RETRACTED ARTICLE: Digital transformation, green innovation, and carbon emission reduction performance of energy-intensive enterprises

Abstract

Digital transformation and green innovation are powerful initiatives to achieve carbon peaking, carbon neutrality targets and high-quality economic development. Using a sample of high energy-consuming listed enterprises from 2012 to 2021, a double fixed-effect model is constructed to verify the effect of green innovation on the carbon emission reduction performance of high energy-consuming enterprises, and digital transformation is used as a moderating variable to analyze the inner mechanism of green innovation affecting the carbon emission reduction performance of high energy-consuming enterprises under the effect of digital transformation. The empirical results show that green innovation can significantly improve the carbon emission reduction performance of energy-consuming enterprises, while digital transformation positively moderates the effect of green innovation on the carbon emission reduction performance of energy-consuming enterprises. When considering the industry heterogeneity, the moderation effect of digital transformation is significant in the chemical raw materials and chemical products manufacturing industry and the electricity and heat production and supply industry, but the petroleum processing and coking and nuclear fuel processing industry, the non-metallic mineral products industry, the ferrous metal smelting and rolling processing industry and the non-ferrous metal smelting and rolling processing industry are not yet significantly affected by green innovation and digital transformation. The findings of the study provide empirical evidence to promote the improvement of carbon emission reduction performance of energy-intensive enterprises in China and to achieve the "double carbon" target.

Introduction

At the 75th session of the United Nations General Assembly in 2020, China solemnly announced that China would strive to reach peak carbon by 2030 and carbon neutrality by 2060. The "double carbon" target reflects China’s determination to comprehensively implement the concept of green development and promote ecological civilization¹. The core of green and low-carbon development is to update and optimise products, processes and services in line with environmental needs, with a view to allocating and utilising resources efficiently, reducing the damage caused by carbon emissions to the environment and ecology, and providing strong support for China to achieve the "double carbon" target and promote high-quality economic development. What is particularly noteworthy is that with the acceleration of the new round of technological revolution and industrial change, digital transformation has become a core driver of new development momentum, an important means to facilitate and promote green development, and a necessary way to achieve the "double carbon" target. As the basic industry of our country, the high energy-consuming industry has made great contribution to the construction of our country's industry and infrastructure. But at the same time, it also brings environmental problems that cannot be ignored. The most prominent one is "high pollution, high energy consumption and high material consumption"². Existing research has found that the output value of China's high-energy-consuming industries is less than 1/3 of the total industrial output value, but energy consumption, material consumption and waste emissions account for more than 1/2 of the total industrial emissions³. As far as China's carbon emissions are concerned, energy-intensive enterprises are the main source of carbon emissions and should be the key target of carbon emission reduction. Therefore, in the context of achieving the "double carbon" target and building a new pattern of economic development in China, high energy-consuming enterprises should focus on transformation and upgrading through the synergy of greening and digitalisation.

Literature review

Green innovation has received much attention from scholars at home and abroad in recent years as an effective means of addressing climate change. Existing studies have found that green innovation may have a complex 'double-edged effect' on carbon reduction performance. On the one hand, green innovation can reduce energy consumption or promote industrial upgrading, thereby improving carbon emission reduction performance. For example, Du & Li and Zhou et al. suggest that green innovation can improve energy efficiency and enhance carbon reduction performance by replacing fossil fuels with clean energy in the production process^4,5. In addition, Du et al. point out that green innovation can promote industrial upgrading by shifting production from low value-added, heavily polluting industries to high value-added, environmentally friendly industries in order to reduce the share of output value of pollution-intensive industries in the overall economic output and accelerate the improvement of carbon emission reduction performance⁶. On the other hand, some scholars argue that the energy rebound effect cannot be ignored. For example, Du et al. points out that when green innovation improves energy efficiency, it instead stimulates producers to consume more energy, making the emission reduction effect much smaller than the carbon emission growth effect, which in turn reduces carbon reduction performance⁷. Shen et al. also argue that the direct effect of green innovation suppresses carbon emissions, but the indirect effect exacerbates them⁸. Therefore, does green innovation by energy-intensive enterprises help to improve carbon emission reduction performance? The answer to this question concerns both the green development of high energy-consuming enterprises and the process of achieving China's "double carbon" goal.

The academic debate on the relationship between digitisation and carbon reduction performance can be traced back to studies on the role of information and communication technologies (ICT) or the internet on carbon reduction performance. These research perspectives fall into two main categories, with some scholars viewing digitalisation as a boon to environmental governance, arguing that it helps to send signals related to environmental protection in order to incentivise companies to implement environmentally friendly technologies. For example, Schulte et al. and Zhang & Wei suggest that ICT can play an important role in the area of curbing the negative effects of climate change by improving energy efficiency and reducing the cost of renewable energy^9,10. In addition, Chen and Xu et al. point out that ICT can effectively prevent and control environmental risks by predicting ecological risks, integrating resources, and environmental regulation^11,12. In contrast to these views, some other scholars argue that digitalisation does not save energy and even generates additional energy consumption and carbon emissions, which hinders the green development process. For example, Salahuddin and Alam argue that the rapid diffusion of ICT or the Internet has stimulated electricity consumption and put pressure on energy use, which in turn has led to increased carbon emissions¹³. Lange et al. and Belkhir & Elmeligi also believed that the "rebound effect" caused by the improvement of energy efficiency offset the impact of digitalization on environmental friendeness, and the energy increase effect in the production, use and disposal of ICT was much higher than the reduction effect, resulting in the increase of carbon emissions^14,15. Therefore, does the digital transformation of energy-intensive companies contribute to green innovation and corporate carbon performance? Will it help companies to achieve their carbon reduction targets? These questions have yet to be tested.

Based on this, this paper focuses on the key issues in the achievement of China's "double carbon" target, and explores the moderating effect of digital transformation in the role of green innovation on the carbon emission reduction performance of energy-consuming enterprises by analysing the mechanism of green innovation on the carbon emission reduction performance of energy-consuming enterprises, and analyses the heterogeneous effect of different industries. The main marginal contributions of this paper are as follows: First, it introduces digital transformation to reveal the inner "black box" of green innovation affecting the carbon emission reduction performance of energy-consuming enterprises under the effect of digital transformation, providing a new perspective to explore the carbon emission reduction performance of enterprises driven by both greening and digital transformation. Secondly, we adopt a text mining method to measure the degree of digital transformation of high energy-consuming enterprises by using digital keywords as a judgment criterion and clearly defining digital transformation events through manual recognition, data visualization and Python's "Jieba" function, and provide ideas for quantifying the digital transformation of enterprises. Thirdly, the heterogeneous effects of green innovation, digital transformation and the carbon reduction performance of enterprises are explored by considering the differences in the industries to which high energy-consuming enterprises belong. The findings of this paper not only expand the theoretical boundaries of the impact of green innovation and digital transformation on the carbon emission reduction performance of enterprises, but also provide a basis for decision-making to promote the achievement of China's carbon peak and carbon neutral goals.

Theoretical mechanisms

Analysis of the effect of green innovation on the carbon reduction performance of enterprises

Green innovation refers to a variety of innovative activities that aim to significantly reduce the environmental burden or improve resource efficiency, including energy-efficient production processes, environmentally friendly products or services, green management and business methods¹⁶. Therefore, green innovation can be mainly divided into green management innovation and green technology innovation¹⁷, i.e. from the perspective of management innovation, enterprises can improve their carbon reduction performance by implementing green innovation strategies and cultivating green innovation awareness. In addition, from the perspective of technological innovation, enterprises can reduce carbon emissions through end-of-pipe treatment technologies, cleaner production technologies and carbon capture, utilisation and storage (CCUS) technologies, thereby improving carbon reduction performance, but the rebound effect of green innovation may inhibit carbon reduction performance¹⁸.

Firstly, based on the perspective of green management innovation, enterprises can implement green innovation strategies and "green" the overall layout of research and development, production and marketing, which is conducive to the rational allocation of resources and the improvement of carbon emission reduction performance. In addition, by developing a green production management system, enterprises can cultivate awareness of energy saving and emission reduction internally and send green signals externally, which can help accelerate the improvement of their carbon emission reduction performance. On the one hand, enterprises, as the practitioners and main force of the "double carbon" target, will undoubtedly make green innovation strategies their leading strategy to fulfil their emission reduction responsibilities, optimising product development, improving production processes, upgrading technology and equipment, and adjusting energy consumption structures in accordance with energy saving and emission reduction requirements, thereby promoting low or zero carbon emissions. On the other hand, based on the meaning-giving mechanism, the "double carbon" target will also drive enterprises to develop and implement green management systems to internally cultivate awareness of energy saving and emission reduction among employees and strengthen environmentally friendly production practices, while externally, through the signalling mechanism, the environmental protection image of enterprises to fulfil their social responsibility will be conveyed to the outside world to gain the recognition of government, financial institutions and other stakeholders. Externally, through the signaling mechanism, the enterprise's image of fulfilling its social responsibility for environmental protection is conveyed to the outside world, gaining recognition from stakeholders such as government and financial institutions, which helps to alleviate the dilemma of financing constraints that enterprises may encounter in green innovation, thus improving their carbon reduction performance¹⁹.

Secondly, based on the perspective of green technology innovation, enterprises can promote carbon emission reduction by adopting end-of-pipe treatment technology, cleaner production technology and CCUS technology. Firstly, end-of-pipe technology innovation focuses on the effective management of pollutants at the end of the enterprise, reducing the emission of pollutants such as sewage and waste gas, and to a certain extent, slowing down the trend of environmental pollution and damage caused by production activities²⁰. In addition, technological innovations in end-of-pipe treatment can also improve the end-of-pipe collection of production waste gases, providing the conditions for enterprises to make rational use of waste and realise the recycling of carbon dioxide, thus promoting carbon emission reduction in enterprises²¹. Secondly, by adopting cleaner production technologies, enterprises can reduce pollutant generation at source by improving process technology, using alternative energy sources and energy-saving equipment, etc.; secondly, by product innovation, using environmentally friendly products to replace non-clean products, reducing the negative impact on the environment throughout the product life cycle, thus improving carbon emission reduction performance²². Finally, the use of CCUS technology allows companies to recycle carbon dioxide, i.e. to purify the carbon dioxide produced in the production process and put it back into production or to store it, thus achieving effective emission reduction²³.

Third, based on the rebound effect perspective. The rebound effect refers to the fact that while green innovation improves the efficiency of energy resources and reduces energy consumption, it also generates new demand for energy, which can even offset the reduction in energy. Under the 'double carbon' objective, energy-intensive enterprises will inevitably limit their behaviour to what is beneficial for carbon reduction, i.e. they will improve energy efficiency and reduce carbon emissions through green innovation, while at the same time they may stimulate energy-intensive enterprises to consume more fossil fuels and other energy sources, driving new energy demand and thus partially, or even completely, offsetting the energy saved. This phenomenon is known as the energy rebound effect of green innovation, i.e. the indirect effect of green innovation on carbon reduction performance. Based on this, this paper proposes hypothesis H₁.

H₁

The direct effect of green innovation helps to improve the carbon reduction performance of firms, but the resulting energy rebound effect inhibits the performance of firms.

Analysis of the moderating effect of digital transformation between green innovation and corporate carbon reduction performance

Digital transformation refers to the continuous deepening of enterprises' application of digital technologies represented by cloud computing, the Internet of Things and big data to accelerate business optimisation, upgrading and innovation transformation, transform traditional kinetic energy and cultivate new kinetic energy to achieve transformation, upgrading and innovation processes. The digital transformation of enterprises can not only improve the efficiency of the flow of innovation factors among enterprises and promote the rapid concentration of innovation factors to enterprises, but also help enterprises break through the boundaries of time and space to form a platform for information sharing and innovation collaboration regarding energy input structure, carbon emissions and green emission reduction technologies, accelerating the improvement of enterprises' carbon emission reduction performance.

First, based on the factor allocation perspective, digital transformation is conducive to improving the efficiency of the allocation of green innovation factors in enterprises and driving the improvement of carbon emission reduction performance. Firstly, it enhances the flow speed of innovation factors. Big data, blockchain and other technologies have triggered a deep change in the flow mechanism of innovation factors. By breaking through the barriers to factor flow among innovation subjects and shortening the flow paths between different innovation factors, innovation factors will be allocated to the process of green innovation and energy saving and emission reduction of enterprises at an accelerated pace, and will lead to the improvement of the emission reduction performance of enterprises²⁴. Secondly, to achieve the precise matching of innovation factors. Digital transformation can solve the information silos and digital divide among enterprises, break the time and space boundaries of innovation links, enable enterprises to precisely match innovation factors and integrate innovation resources in the process of green innovation, achieve energy saving and emission reduction in all aspects of product development and production processes, and thus improve the carbon emission reduction performance of enterprises²⁵. Finally, the combination mode of innovation factors is changed. Digital technology has transformed the combination mode and sequence of innovation factors, allowing different factors to overlap and cross-combine in space and time, providing a more diversified combination of factor supply for enterprise green innovation, thus promoting the efficiency of enterprise innovation resource allocation and enhancing the effect of low carbon emission reduction²⁶.

Secondly, based on the perspective of information sharing, the digital transformation of enterprises can improve the "information power" in green innovation and thus enhance carbon emission reduction performance. Firstly, big data technology improves the ability of enterprises to gather and integrate data and information on energy input structures, carbon emissions and green emission reduction technologies, broadening the depth of information power, enabling green innovation to break through resource constraints and helping enterprises to make carbon emission reduction forecasts and decisions and quickly grasp carbon emission market trends. Secondly, digital technology can help enterprises integrate internal and external information, and transmit, flow and share information on green emission reduction technologies and other aspects through various channels, increasing the breadth of information power and providing opportunities for cross-sectoral green collaborative innovation. Finally, digital technology improves the ability of enterprises to rapidly develop, iterate and deliver, accelerating the speed of their information power and significantly improving the efficiency of their green innovation. Thus, digital transformation can enhance the 'information power' of green innovation in enterprises, through digital technology can effectively track the consumption of raw materials, energy demand and waste output, so that managers can easily control the energy consumption and production situation in the production process, and target to control carbon emissions from the source to the end through clean production technology or end-of-pipe treatment technology, thus achieving the goal of green innovation. The goal of "double carbon" can be achieved by controlling carbon emissions from the source to the end through cleaner production technologies or end-of-pipe management²⁷. Based on this, this paper proposes hypothesis H₂.

H₂

Digital transformation positively moderates the contribution of green innovation to firms' carbon reduction performance.

Study design

Model construction

Drawing on the model of Li et al.²⁸, this paper combines data related to 452 high energy-consuming listed enterprises in China from 2012 to 2021 to verify the effect of green innovation on the carbon emission reduction performance of enterprises, and the baseline econometric model constructed is as follows:

$$Cerp_{it} = \alpha_{0} + \alpha_{1} CreRatio_{it} + \alpha_{2} (\overline{{GreRatio_{it} }} \times \overline{{Energy_{it} }} ) + \alpha_{3} Control_{it} + u_{i} + \upsilon_{i} + \varepsilon_{it} ,$$

(1)

where $Cerp_{it}$ is carbon reduction performance;$CreRatio_{it}$ is green innovation;$Energy_{it}$ is total energy consumption; to avoid multicollinearity problems, an interaction term between centralised green innovation and total energy consumption is added to the model $(\overline{{GreRatio_{it} }} \times \overline{{Energy_{it} }} )$; $Control_{it}$ is a control variable;$u_{i}$ is firm fixed effects;$\upsilon_{i}$ is time fixed effects;$\varepsilon_{it}$ is a random disturbance term; i is an energy-intensive firm; t is a year.

On this basis, in order to further analyze the impact of the interaction between green innovation and digital transformation of high energy-consuming enterprises on their carbon reduction performance, the interaction term of green innovation and digital transformation after centralization was added to the model $(\overline{{GreRatio_{it} }} \times \overline{{DigTra_{it} }} )$, and the following econometric model was obtained:

$$\begin{gathered} Cerp_{it} = \beta_{0} + \beta_{1} CreRatio_{it} + \beta_{2} DigTra_{it} + \beta_{3} (\overline{{GreRatio_{it} }} \times \overline{{DigTra_{it} }} ) \\ + \beta_{4} Control_{it} + u_{i} + \upsilon_{i} + \varepsilon_{it} , \\ \end{gathered}$$

(2)

where $DigTra_{it}$ is the degree of digital transformation,$\beta_{3}$ is the interaction effect, when $\beta_{3}$ > 0, it indicates that digital transformation has a positive moderating effect on the relationship between green innovation and corporate carbon reduction performance, and vice versa, when $\beta_{3}$ < 0, it is a negative moderating effect.

Variable measures

Explanatory variables

Since enterprises rarely disclose their carbon dioxide emissions, this paper collects data on industry carbon emissions from the CEADs database, based on the measurement method of Zhao et al.²⁹, and then estimates the carbon emissions of enterprises. Therefore, this paper uses the operating income per unit of carbon emissions as a proxy variable for carbon emission reduction performance (Cerp), and the larger the value of this indicator, the better the carbon emission reduction performance of the enterprise, which is calculated in Eq. (3). In addition, this paper draws on Li et al.³⁰ to apply relative performance thinking by replacing the explanatory variable with a dummy variable (whether the enterprise has received government environmental recognition) to measure carbon emission reduction performance (EnvPro) for robustness testing.

$$=\frac{\mathrm{Carbon\,\, emmision \,\,reduction\,\, performance }({\text{Cerp}})\mathrm{ corporate \,\,operating\,\, income}}{\left(\frac{\mathrm{Industry \,\,carbon\,\, emissions}}{\mathrm{Industry \,\,main \,\,business\,\, cost}}+1\right)\times \mathrm{Business \,\,operating \,\,cost}}.$$

(3)

Explanatory variables

Existing literature on the measurement of green innovation mostly utilises R&D inputs in environmental protection and the number of green patents, with the former considered as an input to innovation activities and the latter as an output. Given that patent data can more accurately portray the characteristics of technology areas and actual innovation capabilities, this paper selects the number of green patents to measure the green innovation capabilities of enterprises. Drawing on Li and Xiao³¹, we searched the patent application and authorization data of enterprises in the State Intellectual Property Office. Combined with the IPC classification number of green patents, the data of green patent application and authorization of enterprises are finally obtained. Compared with the number of green patents granted, the number of green patent applications reflects the actual time for enterprises to carry out innovation. Moreover, the proportion of green patents is more effective than the number of green patents in controlling other unobservable factors that affect corporate innovation. Therefore, this paper uses the ratio of the number of green patent applications of enterprises to the total number of patent applications (GreRatioA) to measure the green innovation capability of enterprises. In the robustness test, two methods are used to remeasure green innovation, one is to use the method of Yu et al.³² to measure green management innovation (GreMan) by the disclosure of environmental management of enterprises in the current year; the other is to use the method of Lei et al.³³ to measure green process innovation by using the section related to green process innovation ("desulphurisation project" and "desulfurisation project") in the notes to the financial reports of listed enterprises in the construction in progress. ", "desulphurization project", "environmental protection project", "ultra-low emission project", etc.) in the notes to the financial reports of listed enterprises to measure green process innovation (GrePro).

Moderating variables

At present, most traditional enterprises in China are in the initial stage of digital transformation, and there is little literature on quantitative research on enterprise digital transformation at the micro level, and no scientific method has been proposed to assess the degree of digital transformation. Considering that digital transformation is a new engine for high-quality development of enterprises, the decision-making information of listed enterprises in this regard is usually published in their annual reports with guidance, and the key terms in their annual reports can reflect their future development strategies to a certain extent through text mining. Therefore, this paper adopts a text mining method to extract the frequency of words related to "digital transformation" from the annual reports of listed companies, so as to characterise the degree of digital transformation. Based on the measurement method of Wu et al.³⁴, firstly, the annual reports of high energy-consuming listed enterprises were crawled from Juchao Information Website and Oriental Fortune Website using Python. Secondly, based on academic literature³⁵, policy documents and government work reports, keywords related to digital transformation were summarised and a team of digital transformation experts was consulted to identify 40 keywords related to digital transformation, including digitalisation and informatisation. The 40 keywords were summarised by using Python's "Jieba" for each listed company's annual report. Finally, the degree of digital transformation (DigTra) was measured by the ratio of the total frequency of digital transformation keywords to the total frequency of keywords of enterprises in the same industry, as measured by Eq. (4). In the robustness test, two methods are used to measure digital transformation. The first method, borrowing from Yuan et al.³⁶, uses the ratio of the total frequency of digital transformation keywords of listed enterprises to the total number of words in the management discussion and analysis section of the annual report (DigMda) to measure the digital transformation; in the second method, considering that it is impossible to judge the degree of digital transformation achieved by enterprises by extracting keywords from the annual report alone, this paper draws on the approach of Zhang et al.³⁷ to measure digital transformation using the portion of the notes to the financial reports of listed companies that are related to digital transformation in the intangible assets line item ("computer software" "grid access system" "computer software", "ERP systems", "intelligent platforms", etc.) to total intangible assets (DigAss) to measure digital transformation.

$$=\frac{\begin{array}{c}The \,\,degree \,\,of \,\,digital\,\, transformation (DigTra)\\ Total \,\,frequency\,\, of \,\,enterprise \,\,digital\,\,transformation\,\, keywords\end{array}}{\mathrm{Total\,\,frequency\,\, of\,\, digital\,\, transformation\,\, keywords \,\,of \,\,enterprises\,\, in \,\,the\,\, same\,\, industry}}.$$

(4)

Control variables

Drawing on the relevant literature³⁸, this paper controls for variables at both the city and firm levels of influence. The variables at the city level include: ① Industrial structure (Struct): secondary industry/GDP is used to measure. ② Environmental regulation intensity (Ers): a composite index constructed using two indicators of industrial smoke (dust) and sulphur dioxide removal rate is used to measure; ③ Economic development level (Pgdp): the real GDP per capita of each city is used to measure. The variables at the enterprise level include: ① Total energy consumption (Energy): the total energy consumption of the industry is divided by the industry's main business cost and multiplied by the enterprise's operating cost to measure the energy rebound effect caused by green innovation. ② Gearing ratio (Lev): measured as total liabilities/ total assets to control for the solvency of a firm. ③ Net profit growth rate (Npgr): measured by (firm's current net profit—previous net profit)/previous net profit to control for the firm's growth capability. ④ Current Asset Turnover Ratio (Catr): measured by using the enterprise's operating income/end balance of current assets to control the enterprise’s operating capability. ⑤ Research and development expense ratio (Rder): measured by the company's research and development expense/operating revenue to control the profitability of the company.

Data sources

This paper conducts an empirical study using data on China's high energy-consuming listed enterprises from 2012 to 2021, which is mainly sourced from the CSMAR database, the State Intellectual Property Office, the annual reports of listed companies in previous years, the China Urban Statistical Yearbook and the China Energy Statistical Yearbook. The six high-energy-consuming industries of the 2011 National Economic and Social Development Statistical Report and the 2020 SFC Industry Classification Guidelines for Listed Companies were used to match the high-energy-consuming listed companies, and listed companies with ST,  ∗ ST and PT and those that changed industries during the study period were removed. Meanwhile, to eliminate the effect of heteroskedasticity, all continuous variables were log-model transformed; to eliminate the effect of outliers, all continuous variables were shrunken at 1% standard; to eliminate the effect of price changes, corporate financial data and GDP of each city were converted to constant prices in 2012. The descriptive statistics and correlation coefficient matrix of the variables are shown in Table 1. From Table 1, it can be seen that there is a strong correlation between the explanatory and explanatory variables, which provides initial support for the underlying regressions.

Table 1 Descriptive statistics and correlation coefficient of variables.

Empirical analysis

Baseline regression results

In this paper, the VIF test was first conducted and the VIFs were all less than 2, so there was no multicollinearity between any of the variables. Secondly, the Hausman test was conducted and the original hypothesis was rejected, so the fixed effect model was chosen. Finally, the between-group heteroskedasticity test and serial correlation test were conducted, and the results all rejected the original hypothesis, indicating that there is heteroskedasticity and autocorrelation between groups in all types of enterprises. In summary, to eliminate the possible heteroskedasticity among cross-sectional individuals, this paper uses a fixed-effects model and obtains standard errors by randomly sampling 500 repetitions to estimate the impact of green innovation on the carbon emission reduction performance of high energy-consuming enterprises. Column I of Table 2 shows the estimation results in the case where no control variables are introduced, and columns II and III show the estimation results in the case where firm-level, enterprise and city-level control variables are introduced respectively the estimated results.

Table 2 Impact of green innovation on carbon emission reduction performance of energy-intensive enterprises.

First, from column I of Table 2, the direct impact coefficient of green innovation on carbon emission reduction performance of high energy-consuming enterprises is 0.147 and is positively significant at the 1% level, indicating that every 1% increase in green innovation can increase the carbon emission reduction performance of high energy-consuming enterprises by 0.147%. The coefficient of indirect effect of green innovation and total energy consumption on carbon emission reduction performance is -0.007, which is negative and significant at the 10% level, indicating that green innovation also negatively affects carbon emission reduction performance through total energy consumption, indicating that green innovation may lead to a certain degree of energy rebound effect, but the indirect effect of energy rebound is much smaller than the direct effect of green innovation. The hypothesis that H₁ is validated is that green innovation has a positive effect on the carbon emission reduction performance of energy-intensive enterprises. After the introduction of control variables in columns II and III, the coefficient of green innovation is still significantly positive, which shows that green innovation is conducive to improving the carbon emission reduction performance of high energy-consuming enterprises. This empirical result is consistent with the findings of Zhou and Liu³⁹. It can be seen that enterprises in traditional high energy-consuming industries such as petrochemicals and chemicals can effectively mitigate the problem of high carbon emissions by implementing green innovation.

Secondly, the regression results of the control variables in Table 2 show that the increase in net profit growth rate, R&D cost rate, environmental regulation intensity and economic development level are all conducive to improving the carbon emission reduction performance of high energy-consuming enterprises, i.e. the stronger the growth and profitability of enterprises, the more conducive they are to promoting the gradual "decoupling" of economic development from carbon emissions, and the higher the environmental regulation intensity and economic development level of cities, the more conducive they are to improving the carbon emission reduction performance of enterprises. The higher the intensity of environmental regulations and the level of economic development, the more conducive to improving the carbon emission reduction performance of enterprises. The current asset turnover ratio, gearing ratio and industrial structure have a negative impact, as follows: the financial risk faced by enterprises increases with the gearing ratio, which is not conducive to improving the carbon emission reduction performance of energy-consuming enterprises; the operational capacity of enterprises becomes stronger with the increase in current asset turnover ratio, which leads to more carbon emissions from production activities, thus reducing the carbon emission reduction performance of enterprises; the secondary industry consumes more energy and causes more pollution than other The secondary sector consumes more energy and causes more pollution than other industries, and the larger the share of the secondary sector in GDP, the greater the carbon emissions.

Robustness tests

Endogeneity tests

Endogeneity is an issue that cannot be ignored in empirical studies as it can undermine the 'consistency' of parameter estimates. The main sources of endogeneity are measurement error in variables, omitted variable bias and reverse causality, so the following methods are used to deal with this problem to ensure that the baseline regression results are robust.

First, measurement error. In order to analyse whether measurement error affects the regression results, this paper replaces the measurement of green innovation and enterprise carbon emission reduction performance to re-run the regression, green innovation is measured by adding up the environmental management scores disclosed by listed enterprises plus one and taking the logarithm and the sum of all items occurring in green process innovation in the year plus one and taking the logarithm, and enterprise carbon emission reduction performance is measured by using the dummy of whether the enterprise has received environmental protection recognition from the government The regression results are presented in columns I–III of Table 3. The regression results in columns I-II show that the baseline regression is robust to both management and technical measures of green innovation. The regression results in column III show that the baseline regression is still robust after replacing the measure of carbon reduction performance. In summary, the sign and significance of the coefficients for green innovation are generally consistent with the base regression after accounting for the measurement error of the core and explanatory variables, confirming the robustness of the baseline regression results.

Table 3 Robustness test of measurement error and missing variable bias variables.

Secondly, omitted variables. Considering that omitted variables may also pose endogeneity problems, this paper further incorporates firm-level firm age (measured using firm establishment) and firm size (measured using firm actual total assets) as well as city and industry fixed effects, and column IV of Table 3 reports the regression results addressing the omitted variable issue. As can be seen from column IV of Table 3, green innovation still positively affects the carbon reduction performance of energy-intensive firms, indicating the robustness of the baseline regression findings.

Third, reverse causality. Since green innovation may have an inverse causal relationship with the carbon reduction performance of energy-consuming firms and lead to endogeneity problems, instrumental variables are used to overcome this. In the existing literature, green product innovation is a valid instrumental variable. On the one hand, green product innovation by enterprises mainly involves the design and improvement of green ideas for products, while green patent applications by enterprises are mainly inventions and utility patents based on green innovation, so they satisfy the "correlation" assumption of the instrumental variable. "On the other hand, green product innovation is the design and improvement of a product. On the other hand, green product innovation is the development of products with the aim of providing new products to users, and is not directly related to corporate green patent applications, thus satisfying the "exogeneity" hypothesis of the instrumental variable. In addition, the lagged term of the explanatory variable can also be used as an instrumental variable⁴⁰. Therefore, in this paper, green product innovation (GrePro) and the ratio of lagged green patent applications to total patent applications (L.GreRatioA) are selected as instrumental variables for green innovation. Table 4 reports the results of the tests using the instrumental variables approach.

Table 4 Test of instrumental variables of green innovation affecting carbon emission reduction performance of energy-intensive enterprises.

As can be seen from Table 4, the coefficients of the ratio of green product innovation and lagged one-period green patent applications to total patent applications as joint instrumental variables for green innovation are significant for the ratio of green product innovation and lagged one-period green patent applications to total patent applications in column I, and significant for the pseudo-identification, weak instrumental variable and endogeneity tests as well as insignificant for over-identification, indicating that the joint instrumental variables selected are reasonable and valid. In addition, as can be seen from column II, the coefficient of green innovation is still significantly positive and can therefore confirm the robustness of green innovation in promoting the carbon reduction performance of high energy-consuming enterprises.

Other robustness tests

To further ensure the robustness of the baseline regression results, this paper also applies the following two tests. First, as the standard error clustering bias may lead to misspecification of green innovation significance results, the standard errors are re-clustered so that they are locked at the industry and city levels. Secondly, considering that dynamic panel selection bias may lead to biased estimation results, this paper uses the systematic GMM method to further test the impact of green innovation on firms' carbon reduction performance. The regression results are presented in Table 5.

Table 5 Robustness test of standard error clustering deviation and dynamic panel selection deviation.

As can be seen from columns I–II of Table 5, the sign and significance of the coefficients on green innovation did not change significantly after the standard errors were re-clustered at the city and industry levels, confirming the robustness of the findings. In addition, dynamic panel selection bias may also affect the baseline regression results, but as can be seen from column III of Table 5, AR(1) is significant but AR(2) is not, and the Sargan test is not significant, indicating reasonable model selection. Further analysis shows that the coefficient of the first-order lagged term of enterprises' carbon emission reduction performance is significant, indicating that changes in enterprises' carbon emission reduction performance in the previous period have an impact on the current period. The regression coefficient for green innovation remains significantly positive, which also remains consistent with the baseline regression results of this paper.

Testing the moderating effect of digital transformation

Empirical tests of moderating effects

"The 14th Five-Year Plan is a window of opportunity for China to reach its carbon peak, and there is an inherent relationship between green innovation and digital transformation as an effective means to achieve both economic and environmental benefits. Therefore, considering that the impact of green innovation on enterprises' carbon emission reduction performance may be influenced by digital transformation, this paper introduces the moderating variable of digital transformation to investigate its moderating effect on the relationship between green innovation and enterprises' carbon emission reduction performance, and the regression results are shown in Table 6.

Table 6 Regression results of the relationship among green innovation, digital transformation and enterprise carbon emission reduction performance.

As shown in column I of Table 6, the coefficients of green innovation and the coefficient of the interaction term between green innovation and digital transformation are both significantly positive, indicating that digital transformation can promote the positive impact of green innovation on the carbon reduction performance of enterprises, and the hypothesis H₂ is tested. Columns II and III of Table 6 show the regression results by replacing DigTra with DigAss and DigMda respectively. The sign and significance level of the estimated coefficients in columns II and III do not change fundamentally, further indicating that digital transformation positively moderates the impact of green innovation on firms' carbon emission reduction performance. The above results may be attributed to the following reasons:Firstly, digital transformation has given data new resource attributes, weakened the negative factor circulation effect of information asymmetry, helped to accelerate the flow of factors, improved the comprehensive allocation efficiency of enterprise resources, further generated the motivation for green innovation and changed the way of green innovation, thus enhancing the carbon emission reduction performance of enterprises. Secondly, the digital transformation has improved the production and operation control ability of enterprises, which can realize real − time monitoring of the whole process of production and emission, and enterprises can effectively achieve green and low-carbon development through clean production technology or end-of-pipe treatment technology in response to the emission situation.

Heterogeneity tests for moderating effects

Considering that different industries and different types of enterprises have different characteristics, resulting in different paths of green innovation and digital transformation, the impact of digital transformation on their green innovation and carbon emission reduction performance is somewhat heterogeneous. Therefore, this paper estimates the relationship between green innovation, digital transformation and carbon emission reduction performance of six types of industries from the perspective of industry heterogeneity, and the regression results are shown in Table 7.

Table 7 Industry heterogeneity analysis.

As can be seen from Table 7, the coefficients of the interaction terms are insignificant, except for two industries, namely chemical raw materials and chemical products manufacturing and electricity and heat production and supply, where the moderating effect of digital transformation on green innovation and corporate carbon reduction performance is significantly positive, indicating that for companies in different industries, the moderating effect of digital transformation reflects heterogeneity. Among them, C26 and D44 have significant carbon reduction effects in their industries under the effect of greening and digital integration, but C25, C30, C31 and C32 are not yet significantly influenced by green innovation and digital transformation. According to the "China Enterprise Digital Transformation Research Report (2020)", manufacturing enterprises are the "first movers" in digital transformation among the participating enterprises, accounting for 42.3% of the total, with the highest percentage of 20% in the chemical raw materials and chemical products manufacturing industry. The digital transformation of enterprises in the chemical raw materials and chemical products manufacturing industry has entered a virtuous cycle, promoting the green and low-carbon development of enterprises through digital transformation. Compared with other industries, the electricity and heat production and supply industry can generate a large amount of data in real time and has the technical basis to adapt to the Industrial Internet. Based on the cloud platform, it realises intelligent management of power supply and heat supply equipment, monitors power supply and heat supply in real time, and helps enterprises to realise precise regulation of power supply and heat supply and efficient application of energy, and its digital transformation is gradually taking effect. The impact of the transformation is not yet obvious. Therefore, there is an urgent need to apply the new generation of information technology to the production and management processes in a targeted manner according to the characteristics of different high energy-consuming industries and the differences in the degree of digitalisation of each industry and the stage they are in, so as to gradually improve the digital capability of each industry and enhance the efficiency of energy use through digital technology, thereby realising the effect of emission reduction.

Conclusions and policy recommendations

Conclusions

Considering that green innovation and digital transformation are important ways to achieve carbon peaking and carbon neutrality in China, this paper uses 452 high energy-consuming listed enterprises as a research sample and uses a double fixed-effects model to explore the impact effect of green innovation on the carbon emission reduction performance of high energy-consuming enterprises and the moderating effect of digital transformation. The findings show that, firstly, green innovation has a positive effect on the level of carbon emission reduction performance of high energy-consuming enterprises, and the results are robust when tested by endogeneity, standard error clustering bias and dynamic panel selection bias. Secondly, digital transformation can improve the contribution of green innovation to the carbon emission reduction performance of high energy-consuming enterprises. Third, the heterogeneity of the impact of digital transformation on the relationship between green innovation and the carbon emission reduction performance of energy-intensive enterprises is due to the differences in the characteristics of industries. The impact of green innovation and digital transformation on the processing industry is not yet evident.

Policy recommendations

In order to improve the green innovation capability, digital transformation and carbon emission reduction performance of high energy-consuming enterprises, this paper makes the following recommendations: First, the government should formulate a structural reform plan for greening and digitalisation of high energy-consuming industries, accelerate the promotion of new infrastructure with greening and digitalisation as the core, and promote the deep integration of green innovation and digitalisation to provide conditions for high energy-consuming enterprises to promote the transformation of the two. At the same time, the government should also promote basic research related to green innovation and digitalisation in higher education institutions and research institutes, focus on training talents for green innovation and digital transformation, and provide enterprises with talents with green and digital literacy, thereby enhancing the carbon emission reduction performance of enterprises through green innovation and digital transformation and accelerating the major transformation from high carbon to low carbon and then from low carbon to carbon neutral. Secondly, high-energy-consuming industries should formulate targeted green and digital transformation programmes according to their industry characteristics and development stages, and adopt a phased and step-by-step approach to gradually realise the green and digital transformation of each industry under the principles of problem-orientation and urgent use first. In addition, high-energy-consuming industries should also take into full consideration the laws of industry development, promote energy-saving and carbon-reducing transformation in a scientific and orderly manner, avoid "campaign-style" energy-saving and carbon-reducing, and emphasise "establishing first and then breaking", and first do a good job of green innovation and digital transformation infrastructure for carbon-reducing, and then go for coal This will in turn promote a comprehensive shift towards green development in energy-intensive industries. Third, high energy-consuming enterprises should actively build a low-carbon culture, establish a sound green management system and digital management system, and maximize the participation of managers and even employees in the green and digital transformation to promote green and low-carbon development. At the same time, high energy-consuming enterprises should invest in green innovation and digital transformation in terms of capital and talent as soon as possible, promote green and digital transformation from product, production and management in an orderly manner, strive to build a "green and digital" dance platform, fully combine source control and end-of-pipe management, improve the efficiency of green innovation, and ultimately achieve carbon emission reduction. We are committed to fulfilling our responsibilities in energy conservation and emission reduction, and ultimately achieving the goal of carbon peaking and carbon neutrality.