Innovation spillover effect of the pilot carbon emission trading policy in China

Abstract

In today's society, environmental protection and sustainable development have become the focus of global attention. Carbon emission trading, as an effective means to reduce greenhouse gas emissions, has been adopted by many countries and regions. In China, the launch of pilot policies for carbon emissions trading is of great significance in promoting economic transformation, promoting industrial upgrading and protecting the environment. Therefore, it is of great value to study the innovation effect of the pilot carbon trading policy and its transmission path for evaluating the effect of the policy, optimizing the policy design and building the carbon market in the future. Based on the production network theory, this paper constructs a difference in difference model, and uses empirical analysis to evaluate the direct innovation effect and innovation network spillover effect of the pilot carbon emission trading policies. In the empirical analysis, a variety of data sources are used, including public data from the National Bureau of Statistics, annual reports of enterprises, and industry associations, to ensure the reliability and validity of the data. At the same time, SPSS, Stata and other statistical software are applied to process and analyze the data in order to better understand and interpret the research results. The results show that the pilot carbon emission trading policy has a positive effect on improving the innovation level of both regulated and non-regulated industries in the pilot areas. Specifically, this policy can promote the technological innovation and management innovation of the regulated industries, but also promote the technological progress and market development of the non-regulated industries. In addition, the policy will also have an innovation network spillover effect, that is, the innovation results of the policy will be transmitted to the surrounding regions and industries, thus promoting the innovation and development of the entire region and industry. Further, this paper also discusses the transmission path of the innovation spillover effect of pilot carbon emission trading policies. It is found that the policy has a negative impact on the innovation of the relative upstream industry, but significantly promotes the innovation of the downstream industry. This is because the upstream industry, after being impacted by policies, often takes measures to reduce costs and improve efficiency to deal with the challenges, which may have a certain negative impact on the downstream industry. However, downstream industries are more likely to benefit from the innovation spillover effects of policies because they are more closely linked to the carbon emission trading market. Finally, this paper also tests the mechanism of the pilot carbon emission trading policy. It is found that the innovation spillover caused by this production-end shock is mainly realized through the price transmission of finished products. In addition, the inclusion of industries with strong bargaining power and close links with downstream industries in the national pilot carbon emission trading policy is more conducive to the innovation spillover effect of the policy. This is because these industries have stronger market influence and bargaining power, and can better promote the construction and development of the carbon emission trading market.

1. Introduction

China's total carbon emissions rank first in the world, indicating that it bears important responsibilities in global climate change and environmental protection. As the world's largest developing country, China plays an increasingly important role in international affairs; therefore, its commitment to environmental protection and sustainable development has attracted attention world-wide. The Chinese government established a national energy and carbon emissions trading market as an important measure to combat climate change. Such a move can not only promote the efficient use of energy and reduce carbon emissions but also promote economic transformation and upgrading, improve energy efficiency, and promote green and low-carbon development [1]. This was one of the goals established by the Chinese government in the Government Work Report. The report to the 20th National Congress once again stressed the need to “jointly promote carbon reduction” and “green and low-carbon development,” which shows that the Chinese government will continue to step up efforts to promote environmental protection and sustainable development. This is not only to attach importance to China's national economy and people's livelihoods, but also to assume responsibility for global climate change and environmental protection. As a large country, China is obligated to contribute to global environmental protection. China can simultaneously achieve economic transformation and sustainable development through efforts. Global climate change and other environmental issues have become common challenges for humans. As a large country with 1.3 billion people, the success or failure of China's environmental protection and sustainable development efforts will have a profound impact worldwide [2]. Therefore, China must strengthen its cooperation with other countries to jointly deal with climate change and environmental issues. Simultaneously, China must promote the development of global environmental protection through its efforts and contribute to the future of humankind. In 2013, China launched carbon emission trading pilot programs in eight provinces and cities to limit the total carbon emissions of key industries [3]. Since the implementation of the policy, the scale of the carbon market has steadily increased, and its practical effect has been remarkable. According to the “China Carbon Emission Data Report” released by the Ministry of Ecology and Environment of China, by March 2022, China's carbon market had covered 2038 key emitters, representing a total of 440 million tons of carbon emissions, and the cumulative transaction amount was approximately 10.47 billion yuan.

Existing studies have confirmed the innovation incentive effect of environmental regulation, among which governance ability is the key to the existence of an innovation effect [4,5,6,7,8]. Acemoglu et al. (2019) proposed that, in the context of environmental regulation, the government should take the initiative to implement a series of market-incentive policies, such as financial support and research and development support, to help enterprises in energy conservation and emission reduction. Several studies focused on the causal relationship between environmental regulations and enterprise productivity. Considering the differences in regulatory policies in pilot areas and industries, the relationship between the two has not reached consistent conclusions [9,10,11,12]. Chen (2019) found that, compared with other environmental regulatory policies, tradable energy policies have a stronger promoting effect on innovation. Post the signing of the Kyoto Protocol, regional carbon emissions have significantly reduced, and its innovation-promoting effect has attracted widespread attention. Porter's hypothesis states that enterprises adjust their production and investment decisions according to changes in the market environment to obtain higher profits. This theory is widely cited in academia, and has been used to explain the impact of policies on firm innovation. In Porter's original book Competitive Strategy, he proposed important strategies: “cost-driven” and “differentiation-driven.” The cost-driven strategy refers to maximizing profits through the scale effect while pursuing low costs; the differentiation-driven strategy refers to the simultaneous pursuit of high value-added, high-quality products through differentiation to gain competitive advantage. The core idea of Porter's hypothesis states that enterprises choose their own strategies according to market demand and competition. If market demand is stable and competition is fierce, the firm will adopt a cost-driven strategy; if market demand is not stable and competition is not fierce, the enterprise will adopt a differentiation-driven strategy. An enterprise's business strategy also changes under the influence of the policy environment. For example, the government's implementation of environmental protection policies, such as a carbon emission tax, will increase the cost of enterprises, thus forcing them to adopt more environmentally friendly production methods to reduce costs. The impact of such policies on enterprises is indirect, but they can promote the innovation and development of enterprises. Academic circles, citing Porter's hypothesis, often use data from listed companies to demonstrate that policies improve the potential returns of enterprise innovation through carbon prices from the perspective of market-incentive environmental regulation. However, if this effect is extended to pilot areas, relevant empirical evidence shows that Porter's effect did not emerge during this period [13,14,15]. This shows that although the policy has a certain impact on the development of enterprises, it is not immutable but is affected by multiple factors, which require in-depth research and analysis. Therefore, whether policies can promote innovation in the regulatory industry must be further demonstrated through multiple perspectives and robustness tests. According to their own industrial structure and the carbon intensity constraint index provided by the central government, provinces will regulate industries with a high degree of pollution and large emission reduction space, set the threshold of emission control, design and operate monitoring, reporting, and verification mechanisms, and quantify the carbon emissions and emission reductions of enterprises [16]. The carbon trading price essentially reflects enterprises' demand, that is, the difference between their emissions and the given initial quota. The lower the free quota set in pilot areas, the larger the gap for enterprises to reduce emissions or purchase carbon emission rights, and the higher the average annual carbon price in the market [17]. Enterprises with lower marginal innovation costs are given priority in adopting cleaner production with advanced equipment [18]. While achieving the total volume control target of the industry in the current year, the market mechanism should be used to reduce the marginal costs of internal innovation and emissions. Buying enterprises can meet the emission reduction requirements of the year, and selling enterprises can obtain higher economic returns from innovation and finally achieve a win-win situation [19].

With rapid globalization, countries are becoming increasingly connected, and the impact of exogenous policies of specific industries on other production sectors is becoming increasingly obvious. As an important tool for describing the internal structure of economic transmission, production networks have become a key tool for studying the diffusion of micro-shocks and macroeconomic fluctuations. Based on domestic and international research results, this study discusses an empirical analysis method for the combination of specific exogenous shocks and production networks, as well as the transmission effects of carbon emission trading policies on production networks. First, production networks play an important role in the diffusion of micro-shocks. According to Dong (2019),Demirgüç-Kunt(2020), Deng (2021),Yu (2021), Wei et al. (2021), Wang et al. (2022), midstream industries often need to obtain inputs from multiple upstream industries and supply outputs to the downstream sector, and multiple industrial chains are interwoven to form a production network. Since production networks can describe the internal structure of economic transmission well, many studies have applied them to explain the diffusion of micro-shocks and study the path to achieve macroeconomic fluctuations. Second, an empirical analysis is carried out by combining specific exogenous shocks with a production network, and the mechanism and influence of network transmission are quantified from a microscopic perspective using the method of causal inference. Boehm et al. (2020) considered the Great East Japan Earthquake as an exogenous shock to investigate the transmission mechanism of the production network. They empirically demonstrated that the upstream and downstream impacts of natural disasters would have a negative impact on an enterprise's revenue, and the negative impact would be weakened with an increase in the distance between supply chains. Zacchia et al. (2020) regard the adjustment of the Resource Utilization Support Fund Tax for domestic importers in Turkey as an exogenous shock and examine the network effect of the policy in combination with the transaction data of domestic enterprises. Compared with the control group, the increased heterogeneity of the tax rate increases the cost burden on importers and prompts enterprises to change the input-output path, resulting in a significant increase in domestic suppliers for transactions. This cost constraint eventually leads to a cost increase and yield decrease for downstream enterprises through the production network but has no significant impact on upstream enterprises. Additionally, the production network transmission effect of the carbon emission trading policy exists objectively, but few studies have conducted in-depth analyses of the transmission effect of innovation [20,21]. Compared to the pilot period, the expansion of the carbon emission trading volume will improve the breadth and depth of production network transmission, and many industries, such as upstream photovoltaic wind power, midstream electric steelmaking, and downstream new energy vehicles, may benefit from carbon trading. Expanding the coverage of the carbon market has become a key task of the current policy, and based on the steady operation of the current carbon trading market in the power sector, it is expected that more high-emission industries will be gradually included. Therefore, investigating the innovation spillover effect of the pilot carbon emission rights policy and clarifying the innovation transmission path and mechanism of the policy based on the production network are inherent requirements for a comprehensive and objective evaluation of environmental regulation. This study explores the impact of pilot carbon emission trading policies on innovation from the perspective of the production network and the price mechanism of finished products. The specific objectives are as follows: (1) to explore the direct innovation and innovation spillover effects of the pilot carbon emission trading policy. Through theoretical and empirical analyses, it is demonstrated that the policy not only improves innovation in regulated industries but also has innovation spillover effects on non-regulated industries. (2) The effects of industry selection and cross-regional trading policies on the national carbon trading market are analyzed, and it is more conducive for innovation diffusion to include industries with a high degree of centrality and strong price adjustment ability in the national carbon market. At the same time, the research finds that compared to the intra-regional carbon trading market, cross-regional trading significantly improves the direct and spillover effects of innovation. (3) The transmission mechanism of network innovation was analyzed from the perspective of the price mechanism of finished products. In theory, Carvalho (2021) verifies that the cost reduction caused by technological progress will eventually spread to other industries through input-output linkages in the form of price through the endogenous production network framework. On this basis, this study empirically verifies the impact of price rise caused by total carbon emission controls on the innovation of non-regulated industries within the production network and discusses the heterogeneous impact of price transmission from two aspects: industry bargaining power and industry marketization degree, which enriches the research on mechanism transmission in the production network. In conclusion, the goal of this study is to explore the impact of the pilot carbon emissions trading policy and the national carbon trading market on innovation, as well as the role of the price mechanism of finished products. Studying these aspects can provide useful references for policymakers and enterprises to promote industrial upgrades and innovative developments.

2. Literature review

2.1. Economic impact of peak carbon emissions

According to the World Resources Institute (WRI), achieving a carbon peak is an important historical turning point, indicating that carbon emissions will begin to decline, thereby helping to decouple economic development from carbon emissions [22]. Therefore, the sooner the carbon peak target is achieved, the better is the carbon neutrality.

However, the carbon peak also impacts economic development. William Nordhaus integrated the economic system and the ecosystem by building a model in which the operation of the economic system would produce greenhouse gases [23], and a large amount of greenhouse gas emissions would affect the ecosystem, thus affecting the economic system and forming a cycle [24,25]. Gao et al. (2016) analyzed the economic impact of China's carbon emissions peaking based on the dynamic Global Trade and Production Environment with Energy model and found that the earlier the peaking time, the greater the negative impact. The main reasons are as follows. On the one hand, the earlier the peaking time, the earlier the decline of total carbon emissions, and the emissions are mainly from the secondary industry. On the other hand, the imposition of a carbon tax will have a negative impact on the return on capital of enterprises, which will lead to a decline in total investment and capital stock and the contraction of total output. Additionally, a slowdown in economic activity reduces indirect tax revenue, which has a negative impact on the economy. A contraction in aggregate output leads to a decline in demand for imported goods and an appreciation of the real exchange rate; thus, exports and imports have a corresponding negative impact. Emissions reduction has a negative impact on almost all industries in varying degrees. The output decline and price rise of the energy industry are due to the decrease in demand and impact of cost, whereas the output and price changes of non-energy products are due to the impact of emission costs [26,27]. Shao et al. (2022) believe that owing to unavoidable policy constraints and human intervention in the process of peaking carbon emissions, the limited resources originally used for output will be diverted into energy conservation and emission reduction activities, which will impact economic development. The earlier the peaking time, the more enterprises and consumers need to adjust in a short period of time. The higher the requirements for industrial structure, energy structure, and technological improvement caused by the premature replacement of infrastructure, the greater the pressure of emission reduction. Li et al. (2022) used the Time Evolution Computable General Equilibrium Model (TECGE) to quantitatively analyze the economic impact of strengthening carbon peaking commitment, and found that the earlier the carbon peak, the higher the carbon tax price required. The Gross Domestic Product (GDP) and other macroeconomic variables will decrease, but the proportion of the tertiary industry will increase. Therefore, carbon reduction policies will inevitably impact high-carbon industries, leading to a slowdown of economic activities and a loss of total output, investment, employment, and trade. Therefore, in addition to environmental protection considerations, carbon peaking and carbon neutrality goals also imply economic and financial significance.

2.2. Carbon neutrality measures and their economic implications

After the peak of carbon emissions, it is necessary to apply appropriate emission reduction measures, such as tree planting, energy conservation, and emission reduction to change the production and lifestyle of carbon emissions; otherwise, it is difficult to achieve the goal of carbon neutrality and the economy may be locked in a high-carbon state [28]. To achieve carbon neutrality, all carbon emissions should be reduced, mainly through administrative control, quantity permits, and price mechanisms represented by a carbon tax [29]. In addition, the carbon tariff, also known as the Carbon Boundary Adjustment Mechanism (CBAM), is a carbon emission tariff specifically imposed on imported products with high energy consumption [30].

Chen et al. (2020) noted that developed countries advocate the implementation of carbon tariffs to prevent carbon leakage and reduce competitiveness. Carbon leakage means that the implementation of strict emission reduction policies by developed countries will lead to the relocation of enterprises to countries and regions with relatively loose emission reduction policies (industrial outflow), which will then lead to the transfer of carbon emissions to other countries and regions without carbon emission reduction measures, implying that strict emission reduction policies will lead to an increase in domestic production costs. The domestic market and industry are affected by imported products that bear lower emission costs (market share substitution). Consequently, carbon reduction measures may incur additional costs through trade channels. Yuan (2017) pointed out that once unilateral climate trade barriers are implemented, various trade barriers to achieving disguised trade protection will continue to emerge, which may endanger the multilateral trading system. In their simulation study, He et al. (2021) found that the implementation of carbon tariff policies in European and American countries has a serious negative impact on China's industrial exports. A carbon tariff rate of US $30/ton or US $60/ton would lead to a substantial decline in the total output, total exports, and employment in the industrial sector, and the impact may last for five years or even longer.

2.3. Regional integration and environmental protection

Regional economic integration can mitigate the negative impacts of emission reduction policies on the economy. Tao et al. (2021) argue that reducing trade barriers to low-carbon products can promote the import and export of such products by developing countries and improve their economic and climatic benefits. The reduction of trade barriers to low-carbon products and improvements in the level of trade liberalization will also help developing countries improve their technological innovation and adaptability, thus reducing their dependence on technology transfer from developed countries. Ren et al. (2022) pointed out that, contrary to unilateral “low carbon barrier” policies such as the implementation of carbon tariffs, reducing or eliminating trade barriers for low carbon products can promote free trade while promoting the realization of emission reduction goals. International trade and climate governance policies support and integrate with each other, and their coordination is of great practical significance. First, promoting trade liberalization and facilitation under bilateral and multilateral trading systems is conducive to improving the economic welfare levels of countries or regions [31]; Second, global climate change concerns the future and destiny of mankind, and concerted efforts to deal with climate change concern the economic interests of countries or regions and even the sustainable development of mankind, which requires countries to strengthen cooperation rather than take unilateral measures in the process of climate governance [32]. As the issue of climate change has become increasingly prominent, trade policy, as a supplement to global climate negotiations, may be a potential enabler for the agreement and implementation of a global climate regime. This is because trade provides an internalization mechanism for the global externalities represented by climate change [33].

A new round of regional trade agreements is expanding its scope to include policies such as common environmental standards. Regional integration is manifested in deepening and developing economic cooperation, from the economic field to multi-field cooperation such as environmental protection [34,35,36,37]. Consider Regional Comprehensive Economic Partnership as an example. First, the signing of the RCEP reduced the barriers to trade in low-carbon products among countries in the region, which is conducive to promoting the development of a low-carbon economy. Second, two-thirds of RCEP member states have submitted new Intended Nationally Determined Contributions by 2020. Low-carbon economic cooperation based on high international standards is more conducive to energy transition and green development, which will reduce the costs for the society. Third, by negotiating a common environmental protection target and binding mechanism, regional members can avoid carbon leakage and the erosion of the international competitiveness of carbon-intensive products in a few countries by reducing emissions and mitigating the negative impacts of energy conservation and emissions reduction on the economy.

3. Research hypotheses

This section mainly refers to Simpson et al.’s (2017) technological progress model of environmental regulation to discuss the impact of carbon emission trading policies on the innovation level of regulatory industries and the spillover effect of the production network. This study focuses on the decision-making of regulatory enterprises and their upstream and downstream enterprises against the background of carbon reduction. it also analyzes the impact of the carbon emission trading system on the innovation decision-making of representative enterprises in regulated industries and their upstream and downstream representative enterprises.

Assume that there are $\mathrm{n}$ production sectors in the economy, of which sector $\mathrm{d}$ is the fossil energy sector (coal, oil, and natural gas). It is assumed that all production sectors are perfectly competitive, have a Cobb-Douglas production function, and both labor and intermediate goods are used for production. The intermediate goods are the finished products of other production departments, the total output of the department is expressed as $\mathrm{Y}\text{,}$ $\mathrm{L}$ represents the quantity of labor input, $w$ represents wages and is consistent in all production departments, $X_j$ represents the input quantity of intermediate product $j$, the price of product $j$ is represented by $P$, and the average cost of product $j$ is represented by ${\mathrm{c}}_{\mathrm{j}}$.

3.1. Carbon emission trading policy and regulation of enterprise innovation

The total carbon emissions mainly depend on the direct consumption of fossil products as intermediates. Therefore, it is assumed that the representative regulatory enterprise is the direct downstream sector of the fossil energy industry, and its production function is shown as $\mathrm{C}-\mathrm{D}$:

$${\mathrm{Y}}_{\mathrm{i}}={\mathrm{F}}_{\mathrm{i}}\left({\mathrm{A}}_{\mathrm{i}},{\mathrm{L}}_{\mathrm{i}},{\mathrm{X}}_{\mathrm{i}}\right)={\mathrm{A}}_{\mathrm{i}}{{\mathrm{L}}_{\mathrm{i}}}^{\left(1-{\mathrm{\alpha }}_{\text{id}}\sum _{\mathrm{j}\in {\mathrm{S}}_{\mathrm{i}}}{\mathrm{\alpha }}_{\text{ij}}\right)}{\left({\mathrm{X}}_{\text{id}}\right)}^{{\mathrm{\alpha }}_{\text{id}}}\prod _{\mathrm{j}=2}^{\mathrm{n}}{\left({\mathrm{X}}_{\text{ij}}\right)}^{{\mathrm{\alpha }}_{\text{ij}}}$$ (1)

where $Y_i$ is the total output of enterprise $i$, requires the input of both labor and intermediate products; $A_i$ is the production technology of industry $i$, $L_i$ is the input quantity of labor; $X_{id}$ is the input quantity of intermediate products of fossil products of enterprise $i$, and sector $d$ is the primary production sector, $X_{ij}$ is the input quantity of other intermediate products of industry $i$. To distinguish intermediate products from fossil products, set $j$ starting from 2 [38].

According to the calculation method for carbon emissions, the total carbon emissions of enterprise $i$ depend on

$$e\left(Y_i\right)={\epsilon }_d{X}_{id}$$ (2)

where $e\left(Y_i\right)$ represents the total carbon emissions of enterprise $i$ when the production level is $Y_i$, ${\epsilon }_d$ is the carbon emissions coefficient of fossil product $d$, and $X_{id}$ is the total use of fossil products by enterprise $i$ [39].

Under a carbon emissions trading policy, the enterprise must pay for its excess carbon emissions. Given that the price of a unit of carbon is $P_e$, the free emission allowance stipulated by the government is $e^{*}$, and the carbon emission cost to be paid by the enterprise is $P_e\left(e\left(Y_i\right)-e^{*}\right)$. This should be distinguished from a carbon emissions tax policy. In the case of a carbon tax policy, if an enterprise's carbon emissions are lower than the free emission allowance （$e\left(Y_i\right)<e^{*}$）stipulated by the government, the enterprise's carbon emission cost will be 0 [40]. The enterprise's total profit function is:

$${\pi }_i=P_i{Y}_i-w{L}_i-P_d{X}_{id}+\sum _{j=2}^n{P}_j{X}_{ij}-P_e\left(e\left(Y_i\right)-e^{*}\right)$$ (3)

In the equation, $w$ is salary, $P_i$ represents product price of enterprise $i$, and $P_j$ represents product price of intermediate product of $j$. The research of Dong et al. (2019) can be referenced while considering the innovation decision of enterprises:

$${\mu }_i={\lambda }_i{\left(R_i\right)}^{{\phi }_i}$$ (4)

In the equation, ${\lambda }_i$ is the Research and Development (R&D) efficiency parameter. $R_i$ is the innovation input of enterprise $i$, ${\phi }_i$ is the output elasticity of R&D input, and ${\phi }_i$ >l. ${\mu }_i$ is the probability of enterprise $i$'s technology being improved, and the technology level rising from $A_i$ to $\left(1+{\beta }_i\right)A_i$, whereas the probability of enterprise's technology not being improved is $1-{\mu }_i$.

According to the probability of innovation success ${\mu }_i$ and level of innovation progress ${\beta }_i$, the expected technical level after innovation can be obtained as follows [41]

$$E\left(A_i\right)={\mu }_i\left(1+{\beta }_i\right)A_i+\left(1-{\mu }_i\right)A_i$$ (5)

Therefore, when the innovation input is $R_i$, the expected technological progress is [42]:

$$\Delta A_i={\mu }_i{\beta }_i{A}_i$$ (6)

According to the market clearing hypothesis of the perfectly competitive market, to maximize the current profit, all price variables must be assumed to be exogenous, and the current input ratio of intermediate products should be considered as the optimal ratio, and the enterprise will reduce the input of all intermediate products in proportion [to $1/\left({\mu }_i{\beta }_i+1\right)$ of the original input]. To simplify the expression, let $C_i$ represent the production cost of the enterprise, that is, $C_i=w{L}_i+P_d{X}_{id}+\sum _{j=2}^n{P}_j{X}_{ij}$. The expected profit ${\pi }^{,}$ after innovation input is [43]

$${\pi }^{,}=P_i{Y}_i-\frac{C_i}{{\mu }_i{\beta }_i+1}-P_e\left(\frac{{\epsilon }_d{X}_{id}}{{\mu }_i{\beta }_i+1}-e^{*}\right)-R_i$$ (7)

In the equation, ${\pi }^{,}$ represents the expected profit after innovation input, $\frac{C_i}{{\mu }_i{\beta }_i+1}$ represents the expected production cost after innovation input, $\frac{{\epsilon }_d{X}_{id}}{{\mu }_i{\beta }_i+1}$ represents the expected total carbon emission after innovation input, and $R_i$ represents the total innovation input. Thus, the expected net return on innovation input can be obtained as

$${\pi }^{,}-\pi =-\frac{C_i}{{\mu }_i{\beta }_i+1}+C_i-P_e\left(\frac{{\epsilon }_d{X}_{id}}{{\mu }_i{\beta }_i+1}-{\epsilon }_d{X}_{id}\right)-R_i$$ (8)

Equation (8) indicates that the expected net income of innovation depends on the expected change in the production cost, carbon emissions cost, and total input of innovation. If ${\pi }^{,}-\pi$ >0, it means that the enterprise can make profits by investing in innovation; if ${\pi }^{,}-\pi$ <O, it means that the enterprise may face net loss by investing in innovation. By differentiating equation (8) with ${\mu }_i$, we obtain

$$\frac{d\left({\pi }^{,}-\pi \right)}{d{\mu }_i}=\frac{{\beta }_i{C}_i+{\beta }_i{P}_e{\epsilon }_d{X}_{id}}{{\left({\mu }_i{\beta }_i+1\right)}^2}$$ (9)

${\pi }^{,}-\pi$ is equal to 0, and the threshold of innovation success probability can be calculated as [44]:

$${\mu }_i=\frac{R_i}{{\beta }_i{C}_i+{\beta }_i{P}_e{\epsilon }_d{X}_{id}-{\beta }_i{R}_i}$$ (10)

The threshold value of innovation represents an enterprise's expectation of the possibility of profits from innovation. For an innovation, if the probability of success is greater than ${\mu }_i$, it means that the enterprise will gain net profit by investing in innovation; if the probability of success is less than ${\mu }_i$, it means that the enterprise will generate net loss by investing in innovation. Therefore, the greater the ${\mu }_i$, the more conservative the enterprise is in innovation decision-making. The smaller the ${\mu }_i$, the more aggressive the enterprise is in making innovation decisions [45].

When charging only pollutant emission fees to carbon emitters, because excess emission reduction cannot bring direct benefits to enterprises, the optimal emission decision value of enterprises is not lower than the carbon emission quota limited by the government, and there is no economic incentive for further emission reduction. Under the carbon emissions trading pilot policy, enterprises can profit from emission reductions exceeding $e^{*}$ through market trading; thus, the lower limit of the innovation threshold will be lower. Compared to the carbon tax policy, the carbon emission trading policy allows low-emission enterprises to sell emission rights, which has a stronger role in promoting the innovation of low-emission enterprises.

Therefore, H1 is proposed: carbon emission trading policies can promote the innovation of representative enterprises in regulated industries.

3.2. Upstream and downstream transmission of carbon emission trading and innovation

Thus, there is a correlation between carbon emissions trading policies and enterprise innovation regulation. Jaffe (2018) believes that interindustry fluctuations are transmitted through the prices of intermediate goods. To further explore the product price changes in the regulated industry after policy implementation, it is assumed that there are several homogeneous enterprises in a perfectly competitive market. Facing the impact of the policy, they can choose not to pay the independent emission reduction cost and buy a carbon quota from the market or actively reduce emissions and sell excess carbon allowances to offset the cost of innovation through technological and process innovation. To realize market clearing, the carbon emission rights purchased should equal the total amount of carbon emission rights sold, and the enterprise's total production cost is $w{L}_i+P_d{X}_{id}+\sum _{j=2}^n{P}_j{X}_{ij}+P_e\left(e\left(Y_i\right)-e^{*}\right)$. $C_i$ represents the pure production cost of the enterprise, that is, $C_i=w{L}_i+P_d{X}_{id}+\sum _{j=2}^n{P}_j{X}_{ij}\text{.}$ Thus, the enterprise's total production cost is expressed as $C_i$, that is, $C_i+P_e\left(e\left(Y_i\right)-e^{*}\right)$. There is an equilibrium price of $P_e^{*}$ for a number of completely homogeneous enterprises; therefore, the cost of implementing the two types of enterprise decisions is not different. The regulated enterprise's unit cost is the product price in the industry [46].

$$P_i^{,}=c_i^{,}=\frac{C_i}{Y_i}+P_e\left(\frac{{\epsilon }_d{X}_{id}-e^{*}}{Y_i}\right)$$ (11)

The cost of emission reduction for enterprises and the cost of purchasing carbon emission rights are included in the cost of production. Product prices in the regulated industry rise according to the assumption of a perfectly competitive market.

Suppose that the production function of representative enterprise $b$ in the downstream directly related industry is

$$Y_b=F_b\left(S_b,A_b\left(S_b\right),L_b,X_b\right)=A_b\left(S_b\right){L_b}^{\left(1-{\alpha }_{id}\sum _{j\in S_i}{\alpha }_{ij}\right)}{\left(X_{br}\right)}^{{\alpha }_{br}}\prod _{j=2}^n{\left(X_{bj}\right)}^{{\alpha }_{bj}}$$ (12)

where $X_{bj}$ represents the input of products from the upper- and middle-regulation industries produced by Enterprise $b$. The initial price of the product in the regulated industry is $P_i$. After the implementation of the policy, the product price increases from $P_i$ to $P_i^{,}$. Although downstream enterprises do not directly participate in carbon emission rights trading [47], considering the increase in intermediate goods prices caused by an increase in upstream costs, the profits of downstream enterprises after policy implementation are as follows:

$${\pi }_b=P_b{Y}_b-w{L}_b-P_i^{,}{X}_{bi}-\sum _{j=2}^n{P}_j{X}_{bj}$$ (13)

where $P_b{Y}_b$ represents the total output income of enterprise $b$, $w{L}_b$ represents the total labor cost, $P_i^{,}{X}_{bi}$ represents the use cost of intermediate goods in the regulated industry, and $\sum _{j=2}^n{P}_j{X}_{bj}$ represents the use cost of the other intermediate goods.

In the context of carbon emissions trading policies [48], the expected profit function of the downstream representative enterprises after innovation is

$${\pi }_b^{,}=P_b{Y}_b-\frac{w{L}_b+P_i^{,}{X}_{bi}+\sum _{j=2}^n{P}_j{X}_{bj}}{{\mu }_i{\beta }_i+1}-R_b$$ (14)

In the equation, ${\pi }_b^{,}$ represents the expected profit after innovation input, $\frac{w{L}_b+P_i^{,}{X}_{bi}+\sum _{j=2}^n{P}_j{X}_{bj}}{{\mu }_i{\beta }_i+1}$ represents the expected production cost after innovation input, and $R_b$ represents the total innovation input.

The expected profit function of innovation is subtracted from the profit function before innovation to obtain the expected net profit of the innovation input.

$${\pi }_b^{,}-{\pi }_b=-\frac{w{L}_b+P_i^{,}{X}_{bi}+\sum _{j=2}^n{P}_j{X}_{bj}}{{\mu }_i{\beta }_i+1}+w{L}_b+P_i^{,}{X}_{bi}+\sum _{j=2}^n{P}_j{X}_{bj}-R_b$$ (15)

In the equation, ${\pi }_b^{,}-{\pi }_b$ represents the expected net income of innovation, and equation (15) shows that the expected net income of innovation depends on the expected change of production cost, the change of regulatory industry product price, and the total input of innovation.

Based on the analysis in Section 2.1, the threshold value of the innovation success probability can be calculated as [let $P_i^{,}=c_i+P_e\left(\frac{{\epsilon }_d{X}_{id}-e^{*}}{Y_i}\right),C_b=w{L}_b+P_b{X}_{bi}+\sum _{j=2}^n{P}_j{X}_{bj}$]:

$${\mu }_b=\frac{R_b}{\left(w{L}_b+P_i^{,}{X}_{bi}+\sum _{j=2}^n{P}_j{X}_{bj}-R_b\right){\beta }_b}$$ (16)

Assuming that there is no carbon emissions trading policy, that is, the price of regulated industrial products does not change in the profit function （$P_i=P_i^{,}$） [49], the innovation threshold of an enterprise can be calculated as follows:

$${\mu }_b=\frac{R_b}{{\beta }_b{C}_b-{\beta }_b{R}_b}$$ (17)

By comparing equations (16), (17), it can be seen that when carbon emissions trading policies exist, the threshold of innovation success of downstream representative enterprises decreases. The carbon emissions trading policy increases the production cost of the regulated industry and promotes the innovation of downstream enterprises in the regulated industry in the form of cost promotion.

Accordingly, hypothesis H2a is proposed: A pilot carbon emissions trading policy increases the production cost of representative enterprises in regulated industries, which leads to an increase in the price of finished products of representative enterprises in regulated industries and finally forces downstream representative enterprises to innovate through the price mechanism.

The innovation transmission of the policy was investigated from upstream enterprises directly related to the regulation industry. It can be inferred from Section 2.1 that, from the perspective of representative enterprises in the upstream industry, this impact makes the demand for the upstream product $f$ of industry $i$ drop from $Y_f$ to $\frac{Y_f}{{\mu }_i{\beta }_f+1}$ [50]. It can be concluded that when there is a carbon emissions trading policy, the profit function of representative upstream enterprises is

$${\pi }_f=P_f\frac{Y_f}{{\mu }_i{\beta }_f+1}-\frac{C_f}{{\mu }_i{\beta }_f+1}$$ (18)

By introducing the probability of innovation success ${\mu }_f$, the expected net benefits of innovation can be obtained as follows [51]:

$${\pi }_f^{,}-{\pi }_f=\frac{C_f{\mu }_f{\beta }_f}{\left({\mu }_i{\beta }_f+1\right)\left({\mu }_f{\beta }_f+1\right)}-R_f$$ (19)

If there is no carbon emission trading policy [52], the expected net income from innovation of representative upstream enterprise $i$ is

$${\pi }_f^{,}-{\pi }_f=\frac{C_f{\mu }_f{\beta }_f}{{\mu }_f{\beta }_f+1}-R_f$$ (20)

We subtract the two equations to obtain the net innovation income in the presence of carbon emissions trading rights:

$$\frac{-C_i{\mu }_i{\beta }_f{\mu }_k{\beta }_f}{\left({\mu }_k{\beta }_f+1\right)\left({\mu }_i{\beta }_f+1\right)}<0$$ (21)

Equation (21) indicates that the expected net income of upstream enterprises decreases after the pilot program. For regulatory enterprises that achieve independent emission reduction through innovation, the conversion of the old and new driving forces increases the unit input-output. For regulatory enterprises that buy carbon allowances, cost constraints reduce demand for upstream products. In conclusion, pilot policies lead to a lower demand for upstream products in regulated industries, and the total cost of innovation for upstream firms can only be shared by a smaller number of finished products.

On this basis, H2b is proposed: a carbon emission trading policy reduces the innovation level of representative upstream enterprises. The comprehensive theoretical hypothesis is presented in Fig. 1.

Fig. 1.

Theoretical logic of carbon emission trading pilot affecting innovation.

4. Econometric model setting and data sources

4.1. Data sources

The following analysis elaborates on the policies and measures of innovation-driven development in each city and the innovation performance of enterprises. Through an in-depth analysis of these data, it is hoped that the characteristics and advantages of Chinese cities in the field of innovation can be revealed and useful references can be provided to promote the high-quality development of China's economy.

4.1.1. Innovation indicators

This study adopted the four-digit city industry innovation index calculated in the 2022 Venture Capital Report of China's Mainland Provinces and Cities. Compared with other measurement indicators, this innovation level index has two advantages. First, the measurement standard is more objective and fully considers the depreciation and time value of innovation. Second, the specific value is subdivided into four-digit industries of all cities, which is conducive to an in-depth analysis of the spillover effects between industries. As the innovation index obtained according to the above calculation idea is a cumulative quantity, this study further subtracted the innovation index of the previous year from the innovation index of the current year and calculated the increment of the innovation index of each year to measure the change in the innovation level at the city industry level [53,54,55]. Furthermore, in the robustness test, the patent data at the city and listed company levels were introduced as explanatory variables to verify the robustness and credibility of the conclusion to the maximum extent.

4.1.2. Control variables

We can consider the level of innovation in a city from several perspectives. First, we refer to the research of Xie and Hu (2020), which used select indicators such as population, GDP, foreign direct investment, government expenditure on education, and government expenditure on research and development of prefecture-level cities, and take the logarithms of these indicators from the China City Statistical Yearbook over the years. (1) Population size. Population is an important control variable that reflects the accumulation of the labor force in a city. A city with a more abundant labor force may have more concentrated human capital and a higher level of innovation (Wang et al., 2019). For example, China's Guangdong province has a huge labor pool that provides strong support for its economic development and innovation capacity. In addition, with the increasingly serious problem of an aging population, some countries and regions have begun to pay attention to the importance of talent introduction and training to improve their level of innovation. (2) GDP. The GDP is an important indicator of a city's overall level of economic development. The higher the level of economic development, the stronger the demand for innovation and development. For example, Shanghai and Shenzhen, two international metropolitan cities with high GDP, have achieved remarkable results in scientific and technological innovation. In addition, some countries and regions have implemented policies to encourage enterprises to invest more in R&D to improve their overall economic competitiveness [56,57]. (3) Foreign direct investment. Foreign direct investment can promote the innovative development of cities by introducing foreign technology spillovers. For example, Shenzhen has attracted many foreign-funded enterprises since its reform and opening up, and these enterprises have provided a strong impetus for the city's economic development and innovation capacity. In addition, foreign investment can promote market competition and encourage local enterprises to constantly improve their innovation capabilities. (4) Government expenditure on education. The Government's expenditure on education reflects the importance that local governments attach to human capital. Regions with high spending on education are more likely to produce innovative people, and therefore, have higher levels of innovation. For example, the Singaporean government attaches great importance to education and invests a large amount of funds in educational infrastructure construction and personnel training every year, making Singapore highly competitive in terms of scientific and technological innovation. In China, the government is also increasing its investment in education to improve the scientific and cultural quality of its people as well as their ability to innovate. (5) Government expenditure on R&D The Government's R&D expenditure directly reflects government investment in innovation. By increasing investment in R&D, the government can support scientific research institutions and enterprises in making breakthroughs in key technological fields, thus promoting the innovative development of the society. For example, the Silicon Valley region in the United States has become an important engine of global scientific and technological innovation because of the strong support from federal and local governments. In China, the government is also increasing its investment in scientific research, promoting the integration of industry, universities, and research, and fostering several internationally competitive innovative enterprises [58,59]. In summary, by analyzing indicators such as population, GDP, foreign direct investment, government spending on education, and government spending on R&D in prefecture-level cities, we can comprehensively understand the factors that influence the level of innovation in cities. On this basis, policies and measures can be formulated to promote innovative development in cities.

4.1.3. Input-output data

The data is derived from the inter-regional input-output table of the China Urban Statistical Yearbook in 2021, which accurately quantifies the input-output relationship between regional industries based on inter-regional product flow and the input of intermediate goods. The descriptive statistical results are shown in Table 1, where the maximum value of innovation level is approximately 40.9, and the minimum value is about −1.2 (due to the particularity of measurement methods, if the innovation is not enough to make up for the depreciation amount of the current year's stock innovation, the innovation level will be negative), indicating that there is a large gap in the growth of innovation level among industries in different regions. The average innovation level is approximately 0.38, indicating that the overall innovation level is low. It should be noted that the sample size of the control variable is smaller than that of the explanatory and explained variables owing to the missing data of some control variables.

Table 1.

Descriptive statistics.

Variables	$Obs$	Mean	Standard deviation	Minimum value	Maximum value
$Innovation$	1379040	0.3158	2.0696	−1.2358	40.8640
$Region\times Regulated$	1379040	0.0183	0.1341	0	1
$Population$	1178440	4.5684	0.7986	0.0953	7.9411
$Education$	1177080	10.2139	1.3086	0.3284	15.2106
$FDI$	1101600	10.1837	1.7159	2.8332	14.9413
$GDP$	1176400	16.4605	0.9058	14.1063	19.4567
$R\&D$	1177080	12.9972	0.7605	9.2415	15.9986

4.2. Model setting

4.2.1. Direct effect model setting

The selection of regulatory samples and innovation identification strategies is directly related to the robustness of the conclusions. By matching publicly traded enterprise directories in each carbon trading market, this study finds that only 98 of the 1000 companies in the directory are listed companies or their subsidiaries. Considering the systematic differences between the two types of enterprises in terms of innovation ability, the conclusion may be biased when using only the data samples of listed companies. Although regional-level data can effectively describe the innovation level of a place, the results only represent the average effect within the range. It is difficult to clarify the micro mechanisms of policy action and quantify industry heterogeneity. Therefore, this study uses city-industry data to identify the impact of policies on two types of industries and replaces samples with firm- and city-level data in a robustness test. The following dual-variance benchmark model is constructed to examine the direct innovation effect of the pilot policy on regulated industries:

$${Innovation}_{ijt}={\alpha }_1{Region}_j\times Time+{\alpha }_2{Controls}_{jt}+{\mu }_{ij}+{\delta }_t+{\epsilon }_{ijt}$$ (22)

In the equation, ${Innovation}_{ijt}$ is used to measure the innovation level of regulated industry $i$ in region $j$ at time $t$. ${Region}_j$ is the region grouping variable, which is 1 for pilot areas and 0 otherwise. $Time$ is the grouping variable of time, which is 1 in 2014 and after and 0 otherwise. ${Controls}_{jt}$ represents the control variable at the regional level, ${\mu }_{ij}$ is the fixed effect at the regional industrial level, ${\delta }_t$ is the fixed time effect, and ${\epsilon }_{ijt}$ is the residual term. ${\alpha }_1$ is used to measure the direct impact of the carbon emission trading policy on the regulated industry [60,61]. If the value is significantly positive, this indicates that the policy can improve the innovation level of the regulated industry in the pilot area.

4.2.2. Network spillover effect model setting

To investigate the innovation network spillover effect of policies and accurately quantify this spillover effect [62], this study constructs a triple-difference model at the regional industry level:

$${Innovation}_{ijt}={\gamma }_1{Region}_j\times Time\times {Unregulated}_i+{\gamma }_{2}Time\times {Unregulated}_i+{\gamma }_3{Region}_j\times Time+{\gamma }_4{{Controls}_{jt}+{\mu }_{ij}+\delta }_t+{\epsilon }_{ijt}$$ (23)

where ${Unregulated}_i$ is the grouped variable of industries, unregulated industries are equal to 1, and 0 otherwise, and ${Region}_j$ is the grouped variable of regions, which is the same as above. The triple interaction coefficient ${\gamma }_1$ is used to measure the network conduction effect of the carbon emission trading policy. If the coefficient is significantly positive, this indicates that the network conduction effect is stronger. This structural thought is mostly used to verify the indirect influence of the impact. For example, Ye et al. (2022) believed that the promulgation of cleaner production industry standards would not only impact the regulatory industry but also trigger innovation in related industries; therefore, the upstream manufacturing industry was set as the experimental group to investigate the innovation spillover effect of policies. Accordingly, the implicit assumption of this model is that as the industries within the province are more closely related than the industries outside the pilot, the non-regulated industries within the pilot are more closely related to the production of the regulated industries, thus receiving a stronger innovation promotion effect. The degree of correlation based on the input-output correlation between regions also supports this view.

5. Innovation effect test of policy

5.1. Parallel trend test

When analyzing changes in industry innovation levels (Fig. 2), in addition to observing data and trends, we need to use certain methodologies to explain these phenomena. In this study, the above content is elaborated from the perspective of qualitative and quantitative analyses. First, from the qualitative analysis perspective, there was no significant difference in the change in innovation levels between pilot and non-pilot areas before 2014. This finding indicates that there may have been a certain degree of lag in promoting innovation in various regions during the early stages of policy implementation. However, this does not imply that there are essential differences in innovation capacity between regions. During policy implementation, regions may adopt different strategies to promote innovation based on their resources, industrial characteristics, and development needs. Therefore, in the early stages, changes in innovation levels in different regions may be affected by various factors, such as the intensity of policy implementation, local government support, and enterprise input. After 2014, the speed of innovation improvement in regulated industries in the pilot regions increased significantly. This change can be explained from a quantitative analysis perspective. By comparing the data before and after 2014, we observe the following.

• (1)Increased government policy support. With the promotion and implementation of the policy, the government in the pilot areas continuously increased its support in terms of capital, talent, and technology, which helped improve the innovation enthusiasm of enterprises and R&D investment.
• (2)Optimizing the innovation environment. The pilot areas have provided favorable external conditions for the development of enterprises through measures such as optimizing the business environment, strengthening intellectual property protection, and promoting cooperation among enterprises, universities, and research institutes.
• (3)Enterprises' effort. Under the pressure of fierce market competition, enterprises in pilot areas have gradually realized that innovation is the key to maintaining competitiveness. Consequently, companies have increased their investment in R&D to promote technological innovation and product upgrades.
• (4)Regional synergy Certain regional synergies were formed among the pilot regions, and the overall level of innovation was improved by sharing resources and exchanging experiences.

Fig. 2.

Parallel trend test of regulated industries. Note: All data are from China Urban Statistical Yearbook 2013–2022.

In summary, the reasons for significant improvement in the innovation level of regulatory industries in the pilot regions after 2014 include the strengthening of government policy support, innovation environment optimization, enterprises’ efforts, and regional synergies. These factors jointly promoted rapid improvements in the innovation levels in the pilot areas. In future research, the mechanisms of these factors could be discussed in depth to provide theoretical support for the formulation of more effective innovation development strategies in various regions.

5.2. Benchmark regression

The sample used in this study included all regulated industries, and used a multivariate analysis method. The coefficient of the interaction term is significantly positive, indicating that the policy has a significantly positive impact on innovation in regulated industries in pilot areas compared to regulated industries in non-pilot areas. This method can effectively control the influence of other factors to evaluate the direct innovation effect of a policy more accurately. Additionally, a time-series analysis method was used to assess the dynamic effects of policies on regulated industries. In Table 2, the policy effect coefficient increases from 0.0453 to 0.181, indicating that this positive impact increases over time. This method can better capture changing trends during policy implementation to understand the effects of policies more comprehensively. In conclusion, this study uses a variety of methodological means to evaluate the direct innovation effect of the pilot policy of carbon emissions trading and carries out an in-depth analysis and discussion from multiple perspectives and levels. These methods can not only improve the reliability and accuracy of the study but also provide a more scientific and objective basis for policy formulation and practice (see Table 3).

Table 2.

Regression results of direct innovation effect of pilot policies on regulated industries.

Explained variables	$Innovation$	$Innovation$
	（1）	（2）
$Region\times Time$	0.0996*** （0.0163）
$Region\times Year2014$		0.0453** （0.0188）
$Region\times Year2015$		0.0973*** （0.0214）
$Region\times Year2016$		0.163*** （0.0271）
$Region\times Year2017$		0.166*** （0.0282）
$Region\times Year2018$		0.171*** （0.0302）
$Region\times Year2019$		0.174*** （0.0326）
$Region\times Year2020$		0.176*** （0.0331）
$Region\times Year2021$		0.181*** （0.0342）
$Population$	0.0898*** （0.00869）	0.0898*** （0.00870）
$Education$	0.0166*** （0.00393）	0.0121*** （0.00403）
$FDI$	0.00471** （0.00221）	0.00550*** （0.00225）
$GDP$	0.227*** （0.0212）	0.225*** （0.0212）
$R\&D$	0.0103 （0.00917）	0.00876 （0.00924）
$Controls$	Yes	Yes
$YearFE$	Yes	Yes
$CityFE\times IndustryFE$	Yes	Yes
$Constant$	−4.143*** （0.335）	−5.146*** （0.156）
$Observations$	174528	924352
$R-squared$	0.871	0.867

Note: The brackets represent robust standard errors, where *** represents p < 0.01, ** represents p < 0.05, * represents p < 0.1, and the remaining values are the same. source: 2015–2022 Venture Capital Report of Provinces and Cities in Mainland China.

Table 3.

PSM regression results.

Explained variables	$Innovation$	$Innovation$
	（1）	（2）
$Region\times Time$	0.0411*** （0.0116）	0.0696*** （0.0169）
$Controls$	Yes	Yes
$YearFE$	Yes	Yes
$CityFE\times IndustryFE$	Yes	Yes
$Constant$	−3.344*** （0.297）	−6.415*** （0.690）
$Observations$	150876	124740
$R-squared$	0.840	0.880

Data source: Venture Capital Report of China's Mainland Provinces and Cities 2013–2022.

5.3. Robustness test

5.3.1. Propensity score matching-difference-in-difference analysis

Most carbon emissions trading pilots are in the central and eastern regions because these regions have a higher level of economic development and relatively large government R&D expenditure. Therefore, when conducting Propensity score matching-Difference-in-difference (PSM-DID) analysis, it is necessary to consider the systematic differences in the economic level and government R&D expenditure of different regions. To solve this problem, two matching methods are used in this study. The first matching method is based on city control variable data from 2013 and city innovation level data with a lag of one phase. The advantage of this matching method is that the differences between cities can be better controlled to evaluate the impact of pilot policies on regulatory industries more accurately. At the same time, because of the use of lagged data, it can better reflect the effects of policy implementation. The second matching method is based on the city control variable data of 2012 to conduct neighbor matching with calipers of 0.05; matching cities are screened, and column (2) is obtained through regression analysis. The advantage of this matching method is that the matching results can be obtained faster, thereby saving time and resources. However, because only two years of data were used, some important factors may have been overlooked, resulting in less accurate results. Under the two matching methods, the pilot policy has a significant positive impact on regulated industries, which is consistent with the benchmark regression conclusion. This indicates that the pilot carbon emission trading policy can promote the transformation, upgrading, and development of regulated industries. Owing to the small sample size and limited data sources, the results of this study may have limitations. Therefore, it is necessary to further expand the sample size and strengthen the research on data analysis methods to improve the credibility and reliability of this study.

5.3.2. Triple difference

The benchmark regression shows that the innovation level of regulated industries in the pilot has significantly improved. To test whether a spillover effect exists and verify which industries have higher innovation levels, this study adopted a triple difference model. Table 4 (1) shows the results of the triple ${Region}_j\times {Unregulated}_i\times {Time}_t$ difference average effect of the core explanatory variable with a coefficient of 0.0761. This passes the significance test at the 1% level, indicating that the innovation level of non-regulated industries in the pilot area is significantly higher than that of regulated industries in the pilot area. Column (2) introduces the fixed effect of the year of the city and industry, and the result is still significantly positive. Column (3) reports the results of the dynamic effects of the triple difference. The coefficient in the first year is negative and then becomes positive, indicating that the policy's spillover effect is increasing. Column (4) introduces a high-dimensional fixed effect to control for changes in the characteristics of cities and industries over time. The results show that the coefficient of the first stage is negative but not significant, whereas the coefficients of the second to eighth stages are still significantly positive. The reasons for this result may be as follows. First, there is a certain lag in the transmission of innovation in the production network, which leads to an improvement in the innovation level in the regulated industry before the non-regulated industry over time. Second, when the non-regulated industry obtains the clean finished products of the regulated industry, the uncertainty of innovation and the cost of emission reduction can be avoided to a certain extent. Additionally, considering the diffusion of emissions-reduction technology in the industrial chain, non-regulated industries can invest more resources in innovation, and the probability of innovation success is higher.

Table 4.

Results of triple difference regression.

Explained variable	Average effect	Average effect	Dynamic effect	Dynamic effect
	$Innovation$	$Innovation$	$Innovation$	$Innovation$
	（1）	（2）	（3）	（4）
$Region\times Unregulated\times Time$	0.0761*** （0.0179）	0.0747*** （0.0166）
$Region\times Time$	0.0929*** （0.0159）		0.0974*** （0.0160）
$Unregulated\times Time$	0.0149*** （0.00447）		0.0149*** （0.00447）
$Region\times Unregulated\times Year2014$			−0.0788*** （0.0180）	−0.0202 （0.0198）
$Region\times Unregulated\times Year2015$			0.101*** （0.0196）	0.102*** （0.0223）
$Region\times Unregulated\times Year2016$			0.206*** （0.0227）	0.143*** （0.0279）
$Region\times Unregulated\times Year2017$			0.258*** （0.0196）	0.163*** （0.0291）
$Region\times Unregulated\times Year2018$			0.302*** （0.0229）	0.181*** （0.0301）
$Region\times Unregulated\times Year2019$			0.351*** （0.0232）	0.193*** （0.0326）
$Region\times Unregulated\times Year2020$			0.376*** （0.0236）	0.196*** （0.0331）
$Region\times Unregulated\times Year2021$			0.382*** （0.0243）	0.198*** （0.0338）
$Controls$	Yes		Yes
$YearFE$	Yes		Yes
$CityFE\times IndustryFE$	Yes	Yes	Yes	Yes
$CityFE\times YearFE$		Yes		Yes
$IndustryFE\times YearFE$		Yes		Yes
$Constant$	−4.993*** （0.142）	0.386*** （0.00124）	−4.823*** （0.141）	0.386*** （0.00124）
$Observations$	1098880	1098880	1098880	1098880
$R-squared$	0.867	0.890	0.867	0.890

Data source: Venture Capital Report of China's Mainland Provinces and Cities 2015–2022.

5.4. Endogenic treatment

Although the endogenous problems caused by two-way causality are objectively weakened by policy influences, the results may still be biased if there are missing variables that affect both the selection of pilot cities and the innovation level of the industry. To ensure the stability of the conclusions, this study used the regional Ventilation Coefficient as an instrumental variable. The selection of instrumental variables should satisfy exogeneity and correlation requirements. From an exogeneity perspective, the ventilation coefficient as a natural condition does not directly affect the innovation level of urban industries. From the perspective of correlation, the higher the ventilation coefficient, the stronger the dilution ability of pollution, and the environmental pollution caused by high energy consumption is relatively insignificant. Therefore, the pilot will prioritize areas with low ventilation coefficients and relatively prominent pollution, which may be a negative relationship between them [38]. The calculation is as follows:

$${VC}_{it}=\frac{\sum _{m=1}^{m=12}{WH}_{im}\times {BLH}_{im}}{12}$$ (24)

The data were obtained from the reanalysis data provided by the European Center for Medium Range Weather Forecasts (ECMWF). ${VC}_{it}$ is the ventilation coefficient of $t$ year in region $i$, ${WH}_{im}$ is the wind speed at 10 m in $m$ month in region $i$, and ${BLH}_{im}$ is the boundary layer height in $m$ month in region $i$. Spatial information was matched with the urban latitude and longitude. The ventilation coefficient was obtained based on the annual average of the interaction between the regional boundary layer height and the wind speed 10 m before the implementation of the policy. The regression results are presented in Table 5. The first-stage test results proved that the ventilation coefficient was significantly and negatively correlated with the selection of the pilot areas. In the second stage, the results were significantly positive, and the regression results showed that the conclusions remained robust after controlling for endogeneity problems.

Table 5.

Regression results of instrumental variables.

Explained variables	Regulated industry
	$Region\times Time$	$Innovation$
	（1）The first stage	（2）The second stage
$VC$	−0.097*** （0.002）
$Region\times Time$		0.401** （0.168）
$Controls$	Yes	Yes
$YearFE$	Yes	Yes
$IndustryFE$	Yes	Yes
$Observations$		146880
$R-squared$		0.123
$Fvalue$	2618.73	（0.0000）
$Underidentificationtest$	2620.85	（0.0000）
$Weakidentificationtest$	2618.73

Data source: China Urban Statistical Yearbook, 2013–2022.

6. Path test of innovation spillover effect -- production network transmission

6.1. Network transmission model setting and correlation degree calculation

The benchmark regression indicates that the innovation level of non-regulated industries in the pilot sector improved significantly. Existing studies show that the spillover effect of innovation can be transmitted through multiple channels, such as environmental improvement, population flow, and knowledge networks [37]. To verify the transmission effect of the innovation network caused by industry correlations, the following model was constructed:

$${Innovation}_{ijt}={\beta }_1{For}_{ij}\times Time+{\beta }_2{Back}_{ij}\times Time+{\beta }_3{Controls}_{jt}+{\mu }_{ij}+{\delta }_t+{\epsilon }_{ijt}$$ (25)

In the equation, ${Innovation}_{ijt}$ is used to measure the innovation level of industry $i$ in region $j$ at $t$; ${For}_{ij}$ represents the degree of association between industry $i$ of region $j$ as the downstream industry and the regulatory industry; ${Back}_{ij}$ represents the degree of association between industry $i$ of region $j$ as the upstream industry and the regulatory industry; ${\beta }_1$ and ${\beta }_2$ represent the network effect of the policy; ${Controls}_{jt}$ represents the control variable at the regional level; ${\mu }_{ij}$ is the fixed effect at the regional industry level; ${\delta }_t$ is the fixed effect at the time; ${\epsilon }_{ijt}$ is the residual term.

This study refers to the upstream and downstream transmission mechanism of the policy established by Liu et al. (2022), in which the calculation method of the downstream transmission of the regulated industries is as follows (to simplify the explanation, the regional level is not differentiated here, and all the industries in the following are at the regional industry level):

$${For}_i=\sum _{f\ne i}\left(\frac{{input}_{if}}{\sum _f{input}_{if}}\right)\times {Regulated}_f$$ (26)

where $i$ and $f$ represent different industries; $i$ is the relative downstream industry of $f$; ${input}_{if}$ represents the total amount of products of industry $f$ used by industry $i$; $\sum _f{input}_{if}$ represents all intermediate products used in industry $i$; ${Regulated}_f$ represents the regulated situation of industry $f$; if industry $f$ is regulated, ${Regulated}_f$ = 1; and if industry $f$ is not regulated, ${Regulated}_f$ = 0 [63].

Similarly, the upstream conduction of the regulated industry is calculated as follows:

$${Back}_i=\sum _{b\ne i}\left(\frac{{output}_{ib}}{\sum _b{output}_{ib}}\right)\times {Regulated}_b$$ (27)

where $i$ is the relative upstream industry of $b$; ${output}_{ib}$ represents the total amount of industry $i$ input into industry $b$ as an intermediate product; $\sum _b{output}_{ib}$ represents the total amount of all intermediate products used in industry $i$; ${Regulated}_b$ represents the regulated situation of industry $b$. This structure is derived from the transmission of vertical integration in the global value chain. The larger the coefficient, the more finished products of industry $i$ are used in the production of regulated industry $b$, so industry $i$ is more indirectly affected by policy regulation in industry $b$ [52].

6.2. Empirical test of network conduction

To exclude the direct effects of the policies, samples of regulated industries in the pilot areas were excluded from the transmission test. The empirical results are listed in Table 6, where ${For}_{ij}\times Time$ and ${Back}_{ij}\times Time$ represent the upstream and downstream conduction effects, respectively. After the introduction of regional control variables, time fixed effects and city-industry fixed effects, the upstream conduction coefficient was negative and significant, whereas the downstream conduction coefficient was positive and passed the statistical test. The empirical results show that the carbon emission trading policy has a significant downstream transmission effect but inhibits innovation in the upstream industry, which is consistent with the theoretical hypothesis. For the upstream industry, the cost increase and production capacity conversion caused by the policy reduce demand for upstream products and ultimately increase the cost of unit innovation in the upstream industry. Compared to the downstream industry, the price increase of products in the regulated industry forces downstream enterprises to improve the utilization efficiency of intermediate products, actively explore alternatives, and improve the willingness of the downstream industry to innovate.

Table 6.

Test of network conduction effect.

Explained variables	$Innovation$	$Innovation$	$Innovation$	$Innovation$
	（1）	（2）	（3）	（4）
$Back\times Time$	−0.146*** （0.0328）	−0.755*** （0.0786）	−0.307*** （0.0371）	−0.317*** （0.0370）
$For\times Time$	1.316*** （0.0518）	1.571*** （0.102）	1.352*** （0.0568）	1.242*** （0.0579）
$Controls$		Yes	Yes	Yes
$CityFE\times IndustryFE$	Yes		Yes	Yes
$YearFE$	Yes	Yes		Yes
$Constant$	0.292*** （0.000818）	−8.040*** （0.0707）	−10.47*** （0.118）	−5.295*** （0.143）
$Observations$	1353768	1075724	1074364	1074364
$R-squared$	0.862	0.103	0.862	0.862

Data source: China Urban Statistical Yearbook, 2013–2022.

6.3. Mechanism test of production network conduction

6.3.1. Direct impact: the impact of policies on the ex-factory prices of regulated industries

The theoretical mechanism of this study shows that the carbon emissions trading pilot policy increases the production cost of the regulated industry, which leads to an increase in the price of finished products in the regulated industry. As the changes in production costs in various industries cannot be directly observed, Hypothesis 2 is further verified by observing the changes in prices. Compared to production costs, fluctuations in intermediate product prices have a more direct impact on downstream industries. To test the transmission mechanism of prices, the producer ex-factory price index (price of last year = 100) in the China Price Statistical Yearbook was selected as the explained variable to discuss whether the carbon emissions trading pilot policy increased the ex-factory price of regulated industries in the pilot areas. A difference model was constructed for the regression, based on the price index of the two-digit industries in each province from 2012 to 2021. To ensure consistency of the regression, the control variables were fixed at the provincial level, and the data were obtained from the China Statistical Yearbook and industry annual reports. The regression results are presented in Table 7. Column (1) shows that, after the implementation of the carbon emissions trading policy, the ex-factory prices of the regulated industries in the pilot areas increased. The control variables are added in column (2), and the results remain significant. The above results show that the carbon emissions trading pilot policy increased the ex-factory prices of products in regulated industries in the pilot areas. Although pilot carbon trading policies have reduced the cost of emissions reduction for enterprises through market-based mechanisms, strict control of total emissions has typical regulatory characteristics. On the one hand, the policy-forcing mechanism increases the R&D input of enterprises in the short term, and the improvement in the product process and quality is finally reflected in the price, which promotes the price of finished products. On the other hand, strict total quantity control reduces production capacity, and the resulting supply and demand inversion further promotes the price increase of products in regulated industries [64].

Table 7.

Effects of policy shocks on production prices.

Explained variables	Price index	Price index	Price index
	（1）	（2）	（3）
$Region\times Regulated\times Time$	2.301*** （0.290）	2.299*** （0.290）
$Regulated\times Time$			1.269*** （0.519）
$Controls$		Yes	Yes
$ProvinceFE\times IndustryFE$	Yes	Yes	Yes
$YearFE$	Yes	Yes	Yes
$Constant$	99.66*** （0.0923）	80.66** （32.53）	194.1* （100.9）
$Observations$	8938	8938	1762
$R-squared$	0.265	0.266	0.287

Data source: China Price Statistical Yearbook 2014–2022

6.3.2. Spillover effect: network transmission of prices

The impact of pilot carbon emission trading policies on product prices was verified in terms of their direct effects. This section examines the effects of intermediate price shocks on innovation. Forward (downstream) price volatility indicators were constructed using the 2021 inter-regional input-output table, referring to the construction method of the network correlation degree and relevant price volatility data. The specific calculation method was as follows:

$$For\_{price}_{it}=\sum _f\left[\left(\frac{{input}_{if}}{\sum _f{input}_{if}}\right)\times {PriceFluctuation}_{ft}\right]$$ (28)

In the equation, $i$ and $f$ represent different industries; $i$ is the relative downstream industry of $f$; ${input}_{if}$ represents the total amount of products of industry $f$ used by industry $i$; $\sum _f{input}_{if}$ represents all intermediate products used in industry $i$; ${PriceFluctuation}_{ft}$ represents the price fluctuation of industry $f$ in year $t$; the price fluctuation index is calculated as the ex-factory price of goods in the current period divided by the ex-factory price of goods in the previous period [17]. If the price fluctuation index is greater than 100, it means that the commodity price of the current period has increased; if it is lower than 100, it means that the commodity price of the current period has decreased; if it is equal to 100, it means that the price has not fluctuated.

Accordingly, backward (upstream) transmission of price fluctuations was constructed:

$$Back\_{price}_{it}=\sum _b\left[\left(\frac{{output}_{ib}}{\sum _b{output}_{ib}}\right)\times {PriceFluctuation}_{bt}\right]$$ (29)

Furthermore, to explore the heterogeneity of industry price changes on innovation, price fluctuations were divided into those of regulated industries in pilot areas and those of industries in other regions [16]. The upstream price fluctuation in the regulated industry is calculated as follows:

$$Rfor\_{price}_{it}=\sum _f\left[\left(\frac{{input}_{if}}{\sum _f{input}_{if}}\right)\times {PriceFluctuation}_{ft}\times {Regulated}_f\right]$$ (30)

where ${Regulated}_f$ indicates a regulated situation in industry $f$. The regulated industry is 1, and the unregulated industry is 0. The upstream price fluctuations in other industries are calculated as follows:

$$Nonrfor\_{price}_{it}=\sum _f\left[\left(\frac{{output}_{if}}{\sum _f{output}_{if}}\right)\times {PriceFluctuation}_{ft}\times non{Regulated}_f\right]$$ (31)

In the equation, $non{Regulated}_f$ indicates the unregulated situation of industry $f$. The value of the unregulated industry is 1, and the value of the regulated industry is 0. The downstream price fluctuations of the regulated and other industries are calculated in the same manner [65].

Columns (1) and (3) of Table 8 report the impact of upstream and downstream price fluctuations on industry innovation. The results show that a price increase in the upstream industry significantly improves the innovation level of the industry. Simultaneously, the rising price in the downstream industry has a significant positive impact on innovation. The results in columns (2) and (4) show that regardless of whether the core explained variable lags, the price fluctuations of industries not directly affected by policies have significantly positive impacts on upstream and downstream innovation, which is consistent with the conclusions above. However, a price rise in regulated industries has a significantly negative impact on upstream innovation and a stronger positive promoting effect on downstream innovation. For other industries, fluctuations in product prices reflect changes in market demand, and these market-induced co-directional fluctuations stimulate innovation throughout the industrial chain. For regulated industries, product price fluctuations are an increase in production costs caused by policies and an improvement in product quality caused by innovation [56]. On the one hand, this impact on production reduces the demand for upstream products and weakens the incentive for upstream innovation. On the other hand, it forces downstream enterprises to innovate and reconstruct their current industrial structures.

Table 8.

Test of price transmission mechanism.

Explained variables	Current transmission		One phase lag
	$Innovation$	$Innovation$	$Innovation$	$Innovation$
	（1）	（2）	（3）	（4）
$For\times price$	2.469*** （0.122）		2.842*** （0.176）
$Back\times price$	0.585*** （0.130）		1.661*** （0.204）
$Regfor\_price\times Time$		1.380*** （0.0615）		2.040*** （0.0729）
$Nonregfor\_price\times Time$		0.379*** （0.0109）		0.438*** （0.0125）
$Regback\_price\times Time$		−0.413*** （0.0431）		−0.662*** （0.0503）
$Nonregback\_price\times Time$		0.214*** （0.00971）		0.298*** （0.0114）
$YearFE$	Yes	Yes	Yes	Yes
$Controls$	Yes	Yes	Yes	Yes
$CityFE\times IndustryFE$	Yes	Yes	Yes	Yes
$Constant$	−6.631*** （0.174）	−5.731*** （0.161）	−7.606*** （0.229）	−6.474*** （0.208）
$Observations$	892040	892040	709144	709144
$R-squared$	0.881	0.882	0.894	0.894

Data source: China Price Statistical Yearbook 2014–2022.

6.3.3. Heterogeneity analysis of innovation transmission: price distortion

As price is an important mechanism for realizing network transmission, this study further analyzes the factors that affect the price adjustment of products in the regulated industry. According to the study of Liu et al. (2020), the price distortion of the industry can be divided into two parts: market defect $\sigma$ and policy subsidy $\tau$. Among these, the market defect $\sigma$ is derived from the price premium caused by market imperfections. The possible micro-basis includes financial friction, deadweight loss caused by transaction costs, monopoly rent caused by market power, and price premium caused by double marginalization. The policy subsidy $\tau$ comes from non-market factors affecting enterprise pricing, such as direct government subsidy, tax reduction and exemption, cheap land transfer, price limit policy (such as the upward limit of coal power market transaction price). In summary, this study analyzes industrial monopoly and the degree of marketization. On the one hand, industries with market power have stronger bargaining power and are more likely to have price markups. On the other hand, in industries with a high degree of marketization, there are fewer resource misallocation problems caused by market imperfections, and the prices of finished products are less affected by non-market factors. It should be noted that the increase in product prices in regulated industries not only reflects the increase in production costs but also reflects the improvement of product processes or quality after innovation.

Based on data on China's industrial enterprises in 2020, the Herfindahl-Hirschman Index at the two-digit industry level of the province was calculated. To verify the innovation spillover effect of different monopoly industries, the degree of correlation between the middle and downstream regulated industries was calculated as follows:

$${For\_hhi}_i=\sum _{f\ne i}\left(\frac{{input}_{if}}{\sum _f{input}_{if}}\right)\times {Regulated}_f\times {hhi}_f$$ (32)

Among these, ${hhi}_f$ reflects industry monopoly [18]. The results are summarized in Table 9. The upstream and downstream transmission effects were included as control variables in the regression. The empirical results show that the innovation promotion effect of pilot policies is significant in the downstream industries of regulated industries with high degrees of monopoly. Furthermore, HHI reflects the degree of dispersion of manufacturer scale in the market. The greater the HHI, the higher the degree of monopoly in the industry, and the stronger the bargaining power. Faced with the increase in production costs caused by policy impact, industries with a high degree of monopoly are able to transfer this cost constraint to downstream manufacturers through prices to obtain a producer surplus [35]. However, for regulated industries with low degrees of monopoly, it is more difficult to transfer the cost of emission reduction or innovation to downstream manufacturers. The cost constraints of policies and industrial price competition further compress the profit space and eventually accelerate the exit of inefficient enterprises. Supply chain restructuring disrupts the original business order of downstream enterprises and impedes innovation in downstream industries in the short term [17].

Table 9.

Test of the degree of monopoly in regulated industries.

Explained variables	$Innovation$ （1）	$Innovation$ （2）
$For\_hhi\times Time$	0.404*** （0.0250）	0.399*** （0.0261）
$Controls$	No	Yes
$YearFE$	Yes	Yes
$CityFE\times IndustryFE$	Yes	Yes
$Constant$	0.290*** （0.000847）	−5.267*** （0.143）
$Observations$	1353768	1074364
$R-squared$	0.862	0.863

Data source: China Urban Statistical Yearbook, 2013–2022.

Based on the data from China's industrial enterprises in 2020, the index of the proportion of private enterprises at the two-digit industry level of the province was calculated. The index was obtained using the proportion of private enterprises in the total number of enterprises in the two-digit industry of the province. The two-digit industry of the province with a high proportion of private enterprises was less affected by non-market factors, and the price could more effectively reflect the market supply demand relationship [45]. To verify the innovation spillover effect of industries with different proportions of civil enterprises, the correlation between downstream industries and regulated industries is calculated as follows:

$${For\_private}_i=\sum _{f\ne i}\left(\frac{{input}_{if}}{\sum _f{input}_{if}}\right)\times {Regulated}_f\times {private}_f$$ (33)

where ${private}_f$ represents the proportion of private enterprises in $f$ industry. The results are summarized in Table 10. The coefficient of the interaction term is significantly positive, indicating that industries with a higher proportion of private enterprises have a stronger transmission effect on downstream innovation. This also verifies the above point of view: the product price of industries with a high degree of marketization is more sensitive to policies, and the resulting price fluctuations promote the innovation of downstream industries.

Table 10.

Test of the proportion of private enterprises.

Explained variables	$Innovation$ （1）	$Innovation$ （2）
$For\_private\times Time$	0.263*** （0.0106）	0.344*** （0.0145）
$Controls$	No	Yes
$YearFE$	Yes	Yes
$CityFE\times IndustryFE$	Yes	Yes
$Constant$	0.271*** （0.00147）	−5.583*** （0.144）
$Observations$	1353768	1074364
$R-squared$	0.861	0.862

Data source: China Price Statistical Yearbook 2020.

7. Further analyses: industry selection and cross-market transactions

The above has verified the network innovation spillover effect of the carbon emission trading pilot policy, which means that the policy not only has a direct innovation effect, but also will be transmitted through the production network. For the expansion of the policy in the future, which industries will be the first to be included in the national trading will be more conducive to innovation? At present, the electric power industry has been included in the national carbon market trading, so will the innovation effect of cross-regional trading be more prominent? Based on this, this part will focus on the industry selection of carbon emission trading and the expansion of regional scope for analysis.

7.1. The network of innovation transmission: industry correlation

The correlation between an industry and downstream industries can be measured by the Out-degree index, whereas the correlation between an industry and all downstream industries can be measured by the point degree and center degree in a social network analysis, which can be calculated by adding the correlation strength between the industry and all downstream industries [34]. The higher the degree of centrality of an industry, the closer the input-output relationship between the industry and the downstream industry. Therefore, the transmission effect of the policy's impact on the downstream industries is stronger. The degree of centrality of the out-degree is an important standard for selecting an industry for regulation. The innovation promotion effect on the downstream industry may be stronger if the industry with the highest degree of centrality is regulated. To verify the innovation spillover effect of industries with high and low centrality, the correlation situation of downstream industries of regulated industries was calculated as follows:

$${For\_outdegree}_i=\sum _{f\ne i}\left(\frac{{input}_{if}}{\sum _f{input}_{if}}\right)\times {Regulated}_f\times {outdegree}_f$$ (34)

where $i$ is the relative downstream industry of $f$, and ${input}_{if}$ represents the total amount of products of industry $f$ used by industry $i$. $\sum _f{input}_{if}$ represents all intermediate products used in industry $i$. ${Regulated}_f$ indicates whether $f$ industry is regulated, and ${outdegree}_f$ indicates the centrality of $f$ industry [7].

The interaction between ${For\_outdegree}_i$ and the time of policy implementation was included in the regression model as the core explanatory variable. The results are presented in Table 11, where both forward and backward transmissions are included in the model as control variables. The empirical results show that the innovation promotion effect of pilot policies is greater in downstream industries with high-centrality regulations. The calculation results show that the output centrality of the finished products of basic industries such as chemical products, electric power, and metals is the highest (Table 12), the price shock of the finished products of basic industries is more likely to form an innovation backforce mechanism in the production network, and the R&D and technological innovation of the finished products of basic industries can benefit the downstream industries to the greatest extent. Therefore, the carbon emissions trading system should focus on the inclusion of basic industrial products.

Table 11.

Spillover effect of industry correlation heterogeneity.

Explained variables	$Innovation$ （1）	$Innovation$ （2）
$For\_outdegree\times Time$	0.0291*** （0.00105）	0.0169*** （0.00122）
$Controls$	No	Yes
$YearFE$	Yes	Yes
$CityFE\times IndustryFE$	Yes	Yes
$Constant$	0.296*** （0.000742）	−5.168*** （0.143）
$Observations$	1353768	1074364
$R-squared$	0.862	0.862

Data source: China Price Statistical Yearbook 2014–2022.

Table 12.

Out degree and centrality of industry.

Industry	Out degree and centrality
Chemical products	12.16704
Production and supply of electricity and heat	8.297034
Metal smelting and calendering processed goods	8.059205
Paper printing, cultural, educational and sporting goods	5.374374
Petroleum, coking products and processed nuclear fuel	3.235574
Non-metallic mineral products	2.134218

Data source: China Price Statistical Yearbook 2014–2022.

7.2. Connectivity of innovation transmission: a cross-district pilot in Beijing and chengde

Within provincial administrative divisions, considering the agglomeration and isomorphism of industries, the differences in industrial structure and technology within a province are small, whereas the differences in the industrial structure and production technology of each industry across provincial administrative regions are relatively high. Diversified capacity complementarity and innovation can provide more space for carbon emissions trading among enterprises [18,23]. To test the innovation effect of the cross-provincial administrative trading pilot, we constructed a triple-difference term with Beijing and Chengde as the experimental group to investigate the effect of the cross-provincial carbon sink trading pilot policies in Beijing, Chengde, and Hebei in 2021. $Cross\times Treat\times Time$ is the interaction term between cross-provincial administrative regions and the pilot period. The empirical results are shown in Table 13, where column (1) covers the regulated industries in the pilot areas and column (2) shows the non-regulated industries. The core explanatory variable coefficients of the two columns are significantly positive, indicating that the cross-provincial administrative region has a stronger innovation effect, and this enhancement effect is reflected not only in regulated industries but also in non-regulated industries.

Table 13.

Cross-provincial administrative region transaction test.

Explained variables	Regulated industry	Regulated industry
	$Innovation$ （1）	$Innovation$ （2）
$Cross\times Treat\times Time$	1.237*** （0.172）	0.991*** （0.0729）
$Treat\times Time$	0.0301** （0.0139）	0.112*** （0.00787）
$Controls$	Yes	Yes
$YearFE$	Yes	Yes
$CityFE\times IndustryFE$	Yes	Yes
$Constant$	−4.611*** （0.344）	−5.522*** （0.161）
$Observations$	174528	924352
$R-squared$	0.872	0.867

Data source: Beijing Carbon Sink Trading Report 2021 and Hebei Carbon Sink Trading Report 2021.

8. Discussion

8.1. Discussion of the current carbon emissions trading policy

In addition to the findings of this study, many other studies have shown the impact of an emissions trading policies on innovation. Cooke et al. (2019) studied the relationship between carbon markets and innovation in the United Kingdom and found that carbon markets have a positive impact on innovation. They found that the introduction of the carbon market promotes enterprises’ R&D investment and technological innovation, thus improving their competitiveness and innovation abilities. This suggests that carbon emissions trading policies can promote innovation by incentivizing enterprises to make R&D investments and technological innovations. Another study showed that California in the United States implemented an emissions trading policy that required companies to reduce their carbon emissions by 40% by 2020. The implementation of this policy has led some companies to invest more in clean technology and to create more jobs. These results show that carbon emissions trading policies can promote the development of clean energy and low-carbon technologies, thus driving economic transformation and innovation. In addition, several studies have shown that emissions trading policies can promote regional innovation. For example, the European Union has implemented a project called Climate Action, which aims to promote cooperation and development among European countries to combat climate change. The program includes a section called the Climate Innovation Fund, which is used to finance businesses that develop new technologies and solutions. These results show that emissions trading policies can promote innovation by providing financial support. However, some studies suggest that emissions trading policies may negatively impact certain industries. For example, one study found that the automobile manufacturing industry was hit hard by the emissions trading policy implemented in California in the United States. This is because most companies in the industry are producers of petroleum-based engines, and they need to purchase large amounts of carbon emission allowances to continue their operations. Consequently, the policy has forced these companies to find new production methods and technologies that have led to layoffs and plant closures. This shows that the special circumstances of different industries need to be considered when formulating carbon emissions trading policies to avoid excessive impacts on some industries.

8.2. Discussion of the impact of carbon market on innovation

Since the Chinese government proposed the establishment of a national carbon emission trading market in 2011, China's carbon market has undergone remarkable development. By April 2023, China had 28 pilot areas for emissions trading, covering approximately 50% of the country's total carbon emissions. In addition, China established the China National Clearing Center (CNCC) and China Environmental Protection Information System (CIES), which provide technical support and guarantee the operation of the carbon market. During the development of the carbon market, the Chinese government adopted a series of policy measures to promote enterprise participation and innovation. For example, the government encouraged enterprises to invest more in R&D to improve their technological level, supported enterprises in carrying out international cooperation, introduced advanced foreign technology and experience, strengthened the protection of intellectual property rights, and encouraged enterprises to innovate independently. These policies and measures have created favorable conditions for enterprises to participate in the carbon market and provide lessons for other countries.

Etzkowitz et al. (2018) found that China's carbon market has a positive impact on innovation. They believe that the carbon market can promote innovation in the following ways: (1) Encourage enterprises to invest in R&D and technological innovation. Introduction of the carbon market requires enterprises to reduce carbon emissions, thus encouraging them to increase R&D investment and improve their technological levels. This will help improve the competitiveness and innovation abilities of enterprises. (2) Promoting cooperation and knowledge sharing among enterprises. The operation of carbon markets requires information exchange and cooperation among enterprises. This will help enterprises establish cooperative relationships, improve their degree of knowledge sharing, and further promote their innovation ability. (3) Provide financial support for innovation. The operation of carbon markets requires large capital investments. The government provides financial support for innovation through subsidies and loans to enterprises and projects. Based on the above analysis, we can conclude that the carbon emission trading policy can promote innovation by encouraging enterprises to make R&D investment and technological innovation and promoting cooperation and knowledge sharing among enterprises. This has important reference significance for other countries coping with climate change and promoting green development.

To better play the positive role of carbon emission trading policies in promoting innovation, other countries can take the following measures. (1) Improve policy and regulatory systems. Governments should formulate and improve relevant policies and regulations, clarify the policy objectives, implementation steps, and management mechanisms of carbon emissions trading, and provide clear policy guidance for enterprises to participate in the carbon market. (2) Strengthen technology research, development, and dissemination. Governments should increase investment in the R&D of green technologies and clean energy and support enterprises to carry out technological innovation and transfer results. Simultaneously, governments should strengthen the publicity and promotion of clean energy technology and improve social understanding and acceptance of green technology. (3) Optimizing industrial structure and layout. Governments should guide the adjustment of the industrial structure, prioritize the development of low-carbon, environmental protection, and high-value-added industries, and reduce the proportion of high-carbon and high-pollution industries. In addition, governments should optimize the industrial layout and organically integrate the clean and traditional energy industries to achieve coordinated development. (4) Establish an international cooperation mechanism. Governments of all countries should strengthen international cooperation to jointly address the challenges of climate change. It is advised to promote global green development through various means such as technology exchange, financial support, and policy coordination. In summary, carbon emission trading policies play an important role in promoting innovation. Other countries can learn from China's experiences and take corresponding measures to fully utilize the positive role of the carbon market in promoting green development and innovation.

Barunik (2017) shows that carbon markets can promote innovation. He found that the introduction of carbon markets boosted enterprises' R&D investment and technological innovation, which, in turn, increased their competitiveness and innovation capacity. Additionally, they found that the introduction of the carbon market promoted cooperation and knowledge sharing among companies, further improving their innovation abilities. These results suggest that carbon emissions trading policies can promote innovation by incentivizing enterprises to make R&D investments and technological innovations, and facilitate cooperation and knowledge sharing. However, factors such as the industrial structure and policy environment of different countries may influence the impact of carbon markets on innovation. For example, some countries may pay more attention to the development of traditional energy sectors and less to the development of emerging industries. Therefore, the influence of these factors must be considered when conducting cross-country comparisons. To better understand the impact of carbon markets on innovation, we used the following data for our analysis: (1) Carbon market size. According to the International Energy Agency (IEA), global carbon emissions fell from 3.31 billion tons in 2010 to 3.25 billion tons in 2019. Simultaneously, the global carbon market is growing. For example, the European Union's carbon market is expected to reach EUR 68 billion in 2020. These figures show that, as the global carbon market continues to grow, an increasing number of businesses are starting to participate, thus opening up more opportunities for innovation. (2) Corporate research and development investment. According to the World Intellectual Property Organization (WIPO), more than 40,000 companies file patent applications annually worldwide. The United States is one of the most active patent-filing countries, with more than 10,000 companies filing patent applications each year. These figures show that, globally, an increasing number of companies are beginning to value innovation and are willing to devote more resources to it. (3) Cooperation among Enterprises. According to a McKinsey & Company report, more than half of the world's largest companies have partnered with at least one other company over the past ten years. These partnerships often involve the sharing of technology and knowledge to promote innovation.

In conclusion, carbon emission trading policies can promote innovation by incentivizing R&D investment and technological innovation and promoting cooperation and knowledge sharing among enterprises. Factors such as the industrial structure and policy environments of different countries may influence the impact of carbon markets on innovation. Therefore, the impacts of these factors must be considered when making cross-country comparisons.

9. Conclusion and suggestions

The results show that the implementation of a carbon emissions trading policy has a significant impact on various industries. First, through the test of the triple-difference benchmark and its dynamic model, it was found that the network conduction effect lags the direct promotion effect, but the network conduction effect is stronger than the direct promotion effect. This means that the carbon emissions trading policy may not immediately lead to an increase in enterprise output in the short term; however, in the long run, it will promote the upgrading of the entire industrial structure through optimization and adjustment of the production network. Second, by constructing the upstream and downstream weights of the regulated industries, we found that the carbon emissions trading policy has a stronger positive impact on the innovation of the downstream industry, while it has a negative impact on the upstream industry. In the mechanism study, it is proven that an upstream price increase promotes downstream innovation, and there is a production network correlation between innovations. This shows that the carbon emissions trading policy not only regulates the industrial structure, but also affects the competition between various industries to a certain extent. Further research shows that the innovation effect is more pronounced in the cross-regional carbon trading market. Regulatory industries with a high degree of industry correlation and a greater emphasis on prices have stronger innovation spillover effects (Table 14). This indicates that when implementing carbon emissions trading policies in different regions, the industrial characteristics and development level of each region should be fully considered to achieve the optimal policy effect.

Table 14.

Summary of empirical conclusions.

		Direct effect Regulated industries	Diffusion effect Unregulated industries
Benchmark regression	DID	Positive	–
	DDD	Positive	Positive (stronger)
Network analysis	Downstream transmission	–	Positive
	Upstream transmission	–	Negative
Mechanism test - ex-factory price	Ex-factory prices	Positive	–
	Downstream transmission	–	Positive
	Upstream transmission	–	Negative
	Strong bargaining power	–	Positive (stronger)
	Less influenced by non-market factors	–	Positive (stronger)
Further analysis - Industry selection and cross market	DID	Positive	Positive
	High out degree and centrality	–	Positive (stronger)

By analyzing the results, the following countermeasures are proposed.

From a theoretical perspective, these findings are mainly based on neoclassical economics, institutional economics, and econometrics. Neoclassical economics focuses on resource allocation and price mechanisms in a market economy, institutional economics studies the effects of institutions on economic behavior, and econometrics analyzes economic data using statistical methods. These theories provide powerful tools for understanding the implementation of carbon emissions trading policies. From a practical point of view, these findings have important reference values for formulating more effective carbon emissions trading policies. First, the government should consider the adjustment needs of industrial structures when formulating policies to achieve the sustainable development goals. Second, the government should pay attention to the impact of carbon emissions trading policies on various industries to avoid negative effects. Finally, the government should fully utilize the advantages of cross-regional carbon trading markets to promote the flow and optimal allocation of innovative resources. In China, the carbon emission trading policy has achieved remarkable results. For example, state-owned enterprises, such as Sinopec and Petrochina, have actively responded to the national policy by introducing a carbon emissions trading system, which has improved energy efficiency and reduced greenhouse gas emissions. Additionally, the scale of China's carbon market has been expanding, attracting more companies and investors to participate and making a positive contribution to the global response to climate change. In conclusion, this study shows that carbon trading policies have had a significant impact on various industries and that this result has important reference value for formulating more effective carbon trading policies.

First, when promoting the pilot work of carbon emission trading rights, it is necessary to fully consider the spillover and diffusion effect caused by the policy through the production network, and rationally plan the total volume control policy of the industry. Considering the heterogeneous impact of policies on innovation in upstream and downstream industries, reasonable carbon quota targets should be set in light of the actual situation of production network connection. Drawing on the pioneering experience of the European Union and other international carbon trading markets, innovation funds and modernization funds should be set up to facilitate collaborative innovation between regulated industries and upstream and downstream related industries in various regions, and enterprises should be encouraged to flexibly use market mechanisms to achieve the convergence of marginal emission reduction costs across industries and regions under strict emission control standards.

Second, it is necessary to steadily promote a unified national carbon emission trading market and to give priority to the inclusion of regulated industries closely related to downstream input-output and less affected by non-market factors in the pilot. At present, the carbon emission trading policies of the pilot regions differ greatly in the specific implementation path of quota allocation, penalty mechanism, carbon market regulation, etc. Regions with too high emission reduction coefficient will lead to the regulatory industry facing greater cost pressure, and the problem of supply and demand of related finished products is prominent, while too low emission reduction coefficient will lead to the downturn of the carbon trading market and the lack of innovation motivation of enterprises. Therefore, unified quota standards and cross-regional trading can form a more rational and balanced carbon market, and bring more reasonable arbitrage space and economic incentives for innovative enterprises.

Third, establish and improve relevant supporting policies to realize connectivity and coordinated development of various sectors and regions. At present, there are many obstacles in the transmission of network effects of carbon emission trading policies, among which inter-provincial trade barriers are an important reason for the poor transmission effect of production networks. Therefore, the further step is to eliminate non-economic factors that hinder the free flow of factors across regions, promote inter-provincial trade by reducing institutional transaction costs, and unblock the transmission mechanism of production networks to reduce the deadweight loss of environmental regulations as far as possible.

Funding

This work was supported by the Philosophy and Social Science Program of Zhejiang [grant number: 22NDJC343YBM] and the Major Humanities and Social Sciences Research Projects in Zhejiang higher education institutions [grant number: 2023QN167]. The funders played no role in the study design, data collection or analysis, the decision to publish, or manuscript preparation.

Author contribution statement

Junming Lai: Conceived and designed the experiments; Performed the experiments; Wrote the paper.

Yueyan Chen: Performed the experiments; Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data.

Data availability statement

Data included in article/supplementary material/referenced in article.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.