Influence of China's Carbon Emissions Trading Scheme on Green Innovation of Enterprises.

China's economic growth has entered "new normal," and the task of reducing carbon emissions has become more onerous. Hence, this study aimed to explore whether China's carbon emissions trading pilot policy stimulated corporate green innovation capabilities. The data pertained to the green patent data of the listed companies in Shanghai and Shenzhen stock exchanges during 2008-2018. Using a difference-in-difference-in-differences (DDD) method, the study took advantage of the variations across regions, across enterprises, and across years and obtained several novel findings. First, the pilot carbon emissions trading policy significantly stimulated the green innovation capabilities of emission control companies in the pilot areas compared with enterprises in nonpilot areas and the nonemission control list. Second, the effect of the policy on the improvement in corporate green innovation capabilities might be driven by the improvement in corporate input factor allocation efficiency and the additional benefits that could be obtained from the carbon trading market. Third, the positive effect of the policy on the green innovation capabilities of state-owned enterprises was more significant. Therefore, the establishment and promotion of a unified national carbon emissions trading market and supporting mechanisms should be accelerated to achieve the balance of stable economic growth and carbon emission task.

1. Introduction

Different parts of China were attacked by unexpected floods in the summer of 2020, which restarted the discussion on climate change. The intensifying problem of extreme weather not only hampers social operation but also goes contrary to the concept of sustainable development. Excessive greenhouse gases (typically carbon dioxide) are the main precipitating factor for extreme weather. How to reduce greenhouse gas emissions has become a widespread concern [1, 2]. The Chinese government once relied on laws and regulations on energy conservation and emission reduction to force enterprises to reduce emissions [3]. Under the circumstance of increasingly strict environmental regulations, green innovation is the inevitable choice for enterprises to win competitive advantage and gain market position. In practice, these laws and regulations only set macroscopic emission reduction goals without specifying enforceable emission reduction requirements for microcosmic entities. Enterprises prefer paying fines for violating the environmental protection laws and regulations than increasing green technology investment, which is a high-risk investment due to its enormous input and long investment cycle. To maximize the profits and business scale and minimize the loss of benefits, enterprises generally choose to pay fines but continue with the original production model and output featured by high energy consumption. Hence, few enterprises can achieve emission reduction goals. In some situations, the government imposes one strict uniform emissions reduction target for all enterprises. However, enterprises can hardly change the production model in the short term and have to deal with a sudden rise in the production cost. In that case, the cash flow of enterprises may be broken, resulting in a decreased production output or even a complete halt. Under a mandatory emission reduction target, enterprises have to sacrifice their self-interests.

At present, climate change is already considered a non-negligible threat. As China is facing growing pressure to reduce emissions, the national economy undergoes a transition from high-speed to high-quality development. China's economic development has entered a new normal. Command-and-control environmental regulation has proven unsatisfactory in the context of China's transition from high-speed growth to high-quality development. How to safeguard enterprises' benefits while achieving the emission reduction targets has become an urgent issue. During the 13th Five-Year Plan period, the Chinese government gained from the experiences of the European Union and Japan in emission reduction and began an exploration using the market-based approach, that is, pollutant emissions trading scheme. Enterprises are encouraged to step up green technology investment via the carbon emissions trading scheme to achieve the goals of a low-carbon economy and sustainable development. In 2012, the Chinese government issued the Notice on Carbon Emissions Trading Work. In 2020, the promotion of the carbon emissions trading scheme across the country began. During the pilot period, we were concerned about whether the carbon emissions trading scheme had fully motivated enterprises in the pilot regions to strengthen green technology investment as part of the efforts to achieve the emission reduction goals.

In this study, a quasi-natural experiment was conducted to implement the carbon emissions trading scheme pilot project in 2014 (the Chinese government issued the pilot policies in 2012; however, due to the differences in the launch time in different regions, the latest launch time of an individual pilot was treated as the overall implementation time of all pilots). The green innovation capability of enterprises was measured by the number of applied green technology patents. We analyzed the influence of the carbon emissions trading scheme on the green innovation capability of the pilot enterprises. First, a difference-in-difference-in-differences (DDD) model was constructed using the difference method. It was found that, compared with the nonpilot regions, the implementation of the pilot policies not only dramatically improved the green innovation capability of the pilot enterprises but also offered an incentive for all enterprises in the pilot regions to engage in green technology investment. We further analyzed the pathway by which the policies influenced the green innovation capability of enterprises. The promoting effect of policies was mainly derived from improving enterprises' efficiency in input factor allocation and the prospect of gaining extra benefits from carbon emissions trading. Besides, the heterogeneity test showed that the pollutant emissions trading scheme was a more marked positive incentive for green technology investment in state-owned enterprises and large-scale enterprises.

2. Literature Review

Neoclassical economists believe, from the perspective of production cost, that environmental regulation policies push up the production cost of enterprises and impair competitiveness while promoting environmental protection. Thus, environmental regulation hinders economic growth [4-6]. Grossman and Krueger [7] later contradicted this belief by suggesting industrial structure changes and technological progress as other influencing factors for environmental changes apart from economic growth. Grossman's ideas opened up a new perspective for studying the influence of the environment on the economy. Porter and Van-der-Linde [8] incorporated technological progress into the analytical framework of environmental protection versus economic growth. It was hypothesized that appropriate environmental regulation boosted the technological innovation of enterprises or the adoption of innovative technologies. The profits made by enterprises on the market would exceed the cost engendered by environmental regulation, which was conducive to economic growth in the long term. As shown earlier, environmental protection does not necessarily stand in stark opposition to economic growth. Proper environmental regulation can force or motivate enterprises to increase the investment in production technologies and products with high energy efficiency and low pollution. This may serve as a pathway to elevate the green innovation capability and competitiveness of enterprises in the market and an ideal method to achieve business benefits and emission reduction goals. Hence, the question is, “what kind of environmental regulation is appropriate and can motivate enterprises in green technology innovation?” Many scholars conducted empirical studies on this topic, but the conclusions diverged [9-13]. Johnstone et al. [14] took the renewable energy policy as an example. The patent data from 25 countries between 1978 and 2003 were used to examine the influence of environmental policy on technological innovation. They found that public policies played a crucial role in a patent application. Some scholars performed an empirical test after dividing environmental regulation into different intensity intervals for quantifying purposes. They found that environmental regulation increased the total factor productivity of an enterprise only when the intensity of environmental regulation fell within a certain interval [15, 16]. Other researchers performed studies based on the provincial panel data of country. They suggested that we should be cautious with the choice of the environmental regulation tool to promote green technology innovation via environmental regulation [17-19].

The Porter hypothesis does not specify which type of environmental regulation can propel enterprises' efforts in green innovation. Nevertheless, this hypothesis believes in the necessity of the government (policy-makers) to fully utilize the power of the market. The government shall formulate mechanisms that accord with the market principles to guide enterprises in the pursuit of their self-interests and encourage their compliance with the environmental policies. Market incentive environmental regulation policies emerge at a time when China's command-control environmental regulation with governmental dominance proves ineffective. The pollutant emissions trading policy is one form of the market incentive environmental regulation, which enjoys two major advantages over the common-control one. First, the government only needs to set down the total amount of pollutant emissions. The pollutant emissions rights are commercialized and transformed into commodities. The emission reduction target is realized by utilizing the market pricing mechanism and through trading. Second, pollutant emissions trading offers a chance for enterprises to reap potential benefits, which encourages emission reduction. That is, eager to gain potential profits, enterprises are motivated to step up green technology investment and develop the capacity and technology for manufacturing environmental-friendly products (i.e., elevating the green innovation capability), which finally leads to emission reduction [20]. It is important to know whether the pollutant emissions trading scheme can be the right choice for boosting the technological advances of enterprises. China's emissions trading scheme starts from the SO2 emissions trading scheme. Many Chinese scholars chose the SO2 emissions trading scheme and studied its influence on the innovation capacity of enterprises [21, 22]. Qi et al. [23] constructed a DDD model based on the green patent data of China's listed companies from 1990 to 2010. They discussed whether launching the SO2 emissions trading scheme pilot project promoted the green innovation of enterprises. They reported a positive influence exerted by this policy on enterprises' green innovation. Ren et al. [24] used the SO2 emission data of listed enterprises from 2004 to 2015 to examine the influence of the SO2 emissions trading scheme pilot project on the total factor productivity of enterprises in 2007. They found that the SO2 emissions trading scheme pilot project improved the total factor productivity of enterprises in the pilot area by promoting technological innovation and resource allocation efficiency. However, a lag effect was observed in this promoting effect. Carbon emissions trading right is one type of pollutant emissions right. China did not establish the carbon emissions trading policy until recent years. At present, the studies on the relationship between the carbon emissions trading right and innovation are primarily restricted to macroscopic industries and the provincial level [25-28]. Few studies are directed toward individual enterprises at the microscopic level [29]. Shi et al. [30] were interested in the emission reduction effect and the working mechanism of the carbon emissions trading scheme. They constructed a DDD model for an empirical test among the listed enterprises from 2012 to 2015. They suggested that the carbon emissions trading scheme pilot project only motivated enterprises to reduce the production output but not increase the technological innovation investment to reduce emissions. Huang and Yang [31] conducted an empirical study, which showed that the emissions trading scheme dramatically increased the investment in research and development for large-scale enterprises in the pilot industries of the pilot regions.

In summary, only a few studies have been carried out in China regarding the relationship between the carbon emissions trading scheme pilot project and enterprises' innovation. Most of the existing literature chose total factor productivity or investment in research and development to measure enterprises' innovation. This study focused on the influence of environmental policies on the green innovation activities of enterprises. Using the aforementioned indicator as a surrogate indicator for green innovation might bring some biases. Instead, the input or output related to enterprises' development and the adoption of the emissions reduction technology may be appropriate surrogate indicators. Furthermore, an in-depth analysis of the entities truly influenced by the carbon emissions trading scheme is lacking. Many studies discussed whether the carbon emissions trading scheme promoted the green innovation activities of all enterprises in the pilot area. However, enterprises involved in the carbon emissions trading are neglected. Therefore, the empirical analyses yielded less reliable conclusions for informing the assessment of the promoting effect of the carbon emissions trading policy on enterprises' green innovation.

3. Research Method

It is assumed that the products are highly homogeneous in the market where an enterprise is involved. The product price, determined by market supply and demand, is p1. The enterprise's production cost and carbon emissions are determined by the input factors and the production technology. The carbon price on the trading market is p2.

The initial optimal production function of the enterprise is Q1(k1, l1, t1). Thus, the cost function of the enterprise is C1(k1, l1, t1)=VC1(k1, l1, t1)+FC1, the carbon emission is E1(k1, l1, t1), and the profit is R1=p1Q1 × (k1, l1, t1) − C1(k1, l1, t1), where k is the capital input; l is the labor input; t is the technical level; VC is the variable cost function; and FC is the fixed cost. Suppose that (∂E/∂k) > 0, (∂E/∂l) > 0, (∂E/∂t) < 0, (∂Q/∂k) > 0, (∂Q/∂l) > 0, (∂Q/∂t) > 0, (∂C/∂k) > 0, (∂C/∂l) > 0, (∂C/∂t) < 0. With the plants and production equipment fixed, (∂2Q/∂k2) < 0, (∂2Q/∂l2) < 0, (∂2C/∂k2) > 0, (∂2C/∂l2) > 0.

The pollutant emissions trading scheme is imposed to restrict the enterprise's carbon emission to E2(E2 < E1). Under this situation, an enterprise has two choices: the first is to continue the use of the existing production technology t1 and maintain other conditions, such as product price, unchanged. Thus, by reducing the production output to Q2, the enterprise can reduce the carbon emission and gain the profit R2=p1 × Q2(k2, l2, t1) − C2(k2, l2, t1) (Figure 1). Or the enterprise purchases the carbon emission allowance from the market (E1 − E2), which adds to the enterprise spending. The corresponding profit is R2=R1 − p2 × (E1 − E2). Although this choice allows for the achievement of the allocated carbon emission goal, the enterprise suffers from an economic loss R1 − R2. The second choice is to increase the investment in research and development and invent environmental-friendly and green production technologies. The enterprise can optimize the input factor allocation ratio to improve production efficiency, achieve the emissions reduction target, and maintain (or even increase) the current production output. Before the advanced technology comes out, increasing the investment in research and development is equivalent to increasing the fixed production cost (FC2). As the input factor allocation ratio remains unchanged, the variable cost function is unchanged, too, leading to an increase in the total cost for the same production output (C3). Besides, due to carbon emission constraint E2, the enterprise has to reduce the production output to Q2, which is accompanied by a reduction in the profit (R3=p1 × Q2(k2, l2, t1) − C3(k2, l2, t1)) (Figure 1). The total profit of the enterprise remains lower than the original profit if the enterprise chooses the first strategy. Regarding the second choice, although the enterprise profit may be lowered in the short term by increasing the investment in new technology, the enterprise may reap the advanced technology and more profit in the long term. Thus, the second choice is better. We made the following hypotheses:

Figure 1

One choice of the enterprise.

Hypothesis 1 .

The carbon emissions trading scheme offers an incentive for enterprises to engage in green innovation activities.

When new technology t2 is mature and put into use, it is assumed that the product price in the market, the input factor price, and other conditions remain unchanged. As the enterprise's allocation efficiency of the input factors is improved, the output and costs of the enterprise are altered: the input of capital elements and labor elements needed for producing a unit product decreases, while the production efficiency increases. The energy consumption also decreases. The production function becomes Q′(k′, l′, t2), and the cost function is C′(k′, l′, t2)=VC′(k′, l′, t2)+FC2 (Figure 2). That is, under the same amount of input, Q′ > Q; to produce the same amount of products, k′ Q2(k2, l2, t1). The enterprise reaps profits by increasing the production output and reducing the cost after the allocative efficiency of the input factors has increased (R1′ > R3). Thus, the following hypothesis is made:

Figure 2

The other choice of the enterprise.

Hypothesis 2 .

The carbon emissions trading scheme offers an incentive for enterprises to engage in green innovation activities by improving the allocation efficiency of input factors.

For a given production output Q2′, the corresponding carbon emission is E2′. When p2 × E2′ > p1 × Q2′ − C2′, the profit made from selling the carbon emission allowance corresponding to a given production output is higher than the profit made from producing the same amount of products. At this time, while the enterprise maintains a normal production and does not exceed the allotted emission allowance, it may choose to sell the unneeded allowance on the market to gain profit. Regarding the benefits made from selling the products, the consumers are increasingly aware of the quality of the products in use and prefer environmental-friendly products as China undergoes the transition from high-speed to high-quality development [32]. Thus, the enterprise can gain a differential advantage during the competition with other products of the same class. Finally, the enterprise can expand the market or increase or gain more profits by increasing the price of the products. Thus, the following hypothesis is made:

Hypothesis 3 .

The carbon emissions trading scheme offers an incentive for enterprises to engage in green innovation activities by providing a chance for the enterprise to gain extra benefits from carbon emissions trading.

As shown by the aforementioned analysis, the carbon emissions trading scheme pilot project prompts enterprises to increase green technology investment through the following two pathways. The first is to impose a limit on enterprises' carbon emissions based on the original production output and carbon emissions of the enterprise under the dynamic changes in an external competitive environment. Without changing the original production mode, the enterprise should meet the emission reduction goals by continuously reducing the production output or purchasing emission allowance from the market. However, either choice goes contrary to the long-term business goals of the enterprise and the principle of profit maximization. The second is to encourage the enterprise to realize the transition to the green production mode by constructing a carbon emissions trading scheme. Enterprises willing to realize green innovation by increasing the green technology investment may be faced with the difficulty of transforming the production mode in a short time. To survive and maintain normal production, these enterprises can purchase the emission allowance from the market. The emissions trading scheme serves as a buffer during the production decarbonization process. For those that succeed in the low-carbon transition, the emissions trading scheme provides a pathway to sell the unneeded allowance to gain extra profit and maximize business benefits in the long term. In a word, increasing green technology investment and elevating one's green innovation capability are the best choices under the current policy context. Enterprises can hope to enhance the market competitiveness and expand their scale by engaging in green innovation activities.

4. Empirical Analysis

4.1. Data Source, Variable Selection, and Processing

4.1.1. Data Source and Processing

Data samples from China's companies listed on the Shenzhen and Shanghai stock exchanges from 2008 to 2018 were chosen. The data samples were preprocessed as follows. The data were collected only from the normally listed companies. Samples with serious missing data problems of core variables within the sample interval were removed. Explanatory variables were winsorized 1% in each tail to remove the outliers. The final total sample size was 23,739.

The patent data of enterprises came from the China National Intellectual Property Administration. Other data came from the China Stock Market & Accounting Research database.

4.1.2. Variable Selection and Descriptive Statistics

Launching the carbon emissions trading scheme pilot project on October 29, 2011, the China Development and Reform Commission released the notice of the carbon emissions trading scheme pilot project. In this notice, Beijing, Tianjin, Shanghai, Chongqing, Guangdong, Hubei, and Shenzhen were included as the pilot regions. Each pilot region determined its own time to launch the emissions trading scheme project and enterprises to participate in the trading. The core explanatory variable in this study was the interaction term of the three virtual variables. The enterprise included as the participating enterprise in the pilot region within the pilot period was assigned the value of 1 (treatment group). Otherwise, the enterprises were assigned the value of 0 (control group). We chose the enterprises included in the annual lists published by the pilot regions every year from 2014 to 2018.

Green innovation capability of enterprises. The green innovation capability of enterprises was measured by the number of the applied green technology patents. Patent data were retrieved and downloaded from the website of the China National Intellectual Property Administration by province/prefecture-level city/enterprise + year. Data cleaning and screening were performed according to the WIPO's (World Intellectual Property Organization) green patent classification index.

Control variables. The influence of the features of an individual enterprise itself on its green innovation activities was considered. To account for this, we treated the data related to the features of enterprises as the control variables. According to the knowledge production function, the larger the scale of an enterprise, the higher the success rate of green innovation. Enterprises also reap more benefits and revenues. As a result, they invest more heavily in research and development, and the success of such investment also increases. Besides, enterprise liabilities are lower, and enterprises are more willing to engage in high-risk innovation activities and are more likely to succeed. Listing age was another control variable. Innovation is an activity that requires lasting efforts. Enterprises with a longer listing age are more likely to engage in research and development to maintain their competitive edge on the market. Hence, such enterprises have a greater possibility to succeed in innovation. The nature of the enterprise was another control variable of our interest. State-owned enterprises are more inclined to obtain both social and economic benefits. Thus, state-owned enterprises are more likely to engage in the research and development of green products and succeed in it. The explanations of the selected variables are shown in Table 1.

Table 1

Explanations of the variables.

Name of the variable	Explanation
Whether within the pilot period	A value of 0 assigned to the enterprises included as the pilot enterprises from 2008 to 2013; a value of 1 assigned to those included as the pilot enterprises from 2014 to 2018
Whether being a pilot region or city	A value of 1 assigned to Shenzhen, Beijing, Shanghai, Guangdong, and Tianjin; otherwise, a value of 0 assigned
Whether being an enterprise included in the annual list	A value of 1 assigned to the enterprises included in the annual lists of the pilot regions and cities every year from 2014 to 2018; otherwise, the value of 0 assigned
Nature of the enterprise	A value of 1 assigned to the state-owned enterprises; otherwise, the value of 0 assigned
Listing age	Difference between the sample year and the listing age
Net profit-to-net worth ratio	Ratio of net profit to balance of total assets
Ratio of expenses to sales	Ratio of the selling expense to the revenue
Enterprises' scale	Logarithm of the total assets
Asset-to-liability ratio	Ratio of total liabilities to total assets
Investment level of the enterprise	(Cash paid to acquire fixed assets, intangible assets, and other long-term assets and cash recovered from disposing of fixed assets, intangible assets, and other long-term assets)/total assets at the end of the reporting period

4.2. Parallel Trend Test

For a difference-in-difference (DID) model to have explanatory power, the treatment and control groups must have the same temporal variation trend before policy intervention. In the present study, this requirement also applied to the DDD benchmarking model. Here, the event study method was adopted [33]. Using the total number of green patents as an explained variable, fixed-effects estimation was performed for the triple interaction term consisting of the year of pilot project implementation, pilot region, and pilot enterprises and the control variables. Figure 3 shows the coefficients of the triple interaction term. The coefficients were insignificant before the year 2014. This finding indicated that the emissions trading scheme pilot project dramatically spurred the green innovation activities of the pilot enterprises, promoting the elevation of the green research and development level. The treatment effect existed.

Figure 3

The result of parallel trend test.

4.3. Model Construction

As shown by the information collected from the carbon emissions trading center in the pilot regions, more than 90% of the regions successfully fulfilled the emission reduction goals. Apart from the results of fulfillment, we were also concerned about how these enterprises achieved emission reduction. As stated by the aforementioned theories, green innovation propels enterprises to save energy and reduce emissions in the long term, optimizes input factor allocation, and increases potential benefits for enterprises. The present study examined the changes in the number of green patents enjoyed by the enterprises during the sample period to determine whether the green innovation capability of enterprises was elevated.

4.3.1. Influence of the ETS on the Green Innovation Capability: Provincial DID Model

We first considered the influence of carbon emissions trading on the green innovation activities of all enterprises in the pilot regions. Here, a DID model was established:where i, j, r, and t are the listed enterprise, industry, province, and time, respectively; ε is the random disturbance term; the explained variable Greenpatent is the number of green patents enjoyed by the listed enterprise i within the year t; and Treatr is the pilot region dummy variable. When the enterprise was in one of the seven pilot regions of the carbon emissions trading scheme pilot project, a value of 1 was assigned; otherwise, a value of 0 was assigned. Timet is the time dummy variable (the time when the pilot project is launched). The latest time of launching the pilot project among the seven pilot regions and cities was chosen as a threshold. The time period after it (2014 and after) was assigned the value of 1; otherwise, the value of 0 was assigned. The core explanatory variable was the interaction term between Treatr and Timet, which measured the influence of policy implementation on the green innovation activities of the enterprises in the pilot regions. The control variables Xit included the listing age, ratio of net profit to net worth, ratio of expenses to sales, nature of the enterprise, scale of the enterprise, and ratio of asset to liability. These variables representing the internal features of the enterprises were used to control for other factors that might influence the green innovation activities. Finally, we controlled for the time fixed effects, entity fixed effects, region fixed effects, and industry fixed effects.

4.3.2. Influence of the ETS on the Green Innovation Capability: DDD Model

In the practice of implementing the pilot project, enterprises with high carbon emissions were usually included in the lists and reviewed annually. Compared with nonincluded enterprises, those included in the lists were required to reach the emission reduction targets. To achieve this goal, these enterprises were particularly active on the carbon emissions trading market and were more willing to engage in green innovation activities. Observing whether a difference existed between the included and nonincluded enterprises in green innovation activities offered a pathway to verify the effectiveness of the carbon emissions trading scheme. Based on the DID model, we built the DDD benchmarking model by introducing the pilot enterprise dummy variable:where Control is the pilot enterprise dummy variable. If the enterprise was a pilot enterprise, the value of 1 was assigned; otherwise, the value of 0 was assigned. The core explanatory variable Timet × Treatr × Control is the interaction term between the pilot period, pilot region, and pilot enterprise. The coefficient estimates β1 also attracted our attention. The definitions of the remaining explanatory variables were the same as in formula (1).

4.4. Results of the DID Model

The regression results on the influence of the emissions trading scheme pilot project on the green innovation capability of enterprises are shown in Table 2. The first column shows the average influence of the pilot project on the number of green patents after controlling for the industry fixed effects, region fixed effects, and time fixed effects. The regression coefficient for the influence of the pilot project on the number of green patents was 1.5046 (significant at the 1% level). Thus, the pilot project significantly increased the number of green patents applied by the listed enterprises in the pilot regions. The second column includes the control variables based on the first column. The regression results basically remained unchanged, the regression coefficient being 1.5640 (significant at the 1% level). The third and the fourth columns include the time-invariant factors. The fixed-effects model was used. The regression coefficient decreased compared with the baseline regression model, although it was still positive and significant (at the 1% level). Both models indicated a significant positive influence. It was thus inferred that the regression results had certain robustness. The results showed that the emissions trading scheme pilot project increased, to a certain extent, the number of green patents applied by all enterprises in the pilot regions. This result testified the aforementioned hypothesis about the positive influence of the emissions trading scheme pilot project on the long-term business decisions of enterprises.

Table 2

Results of the DID model.

Variables	1	2	3	4
Timet × Treatr	1.5046∗∗∗	1.5640∗∗∗	0.4345∗∗	0.4248∗∗
	(0.5805)	(0.5795)	(0.1799)	(0.1815)
Whether or not the pilot period	1.6543∗∗∗	−0.5896	1.0739∗∗∗	0.0970
	(0.4278)	（0.4523)	（0.1276)	（0.7199)
Whether or not the pilot region	0.1666	−1.4655∗∗	——	——
	(0.5029)	(0.6243)
Constant	−0.9991	−54.2733∗∗∗	0.4170∗∗∗	−2.3055∗∗
	(0.6185)	(4.8920)	(0.1078)	(1.0440)
Control variables	No	Yes	No	Yes
Time fixed effect	Yes	Yes	Yes	Yes
Individuals fixed effect	No	No	Yes	Yes
Region fixed effect	Yes	Yes	No	No
Industry fixed effect	Yes	Yes	No	No
Observations	23 739	23 726	23 739	23 726
R-squared	0.2187	0.2408	0.0164	0.0168

Note. Robust clustering standard error is indicated in brackets; , , and indicate significant at the level of 10%, 5%, and 1%, respectively, which are the same in the following tables.

4.5. Results of the DDD Benchmarking Model

The results of the DDD model are shown in Table 3. First, the industry fixed effects, year fixed effects, and region fixed effects were introduced into the first column and the control variables into the second column. Then, the fixed-effects model was adopted in the third column to control the time fixed effects and entity fixed effects. The control variables were introduced into the fourth column. As shown by the results of baseline regression in the first to the second columns of Table 3, the coefficient of the triple interaction term Timet × Treatr × Control was positive and significant (at the 1% level). This indicated that the emissions trading scheme pilot project positively promoted the green innovation activities of the pilot enterprises in the pilot regions. Further, the fixed-effects regression in the third to the fourth columns showed that the coefficient of the triple interaction term Timet × Treatr × Control was still positive and significant (at the 1% level). Besides, this coefficient estimate was larger than that in the baseline coefficient. Hence, it was inferred that the pilot project reasonably promoted the green innovation activities of enterprises. It also testified to the hypothesis that enterprises were motivated to step up green technology and manufacture the energy-conserving and emission-reducing products to gain a long-term competitive edge and increase business benefits. The emissions trading scheme system was proposed to coordinate economic stability and environmental optimization by combining government and market regulations. Its effectiveness reflected whether the national top-level design was reasonable.

Table 3

Baseline results of the DDD model.

Variables	1	2	3	4
Timet × Treatr × Control	2.0530∗∗∗	2.1857∗∗∗	4.2393∗∗∗	4.2570∗∗∗
	(0.3531)	(0.3475)	(1.5480)	(1.5467)
Timet × Treatr	−0.0313	−0.0456	0.0367	1.1305
	(0.1311)	(0.1291)	(0.1010)	(0.0236)
Timet × Control	4.4433∗∗∗	3.7719∗∗∗	——	——
	(0.2623)	(0.2591)
Constant	−0.4802	−17.0066∗∗∗	0.4426	−2.7643∗∗∗
	(0.4429)	(0.7353)	(0.0981)	(1.0011)
Control variables	No	Yes	No	Yes
Time fixed effect	Yes	Yes	Yes	Yes
Individuals fixed effect	No	No	Yes	Yes
Region fixed effect	Yes	Yes	No	No
Industry fixed effect	Yes	Yes	No	No
Observations	23 739	23 726	23 739	23 726
R-squared	0.1038	0.1333	0.0326	0.0331

After controlling for the entity fixed effects and time fixed effects, the scale of the enterprise and listing age were found to have a promoting effect on the number of green patents applied. Each additional year of being listed increased the number of green patents by 0.087. Each additional 1% expansion of the scale increased the number of green patents by 0.136%. The coefficients of these two control variables were significant at the 1% level (details not shown due to limited space). The remaining variables were insignificant at the 10% level.

5. Discussion

5.1. Placebo Test

We performed a placebo test for time and region to ensure the robustness of the empirical results. First, a conceptual year of policy implementation was introduced for the placebo test (the Chinese government launched the emissions trading scheme pilot project in 2012; however, it was not until 2014 that the pilots were actually established in different pilot regions. To eliminate the possibility that some enterprises launched the pilots soon after the Chinese government published its decision, we set the conceptual year of policy implementation to 2011). It was then assumed that the elevation of the green innovation capability of enterprises was not due to the implementation of the pilot project but due to enterprises' own initiative and the influence of external factors. In that case, the elevation of the green innovation capability of enterprises had nothing to do with the pilot project. We ran the placebo test by assuming this conceptual year of project implementation. The results are shown in Table 4.

Table 4

Result of placebo test.

Variables	1	2
Timet × Treatr × Control	2.5221	3.4597
	(1.6475)	(2.3564)
Timet × Treatr	0.0966	−0.0393
	(0.1098)	(0.1431)
Timet × Control	2.9118∗∗∗	——
	(0.6965)
Constant	−17.0181∗∗∗	−2.5730∗∗
	(2.7491)	(1.1980)
Control variables	Yes	Yes
Time fixed effect	Yes	Yes
Individuals fixed effect	No	Yes
Region fixed effect	Yes	No
Industry fixed effect	Yes	No
Observations	23 726	23 726
R-squared	0.138	0.0231

The sample interval was from 2010 to 2018 (Time was 0 from 2008 to 2010; and Time was 1 from 2011 to 2018). The first column represents baseline regression with the industry fixed effects, region fixed effects, and time fixed effects introduced; the second column represents the fixed-effects model controlling for the entity fixed effects and time fixed effects. Control variables were introduced into both models. The results showed that, in either the baseline regression model or the fixed-effects model, the coefficient of the interaction term between the conceptual year of pilot project implementation, pilot region, and pilot enterprise was insignificant. Except for the year 2014, no significant results were obtained if any other year was taken as the year of project implementation. Therefore, the empirical results were not due to other random factors.

The pilot region selection may be random; however, some unobservable factors may influence the central government's selection of the pilot regions. Addressing this effect, we followed the method used by some other scholars [34, 35]. A random simulation was run 1000 times for randomization of the pilot region selection. The mean coefficient estimate of the interaction term between the pilot period and the pilot region was −0.00095 (using the fixed-effects model) versus –0.00279 (using the ordinary least squares regression). Both were close to zero (Figures 4 and 5). The results indicated that the unobservable region factor hardly had an impact on the estimation. Thus, the benchmarking model was robust.

Figure 4

Coefficient of FE running 1000 times.

Figure 5

Coefficient of OLS running 1000 times.

5.2. Instrumental Variable Test

During the quasi-natural experiment of the emissions trading scheme pilot project, the intervention of other policy factors might occur during the sampling period. We adopted the method proposed by Hu and Ding [36], in which the annual average temperature of each region in China was considered an instrumental variable for determining whether to include the region as the pilot region. The data of the annual average temperature came from the China Statistical Yearbook. One direct outcome of excess carbon dioxide was the greenhouse effect, which further led to a yearly rise in temperature. Thus, regions with a higher average temperature were more likely to be chosen as pilot regions for the emissions trading scheme. Besides, no potential connections were discovered between the average temperature and enterprises' application for green patents. Thus, the average temperature was an exogenous variable.

The results of the instrumental variable test are shown in Table 5. In the first stage, the coefficient estimates of Temp × Time was positive and significant (at the 1% level), indicating that the average temperature of the region had a positive impact on the launch of the pilot project. In the second stage, the coefficient of the triple interaction term Timet × Treatr × Control was positive and significant (at the 1% level), indicating that the launch of the pilot project significantly increased the number of green patents applied. The tests demonstrated that the empirical results of the benchmarking model were not due to sample self-selection.

Table 5

Results of instrumental variable test.

Variables	The first stage		The second stage
	1	2	1	2
	Timet × Treatr × Control	Timet × Treatr × Control	The patents	The patents
Temp × Time	0.0031∗∗∗	0.0034∗∗∗
	(0.0004)	(0.0003)
Timet × Treatr × Control			7.9862∗∗∗	7.1356∗∗∗
			(2.8196)	(2.0442)
Constant	−0.1701∗∗∗	0.0757∗∗∗	−10.6924∗∗∗	−3.0606∗∗∗
	(0.0182)	(0.0293)	(0.6748)	(0.6356)
Control variables	Yes	Yes	Yes	Yes
Time fixed effect	Yes	Yes	Yes	Yes
Individuals fixed effect	No	Yes	No	Yes
Region fixed effect	Yes	No	Yes	No
Industry fixed effect	Yes	No	Yes	No
Observations	23726	23726	23726	23726
R-squared	0.0868	0.041	0.0979	0.0163

5.3. Heterogeneity Test

In some previous studies on the influence of the emissions trading scheme on the innovation capability of enterprises with different ownership, it was believed that the state-owned enterprises enjoyed more favorable resources and fiscal policies from the government. In contrast, non-state-owned enterprises assumed sole responsibility for their profits or losses. Thus, these enterprises were more strongly driven by the pressure from increased costs and the pursuit of economic benefits and more positively motivated by the emissions trading scheme. As a result, the non-state-owned enterprises in the pilot regions finally developed higher green innovation capability [23, 24]. However, the carbon emissions trading scheme is unlike the emissions trading schemes for other pollutants in China mainly in terms of the implementation time (from the year 2014). After the 18th National Congress of the Communist Party of China in 2012, the Party Central Committee with General Secretary Xi Jinping at the core has been strongly determined on eco-environmental protection. It is after the year 2012 that China has been unprecedentedly devoted to environmental regulations in nearly every aspect: from concept to action, from top-level design to implementation, and from legislation to enforcement. The policies are now obviously skewed toward environmental protection, forcing enterprises to be aware of the environmental impact during the production process. The non-state-owned enterprises are in pursuit of profits, whereas the state-owned enterprises have more intimate connections with the government and attach greater importance to social benefits. The state-owned enterprises are often in the vanguard of environmental reform. They are more concerned about the external of their production and business activities. The present study investigated the influence of the carbon emissions trading scheme on the green innovation capability of enterprises with different ownership.

As shown by the first column in Table 6, the regression coefficient of the interaction term involving the state ownership was significant at the 5% level. However, the coefficient of the interaction term involving non-state ownership was not significant. This result indicated that the carbon emissions trading scheme promoted the green innovation activities of the state-owned enterprises in the entire pilot region at the regional level. According to the second column, the coefficients of the interaction terms involving either the state or non-state ownership were positive and significant. The coefficient estimates of the interaction term involving the state ownership was far higher than that involving the non-state ownership. Moreover, the research and development in green technology was more considerably elevated in state-owned enterprises. The findings were in accordance with China's strengthening efforts in constructing ecological civilization.

Table 6

Results of enterprises' nature.

Variables	1		Variables	2
	State	Non-state		State	Non-state
Timet × Treatr × Control			Timet × Treatr × Control	6.8461∗∗	2.0375∗∗∗
				2.9480	0.3027
Timet × Treatr	0.7812∗∗	0.1533	Timet × Treatr	0.0433	−0.0112
	(0.3731)	(0.1589)		(0.1623)	(0.1405)
Whether or not the pilot period	−2.8193	1.0688	Timet × Control
	(1.9272)	(0.6561)
Constant	−3.4630	−3.9597	Constant	−3.1556∗∗∗	−4.0391∗∗
	(2.5055)	(1.3161)		(1.9662)	(1.7047)
Control variables	Yes	Yes	Control variables	Yes	Yes
Time fixed effect	Yes	Yes	Time fixed effect	Yes	Yes
Individuals fixed effect	Yes	Yes	Individuals fixed effect	Yes	Yes
Observations	9 225	14 501	Observations	9 225	14 501
R-squared	0.0195	0.019	R-squared	0.052	0.0236

5.4. Mechanism Analysis

In the face of carbon constraints imposed by the emissions trading scheme pilot project, expanding the green innovation activities is not the only option for the pilot enterprises. However, the empirical results showed that the number of green patents applied by pilot enterprises increased dramatically during the sample period. The question was what caused the pilot enterprises to step up the green innovation input under pressure from the emissions trading scheme pilot project.

First, enterprises pursue profits. As a type of environmental regulation, the carbon emissions trading scheme has a certain negative impact on enterprise production and costs. Blindly reducing the production output goes contrary to the long-term benefits and profits of enterprises. To find a sustainable solution, enterprises consider continuously optimizing the allocation of production factor inputs to increase productivity as the carbon emission reduction requirement is being tightened. They may choose green, advanced production equipment. Continuously increasing the research and development activities offers a pathway to achieve this goal. By referring to Ren et al. [24]; we further studied the influence of the carbon emissions trading scheme pilot project on the efficiency of input factor allocation. The following model was constructed:where investment is the enterprise's investment level, and roa is the net profit to net worth ratio. The definitions of other variables are the same as in formula (1).

Second, under the emissions trading scheme, enterprises can sell the unneeded carbon emission allowance on the market to increase their extra profits and improve the cash flow. The extra profits and improvement of the cash flow, in turn, promote enterprises' research and development in green technology [37]. By referring to Liu and Zhang, we examined the influence of the emissions trading scheme pilot project on enterprises' cash flow and return on assets.

Taken together, the carbon emissions trading scheme pilot project motivated enterprises to engage in green innovation activities. Enterprises responded to this scheme by optimizing the input factor allocation, which further improved enterprises' cash flow and the expected return (Table 7).

Table 7

Results of mechanism analysis.

Variables	Cash flow	Return on assets	Factor allocation
Timet × Treatr × Control	0.1125	0.0096∗∗∗	−0.0068
	(0.1930)	(0.0033)	(0.0045)
Timet × Treatr	−0.0293	−0.0015
	(0.1191)	(0.0053)
Timet × Treatr × Control × roa			0.0968∗∗∗
			(0.0184)
roa			−0.0257
			(0.0174)
Constant	5.2714∗∗∗	0.0055	−0.0049
	(1.1873)	(0.1713)	(0.0307)
Control variables	Yes	Yes	Yes
Time fixed effect	Yes	Yes	Yes
Individuals fixed effect	Yes	Yes	Yes
Observations	7 356	23 739	23 739
R-squared	0.0371	0.0876	0.0952

As shown in Table 7, the regression results with cash flow as the explained variable show that the positive coefficient of interaction term Timet × Treatr × Control is not significant, indicating that the cash flow of included enterprises has not increased significantly after the implementation of the carbon trading pilot scheme. The regression results with the return on assets as the explained variable show that the coefficient of interaction term Timet × Treatr × Control is significantly positive (at the 1% level), indicating that the implementation of carbon emission trading scheme has improved the net return on assets of included enterprises. The regression results of the model with enterprise's investment as the explained variable show that the coefficient of interaction term Timet × Treatr × Control × roa is significantly positive (at the 1% level), indicating that the implementation of carbon emission trading scheme promotes enterprises to optimize the allocation of input factors. Based on the sample regression results of this paper, it is found that the carbon trading pilot scheme can encourage enterprises to carry out green technology innovation by improving the allocation efficiency of enterprises' input factors and setting up a market to allow enterprises to obtain additional benefits through carbon trading.

6. Conclusive Remarks

In the first 15 years of the 21st century, China developed at a stunningly high speed at the expense of the environment. China's economic development has entered a new normal. The Chinese government has come up with the slogan “green water and green mountains are golden mountains and silver mountains,” which marks the beginning of the road toward high-quality, sustainable development. Which type of policies can guarantee a win-win situation where the goals of national economic stability and emissions reduction can be met simultaneously? The Chinese government has launched the carbon emissions trading scheme pilot project by learning from the successful experience of other countries and by resorting to the market-oriented approach. This attempt represents China's initial exploration in emissions reduction. In this context, we performed an empirical test of the influence of the carbon emissions trading scheme pilot project on the green innovation capability of enterprises. The reliability of the research findings was proven by the placebo test, robust test, and heterogeneity test. The following conclusions were drawn. First, the carbon emissions trading scheme pilot project significantly elevated the green innovation level of the pilot enterprises in the pilot regions. Second, the carbon emissions trading scheme pilot project forced enterprises to optimize factor input allocation. Enterprises could reap extra benefits from the emission trading market, which, in turn, motivated enterprises to engage in green innovation activities. Third, in the face of carbon constraints imposed by the emissions trading scheme pilot project, the pilot enterprises of a larger scale were more willing to and more capable of increasing the research and development in green technology compared with those of a smaller scale. Thus, the green innovation capability of these enterprises was elevated more considerably. Fourth, under the guidance of the national macroscopic policies, state-owned enterprises were more inclined to balance between economic and social benefits due to their ownership. Consequently, state-owned enterprises invested more heavily in the research and development of green technologies. Their green innovation output also increased more dramatically.