How does carbon emission trading scheme affect enterprise green technology innovation: evidence from China’s A-share non-financial listed companies

More and more emphasis is placed on the common development of economy and ecological environment in China’s development strategy, and one of the key solution is green technology innovation of enterprises. This paper takes the carbon emission trading scheme carried out in China in 2013 as a quasi-natural experiment and uses the data of China’s A-share non-financial listed companies and the DID method to empirically test the impact of the scheme on enterprise green technology innovation from the micro level. The results suggest that the carbon emission trading scheme has a significant role in promoting enterprise green technology innovation, mainly through the innovation of green practical patents and alternative energy-based patent. Using a series of robustness tests such as dynamic effect test, placebo test, and PSM-DID, it is found that the results are still valid. Further analysis finds that debt financing will weaken the role of carbon emission trading scheme in promoting enterprise green technology innovation. And the carbon emission trading scheme plays a greater role in promoting green technology innovation in state-owned enterprises, enterprises belonging to areas with strong pollution control intensity and high pollution industry.


Introduction
As China's contribution to the world economy grows, it has become the largest contributor to global carbon emissions. The enterprise green technology innovation is essential to sustainable development of economy and ecological environment. Since China's economic growth has entered the new normal, its development concept also altered from "focusing only on GDP" to the new concept of "innovation, coordination, green, openness, and shared." However, the increase in carbon dioxide emissions not only reduce social welfare but also cause serious damage to national health and ecological environment during the process of economic transformation. In the long run, they have a negative impact on the sustainable development of China's economy. Based on CSMAR database, Chinese total carbon dioxide emissions increased from 3.35 gigaton in 2000 to 9.54 gigaton in 2012, representing more than twice of total carbon dioxide emissions. And Chinese total carbon emission has ranked first in the world since 2005. To achieve the goal of lowcarbon emission reduction, China implemented the carbon emission trading market policy during 2013-2014, and successively established pilot projects in Beijing, Shanghai, Guangdong, Shenzhen, Tianjin, Chongqing, and Hubei. "The 13th Five-Year Plan" clearly states: "China will establish and improve the initial allocation system for energy use rights, water rights, pollution rights, and carbon emission rights, and cultivate and develop the trading market." In September 2020, General Secretary Xi Jinping formally proposed the "dual carbon goal" at the UN General Assembly. These policy show that resource and environmental issues have become the bottleneck restricting the high-quality development of China's economy, and have risen to major issues in national politics, people's livelihood, diplomacy, and strategic development (Wang and Wang 2011).
Traditional technology innovation fails to achieve the goal of reducing environmental pollution, and cannot meet the needs of green economy and sustainable development. As a method of technology innovation, green technology innovation has the dual advantages of environmental protection and high-quality economic development (Yuan and Chen 2019;Liu et al. 2022b). Braun and Wield (1994) firstly proposed the concept of green technology innovation, which refers to the sum of technologies or products that reduce energy, raw material consumption, and pollutant emissions. Environmental innovation is complicated, and external knowledge is an important driver (Bai et al. 2020). At present, existing literature explore the factors affecting green technology innovation from various aspects, such as environmental regulation, green credit, foreign direct investment, and board governance Lu et al. 2021;Li et al. 2016;Wang and Chen 2018). In order to implement the ecological concept of "Clear waters and green mountains are as good as mountains of gold and silver," the Chinese government also vigorously encourages green technology innovation. The report of the 19th CPC National Congress in 2017 proposed "building a market-oriented green technology innovation system," which shows that the concept of green technology innovation entered the highest programmatic document of the party for the first time; In 2019, the National Development and Reform Commission and the Ministry of Science and Technology jointly issued the "Guiding Opinions on Building a Market-Oriented Green Technology Innovation System," which marks that green technology innovation has officially become a special government policy document. To sum up, green technology innovation has become topic of common concern in academia and practice.
China is facing the dual challenges of curbing its everincreasing carbon emissions and maintaining high-quality economic growth. Enterprise green technology innovation is an indispensable factor to solve problems. However, only relying on enterprises to carry out innovation lacks motivation in real life and the government can influence enterprise decision-making through environmental regulation. As a market environmental regulation, carbon emission trading scheme has developed rapidly in recent years, which is more flexible and effective. Therefore, this paper discusses the relationship between carbon trading scheme and enterprise green technology innovation in detail. This paper seeks to address the following three questions. First, can carbon emission trading scheme stimulate enterprise green technology innovation? Second, if it is, what are the influencing mechanisms? Third, does the positive effect of carbon emission trading scheme stimulate on enterprise green technology innovation vary based on different corporate characteristics?
To answer these questions, this paper empirically testes the impact of the scheme on enterprise green technology innovation by using the data of China's A-share non-financial listed companies from 2009 to 2017.
The existing researches on carbon emission trading scheme and enterprise green technology innovation mainly analyzed from the macro perspective (Zhou and Wang 2022;Wang and He 2022;Liu et al. 2022a;Liu and Sun 2021), and the researches on enterprise level are shallow . The main contributions of this paper are as follows: First, this paper uses China's A-share non-financial listed companies data to discuss the relationship between carbon emission trading and green technology innovation of enterprises from the micro level, which enriches the research in carbon emission trading scheme. Second, this paper explores deeply the structure of enterprises green technology innovation in terms of form and content. Third, this paper starts from the perspective of corporate funds, and studies the moderating effect of debt financing in the carbon trading scheme and corporate green technology innovation. Finally, this paper evaluates the policy effect of the carbon emission trading scheme objectively and roundly, from the aspects of enterprise nature, environmental regulation intensity, industry nature and so on. This paper has important policy implications for further improving and deepening the carbon emission trading scheme.

Literature review and theoretical hypothesis
To a large extent, carbon emission problems come from technologies in the production process of industry especially the manufacturing industry. But green technology innovation is the key to fundamentally solve this environmental problem and a significant approach to a win-win path of economic development and environmental protection (Liao et al. 2018). At present, there is a lot of researches on green technology innovation, and scholars start from internal and external factors such as digital finance (Cao et al. 2021), political connections , energy poverty eradication , information infrastructure , green credit , directors with Overseas Background (Shen et al. 2022), and environmental awards (Lai et al. 2022). However, from the perspective of the carbon emission trading scheme, there is little literature that deeply analyzes the internal logic of the carbon trading scheme and enterprise green technology innovation. Most of the technological innovation of enterprises is based on the perspective of cost-effectiveness and strategic development. However, the green technology innovation of enterprises generally has characteristics such as large capital investment, long construction period, high adjustment cost, and low initial income. Enterprise green technology innovation has dual positive externalities of environment and innovation leads to market failure (Bian et al. 2021), and the short-term economic and long-term social benefits brought about by enterprises' green environmental protection strategies are not always achieved concurrently (Peng et al. 2021). They lead to insufficient motivation to rely solely on enterprises for green technology innovation and needs external forces, in other word, the government influences enterprise decisionmaking through environmental regulation. But the traditional economics believe that environmental regulation will lead to the rising transaction cost of enterprises, which is not conducive to the improvement of productivity. It will also reduce market competitiveness, financial performance and produces follow-cost effect which inhibit green technology innovation to a certain extent. Some scholars find that environmental regulation has a negative effect on enterprise technology innovation (Blackman and Kildegaard 2010;Kneller and Manderson 2012;Yuan and Xiang 2018). The Porter Hypothesis (Porter 1991) holds that strict and appropriate environmental regulation helps enterprises to innovate technologically and reduce production costs. Besides, the compensation effect of innovation can partially or completely offset environmental regulation. Environmental regulation does not cause a decline in corporate productivity. Meanwhile, it generates net benefits and achieves the win-wins of environment and economy (Chakraborty and Chatterjee 2017). The porter effect can be measured from technological innovation (Yang et al. 2020), economic dividend, and environmental dividend (Dong et al. 2019). The existing literature confirms the Porter hypothesis from SO2 Emission rights trading policy, low-carbon city policy, policy of raising emission fees, the monitoring policy, disclosure program, and regulations of air pollution (Qi et al. 2018;Xu and Cui 2020;Chen et al. 2021;Sun et al. 2022;Zhang and Zhao 2022;. These articles fully demonstrate that the direction of technological progress is pathdependent, as well as reasonable environmental regulation can change the direction of technological progress. There are three different types of environmental regulation, namely imperative type, market incentive type, and voluntary type, which are valid within limits (Peng et al. 2021). They have obvious differences in the mechanism on technological innovation, among which the market type is more flexible and effective (Anderson et al. 2010;Johnstone et al. 2017). Market incentive environmental regulation is different from strict environmental management and control, which guides the environmental management behavior of enterprises based on clear price signals. It can be realized through direct price control, such as environmental protection tax (Bovenberg et al. 1997) and green technology subsidies. Besides, it can also be realized through quantity control such as emission rights trading (Faere et al. 2007). The main market-based environmental regulatory policies are commonly adopted in China include the following: environmental taxes and emissions trading mechanisms (Zhu et al. 2022). As a market environmental regulation, carbon emission trading scheme has developed rapidly in recent years. It defines the subject and quota of carbon emission rights, and sets up a corresponding punishment mechanism. Enterprises can obtain the emission allowances through the initial allocation by the government as well as sell or buy emission allowances through the carbon emission rights trading market. The carbon emission trading scheme can induce green technology innovation of enterprises because it brings the firm significant cost pressure or economic incentives (Yuan and Chen 2019). In terms of production cost, the carbon emission trading scheme increases the firm's production cost by increasing the price of exogenous energy, forcing enterprises to carry out green technology innovation. In terms of incentives, enterprises can store excess pollution indicators for backup or sell them for profit. Driven by profit maximization, enterprises will consciously seek green technology innovation to reduce the production and emission of their own pollutants; Enterprises trade the remaining emission rights to other enterprises with sewage needs, which provide continuous "dynamic incentives" for enterprises (Ren et al. 2019). Although the carbon trading mechanism will bring certain cost pressure to enterprises in the short term, it can stimulate enterprises through compliance pressure and economic compensation effects in the long run.
Schumpeter's innovation theory finds that the availability of funds plays an important role in innovation: sufficient and continuous supply of funds is a prerequisite for technological innovation of enterprises. The Priority-Order Financing Theory (Myers and Majluf 1984) further finds that equity financing can effectively make up for the shortage of funds for enterprise innovation activities, and pay more attention to the sustainable growth of enterprises brought about by technology research and development. Eventually, it ensure the continuity of future innovation investment of enterprises. However, debt financing relying on fixed income has a fierce contradiction with the high-risk characteristics of innovation, thus plays an inhibitory effect on technological innovation. The specific manifestation is that the contract rigidity of debt financing puts forward higher requirements on the solvency of the company. When a company has problems paying off its debt, creditors tend to resort to the law instead of tolerance. This behavior increases the risk of company bankruptcy and the career anxiety of company executives. Finally, it reduces the firm's motivation to carry out green technology innovation. When debt financing gives company executives higher control rights and less supervision, managers are in pursuit of opportunistic behavior and pleasure, while enterprise results in lower corporate governance. In conclusion, debt financing is adverse to the development of corporate green technology innovation (Jiang et al. 2021).
The financing matching mechanism points out that the inherent characteristics of debt financing matches the innovation activities of enterprises without considering other factors, and debt financing can effectively promote enterprise green technology innovation. The debt financing is in line with the long-term technological innovation cycle of enterprises, and meets the needs of enterprises for R&D funds in a certain period of time because of low financing cost. Compared with bank loans, corporate debts meet the needs of most investors on account of stronger liquidity and flexibility. At the same time, relational creditors are gradually showing the characteristics of innovation as well as inclusiveness, and achieve win-win goal by promoting firm green technology innovation (Wen et al. 2011;David et al. 2008). In order to achieve the dual goals of protecting the ecological environment and economic development, the carbon trading scheme should be effectively passed on to enterprises to guide enterprise green technology innovation. This requires not only the continuous improvement of the carbon trading scheme and the strict law enforcement by local governments, but also the active cooperation of enterprises. On the one hand, the debt financing weakens the role of the carbon trading scheme in promoting enterprises green technology innovation because of its characteristics of fixed income and low corporate governance. On the other hand, the debt financing enhances the role of the carbon trading scheme in promoting enterprises green technology innovation due to its characteristics of long term and low cost.
Based on the above analysis, this paper proposes the following hypotheses: • Hypothesis 1: Carbon trading scheme will promote enterprise green technology innovation. • Hypothesis 2-1: Debt financing will weaken the role of carbon emission trading scheme in promoting enterprise green technology innovation. • Hypothesis 2-2: Debt financing will strengthen the role of carbon emission trading scheme in promoting enterprise green technology innovation

Explained variables
This paper used the number of green patent authorizations to represent enterprise green technology innovation for three reasons. First, availability and accuracy of the number of green patent authorizations data is excellent. At present, there are many indicators to measure enterprise innovation, such as R&D expenses, innovation subsidies, the number of researchers. They have difficulty distinguishing whether green technology innovation is green or not, but the data on green patents of companies released by the State Intellectual Property Office is more accurate. Second, patent data is the direct result of enterprise technology innovation, it can better measure the degree and quality of enterprise innovation activities. Third, there are some speculations in patent applications, while patent authorization can effectively discriminate the low-quality enterprise patent application to a certain extent. Thus, the high-quality characteristics of enterprise invention patent are guaranteed. Therefore, this paper refers to the practice of Tao et al. (2021), and uses the number of green patent authorizations of enterprises to expresses it.

Explanatory variables
From 2013 to 2014, China implemented the carbon emission trading market policy and successively established pilot projects in seven provinces and cities, including Beijing, Shanghai, Guangdong, Shenzhen, Tianjin, Chongqing, and Hubei. This paper uses treat as a dummy variable. If the enterprise is located in a pilot area where the carbon trading scheme implemented in 2013 and 2014, the value is 1; otherwise, it is 0; time is also a dummy variable, which is assigned to each year after 2013 is 1; otherwise, it is 0.

Control variables
In order to eliminate the interference of other variables and obtain the objective estimate of policy effect, this article refers to relevant literature (Qi et al. 2018;Jiang et al. 2021;Wang and Wang 2011) and select the following control variables, including: (1) Tobin Q value (Tobin Q); (2) Enterprise size (Cap), measured by the logarithm of the total of each item of shareholders' equity; (3) Age of the enterprise (Age); (4) Asset-liability ratio (Debt); (5) Enterprise number of employees (Labor); (6) Net profit margin on assets (ROA); (7) Capital intensity (Cap_inten), expressed by the ratio of total assets to operating income; (8) Cash ratio (Cash); (9) Board size (Board); and (10) Proportion of independent directors (Ind).

Data
The green patent data and financial data of listed companies used in this paper are from the CSMAR database, and the content of the pilot policy of the carbon emission trading scheme is from the China Carbon Emissions Trading Network. This paper takes 2009-2017 Chinese A-share non-financial listed companies as a sample and excludes ST and * ST listed companies. In data processing, the data of corporate green patents as null values are assigned as 0, and the corporate financial data is null values are imputed by interpolation, resulting in 13,851 company-year sample data.

Model establishment
In the field of environmental economics, international cutting-edge research usually adopts regression discontinuity, synthetic control method, matching method, and differencein-differences model. The difference-in-differences model compares the difference between the influence of the change before and after the policy on the pilot area (experimental group) and the non-pilot area (control group). The net effect of policy disposition is obtained by removing the factors that change over time, so as to evaluate the causal promotion effect of policy. To be specific, firstly the difference between each group before and after the implementation of the policy is made, and then, the difference between the treatment group and the control group is made by second-order difference to eliminate the common trend of the two groups, and finally get the implementation of the policy. Therefore, this paper constructs the following difference-in-differences model In order to examine the impact of the carbon trading scheme on firm green technology innovation.
Among them, Green i,t is the dependent variables, which indicate the degree of green technology innovation of enterprise in year t, and it is expressed by the number of green patents authorized by the enterprise in that year; the double difference item treat i × time t is explanatory variable to measure whether the enterprise is affected by the carbon trading scheme; CVs i,t is the control variables, this paper also controls year (Year) and industry (Indcd), and i,t is random error term.

Empirical results and analysis
Descriptive statistics Table 1 reports the descriptive statistical analysis of the sample variables. Because this paper selects balanced panel data, the numbers of different variables' observations are all same. The mean and standard deviation of the number of green patent authorizations are 3.05 and 23.69 respectively, implying that substantial differences exist in enterprise green technology innovation among the sample firms. The mean of green invention patent (Green_I) and green utility model patent (Green_P) are 1.05 and 2.00 respectively, which shows that different forms of green technology innovation vary greatly; The mean values of alternative energy (Green_Sub) and energy conservation (Green_Eco) patents are 0.80 and 0.87 respectively, which indicates that enterprises have the same (1) choice of these two types of green technology innovation. Other control variables are not repeated here.

Benchmark regression results
This paper uses the double-difference model to evaluate the impact of carbon emission trading scheme on firm green technology innovation. The benchmark regression results are shown in Table 2. The estimated coefficient of column (1) is significantly positive. Hypothesis 1 is thus verified; in other word, carbon emission trading scheme will promote enterprise green technology innovation.
The carbon emission trading scheme affects not only the number of green patents granted by enterprises but also their structure. According to the practice of Xu and Cui (2020) and Qi et al. (2018), green patents include green invention patents (Green_I) and green practical patents (Green_P), and green invention patents have more requirements and higher innovation content. The other classification is the based on the United Nations Framework Convention on Climate Change, where green patents include seven types and the green technology innovation effect of enterprises is mainly reflected in alternative energy patents (Green_Sub) and energy saving patents (Green_Eco). As shown in Table 2, the regression coefficient in columns (2) and (3) is positive and significant at the 1% level, which indicates that the carbon emission trading scheme has a positive effect on both green invention patents and green practical patents, and the positive effect on green practical patents is stronger. The regression coefficients in columns (4) and (5) are also significantly positive, and the regression coefficient in column (4) is larger, which indicates that the carbon

Dynamic effect test
Through the dynamic effect test, this paper not only verifies the parallel trend hypothesis, but also studies the dynamic effect of the carbon emission trading scheme on enterprise green technology innovation. Figure 1 shows that there was no significant difference between the experimental group and the treatment group in terms of enterprise green patent authorization before 2013, which satisfies the parallel trend assumption. After the implementation of carbon emission trading scheme, there is a significant difference between the experimental group and the treatment group in terms of enterprise green technology innovation. The influence of carbon emission trading scheme on enterprise green technology innovation reached a peak after four years, which indicates the implementation of the policy has a certain time lag.

Placebo test
The policy implementation time in the research sample is 2013, and the number of enterprises in the experimental group is 576. In order to eliminate the influence of omitted variables or artificial settings, this paper conducts a placebo test. The implementation time of the carbon emission trading scheme remains unchanged, and 576 listed companies are randomly selected as the experimental group to reestimate the model (1). This paper repeats the above process 500 times, thus obtains 500 estimated coefficients of treat i × time t . Figure 2 shows that the mean of the regression coefficients of the double difference term is close to 0. The result shows that the random selection of treatment groups makes no difference on enterprise green technology innovation, and confirms that the carbon emission trading scheme has a positive effect on enterprise green technology innovation from a counterfactual perspective.

PSM-DID robustness test
In order to overcome the bias caused by sample selection, this paper refers to the method of Shi et al. (2018) and uses the PSM-DID model. In this paper, banks in the treatment group and control group are matched according to 10 observable variables such as Tobin Q, Cap, Age, Debt, Labor, ROA, Cap_inten, Cash, Board, and Ind. When the PSM-DID method is used in this paper, the logit regression is performed on the control variables through the dummy variable of whether it is an enterprise in the pilot area of carbon trading, and the propensity score value is obtained. The enterprise with the closest propensity score value is the enterprise paired in the carbon trading pilot area. Figures 3 and 4 show that the total bias decreases significantly after sample matching and is less than the 10% red line standard stipulated by the balance test, and there is no systematically significant difference between the new samples after matching by the PSM method. The propensity score matching method can effectively reduce the difference in the distribution of explanatory variables between the control group and the treatment group, and eliminate the sample estimation bias caused by self-selection. The coefficient of treat i × time t is still positive at the 1% level in Table 3, and supports the results of the benchmark regression that carbon emission trading scheme will promote enterprise green technology innovation.

Excluding the influence of other policies
In order to achieve the coordinated development of environment and economy, the Chinese government has formulated various environmental regulations to promote energy conservation and emission reduction of enterprises. However, different policies may overlap and the level of corporate green technology innovation may be affected by multiple policies. To exclude the influence of other policies, this paper mainly considers the following two environmental regulations: (1) 2010 Pilot Policies for Low-Carbon Provinces and Low-Carbon Cities. Its goal is to reduce carbon dioxide emissions in line with the carbon trading scheme; thus, it is important to exclude the policy.
(2) 2007 SO 2 emissions trading pilot policy. As the first market-based environmental regulation, the pilot policy of SO 2 emissions trading in 2007 will inevitably affect carbon emission trading scheme, thus it is necessary to exclude the policy. In order to reduce the interference of the above two (2) In model (2), DID01 and DID02 are the difference-differences estimates of the pilot policies for low-carbon provinces and cities in 2010, and 2007 SO 2 emissions trading pilot policy respectively (Table 4). If province i becomes a lowcarbon province and low-carbon city pilot area in year t, then province i in year t and later years DID01 i,t =1; otherwise, it is 0. DID02 i,t are obtained based on the same approach. Table 5 reports the corresponding estimation results. This paper finds that the net effect of the carbon emission trading scheme on enterprise green technology innovation is still significantly positive, and it is promoted.

Substitution of dependent variable
Referring to the practice of Tao et al. (2021) and He et al. (2019), this paper chooses to use the number of green patent applications (Green Apply) and the ratio of R&D investment to operating income (RD) as the dependent variables to replace the number of green patent authorizations (Green). They are substituted into model (1) for regression respectively, and the results are shown in Table 5. The estimated coefficients in columns (1) and (2) are all significantly positive, which indicates that the carbon trading scheme has a significant positive effect on enterprise green technology innovation.

Heterogeneity analysis
The impact of the implementation of the carbon trading scheme on enterprises with different ownership varies wildly. Relying on their strong financial strength and strong government support, state-owned enterprises have less financing constraints in obtaining green credit support. The capital advantages, talent advantages, and standardized technology paradigms owned by state-owned enterprises ensure the multi-level and continuous investment in enterprise green technology innovation, and achieve good control of innovation risks. In the process of pursuing wealth maximization, state-owned enterprises often face greater public opinion and supervision, so they are more sensitive to environmental problems. However, state-owned enterprises are often insensitive to efficiency improvement information and technological innovation incentives provided by external market in the implementation process of environmental regulation policy, thus they have insufficient motivation for green technology innovation. Therefore, according to the different nature of enterprise ownership, this paper divides the research sample into state-owned enterprises and non-stateowned enterprises and conducts regressions respectively. In empirical analysis, grouping test is often needed to test the influence of explanatory variables on the explained variables under different circumstances, and then the coefficient difference between the two samples is compared. Therefore,    (1) and (2) of Table 6. The P-value of SUEST test for the regression coefficient of treat i × time t is significant under the level of 1%. The regression coefficient of column (1) is 5.5135, which is significant at the level of 1%. The regression coefficient of column (2) is significantly positive at the level of 10%. The results show that there are significant differences in the carbon emission trading scheme between state-owned enterprises and nonstate-owned enterprises and carbon emission trading scheme promotes enterprise green technology innovation mainly through state-owned enterprise.
In order to protect local interests, there are some local officials who cover up stealthily emissions and leaks of high pollution companies, which results in the failure of the normal implementation of emission rights. Therefore, this paper argues that a high level of environmental regulation is required for the carbon emission trading scheme to fully play its role. The greater the intensity of environmental regulation faced by enterprises, the lower the chance of choosing illegal environmental behaviors. In this paper, the 30 provinces are divided into two groups: strong and weak pollution control intensity based on the difference in pollution control level. The division standard is the completed investment in industrial pollution control per unit of GDP. Provinces above the average are areas with strong pollution control intensity, while provinces below the average are areas with weak pollution control intensity. They are regressed respectively and the results are shown in columns (3) and (4) of Table 6. The P-value of SUEST test for the regression coefficient of treat i × time t is significant under the level of 1%. It shows that there are significant differences in the carbon emission trading scheme between strong pollution control and weak pollution control areas The results illustrates that carbon emission trading scheme promotes enterprise green technology innovation and have a positive effect on both strong pollution control and weak pollution control areas, and the effect is larger in strong pollution control areas.
The implementation of the carbon emission trading scheme leads to an increase in the production cost of enterprises, which is solved by cost transfer or self-digestion. The difficulty of cost transfer depends on the price elasticity of demand and economic growth. The short-term demand of high pollution industries is rigid and the rising costs are passed on to consumers. However, consumers will change their consumption behavior and choose cleaner products in the long run. Thus the carbon emission trading scheme would cause the result that enterprises have to bear the increased costs by themselves and would have a negative impact on improving productivity and competitiveness, which forces enterprises to carry out green technology innovations. In order to examine the industries differences, this paper divides the samples.  Table 6. The P-value of SUEST test for the regression coefficient of treat i × time t is 0.0130, which is significant under the level of 5%. The results show that there are significant differences in the carbon emission trading scheme between high pollution industry enterprises and low pollution industry enterprises. The coefficient of treat i × time t is significant and positive at the level of 1% in columns (5) and (6). But the coefficient in the high pollution industry is higher than in the low pollution industry, which indicates that the green technology innovation effect of enterprises in the high pollution industry is higher than that in the low pollution industry.

Moderating effect analysis
Another important area of enterprise green technology innovation research that needs to be more discussed is the moderating effect between various influencing factors (Liao et al. 2018). In order to examine the moderating effect of debt financing with environmental regulation on enterprise green technology innovation, this paper constructs model (3) and focuses on the regression coefficient of the interaction term treat i × time t × DebtFin.
Model (3) introduces the new variable debt financing on the basis of model (1). Referring to the practice of Wang and Chu (2019), this paper selects the asset-liability ratio to measure debt financing, which is the ratio of liabilities to total assets. As shown in Table 7, the regression coefficient of treat i × time t × DebtFin is significant and negative at the level of 1%, which indicates that the increase of debt financing will weaken the promotion effect of carbon emission trading scheme on enterprise green technology innovation. Hypothesis 2-1 is thus verified. At different quantile levels of debt financing, the role of treat i × time t in promoting enterprise green technology innovation continues to decline with the increase of debt financing. It means that the smaller the debt financing is for enterprises, the greater the role of carbon emission trading scheme is in promoting green technology innovation. Therefore, changing the financing structure of enterprises and reducing their reliance on debt financing can effectively ensure the positive effect of carbon emission trading scheme.

Conclusions and policy implications
This paper takes the 2013 carbon emission trading market pilot policy as a quasi-natural experiment and studies its impacts by establishing a DID model. The research finds that (1) The carbon trading scheme has a significant role in promoting enterprise green technology innovation, mainly through green practical patents and alternative energy patents, and the results are still significant after a series of robustness tests. (2) In terms of heterogeneity analysis, this paper finds that the carbon trading scheme has a more significant role in promoting green technology innovation in state-owned enterprises, enterprises belonging to areas with strong pollution control intensity (3) and high pollution industry. (3) Deep research finds that the increase of debt financing will weaken the promotion effect of carbon emission trading scheme on enterprise green technology innovation. This paper highlights specific recommendations in order to ensure the effectiveness of the implementation of carbon emission trading scheme and improve enterprise green technology innovation.
• Establish a comprehensive combination system of environmental regulation with market incentives environmental regulations as the dominance and the others as supplement. The carbon trading scheme is of great significance to the realization of China's "dual carbon" goal and the construction on the emerging development concept of coordinated development of environment and economy. However, China carbon emissions trading market started late and is still developing in comparison to the degree of marketization in other developed countries. So it requires imperative and voluntary environmental regulation in conjunction with market incentive type. • Reduce dependence on external financing and promote sustainable enterprise green technology innovation. In terms of financing, enterprises can reduce their reliance on debt financing and use more internal financing to ensure the sustainability of R&D investment. At the same time, the government should vigorously develop green finance to help enterprises gain stable and sufficient R&D funds. According to the heterogeneity of enterprise, they can formulate differentiated environmental policies and transform environmental regulation into advantageous resources to enhance the initiative of enterprise green technology innovation. • According to the heterogeneity of enterprises, Chinese government can formulate differentiated carbon emission trading system scheme to achieve precise positioning of environmental regulation policies. For example: Chinese non-state-owned enterprises contribute more than 60% of GDP, more than 70% of state tax revenue, and more than 90% of new jobs. Compared with stateowned enterprises with strong financial strength, nonstate-owned enterprises have greater financing constraints. In the face of strict environmental regulations, they adopt more strategic corporate behaviors and are powerless to green technology innovation. Therefore, local governments should fully consider the heterogeneity of enterprises when implementing the carbon emission trading scheme and propose a carbon trading scheme in a targeted manner according to the different characteristics of enterprises. Meanwhile, governments actively encourage and guide different social entities to carry out green technology innovation, and give full play to the positive effect of the carbon emission trading scheme on green technology innovation of regulatory entities with different characteristics, rather than a "onesize-fits-all" approach. • Strengthen environmental law enforcement and improve the policy supervision. To a large extent, whether the carbon emission trading scheme is effective depends on supervision and punishment system. After the implementation of the carbon trading scheme, the government should strictly enforce the law to ensure the effectiveness of policy. Meanwhile, the government continuously improve the supervision and punishment system, and increase violation cost to force enterprises to embark on enterprises green technology innovation.