Green Technology Innovation and Corporate Carbon Performance: Evidence from China

Abstract

Against global carbon neutrality goals and China’s “dual carbon” strategy, this study examines how green technology innovation shapes corporate carbon performance through a dual-path mechanism—improving enterprises’ resource utilization efficiency and environmental governance capabilities. Leveraging data from Chinese A-share listed firms (2007–2022) and methods including fixed effects, instrumental variables, and Heckman two-stage models, key findings include: (1) Green technology innovation significantly improves carbon performance. (2) This effect operates through two pathways: enhancing total factor productivity (TFP) and strengthening environmental governance. (3) Green media and investor attention amplify the positive impact of green innovation on carbon performance. (4) The effect remains significant but shows diminishing marginal returns over 1–4 future periods. (5) Non-state-owned enterprises and non-high-carbon industries exhibit more pronounced improvements. This research provides micro-level evidence for “technology-driven low-carbon transformation”, offering theoretical support for policy differentiation and corporate green technology strategies, with practical implications for achieving China’s “dual carbon” objectives.

1. Introduction

In the global sustainable development paradigm, addressing climate change and fostering inclusive economic transformation have emerged as interlocking imperatives for nations and corporations alike. As the world’s largest carbon emitter and manufacturing hub, China’s 2020 pledge to achieve “carbon peak and carbon neutrality” (the “dual carbon” goals) exemplifies a national strategy explicitly aligned with the United Nations Sustainable Development Goals (SDGs). This commitment underscores green technology innovation as the linchpin for realizing a triple dividend of environmental sustainability, economic competitiveness, and social responsibility [1,2].

Green technology innovation serves as the cornerstone of sustainable development by reconciling environmental stewardship with economic growth. It enables firms to reduce carbon footprints through energy-efficient technologies while reshaping industrial ecosystems via circular economy models, directly contributing to SDG 7 (affordable and clean energy) and SDG 13 (climate action). However, as China navigates its sustainable development transition, critical gaps remain in understanding how green innovation translates into tangible carbon performance improvements, particularly through mechanisms of resource allocation and organizational capability building. Equally important is unpacking how industry-specific contexts and firm characteristics moderate this relationship, insights that are essential for refining the theoretical foundations of sustainable development.

Answering these questions not only advances scholarly understanding of the innovation–sustainability nexus but also meets pressing practical needs. For enterprises, clarifying these linkages can guide strategic investments in green technologies to fulfill environmental, social, and governance (ESG) responsibilities. For policymakers, evidence-based insights into effective green innovation pathways can strengthen regulatory frameworks supporting global climate governance. This study thus contributes to both the academic discourse on sustainable development and the practical implementation of SDG-aligned strategies in emerging economies.

Existing studies have partially confirmed the positive effects of green technology innovation on environmental performance, but two key limitations persist. First, most research focuses on the relationship between green technology innovation and financial performance or traditional environmental metrics [3,4,5]. Less attention has been paid to “carbon performance”, a core dimension directly aligned with global climate goals, especially regarding subdivided indicators such as carbon emission intensity and carbon reduction efficiency. Second, the literature lacks systematic analysis of institutional specificities in the Chinese context, such as carbon market development and variations in environmental regulation intensity, as well as enterprise heterogeneity, including ownership structures and industry-specific carbon emission characteristics. This limits our understanding of the internal logic through which green technology innovation drives improvements in carbon performance.

Against China’s transitional economic background, this study situates green technology innovation within a multidimensional sustainable development framework. By exploring its impact pathways on corporate carbon performance, this research aims to bridge the explanatory gap in existing theories under specific institutional contexts. It also seeks to provide transferable “Chinese experience” for global sustainable development.

Using microdata from Chinese listed companies, this study empirically examines the impact of green technology innovation on corporate carbon performance and its underlying mechanisms. It further investigates the moderating effects of external supervision (e.g., green news coverage, investor attention) and the influences of heterogeneous enterprise conditions (e.g., ownership structure, industry characteristics). The findings reveal that green technology innovation improves corporate carbon performance by enhancing total factor productivity and environmental governance capabilities, with external supervision strengthening this effect. Heterogeneous enterprise conditions also shape the relationship significantly.

The research provides empirical support for targeted policy-making under China’s “dual carbon” goals, facilitating the construction of a low-carbon development pathway characterized by “technology-driven innovation—external supervision empowerment—heterogeneous entity adaptation.” This contributes to advancing the Sustainable Development Goals and promoting practical innovation at the micro-enterprise level.

The remainder of the paper is structured as follows: Section 2 reviews the literature; Section 3 develops research hypotheses; Section 4 describes the research design; Section 5 reports empirical results; and Section 6 concludes with discussions, research findings, and policy implications.

2. Literature Review

Against the intertwined backdrop of the SDGs and China’s “dual carbon” strategy, green technology innovation has emerged as a core issue in resolving the contradiction between economic growth and carbon emissions. As the primary driver of corporate low-carbon transformation, green technology innovation not only embodies expectations for technological breakthroughs and industrial upgrading but also shoulders the dual mission of environmental improvement and sustainable development. However, existing research has a theoretical gap in understanding how green technology innovation influences enterprises’ core environmental performance, especially “carbon performance”, which directly links to climate goals. This literature review proceeds from three aspects to clarify the logical connections between green technology innovation and corporate carbon performance.

2.1. Consequences of Green Technology Innovation

As a core driver integrating environmental goals and technological progress, research on the consequences of green technology innovation focuses primarily on its dual impacts on environmental and economic performance. From the environmental perspective, existing literature widely confirms that green technology innovation can reduce corporate pollutant emissions through energy-saving and emission-reduction technologies (e.g., clean energy utilization, carbon capture) [6], improve resource efficiency [7], and thereby enhance environmental performance [8]. Economic studies center on the “Porter Hypothesis,” exploring whether green technology innovation can boost financial performance via cost savings and differentiated competitive advantages [9,10]. Yet, debates persist: some scholars argue that environmental investments may crowd out R&D resources, negatively impacting short-term financial performance [11], while long-term synergies gain more empirical support [12].

2.2. Determinants of Corporate Carbon Performance

Enhancing corporate carbon performance results from the combined effects of multiple internal and external factors. Internally, the resource-based view (RBV) highlights the role of heterogeneous resources, including technological innovation capabilities [13], environmental management systems [14], and executives’ environmental strategic orientation [15]. Externally, institutional environments and market pressures are key: environmental regulations (e.g., carbon pricing, emissions trading schemes) drive emission reduction through compliance pressures [16], while stakeholder supervision (e.g., investors’ green preferences, consumers’ environmental awareness) incentivizes reductions via reputation mechanisms [17]. Notably, China-specific institutional factors, such as state-owned enterprises’ policy response mechanisms [18] and industry carbon emission characteristics [19], are increasingly becoming research focal points.

2.3. The Relationship Between Green Technology Innovation and Corporate Carbon Performance

Early research on the direct link between green technology innovation and carbon performance often adopted a “technological determinism” perspective, assuming a linear positive correlation between R&D intensity and carbon reduction effects [20]. As studies deepened, scholars began examining intermediary mechanisms, such as green technology innovation reducing emissions by optimizing energy structures [21], improving production processes [22], or achieving carbon reduction and efficiency enhancement through total factor productivity improvements [23]. Some research introduces moderating variables, finding that environmental regulation intensity [24] and enterprise absorptive capacity [25] influence the effect of green technology innovation on carbon performance.

While existing literature provides important foundations for understanding this relationship, especially regarding the environmental effects of technology innovation and multidimensional determinants of carbon performance, gaps remain: First, most studies focus on single impact mechanisms (e.g., technological efficiency improvements), ignoring synergies between green technology innovation, environmental governance capabilities, and resource integration, which limits explanations for why some high-tech-input firms exhibit poor emission reduction outcomes. Second, there is no systematic conclusion on how external stakeholders like media and investors amplify or dampen the emission reduction effects of green technology innovation through supervision mechanisms, particularly the impact of market supervision innovations under China’s “dual carbon” goals. Third, the lack of in-depth analysis of Chinese enterprises’ ownership differences and industry carbon emission characteristics restricts the practical guidance of research conclusions for local practices. Therefore, this study examines the impact of green technology innovation on corporate carbon performance in the Chinese context, aiming to expand research boundaries and provide theoretical insights for low-carbon transitions in emerging economies.

3. Research Hypotheses

Against the backdrop of global low-carbon transformation and the strengthening of corporate environmental responsibilities, green technology innovation, as a crucial link connecting technological progress and sustainable development, urgently requires an in-depth theoretical deconstruction of its impact mechanism on corporate carbon performance. Although existing studies have preliminarily confirmed the correlation between them, the answers to core questions such as “why innovation can enhance carbon performance” and “through what pathways it can be achieved” remain fragmented. This work, based on the technological innovation theory, endogenous growth theory, and the Resource-Based View (RBV), systematically expounds the internal logic of how green technology innovation acts on carbon performance.

Starting from the essence of the technological innovation theory, green technology innovation forms a direct driving effect on carbon performance by integrating environmental goals with technological research and development. As a core element for enterprises to build low-carbon competitiveness, this work argues that green technology innovation has a significant positive impact on carbon performance for the following reasons.

First, in terms of technological attributes, green technology innovation is oriented towards reducing environmental negative externalities and covers core fields such as pollution control, energy efficiency improvement, and utilization of renewable energy [26]. For example, clean production technologies reduce carbon emissions in intermediate processes by optimizing the production flow, and Carbon Capture and Storage (CCS) technologies directly block the emission of greenhouse gases in industrial waste gases [27]. The research, development, and application of such technologies can reduce the carbon footprint from the source of production, forming a direct conversion chain of “technological input–emission reduction output”.

Second, in terms of institutional response, based on the policy incentive logic of the “Porter Hypothesis” [28], environmental regulations and resource favoritism under China’s “dual carbon” goals prompt enterprises to concentrate their innovation resources in the green field. This institutional drive not only reduces the innovation costs of enterprises but also forces technological upgrading through compliance pressure, forming a closed loop of “policy guidance–innovation input–performance improvement”.

Third, in terms of stakeholder interaction, according to the stakeholder theory, consumers’ preference for low-carbon products, investors’ attention to ESG performance, and the media’s supervision of environmental responsibilities together constitute the external pressure for enterprises to improve their carbon performance. As a core means to respond to the above demands [29], green technology innovation can not only obtain market share through product differentiation but also enhance corporate reputation through environmental legitimacy, ultimately forming a virtuous cycle of “external pressure–innovation response–performance improvement”.

The endogenous growth theory reveals that technological innovation enhances Total Factor Productivity (TFP) through knowledge accumulation and optimization of factor allocation, thereby achieving the dual goals of “carbon reduction and efficiency enhancement” [30]. We argue that green technology innovation improves corporate carbon performance through the mechanism of TFP by promoting knowledge spillover and production efficiency improvement.

On the one hand, in terms of process optimization in process innovation, green process innovation reduces energy consumption and waste generation by reconstructing the production process [31]. For example, the application of dry quenching technology in steel enterprises can reduce energy loss by more than 30%, and chemical enterprises can achieve real-time monitoring and dynamic adjustment of carbon emissions through process digitization. Such innovations directly reduce the carbon emission intensity per unit of output value by improving the input–output efficiency.

On the other hand, in terms of the linkage on the demand side of product innovation, green product innovation not only meets the low-carbon needs of end consumers but also drives upstream suppliers to improve production technologies through the conduction effect of the industrial chain [32]. For example, the breakthrough in new energy vehicle battery technology has prompted cathode material manufacturers to develop low-energy consumption preparation processes, forming a full-chain efficiency improvement of “downstream demand guidance–upstream technology response”. This technology diffusion effect is ultimately reflected in the optimization of the carbon efficiency of the entire industrial system.

The Resource-Based View (RBV) emphasizes that the heterogeneous resources and capabilities of enterprises are the sources of competitive advantages. As a core resource, green technology innovation forms a long-term impact on carbon performance by promoting the upgrading of environmental governance capabilities [33]. This study argues that green technology innovation can systematically enhance the effectiveness of corporate carbon governance by promoting institutional arrangements, the iteration of management tools, and the formation of organizational routines.

First, when enterprises carry out green technology research and development, they often establish supporting environmental management systems simultaneously. These institutional arrangements provide organizational guarantees for carbon emission monitoring and the decomposition of emission reduction targets [34]. For example, enterprises implementing carbon footprint accounting technology usually establish cross-departmental collaboration mechanisms to ensure the deep integration of technological achievements and management processes.

Second, the achievements of green technology innovation (such as carbon accounting software and environmental data analysis platforms) provide precise tools for environmental governance. For instance, the application of blockchain technology in carbon emission traceability improves data transparency, and the application of artificial intelligence algorithms in energy scheduling optimizes the efficiency of emission reduction decision-making [35]. The application of these tools enables enterprises to shift from “passive compliance” to “active governance”.

Third, the environmental knowledge accumulated during the technological innovation process gradually internalizes into organizational routines [36], forming a closed-loop governance system of “monitoring–evaluation–improvement”. For example, enterprises that continuously invest in green technology often have stronger environmental learning capabilities and can achieve a spiral improvement of carbon performance through technological iteration and management innovation.

Based on the above analysis, we propose the following hypothesis:

Hypothesis 1.

Green technology innovation is significantly positively correlated with corporate carbon performance.

4. Research Design

4.1. Sample Selection and Data Sources

This paper takes Chinese listed companies on the Shanghai and Shenzhen A-share markets from 2007 to 2022 as the research objects. Referring to existing studies [37], this section processes the samples as follows: (1) Since the financial data, structures, and indicators of financial enterprises such as banks and securities firms differ significantly from those of other industries, financial enterprises with industry codes starting with “J” are excluded according to the Industry Classification Guidelines for Listed Companies issued by the China Securities Regulatory Commission in 2012; (2) Companies with abnormal listings such as ST and PT are excluded; (3) Samples with missing values and outliers are removed; (4) Samples in which the data on green technology innovation and corporate carbon performance are only available for one year are deleted; (5) To avoid the impact of extreme values on the estimation results, all continuous variables are winsorized at the 1% level at both the upper and lower ends. Finally, a total of 22,312 annual-company observations are obtained. In this section, except for the data on local fiscal environmental protection expenditures and regional industrial structures, which are sourced from the National Bureau of Statistics, all other data come from the China Stock Market & Accounting Research Database (CSMAR).

4.2. Variable Definition

4.2.1. Explained Variable

The explanatory variable in this paper is corporate carbon performance (CP). Existing studies mainly measure corporate carbon performance in terms of the absolute value of carbon emissions, carbon intensity, and carbon efficiency [38,39]. Among them, the absolute value of carbon emissions is easily affected by the scale of the enterprise, making it difficult to conduct horizontal comparisons; although carbon intensity takes into account the factor of output value, it only reflects the relative scale of carbon emissions and fails to comprehensively reflect the economic and environmental synergistic effects of low-carbon development. We use the ratio of operating revenue to carbon emissions as a proxy indicator for corporate carbon performance. This indicator, through the operating revenue created per unit of carbon emissions, directly reflects the enterprise’s ability to achieve value addition while reducing carbon emissions, which is in line with the connotation of “low consumption and high output” in green development. The calculation formula is as follows:

$${CP}_{i,t}={OR}_{i,t}/{CE}_{i,t}$$

(1)

Here, ${CP}_{i,t}$ represents the corporate carbon performance; ${OR}_{i,t}$ represents the operating revenue of the enterprise; and ${CE}_{i,t}$ represents the carbon emissions of the enterprise, which is calculated based on the measurement of industry carbon emissions. The calculation formula is as follows:

$${CE}_{i,t}={OC}_{i,t}/{OC}_{j,t}\times {CE}_{j,t}$$

(2)

Here, ${OC}_{i,t}$ is the operating cost of the enterprise; ${OC}_{j,t}$ is the total operating cost of the industry in which the enterprise is located; and ${CE}_{j,t}$ is the carbon emissions of the industry in which the enterprise is located. Industry carbon emissions are calculated by multiplying the consumption of energy sources such as coal, coke, crude oil, gasoline, kerosene, diesel, fuel oil, and natural gas by their respective carbon dioxide emission factors.

The Carbon Performance (CP) indicator effectively captures both short-term and long-term effects through its “revenue–emission” ratio. In the short term, enterprises can enhance revenue generation per unit of carbon emission by optimizing production processes and controlling emissions without significantly altering emission scales, thereby reflecting resource utilization efficiency and policy responsiveness. Over the long term, the indicator measures the lagged impacts of strategic initiatives (e.g., low-carbon technology investments, industrial transformation) on sustained emission reduction and revenue growth via intertemporal dynamic equilibrium. Additionally, it tracks enterprises’ relative performance in industry-wide emission reduction trends through sectoral emission allocation mechanisms, aligning with the long-term goal of “economic–environmental synergistic development”.

The choice of using the operating revenue-to-carbon emissions ratio as a proxy for CP is grounded in environmental economics and resource-based view (RBV) theory. From an economic efficiency perspective, this indicator addresses the limitations of traditional carbon performance metrics: absolute emissions are confounded by firm scale (violating the law of comparative advantage in microeconomics), while carbon intensity ignores the “value-added” dimension of low-carbon development. By integrating revenue (a core economic outcome) with emissions (an environmental cost), CP reflects the “double dividend” of green development, measuring how effectively enterprises convert carbon inputs into economic outputs. This aligns with the RBV, where sustainable competitive advantage stems from efficiently combining environmental responsibility with resource utilization.

4.2.2. Explanatory Variable

The explanatory variable in this work is green technology innovation (GP). Existing literature mainly measures green technology innovation in two categories: one is input-side indicators, such as green R&D investment and the number of R&D personnel [40]; the other is output-side indicators, such as the number of green patents and the number of low-carbon technology papers [41]. Considering the uncertainty in the transformation between innovation input and output, and that patents are the legal carriers of technological innovation achievements and the core symbols of market-oriented applications, we use the number of green patents independently applied for by enterprises as the core explanatory variable for the following reasons: Compared with R&D investment, the number of patents, as the “hard output” of innovation activities, can more intuitively measure the enterprise’s ability to transform R&D resources into actual technological achievements. Moreover, patent data can be accurately obtained through the China National Intellectual Property Administration or commercial databases, avoiding possible accounting treatment biases of input indicators.

The use of green patent counts as a measure of GP is justified by innovation economics and property rights theory. Drawing on Schumpeterian innovation theory, patents symbolize market-oriented technological achievements that directly impact production processes and product differentiation, thus affecting carbon performance through efficiency gains (e.g., energy-saving technologies) or substitution effects (e.g., clean energy alternatives). Unlike R&D investment (an input-side indicator prone to measurement noise due to accounting practices), patents serve as “hard” output indicators protected by intellectual property rights, which align with the incentive compatibility principle in institutional economics—enterprises are motivated to disclose patent data truthfully due to legal and competitive pressures. This choice also reflects the marketization of innovation, where patented green technologies can be licensed, sold, or integrated into production systems, creating tangible economic value alongside environmental benefits.

4.2.3. Control Variable

To eliminate the interference of corporate characteristics and the external environment on carbon performance, drawing on existing research, this work selects the following control variables: (1) Management characteristics, including two variables: the environmental protection background of the management and the proportion of management shareholding. Among them, the original data of the environmental protection background of the management comes from the personal resume information of the regulatory layer announced by CSMAR. For example, if the personal resume includes keywords such as “environment”, “environmental protection”, “ecology”, “low-carbon”, “new energy”, “energy conservation”, “green”, “sustainable”, “clean energy”, etc., then the sample is identified as having an environmental protection background, and on this basis, the number of management personnel with an environmental protection background is counted [42]. (2) Corporate characteristics, including variables such as company size, asset–liability ratio, establishment duration, return on assets, proportion of intangible assets, equity balance degree, proportion of independent directors, etc. (3) Industry characteristics, choosing the number of companies within the same industry as the control variable. (4) Regional characteristics, selecting the proportion of environmental protection expenditure in fiscal expenditure and the proportion of the added value of the secondary industry in GDP as control variables.

Management characteristics (environmental background and shareholding proportion) are rooted in principal-agent theory, where managers’ personal expertise (environmental background) and equity stakes (shareholding) influence their preference for green strategies. A management team with environmental expertise is more likely to recognize the long-term economic value of carbon reduction (e.g., avoiding future regulatory costs), while shareholding aligns managers’ interests with shareholders’, reducing short-termism in low-carbon investments. Corporate characteristics (size, leverage, profitability, etc.) draw on finance and industrial organization theories. Firm size captures scale economies in pollution abatement (larger firms may have lower unit emission reduction costs), while leverage reflects financing constraints (high-debt firms may face tighter creditor scrutiny, affecting green investment capacity). Return on assets and intangible asset proportion proxy for resource allocation efficiency and technological readiness, respectively, are both critical for translating green innovation into carbon performance. Industry and regional characteristics are informed by new institutional economics and spatial economics. Industry competition (number of peer firms) influences the “pollution haven” effect or technological spillover dynamics, while regional environmental expenditure and industrial structure (secondary industry GDP share) reflect formal institutional pressures (policy support) and structural constraints (e.g., heavy industry’s inherent carbon intensity), which shape firms’ innovation incentives and emission reduction potentials.

Table 1. Variable definitions.

Type	Symbol	Measurement
Dependent variable	CP	The ratio of the enterprise’s operating revenue to its carbon emissions
Explained variable	GP	The number of green patents independently applied for by the enterprise
Control variable	Menv	The number of board members and senior executives with an environmental protection background
	Mshare	The ratio of the number of shares held by directors, supervisors, and senior executives to the total number of shares
	Size	The natural logarithm of the total assets
	Lev	The ratio of total liabilities to total assets
	Age	The natural logarithm of the enterprise’s establishment duration plus 1
	ROA	The ratio of the net profit to the total assets
	Intan	The ratio of the net amount of intangible assets to the total assets
	Equity	The shareholding ratio of the 2nd to 5th largest shareholders/the shareholding ratio of the largest shareholder
	Indep	The ratio of the number of independent directors to the size of the board of directors
	Comp	The natural logarithm of the number of enterprises within the same industry plus 1
	Gexp	The ratio of the local fiscal environmental protection expenditure of the province where the listed enterprise is registered to the GDP
	Str	The proportion of the added value of the secondary industry in the GDP of the province where the listed enterprise is registered

reports the variable types, symbols, and measurement methods.

4.3. Model Setting

In examining the relationship between green technology innovation and corporate carbon performance, this study employs a Fixed Effects (FE) model to control for unobserved individual heterogeneity that may correlate with explanatory variables. To formally justify the choice of the FE model over the Random Effects (RE) model, a Hausman Test was conducted. The test’s null hypothesis posits that there is no systematic difference between the coefficient estimates of the two models. The results show a chi-squared statistic of 400.20 with 13 degrees of freedom and a p-value of 0.000, strongly rejecting the null hypothesis at the 1% significance level. This indicates that the RE model’s estimates would be inconsistent due to the correlation between individual effects and explanatory variables, whereas the FE model effectively addresses this endogeneity by eliminating time-invariant unobserved heterogeneity. The benchmark model is as follows:

$${CP}_{i,t}={\alpha }_0+{\alpha }_1{GP}_{i,t}+\gamma \sum {Controls}_{i,t}+u_i+v_t+{\epsilon }_{i,t}$$

(3)

Here, ${CP}_{i,t}$ represents the corporate carbon performance; ${GP}_{i,t}$ represents green technology innovation; ${Controls}_{i,t}$ represents the control variables, $u_i$ represents the individual fixed effect; $v_t$ represents the time fixed effect; and ${\epsilon }_{i,t}$ represents the random disturbance term.

5. Empirical Results

5.1. Descriptive Statistics

Table 2. Descriptive statistics.

Variables	N	Mean	sd	Min	p50	Max	Skewness	Kurtosis
CP	22,312	6.715	10.962	0.006	1.312	49.162	2.054	6.552
GP	22,312	0.740	2.801	0.000	0.000	20.000	5.416	34.298
Menv	22,312	1.394	1.976	0.000	1.000	10.000	2.310	8.946
Mshare	22,312	0.114	0.190	0.000	0.001	0.687	1.599	4.276
Size	22,312	22.189	1.263	19.659	22.023	25.922	0.663	3.276
Lev	22,312	0.428	0.194	0.059	0.425	0.893	0.136	2.294
Age	22,312	2.862	0.363	1.386	2.890	3.526	−0.885	4.242
ROA	22,312	0.039	0.061	−0.228	0.038	0.197	−0.993	7.381
Intan	22,312	0.048	0.043	0.000	0.037	0.260	2.548	11.424
Equity	22,312	0.682	0.572	0.018	0.529	2.616	1.188	4.076
Indep	22,312	0.372	0.053	0.000	0.333	0.571	1.226	5.194
Comp	22,312	4.179	0.849	1.609	4.331	5.481	−0.656	2.846
Gexp	22,312	0.028	0.009	0.011	0.026	0.057	1.059	4.649
Str	22,312	0.422	0.081	0.160	0.434	0.562	−1.296	4.970

reports the descriptive statistics of the variables. Regarding the explained variable, the mean value of CP is 6.715, indicating that on average, each unit of carbon emissions corresponds to an operating revenue of CNY 67,150. The standard deviation is 10.962, showing significant differences in carbon performance among enterprises. The minimum value is 0.006, and the maximum value is 49.162. The huge gap between the extreme values reflects the obvious differentiation of the low-carbon development level within the industry. The median of 1.312 is lower than the mean value, the skewness is 2.054, and the kurtosis is 6.552, indicating that the carbon performance of most enterprises is concentrated in the lower range. Only a few high-performance enterprises have achieved an efficient balance between carbon emissions and economic output through technological innovation or efficiency improvement, presenting an overall characteristic of a long-tail distribution.

Regarding the explanatory variable, the mean value of GP is 0.740, and the median is 0, indicating that more than half of the enterprises have not independently applied for green patents, and the innovation achievements are concentrated in a few leading enterprises. The standard deviation of 2.801 is much larger than the mean value, with a skewness of 5.416 and a kurtosis of 34.298, further confirming the significant “Matthew effect” in green technology innovation: a few enterprises dominate the output of green patents, while most enterprises lag behind relatively in green technology research, development, and transformation. This distribution characteristic is highly consistent with the heterogeneity of enterprises’ innovation resource endowments in reality.

5.2. Correlation Analysis

Figure 1 shows the correlation coefficients between variables. Among them, green indicates a significant positive correlation, and red indicates a significant negative correlation. The darker the color, the closer the correlation coefficient is to 1 (or −1), and blue indicates insignificance.

Regarding the explanatory variable, GP has a significant positive correlation with CP, indicating that the more green patents an enterprise independently applies for, the better its corporate carbon performance, which is consistent with theoretical expectations.

Regarding the control variables, governance factors such as industry competition, management shareholding, and equity balance degree, as well as fundamental characteristics such as company size and establishment duration, are all positively correlated with carbon performance, while the regional industrial structure significantly inhibits carbon performance, reflecting the transformation pressure of high-carbon industries. To avoid the influence of multicollinearity on the estimation results, this paper statistically analyzed the variance inflation factors of the variables. The results show that the maximum value of VIF is 1.63, which is far lower than the empirical level of 5, indicating that these variables can be included in the same model for regression analysis.

5.3. Baseline Regression Analysis

Table 3. The impact of green technology innovation on corporate carbon performance.

Variables	Baseline		Sample Matching
	(1)	(2)	(3)	(4)
	CP	CP	CP	CP
GP	0.236 ***	0.211 ***	0.146 ***	0.158 ***
	(4.322)	(4.156)	(3.104)	(2.799)
Menv		−0.053	−0.285 **	−0.287 *
		(−0.572)	(−2.092)	(−1.910)
Mshare		−2.770 *	−2.778	−3.953
		(−1.898)	(−1.477)	(−1.471)
Size		0.981 ***	1.081 ***	0.866 **
		(4.295)	(3.085)	(1.991)
Lev		2.967 ***	4.789 ***	4.687 ***
		(3.643)	(3.736)	(2.929)
Age		1.971	2.153	3.686
		(1.251)	(0.935)	(1.332)
ROA		2.229	−0.069	−2.744
		(1.599)	(−0.029)	(−0.876)
Intan		−5.556 **	−6.254	−10.226 **
		(−2.120)	(−1.441)	(−2.014)
Equity		0.701 **	0.938 **	0.275
		(2.333)	(2.197)	(0.507)
Indep		−0.244	−1.813	0.001
		(−0.133)	(−0.646)	(0.000)
Comp		4.808 ***	6.697 ***	6.664 ***
		(12.166)	(10.721)	(7.006)
Gexp		14.805	−1.443	7.745
		(1.221)	(−0.093)	(0.359)
Str		−1.034	−1.231	1.295
		(−0.293)	(−0.248)	(0.214)
Cons	6.540 ***	−41.995 ***	−52.177 ***	−52.868 ***
	(161.732)	(−6.430)	(−5.475)	(−4.351)
Individual fixed effect	Yes	Yes	Yes	Yes
Time fixed effect	Yes	Yes	Yes	Yes
Observations	22,312	22,312	22,312	5597
R-squared	0.768	0.782	0.814	0.830

Note: ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively; the t-values in parentheses are based on heteroscedasticity-robust standard errors.

presents the benchmark regression results examining the impact of green technology innovation on corporate carbon performance, directly linking to Hypothesis 1 (green technology innovation improves corporate carbon performance). Column (1) reports estimates from the baseline model without control variables, while Column (2) includes controls and year-fixed effects for robustness.

In Column (1), the coefficient of GP is 0.236 (p < 0.01), indicating that a one-unit increase in GP is associated with a significant improvement in carbon performance. When control variables are added in Column (2), the GP coefficient remains strongly positive at 0.211 (p < 0.01), demonstrating the stability of the relationship. These results confirm that green technology innovation enhances energy efficiency through low-carbon technologies (e.g., energy-saving equipment, clean energy adoption), reducing unit carbon emissions while sustaining economic output, thereby supporting Hypothesis 1.

5.4. Endogeneity Test

5.4.1. Sample Matching

Endogeneity problems often stem from the correlation between the explanatory variables and the error term, and this correlation may be caused by the presence of unobserved confounding factors. When studying the relationship between green technology innovation and corporate carbon performance, factors such as the management style of enterprises and the industry competition situation may affect both corporate carbon performance and be related to technological innovation decisions, thus giving rise to endogeneity. Therefore, entropy balancing matching and propensity score matching are adopted to make the treatment group and the control group as similar as possible in terms of these potential confounding factors.

Entropy Balancing Match (EBM) is a statistical method used to balance the distribution of covariates in observational data. The aim is to reduce sample selection bias, making the treatment group and the control group as similar as possible in terms of covariates to estimate the causal effect more accurately. Firstly, all control variables are set as covariates that need to be balanced; secondly, an optimization algorithm is used to find a set of weights so that the distributions of the treatment group and the control group on these covariates reach equilibrium; finally, these weights are used for regression analysis. The results are shown in column (3) of Table 3. The coefficient of GP is significantly positive at the 1% level, which is consistent with the previous conclusion.

Propensity Score Matching (PSM) is a commonly used statistical method in causal inference. Its basic idea is to find individuals with similar propensity scores for matching in the treatment group and the control group. The processing procedure is as follows: Set the treatment group and the control group according to whether the enterprise applies for green patents, and select the management’s environmental protection background, management shareholding, company size, asset–liability ratio, establishment duration, return on assets, etc., as covariates to estimate the propensity score of each individual. Secondly, according to the calculated propensity scores, use a 1:1 matching ratio to find suitable matching objects in the treatment group and the control group, respectively. Finally, use the matched samples for regression analysis. The results are shown in column (4) of Table 3. Neither the sign nor the significance level of the coefficient of GP has changed substantially, indicating that the conclusion of this work is robust.

5.4.2. Instrumental Variable Method

There may be reverse causality between green technology innovation and corporate carbon performance (that is, enterprises with higher carbon performance have more resources to invest in innovation) or omitted variables (such as unobserved corporate strategic environment), leading to estimation bias in the benchmark model [43]. Therefore, this paper uses the number of universities in the province where the listed enterprise is registered as an instrumental variable for green technology innovation and solves the endogeneity problem through two-stage least squares (2SLS). The rationality lies in the following:

On the one hand, the number of universities in a province is theoretically and empirically correlated with firms’ green technological innovation. First, provinces with more universities possess abundant scientific research talents and technology transfer platforms, enabling firms to access knowledge and technical support for green innovation through talent recruitment and industry–university–research collaboration. Second, university-dense regions foster vibrant innovation ecosystems, often accompanied by supportive policies and capital inflows, which provide resource support for firms’ green R&D investment and patent output. This “talent supply–research collaboration–policy incentive” chain confirms the causal link between university quantity and green innovation, satisfying the relevance condition of instrumental variables.

On the other hand, the number of universities affects corporate carbon performance only through green technological innovation, without direct influence. Universities, as educational institutions, do not directly participate in firms’ production or carbon management; carbon performance improvement relies on firms’ technical upgrades. Even if university-dense provinces have stricter environmental regulations or better industrial structures, the study controls for province-level confounders via time–province interactive fixed effects. This ensures that university quantity impacts carbon performance solely through green innovation, rather than other channels, thus satisfying the exclusion restriction.

Columns (1) and (2) in

Table 4. Endogeneity test.

Variables	IV				Heckman
	(1)	(2)	(3)	(4)	(5)	(6)
	GP	CP	GP	CP	High_CP	CP
GP		5.498 ***		0.366 **		0.210 ***
		(2.879)		(2.360)		(4.123)
IV	−1.195 **		0.184 ***		−0.016
	(−1.977)		(4.535)		(−0.053)
IMR						1.562
						(1.613)
Menv	0.073 ***	−0.427 **	0.080	−0.093	0.008	−0.036
	(2.860)	(−2.567)	(1.500)	(−1.452)	(0.709)	(−0.388)
Mshare	0.501 **	−5.400 ***	−0.432	0.018	0.634 ***	−2.262
	(2.178)	(−3.087)	(−1.146)	(0.016)	(5.395)	(−1.490)
Size	0.173 ***	0.106	0.236 *	−0.401 *	−0.052 **	0.989 ***
	(2.620)	(0.264)	(1.945)	(−1.930)	(−2.408)	(4.355)
Lev	−0.036	3.041 ***	0.323	1.468	−1.188 ***	1.615
	(−0.187)	(3.741)	(0.513)	(1.390)	(−9.298)	(1.392)
Age	−0.319	3.670 **	−0.291	0.004	−0.235 ***	1.965
	(−0.743)	(2.191)	(−0.694)	(0.006)	(−3.108)	(1.246)
ROA	0.424	0.134	0.675	5.480 *	8.571 ***	11.548 *
	(1.223)	(0.084)	(0.488)	(1.849)	(21.335)	(1.904)
Intan	1.179	−11.569 ***	1.358	0.443	1.529 ***	−3.736
	(1.620)	(−3.330)	(0.605)	(0.126)	(3.140)	(−1.312)
Equity	0.012	0.623 **	−0.244	0.784 ***	0.122 ***	0.790 **
	(0.173)	(2.060)	(−1.626)	(2.614)	(3.394)	(2.534)
Indep	0.971	−5.150 **	−1.352	−7.510 ***	0.153	0.154
	(1.414)	(−1.997)	(−0.678)	(−2.699)	(0.462)	(0.084)
Comp	0.205 **	3.820 ***	−0.054	11.670 ***	0.181 *	4.982 ***
	(2.536)	(6.915)	(−0.073)	(7.016)	(1.868)	(12.543)
Gexp	4.713	−19.396	14.186	−6.660	4.386 **	19.719
	(1.134)	(−1.255)	(1.101)	(−0.383)	(2.272)	(1.579)
Str	−0.076	1.583	2.363	−1.957	−0.228	−0.113
	(−0.061)	(0.369)	(1.341)	(−0.954)	(−0.344)	(−0.027)
Cons	1.805	−24.364 **	−4.710	−31.418 ***	1.154	−44.924 ***
	(0.555)	(−2.526)	(−1.153)	(−3.739)	(0.835)	(−6.766)
Individual fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Time fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Regional fixed effect	Yes	Yes			Yes	Yes
Observations	22,312	22,312	2367	2367	22,311	22,311
R-squared	0.675	0.783	0.352	0.783	-	0.783
Kleibergen–Paap rk LM statistic	3.908 **		12.043 ***
Cragg–Donald Wald F statistic	18.733		355.933
Hansen J statistic	0.000		0.000
Endogeneity test	7.634 ***		3.137 *

Note: ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. The t-values in parentheses are based on heteroscedasticity-robust standard errors. Since this study only selects one instrumental variable and there is no problem of over-identification, the Hansen J statistic is 0.000.

report the 2SLS test results of the instrumental variable method. In the first stage, the coefficient of the IV is significant at the 5% level, and the Kleibergen–Paap rk LM statistic rejects the null hypothesis at the 5% significance level, indicating that there is a significant correlation between the instrumental variable and the endogenous explanatory variable, and the model is “identifiable”. The Cragg–Donald Wald F statistic of 18.733 is greater than 16.38, indicating a strong correlation between the instrumental variable and the endogenous variable, rejecting the hypothesis of “weak instrumental variable”, and the IV estimation results are reliable. The Wald Chi-sq of the Endogeneity test is 7.634, which rejects the null hypothesis at the 1% significance level, confirming that green technology innovation (GP) is an endogenous variable and there is reverse causality, so it is necessary to use the IV method to correct the endogeneity. In the second stage, the coefficient of the IV on CP is significantly positive at the 1% level, indicating that after using the instrumental variable to alleviate the endogeneity problem, green technology innovation still significantly improves corporate carbon performance, and the hypothesis of our study still holds.

Additionally, we select low-carbon R&D investment as an instrumental variable (IV). When examining the impact of green technology innovation on corporate carbon performance, using low-carbon R&D investment as an IV for green technology innovation is reasonable for the following reasons. In terms of relevance, low-carbon R&D investment is closely linked to green technology innovation, as increased corporate investment in low-carbon R&D directly promotes the research, development, and application of green technologies, thereby driving green technology innovation. Regarding exogeneity, low-carbon R&D investment can be considered partially exogenous, as it is influenced by external factors such as policy incentives and market demand, rather than being directly determined by a firm’s carbon performance. Therefore, low-carbon R&D investment serves as an effective IV for green technology innovation, addressing endogeneity issues and enabling more accurate estimation of the impact of green technology innovation on corporate carbon performance. The test results are reported in columns (3) and (4) of Table 4. Neither the relevant statistics nor the coefficient of GP exhibit substantial changes, confirming the robustness of our conclusions.

5.4.3. Heckman Two-Stage Method

The relationship between green technology innovation and corporate carbon performance may be affected by sample selection bias, as firms with certain characteristics (e.g., resource endowments or regional environmental regulations) may systematically differ in their likelihood of being included in the analysis or achieving specific carbon performance levels. To address this, we employ the Heckman two-stage selection model, which explicitly models the sample selection mechanism and corrects for potential bias in ordinary least squares (OLS) estimates.

We define a binary selection variable High_CP, which equals 1 if a firm’s carbon performance exceeds the annual industry median and 0 otherwise. The selection equation is estimated using a Probit model, with the following specification:

$${\mathrm{P}\mathrm{r}(High\_CP}_{i,t}=1)=\mathsf{\Phi }({\beta }_0+{\beta }_{1}IV+\gamma X_{i,t}+u_i+v_t+{\epsilon }_{i,t})$$

(4)

Here, $IV$ (number of universities in the firm’s registered province) serves as an exclusive exclusion restriction variable, included in the selection equation but excluded from the outcome equation (second stage). This choice is justified by the theoretical rationale that university density influences a firm’s access to innovation resources (e.g., talent, research collaborations) that affect its likelihood of achieving high carbon performance via green technology innovation, yet it does not directly impact carbon performance through channels unrelated to innovation (e.g., operational emission reduction measures). This satisfies the key exclusion restriction requirement for the Heckman model: the variable affects the selection process (probability of being in the high-carbon-performance group) but not the outcome (carbon performance) directly, except through its influence on the endogenous regressor (green technology innovation).

Additional controls in the selection equation (denoted $X_{i,t}$) include firm-level characteristics (size, leverage, profitability), individual fixed effects ($u_i$) and year fixed effects ($v_t$) to capture time-varying individual trends. The Probit estimation generates the Inverse Mills Ratio (IMR), which measures the correlation between the error terms of the selection and outcome equations.

The IMR is incorporated into the benchmark regression model to adjust for selection bias:

$${CP}_{i,t}={\alpha }_0+{\alpha }_{1}GP+{\alpha }_2{IMR}_{i,t}+\gamma \sum {Controls}_{i,t}+u_i+v_t+{\epsilon }_{i,t}$$

(5)

As reported in column (6) of Table 4, the estimated coefficient of the IMR is statistically insignificant (p > 0.10), indicating that the error terms of the selection and outcome equations are not systematically correlated, formally confirming the absence of sample selection bias in our dataset. Importantly, the coefficient of green technology innovation (GP) remains significantly positive at the 1% level, underscoring the robustness of our core finding that GP improves corporate carbon performance, even after accounting for potential selection effects.

The exclusion of universities from the outcome equation is critical for model validity. Universities primarily influence carbon performance by fostering green technology innovation (e.g., through knowledge spillovers and talent supply), which is already captured by the endogenous variable GP in the outcome equation. By design, university density does not directly affect emissions reductions through non-innovative channels (e.g., mandatory emission caps or market-based trading mechanisms), ensuring it meets the exclusion restriction. This theoretical alignment, combined with the insignificant IMR result, supports the appropriateness of the Heckman correction and mitigates concerns about model misspecification.

5.4.4. Placebo Test

The Placebo test is an important method used to verify the reliability of empirical results and the robustness of causal relationships. Its core purpose is to rule out the interference of other potential factors on the research results and ensure that the discovered effects are not caused by accidental factors, model specification problems, or other spurious relationships.

Firstly, the values of green technology innovation are randomly assigned to the samples to generate virtual explanatory variables. Secondly, OLS estimation is carried out, and the coefficients and p-values of the virtual GP are retained. Finally, the above steps are repeated 500 times. If the coefficients and p-values of the virtual GP approximately conform to the normal distribution hypothesis, and passing the significance test is a low-probability event, it indirectly indicates the reliability of the research conclusions of this paper.

Figure 2 shows the distribution of the virtual coefficient values and p-values. It can be seen that the virtual coefficient values of GP all approximately follow a normal distribution, and the vast majority are far from the real coefficient value of 0.211. It is a low-probability event that the p-value is less than 0.1, indicating that the research conclusions of our work have good robustness and are not caused by random factors.

5.5. Robustness Test

5.5.1. GMM Dynamic Model

To avoid the influence of time trends, we introduced a dynamic Generalized Method of Moments (GMM) model for robustness testing. The model is specified as follows:

$${CP}_{i,t}={\alpha }_0+{\alpha }_1{CP}_{i,t-1}+{\alpha }_2{GP}_{i,t}+\gamma \sum {Controls}_{i,t}+u_i+v_t+{\epsilon }_{i,t}$$

(6)

where ${CP}_{i,t-1}$ is the lagged term of the dependent variable. The meanings of the remaining symbols are consistent with those in Equation (3).

Column (1) of

Table 5. Robustness test (1).

Variables	GMM	Replace the Core Variable		Panel Tobit	Add Omitted Variables
	(1)	(2)	(3)	(4)	(5)
	CP	CP	CP	CP	CP
L.CP	0.963 ***
	(117.473)
GP	0.021 ***	0.043 ***	0.006 *	0.144 ***	0.206 ***
	(3.001)	(2.920)	(1.802)	(7.470)	(4.008)
Menv	0.001	−0.448 **	0.001	−0.110 ***	−0.048
	(0.659)	(−2.435)	(0.277)	(−3.543)	(−0.503)
Mshare	0.001	−3.257	−0.085	0.354	−2.313
	(0.034)	(−1.121)	(−1.397)	(1.020)	(−1.569)
Size	−0.012 ***	1.918 ***	0.229 ***	0.009	0.945 ***
	(−4.228)	(3.970)	(22.074)	(0.145)	(4.042)
Lev	0.043 **	2.783	−0.090	1.525 ***	3.014 ***
	(2.386)	(1.413)	(−1.337)	(4.458)	(3.570)
Age	−0.020 *	1.444	0.010	−0.159	2.380
	(−1.914)	(0.369)	(0.291)	(−0.658)	(1.423)
ROA	0.439 ***	−4.255	0.503 ***	4.550 ***	1.953
	(8.709)	(−1.362)	(2.669)	(5.659)	(1.380)
Intan	−0.040	−15.985 **	−0.251	−2.517 **	−6.233 **
	(−0.606)	(−2.192)	(−0.977)	(−1.967)	(−2.261)
Equity	0.009 **	0.786	0.010	0.449 ***	0.770 **
	(2.061)	(1.398)	(0.589)	(4.543)	(2.488)
Indep	0.150 ***	−1.328	0.016	0.334	−0.306
	(2.987)	(−0.366)	(0.087)	(0.364)	(−0.162)
Comp	0.057 ***	7.101 ***	0.179 **	9.492 ***	4.795 ***
	(7.977)	(6.484)	(2.256)	(32.859)	(11.890)
Gexp	−0.820 **	−8.098	−3.286 ***	18.478 ***	7.819
	(−2.511)	(−0.353)	(−2.594)	(3.045)	(0.618)
Str	−0.058	−1.536	0.385 ***	−1.480	−0.200
	(−1.575)	(−0.192)	(2.759)	(−1.527)	(−0.052)
Ginsti		−67.052 ***			0.124 ***
		(−4.339)			(2.667)
ER		6545			0.298 ***
		0.822			(4.815)
Cons		−67.052 ***	-	−29.346 ***	−42.798 ***
		(−4.339)	-	(−15.290)	(−6.246)
Individual fixed effect		Yes	Yes	Yes	Yes
Time fixed effect		Yes	Yes	Yes	Yes
AR(1)	−15.76 ***
AR(2)	−0.98
Hansen test	0.276
/sigma_u				3.269 ***
				(42.950)
/sigma_e				5.233 ***
				(195.389)
Observations	17,111	6545	21,560	22,312	21,544
R-squared	-	0.822	-	-	0.786

Note: ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively; the t-values in parentheses are based on heteroscedasticity-robust standard errors.

reports the estimation results of the GMM dynamic model. The coefficient of GP remains significantly positive (indicating improved carbon performance), and the lagged term of carbon performance is statistically significant, confirming the dynamic relationship. The Arellano–Bond test for autocorrelation (p > 0.1 for second-order autocorrelation) and the Hansen test for instrument validity (p > 0.1) validate the model’s reliability.

5.5.2. Replace the Core Variable

To avoid the impact of measurement errors on the regression results, we conducted a robustness test by replacing variables. Referring to existing literature, we first measured green technology innovation based on the number of citations of the green patents that enterprises have been granted in the corresponding year [44]. Second, this work uses corporate waste gas emission reduction governance as a proxy variable for carbon performance. If there is no description, it is assigned a value of 0; if there is a qualitative description, it is assigned a value of 1; and if there is a quantitative description, it is assigned a value of 2. An ordinal regression model was employed for coefficient estimation. The regression results are shown in Columns (2) and (3) of Table 5. The coefficients of GP are significantly positive at least at the 10% level. Green technology innovation still has a significant positive effect on corporate carbon performance, once again supporting the hypothesis of this paper.

5.5.3. Using the Panel Tobit Model

The values of the explained variable, corporate carbon performance, are all greater than 0, showing a truncated characteristic. OLS assumes that the dependent variable can take values within the entire real number range. If it is directly used to analyze this type of data, it will lead to bias in the estimation results. The panel Tobit model is specifically designed to handle this kind of truncated data. It can take into account the truncated nature of the data and estimate parameters through an appropriate likelihood function, thus more accurately describing the relationship between green technology innovation and corporate carbon performance. Therefore, this paper uses the panel Tobit model to re-estimate the coefficient of GP. The results are shown in Column (4) of Table 5. The coefficient of GP is 0.144, which is significant at the 1% level and consistent with the benchmark conclusion.

5.5.4. Consider Omitted Variables

To address potential omitted variable bias, this study adds two control variables: green investors and the government environment. Green investors directly influence the green technology innovation decisions of enterprises through their capital allocation preferences, and differences in their participation levels may systematically affect carbon performance. As an external institutional pressure, government environmental regulation may not only force enterprises to carry out green technology innovation to reduce compliance costs but also directly promote the improvement of carbon emission efficiency. These two factors represent the impacts of market mechanisms and policy tools on corporate behavior, respectively. If they are not controlled, it may lead to bias in the estimation of the effect of green technology innovation on carbon performance. In this work, the number of green investors (Ginsti) and the number of environmental penalty cases in the province where the listed enterprise is registered divided by 1000 (ER) are used to measure green investors and government environmental regulation, respectively. The regression results are shown in column (5) of Table 5. The coefficients of Ginsti and ER are both significantly positive, indicating that green investors and government environmental regulation are conducive to improving corporate carbon performance. At this time, the estimated coefficient of GP is still significantly positive at the 1% level, and the hypothesis of this study still holds.

5.5.5. Consider the Impact of Policies

To exclude the impact of policy shocks on the relationship between green technology innovation and corporate carbon performance, this paper conducts grouped regression according to whether it is before or after the implementation of the new Environmental Protection Law and whether the enterprises are located in low-carbon city pilot areas to verify the robustness of the conclusions. Through mechanisms such as increasing the cost of pollution and strengthening information disclosure, the new Environmental Protection Law may force enterprises to increase green technology innovation to improve carbon performance; low-carbon city pilots directly affect enterprises’ innovation decisions and emission reduction behaviors through policy preferences. If under different policy groupings, the positive impact of green technology innovation on carbon performance is significant and the directions of the coefficients are consistent, this indicates that the core conclusion is not affected by changes in the policy environment and can effectively reflect the internal relationship between the two.

Table 6. Robustness test (2).

Variables	The Implementation of the New Environmental Protection Law		The Policy of Low-Carbon City Pilot
	(1)	(2)	(3)	(4)
	CP	CP	CP	CP
GP	0.099 ***	0.107 *	0.187 **	0.201 ***
	(3.250)	(1.796)	(2.124)	(3.257)
Menv	0.054	−0.116	0.123	−0.214
	(1.110)	(−1.114)	(0.996)	(−1.565)
Mshare	0.424	−2.334	−1.545	−4.666 **
	(0.338)	(−1.533)	(−0.814)	(−2.205)
Size	0.419 **	0.509 *	1.139 ***	0.797 **
	(2.027)	(1.666)	(3.851)	(2.361)
Lev	−0.088	1.975 *	1.210	4.568 ***
	(−0.195)	(1.800)	(1.184)	(3.697)
Age	1.521 *	4.955	0.774	3.068
	(1.688)	(1.536)	(0.353)	(1.354)
ROA	2.505 ***	3.485 **	0.621	4.897 **
	(3.819)	(2.095)	(0.337)	(2.407)
Intan	0.029	−2.839	−3.674	−6.902 *
	(0.022)	(−0.667)	(−1.116)	(−1.715)
Equity	0.013	0.883 ***	1.248 ***	0.224
	(0.071)	(2.593)	(3.142)	(0.513)
Indep	2.122 **	−0.788	−0.820	1.290
	(2.525)	(−0.388)	(−0.395)	(0.436)
Comp	1.356 ***	4.682 ***	4.769 ***	4.965 ***
	(6.543)	(5.916)	(8.235)	(9.103)
Gexp	22.093 ***	−32.500 *	2.073	9.046
	(2.850)	(−1.914)	(0.110)	(0.570)
Str	−1.873	−12.339 **	−3.479	2.650
	(−1.051)	(−2.026)	(−0.711)	(0.445)
Cons	−17.161 ***	−31.926 ***	−41.149 ***	−42.782 ***
	(−3.403)	(−2.632)	(−5.022)	(−4.353)
Individual fixed effect	Yes	Yes	Yes	Yes
Time fixed effect	Yes	Yes	Yes	Yes
Observations	8602	13,507	10,629	11,671
R-squared	0.795	0.860	0.781	0.788

Note: ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively; the t-values in parentheses are based on heteroscedasticity-robust standard errors.

reports the impact of green technology innovation on corporate carbon performance under policy shocks. The estimated coefficients of GP are significantly positive at least at the 10% level, and the differences in coefficients between groups are not large, indicating that the impact of green technology innovation on corporate carbon performance has a certain degree of stability.

5.6. Further Analysis

5.6.1. Mechanism Analysis

(1)

Impact Mechanism of Total Factor Productivity

Green technology innovation enhances total factor productivity (TFP) through multi-dimensional pathways, thereby improving corporate carbon performance.

First, green technology innovation directly boosts energy efficiency and reduces pollution emissions by optimizing resource allocation and production processes. For example, the application of clean production technologies and renewable energy can reduce resource consumption and carbon emissions per unit of output, creating a direct effect of “cost reduction and efficiency improvement”.

Second, green technology innovation has significant technological spillover effects: by developing or introducing green technologies, enterprises not only enhance their own production efficiency but also drive collaborative innovation upstream and downstream of industry chains. For instance, digital energy management systems can optimize the overall energy consumption of supply chains, fostering technological diffusion and efficiency improvements at the industry level.

Third, green technology innovation promotes changes in corporate management models, achieving fine-grained control over the entire production process through intelligent monitoring, circular economy models, etc. For example, real-time carbon emission monitoring systems can accurately identify emission reduction potentials while enhancing the scientific basis of production decisions, indirectly boosting TFP.

Additionally, green technology innovation restructures the combination of productivity factors by integrating environmental factors into the TFP accounting framework, creating synergies among technological progress, management optimization, and ecological benefits. For example, technological iterations in the new energy industry not only reduce carbon emissions but also enhance overall economic efficiency through economies of scale. Empirical studies show that green technology innovation by enterprises in heavily polluting industries (especially practical green patents) have a significant positive impact on green total factor productivity (GTFP) [45], and improvements in GTFP are directly linked to better corporate carbon performance.

In essence, this mechanism transforms the relationship between ecological protection and economic growth from “opposition” to “symbiosis” through technology-driven productivity changes, ultimately achieving the dual goals of corporate low-carbon transformation and competitiveness enhancement.

Referring to research by Lu and Lian [46], this work uses the total factor productivity (TFP) as the explained variable. In the estimation of TFP, both the OP method and the LP method are employed to address the endogeneity issue in production function estimation, each with its own advantages and disadvantages. Given that the LP method offers superior data availability, broader sample coverage, and greater flexibility in assumptions, making it particularly suitable for TFP estimation scenarios where investment data quality is limited or full sample information needs to be preserved, this work utilizes the LP method to calculate TFP. The results are shown in Columns (1)–(2) of

Table 7. Mechanism analysis.

Variables	(1)	(2)	(3)	(4)	(5)	(6)
	TFP	TFPt+1	Egov	Egovt+1	TFP	TFPt+1
GP	0.003 *	0.008 ***	0.017 **	0.015 *	0.003 *	0.008 ***
	(1.810)	(3.788)	(2.348)	(1.848)	(1.733)	(3.771)
Egov					0.009 ***	0.012 ***
					(4.017)	(4.471)
Menv	−0.008	−0.013 ***	−0.005	−0.004	−0.008	−0.013 ***
	(−1.574)	(−2.645)	(−0.334)	(−0.294)	(−1.568)	(−2.644)
Mshare	−0.072	−0.041	0.849 ***	0.654 ***	−0.080	−0.050
	(−1.365)	(−0.643)	(5.410)	(3.950)	(−1.509)	(−0.772)
Size	0.662 ***	0.590 ***	0.308 ***	0.214 ***	0.659 ***	0.587 ***
	(46.697)	(36.189)	(8.460)	(5.410)	(46.349)	(36.039)
Lev	0.406 ***	0.028	−0.358 ***	−0.356 **	0.410 ***	0.034
	(7.986)	(0.526)	(−2.667)	(−2.526)	(8.076)	(0.638)
Age	0.212 ***	0.326 ***	0.279	0.378 *	0.208 ***	0.322 ***
	(3.121)	(3.575)	(1.427)	(1.672)	(3.075)	(3.535)
ROA	2.157 ***	0.124	0.109	−0.612 **	2.156 ***	0.121
	(24.517)	(1.436)	(0.441)	(−2.317)	(24.528)	(1.404)
Intan	−0.506 ***	−0.538 **	1.130 **	0.695	−0.517 ***	−0.549 ***
	(−2.588)	(−2.539)	(2.216)	(1.254)	(−2.645)	(−2.590)
Equity	−0.021	−0.036 **	−0.099 **	−0.077 *	−0.020	−0.035 **
	(−1.496)	(−2.281)	(−2.401)	(−1.795)	(−1.435)	(−2.202)
Indep	−0.026	−0.059	0.162	−0.253	−0.027	−0.060
	(−0.294)	(−0.607)	(0.473)	(−0.714)	(−0.308)	(−0.616)
Comp	−0.010	−0.012	−0.094 *	−0.103	−0.009	−0.011
	(−0.433)	(−0.413)	(−1.654)	(−1.632)	(−0.396)	(−0.383)
Gexp	−0.467	−0.778	1.414	−0.014	−0.483	−0.780
	(−0.717)	(−1.063)	(0.664)	(−0.006)	(−0.743)	(−1.070)
Str	0.231	0.293	−0.660	−1.422 **	0.237	0.304
	(1.172)	(1.339)	(−1.061)	(−2.032)	(1.204)	(1.392)
Cons	−4.870 ***	−3.434 ***	−5.345 ***	−3.046 ***	−4.819 ***	−3.371 ***
	(−14.252)	(−7.993)	(−5.322)	(−2.698)	(−14.071)	(−7.875)
Individual fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Time fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Observations	20,783	17,891	22,312	19,214	20,783	17,891
R-squared	0.953	0.941	0.659	0.658	0.953	0.941

Note: ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively; the t-values in parentheses are based on heteroscedasticity-robust standard errors.

.

The coefficient of GP on TFP is 0.003, significant at the 10% level, and the coefficient on subsequent TFP is 0.008, significant at the 1% level. This indicates that the “productivity dividend” of green technology innovation serves as a dual-driven mechanism for carbon performance improvement, with distinct temporal dynamics. The statistically significant coefficient of 0.003 at the 10% level for current TFP suggests that green technological advancements already exert marginal yet discernible effects on enhancing production efficiency through immediate channels such as resource allocation optimization and energy-saving technology adoption.

More notably, the substantially larger coefficient of 0.008, significant at the 1% level for subsequent TFP, highlights a cumulative amplification effect over time, implying that GP fosters deeper structural transformations in production processes—such as the gradual integration of low-carbon innovation ecosystems, long-term R&D spillover effects, and organizational learning toward sustainable practices. This temporal gradient underscores that green technology innovation not only delivers short-term productivity gains but also cultivates enduring “capacity dividends” for carbon performance, as improved TFP over time likely translates into higher energy efficiency, reduced carbon intensity per unit output, and more sustainable supply chain management. Together, these findings reveal that GP acts as a persistent engine for carbon reduction, with its productivity-enhancing effects unfolding progressively to support long-term environmental sustainability goals.

(2)

Impact Mechanism of Environmental Governance Capabilities

Green technology innovation forms a core mechanism for improving corporate carbon performance by empowering multi-dimensional enhancements in environmental governance capabilities.

First, green technology innovation provides critical tools and technical support for environmental governance. For example, carbon capture, utilization, and storage (CCUS) technologies directly reduce carbon emissions in production processes, while intelligent monitoring systems enable precise traceability and real-time control of pollutant emissions, enhancing the effectiveness and efficiency of corporate pollution treatment at the technical level.

Second, green technology innovation drives the upgrading of environmental governance models. By integrating environmental data (such as energy consumption and emission indicators) through digital management platforms, it establishes a closed-loop management system of “monitoring–analysis–optimization”, improving the scientific basis of corporate environmental decisions (e.g., dynamically adjusting emission reduction targets and optimizing resource allocation), thereby systematically reducing carbon intensity.

Furthermore, green technology innovation promotes the integration and coordination of environmental governance resources. For instance, circular economy technologies enable the reuse and recycling of waste, reducing end-of-pipe governance pressures. Meanwhile, through technological diffusion in industry chains (such as green supply chain management technologies), they drive upstream and downstream enterprises to jointly reduce emissions, forming regional or industrial environmental governance network effects.

Additionally, green technology innovation indirectly improves carbon performance by enhancing environmental compliance capabilities. Enterprises leverage technological innovation to meet stringent environmental standards (such as carbon emission standards and ESG disclosure requirements), avoiding compliance risks while accumulating environmental credibility, attracting green investments, and driving internal governance capacity upgrades.

In essence, this mechanism represents a “two-way empowerment” between technological innovation and governance capabilities: green technology innovation provides hard-power support for environmental governance, while enhanced governance capabilities reciprocate by deepening and broadening technological applications. Ultimately, through the collaborative optimization of pollution control capabilities, resource utilization efficiency, and compliance management standards, it achieves sustained improvements in corporate carbon performance.

Corporate environmental governance capability refers to the comprehensive ability of enterprises to control the negative impacts of production and operations on the ecological environment through the formulation and implementation of environmental management strategies and measures to achieve synergistic development between economy and environmental protection. Drawing on existing research [47,48], we evaluate the environmental governance capabilities (Egov) of listed enterprises from eight aspects, including environmental protection concepts, environmental protection goals, environmental protection management system, environmental protection education and training, special environmental protection actions, emergency response mechanisms for environmental incidents, environmental protection honors and awards, and the “three simultaneous” system. The test results are shown in Columns (3)–(4) of Table 7.

The coefficient of GP on Egov is 0.017 (significant at the 5% level), while that on next-period Egov is 0.015 (significant at the 10% level), revealing a dual impact of green technology innovation on corporate environmental governance capabilities for “active emission reduction”. The immediate significant coefficient indicates that GP drives firms to rapidly deploy green technologies (e.g., energy-efficient systems) and adopt governance practices (e.g., emissions monitoring) to reduce carbon footprints, while the lagged significant coefficient reflects cumulative learning effects and long-term integration of green innovation into operational strategies, such as establishing sustainability departments or embedding low-carbon goals in executive compensation. Together, these findings show that GP acts as both a short-term catalyst and long-term enabler, fostering immediate governance improvements and enduring mechanisms for sustained carbon performance enhancement.

Additionally, we examined the interaction between governance effects and productivity effects, with the results reported in Columns (5)–(6) of Table 7. The coefficient of Egov on current TFP is 0.009, and the coefficient on subsequent TFP is 0.012, both significant at the 1% level. This reveals a dynamic complementary relationship between environmental governance and productivity: enhanced Egov not only improves current production efficiency but also amplifies long-term productivity gains. The mechanism involves two pathways: first, proactive environmental governance (e.g., adopting low-carbon management systems, establishing emissions reduction accountability mechanisms) reduces resource waste and operational inefficiencies, directly boosting TFP through optimized input–output ratios; second, sustained investment in environmental capabilities fosters organizational learning and technological absorption capacity, enabling firms to better integrate green innovations (e.g., smart energy systems, circular economy technologies) into production processes over time, thereby generating cumulative productivity improvements (as evidenced by the larger lagged coefficient).

These findings highlight that environmental governance and productivity growth are not mutually exclusive but rather reinforce each other in driving carbon performance. For instance, improved TFP from governance-driven efficiency gains reduces carbon intensity per unit of output, while enhanced productivity provides financial and technological resources to further upgrade environmental practices (e.g., R&D in cleaner production). The significant coefficients at the 1% level underscore the robustness of this synergy, aligning with the “resource-based view” of sustainability, where environmental capabilities evolve into core competencies that deliver both ecological and economic benefits. Collectively, these results suggest that integrating governance-focused “active emission reduction” strategies with productivity-enhancing green innovations creates a self-reinforcing cycle, offering empirical support for the dual dividend of sustainability—simultaneously improving environmental and operational performance.

5.6.2. Moderating Effect Analysis

(1)

Moderating Effect of Green Media Attention

As the main disseminator of environmental information, green media constructs a “public opinion field” for corporate environmental responsibility by exposing high-carbon emission behaviors, interpreting environmental protection policies, and promoting cases of low-carbon technologies. When the attention of green media is high, the “signal value” of enterprises’ green technology innovation is amplified. On the one hand, the media continuously tracks the actual implementation of innovation achievements (such as the progress of patents being transformed into emission reduction equipment), forming implicit supervision of the management and prompting innovation resources to tilt towards areas that improve carbon performance [49]. On the other hand, the public forms expectations of enterprises’ environmental performance through media reports, forcing enterprises to transform green patents into specific carbon emission management measures to avoid “innovation hoarding” or “greenwashing” behaviors [50].

To examine the moderating effect of green media attention, this paper uses the number of environmental protection news reports at the enterprise level (Gmedia) to measure green media attention and generates the interaction term Gmedia*GP. The regression result is shown in Column (1) of

Table 8. Moderating effect analysis.

Variables	(1)	(2)
	CP	CP
GP	0.172 ***	0.155 ***
	(3.123)	(2.907)
Gmedia	−0.147 ***
	(−3.099)
GP*Gmedia	0.023 **
	(2.201)
Ginvest		0.021
		(1.358)
GP*Ginvest		0.007 ***
		(3.510)
Menv	−0.049	−0.067
	(−0.530)	(−0.735)
Mshare	−2.819 *	−2.790 *
	(−1.934)	(−1.913)
Size	1.025 ***	0.960 ***
	(4.519)	(4.216)
Lev	2.942 ***	2.990 ***
	(3.619)	(3.680)
Age	1.960	1.962
	(1.247)	(1.247)
ROA	2.310 *	2.290
	(1.657)	(1.644)
Intan	−5.579 **	−5.719 **
	(−2.139)	(−2.186)
Equity	0.687 **	0.696 **
	(2.288)	(2.322)
Indep	−0.249	−0.422
	(−0.136)	(−0.231)
Comp	4.786 ***	4.791 ***
	(12.134)	(12.171)
Gexp	14.629	13.856
	(1.205)	(1.141)
Str	−0.831	−1.149
	(−0.234)	(−0.327)
Cons	−42.829 ***	−41.338 ***
	(−6.558)	(−6.322)
Individual fixed effect	Yes	Yes
Time fixed effect	Yes	Yes
Observations	22,312	22,312
R-squared	0.783	0.783

Note: ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively; the t-values in parentheses are based on heteroscedasticity-robust standard errors.

. The coefficient of Gmedia*GP is significantly positive at the 5% level, indicating that green media attention positively moderates the relationship between green technology innovation and corporate carbon performance. This moderating effect is also in line with the “legitimacy theory” [51]; that is, in order to obtain social legitimacy, enterprises are more likely to transform technological innovation into substantial emission reduction actions under media supervision, thereby strengthening the positive impact of green technology innovation on carbon performance.

(2)

Moderating Effect of Investors’ Green Attention

Investor attention plays a critical moderating role between green technology innovation and corporate carbon performance, primarily through two complementary pathways: on the one hand, institutional investors directly pressure companies to disclose the impact of green technology innovation on carbon performance through governance participation mechanisms like shareholder proposals and board interventions, particularly enhancing the transformation of innovation outcomes into carbon reductions in high-carbon industries such as energy [52]. Second, the carbon risk pricing mechanism drives market capital to favor low-emission innovative enterprises, as high-emission firms are compelled to increase green technology investments to avoid valuation discounts due to bearing carbon risk premiums [43]. These two mechanisms form a synergistic effect of “governance pressure + market incentives,” which elevates the priority of innovation through institutional intervention in the short term and forces improvements in carbon performance through capital flows in the long term, jointly driving enterprises to translate green technologies into tangible emission reduction outcomes. With the popularization of the ESG investment concept in China, investors’ attention to enterprises’ environmental performance has significantly increased, forming a “demand-side incentive” for green technology innovation [53]. When investors’ green attention is high, they are more inclined to provide long-term capital support for green technology innovation, alleviating the financing constraints of enterprises’ innovation investments. At the same time, shareholder’s proposals or participation in corporate governance require management to disclose the specific impact of innovation achievements on carbon performance.

In this context, green technology innovation is no longer an isolated R&D behavior. Instead, it is deeply bound to investors’ value demands, encouraging enterprises to integrate innovation activities with the carbon emission management system, thereby amplifying the effect of innovation on improving carbon performance. This study uses the number of environmental protection issues of listed enterprises raised by investors through channels such as websites, WeChat, and APPs (Ginvest) as a proxy variable for investors’ green attention and generates the interaction term Ginvest*GP. The regression results are shown in Column (2) of Table 8. The coefficient of Ginvest*GP is significantly positive at the 1% level, indicating that investors’ green attention strengthens the positive impact of green technology innovation on corporate carbon performance. This moderating effect is also in line with the “signaling theory” [54]; that is, the higher the investors’ green attention, the stronger the motivation of enterprises to send signals of low-carbon transformation through green technology innovation. Then, a positive cycle is formed at the levels of resource acquisition and strategic implementation.

5.6.3. Long-Term Effect Analysis

From the patent application of green technology innovation to the manifestation of actual emission reduction effects, a complete cycle of “technological research and development–achievement transformation–production application–efficiency optimization” is required. In the short term, the acquisition of green patents only represents technological reserves and has not been transformed into actual productive forces. As time goes by, enterprises embed innovation achievements into the production system through equipment renovation, process reorganization, and management support (such as establishing a carbon monitoring system), initially releasing substantial emission reduction potential with significant marginal effects. However, as technology diffuses across industries or the physical limits of emission reduction are approached, the incremental benefits of each additional unit of technological input gradually diminish.

To reveal the persistent impact of green technology innovation on corporate carbon performance, we estimate the regression coefficients of GP on corporate carbon performance in the next 1–4 periods, as shown in

Table 9. Long-term effect analysis.

Variables	(1)	(2)	(3)	(4)
	CPt+1	CPt+2	CPt+3	CPt+4
GP	0.150 ***	0.149 **	0.134 **	0.094 **
	(2.611)	(2.457)	(2.393)	(2.005)
Menv	−0.091	−0.134	−0.194 **	−0.218 ***
	(−0.969)	(−1.439)	(−2.203)	(−2.719)
Mshare	−1.956	−1.269	−1.104	−1.387
	(−1.362)	(−0.909)	(−0.834)	(−1.078)
Size	0.921 ***	0.695 ***	0.574 **	0.575 **
	(3.858)	(2.838)	(2.410)	(2.463)
Lev	2.419 ***	2.690 ***	2.334 ***	1.345 *
	(3.050)	(3.371)	(2.848)	(1.692)
Age	2.381	2.404	2.586	3.170
	(1.345)	(1.248)	(1.236)	(1.334)
ROA	−2.314	−2.951 **	−3.678 ***	−3.521 ***
	(−1.638)	(−2.065)	(−2.697)	(−2.825)
Intan	−3.798	−2.530	−2.290	−1.675
	(−1.451)	(−0.951)	(−0.887)	(−0.631)
Equity	0.735 **	0.568 *	0.463	0.396
	(2.347)	(1.838)	(1.627)	(1.534)
Indep	−0.931	0.217	1.239	1.040
	(−0.512)	(0.121)	(0.696)	(0.616)
Comp	3.686 ***	2.694 ***	2.118 ***	1.791 ***
	(10.735)	(8.705)	(6.934)	(5.501)
Gexp	22.128 *	16.554	6.118	1.557
	(1.798)	(1.361)	(0.523)	(0.154)
Str	−0.826	−0.012	2.295	1.366
	(−0.233)	(−0.003)	(0.552)	(0.331)
Cons	−37.799 ***	−30.296 ***	−27.327 ***	−27.431 ***
	(−5.401)	(−4.120)	(−3.524)	(−3.186)
Individual fixed effect	Yes	Yes	Yes	Yes
Time fixed effect	Yes	Yes	Yes	Yes
Observations	19,214	16,521	14,191	12,025
R-squared	0.772	0.757	0.753	0.767

Note: ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively; the t-values in parentheses are based on heteroscedasticity-robust standard errors.

. The impacts of GP on CP in the next 1–4 periods are all positively significant and show a trend of diminishing marginal effects, which is consistent with the “time lag effect” (where initial adoption drives rapid improvements) and “diffusion saturation effect” (where subsequent gains attenuate due to technology spillover or narrowed reduction space) of technological innovation. For example, consider a leading Chinese new-energy vehicle manufacturer that, upon securing a breakthrough patent for battery recycling technology in 2018, initially achieved a 15% reduction in carbon intensity within two years as the technology was integrated into production lines. However, by the fourth year, as the technology diffused across the industry and the scope for incremental process optimization narrowed, the marginal reduction in carbon intensity declined to 5%. This pattern reflects that while green technology innovation generates sustained positive impacts, its marginal contributions to carbon performance diminish over time as the low-hanging fruits of emission reduction are exhausted and competitive diffusion erodes proprietary advantages. This analysis verifies that green technology innovation is not a short-term “greenwashing” behavior; instead, through continuous technological penetration and organizational adaptation, it forms a long-term improvement mechanism for carbon performance, albeit one characterized by diminishing marginal returns as the innovation lifecycle matures.

5.6.4. Heterogeneity Analysis

(1)

Heterogeneity Analysis Based on the Nature of Property Rights

There are obvious differences between state-owned enterprises and non-state-owned enterprises in terms of governance mechanisms, innovation incentives, and environmental responsibility drivers. To examine the heterogeneous impact of the nature of property rights, this work conducts grouped regression according to whether the samples are state-owned enterprises. The results are shown in Columns (1) and (2) of

Table 10. Heterogeneity analysis (1).

Variables	State = 0	State = 1	High_Carbon = 0	High_Carbon = 1
	(1)	(2)	(3)	(4)
	CP	CP	CP	CP
GP	0.253 ***	0.181 ***	0.218 ***	−0.001
	(3.065)	(2.964)	(3.787)	(−1.297)
Menv	0.158	−0.282 **	0.116	−0.002
	(1.154)	(−2.455)	(0.948)	(−0.896)
Mshare	−1.020	−26.167 **	−2.808 *	−0.035
	(−0.664)	(−2.194)	(−1.751)	(−1.499)
Size	1.022 ***	0.428	1.204 ***	0.022
	(3.040)	(1.480)	(4.299)	(1.579)
Lev	1.049	4.354 ***	1.429	−0.024
	(0.928)	(3.809)	(1.444)	(−1.119)
Age	1.671	1.038	2.323	0.037
	(0.817)	(0.382)	(1.310)	(1.021)
ROA	2.929	3.968 *	8.220 ***	0.257 ***
	(1.628)	(1.856)	(4.748)	(5.061)
Intan	−5.147	−2.265	0.474	−0.041
	(−1.370)	(−0.621)	(0.140)	(−0.325)
Equity	0.740 *	0.504	0.497	0.006
	(1.748)	(1.268)	(1.401)	(0.881)
Indep	1.059	−1.164	−1.253	0.153 **
	(0.401)	(−0.466)	(−0.555)	(2.076)
Comp	4.887 ***	4.522 ***	5.655 ***	0.047
	(9.440)	(7.905)	(11.736)	(1.310)
Gexp	−2.957	35.645 **	15.112	−0.586
	(−0.143)	(2.555)	(1.036)	(−1.563)
Str	4.783	−3.520	2.896	−0.060
	(0.720)	(−0.922)	(0.664)	(−0.605)
Cons	−43.073 ***	−27.994 ***	−50.113 ***	−0.662 **
	(−4.633)	(−2.936)	(−6.501)	(−2.577)
Individual fixed effect	Yes	Yes	Yes	Yes
Time fixed effect	Yes	Yes	Yes	Yes
Observations	13,645	8626	16,406	5888
R-squared	0.798	0.752	0.798	0.683
The inter-group differences of GP	0.0719 ***		0.2199 ***

Note: ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively; the t-values in parentheses are based on heteroscedasticity-robust standard errors. The p-value of the inter-group difference test of the GP is calculated using Fisher’s combination test (with 1000 samplings).

. In non-state-owned enterprises (State = 0), the coefficient of GP is 0.253, which is significant at the 1% level; in state-owned enterprises (State = 1), the coefficient of GP is 0.181, which is significant at the 1% level; the coefficient difference of GP between the two groups is 0.0719, which is significant at the 1% level, indicating that green technology innovation has a greater promoting effect on carbon performance in non-state-owned enterprises.

The possible reason is that as “policy implementers”, state-owned enterprises have advantages in resource acquisition in green technology innovation, but their innovation behaviors may be restricted by multiple objectives, resulting in relatively low technology transformation efficiency [18]. In contrast, non-state-owned enterprises face stronger market competition pressure and survival drivers, and their green technology innovation is more likely to follow the “cost–benefit” logic: the management decision-making chain is short, enabling them to quickly transform patent achievements into emission reduction practices; at the same time, the high sensitivity of non-state-owned enterprises to shareholder value prompts them to enhance their market reputation and financing capabilities by improving carbon performance.

(2)

Heterogeneity Analysis Based on the Nature of the Industry

There are differences in technological paradigms and emission reduction paths among different industries. To examine the heterogeneous impact of the nature of the industry, referring to research by Zhang et al. [55], this study classifies industries with the 2012 edition of the industry classification codes of the China Securities Regulatory Commission as C22, C25, C26, C30-C32, D44, D45, and G56 as high-carbon industries, and the rest as non-high-carbon industries. Subsequently, we conduct grouped regression according to whether the samples belong to high-carbon industries. The results are shown in Columns (3) and (4) of Table 10. In non-high-carbon industries (High_Carbon = 0), the coefficient of GP is significantly positive; in high-carbon industries (High_Carbon = 1), the coefficient of GP is not significant; the coefficient difference between the two groups is significant at the 1% level, indicating that the promoting effect of green technology innovation on carbon performance only exists in non-high-carbon industries.

The possible reason is that high-carbon industries have significant “path dependence” characteristics. On the one hand, their production processes are highly dependent on fossil energy, and the specificity of fixed assets is strong. Green technology innovation needs to break through the traditional technical system and faces high sunk costs and transformation risks [19]. On the other hand, the carbon emission base of high-carbon industries is large. Even if green technologies are invested, in the short term, in the performance calculation of “operating revenue/carbon emissions”, the marginal decrease effect of the denominator may be diluted by the large base, resulting in insignificant innovation effects. In contrast, the production models of non-high-carbon industries are more asset-light, with lower energy consumption and carbon emission intensities. Green technology innovation is more likely to achieve “small investment, high return”. For example, software enterprises reduce the energy consumption of data centers through algorithm optimization, and the improvement of their carbon performance is directly reflected as high-value-added output corresponding to unit carbon emissions. At the same time, non-high-carbon industries are less restricted by the pressure of eliminating traditional production capacity, and innovation resources can be directly used to build low-carbon competitive advantages, forming a rapid transmission mechanism of “innovation–performance”.

The existing study shows that the impact of technological innovation also exhibits heterogeneity within high-carbon industries. Therefore, we conducted a more meticulous examination of the relationship between green technology innovation and corporate carbon performance in various high-carbon industries, and the results are shown in

Table 11. Intra-industry heterogeneity analysis within high-carbon industries.

Variables	C22	C25	C26	C30	C31	C32	D44	D45
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	CP	CP	CP	CP	CP	CP	CP	CP
GP	−0.001 *	−0.001 *	0.001	−0.001 *	−0.001 ***	−0.006	0.001 **	−0.002
	(−1.685)	(−1.734)	(0.715)	(−1.962)	(−3.129)	(−1.180)	(2.078)	(−0.452)
Menv	0.001	0.001	−0.001	−0.001	0.001	0.001	−0.001	0.001
	(0.649)	(0.108)	(−0.766)	(−0.632)	(0.299)	(0.148)	(−0.395)	(0.105)
Mshare	−0.017	−0.001	−0.009	−0.004	−0.086	0.010	0.003	−0.135
	(−1.452)	(−1.279)	(−0.725)	(−0.414)	(−1.290)	(0.318)	(1.252)	(−0.351)
Size	0.004	0.001	0.008 ***	0.006 *	0.005 *	0.054 **	0.001	−0.012
	(1.240)	(1.633)	(3.315)	(1.740)	(1.836)	(2.032)	(1.487)	(−0.209)
Lev	0.007	0.001	−0.009	−0.010	−0.009	−0.181	−0.003	0.062
	(0.703)	(0.508)	(−1.534)	(−0.983)	(−1.481)	(−1.132)	(−1.286)	(0.602)
Age	−0.011	−0.002 **	0.001	−0.010	−0.002	0.006	0.006	−0.076 *
	(−0.615)	(−2.189)	(0.053)	(−0.865)	(−0.205)	(0.105)	(1.454)	(−1.714)
ROA	0.096 ***	0.004 ***	0.124 ***	0.184 ***	0.097 ***	0.509 ***	0.036 ***	0.908 ***
	(4.500)	(5.086)	(5.580)	(3.788)	(5.013)	(3.769)	(4.934)	(2.789)
Intan	−0.050 **	0.001	0.043	−0.080 *	−0.018	−0.215	0.016 *	−0.673
	(−2.387)	(0.185)	(1.511)	(−1.921)	(−0.522)	(−0.747)	(1.685)	(−0.974)
Equity	0.001	0.001 ***	0.007	0.002	−0.002	0.014	−0.001	0.012
	(0.003)	(3.553)	(1.518)	(0.945)	(−1.597)	(1.174)	(−0.223)	(0.423)
Indep	−0.042 *	−0.002	0.030	0.007	−0.005	0.279 *	−0.006	0.384 *
	(−1.907)	(−0.950)	(1.458)	(0.391)	(−0.551)	(1.897)	(−1.171)	(1.975)
Gexp	0.264	0.002	0.213	−0.050	0.057	1.566	0.004	−0.516
	(1.517)	(0.280)	(1.083)	(−0.490)	(0.739)	(1.542)	(0.171)	(−0.396)
Str	−0.085 **	0.002	0.038	−0.030	−0.028	−0.120	−0.004	1.233 *
	(−2.090)	(0.818)	(0.428)	(−0.819)	(−1.126)	(−0.713)	(−0.342)	(1.974)
Cons	0.159 *	0.010 *	−0.092 **	0.004	0.019	−0.910	−0.016	0.351
	(1.969)	(2.043)	(−2.163)	(0.054)	(0.286)	(−1.437)	(−0.860)	(0.261)
Individual fixed effect	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Time fixed effect	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	387	237	2248	731	437	760	899	183
R-squared	0.990	0.900	0.915	0.943	0.990	0.830	0.936	0.968

Note: ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively; the t-values in parentheses are based on heteroscedasticity-robust standard errors. The industry competition variable (Comp) was deleted due to the collinearity problem; the test results of the air transportation industry (G56) were not reported due to the insufficient estimation sample.

. As can be seen, in the paper and paper products industry (C22), the petroleum processing, coking, and nuclear fuel processing industry (C25), the non-metallic mineral products industry (C30), and the ferrous metal smelting and rolling processing industry (C31), the estimated coefficients of GP are significantly negative. In the electricity, heat production, and supply industry (D44), the estimated coefficient of GP is significantly positive. In the chemical raw materials and chemical products manufacturing industry (C26), the non-ferrous metal smelting and rolling processing industry (C32), and the gas production and supply industry (D45), the estimated coefficients of GP are not significant. The above results indicate that even within high-carbon industries, the impact of green technology innovation on carbon performance varies.

Most of the industries where the impact is significantly negative are traditional industries with high energy consumption and high pollution. Their production processes rely on fossil energy, and it is difficult to transform their technologies. In the initial stage of green technology innovation, a large amount of capital investment is required. In the short term, it may drive up production costs instead of directly reducing carbon emissions. Even due to the long technology transformation cycle, the carbon performance may be temporarily under pressure. The industries where the impact is significantly positive are mainly the electricity and heat production industries. Green technologies in this field (such as clean energy power generation and carbon capture technology) have the attribute of direct emission reduction, and policy-driven measures promote the rapid implementation of these technologies. Under the effect of economies of scale, innovation has a more significant improvement on carbon performance. In industries where the impact is not significant, such as the chemical raw materials, non-ferrous metal smelting, and gas industries, their technological innovation may be limited by the characteristics of the industries. In the chemical industry, there are diverse technological paths. The non-ferrous metal smelting industry is restricted by the endowment of mineral resources and the complexity of the smelting process. The gas industry is affected by the speed of energy structure transformation, resulting in the emission reduction effect of green innovation not being fully manifested or being offset by other factors (such as the expansion of production scale).

(3)

Heterogeneous Analysis Based on Innovative City Policy

Innovative pilot cities typically enjoy policy dividends (such as fiscal subsidies, tax incentives, and industry–university–research cooperation platforms), which can provide more abundant financial, talent, and technical resource support for enterprises’ green technology innovation, thereby potentially strengthening the improvement effect of green technology innovation on carbon performance. At the same time, pilot cities often have more complete environmental governance institutions, stricter carbon emission regulatory systems, and higher social environmental awareness, forming a dual mechanism of “policy incentives + market backward pressure” to urge enterprises to more actively translate green technology innovation into actual carbon reduction outcomes. Theoretically, this analysis aligns with the logic of the Resource-Based View (RBV) and institutional theory—the differentiated resource acquisition capabilities empowered by policies and institutional environments lead to heterogeneity in the output of green technology innovation and environmental governance efficiency among enterprises in different regions. Additionally, as an exogenously given regional development strategy, the innovative city policy has clear boundaries in terms of pilot scope and timelines, providing an objective and observable basis for group comparisons. This helps identify differentiated pathways through which green technology innovation influences carbon performance under policy drivers, offering empirical evidence for optimizing regional low-carbon policy supply.

To examine the heterogeneous effects of the Innovative City Policy, this study groups listed firms based on whether their registered addresses are located in innovative pilot cities. The test results are presented in

Table 12. Heterogeneity analysis (2).

Variables	ICP = 0	ICP = 1
	(5)	(6)
	CP	CP
GP	0.199 **	0.211 ***
	(2.327)	(3.416)
Menv	0.021	−0.084
	(0.160)	(−0.647)
Mshare	−0.728	−4.728 **
	(−0.346)	(−2.428)
Size	1.175 ***	0.749 **
	(3.339)	(2.526)
Lev	2.457 **	3.042 ***
	(2.102)	(2.714)
Age	−1.871	4.227 **
	(−0.781)	(2.043)
ROA	0.648	4.072 **
	(0.324)	(2.154)
Intan	−3.550	−6.590 *
	(−0.952)	(−1.808)
Equity	0.847 **	0.619
	(1.967)	(1.519)
Indep	2.116	−1.754
	(0.841)	(−0.676)
Comp	4.921 ***	4.867 ***
	(7.792)	(9.511)
Gexp	42.390 **	−8.998
	(2.523)	(−0.509)
Str	−2.419	1.262
	(−0.486)	(0.221)
Cons	−37.642 ***	−42.695 ***
	(−3.878)	(−4.887)
Individual fixed effect	Yes	Yes
Time fixed effect	Yes	Yes
Observations	9510	12,790
R-squared	0.780	0.789
Inter-group differences in GP	−0.0124 **

Note: ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively; the t-values in parentheses are based on heteroscedasticity-robust standard errors. The p-value of the inter-group difference test of the GP is calculated using Fisher’s combination test (with 1000 samplings).

. In non-pilot cities (ICP = 0), the coefficient of GP is 0.199, significant at the 5% level; in innovative pilot cities (ICP = 1), the coefficient of GP is 0.211, significant at the 1% level. The inter-group coefficient difference of GP is −0.0124, significant at the 5% level. These results indicate that the effect of green technology innovation on corporate carbon performance is more pronounced in innovative pilot cities.

The core reasons for this difference are threefold: First, innovative pilot cities accelerate the practical transformation of green technologies by establishing more comprehensive green innovation ecosystems (e.g., intra-cluster technology sharing and supply chain collaborative emission reduction mechanisms). Second, the first-mover advantage of pilot cities prompts enterprises to integrate green innovation into their strategic cores earlier, forming a virtuous cycle of “innovation–emission reduction”. Third, the benchmarking effect of policy pilots may attract more cross-regional collaborations and green investments, further amplifying the environmental governance efficacy of green technologies. This finding suggests that the Innovative City Policy significantly enhances the marginal contribution of green innovation to carbon performance by optimizing the allocation of innovation factors and industrial collaboration networks, providing new insights for the precise implementation of regional low-carbon policies.

6. Discussion and Conclusions

6.1. Discussion

The core findings of this study indicate that the promotion effect of green technology innovation on corporate carbon performance has significant robustness and a multi-dimensional transmission mechanism. This conclusion not only echoes the micro-logic of the “Porter Hypothesis” but also provides an action path for enterprises’ low-carbon transformation.

From a theoretical perspective, green technology innovation improves carbon performance through a dual path of enhancing total factor productivity and environmental governance capabilities, confirming the view that technological innovation is not only a “cost consumption” but also “ability building”; the former reflects the direct optimization of innovation on resource allocation efficiency, and the latter reflects the systematic shaping of innovation on the organization’s emission reduction capabilities. The two together constitute a low-carbon transformation model of “technology–ability” co-evolution.

Green media attention and investors’ green attention in the external context play a positive moderating role, revealing the “catalytic effect” of social supervision and the capital market in innovation transformation: the media forces the implementation of innovation by constructing a public opinion field of environmental responsibility, and investors guide the flow of resources through their ESG investment preferences. The two jointly reduce the “idle cost” of innovation achievements and strengthen the transformation efficiency of technological advantages into carbon performance.

The analysis of the long-term effect shows that the impact of green technology innovation on carbon performance presents the characteristics of “continuously significant but with diminishing marginal effects”. This is not only in line with the “time lag” law of technological innovation (it takes a transformation cycle from patent application to production application) but also reflects the phased release of emission reduction potential—initial innovation breaks through technical bottlenecks—bringing significant efficiency improvements, and in the long term, as the space for emission reduction narrows, the effect stabilizes.

The discovery of the “advantages of non-state-owned enterprises and non-high-carbon industries” in the heterogeneity analysis is essentially the result of differences in governance mechanisms and industry technological paradigms: the market-oriented incentive mechanism of non-state-owned enterprises reduces agency costs, making it easier for them to transform innovation investments into substantial emission reduction actions; the asset-light model of non-high-carbon industries is more adaptable to low-carbon technologies, avoiding the transformation resistance caused by the “path dependence” of high-carbon industries. These conclusions provide a basis for the precise implementation of the “dual carbon” policy. For example, for state-owned enterprises, it is necessary to optimize the innovation assessment system to reduce the impact of administrative intervention, and for high-carbon industries, it is necessary to increase support for the research and development of common technologies to break through the capacity locking effect.

This study makes three key contributions to the literature. First, it advances the theoretical understanding of green technology’s dual-role in carbon performance by proposing a “productivity–governance synergy framework”. Unlike prior research that often isolates single mechanisms, this study demonstrates that green technology innovation improves corporate carbon performance through two interdependent pathways, enhancing total factor productivity (TFP) to drive operational efficiency and strengthening environmental governance capabilities to ensure regulatory compliance. This dual-path model enriches the discourse on how technological innovation fosters sustainable performance by integrating both efficiency improvements and institutional adaptation.

Second, this research introduces a novel perspective on external supervision’s moderating role in low-carbon transformation. By incorporating green media attention and investor green attention as moderators, it reveals how public opinion and capital market dynamics amplify the impact of green technology innovation on carbon performance. This highlights the synergistic effect between market intermediaries and corporate innovation, offering new insights into how external pressure activates internal innovation effectiveness; a dimension underdeveloped in existing studies.

Third, the analysis uncovers the dynamic and context-dependent nature of innovation impacts by examining both long-term marginal returns and heterogeneous enterprise conditions. It shows that while green technology generates sustained positive effects on carbon performance, its marginal benefits diminish over time due to technology diffusion, and its efficacy varies significantly across ownership structures (e.g., stronger effects in non-state-owned enterprises) and industry carbon intensities (more pronounced in non-high-carbon sectors). This nuanced understanding of situational dependencies addresses a critical gap in research that has historically overlooked temporal and contextual complexities.

6.2. Conclusions

Against the backdrop of the intensification of global climate change and driven by China’s “dual carbon” goals, green technology innovation, as the core driving force for enterprises’ low-carbon transformation, has become the focus of attention in both academic and practical circles regarding its impact mechanism on carbon performance and situational effects. This study takes Chinese A-share listed companies on the Shanghai and Shenzhen stock exchanges from 2007 to 2022 as the research objects. Through theoretical analysis and empirical testing, it systematically explores the impact paths, moderating factors, long-term effects, and heterogeneous characteristics of green technology innovation on corporate carbon performance.

We find that green technology innovation can significantly enhance corporate carbon performance. This conclusion remains valid after multiple robustness tests, such as balancing sample characteristics through propensity score matching, addressing reverse causality through the instrumental variable method, and correcting sample selection bias through the Heckman two-stage method. Further analysis shows that green technology innovation improves carbon performance through two core mechanisms: Firstly, by enhancing total factor productivity, it optimizes resource allocation efficiency, achieving higher economic output with fewer factor inputs. Secondly, by strengthening environmental governance capabilities (such as the reserve of pollution prevention and control technologies and the construction of environmental management systems), it builds systematic emission reduction capabilities. In terms of the influence of the external situation, green media attention and investors’ green attention play a positive moderating role. The former strengthens the actual transformation of innovation achievements through social supervision, and the latter amplifies the emission reduction effect of innovation through the resource allocation of the capital market. The analysis of the long-term effect shows that green technology innovation has a significant effect on improving carbon performance in the next 1–4 periods. However, the marginal effect decreases over time, reflecting the time lag in technology transformation and the phased release characteristics of emission reduction potential. The heterogeneity analysis reveals that the effect of green technology innovation on improving carbon performance is more significant in non-state-owned enterprises, non-high-carbon industries, and innovative cities. The former benefits from higher innovation transformation efficiency under the market-oriented governance mechanism, and the latter stems from the stronger adaptability of the asset-light model to low-carbon technologies.

6.3. Policy Implications

Based on the research conclusions, we put forward the following policy recommendations:

Firstly, we recommend strengthening the core driving role of green technology innovation and construct a “efficiency-ability” two-wheel policy system. This research confirms that green technology innovation significantly improves corporate carbon performance by enhancing total factor productivity and environmental governance capabilities, and the long-term effect persists, although with diminishing marginal returns. It is necessary to construct a policy support framework covering the entire cycle of “research and development–transformation–application”. Specific measures include: increasing investment in research and development and the construction of common technology platforms, providing special subsidies for key links in the transformation of green patents, reducing the cost of enterprise innovation transformation, and especially focusing on the breakthrough of common technologies in high-carbon industries to alleviate the transformation resistance caused by their “path dependence”; establishing an evaluation system for enterprises’ environmental governance capabilities, incorporating indicators such as investment in environmental protection facilities and certification of carbon emission management systems into the assessment indicators for policy support, and guiding enterprises to shift from “end-of-pipe treatment” to “construction of systematic emission reduction capabilities”. For example, enterprises that have passed the ISO 14001 certification or established a carbon footprint tracking system can be given tax incentives; implementing a dynamic linkage subsidy of “innovation–performance”, considering the characteristic of diminishing marginal returns in the long-term, providing step-by-step rewards for enterprises whose continuous transformation of green technology innovation achievements meets the standards, encouraging them to form a long-term emission reduction mechanism, and avoiding short-term “greenwashing” behaviors.

Secondly, we recommend optimizing the external governance environment and releasing the synergistic catalytic effect of green media attention and the capital market. Green media attention and investors’ green attention positively moderate innovation efficiency through public opinion supervision and resource allocation. It is necessary to strengthen the guiding role of the two in enterprises’ low-carbon transformation from the institutional level. Policy design should focus on improving the environmental information disclosure system; requiring enterprises to disclose data related to green technology innovation investment, patent transformation progress, and carbon performance; reducing information asymmetry in media supervision and promoting the shift of green media attention from “issue exposure” to “effect tracking”; and guiding the capital market to construct an ESG investment ecosystem. By setting up green transformation guiding funds and improving ESG rating standards, institutional investors are encouraged to increase their long-term shareholdings in green technology innovation enterprises. At the same time, listed companies are encouraged to incorporate carbon performance into the performance evaluation system of senior executives’ salaries, and the governance intervention of investors’ attention in innovation transformation is strengthened. This involves establishing a “media–regulatory–market” linkage mechanism, providing financing convenience or preferential market access for enterprises that actively respond to green media supervision and achieve an improvement in carbon performance, forming a positive cycle of “social supervision–market rewards–innovation investment”, and amplifying the activation effect of the external governance environment on innovation efficiency.

Thirdly, we recommend implementing differentiated policy guidance to break through the transformation constraints of the nature of property rights and industry characteristics. The heterogeneity analysis shows that non-state-owned enterprises and non-high-carbon industries have more advantages in innovation transformation. It is necessary to design precise policies for the special constraints of state-owned enterprises and high-carbon industries. The key directions include optimizing the innovation assessment mechanism of state-owned enterprises, adding market-oriented indicators such as the “conversion rate of green patents” and the “improvement range of carbon performance” to the performance evaluation system, and reducing the impact of administrative intervention on the efficiency of technology transformation; for example, linking the emission reduction effect with the promotion of management and promoting the transformation of state-owned enterprises from “policy-driven innovation” to “efficiency-driven innovation”, and establishing a special support plan for the green technology transformation of high-carbon industries. In view of the characteristics of strong specificity of fixed assets and high transformation costs in industries such as steel and chemical engineering, policies such as subsidies for equipment renewal and preferential allocation of carbon emission trading rights should be provided to reduce the sunk costs of their technology application. At the same time, industry common technology platforms should be constructed to accelerate the large-scale application of low-carbon technologies, cultivating the low-carbon competitive advantages of non-high-carbon industries. Systems such as green certification and carbon footprint labeling can help industries such as information technology and high-end manufacturing to transform their innovation achievements into market premiums; for example, giving priority to purchasing goods from enterprises that have passed low-carbon product certification, guiding consumers to choose high-value-added products driven by low-carbon technologies, and strengthening the innovation incentives for non-high-carbon industries.