33	0.007	2668.811	0.416	[0.3349,0.3761]	6.438	7.331	11.293
GTTE	Singe	0.2684	10.83	0.013	2669.848	0.417	[0.2268,0.2733]	6.611	7.42	11.439

Notes: The p value and the critical value were generated by repeated sampling 300 times with the Bootstrap. The fixed-effect model was used.

and

Table 11. Excluding observations from 2020–2021: threshold regression results.

Variables	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
	(1)	(2)	(3)	(4)	(5)	(6)
ESG (ESG ≤ 27.1502)	0.00526 ***
	(3.25)
ESG (ESG > 27.1502)	0.00887 ***
	(7.96)
ESG (E ≤ 13.9535)		0.0128 ***
		(10.44)
ESG (13.9535 < E ≤ 15.4636)		0.0143 ***
		(10.40)
ESG (E > 15.4636)		0.0109 ***
		(11.28)
ESG (S ≤ 7.3761)			0.0172 ***
			(8.76)
ESG (7.3761 < S ≤ 12.0314)			0.0143 ***
			(11.50)
ESG (S > 12.0314)			0.0122 ***
			(12.25)
ESG (G ≤ 39.2857)				0.0121 ***
				(5.10)
ESG (39.2857 < G ≤ 77.6942)				0.0117 ***
				(10.59)
ESG (G > 77.6942)				0.0104 ***
				(10.68)
ESG (GRDE ≤ 0.3661)					0.00777 ***
					(6.49)
ESG (GRDE > 0.3661)					0.0109 ***
					(11.16)
ESG (GTTE ≤ 0.2684)						0.00616 ***
						(4.08)
ESG (GTTE > 0.2684)						0.0113 ***
						(11.55)
Constant	9.616 ***	8.631 ***	9.174 ***	8.621 ***	9.717 ***	9.695 ***
	(29.22)	(27.87)	(28.26)	(27.33)	(29.30)	(29.07)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Adj.R2	0.0372	0.0384	0.0326	0.0404	0.0380	0.0369

Notes: E represents the E score; S represents the S score; G represents the G score; and t statistics are reported in parentheses. *** represents statistical significance at the 1% level.

confirm that the key findings of the threshold results remain unchanged. Although the results in Table 10 show that when the 2020 sample is excluded, the S score and the GTTE differ from the threshold test results of the original model, where the S score passes the double-threshold test (the threshold variables are as follows: S score = 7.3761 and S score = 12.0314, for the S score), and the GTTE only passes the single-threshold test (the threshold variable is as follows: the GTTE = 0.2684); however, combined with the results of Table 11, the results of constantly crossing the threshold reflect the same trend as the original model. Model 5 shows that the regression coefficient of ESG on EE is 0.0172 when the S score is ≤7.3761. When the S score is between 7.3761 and 12.0314, the regression coefficient of ESG on EE decreases to 0.0143. As the S score exceeds 12.0314, the magnitude of the positive effect of ESG on EE further decreases, at which point the regression coefficient of ESG on EE is 0.0122. Model 6 shows that the regression coefficient of ESG on EE also varies as the GTTE increases. When the GTTE crosses the threshold value of 0.2684, the regression coefficient increases from 0.00616 to 0.0113, which is still the same trend of increasing marginal effects as the original model. Thus, although the threshold tests for the S score and GTTE differ from those of the original model, the coefficients change in the same trend. These results suggest that our main results remain robust when observations from this particular period are excluded.

4.2.4. Heterogeneity Discussion

To further analyze the heterogeneity of the ESG threshold effects on EE, we focused on the ownership structure of listed enterprises in China by classifying the sample into SOEs and non-SOEs.

The threshold test in

Table 12. Test of the threshold effect for SOEs.

Variables	Number of Thresholds	Value	F-Value	p-Value	RSS	MSE	Confidence Interval	10%	5%	1%
ESG	Double	27.1301	22.25	0.000	710.406	0.209	[26.2851,27.3313]	9.418	11.93	14.778
		32.542	12.34	0.050	707.833	0.209	[31.6165,32.7834]	9.438	11.575	15.833
E score	Double	14.3763	11.88	0.020	701.613	0.211	[13.4551,14.6179]	6.461	8.003	12.329
		15.6448	20.86	0.003	697.235	0.210	[15.1918,15.9468]	5.209	7.059	11.003
S score	Single	10.6106	41.76	0.000	784.757	0.233	[10.399,11.0036]	13.049	16.943	19.154
G score	Double	66.5563	15.28	0.003	808.128	0.243	[66.165,72.2456]	6.569	8.029	10.701
		64.419	9.11	0.023	805.926	0.242	[57.1192,66.165]	6.822	8.315	11.99
GRDE	None
GTTE	Single	0.4084	17.48	0.000	711.398	0.210	[0.3811,0.4139]	8.606	10.162	12.705

Notes: The p value and the critical value were generated by repeated sampling 300 times with the Bootstrap. The fixed-effect model was used.

shows that for SOEs, there are double-threshold effects for ESG, the E score, and the G score for state-owned enterprises; the S score and the GTTE pass only the single-threshold effects, whereas the GRDE fails the threshold test.

Table 13. Threshold effect test for non-SOEs.

Variables	Number of Thresholds	Value	F-Value	p-Value	RSS	MSE	Confidence Interval	10%	5%	1%
ESG	Single	39.5534	13.33	0.070	1402.523	0.484	[37.9036,40.2877]	11.675	14.701	16.94
E score	Double	14.6179	15.29	0.003	1186.770	0.432	[14.3763,14.7992]	8.219	10.221	12.223
		14.7387	37.66	0.013	1170.738	0.426	[14.4971,14.7992]	8.237	10.769	49.852
S score	None
G score	Single	39.2857	11.57	0.013	1155.874	0.411	[37.5,41.0714]	7.507	9.095	11.847
GRDE	Single	0.3788	18.2	0.000	1400.180	0.483	[0.335,0.391]	7.901	8.88	13.252
GTTE	Double	0.6271	13.02	0.007	1397.946	0.482	[0.6225,0.6322]	6.523	8.034	10.041
		0.6322	9.8	0.043	1404.222	0.484	[0.6271,0.6388]	7.967	9.48	11.659

Notes: The p value and the critical value were generated by repeated sampling 300 times with the Bootstrap. The fixed-effect model was used.

shows the threshold effects test for non-SOEs, indicating that ESG, the G score, and the GRDE pass the single-threshold test, whereas the E score and GTTE pass the double-threshold test, and the S score fails the threshold test. We combined the threshold regression results to analyze the heterogeneity.

We compared and analyzed the results of

Table 14. Threshold regression results for SOEs.

Variables	Model 1	Model 2	Model 3	Model 4	Model 5
	(1)	(2)	(3)	(4)	(5)
ESG (ESG ≤ 27.1301)	0.00481 **
	(2.18)
ESG (27.1301 < ESG≤ 32.542)	0.00831 ***
	(4.86)
ESG (ESG > 32.542)	0.00684 ***
	(5.14)
ESG (E ≤ 14.3763)		0.0105 ***
		(7.59)
ESG (14.3763 < E ≤ 15.6448)		0.0131 ***
		(8.05)
ESG (E > 15.464)		0.00885 ***
		(8.11)
ESG (S ≤ 10.6106)			0.00983 ***
			(6.89)
ESG (S > 10.6106)			0.00867 ***
			(8.06)
ESG (G ≤ 64.419)				0.00604 ***
				(3.78)
ESG (64.419 < G ≤ 66.5563)				0.0138 ***
				(6.05)
ESG (G > 66.5563)				0.00738 ***
				(6.41)
ESG (GTTE ≤ 0.4084)					0.00618 ***
					(4.84)
ESG (GTTE > 0.4084)					0.00795 ***
					(7.54)
Constant	6.880 ***	6.691 ***	6.566 ***	6.585 ***	6.849 ***
	(17.53)	(17.33)	(17.09)	(17.12)	(17.47)
Controls	Yes	Yes	Yes	Yes	Yes
Adj.R2	−0.00283	−0.00280	−0.00707	0.0889	−0.00639

Notes: E represents the E score; S represents the S score; G represents the G score; and t statistics are reported in parentheses. **, and *** represent statistical significance at the 5%, and 1% levels, respectively.

and

Table 15. Threshold regression results for non-SOEs.

Variables	Model 1	Model 2	Model 3	Model 4	Model 5
	(1)	(2)	(3)	(4)	(5)
ESG (ESG ≤ 39.5534)	0.0202 ***
	(9.05)
ESG (ESG > 39.5534)	0.0169 ***
	(9.78)
ESG (E ≤ 14.6179)		0.0138 ***
		(7.22)
ESG (14.6179 < E ≤ 14.7387)		0.00719
		(0.97)
ESG (E > 14.7387)		0.0126 ***
		(8.07)
ESG (G ≤ 39.2857)			0.00998 **
			(2.35)
ESG (G > 39.2857)			0.0149 ***
			(8.79)
ESG (GRDE ≤ 0.3788)				0.00993 ***
				(5.01)
ESG (GRDE > 0.3788)				0.0144 ***
				(8.77)
ESG (GTTE ≤ 0.6271)					0.0141 ***
					(8.23)
ESG (0.6271 < GTTE ≤ 0.6322)					0.0268 ***
					(8.45)
ESG (GTTE > 0.6322)					0.0148 ***
					(9.05)
Constant	11.54 ***	9.896 ***	11.52 ***	12.03 ***	11.52 ***
	(23.39)	(21.75)	(22.93)	(23.73)	(23.21)
Controls	Yes	Yes	Yes	Yes	Yes
Adj.R2	0.0878	0.0945	0.0889	0.0903	0.0901

Notes: E represents the E score; S represents the S score; G represents the G score; and t statistics are reported in parentheses. **, and *** represent statistical significance at the 5%, and 1% levels, respectively.

. The results of Model 1 of Table 14 show that for SOEs, the coefficient is 0.00481 when ESG ≤ 27.1301, the regression coefficient increases to 0.00831 when ESG is between 27.1301 and 32.542, and the coefficient decreases again to 0.00684 when crossing the second threshold of 32.542. Model 1 of Table 15 shows that for non-SOEs, the coefficient is 0.0202 when ESG ≤ 39.5534, and the coefficient is reduced to 0.0169 when ESG crosses the threshold. A comparison of the coefficient values reveals that the coefficients of 0.0138 and 0.0126 for non-SOEs are greater than the three coefficient values for SOEs. This finding indicates that SOEs are subject to stricter government regulation than non-SOEs and are therefore more concerned about the effectiveness of ESG practices and follow the principles of ESG more [125]. Whereas non-SOEs are less regulated by the government because of fewer factors, once investing energy into ESG practices, they show more obvious promotion, are better able to benefit from ESG practices, and will also realize the effect of better sustainability [121].

According to Model 2 of Table 14, the regression coefficient of ESG on EE for SOEs varies as the E score increases. The coefficient is 0.0105 when the E score is ≤14.3763, 0.0131 when the E score is between 14.3763 and 15.6448, and decreases again to 0.00885 when it exceeds 15.6448. Following Model 2 of Table 15, we find that the coefficient for non-SOEs similarly varies as the E score increases. The coefficient is 0.0138 when the E score is ≤14.6179. When the E score is between 14.6179 and 14.7387, there is no significant effect of ESG on EE. However, as the E score increases further, when it exceeds 14.7387, the coefficient is 0.0126 and significant at the 1% significance level. These results show that although SOEs and non-SOEs have heterogeneous threshold effects on the E score, their coefficient trends after crossing the threshold are consistent with those of the original model. Therefore, the results also suggest that environmental performance should be controlled within the appropriate range to maximize the effect of ESG on EE promotion.

From the perspective of externalities in terms of environmental performance, since the purpose of environmental regulation is ultimately to achieve the goals of environmental and economic sustainability, the environmental performance of non-SOEs should always be better than that of SOEs if non-SOEs are not considered to internalize environmental externalities [126]. However, SOEs are continuously subjected to pressure from government policies, citizen monitoring, and technology costs during their continuous ESG practices, and their resource utilization efficiency cannot be improved in a short period. When undesirable environmental pollution occurs, SOEs receive lower penalties than non-SOEs do; thus, their incentive effect for environmental management is even lower. Therefore, our results also suggest that SOEs should be integrated and comprehensively developed under all types of pressure and should reasonably improve environmental performance to achieve better effects. For non-SOEs, ESG performance has the best facilitating effect on EE when the E score is lower than 14.6179. When the pressure from local government regulation and regulatory policies gradually intensifies, it increases the motivation of non-SOEs to passively internalize the effects of the interaction between economic activities and environmental governance, and the corresponding environmental management costs also increase, at which point ESG practices no longer demonstrate a catalytic effect on EE. For non-SOEs to truly achieve the SDGs, there is inevitability in this stage, and once the pressure from all parties is withstood, the continued focus on the E score and good performance ultimately leads to a significant facilitating effect on EE.

According to Model 3 in Table 14, the S score of SOEs has a single-threshold effect on ESG and EE, and the coefficient decreases from 0.00983 to 0.00867 when the S score exceeds the threshold value of 10.6106; this tendency toward a diminishing marginal effect is the same as that for the original model. However, the threshold test in Table 13 shows that there is no threshold effect on the S score of non-SOEs, which may be because nonSOEs, compared with SOEs, are aimed at maximizing economic benefits, have a limited ability to own resource advantages, have relatively weak awareness of social responsibility, and do not assume more environmental responsibility and social responsibility contributions with government intervention [127,128].

The results of Model 4 of Table 14 reveal that, for SOEs, the coefficient is 0.00604 when the G score is ≤64.419; the coefficient increases to 0.0138 when the G score is between 64.419 and 66.5563, but the coefficient decreases to 0.00738 when the G score is >63.7267. Model 3 of Table 15 shows that the coefficient increases from 0.00998 to 0.0149 when the G score of non-SOEs crosses the threshold value of 39.2857. First, SOEs face greater market competition for factors such as energy and other resource allocation due to greater government policy factors, greater public attention, and greater analyst attention, and are more focused on reputational capital [129]. ESG performance contributes best to EE only when the G score remains within a certain range. This also implies that the corporate governance of SOEs will be subject to greater requirements in the process of sustainable development. In addition, comparing the sizes of the above regression coefficients, we obtain the coefficient value of non-SOEs, 0.0149, which is the largest among the other four coefficients (0.00604, 0.0138, and 0.00738 for SOEs, 0.00998 for non-SOEs), and this facilitation is more intuitive. Undoubtedly, non-SOEs are not subject to the same level of regulation and capital attention in the market as SOEs are, and are often subject to managerial short-sightedness driven by the inherent goal of profit maximization; however, once non-SOEs focus on optimizing the corporate governance level, they are rewarded with higher returns [130]. Another reason is that investment projects tend to gravitate toward greener project attributes that are more environmentally friendly when the management of non-SOEs pays more attention to environmental protection and green and low-carbon practices, and because of the dual attraction of direct profit returns and enhanced corporate reputation, this factor has been more effectively promoted among non-SOEs [114].

From Table 12, we find that there is no threshold effect of the GRDE on ESG and EE in SOEs, whereas the GRDE in non-SOEs has a single-threshold effect. On the basis of Model 4 of Table 15, the coefficient increases from 0.00993 to 0.0144 after the GRDE of non-SOEs crosses the threshold value of 0.3788. We attempt to analyze such heterogeneous results. SOEs are subject to stricter environmental regulations in the long run [131], and such exogenous shocks force enterprises to adjust their production and marketing strategies, leading to an increase in their costs [132]. In a short period of time, SOEs are unable to balance the dual benefits of ESG practices with a significant increase in EE. Owing to the higher costs and risks associated with technological breakthroughs in green patent research and development and the fact that the benefits of patented inventions do not make a significant difference in the compensation of managers in SOEs and that managers in SOEs are more risk averse and usually do not advance in their positions through green patent innovations, there is insufficient incentive to develop green patented technologies in SOEs on a combined basis [133]. In contrast, non-SOEs do have a competitive advantage in the marketization and commercialization of green technology innovations [98]. In general, managers of non-SOEs, with the goal of maximizing profits, are more likely to maximize benefits and position advancement through patented inventions than SOEs are, which is also related to the form of the corporate system of non-SOEs. Beneficial factors increase motivation. With increasing financing opportunities in market competition, there is more technology capital for the integration of the energy market to improve the energy efficiency within the enterprise, which is the reason why the green technology research and development efficiency can cross the threshold, and the effect constantly increases.

For the GTTE, the results of Model 5 of Table 14 show that for SOEs, the coefficient increases from 0.00618 to 0.00795 when the GTTE crosses the threshold value of 0.4084. Model 5 of Table 15 shows that for non-SOEs, the relationship between ESG and EE varies as the GTTE increases. When the GTTE ≤ 0.6271, the regression coefficient of ESG on EE is 0.0141. When the value of the GTTE is between 0.6271 and 0.6322, the coefficient increases to 0.0268. When the value of the GTTE is greater than 0.6322, the promotion of EE by ESG slows at this point but remains high, with an estimated coefficient of 0.0148. The emergence of the GTTE often has a direct relationship with the advantages of resources such as green credit support, government subsidies and tax incentives, and SOEs themselves have the ability to access these resources, which is determined by the ownership structure. After crossing the threshold, the smooth increase in the promotion effect indicates that SOEs tend to be the best channel in the market to undertake emerging industries such as large-scale, high-risk, and high-tech industries, and the resource advantages are demonstrated at this stage [79,134]. However, the resulting coefficient values of the three intervals of non-SOEs are greater than the estimated coefficients of SOEs, although they vary with increasing thresholds. That is, once non-SOEs obtain more resources, such as green credit support, government subsidies, and tax incentives, ESG practices will improve the EE of this type of enterprise.

Overall, Chinese listed companies have focused more on the integration of innovation elements with social development, which coincides with the principles of ESG strategic planning [120]. However, sustainable development practices often do not necessarily increase the efficiency of green innovation and are related to numerous factors [135]. The contribution of ESG performance to EE exhibits significant heterogeneity across ownership structures, indicating that ESG performance may also make the market a Matthew effect competitive situation [114]. Nevertheless, this is no less than the best time to identify the sustainable green development of enterprises, as both SOEs and non-SOEs will seek the best development path in the implementation of ESG concepts in practice.

4.3. The Relationship Between EE and TFP

This paper employs Equations (10) and (11) to verify causal effects, and the results are shown in Columns (1) and (2) of

Table 16. OLS regressions for EE and TFP and the robustness check.

Variables	OLS Regression		Variable Replacement		GMM Estimation
	TFP_OPt+1	EEt+1	TFP_LPt+1	EEt+1	TFP_OPt+1	EEt+1
	(1)	(2)	(3)	(4)	(5)	(6)
EE	0.263 **	0.921 ***	2.083 **	0.913 ***	1.673 ***	0.814 ***
	(2.32)	(124.57)	(2.46)	(108.83)	(5.86)	(42.57)
TFP_OP	0.924 ***	0.00120 ***			0.732 ***	0.00925 ***
	(133.73)	(2.68)			(33.68)	(6.35)
TFP_LP			0.946 ***	0.000417 ***
			(124.93)	(3.42)
Constant	0.00697	0.00472 ***	−0.464 ***	0.00960 ***	−0.172 ***	0.00966 ***
	(0.46)	(4.75)	(−4.90)	(8.80)	(−4.86)	(4.08)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Adj.R2	0.912	0.875	0.937	0.877

Notes: t statistics are reported in parentheses. **, and *** represent statistical significance at the 5%, and 1% levels, respectively.

. As shown in Column (1) of Table 16, EE has a positive effect on TFP (β = 0.263 **), and TFP has a significant positive effect on EE (β = 0.00120 ***), as shown in Column (2). To ensure robustness, we applied two methods for the robustness test, including: (1) Replacing the TFP calculated by the OP method with the TFP calculated by the LP method [99], and the robustness results are shown in Columns (3) and (4). (2) To avoid the endogeneity problem, the GMM (generalized method of moments) is usually applied, and the results are shown in Columns (5) and (6). The results of both of these checks verify the robustness of the regression results. Our study reveals that the increase in TFP may be associated with changes in energy use efficiency and that technological progress and reforms represented by TFP are usually inseparable from EE in life activities, whereas EE, in turn, leads to an increase in TFP [3]. Thus, Hypothesis 4 is supported. We also find that this bidirectional causality has a stronger contribution of EE to TFP (0.263 > 0.00120), and such results are equally robust to the other two robustness checks (2.083 > 0.000417 and 1.673 > 0.00925).

5. Research Conclusions and Policy Recommendations

5.1. Research Conclusions

In this work, we utilize the data of China’s A-share listed companies from 2010–2022 to explore the relationship between ESG performance and TFP, perform mechanism tests and heterogeneity tests on the relationship between ESG performance and the threshold effect of EE as well as the heterogeneity of the ownership structure, and reveal the relationship between EE and TFP. Our conclusions and recommendations are as follows.

The results of this study prove that ESG performance significantly contributes to TFP. Firms scored by ESG rating agencies identify their potential risks more intuitively while directly affecting TFP. The heterogeneity analysis results suggest that SOEs and heavily polluting enterprises are more likely to promote ESG governance as a long-term sustainable development goal to improve their productivity. In addition, for this positive relationship, we tested the mechanism at both the internal and external levels. We find that firms with better ESG performance tend to be stronger in terms of increasing TFP by easing financing constraints and reducing inefficient investment behavior, whereas mitigating inefficient investment behavior plays a more significant role. Moreover, external monitoring should not be ignored. At the productivity level, formal environmental regulation and ESG performance tend to be complementary. However, informal environmental regulation has the opposite effect. The emergence of negative news causes many more polluting enterprises to bear more pollution control costs, and their reputation is negatively affected, which also has a dampening effect on TFP. A combined effect of formal environmental regulation and informal environmental regulation on TFP was demonstrated, but when ESG was combined with both of them, TFP was instead inhibited.

ESG performance exhibits a nonlinear relationship with EE, with marginal effects gradually increasing as the threshold increases. Additionally, the E score, S score, and G score derived from the ESG score decomposition, as well as the GRDE and GTTE, two technology-related elements, exacerbate this nonlinear feature. All these results support the existence of a threshold effect of ESG performance on EE. In addition, in the empirical analysis, we find that the threshold regression coefficient of ESG on EE is the highest and has the strongest facilitating effect in a series of outcome comparisons when the E score is between 14.268 and 15.464. Environmental performance has positive externality, and too much attention may bring about an increase in operating costs and other controversial risks, while there is a limit to the amount of money that can be wagered for energy efficiency improvement. ESG strategy, as a strategy for the long-term sustainable development of enterprises, requires that the attention of enterprises to environmental performance should be controlled within a specific range to achieve the maximum facilitating effect on the improvement of EE. Similarly, social responsibility is an important aspect that encompasses factors such as stakeholders and regulators and should be strategized more carefully. Social responsibility manifests itself in the consideration of communication being effective and should consider the true beneficial value of such behavior, and once the boundaries are exceeded, the best effect on EE promotion will be lost. We also find that the continuous mapping of enterprises at the governance level does make sense. ESG has the strongest effect on EE promotion when the performance of corporate governance exceeds the second threshold. The best ESG strategy approach can be realized only if the management of a company continuously improves its cognition of environmental thinking, sustainable development strategy, and global thinking, improves its internal governance mechanism, and utilizes its internal growth potential from the top down. Furthermore, we demonstrate that the green technology innovation of Chinese listed companies can indeed promote the effect of ESG on EE, and when the green technology research and development efficiency and the green technology achievement transformation efficiency continuously exceed the thresholds, a better promotion effect of ESG on EE can be realized.

In the heterogeneity analysis of the relationship between ESG and EE, we demonstrate that SOEs and non-SOEs do differ in their presentation effects in the implementation of ESG strategies. Analyzing ESG as a whole, once non-SOEs bet on the overall deployment of ESG strategies, they show a more pronounced contribution to EE relative to SOEs and are better able to benefit from ESG practices. The trend of the coefficients of both types of enterprises in terms of environmental performance after crossing the threshold is consistent with the original model. These results all demonstrate that current Chinese listed companies are more heavily pressured by government regulation, so much of the relevant work may only focus on environmental responsibility, which does not yield the best results when environmental management costs increase. To improve EE, the social responsibility performance of SOEs should also be controlled. For non-SOEs with relatively poor environmental sustainability, a weak awareness of social responsibility, and relatively imperfect governance mechanisms, the current performance of social responsibility does not have a significantly positive effect on EE. Therefore, the social regulation of non-SOEs should be more comprehensive and stronger. However, the heterogeneous characteristics are maximized at the corporate governance level. SOEs are inherently subject to more government regulatory factors, receive more attention from the capital market, and have greater reputational capital, thus targeting higher requirements at the corporate governance level. Non-SOEs, on the other hand, are driven by the inherent economic goal of profit maximization, their governance mechanisms are not sound, and their management’s awareness of risk identification is low, making them more prone to managerial shortsightedness. In the context of sustainable development related to energy policy, carbon emission policy and other environmental policies emerging in the market, enterprises with green investment programs are more likely to obtain policy recognition from authorities. With these factors, the management of non-state-owned enterprises is more likely to be attracted by green investment programs, and the promotion of ESG strategies for EE is more intuitive and effective once the effective improvement of the governance mechanism has been strengthened. It is worth noting that non-SOEs do have a competitive advantage over SOEs in terms of green technological innovation because SOEs do not have sufficient research and development incentives. If non-SOEs receive more policy subsidies and tax incentives, they will exert the best effect of technological factors on energy efficiency. This also reflects that the next path of ESG practices for non-SOEs will be more complicated.

This study also discusses the bidirectional causality between EE and TFP. Technological progress and reforms within enterprises, represented by TFP, are usually inseparable from EE, while EE is an important driver of TFP, and in this bidirectional causality, EE contributes more strongly to TFP.

5.2. Policy Recommendations

This research has theoretical and practical implications. This study provides a new theoretical framework and research perspective for subsequent scholars to explore the interaction of ESG, TFP, and EE in different industries, different ownership systems, and different stages of development. ESG performance can significantly improve TFP, whereas ESG has a significant nonlinear effect on EE, which is confirmed across the three components of ESG and the two technology-related elements, and that there is again a bidirectional causal relationship between EE and TFP. At the firm level, this study provides a theoretical basis and methodological path for enterprises to explore both productivity sustainability and energy market reform in ESG practices, which can help enterprises identify potential shortcomings in the implementation of ESG strategies and maximize the avoidance of potential risk points.

At the firm level, firms should focus on building a better ESG risk assessment and disclosure system. On the one hand, they should regularly conduct a comprehensive assessment of the ESG risks faced by the firm, and on the other hand, they should incorporate ESG risks into the overall risk management framework of the firm in order to formulate detailed risk response strategies. At the same time, ESG disclosure should be strengthened to show investors, consumers and other stakeholders the results and effectiveness of ESG practices of the firm and enhance market trust in the firm. In terms of investment decision-making, enterprises should prioritize investment in projects with good ESG performance based on the ESG standard investment project screening mechanism, avoiding investment in high-pollution and high-risk projects, and avoiding inefficient investment due to the neglect of ESG factors at the source. In terms of financing decisions, firms should actively create financing advantages based on ESG concepts. By issuing green bonds, sustainable development-linked bonds and other innovative financing tools, they can further enhance their financing competitiveness in the capital market and broaden financing channels.

At the level of policymakers, local governments should improve the environmental regulatory policy system and promote synergistic development of formal environmental regulation and informal environmental regulation. In terms of formal environmental regulation, policymakers should establish strict environmental performance standards by industry and by phase. At the same time, they should establish an incentive mechanism for environmental regulation, and give policy incentives such as tax breaks, financial subsidies, and prioritized project approvals to firms with excellent ESG performance and remarkable environmental governance, so as to motivate firms to actively improve their ESG performance. In terms of informal environmental regulation, they should recognize that firms’ ESG practices under informal environmental regulation will be detrimental to TFP. Therefore, policymakers should focus on guiding the media, the public, and other social forces to play an active role in establishing a sound platform for environmental information disclosure. At the same time, they should pay more attention to strengthening the guidance and regulation of the media, avoiding excessive negative media coverage of firms’ environmental problems, and creating a favorable public opinion environment.

Second, policymakers should implement differentiated policy guidance to promote the improvement of ESG practices among firms of different ownership. Policymakers should realize that SOEs themselves have the advantages of policy support, financing project resources, and talent introduction, and at the same time, they are the optimal channel for undertaking large-scale, high-risk, and other emerging industries. Therefore, the next policy guidance of the government for SOEs should tend to help enterprises from the internal level to be incentivized from the top down for green technological innovation and maximize the key deployment of green technological factors in their ESG strategies to achieve the best promotion effect for EE. Moreover, the government should increase its policy support for the green technology innovation of non-SOEs. Local governments should also increase effective market communication with SOEs, with non-SOEs, and with each other in both types of enterprises on the basis of policy regulation to guide policy makers, investors, and leaders of both types of enterprises to realize and balance the adaptation of their optimal ESG strategies.

Third, policymakers should formulate policy measures to promote green technology innovation, including increasing financial investment in green technology R&D as well as increasing support for the circulation and transformation of green technologies. For example, a trading platform for green technology achievements should be constructed, and modern information technology should be utilized to realize accurate docking between the supply and demand sides of technology. In the area of taxation, tax credits will be given to enterprises for the increased costs of equipment purchases and research and development inputs resulting from the adoption of green technologies, thereby incentivizing enterprises to widely apply green technologies and accelerating the transformation of green technologies into production practices in firms.