Carbon finance and low-carbon technological change: Evidence from China

Introduction

“Fighting climate change has become an imperative at the top of policy agendas around the world.” The adoption of The Paris Agreement in December 2015 has global implications, with ambitious carbon mitigation targets being expressed by countries in the Intended Nationally Determined Contributions (INDCs). These targets can be achieved by a transformative shift and massive adoptions of low-carbon technologies (LCTs).^1,2 China's INDC target is to reduce emission intensity by 60–65% by 2030 (reference: 2005). The Chinese government has pushed the development of LCT through a series of policies, but the quality is still behind that of advanced international standards. This quality can be illustrated by carbon intensity in a straightforward way, as it is measured by the amount of CO₂ emitted per unit of energy production. In 2021, China's carbon intensity was 0.26, which is higher than the world average of 0.22, indicating a high-carbon energy mix in China. ^a This means that China is strongly dependent on fossil fuels relative to low-carbon sources.

Studies of climate change policies tend to agree that the inducement mechanisms of technological innovation that work best are market-based.^3,4 However, the Chinese government's initial desire to promote the green transformation of the economy through command-and-control environmental regulation and green industrial policy is essentially a mandatory institutional change and runs counter to market mechanisms. ⁵ Command-and-control environmental regulation and green industrial policy shifted government officials’ performance orientation to “green patent quantity determinism.” ⁶ The “pro-patent” approach by the government distorts the incentives for firms to issue patents, and hence “strategic patenting” appeared with the intention of being a barrier to market competition to reduce the risk of being held up by other patent owners and to gain stronger contractual power toward competitors in cross-licensing settings.^7–9 Moreover, the innovation subsidies and tax incentives frequently used by the Chinese government make the distortion worse because of rent-seeking by firms. ¹⁰ The result is the “innovation illusion” of green technology, with explosive growth in quantity and a decline in quality.

As of now, China has initiated carbon emission trading (CET) and established a tightly coordinated low-carbon investment and financing (LCF) market. In our study, we refer to this as “carbon finance,” which is a broader concept than the definition of carbon finance given by the World Bank Carbon Finance Unit. ^b Carbon finance is internationally regarded as a new branch of environmental finance (from Wikipedia) ^c and a component of green finance in China. ^d According to a diverse and collaborative effort, the features of carbon finance can be summarized as follows: carbon constraints, emission trading, clean energy deal investment and financing, and diligence assessments of firms.^11–13 As a market-based instrument, carbon finance is highly expected to spur the development of new LCTs. ¹⁴ When regulated firms expect to face a higher price on emissions relative to other costs of production, CET provides them with an incentive to make operational changes and investments that reduce the intensity of their emissions. ¹⁵ Low-carbon investment and the financing market provide a framework of two key pillars for financing an effective low-carbon transition: (1) increasing low-carbon investment and (2) supporting the transition of carbon-intensive sectors. ^e Indeed, the driving force of carbon finance in terms of low-carbon technological change has been foreseen and articulated in developed countries, ¹⁶ but the Chinese government has not given it adequate consideration. More attention remains with respect to intuitive carbon reduction outcomes, ¹⁷ which may only be seen by enterprises as a way to improve ties with governments and earn a good social reputation, but still have the ineffectiveness of the above cost-effective mechanism. ¹⁸

Based on the national conditions of China's “innovation bubble,” this article aims to fill the gap in terms of the impact of carbon finance on the quality of low-carbon innovation and highlight the role of carbon finance in low-carbon technological change, as this role should be taken into further consideration by China's policymakers in the carbon market system design. There are now established research questions:

RQ1: Does carbon finance improve the quality of enterprises’ low-carbon innovation?

This article contributes to the literature in the following ways. (1) We obtain 2953 CO₂ emission allowance enterprises with 40490 low-carbon patent information by employing web crawler technology and manual reading from 8 provincial and municipal pilots: Shenzhen, Shanghai, Beijing, Guangdong, Tianjin, Hubei, Chongqing, and Fujian. To our knowledge, existing studies have only used listed emission allowance enterprises to examine the function of CET. Our paper is the first to collect and estimate the low-carbon patents of all emission allowance enterprises. (2) A new attempt to measure the quality of low-carbon technological innovation is offered. Citations received, which constitute the most frequently used proxy for patent quality, suffer from right truncation issues (i.e. more recent patents mechanically have less time to accumulate citations), so researchers typically award 6.5 years to a patent application to collect citations, ^{7 ,19–21} which is an important limitation for emerging research disciplines, such as LCT. In this paper, textual analysis technology is used to measure the quality of 40490 low-carbon patents from 2953 CO₂ emission allowance enterprises. (3) We evaluate the development level of eight carbon finance pilots and their respective components in terms of CET and LCF markets. Our instrumental variable (IV) approach confirms their diversified contribution to low-carbon technological change. We investigate the heterogeneity and possible channels of carbon finance in terms of their effect on low-carbon technical improvement.

The rest of the paper is organized as follows: Section “Literature review” discusses the institutional background and literature review; Section “Carbon finance: evaluation and analysis” provides an evaluation and analysis of the carbon finance, the CET, and the LCF markets; Section “Empirical design” introduces the model specification and data; Section “Empirical results” discusses the empirical results, including IV estimation, robustness checks, two constituent markets’ roles, and heterogeneity tests; Section “Possible channels” outlines possible channels of carbon finance affecting low-carbon innovation quality, i.e. stronger R&D intensity and external financing capability; and the last section provides the conclusion.

Literature review

China has been involved in the construction of the international carbon trading market since 2005, when it was mainly the Clean Development Mechanism (CDM). Carbon emissions trading was announced in 2011 during the 12th Five-Year Planning process as part of a broader strategy for enhancing green development. ²² It builds upon several regional pilot programs that were launched in 2013–2014 with seven provincial and municipal pilots: Shenzhen, Shanghai, Beijing, Guangdong, Tianjin, Hubei, Chongqing, and launched in 2016 in Fujian. The eight regions, spanning from the north to the south and the coastal to the inland regions, were at various stages of development and therefore covered different income per capita levels and diverse economic compositions. ²³ Eight high-carbon sectors are slated for inclusion in eight carbon emissions trading pilots: electricity (including power generation and power and heat cogeneration), buildings, iron and steel, nonferrous metal processing, petroleum refining, chemicals, pulp and paper, and aviation.^24–26 The pilot adopts a system design like that of the EU Emissions Trading System (ETS), i.e. emissions trading under cap-control, while accepting Certified Emission Reductions generated by domestic voluntary emission reduction projects, i.e. Chinese Certified Emission Reduction (CCER). The essence of the pilot programs was to determine which policies were most effective in incentivizing lower emissions to accumulate enough successes and failures for establishing the national ETS. This kind of microlevel policy experiment has long been proven to be effective in China before implementing an innovative, disruptive, and market-based policy.^27,28 The national carbon market, the world's largest in terms of the amount of greenhouse gas emissions covered, was launched on July 16, 2021, and this has become an even more important moment for China's policy makers to embrace the expanded use of market mechanisms. ²⁹ As of 2022, the cumulative trading volume of carbon allowances has exceeded 194 million tons, with a cumulative turnover of approximately 8.5 billion yuan. ^f However, the national carbon market is still in its infancy and only covers the electricity sector, thus needing to be deepened. ³⁰ In this paper, eight carbon trading pilots are used as our research objects to study the role of carbon finance in low-carbon technological change, hoping to provide enlightenment for China's policy makers in terms of the construction of a national carbon market.

Following Jaffe et al. ³¹ and Groba and Breitschopf, ³² we understand the process of technological change to be the cumulative economic and environmental impact of an invention as it progresses through innovation and diffusion. Environmental economics research on technological change has grown considerably, motivated by “the induced innovation hypothesis,” which proposes that the direction of technological change corresponds to the direction of change in relative prices. ³³ Recent research indicates that this theory also applies to materials that play a critical role in the global green transition and the reduction of global warming. ^{g , 34} Nevertheless, an important peculiarity of green technology innovation (a.k.a. environmentally friendly technology) is that it produces positive spillover in both the innovation and diffusion phases. ³⁵ This “double externality problem” leads to significant underinvestment from a social point of view.^36,37 In view of this, it is widely accepted that government intervention is required to correct the market. ³⁸ Environmental regulations began to be emphasized and are seen as a motivator for firms to develop new methods of production that eliminate particular types of emissions. ³⁹ “The Porter hypothesis,” which is the most representative manifestation, must be mentioned here; it was originally proposed by Porter ⁴⁰ to describe the relationship between environmental regulation and economic performance at the national level, and has since been developed by scholars, and is widely used to stimulate innovation at the enterprise level.^41,42 However, in fact, the concepts related to “The Porter hypothesis” are ambiguous, ^{h , 42} and the policy choice and design with decisive influence have been widely explored. ⁴³ These policies can generally be differentiated into market-based approaches and command-and-control measures. Market-based approaches encourage firms to innovate through market signals and incentive setting, leaving them with flexibility to choose the least costly options. Command-and-control policies, such as technology standards, leave relatively little room to maneuver, as they principally comprise explicit directives and performance standards.^3,28,44

Anderson et al.'s ⁴⁵ survey results based on the Irish EU ETS firms showed that nearly half of the firms employed new machinery or equipment, and nearly three-fourths of the firms made process or behavioral changes. Empirical research by Johnstone et al. ⁴⁶ using patent data on a panel of 25 OECD countries over 26 years found that tradable energy certificates are more likely to induce innovation in technologies that are close to being competitive with fossil fuels. Calel and Dechezleprêtre ¹⁵ provided the first comprehensive empirical estimates of the EU ETS's impact on low-carbon technological change and found that the EU ETS has increased low-carbon innovation among regulated firms by as much as 10%. China's Ministry of Environmental Protection put into place an experimental SO₂ emissions trading program in 2002. According to a study by Qi et al., ⁴⁷ the SO₂ ETS encourages more green innovation in enterprises located in SO₂ pollution-intensive industries and ETS regions than it does in businesses located in non-ETS locations or non-SO₂ pollution-intensive sectors. However, not only the quantity but also the quality of innovation should be accounted for in the context of technological change. The quality of innovation is highly positively correlated with the process of market diffusion. The intuitive understanding is that patent citations that show the diffusion of this technology are often regarded by scholars as an indicator of patent quality (economic value). ^{7 ,19,48, 49} Specifically, China's current low proportion of high-level invention patents and the relative lack of green technology patents have not yet been completely reversed. ⁵⁰ The fact that the quality of low-carbon innovation can be effectively incentivized in China's carbon market should be highlighted. Hu et al. ⁵¹ found that China's CET system stimulates more green quantity than quality. The results, however, are not totally compelling due to the crude measurement of innovation quality (the proportion of invention patents) and the absence of a systematic strategy for developing incentives for innovation quality.

In addition, for knowledge-generating firms, the “double externality problem” prevents them from capturing the full economic benefit of their investment in low-carbon innovation. ³² Moreover, traditional financial business is interest-oriented, leading to the flow of financial funds to projects with high yields and short return cycles, thus bringing about financing difficulties for firms in low-carbon innovation. ⁵² The Chinese growth is characterized by a strategy which is known as “pollute first, clean up later” over the past four decades. ^{53 , 54} The transition towards greener growth will require mobilizing large amounts of financing. ⁱ While China's financial sector has deepened rapidly in recent years, the lack of access to equity finance often constrains the expansion of green, innovative businesses. Therefore, a LCF market is needed to increase the credit limit for high-emission and high-pollution activities while offering low-interest rates to low-carbon enterprises to satisfy their financing requirements.^55,56 In this paper, the respective roles of CET and LCF markets in low-carbon technological change are estimated. More importantly, privately owned firms suffer from explicit and implicit system discrimination in capital markets compared with state-owned enterprises (SOEs). ⁵⁷ When the operating environment undergoes major changes, economic entities need to make tradeoffs between short-term benefits and long-term development and make appropriate investment horizon choices. ²⁶ High-tech enterprises have more innovation activities and are more skilled in technological means, which may effectively embed transformation deeply into their organizational structure, decision-making system, and production process. ⁵⁸ Further explorations of firm heterogeneity in terms of the effect of carbon finance on low-carbon technological change are presented in the study.

Carbon finance: evaluation and analysis

Eight CET pilots conducted extensive research in the areas of environmental binding design, market operation mechanism construction, policy and regulation system construction, etc., while according to recent studies, the environment, market, and policy of the carbon market are still in a fragmented state, and the fundamental goal of “using market mechanisms to address environmental issues” has not yet been fully achieved.^59,60 However, the construction and improvement of the carbon market must adhere to the progressive improvement approach. When compared to command-and-control environmental policies, the carbon market is highly anticipated to accelerate low-carbon technological change and solve environmental issues. In this section, we construct a carbon finance development index and evaluate the regional differences presented to understand the process of carbon financial marketization of the eight pilots and, subsequently, to study the relationship between carbon finance and low-carbon technology change.

According to Labatt and White, ¹¹ “carbon finance explores the financial implications of living in a carbon-constrained world, in which emissions of carbon dioxide carry a price.” Garcia and Roberts ¹² summarized several different features of carbon finance, which provided inspiration for our work to construct the carbon finance index (CFI). The first feature they considered involved a market that trades carbon allowances. Second, it relates to the investment and financing of clean energy deals. Finally, it lends a new thematic focus to due diligence assessments of firms (i.e. one that will lead to a better understanding of how firms have positioned themselves for success in a clean energy economy, or more troubling, how they have failed to do so). In this research, diligence assessments by enterprises are seen as a success in low-carbon technological change. Based on the characteristics of carbon finance, we construct the development index of CET and LCF markets, respectively. Carbon allowance turnover, trading price, and trading day are key indicators of China's CET,^61,62 and the variation in them is immense across eight pilots. ⁶³ Making penetrating analysis of carbon allowance turnover, trading price, and trading day allows us to identify the structure, volatility, and efficiency of the CET. ⁶⁴ In particular, the Chinese government pursues goal of a high compliance rate near to 100%. ⁶⁵ The annual compliance rate of allowances for pilots announced by the Local Development and Reform Commission provides concrete evidence of market efficiency. In addition to carbon emission rights, the fundamental trading product, there are numerous varieties of carbon finance derivative instruments, including carbon forwards, futures, options, and swaps. ⁶⁶ Despite the light trading volume of the above derivatives, the variety of their types does offer us proof of market financialization. As the meaning of low-carbon is not fully specified in China, the term “green low-carbon” is really viewed as a general concept (see, for instance, the State Council's announcement of the “green and low-carbon circular economic development system”). Some scholars have constructed comprehensive index systems to measure the level of green finance development.^67,68 Key components of their system include the level of green credit and green bonds as well as the status of green enterprises. In light of Chen et al.'s research, ⁶⁰ financial service efficiency has also been included in the LCF index. Furthermore, the Chinese government is an important promoter of the development of the low-carbon finance and investment markets, ⁶⁶ which might’ve produced an impact on the low carbon index. ⁶⁹ As a result, we finally incorporate government oversight and support for environmental preservation into the LCF index.

The evaluation of carbon finance

As shown in Table 1, the evaluation index system for carbon finance consists of two evaluation objectives: CET and LCF. Among these objectives, CET includes four evaluation levels: (1) Market size, as determined by transactions in the Carbon Allowance Market and CCER Market, is a clear indicator of the capacity for resource allocation. (2) Market structure. The percentage of trading turnover in the top 20% of the daily trading turnover of the performance year serves as a proxy for market structure, denoting the market transaction structure and representing the activity of transactions. (3) Market volatility, reflected by trading price volatility, is equal to “1／Standard deviation of average daily trading price.” (4) Market efficiency, which refers to the efficiency of resource allocation, includes three factors: compliance rate, sensitivity, and financialization. Sensitivity is measured by the proportion of trading days during the performance period. The trading and price data we use for the Carbon Allowance Market are obtained from the Wind dataset, the CCER turnover is obtained from the exposure data of the China Hubei Emission Exchange and the list of CO₂ emission allowance enterprises, as well as the compliance rate; and the types of carbon finance derivatives are derived from the public data of the National Development and Reform Commission of each pilot region.

OPEN IN VIEWER

Table 1.

Evaluation index system of carbon finance.

Objective	Level	Factor	Indicator	Direction
Carbon emission trading (CET)	Market size	Allowance trading	Allowances turnover (ton)/Counts of CO2 emission allowances enterprises	Direct
			Allowances turnover (yuan)/Counts of CO2 emission allowances enterprises	Direct
		Voluntary trading (CCER)	CCER turnover (ton)/Counts of CO2 emission allowances enterprises	Direct
	Market structure	Concentration Ratio	Allowances turnover (ton) concentration ratio	Inverse
			Allowances turnover (yuan) concentration ratio	Inverse
			CCER turnover (ton) concentration ratio	Inverse
	Market volatility	Price	1/Standard deviation of average daily trading price	Direct
	Market efficiency	Compliance rate	Compliance rate of CO2 emission allowances enterprises	Direct
		Sensitivity	Effective trading day ratio	Direct
		Financialization	Types of carbon finance derivatives	Direct
Low-carbon investment and financing (LCF)	Market depth	Direct and indirect financing scale	IPO and refinancing scale of green listed enterprises/Counts of CO2 emission allowances enterprises	Direct
			Green bonds balance/Counts of CO2 emission allowances enterprises	Direct
			Green credit balance/Counts of CO2 emission allowances enterprises	Direct
	Service efficiency	Financial service cost	RMB-weighted average lending rate	Inverse
		Financial service availability	Total number of financial institutions/Counts of CO2 emission allowances enterprises	Direct
	Government management	Support	Local Governments Expenditure, proportion of environmental protection	Direct
		Supervision	Pollution Information Transparency Index (PITI)	Direct

Market depth, service efficiency, and government management are the three levels of evaluation that make up LCF. (1) Market depth is a measure of how much enterprises invest in finance efforts to reduce carbon emissions; the greater the level of engagement, the deeper the LCF market depth. The ratio of the size of financial instruments (stocks, bonds, and credit) associated with the low-carbon economy to the counts of CO₂ emission allowance enterprises represents the average financing size of enterprises after the effect of quantity has been removed. (2) There are two components in terms of financial service efficiency. The annual RMB-weighted average loan rate is a measure of financial service cost, which represents the opportunity cost of low-carbon investment by financial institutions like commercial banks. Financial service availability represents the situation in which CO₂ emission allowance enterprises may acquire investment and financing services; the more financial services that are available, the more investment and financing services that a unit enterprise can obtain. (3) Government management consists of the factors of support and supervision, measured by the proportion of environmental protection local government expenditure and the Pollution Information Transparency Index (PITI), respectively. We obtain direct financing and green bond data from the China Stock Market & Accounting Research Database (CSMAR), and green credit data is collected by the local banking social responsibility reports, statistical information of the local office of the Banking and Insurance Regulatory Commission, etc. The Service efficiency data are obtained from the Regional Financial Operations Report 2014–2021. The data on local government expenditure come from the National Bureau of Statistics, and the PITI data are obtained from the Institute of Public and Environmental Affairs (IPE).

Our research sample period spans the years 2014 to 2021 and includes data from eight provincial and municipal pilots: Shenzhen, Shanghai, Beijing, Guangdong, Tianjin, Hubei, Chongqing, and Fujian. Following Chen et al., ⁶⁰ the indicators in Table 1 are weighted using the coefficient of variation technique, and the measurement procedures are as follows:

As seen in Equations (1) and (2), the indicators are first standardized to control the dimensional inconsistencies.

X i j t + = x i j t − m i j t M i j t − m i j t

(1)

X i j t − = M i j t − x i j t M i j t − m i j t

(2)

where i denotes the evaluation indicator of j pilots in year t, and X i j t + and X i j t − refer to the dimensionless approaches for direct and inverse indicators, respectively. The initial value is represented by x i j t , and the maximum and minimum values of the indicator in each pilot in year t are represented by M i j t and m i j t , respectively.

ω i t = V i j t ∑ i V i j t

(3)

V i j t = σ i j t X ¯ i j t

(4)

CFI n j t = 1 − ∑ i ( 1 − X i j t ) 2 ω i t 2 ∑ i ω i t 2

(5)

Equations (3) and (4) present a weighting method, with ω i t standing for the weight of indicator i in year t, V i j t is the coefficient of variation of the indicator, and σ i j t , X ¯ i j t represent the standard deviation and mean of dimensionless indicators, respectively. Then, Equation (5) is used to hierarchically determine the weighted index of the evaluation factor n, which is a further dimensionless indicator and is weighted to hierarchically determine the weighted index of the evaluation level v. By that analogy, we can gradually obtain the index of CET and LCF and end up with the CFI. The range of CET, LCF, and CFI values is [0, 1], with higher values indicating a better developed carbon finance system.

The analysis of carbon finance

The evaluation results of CET, LCF, and CFI of the eight pilots in 2014–2021 are shown in Table 2, and they eventually lead us to the following conclusions. First, the average CFI, CET, and LCF in the eight pilots between 2014 and 2021 ranged between 0.633 and 0.898, 0.596 to 0.890, and 0.713 to 0.947, respectively, demonstrating a moderate level of carbon finance development. Overall, LFC is larger than CET because China's CET has only been operating as a trial since 2013, whereas the investment and financing industry for the low-carbon economy first emerged in 1990, and after years of operation, it has developed a stable scale and trading volume with a higher market maturity. ⁶⁰ Second, from 2014 to 2017, CFI, CET, and LCF all showed an upward trend, suggesting a steadily improving market mechanism and carbon finance development level. However, since 2018, the values of CFI, CET, and LCF have started to fluctuate. One explanation for this may include that China started to set up a national carbon market in 2018, which perhaps influenced the initial regional carbon market. The traits and distinctions of eight pilots are our third point of emphasis. The eastern pilots’ overall level of development is quite high, as evidenced by the fact that the average value of the Shanghai CFI, which came in first among the eight pilots, was followed by those of Fujian and Guangdong (0.857 and 0.851, respectively). These features are a result of the more advanced financial market infrastructure and friendly business environment in the eastern area. Having an average CFI of 0.633, the western area, represented by Chongqing, has the lowest CFI compared to the eastern region, with Beijing (0.824) and Hubei (0.8) in the central region, being compatible with China's current interregional economic, financial, and social development. The highest mean CET, 0.89, is found in Hubei. As a significant industrial province in the central region, the enormous industrial base supports the large turnover of carbon trading, which is further supported by more than 90% of the actual trading days. This results in an efficient allocation of carbon trading resources. With LCF values exceeding 0.9 from 2014 to 2021, Beijing has the highest level of low-carbon investment and finance development. The principal benefits are as follows: First, the green finance scale has consistently come first among all pilots, indicating the highest market depth; second, there is a high availability of financial services from numerous financial institutions, making it easy for CO₂ emission allowance enterprises to look for financial support to lower their carbon emissions; third, with the lowest RMB weighted average loan rate, implying the lowest financial service cost, financial institutions are more active in the carbon finance market.

OPEN IN VIEWER

Table 2.

Evaluation results of CET, LCF, and CFI of eight pilots in 2014–2021.

	2014	2015	2016	2017	2018	2019	2020	2021	Mean
CFI
Shenzhen	0.678	0.617	0.718	0.720	0.708	0.617	0.769	0.521	0.668
Shanghai	0.891	0.669	0.984	0.957	0.966	0.867	0.881	0.968	0.898
Beijing	0.905	0.635	0.963	0.985	0.771	0.758	0.911	0.667	0.824
Guangdong	0.582	0.784	0.860	0.922	0.974	0.894	0.893	0.900	0.851
Tianjin	0.818	0.764	0.600	0.894	0.812	0.942	0.673	0.847	0.794
Hubei	0.859	0.954	0.779	0.769	0.808	0.678	0.764	0.793	0.800
Chongqing	0.569	0.500	0.634	0.500	0.668	0.654	0.769	0.765	0.633
Fujian				0.797	0.951	0.767	0.920	0.850	0.857
CET
Shenzhen	0.634	0.737	0.754	0.766	0.656	0.634	0.616	0.636	0.679
Shanghai	0.805	0.700	0.841	0.799	0.826	0.747	0.873	0.997	0.823
Beijing	0.705	0.686	0.762	0.878	0.648	0.606	0.715	0.624	0.703
Guangdong	0.628	0.778	0.992	0.942	0.879	0.710	0.903	0.792	0.828
Tianjin	0.693	0.764	0.762	0.753	0.920	0.915	0.769	0.795	0.797
Hubei	0.999	0.997	0.952	0.997	0.743	0.690	0.972	0.769	0.890
Chongqing	0.538	0.684	0.549	0.582	0.603	0.567	0.561	0.687	0.596
Fujian				0.743	0.825	0.626	0.764	0.794	0.750
LCF
Shenzhen	0.726	0.645	0.659	0.748	0.754	0.696	0.809	0.670	0.713
Shanghai	0.829	0.957	0.919	0.962	0.888	0.790	0.764	0.845	0.869
Beijing	0.994	0.914	0.910	0.985	0.944	0.918	0.948	0.963	0.947
Guangdong	0.664	0.792	0.721	0.848	0.845	0.920	0.768	0.980	0.817
Tianjin	0.814	0.801	0.600	0.879	0.645	0.794	0.652	0.800	0.748
Hubei	0.754	0.769	0.671	0.745	0.732	0.689	0.673	0.762	0.724
Chongqing	0.711	0.641	0.666	0.701	0.845	0.815	0.922	0.922	0.778
Fujian				0.800	0.843	0.844	0.864	0.807	0.832

Empirical design

Empirical model

To formally test the effect of carbon finance on low-carbon technological change, following Hu and Jefferson, ⁷⁰ we begin with our baseline estimation of Equation (6):

Lpatent _ qualit y i t = α 1 + β 1 CFI k t + λ 1 X + γ year + μ firm + ε i t

(6)

where i, k and t indexes the firms, provincial and municipal, years, respectively; CFI k t measured in Section “Carbon finance: evaluation and analysis” stands for the level of carbon finance; Lpatent _ qualit y i t denotes the quality of low-carbon patents measured by textual analysis technology. X i t represents a vector of city-level control variables. The shareholding proportion of the top shareholder (top) and total executives (ln exes) are proxy variables of the corporate governance structure and the principal force in making innovation decisions. It has been found that enterprise scale is one of the most important factors influencing corporate patenting activity. ⁷¹ Total employees (ln people) and registered capital (cap) are controlled. Considering the time invariance of registered capital (cap), we use (cap*year) to make it time-variant. The number of years that the enterprise has existed (ln age), implying the development stage of the enterprise, is also controlled, and thus it is closely related to the innovation strategy. Finally, industry factors, the degree of industry competition, and industrial policy will affect the motivation of enterprises to apply for patent applications and their ability to obtain patents. The dummy variable of whether it is supported by industrial policy (policy) is also controlled. γ year and μ firm are the fixed time and firm effects, respectively. ε i t is an error item.

Data

To construct our sample, first, 3716 CO₂ emission allowance institutions were collected and counted manually from eight provincial and municipal pilots: Shenzhen, Shanghai, Beijing, Guangdong, Tianjin, Hubei, Chongqing, and Fujian. After excluding public institutions, cancellation and revocation enterprises, financial, hotel, and other service enterprises, wholesale enterprises, and retail enterprises, we identified 2953 effective CO₂ emission allowance enterprises. With manual reading and web crawler technology, we obtain the patent information of 2953 CO₂ emission allowance enterprises from Tianyancha (www.tianyancha.com), an official filing online commercial query website based on government open data and the computer language learning ability of graph database technology. A total of 313114 patents of 2953 enterprises from 2014 to 2021 were obtained from Tianyancha, including basic information such as invention title, type, legal status, application number, and publication number, as well as detailed information such as International Patent Classification (IPC) codes and abstracts. The “IPC Green Inventory,” developed by the IPC Committee of Experts, facilitates searches for patent information relating to Environmentally Sound Technologies (ESTs), as listed by the United Nations Framework Convention on Climate Change. ESTs are currently scattered widely across IPCs in numerous technical fields. ^j LCTs are not clearly defined in China, and the Chinese government usually treats green and LCTs as the same category. In this paper, we follow the “IPC Green Inventory” and identify 40490 patents from 313114 pieces of patent information for CO₂ emission allowance enterprises. Existing papers have investigated the impact of carbon finance on low-carbon innovation based on listed CO₂ emission allowance enterprises.^{22,27,51, 72} Our paper was the first to employ all CO₂ emission allowances enterprises to investigate the relationship between China's carbon finance and low-carbon innovation quality, as listed enterprises comprise a very small portion of all allowance enterprises. Considering the lack of channel variables for unlisted enterprises, we still regard listed enterprises with CO₂ emission allowances as the primary sample. In the robustness checks, the whole sample of 2953 enterprises with CO₂ emission allowance is examined. Table 3 shows the correspondence between our sample screening, geographical location and low carbon innovation. The Wind database is the source of the financial information for listed enterprises. Financial data for unlisted enterprises is still obtained from Tianyancha using manual scanning and web crawler technology.

OPEN IN VIEWER

Table 3.

Correspondence between sample screening, geographical location, and low carbon innovation.

	CO2 emission allowance institutions	Effective CO2 emission allowance enterprises	Effective CO2 emission allowance enterprises with low-carbon innovation	Listed CO2 emission allowance enterprises	Listed CO2 emission allowance enterprises with low-carbon innovation
Shenzhen	866	807	279	71	59
Shanghai	459	402	180	28	21
Beijing	959	350	143	26	16
Guangdong	285	274	111	5	3
Tianjin	272	258	95	8	5
Hubei	372	366	183	18	14
Chongqing	207	200	71	8	4
Fujian	296	296	74	5	3
Total	3716	2953	1136	169	125

Variables and definition

Low-carbon innovation quality

The output of innovation development can be accurately measured with patent data. The number of patents is generally used as a quantitative indicator of technological innovation, while patents vary enormously in their importance or value, and hence, simple patent counts cannot be informative about innovative output. ¹⁹ Citations received have traditionally been used as measures of the importance or value of patents. ^{7 ,20,49, 73} However, we cannot overlook that citations suffer from right truncation issues (i.e. more recent patents mechanically have less time to accumulate citations). ²⁰ Thus, the sample must typically be limited after 6.5 years to collect patent citations, ²¹ which is an important limitation for emerging research disciplines; specifically, China's low-carbon patent filing activity has exploded in recent years. Against this background, we attempt to use text analysis based on machine learning to measure the low-carbon technical complexity of patents (LCTcpx) as a proxy variable for the quality of low-carbon innovation. First, text preprocessing. We use the word segmentation technique to analyze all the abstracts in our collection of low-carbon patents because an abstract is a succinct and condensed version of the patent title, technology purpose, and scope. The “Jieba module” is called when all the abstracts of patents have been imported into Python. The “Baidu Stop Vocabulary Table” is then used to remove words such as title number, interjection, and conjunction that lack a practical analytical meaning. Second, low-carbon technical complexity calculation. We only measure nouns in the abstract—not all the words are counted—because nouns are considered more significant technological aspects. ¹⁰ The length of the patent abstracts is another factor. The computation of LCT complexity varies depending on how long the patent abstract is and how many LCT nouns may be included. According to Li et al.'s (2020) research, the percentage of LCT nouns in the overall word frequency of patent abstracts is utilized to measure low-carbon technological complexity to solve this issue. Finally, firm-year-level variable generation. Because the patent application can more accurately reflect a company's current innovation plan, the low-carbon technical complexity of the patent is aggregated to the enterprise level according to the “enterprise-application year.” We employ the median (LCTcpx_median) and mean (LCTcpx_mean) aggregation approaches, which are based on the studies of Zhang and Zheng ⁷⁴ and Tao et al. ⁶ The fact that only utility models and invention patents are covered by this study needs to be emphasized. The “IPC Green Inventory” cannot be used to identify design patent classification numbers because they are fundamentally different from invention and utility model patent classification numbers.

Descriptive statistics

The summary statistics of all variables based on the listed enterprise sample and full sample, respectively, are provided for readability and clarity (as shown in Table 4). Panel A relates to our key dependent variable, firm-level low-carbon innovation quality, which suggests that the mean of firms’ low-carbon patent quality, LCTcpx_median, LCTcpx_mean, is 0.523, with standard deviations of 0.083 and 0.079, respectively. LCTknowidth_median and LCTknowidth_mean measured by Zhang and Zheng's ⁷⁴ knowledge width approach are used in the robustness check, and the corresponding mean values are 0.359 and 0.361, with standard deviations of 0.268 and 0.234, respectively. Panel B1 refers to the controlling variables. Panel B2 relates to additional controls for robustness checks of the listed sample, as shown in 5.3.4. Panels C and D describe channel variables and IV, respectively. The descriptive statistics of the dependent variables, controls, and instrument variable of listed enterprises clearly show a close resemblance to the full sample results, suggesting that the sample of listed enterprises is rather representative.

OPEN IN VIEWER

Table 4.

Summary statistics.

	Listed sample					Full sample
Variable	Obs	Mean	StDev	Min	Max	Obs	Mean	StDev	Min	Max
Panel A dependent variables
LCTcpx_median	563	0.523	0.083	0.272	0.747	3579	0.514	0.086	0.174	0.841
LCTcpx_mean	563	0.523	0.079	0.278	0.747	3579	0.513	0.083	0.174	0.841
LCTknowidth_median	563	0.359	0.268	0	0.864	3579	0.316	0.271	0	0.875
LCTknowidth_mean	563	0.361	0.234	0	0.864	3579	0.316	0.239	0	0.875
Panel B1 controls
top	1344	1.041	5.689	0.063	62.99	4298	0.880	3.188	0.060	62.99
ln exes	1344	1.980	0.447	0	3.611	4298	1.851	0.608	0	4.357
ln people	1310	8.223	1.342	5.106	12.57	3994	7.186	1.614	0.693	12.57
cap*year	1352	4.608	9.442	0.129	77.70	4303	6.335	25.07	0.002	503.2
ln age	1344	2.932	0.357	0.693	3.738	4363	2.748	0.558	0	4.174
policy	1352	0.373	0.484	0	1	4371	0.277	0.447	0	1
Panel B2 additional controls
HHI	1314	2.512	5.912	0.002	65.73	–	–	–	–	–
ROA	1314	2.314	4.254	0.018	35.25	–	–	–	–	–
leverage	1303	0.069	0.076	−0.269	0.400	–	–	–	–	–
boardsize	1309	0.477	0.198	0.059	0.959	–	–	–	–	–
duality	1352	7.907	3.084	0	17	–	–	–	–	–
Panel C channel variables
RDpeople	944	0.139	0.484	0	0.902	–	–	–	–	–
RDexpand	1152	0.265	0.218	0	0.187	–	–	–	–	–
KZ	1075	1.253	2.084	−0.927	10.36	–	–	–	–	–
Panel D instrument variable
Thermal inversions	1352	76.375	53.461	19	194	4371	87.15	53.051	19	194

Empirical results

Basic results

The baseline results are reported in Table 5. In every regression, we control for all fixed effects (year, firm, and province). We find that in all regressions, carbon finance at eight provincial and municipal pilots has a significant and positive effect on the low-carbon innovation quality of CO₂ emission allowance enterprises. The effect is statistically and economically significant. Specifically, from the full Equation (6) estimation in columns (1) and (4), an increase of one unit in carbon finance is associated with a 14.26 (13.13) percent change in low-carbon innovation quality. This is again proof of the effect of the development of carbon finance on low-carbon technological change, in line with Calel and Dechezleprêtre's, ¹⁵ discovery that the EU ETS has increased low-carbon innovation among regulated firms by as much as 10%. The regressions of low-carbon invention patent quality are shown in columns (2) and (5), and the utility model patents are shown in columns (3) and (6). The coefficient β 1 , on the CFI of utility model patents in columns (3) and (6) is greater than that of invention patents, indicating that carbon finance may have a stronger incentive effect on utility model patent quality. For control variables, top is significantly positively related to low-carbon innovation quality, implying a successful regulatory role. Noteworthy is the negative sign of policy, which has been confirmed in that the implementation of Chinese industrial policy has induced enterprises to innovate for “support seeking,” thus reducing “strategic innovation” instead of “substantial innovation”. ^{74 , 75} Overall, the coefficient of interest β 1 is statistically significant, indicating that a change in LCT might be sparked by more market-based carbon finance.

OPEN IN VIEWER

Table 5.

Baseline regressions: carbon finance and low-carbon innovation quality.

	LCTcpx_median			LCTcpx_mean
	Application (1)	Invention (2)	Utility Model (3)	Application (4)	Invention (5)	Utility Model (6)
CFI	0.1426***	0.1112*	0.1516*	0.1313***	0.1074*	0.1590**
	(3.94)	(1.83)	(2.15)	(3.18)	(1.84)	(2.18)
top	0.0423***	0.1289**	0.0163***	0.0427***	0.1390***	0.0187***
	(5.98)	(2.94)	(3.68)	(4.66)	(3.45)	(4.35)
ln exes	−0.0153**	−0.0031	−0.0041	−0.0149**	0.0044	−0.0033
	(−2.61)	(−0.38)	(−0.30)	(−2.51)	(0.62)	(−0.24)
ln age	0.0776**	0.1291***	−0.1798*	0.0572*	0.1088***	−0.2069*
	(2.61)	(6.83)	(−2.02)	(2.01)	(5.22)	(−2.13)
ln people	0.0014	−0.0140	0.0053	0.0021	−0.0166	0.0017
	(0.16)	(−0.97)	(0.63)	(0.26)	(−1.09)	(0.14)
reass*year	−0.2393	0.1119	0.4465	−0.2890	0.2136	0.5866
	(−1.06)	(0.50)	(0.67)	(−1.35)	(0.88)	(0.86)
policy	−0.0219	−0.0767**	−0.0750	−0.0343	−0.0895**	−0.0824
	(−0.77)	(−2.37)	(−0.93)	(−1.67)	(−2.82)	(−1.15)
Constant	1.7453	−0.7673	−0.9647	2.1406	−1.4997	−1.4461
	(1.22)	(−0.45)	(−0.33)	(1.58)	(−0.83)	(−0.49)
Year FE	YES	YES	YES	YES	YES	YES
Firm FE	YES	YES	YES	YES	YES	YES
Province FE	YES	YES	YES	YES	YES	YES
Observations	541	399	351	541	399	351
Adj R-squared	0.2554	0.2827	0.1110	0.2558	0.2844	0.1169

Robust standard errors cluster in industry.

***p < 0.01.

**p < 0.05.

*p < 0.1.

Endogeneity and instrumental variable estimation

The endogeneity problem could arise from classical measurement error and missing variables. To mitigate this potential endogeneity problem, we instrument the possible endogenous variable, the city-level CFI index, with a meteorological phenomenon: the annual number of days with at least one thermal inversion for each city. In general, the temperature of the atmosphere decreases with altitude. This atmospheric knot is conducive to pollutant rise and discharge, thereby reducing the degree of air pollution. However, sometimes there is a mass of hot air on top of a mass of cold air; this is called a thermal inversion. Under the “lid” effect of hot air, pollutants are trapped beneath the inversion layer. ⁷⁶ The accumulation of pollutants is bound to exacerbate air pollution. Thermal inversion has been widely applied to estimate the effects of air pollution on various outcomes.^76–79 From the perspective of correlation, when the levels of pollution in the atmosphere increase, local governments tend to have stricter ETSs that include more CO₂ emission allowance enterprises and stricter carbon emission caps. The typical conditions under which thermal inversion can occur are radiation inversion, advection inversion, and subsidence inversion, and these do not directly act on the green innovation level of the enterprise. The two-stage IV model is expressed as shown in Equations (7) and (8) below:

CFI k t = α 2 + β 2 Thermal _ inversion s k t + λ 2 X + γ year + μ firm + ε i t

(7)

Lpatent _ qualit y i t = α 3 + β 3 CFI k t ^ + λ 3 X + γ year + μ firm + ε i t

(8)

In the aforementioned equations, Thermal _ inversion s k t is the number of thermal inversion days in firm i ′ s provincial and municipal in year t, and CFI k t ^ represents the predictive value of CFI k t . The thermal inversion data were obtained from MERRA-2, ^{i , j} reporting air temperatures every 6 h at 42 vertical layers from 110 m. To be robust, the existence of a thermal inversion is defined as the temperature of the third layer being higher than that of the first layer.

We report the above 2SLS estimations in Table 6. The results of the first stage show that our IV, thermal inversions, is significantly positively correlated with it at the 1% level, regardless of whether the explained variables are CFI, CET, or LCF. The Wald F-statistics are above the standard threshold of 10 suggested by Staiger and Stock, ⁸⁰ implying that our instruments are not weak and satisfy the correlation assumption. Controlling for endogeneity, we find that the 2SLS estimate of the coefficient of CFI is 0.3397 (0.3180), which is statistically significant at the 10% level, positive, and greater than those in Table 5.

OPEN IN VIEWER

Table 6.

Carbon finance and low-carbon innovation quality: 2SLS estimates.

	LCTcpx_median(1)	LCTcpx_mean(2)	LCTcpx_median(3)	LCTcpx_mean(4)	LCTcpx_median(5)	LCTcpx_mean(6)
CFI	0.3397*	0.3180*
	(2.08)	(1.83)
CET			1.0967**	1.0268*
			(2.19)	(1.91)
LCF					0.6116*	0.5726
					(1.78)	(1.61)
top	0.0396***	0.0401***	0.0242	0.0257	0.0512***	0.0510***
	(4.14)	(3.48)	(1.04)	(1.03)	(3.82)	(3.43)
ln exes	−0.0175**	−0.0170**	−0.0382*	−0.0364	−0.0130**	−0.0127**
	(−2.68)	(−2.55)	(−1.78)	(−1.69)	(−2.16)	(−2.14)
ln age	0.0604*	0.0409	0.0449	0.0264	0.0503	0.0315
	(1.90)	(1.42)	(0.86)	(0.54)	(0.62)	(0.41)
ln people	0.0008	0.0015	0.0036	0.0041	0.0046	0.0051
	(0.09)	(0.18)	(0.17)	(0.21)	(0.46)	(0.56)
reass*year	−0.3540	−0.3976	−1.6553	−1.6159	0.2157	0.1358
	(−1.34)	(−1.58)	(−1.73)	(−1.65)	(0.81)	(0.51)
policy	−0.0165	−0.0291	0.0171	0.0023	−0.0306	−0.0424*
	(−0.56)	(−1.28)	(0.49)	(0.08)	(−1.07)	(−1.82)
First stage
	Dependent variable: CFI		Dependent variable: CET		Dependent variable: LCF
Thermal inversions	0.0034***	0.0034***	0.0011***	0.0011***	0.0019***	0.0019***
	(7.71)	(7.71)	(3.56)	(3.56)	(5.34)	(5.34)
F statistics	59.457	59.457	12.657	12.657	28.496	28.596
Year FE	YES	YES	YES	YES	YES	YES
Firm FE	YES	YES	YES	YES	YES	YES
Province FE	YES	YES	YES	YES	YES	YES
Observations	541	541	541	541	541	541

Robust standard errors cluster in industry.

***p < 0.01.

**p < 0.05.

*p < 0.1.

Robustness checks

In this section, we conduct the following three sets of robustness checks of the impact of carbon finance at eight provincial and municipal pilots on CO₂ emission allowances enterprises’ low-carbon innovation quality. The results are all reported in Tables 7–11.

OPEN IN VIEWER

Table 7.

Robustness check: full sample retest.

	LCTcpx_median				LCTcpx_mean
	Application(1)	Invention(2)	Utility Model(3)		Application(4)	Invention(5)	Utility Model(6)
CFI	0.0587**	0.0498**	0.0691*		0.0599**	0.0524**	0.0686*
	(2.36)	(2.40)	(1.90)	(2.35)		(2.39)	(1.95)
top	0.0287***	0.0690**	0.0025	0.0284**		0.0697**	0.0025
	(3.57)	(2.15)	(0.25)	(2.75)		(2.24)	(0.25)
ln exes	−0.0034	−0.0015	0.0189	−0.0033		0.0046	0.0192
	(−0.28)	(−0.09)	(1.70)	(−0.31)		(0.34)	(1.66)
ln age	−0.0169	−0.0207	−0.0023	−0.0153		−0.0107	0.0016
	(−1.13)	(−1.64)	(−0.10)	(−1.07)		(−0.85)	(0.07)
ln people	0.0054	0.0038	−0.0006	0.0033		0.0012	−0.0016
	(1.27)	(0.86)	(−0.07)	(0.87)		(0.28)	(−0.20)
reass*year	0.0084***	0.0112***	−0.0018	0.0079***		0.0083***	−0.0010
	(3.10)	(5.12)	(−0.23)	(3.09)		(3.79)	(−0.12)
policy	−0.0141	−0.0038	−0.0519	−0.0181		−0.0157	−0.0474
	(−0.71)	(−0.21)	(−1.31)	(−0.87)		(−0.72)	(−1.11)
Constant	0.4008***	0.3515***	0.4516***	0.4155***		0.3634***	0.4414***
	(4.62)	(6.06)	(3.64)	(4.90)		(6.78)	(3.39)
Year FE	YES	YES	YES	YES		YES	YES
Firm FE	YES	YES	YES	YES		YES	YES
Province FE	YES	YES	YES	YES		YES	YES
Observations	2839	1790	1935	2839		1790	1935
Adj R-squared	0.1932	0.1920	0.1588	0.2019		0.2067	0.1625

Robust standard errors cluster in industry.

***p < 0.01.

**p < 0.05.

*p < 0.1.

OPEN IN VIEWER

Table 8.

Robustness check: full sample retest (2SLS estimates).

	LCTcpx_median		LCTcpx_mean
	(1)	(2)	(3)	(4)
CFI	0.1218**	0.1246*	0.1187*	0.1185*
	(2.04)	(2.04)	(1.97)	(1.95)
top		0.0275***		0.0273**
		(3.73)		(2.79)
ln exes		−0.0047		−0.0045
		(−0.41)		(−0.45)
ln age		−0.0179		−0.0162
		(−1.09)		(−1.04)
ln people		0.0051		0.0031
		(1.21)		(0.80)
reass*year		0.0085***		0.0080***
		(2.99)		(2.99)
policy		−0.0125		−0.0168
		(−0.68)		(−0.86)
First stage	The dependent variable is CFI
Thermal inversions	0.0030***	0.0030***	0.0030***	0.0030***
	(11.68)	(11.81)	(11.68)	(11.81)
F statistics	136.538	139.418	136.538	139.418
Year FE	YES	YES	YES	YES
Firm FE	YES	YES	YES	YES
Province FE	YES	YES	YES	YES
Observations	3188	2839	3188	2839

Robust standard errors cluster in industry.

***p < 0.01.

**p < 0.05.

*p < 0.1.

OPEN IN VIEWER

Table 9.

Robustness check: replacement patent application with grant.

	LCTcpx_median			LCTcpx_mean
	Grant (1)	Invention (2)	Utility Model (3)	Grant (4)	Invention (5)	Utility Model (6)
CFI	0.1418***	0.1888*	0.1255**	0.1373**	0.1927*	0.1258**
	(3.14)	(2.02)	(2.23)	(2.92)	(2.00)	(2.38)
top	0.0524***	0.4018***	0.0419***	0.0557**	0.4160***	0.0446***
	(3.24)	(3.71)	(3.81)	(2.80)	(4.32)	(3.31)
ln exes	−0.0458*	−0.0030	−0.0437	−0.0461*	−0.0005	−0.0407
	(−2.05)	(−0.10)	(−1.60)	(−1.99)	(−0.02)	(−1.53)
ln age	0.0405	0.1151**	−0.1938*	0.0330	0.1184**	−0.2002*
	(0.77)	(2.20)	(−1.85)	(0.68)	(2.92)	(−1.92)
ln people	−0.0108	−0.0693**	0.0000	−0.0036	−0.0579**	0.0029
	(−0.75)	(−2.63)	(0.00)	(−0.28)	(−2.21)	(0.22)
reass*year	0.3295	0.3380	0.7979	0.4160**	0.4033*	1.0845**
	(1.44)	(1.19)	(1.39)	(2.55)	(1.89)	(2.23)
policy	−0.0878***	−0.2053***	−0.0966***	−0.0868***	−0.1969***	−0.0844***
	(−7.86)	(−11.63)	(−4.17)	(−7.43)	(−11.31)	(−4.64)
Constant	−1.7296	−2.9840	−2.9006	−2.3475*	−3.8281	−4.3498
	(−1.13)	(−0.98)	(−0.98)	(−2.10)	(−1.67)	(−1.72)
Year FE	YES	YES	YES	YES	YES	YES
Firm FE	YES	YES	YES	YES	YES	YES
Province FE	YES	YES	YES	YES	YES	YES
Observations	438	181	339	438	181	339
Adj R-squared	0.2378	0.2925	0.1202	0.2659	0.3395	0.1537

Robust standard errors cluster in industry.

***p < 0.01.

**p < 0.05.

*p < 0.1.

OPEN IN VIEWER

Table 10.

Robustness check: alternative innovation quality measure.

	LCTknowidth_median		LCTknowidth_mean
	(1)	(2)	(3)	(4)
CFI	0.4910*	0.4838*	0.4854*	0.4756*
	(1.87)	(1.82)	(1.89)	(1.79)
top		−0.1506***		−0.1457***
		(−5.52)		(−6.20)
ln exes		−0.0146		−0.0202
		(−0.28)		(−0.36)
ln age		0.3292**		0.3210***
		(2.86)		(3.15)
ln people		0.0208		0.0170
		(0.89)		(0.89)
reass*year		−1.2827***		−0.9067***
		(−3.45)		(−3.42)
policy		−0.0629		−0.0848***
		(−1.62)		(−4.03)
First stage	The dependent variable is CFI
Thermal inversions	0.0034***	0.0034***	0.0034***	0.0034***
	(7.38)	(7.71)	(7.38)	(7.71)
F statistics	54.449	59.457	54.449	59.457
Year FE	YES	YES	YES	YES
Firm FE	YES	YES	YES	YES
Province FE	YES	YES	YES	YES
Observations	541	541	541	541

Robust standard errors cluster in industry.

***p < 0.01.

**p < 0.05.

*p < 0.1.

OPEN IN VIEWER

Table 11.

Robustness check: extensive control variables.

	LCTcpx_median			LCTcpx_mean
	Application(1)	Invention(2)	Utility Model(3)	Application(4)	Invention(5)	Utility Model(6)
CFI	0.1444***	0.1169*	0.1415**	0.1303***	0.1113*	0.1521**
	(4.35)	(1.88)	(2.24)	(3.43)	(1.89)	(2.28)
top	0.0398***	0.1442***	0.0079	0.0414***	0.1581***	0.0124
	(4.16)	(3.44)	(1.21)	(3.54)	(4.17)	(1.75)
ln exes	−0.0040	0.0003	0.0094	−0.0030	0.0076	0.0127
	(−0.53)	(0.03)	(0.50)	(−0.36)	(1.11)	(0.62)
ln age	−0.0740	0.0641	−0.2443**	−0.0862	0.0530	−0.2804**
	(−0.98)	(1.19)	(−2.47)	(−1.25)	(0.95)	(−2.49)
ln people	0.0066	−0.0101	0.0098	0.0063	−0.0124	0.0009
	(0.72)	(−0.62)	(0.77)	(0.70)	(−0.71)	(0.05)
reass*year	−0.1758	0.1914	0.5520	−0.2339	0.2972	0.7416
	(−0.73)	(0.80)	(0.77)	(−1.02)	(1.26)	(1.03)
policy	−0.0080	−0.0751*	−0.0603	−0.0222	−0.0888**	−0.0738
	(−0.32)	(−2.04)	(−0.84)	(−1.18)	(−2.50)	(−1.11)
HHI	−0.0015***	−0.0006	−0.0004	−0.0006	−0.0003	0.0006
	(−3.35)	(−0.36)	(−0.30)	(−1.20)	(−0.15)	(0.46)
ROA	−0.1484**	−0.0931	−0.2252***	−0.1469**	−0.1135	−0.1885**
	(−2.34)	(−1.39)	(−3.24)	(−2.32)	(−1.62)	(−2.55)
leverage	−0.0621	−0.0444	−0.0872	−0.0785*	−0.0706	−0.0747
	(−1.42)	(−0.80)	(−1.71)	(−1.97)	(−1.32)	(−1.46)
boardsize	0.0003	0.0004	−0.0010	0.0007	0.0007	0.0003
	(0.18)	(0.19)	(−0.35)	(0.41)	(0.38)	(0.09)
duality	−0.0112	−0.0262	−0.0110	−0.0076	−0.0257	−0.0105
	(−0.69)	(−1.51)	(−0.73)	(−0.46)	(−1.33)	(−0.74)
Constant	1.7498	−1.2046	−1.2567	2.1845	−1.9780	−1.9239
	(1.16)	(−0.65)	(−0.40)	(1.52)	(−1.16)	(−0.62)
Year FE	YES	YES	YES	YES	YES	YES
Firm FE	YES	YES	YES	YES	YES	YES
Province FE	YES	YES	YES	YES	YES	YES
Observations	529	394	341	529	394	341
Adj R-squared	0.2718	0.2842	0.1201	0.2733	0.2890	0.1227

Robust standard errors cluster in industry.

***p < 0.01.

**p < 0.05.

*p < 0.1.

Full sample retest

CO₂ emission allowance listed enterprises are used as our primary sample due to the availability of channel data. The results of the first robustness check, which retested the entire sample of 2953 CO₂ emission allowance enterprises and 40490 LCT patents, are shown in Table 7. The estimated coefficients on CFI are significantly positive at the 5% level, suggesting a strongly positive association between carbon finance and the low-carbon innovation quality of all enterprises with CO₂ emission allowances. We report the results of the 2SLS estimations in Table 8, which are consistent with the above baseline results regardless of the inclusion of control variables; that is, the coefficient of CFI is always significantly positive.

Replacement patent application with grant

Patent granted is used to refer to the patent examiner having approved it after an applicant has successfully prosecuted his patent application, which means it is more useful and novel. Fewer patents have been issued than have been applied for, and there is typically a review period of approximately 10 months between the filing of a patent application and its granting. Therefore, there should be a delay when using the patent granted in place of an application, and the delay period is chosen to be one year in our paper. In this case, we further manually collected the low-carbon patents of CO₂ emission allowance listed enterprises in 2022. We present the results in Table 9. The estimated coefficients on CFI are significantly positive at the 1% (5%) level, as shown in columns (1) and (4), which is highly consistent with the baseline results, suggesting a strongly positive association between carbon finance and the quality of low-carbon innovation.

Alternative innovation quality measure

In the third robustness check, the measure of innovation quality is replaced. Specifically, Gtknowidth, defined as the knowledge width of patents, measured by the diversity of patents’ IPC codes, is employed. IPC codes generally adopt the format “Section-Class-Subclass-Main group-Group,” i.e. “C10B53/02.” Following Zhang and Zheng, ⁷⁴ it is weighted by the logical idea of the Herfindahl–Hirschman Index at the level of Main group, and the specific calculation method of enterprise patent knowledge width is as follows: GTknowidth = 1 − ∑ α 2 , where α 2 represents the proportion of each Main group in the patent IPC number, i.e. the proportion of the part before “/.” After determining each patent's knowledge width, it is aggregated to the enterprise level using the median and mean approaches, adopting a similar approach to low-carbon technical complexity (LCTcpx). As shown in Table 10, the 2SLS estimate of the coefficient of CFI is 0.4838 (0.4756), which is statistically significant at the 10% level and positive and slightly greater than those in the baseline regression, confirming the robustness of our main results.

Extensive control variables

Suffering from the limitation of the availability of non-listed enterprises’ financial data, our controls are relatively insufficient. In order to avoid the problem of missing variables in the previous regression results, we choose industry Herfindhal–Hirschman Index (HHI), return on an asset (ROA), return on an asset (leverage), CEO duality (duality), and board size (boardsize) ⁸¹ as the control variables to regress again. Table 11 shows the regression results of the missed important variables, and it is found that the symbols and the significance of the relationship coefficient of the fellowship are basically consistent with the previous ones, indicating that the conclusions of this study are still stable. We can now say with certainty that there is a positive correlation between the growth of carbon finance and the advancement of LCT change in answer to RQ1 (i.e. does carbon finance improve the quality of enterprises’ low-carbon innovation?).

The effect of two constituent markets on low-carbon technological change

The two components of the integrated development of carbon finance are carbon emissions trading and LCF markets. Supporting low-carbon projects and industries through investment and financing activities is an essential goal of LCF markets. The carbon emissions trade promotes firms to accomplish their emission reduction goals and profit from lower emissions by providing incentives for LCF. These benefits can then be reinvested in related projects. This section investigates the effects of CET as well as the LCF markets on the development of LCT as a response to RQ2, and the results are reported in Tables 12 and 13.

OPEN IN VIEWER

Table 12.

Carbon emission trading and low-carbon innovation quality.

	LCTcpx_median			LCTcpx_mean
	Application(1)	Invention(2)	Utility Model(3)	Application(4)	Invention(5)	Utility Model(6)
CET	0.1276**	0.0674	0.2748***	0.1236**	0.0546	0.2867***
	(2.30)	(1.08)	(3.44)	(2.17)	(0.85)	(3.51)
Top	0.0420***	0.1237**	0.0104*	0.0423***	0.1337***	0.0126**
	(5.42)	(2.87)	(2.05)	(4.27)	(3.35)	(2.22)
ln exes	−0.0165**	−0.0040	−0.0105	−0.0162**	0.0034	−0.0100
	(−2.46)	(−0.50)	(−0.65)	(−2.32)	(0.48)	(−0.60)
ln age	0.0848***	0.1343***	−0.1670*	0.0636**	0.1141***	−0.1934*
	(3.06)	(6.56)	(−2.02)	(2.41)	(5.17)	(−2.07)
ln people	0.0021	−0.0133	0.0083	0.0027	−0.0160	0.0048
	(0.21)	(−0.89)	(1.05)	(0.29)	(−1.01)	(0.44)
reass*year	−0.3308	0.0807	−0.1680	−0.3815	0.1956	−0.0539
	(−1.28)	(0.35)	(−0.24)	(−1.49)	(0.75)	(−0.08)
policy	−0.0209	−0.0789**	−0.0692	−0.0330	−0.0917***	−0.0763
	(−0.73)	(−2.55)	(−0.80)	(−1.67)	(−2.99)	(−0.98)
Constant	2.3375	−0.5010	1.4857	2.7370	−1.3308	1.1079
	(1.44)	(−0.29)	(0.49)	(1.71)	(−0.68)	(0.36)
Year FE	YES	YES	YES	YES	YES	YES
Firm FE	YES	YES	YES	YES	YES	YES
Province FE	YES	YES	YES	YES	YES	YES
Observations	541	399	351	541	399	351
Adj R-squared	0.2471	0.2746	0.1295	0.2491	0.2759	0.1376

Robust standard errors cluster in industry.

***p < 0.01.

**p < 0.05.

*p < 0.1.

OPEN IN VIEWER

Table 13.

Low-carbon investment and financing and low-carbon innovation quality.

	LCTcpx_median			LCTcpx_mean
	Application(1)	Invention(2)	Utility Model(3)	Application(4)	Invention(5)	Utility Model(6)
LCF	0.1252**	0.1439*	0.0792	0.1108**	0.1376*	0.0982
	(2.52)	(1.81)	(0.91)	(2.17)	(1.80)	(1.04)
Top	0.0457***	0.1329**	0.0214***	0.0458***	0.1428***	0.0241***
	(6.27)	(2.84)	(4.90)	(4.93)	(3.27)	(6.45)
ln exes	−0.0135**	−0.0047	−0.0012	−0.0133**	0.0028	−0.0004
	(−2.26)	(−0.61)	(−0.10)	(−2.20)	(0.38)	(−0.03)
ln age	0.0819**	0.1313***	−0.1713	0.0615*	0.1109***	−0.2024*
	(2.33)	(4.96)	(−1.71)	(1.80)	(4.18)	(−1.92)
ln people	0.0025	−0.0136	0.0055	0.0030	−0.0162	0.0018
	(0.27)	(−0.96)	(0.62)	(0.37)	(−1.08)	(0.15)
reass*year	−0.0802	0.2527	0.6664	−0.1452	0.3487*	0.8371
	(−0.39)	(1.33)	(0.92)	(−0.74)	(1.89)	(1.14)
policy	−0.0268	−0.0817**	−0.0750	−0.0388*	−0.0943***	−0.0830
	(−0.98)	(−2.61)	(−0.89)	(−2.00)	(−3.07)	(−1.10)
Constant	0.6777	−1.9107	−1.8778	1.1784	−2.5958*	−2.4854
	(0.52)	(−1.31)	(−0.60)	(0.94)	(−1.85)	(−0.78)
Year FE	YES	YES	YES	YES	YES	YES
Firm FE	YES	YES	YES	YES	YES	YES
Province FE	YES	YES	YES	YES	YES	YES
Observations	541	399	351	541	399	351
Adj R-squared	0.2437	0.2813	0.0938	0.2437	0.2813	0.0938

Robust standard errors cluster in industry.

***p < 0.01.

**p < 0.05.

*p < 0.1.

The carbon emissions trade does not appear to have a substantial incentive to improve the quality of invention patents, but the LCF market does show a positive association that is significant at the 10% level in columns (2) and (5) in Table 11. Patents for utility models work the other way around. The regression in columns (3) and (6) in Table 12 shows that the carbon emissions trade has a statistically significant and positive impact at the 1% level, and the coefficient of LCF is nonsignificant in columns (3) and (6) in Table 13. The technical bar for invention patents is higher than utility model patents, necessitating a greater expenditure in R&D. Therefore, the process of invention is more dependent on LCF market. The improvement of the quality of utility model patents demonstrates that carbon emissions trading effectively inhibits “strategic innovation,” which frequently manifests itself in the growth of utility model patents. This is where carbon finance varies from green business policies and command-and-control environmental regulations. Under carbon emissions trading, technical decisions are made in regulated enterprises by purposefully comparing emission prices with production costs ¹⁵ instead of mechanically catering to policies.

Heterogeneity checks

RQ3: Does carbon finance have a heterogeneous effect on the quality of low-carbon innovation? The discussion of the RQ3 is beneficial for fully comprehending the operation of the carbon market in China.

Ownership

Two mechanisms have been paralleled because China's carbon market mechanism has not yet been fully and effectively established. The first is the market mechanism, with carbon trading as its central component. The second is administrative intervention, i.e. local governments using administrative control to maintain the emission fulfillment of allowance enterprises.^59,82 SOEs are more susceptible to government control than non-SOEs. For instance, the government can link emission fulfillment behavior to a state-owned enterprise performance appraisal and evaluation system. Therefore, under the administrative intervention mechanism, the sample of SOEs can be utilized to assess the effect of carbon finance on the quality of low-carbon innovation. The market mechanism of carbon finance is tested using a sample of non-SOEs. The estimates are reported in columns (1) - (4) in Table 14. In columns (1) and (2), the estimate of CFI related to non-SOEs is highly significant and positive, while the estimates in columns (3) and (4), reflecting SOEs, show the phenomenon of “focusing on fulfillment rather than trading,” with no significant effect of carbon finance on low-carbon technological change. Non-SOEs, in contrast, are less affected by administrative intervention and can efficiently use market mechanisms. Emissions’ trading has fueled its incentives for low-carbon technological change.

OPEN IN VIEWER

Table 14.

Heterogeneity checks: ownership.

	Non-state-owned		State-owned
	LCTcpx_median（1）	LCTcpx_mean（2）	LCTcpx_median（3）	LCTcpx_mean（4）
CFI	0.8652***	0.8229***	0.1799	0.1564
	(3.79)	(3.75)	(1.07)	(0.87)
top	−0.0066	0.0086	0.0421***	0.0434***
	(−0.07)	(0.09)	(8.27)	(7.81)
lnexes	−0.0523**	−0.0465**	−0.0034	−0.0080
	(−2.70)	(−2.31)	(−0.20)	(−0.49)
lnage	−0.3819***	−0.3613***	0.1178***	0.0865**
	(−3.39)	(−3.49)	(3.92)	(2.21)
lnpeople	0.0003	0.0071	−0.0012	−0.0027
	(0.02)	(0.54)	(−0.09)	(−0.19)
reassyear	−1.9116	−2.4057*	−0.1808	−0.1931
	(−1.38)	(−1.84)	(−0.74)	(−0.81)
policy	0.0342	0.0395*	−0.0334	−0.0472
	(1.62)	(1.90)	(−0.92)	(−1.51)
First stage	The dependent variable is CFI
Thermal inversions	0.0033***	0.0033***	0.0033***	0.0033***
	(6.44)	(6.44)	(7.79)	(7.79)
F statistics	41.434	41.434	65.730	65.730
Year FE	YES	YES	YES	YES
Firm FE	YES	YES	YES	YES
Province FE	YES	YES	YES	YES
Observations	227	227	314	314

Robust standard errors cluster in industry.

***p < 0.01.

**p < 0.05.

*p < 0.1.

Technology intensity

Characterized by knowledge intensity and technology intensity, high-tech enterprises that continue to conduct research and development transform technological advances and create the core of autonomous intellectual property rights. ^{k , l} High-tech enterprises are more responsive to the relationship between carbon finance and the quality of low-carbon innovation. The results are reported in columns (1) - (4) in Table 15, demonstrating that high-tech enterprises upgrade their low-carbon innovation quality in response to market-oriented carbon finance, while non-high-tech enterprises have no apparent response in their low-carbon innovation quality.

OPEN IN VIEWER

Table 15.

Heterogeneity checks: technology intensity.

	High-tech		Non-high-tech
	LCTcpx_median（1）	LCTcpx_mean（2）	LCTcpx_median（3）	LCTcpx_mean（4）
CFI	0.4232**	0.4073**	−0.1116	−0.1512
	(2.47)	(2.37)	(−0.31)	(−0.43)
Top	0.0584	0.0838*	0.0392*	0.0402*
	(1.41)	(2.05)	(1.82)	(1.84)
lnexes	0.0107	0.0140	−0.0437*	−0.0464**
	(1.27)	(1.58)	(−2.14)	(−2.50)
lnage	−0.1163	−0.1210**	0.1253	0.1222
	(−1.68)	(−2.48)	(1.33)	(1.44)
lnpeople	−0.0072	−0.0043	0.0066	−0.0090
	(−0.67)	(−0.48)	(0.13)	(−0.18)
reassyear	−0.3519	−0.3532	−0.5309	−0.6593
	(−1.04)	(−1.19)	(−1.08)	(−1.35)
policy	−0.0274	−0.0397***	−0.0248	−0.0366
	(−1.30)	(−5.89)	(−0.42)	(−0.65)
First stage	The dependent variable is CFI
Thermal inversions	0.0038***	0.0038***	0.0027***	0.0027***
	(5.81)	(5.81)	(5.43)	(5.43)
F statistics	33.803	33.803	29.499	29.499
Year FE	YES	YES	YES	YES
Firm FE	YES	YES	YES	YES
Province FE	YES	YES	YES	YES
Observations	410	410	131	131

Robust standard errors cluster in industry.

***p < 0.01.

**p < 0.05.

*p < 0.1.

Possible channels

Up to this point, the role of carbon finance in promoting low-carbon technological change has been confirmed in specific enterprises. The continuous advancement of carbon marketization at the city level positively affects firm-level low-carbon innovation quality. In order to response to RQ4, we take advantage of two potential avenues: (1) to determine whether the carbon market promotes increased R&D intensity, taking into account the two levels of R&D personnel (RDpersonnel) and R&D expenditure (RDexpenditure); and (2) to determine whether market-oriented carbon finance can reduce the external financial constraints faced by enterprises (KZ index). The results are all reported in Table 16.

OPEN IN VIEWER

Table 16.

Possible channels.

	RDpersonnel		RDexpenditure		KZindex
	LCTcpx_median(1)	LCTcpx_mean(2)	LCTcpx_median(3)	LCTcpx_mean(4)	LCTcpx_median(5)	LCTcpx_mean(6)
CFI	0.1197*	0.1257*	0.0994**	0.0780*	0.1920***	0.1671***
	(2.07)	(2.02)	(2.79)	(1.93)	(4.29)	(3.34)
Top	0.0423***	0.0410***	0.0519***	0.0519***	0.0404***	0.0422***
	(4.84)	(5.62)	(9.05)	(8.28)	(4.04)	(4.64)
ln exes	0.0007	0.0026	−0.0006	−0.0001	−0.0005	0.0013
	(0.03)	(0.10)	(−0.06)	(−0.01)	(−0.02)	(0.05)
ln age	−0.2846*	−0.3199**	0.0039	−0.0104	−0.1404	−0.1429*
	(−2.02)	(−2.71)	(0.05)	(−0.15)	(−1.71)	(−1.78)
ln people	0.0038	0.0085	−0.0041	−0.0020	−0.0066	−0.0059
	(0.35)	(0.97)	(−0.36)	(−0.20)	(−0.46)	(−0.42)
reass*year	−0.3357	−0.3185	−0.1983	−0.2624	−0.2153	−0.2268
	(−1.43)	(−1.43)	(−0.84)	(−1.22)	(−1.01)	(−1.11)
policy	0.0173	−0.0226**	−0.0198	−0.0309	0.0378***	0.0120
	(1.70)	(−2.69)	(−0.56)	(−1.30)	(3.00)	(0.89)
CFI*RDpersonnel	0.4224*	0.2730
	(1.85)	(1.21)
RDpersonnel	−0.1402	−0.0281
	(−0.94)	(−0.19)
CFI*RDexpenditure			1.7800*	2.1886**
			(2.05)	(2.92)
RDexpenditure			−0.6187	−0.9341**
			(−1.18)	(−2.77)
CFI*KZindex					−0.0277***	−0.0254***
					(−3.46)	(−3.07)
KZindex					0.0253***	0.0216***
					(5.45)	(3.96)
Constant	3.3224**	3.2894**	1.6909	2.1542	2.3193	2.4295
	(2.21)	(2.36)	(1.11)	(1.57)	(1.54)	(1.67)
Year FE	YES	YES	YES	YES	YES	YES
Firm FE	YES	YES	YES	YES	YES	YES
Province FE	YES	YES	YES	YES	YES	YES
Observations	421	421	490	490	464	464
Adj R-squared	0.2809	0.2857	0.2568	0.2565	0.3095	0.2987

Robust standard errors cluster in industry.

***p < 0.01.

**p < 0.05.

*p < 0.1.

R&D intensity

Many R&D inputs are entailed for enterprises to develop advanced technology and modify product designs. ⁸³ According to Pakes and Griliches ⁸⁴ and Hu and Jefferson, ⁷⁰ the association between R&D inputs and patent production exists only at the contemporaneous level. To identify the possible channel of R&D input, the estimating equation is set as Equation (9) as follows:

Lpatent _ qualit y i t = α 4 + β 4 CFI k t + θ 1 RDintensity i t + φ 1 CFI k t * RDintensity i t + λ 4 X + γ year + μ firm + ε i t

(9)

where RDintensity i t refers to R&D personnel and R&D expenditure, measured by the ratio of R&D personnel to all employees and the ratio of R&D expense stock to the enterprise's total assets, respectively. The decision to invest in the R&D of CO₂ emission allowance enterprises is made by weighing the long-term costs of purchasing carbon allowances against the anticipated gains from modifying technology to be low-carbon. Instead of mechanical pandering to environmental regulations, low-carbon innovation was driven by carbon finance benefits from the subjective initiative of CO₂ emission allowance enterprises. The result of this channel is reported in columns (1) to (4) in Table 16. In columns (1), (3), and (4), the coefficients of the interaction term, CFI * RDintensity , as the key parameter of interest to be estimated, are positive and significant, suggesting that the development of carbon finance encourages enterprises to conduct more R&D activity, which is conducive to low-carbon innovation quality improvement.

Financial constraints

Furthermore, it is undeniable that most Chinese enterprises, particularly private ones, face a challenging and expensive financing environment, ⁵⁷ possibly preventing private enterprises from implementing technological change and advancement. Promoting green transformation to establish an ecological civilization, China has developed a program of reforms aimed specifically at shifting investment patterns and making its entire financial system green, called the green financial system, which is a leap that many other countries have not yet taken. ⁸⁵ Green finance has undertaken the important mission of guiding social funds to flow to energy-saving, low-carbon, and environmental protection industries. LCF are anticipated to finance low-carbon technological change as a market segment of green finance. In this part, we further consider whether financial constraints will be a channel through which de facto carbon finance affects the quality of low-carbon innovation.

Following Kaplan and Zingales, ⁸⁶ we construct the KZindex of listed enterprises, which denotes the degree of financial constraints experienced by enterprises: the stronger the KZindex is, the greater the degree of financing constraints faced by an enterprise, as shown in Equation (10) below:

Lpatent _ qualit y i t = α 5 + β 5 CFI k t + δ 1 KZindex i t + τ 1 CFI k t * KZindex i t + λ 4 X + γ year + μ firm + ε i t

(10)

The result of this channel is reported in columns (5) to (6) in Table 16. The estimates of the coefficients of the interaction term, CFI * KZindex , are negative and significant, demonstrating the role of carbon finance in enhancing the financing capacity of enterprises. A possible channel for this is that CO₂ emission allowance enterprises can spend more on the quality improvement of low-carbon innovation if they face fewer external financial constraints.

Concluding remarks

“Addressing climate change will require a significant shift in the global economy: how we generate energy; how we produce goods…”; one of the few tools at our disposal is carbon finance, an environmental market solution. ¹² In this paper, we have investigated China's carbon finance system's partial success in low-carbon technological change through the following processes: The first step is to obtain 2953 CO₂ emission allowance enterprise patents by means of web crawler technology and manual reading. Leveraging textual analysis technology, we then provide a new attempt at measuring the quality of low-carbon innovation. The third is to evaluate the development of carbon finance of eight pilots, showing that carbon finance is not fully marketable and operates with regional dissonance. To control endogeneity, we use the thermal inversions of eight pilots to instrument carbon finance. Up to now, we have conducted methodical empirical research; therefore, we are able to fully respond to RQ1 to RQ4.

As a whole, our evidence from China coincides with the encouraging results of Calel and Dechezlepretre's ¹⁵ European investigation and lends important support to the existing literature^12,14 that emphasizes the crucial role of carbon finance in new LCTs. Our paper's full sample test from 2953 CO₂ emission allowance enterprises have been used to supplement the field of research on China's emission trading.^22,27,51,72. The results of Dang and Motohashi ¹⁰ demonstrate how unreliable it is to use invention patent share as a stand-in for innovation quality. A new attempt to measure the quality of low-carbon technological innovation is offered in this study, which may be viewed as an improvement above previous study. ⁵¹

The national ETS builds on the successful experience of pilot carbon markets implemented in eight regions and has been in place for two years. It is hoped that the study's findings can provide enlightenment for the further development of the national carbon market. First, as the coverage of CET expands, the growth and development of the LCF market ought to also emphasize cultivating in order to support high-quality innovation patents and PCT patents. Second, the establishment of the carbon market needs to appropriately address how the market and government should position their respective roles, define the boundaries of their respective behaviors in a scientific and reasonable manner, minimize the intervention of state-owned enterprises, and optimize the function of market mechanisms within these enterprises. Briefly speaking, carbon markets are trading systems in which carbon credits are sold and bought. Selling carbon credits indicates that the enterprises’ emissions are below historical averages or industry averages, which may suggest the development of LCTs. ^m Therefore, third, the government can suitably reward these enterprises for their efforts by providing staff and capital to boost R&D spending and mitigate financial constraints.

Frankly speaking, there are also two limitations in the research of this paper. First and foremost, we developed an index system in order to examine the distinct roles that CET and LCF play in the transition to LCTs. Our indicators lack persuasiveness since they are ultimately subjective, even though they are founded on earlier research. Second, due to the limitations of the data, it is impossible to collect more financial data of unlisted CO₂ emission allowance enterprises. It is hoped that more representative variables for CET and LCF can be selected and higher information disclosure of CO₂ emission allowance enterprises in future research to further test the relationship between carbon finance and low-carbon technological change.