When enterprises engage in green innovation, two primary approaches in the technological domain are observed. The first involves upgrading and modernizing existing technologies. For instance, enterprises focus on updating current equipment through continuous collaboration with technology suppliers, ensuring ongoing advancements in environmental technologies. Green technological innovation also facilitates the restructuring and optimization of industrial sectors, promoting the transformation of energy-intensive industries. For example, introducing advanced energy-saving technologies and equipment enables sectors such as steel and chemicals to significantly reduce energy consumption per unit of output, achieving a green transition. By consistently upgrading existing clean energy technologies, enterprises can effectively decrease their reliance on fossil fuels and enhance overall energy efficiency (Akther et al., 2024). Additionally, in the realm of production processes, enterprises leverage digital technologies—such as the industrial internet and artificial intelligence—to improve production design processes, retrofit existing equipment, optimize facility management, and achieve intelligent production. These advancements reduce marginal costs and further enhance energy efficiency.

The second approach involves adopting new technologies. This is exemplified by introducing cutting-edge energy-efficient equipment and employing the latest energy-saving technologies to reduce energy consumption. The integration of digital and intelligent technologies is also promoted (Gao et al., 2024). By utilizing digital technologies such as the Internet of Things (IoT), big data, and cloud computing, enterprises can monitor and control energy usage in real-time. Data analytics and intelligent algorithms optimize energy allocation (Hu J. et al., 2024; Hu S. et al., 2024), while advanced AI algorithms predict energy demand, enabling dynamic adjustments to energy supply strategies and minimizing unnecessary waste. Moreover, enterprises actively participate in the research and development of clean energy technologies, such as hydrogen and biomass energy, to explore diversified clean energy solutions. The development of efficient energy conversion technologies, including advanced internal combustion engines, high-efficiency motors, and inverters, also significantly reduces energy losses during conversion processes.

In general, industrial carbon emissions primarily originate from the combustion of fossil fuels (Sun and Onuh, 2024), which are used either as energy sources or as raw materials in production processes. Clearly, improving energy utilization efficiency can reduce carbon emissions at the source (Liu et al., 2022b). Based on this, the following hypothesis is proposed:

H2

Green innovation reduces enterprise carbon emissions by enhancing energy efficiency.

Enterprises engaged in green innovation typically develop systematic strategic plans that define clear goals, steps, and divisions of responsibility. Such well-defined planning enhances internal control mechanisms, facilitating more effective managerial coordination (Wang et al., 2023). During this process, enterprises perform accurate assessments of their resources, gaining a clearer understanding of their green innovation directions. This enables the adoption of targeted management measures to achieve strategic objectives. Additionally, green management innovation effectively mitigates environmental risks. By optimizing processes, enterprises improve their resilience to risks, enabling more efficient and timely responses to policy changes and market demands, thereby increasing the flexibility and precision of managerial specialization. Moreover, when implementing green innovation, enterprises enhance execution methods and performance standards, further professionalizing managerial practices. Green innovation also extends beyond internal operations to encompass external supply chain networks. Green initiatives along the supply chain are often achieved through collaboration among multiple enterprises or departments. To ensure these initiatives are realized, comprehensive inter-organizational cooperation is fostered, providing organizational support for activities. The coordination, task allocation, progress monitoring, and performance evaluation among different enterprises or departments compel enterprises to improve managerial specialization.

As managerial specialization advances, production strategies are adjusted to prioritize environmental impacts over purely economic objectives, promoting low-carbon production practices. Enhanced managerial specialization also leads to more stringent employee selection and training criteria, incorporating dimensions such as green experience and awareness. Enterprises allocate more budgets to employee incentives, fostering creativity and motivation for green initiatives and deepening efforts in low-carbon and emission reduction practices. Employees selected under such criteria are more committed to sustainable development, proactively adopting green technologies to reduce carbon emissions and support long-term sustainability goals. Implementing green supply chain management enhances the environmental performance of the entire production network, satisfying consumer demands for green products across diverse economies and complying with environmental regulations in various export markets. This optimizes supply chain structures, reduces carbon emissions across the supply chain, and fundamentally improves carbon emission performance. Based on this, the following hypothesis is proposed:

H3

Green innovation reduces enterprise carbon emissions by enhancing the professionalism of enterprise management.

The relationships among the three hypotheses proposed in this study are shown in Figure 1.

Considering data availability, this study examines A-share listed enterprises from 2000 to 2022. The data sources for this study are as follows:

First, Green innovation data from 2000 to 2022 year: this data originates from the State Intellectual Property Office (SIPO) of the People’s Republic of China. In 2010, the World Intellectual Property Organization (WIPO) introduced the International Patent Classification Green Inventory, an online tool to facilitate the search of environmentally friendly technology patents. Based on the United Nations Framework Convention on Climate Change, this classification divides green patents into seven main categories. Following this classification, we calculate the annual number of green patents for each enterprise. Second, Carbon emission data from 2000 to 2022 year: the annual carbon emissions data are calculated based on enterprises’ disclosures in social responsibility reports, sustainability reports, environmental reports, and the “Guidelines for Enterprise Greenhouse Gas Emissions Accounting and Reporting.” Third, Enterprise characteristic variable from 2000 to 2022 year: these variables are sourced from the China Stock Market and Accounting Research (CSMAR) database. Given data availability, this study selects A-share listed companies on the Shanghai and Shenzhen stock exchanges from 2000 to 2022 as the research sample. To ensure the reliability of the sample data and the accuracy of the regression results, the following procedures are implemented:

(1) Companies in the financial and insurance sectors are removed, as these enterprises do not operate within the real economy and have unique business models.

To investigate the impact of green innovation on enterprise carbon emissions, this study constructs the following econometric model for empirical testing: C E f t = a 0 + a 1 G I f t + a 2   Z + φ t + θ f +   ε f t (1)

In Equation 1, f and t represent the enterprise and year, respectively. The dependent variable, CE _ft , indicates the carbon emissions level of enterprise f in year t. The explanatory variable, GI _ft , reflects the overall internal control status of enterprise f in year t. Z denotes a set of control variables, φ _t represents the year fixed effects, θ _f denotes the enterprise fixed effects, and ε _ft refers to the random error term.

Table 2 presents the baseline regression results. Column (1) shows the regression outcome of GI and CE without any control variables, while controlling for year and enterprise fixed effects. Column (2) incorporates a series of control variables on this basis. The empirical results indicate that, whether or not control variables are included, the estimated coefficient of the explanatory variable GI is significantly negative at the 1% level. This finding demonstrates that green innovation by enterprises can significantly reduce carbon emissions among Chinese enterprises.

To verify the stability of the empirical results and enhance their credibility, four robustness tests are conducted in this study: substituting explanatory variables, controlling for industry fixed effects, clustering robust standard errors, and applying a 1% bilateral winsorization.

To ensure the accuracy of baseline regression, it is essential to address potential endogeneity issues in the econometric model, primarily arising from reverse causality and omitted variables. This paper employs the number of green trade agreements signed by enterprises’ export trade partner countries (Number) as an instrumental variable for green innovation, estimating the relationship using two-stage least squares. Given that the provisions included in international green trade agreements explicitly define environmental standards for various traded products, these agreements influence the types and quantities of green products exported by economies, thereby affecting the level of green innovation within enterprises located in those economies. This fully aligns with the relevance assumption of instrumental variables. Moreover, green trade agreements, signed by different economies to establish, maintain, and develop economic and trade relations while pursuing green benefits, are outcomes of political and economic equilibrium among nations. This characteristic satisfies the exclusivity requirement of instrumental variables. This study uses data from the WTO RTAIS (Regional Trade Agreements Information System) database, collating and processing the number of trade agreements that contain green clauses. As the number of green trade agreements is measured at the national level, this study also weights this variable by the trade volume of each enterprise’s exports to different economies, matching it to the enterprise level.

Table 4 reports the results of two-stage estimations. Column (1) presents the first-stage regression outcomes, where the estimated coefficients of the instrumental variables are all significantly negative at the 1% level. Given that the number of instrumental variables does not exceed the number of explanatory variables and the F-statistic values of the instrumental variables are well above the critical value at the 10% significance level in the Stock-Yogo test, weak correlation issues are excluded. Column (2) provides the second-stage regression results, showing that the sign and significance of the estimated coefficient for GI remain consistent with those in the baseline regression, further validating the findings of this study.

Green innovation includes applications for both green invention patents and green utility patents, each differing in technological innovation level and scope of protection, which may affect enterprises’ carbon emissions reduction in varying ways. Compared to green utility patents, green invention patents generally involve higher levels of technological R&D, a more complex review process, and a broader scope of protection, encompassing various aspects such as manufacturing methods, usage methods, and product structure (Li and Zheng, 2016). Based on this, we hypothesize that applying for green invention patents is more effective in reducing enterprises’ carbon emissions than applying for green utility patents.

This study explores the heterogeneous effects of different types of green innovation on reducing enterprises’ carbon emissions by distinguishing between applications for green invention patents and green utility patents. The quantity of green invention patent applications (INVA) and green utility patent applications (UMA) are used to measure the respective application activity for these types of patents. As shown in Table 5, Columns (1) and (2) report the effects of green invention patents and green utility patents on enterprises’ carbon emissions. Notably, the estimated coefficient for green invention patents is higher than that for green utility patents, supporting our hypothesis.

Given the variations in internal control strength among enterprises with different characteristics (Lin et al., 2024), enterprises may exhibit distinct responses when engaging in green innovation. Compared to non-state-owned enterprises (Non SOEs), state-owned enterprises (SOEs) benefit from stronger administrative protections, operate in more stable market positions, and face less competitive pressure. This, to some extent, reduces their inclination toward green innovation, as the implementation of policies is often limited by management structures and vested interests, resulting in outcomes that fall short of expectations. Consequently, we hypothesize that non-state-owned enterprises achieve greater carbon emissions reductions through enhanced green innovation than their state-owned counterparts.

In this study, following Hsieh and Song (2015), we construct a binary variable for enterprise ownership, defining an enterprise as state-owned if its state capital constitutes 50% or more of its paid-in capital, assigned a value of 1; otherwise, it is categorized as non-state-owned with a value of 0. Columns (1) and (2) of Table 6 report the impact of green innovation GI on CE, showing that the estimated coefficient for non-state-owned enterprises is higher than that for state-owned enterprises, thereby validating the above hypothesis.

Enterprises in industries with varying pollution levels may experience heterogeneity in the impact of green innovation on carbon emissions due to differences in pollution emission structures and policy support across industries. Enterprises in high-pollution industries typically engage in daily production activities associated with significant energy consumption and pollutant emissions, often resulting in higher levels of carbon output. Additionally, the government generally places stricter environmental policies and emission standards on high-pollution industries, incentivizing these enterprises to pursue green innovation more actively. Thus, we propose that green innovation is more effective in reducing carbon emissions for enterprises in high-pollution industries than for those in low-pollution sectors.

This study examines the heterogeneity in the impact of green innovation on enterprise carbon emissions by distinguishing the pollution levels of their respective industries. The dummy variable, POLLUTION, is used to measure pollution levels, with the industry median as the cutoff. Industries with pollution levels above the median are classified as high-pollution (POLLUTION = 1), while those below the median are classified as low-pollution (POLLUTION = 0). Results are presented in Table 7; columns (1) and (2) report the impact of GI on CE, revealing that the estimated coefficient for high-pollution industries exceeds that for low-pollution industries, thereby confirming the hypothesis outlined above.

Enterprises operating in industries with different levels of technological intensity may exhibit varying carbon reduction effects from green innovation due to differences in technological innovation capacity and resource consumption. Compared to enterprises in low-technological-intensity industries, those in high-technological-intensity industries typically possess stronger R&D capabilities and higher innovation capacity, enabling more comprehensive and advanced green innovation. Consequently, the carbon reduction effects of green innovation are more pronounced in high-technological-intensity industries. Furthermore, enterprises in high-technological-intensity industries often face substantial technological investments and higher levels of energy and resource consumption, which provide greater overall emission reduction potential. Based on this, it is hypothesized that the effect of green innovation on reducing carbon emissions is greater for enterprises in high-technological-intensity industries than for those in low-technological-intensity industries.

This study explores the heterogeneous impact of green innovation on enterprise carbon emissions by distinguishing enterprises based on their industry’s technological intensity. Following Lu and Dang (2014), a dummy variable TECH is used to measure technological intensity, calculated as the ratio of R&D expenditure to employee compensation. Industries are classified by the median technological intensity: industries above the median are categorized as high-technological-intensity (coded as 1), while those below the median are categorized as low-technological-intensity (coded as 0). The results, presented in Table 8, show that the estimated coefficients of GI on carbon emissions CE are higher for enterprises in high-technological-intensity industries than for those in low-technological-intensity industries. These regression results confirm the proposed hypothesis.

This study uses the intensity of energy consumption per unit of GDP to assess energy efficiency (Energy). A higher Energy value indicates a greater energy utilization efficiency for the enterprise. Table 9 presents the results of testing this mechanism. Column (1) estimates the impact of GI on Energy, with the coefficient showing a significant positive effect at the 1% level. This finding suggests that enhancing green innovation in enterprises contributes to improved energy utilization efficiency. Column (2) displays the estimated coefficient of Energy on CE, showing a significant negative effect at the 1% level, indicating that increasing an enterprise’s RISK level can significantly reduce its carbon emissions. Therefore, it can be inferred that strengthening green innovation in enterprises can reduce carbon emissions by improving energy utilization efficiency.

Based on the matching data from the WIOD and labor occupational databases, as referenced by Zhong et al. (2021), the forward decomposition method of trade value-added is applied to measure industry-specific management specialization levels through industry factor returns. Following the approach of Chor et al. (2021), management specialization levels for enterprises (Manage) are calculated using a weighted method, where higher values indicate stronger and more effective management. Table 10 presents the results of tests on the above influence mechanism. Column (1) estimates the effect of GI on Manage, with a positive coefficient significant at the 1% level, indicating that enhancing green innovation in enterprises promotes higher management specialization. Column (2) shows a negative and significant coefficient of Manage on CE at the 1% level, suggesting that improving enterprise management specialization significantly reduces carbon emissions. Thus, it can be inferred that strengthening green innovation in enterprises can lower carbon emissions by enhancing management specialization.

Based on panel data from A-share listed enterprises in China from 2000 to 2021, this study examines the impact of green innovation on enterprise carbon emissions. It also explores the heterogeneity of these effects by green innovation type, industry pollution level, and enterprise ownership, and investigates the specific mechanisms through which green innovation affects carbon emissions. The findings reveal that:

(1) Green innovation significantly reduces enterprise carbon emissions, a conclusion that remains robust across various analyses. This reduction primarily occurs through two channels: improved energy efficiency and enhanced enterprise management specialization.

This paper provides empirical evidence identifying the impact of green innovation on enterprises’ carbon emissions, offering new insights for promoting sustainable enterprise development. Based on these findings, the following recommendations are proposed:

(1) The results of this study indicate that green innovation significantly reduces enterprise carbon emissions and enhances environmental benefits. Therefore, enterprises should proactively implement green innovation in the future. Internally, enterprises are encouraged to integrate the concept of green innovation into their corporate culture, embedding green elements such as environmental protection and low carbon into their core values and business strategies. This approach should promote awareness of green innovation among all employees. When formulating development strategies, enterprises are advised to move beyond solely pursuing economic benefits and increasingly prioritize environmental performance, aiming for the synergistic development of economic and environmental benefits. Externally, enterprises can consider learning from international best practices in green innovation and enhancing their own capabilities in this field. Collaboration with trading partners on research and application development in green innovation is also recommended.