Clean energy,emission trading policy,and CO 2 emissions: Evidence from China

Abstract

This paper constructs a unique dataset of clean energy and adopts static panel models and dynamic panel specifications to explore the correlation between clean energy and CO₂ emissions. Furthermore, this paper employs the interaction term of pilot areas and pilot time as the proxy of emission trading policy to examine the effect of China’s emission trading pilot on clean energy. Then, this paper conducts quasi-natural experiments on CO₂ emissions. Our findings show a negative correlation between CO₂ emissions and clean energy. We also find that China’s emission trading pilot has a significant impact on promoting clean energy. In addition, empirical results affirm that emission trading market pilots can help decrease CO₂ emissions. Finally, we put forwards relevant policy recommendations.

Keywords

emission trading policy clean energy GMM DID-PSM

1. Introduction

China has had rapid and sustained growth at the expense of excessive energy consumption and environmental pollution since its economic reforms in the late 1970s.¹ By the early 1980s, China was suffering from black smoke from heavy industry. In the 21st century, the rapid advancement of industrialization and economic development has led to increased energy use and greenhouse gas emissions. Thus, China’s air pollution problem has become more acute. Fossil energy use is a core source of carbon dioxide emissions, and approximately 80% of carbon dioxide emissions are caused by fossil energy.² The extreme climate change and environmental crisis caused by carbon emissions have attracted the Chinese government’s attention. China’s central and local governments have already planned to replace fossil energy use and reduce emissions of greenhouse gases in response to environmental pollution and global warming.¹ On the one hand, the China Energy Administration has said that China would vigorously develop clean energy during the 14th Five-Year Plan period. Clean energy refers to energy that does not emit pollutants and can be directly used for production and life. It includes nuclear energy and renewable energy.³ In addition, the definition of clean energy has three aspects: one, clean energy describes the technical system of energy utilization, which is not a simple classification of energy; two, clean energy values cleanliness and economy; and three, clean energy complies with certain emission standards.⁴ According to the China Energy Administration, clean energy consumption in China will account for 80% of the total energy consumption in the next five years.² On the other hand, the carbon emission trading mechanism based on the market mechanism effectively reduces CO₂ and other greenhouse gas emissions. In October 2011, the China Nation Development and Reform Commission issued the Notice on Carbon Emission Trading Pilot Work, which approved Chongqing, Beijing, Shenzhen, Shanghai, Hubei, Guangdong, and Tianjin to launch carbon trading pilots.⁵ After nearly ten years of piloting, on July 16, 2021, the emissions trading market was officially launched in China.³ The emissions trading market allows firms with lower carbon emissions to trade on the premise that the total emissions specified by carbon emissions trading are not exceeded. A firm’s right to emit a certain amount of carbon dioxide turns into a trading commodity. Carbon emissions trading is an important mechanism for employing a market economy to reduce emissions of greenhouse gases and promote environmental protection. Thus, research on emission trading policies, clean energy, and CO₂ emission reduction has crucial strategic significance.

In the following research, we use the following methods: One, we employ static panel models and dynamic panel specifications to study the correlation between clean energy and CO₂ emissions; two, we use the interaction term of pilot areas and pilot time as the proxy of emission trading policy; and three, we exploit the difference-in-differences (DID) model and the propensity score matching-DID (PSM-DID) method to study the impact of the emission trading pilot on CO₂ emissions.

This paper is dedicated to finding effective ways to reduce carbon emissions, thereby protecting the environment. First, we exploit static panel models and dynamic panel specifications to study the correlation between clean energy and CO₂ emissions. We prove that clean energy development can help reduce carbon dioxide emissions. Since clean energy can help reduce carbon dioxide emissions, we further study how to accelerate clean energy development to reduce carbon dioxide emissions effectively and whether the emission trading policy is conducive to reducing carbon emissions. Thus, in the next sections, we use the interaction term of pilot areas and pilot time as the proxy of emission trading policy to study the effects of emission trading policy on clean energy and the policy effect of emission trading policy on CO₂ emissions. We prove through empirical analysis that the carbon emissions trading policy can promote clean energy development. In addition, we further find that the carbon emissions trading policy can not only directly reduce CO₂ emissions but also indirectly reduce carbon emissions by promoting the development of clean energy. The innovations in this paper are as follows: First, this paper combines clean energy, emissions trading policy, and CO₂ emission reduction to explore the relationship between energy use and environmental protection, providing a new perspective for related research. Second, this paper uses the interaction term of pilot areas and pilot time as the proxy of emission trading policy and introduces it into the fixed effects model, providing a way to measure specific policies. Third, this paper uses PSM-DID methods and employs (difference and system) GMM to estimate the dynamic panel model, reducing the impact of endogeneity on the results.

2. Related literature

2.1. Emission trading scheme, energy, and Co₂ emissions

Several studies have been published on the correlation between emission trading policy, energy, and renewable energy. Streimikiene and Roos⁶ found that the emission trading policy would increase investments in renewable energy sources. Möst and Fichtner⁷ indicated that the promotion of renewable energy led to a decrease in the scarcity of CO₂-emission allowances. Lin and Jia⁸ used a dynamic recursive computable general equilibrium model with multiple sectors to examine different emission trading scheme (ETS) revenue distribution strategies and found that an ETS with no renewable energy subsidy would reduce energy demand, increase the cost of renewable energy sources, and reduce generation. Ma et al.⁹ developed an analysis of the dynamic linkage mechanism of China’s carbon emission trading market, energy market, and capital market based on the DCC-MVGARCH model. Many recent studies have shown ETS’s impact on traditional energy and renewable energy; however, minimal attention has been devoted to the impact of ETS on clean energy.

Some scholars have mainly been interested in questions concerning CO₂ emissions from energy use. Al-mulali and Binti Che Sab¹⁰ indicated that energy consumption enables high economic and financial development. However, high levels of development increased CO₂ emissions. Shahbaz et al.¹¹ pointed out that financial development and trade openness shrink CO₂ emissions, while economic growth and energy consumption increase CO₂ emissions. Mahdi Ziaei¹² used panel vector autoregression models to explore how financial indicators can affect energy consumption and CO₂ emissions, highlighting the role of both variables in explaining deviations, with only the degree of impact varying. Shahzad et al.¹³ revealed an inverted U-shaped relationship between carbon emissions and energy consumption. Ehigiamusoe and Lean¹⁴ found that energy consumption, economic growth, and financial development negatively affect carbon emissions in 122 countries. Khan et al.¹⁵ exploited the dynamic ARDL simulation model to examine how globalization, economic issues, and energy use affect CO₂ emissions in Pakistan. According to them, energy consumption in Pakistan positively impacts CO₂ emissions.

2.2. Clean energy and Co₂ emissions

Some recent studies^16–18 have shown that clean energy favours CO₂ emission reduction. Maji¹⁷ studied the effect of clean energy on CO₂ emissions using the renewable energy index as a clean energy measure. They analysed data for 42 sub-Saharan countries using the system generalized method of moments (GMM), indicating that increased renewable energy usage in the area would dampen CO₂ emissions. Ahmed et al.,¹⁶ used the ARDL technique to show that an increase in clean energy consumption reduces per capita CO₂ emissions. Ummalla and Goyari¹⁸ investigated the economic growth and CO₂ emissions effects of clean energy consumption within the environmental Kuznets curve hypothesis framework. The results showed that clean energy consumption and economic growth significantly negatively impact CO₂ emissions. However,¹⁹ employed a newly developed bootstrap ARDL bounds test and found no cointegration among real GDP per capita, clean energy consumption, and CO₂ emissions in Canada, France, Italy, the US, and the UK. In addition, using data for Norway and New Zealand,²⁰ revisited the energy-growth nexus by taking into account the effect of clean energy, CO₂ emissions, and technological innovations. Clean energy use does reduce emissions of greenhouse gases at the expense of economic growth, according to researchers. Rahman and Alam²¹ found that the use of clean energy improved the environmental quality.

2.3. Emission trading scheme and Co₂ emissions

Researchers have attempted to assess the effect of ETCs on CO₂ emissions. Anger²² found that the European Emission Trading Scheme would result in a 7.4% reduction in air transport CO₂ emissions based on the Energy-Environment-Economy Model for Europe. Dutch power production and generation portfolios have relatively small and late effects on CO₂ emissions from emission trading.²³ Jiang et al.,²⁴ concluded that Shenzhen’s ETS was designed with a mechanism for postadjustment caps, which minimizes the impact of economic fluctuations on the scheme’s stability by slowing emission growth. Nguyen et al.,²⁵ identified that the carbon trading scheme of Japan could engender a 42% reduction in emissions without resilience constraints and 34% incorporating a best-mix, resilient approach. Gu et al.,²⁶ studied the effect of the Maritime Emissions Trading Scheme (METS) on fleet operations and their CO₂ emissions. In the short term, they found that low bunker prices, high allowance costs, or global METS coverage can all lead to a significant decrease in CO₂ emissions. Using the DID approach, Hu et al.,¹ examined how China’s CO₂ ETS affects energy conservation and emissions. According to the study, the carbon ETS reduces the energy use and CO₂ emissions of the regulated industries in pilot areas by 22.8% and 15.5%, respectively, compared to nonpilot areas. Shen et al.,²⁷ employed the PSM-DID to explore the causal effect of ETS. They suggested that ETS is conducive to carbon emissions reduction, but the effect attenuates over time. Zhang et al.,⁵ developed the full DEA evaluation model for estimating the efficiency of carbon ETMs. They showed that implementing the carbon trading policy significantly reduces the emissions (24.2%) of industrial CO₂. Zhang et al.,²⁸ found that China’s ETS pilot policy significantly improves regional carbon equality. Chen and Lin²⁹ identified the role of China’s ETS in promoting and reducing emissions, which is regarded as a valid policy tool to facilitate carbon neutrality.

2.4. Discussion

In summary, most scholars are concerned about the impact of traditional energy use on CO₂ emissions. A few scholars have studied the association between emission trading policy and investments in renewable energy use, the influence of renewable energy on CO₂ emission allowances, and the relationship between the ETS’s renewable energy subsidy and energy demand. In addition, little research has explored the influence of CO₂ emissions under China’s clean energy and emission trading policy. On the one hand, very few researchers are devoted to the impact of ETCs on clean energy. Existing research shows that much of the energy literature omits clean energy, which can be directly used for production and life. Some scholars conduct related research on renewable energy as a proxy indicator of clean energy. More energy studies mention only renewable and fossil energy use rather than clean energy use. Current research is not mature in this area, requiring more reliable research. On the other hand, the literature on ETCs and CO₂ emissions pays attention to specific industries, such as shipping, power production, and air transport.^22,23,26 However, literature related to the association between ETC and regional CO₂ emissions is scant. Thus, this paper devotes itself to the literature on the correlation between emission trading policy, clean energy, and regional CO₂ emissions.

The contribution of this paper is as follows: we first manually collect a unique dataset of clean energy and employ static panel models and dynamic panel specifications estimated by the (difference) and (system) GMM to explore the correlation between clean energy and CO₂ emissions. Second, we use the interaction term of pilot areas and pilot time as the proxy of emission trading policy to explore whether China’s current emission trading policy has a significant promoting effect on clean energy. Third, we study the impact of emission trading on CO₂ emissions and explore whether China’s current policy can help CO₂ emissions reduction. Finally, we make policy recommendations to policy-makers.

3. Data description

3.1. Main variables

(1) Clean energy (CE). We use the product of the total energy output of each province (municipalities and autonomous regions) and the proportion of clean energy to express clean energy production, which measures the scale of clean energy production (100 million kWh). Because the statistical yearbooks of China’s provinces (municipalities and autonomous regions) contain the total energy production of each region over the years, primary energy (hydropower) accounts for the proportion of total energy production, and other energy sources (nuclear energy, wind energy, geothermal energy, and other clean energy) account for the total energy production proportion. Given the incomplete clean energy data in the statistical yearbooks of some provinces (municipalities and autonomous regions), for the missing data, we use the clean energy production data calculated from the hydropower, nuclear power, wind power, solar, and other clean energy generation data of the provinces (municipalities and autonomous regions) in the China Energy Statistical Yearbook instead.^3,30

(2) Carbon dioxide emissions (CO₂). The total carbon dioxide emissions include natural gas emissions, raw coal emissions, crude oil emissions, and emissions during cement production. We obtain carbon dioxide data from the Carbon Emission Accounts and Datasets (CEADs).³¹

(3) Interaction term $C O R E$ . Pilot time dummy variable: This paper sets 2011 as the initial pilot time (year = 1, if t ≥ 2011; year = 0, if t < 2011). Pilot area dummy variable: We choose provinces (or municipalities) that implemented the pilot policy since 2011 as pilot areas ( $d μ = 1$ , if $μ$ denotes the pilot province; $d μ = 0$ , if $μ$ denotes not the pilot province). The interaction term $C O R E = d μ * y e a r$ denotes the “policy treatment effect”.³² An emission trading policy pilot is a localized decision-making and implementation activity to verify the correctness and feasibility of the policy proposal. To quantify the emissions trading policy, we use the interaction term of pilot areas and pilot time as the proxy for the emission trading policy.

3.2. Control variables

(1) Real GDP per capita (RGC). The level of economic development in each region is different. Understanding and grasping a region’s macroeconomic operation can be done effectively by the real GDP per capita. We divide each province’s actual GDP by its total population to determine its real GDP per capita.

(2) Research and development (R&D). A region’s scientific and technological strength and core competitiveness can be evident in the R&D activities it undertakes, including basic research, applied research, and experimental development. The indicator calculation method³³ is $S_{i t}^{d} = R D_{i t}^{d} + (1 - δ) * S_{i, t - 1}^{d}$ . $S_{i t}^{d}$ represents the domestic R&D stock, $R D_{i t}^{d}$ denotes the domestic R&D expenditure, and $δ$ indicates the depreciation rate ( $δ = 0.05$ ). The formula for $S_{i 0}^{d}$ is $S_{i 0}^{d} = R D_{i 0}^{d} \div (g + δ)$ . $S_{i 0}^{d}$ denotes the base period’s domestic R&D stock, $R D_{i 0}^{d}$ denotes the base period’s domestic R&D expenditure, and $g$ means the average growth rate of R&D expenditure.

(3) Education level (EDU). Regional energy mix transformation and carbon emission reduction are influenced by the amount of talent a region can provide, and differences in education levels affect the efficiency of regional energy efficiency and emission reduction. As a proxy for education level, we exploit the ratio of the population with a college degree or higher (6 years old and over).³²

(4) Government intervention (GIN). As a proxy for government intervention, we exploit the ratio of local government budget expenditure to GDP as a measure. Local governments can use taxation and subsidies to influence the carbon emissions trading market and clean energy development. The degree of government intervention and regulation on the market varies from region to region, so we use it as a control variable.

According to the data availability, this paper employs panel data from 28 provinces (municipalities and autonomous regions) in China from 2000 to 2016 and takes the logarithm of all variables. Because Ningxia, Hainan, and Tibet data are seriously lacking, these three provinces are not included in the 28 provinces in the sample of this paper. The specific databases used in this paper include the Chinese Statistical Yearbook, CEADs, and the China Energy Statistical Yearbook. Table 1 displays the descriptive statistics and correlation matrix.

Table 1.
Descriptive statistics and correlation matrix of variables.

Variable Mean Units Std. Dev. Min Max

CO₂ 0.073 Ton 0.137 −0.599 0.914

CE 0.145 100 million kWh 0.233 −0.638 1.772

CORE 0.076 - 0.265 0 1

RGC 0.123 Ten thousand Yuan/person 0.059 −0.252 0.315

EDU 0.213 % 2.132 −14.794 6.699

GIN 0.783 % 1.539 −3.578 10.033

R&D 0.012 Ten thousand Yuan 0.337 −1.513 1.513

Correlation matrix of variables

CO₂ CE CORE RGC EDU GIN R&D

CO₂ 1.000

CE −0.038 1.000

CORE −0.172 −0.029 1.000

RGC 0.279 0.231 −0.156 1.000

EDU −0.035 0.080 0.101 0.052 1.000

GIN −0.044 −0.072 −0.008 −0.205 0.046 1.000

R&D −0.066 −0.039 0.011 −0.109 0.040 0.006 1.000

Variable	Mean	Units	Std. Dev.	Min		Max
CO₂	0.073	Ton	0.137	−0.599		0.914
CE	0.145	100 million kWh	0.233	−0.638		1.772
CORE	0.076	-	0.265	0		1
RGC	0.123	Ten thousand Yuan/person	0.059	−0.252		0.315
EDU	0.213	%	2.132	−14.794		6.699
GIN	0.783	%	1.539	−3.578		10.033
R&D	0.012	Ten thousand Yuan	0.337	−1.513		1.513
Correlation matrix of variables
	CO₂	CE	CORE	RGC	EDU	GIN	R&D
CO₂	1.000
CE	−0.038	1.000
CORE	−0.172	−0.029	1.000
RGC	0.279	0.231	−0.156	1.000
EDU	−0.035	0.080	0.101	0.052	1.000
GIN	−0.044	−0.072	−0.008	−0.205	0.046	1.000
R&D	−0.066	−0.039	0.011	−0.109	0.040	0.006	1.000

Economic data often have nonstationary problems. Using nonstationary data for research will cause pseudoregression problems. Therefore, the data should be tested for stationarity before studying panel data. For nonstationary panel data, it is necessary to make the first difference smooth.³⁴ The results are shown in Table 2, which uses the Levin-Lin-Chu (LLC) and the Im-Pesaran-Shin (IPS) unit-root tests.

Table 2.

Panel-data unit-root tests.

Variable	LLC		IPS
Variable	Adjusted t	P-Value	t-bar	P-Value
CO₂	−10.327***	0.000	−1.902**	0.034
d(CO₂)	−2.662***	0.004	−3.224***	0.000
CE	−2.887***	0.002	−1.099	0.998
d(CE)	−5.960***	0.000	−4.645***	0.000
RGC	−9.671***	0.000	−1.443	0.612
d(RGC)	−1.572*	0.058	−2.058***	0.001
EDU	3.516	1.000	−0.913	0.999
d(EDU)	−19.314***	0.000	−3.991***	0.000
GIN	2.581	0.995	−0.554	1.000
d(GIN)	−6.123***	0.000	−3.972***	0.000
R&D	−2.769***	0.003	−3.448***	0.000
d(R&D)	−14.063***	0.000	−5.446***	0.000

Note: d(x) indicates the first difference of x, * p < 0.1, ** p < 0.05, *** p < 0.01.

4. Empirical results

4.1. Clean energy promotes Co₂ emission reduction

As shown in Table 3, a negative relationship is found between carbon dioxide emissions and clean energy. Columns 1 and 3 denote the random and fixed effects model without adding control variables. The coefficients of CE in Columns 1 and 3 are −0.269 and −0.286, respectively. Columns 2 and 4 denote the random and fixed effects model by adding control variables. The coefficients of CE in Columns 2 and 4 are −0.225 and −0.236, respectively. Column 5 represents a dynamic panel specification estimated by the (difference) GMM methodology. Column 6 represents a dynamic panel specification estimated by the (system) GMM methodology. The coefficients of CE in Columns 5 and 6 are −0.124 and −0.134, respectively. The results of Columns 5 and 6 are significant at the p = 0.01 level. First, we consider the results of static panel regression. For the fixed effects and random effects in the static panel, it can be seen from the results of the Hausman test in Column 4 that the p value is 0.002, which refuses the null hypothesis that the random effects are better, so the fixed effects are better than the random effects. Next, we consider the results of dynamic panel GMM regression. The consistency of the difference and system GMM estimators relies on the assumption that the error term does not have a serial correlation problem and the validity of instrument variables.³⁵

Table 3.
The effect of clean energy on CO₂ emissions.

CO₂ emissions

RE (1) RE (2) FE (3) FE (4) DGMM (5) SGMM (6)

CE −0.269** (0.132) −0.225* (0.117) −0.283** (0.134) −0.236* (0.116) −0.124* (0.042) −0.134* (0.049)

RGC −1.254* (0.450) −1.106 (0.487) 0.397* (0.093) 0.351 (0.175)

EDU 0.067* (0.012) 0.066* (0.012) 0.005** (0.002) 0.004 (0.003)

GIN −0.022* (0.008) −0.018 (0.008) −0.001 (0.006) −0.003 (0.007)

R&D 0.226 (0.095) 0.310 (0.147) −0.068** (0.029) −0.036* (0.021)

L1. y 0.799* (0.067) 0.860* (0.065)

L2. y 0.101* (0.057) 0.053 (0.065)

Intercept 5.361* (0.136) 2.698 (1.175) 5.363* (0.019) 1.631 (1.862) 1.429* (0.340) 0.968*** (0.267)

Hausman 5.350 21.030

p 0.069 0.002

Sargan 88.250 123.820

p 0.000 0.000

AR(2) 1.025 1.091

p 0.305 0.275

N 448 448 448 448 448 448

	CO₂ emissions
CE	−0.269** (0.132)	−0.225* (0.117)	−0.283** (0.134)	−0.236* (0.116)	−0.124*** (0.042)	−0.134*** (0.049)
RGC		−1.254*** (0.450)		−1.106** (0.487)	0.397*** (0.093)	0.351** (0.175)
EDU		0.067*** (0.012)		0.066*** (0.012)	0.005** (0.002)	0.004 (0.003)
GIN		−0.022*** (0.008)		−0.018** (0.008)	−0.001 (0.006)	−0.003 (0.007)
R&D		0.226** (0.095)		0.310** (0.147)	−0.068** (0.029)	−0.036* (0.021)
L1. y					0.799*** (0.067)	0.860*** (0.065)
L2. y					0.101* (0.057)	0.053 (0.065)
Intercept	5.361*** (0.136)	2.698** (1.175)	5.363*** (0.019)	1.631 (1.862)	1.429*** (0.340)	0.968*** (0.267)
Hausman			5.350	21.030
p			0.069	0.002
Sargan					88.250	123.820
p					0.000	0.000
AR(2)					1.025	1.091
p					0.305	0.275
N	448	448	448	448	448	448

Note: L1.y and L2.y denote the lag of the dependent variable. Reported standard errors in parentheses, we employ ***, **, and * to represent statistical significance at the 0.01, 0.05 and 0.1 levels.

The Arellano–Bond test is employed to detect no serial correlations in the error terms, and the Sargan test is employed to detect instrument variables.³⁵ At the bottom of Table 3, we display Arellano–Bond and Sargan test results for the difference and system GMM estimations. The results of the Arellano–Bond test indicate that the p value of AR (2) is greater than 0.1, accepting the null hypothesis of “no autocorrelation of disturbance items.” The p values of the Sargan test are all equal to 0, rejecting the null hypothesis that “all instrumental variables are valid” at the 1% significance level, so the instrumental variables selected by the (difference) and (system) GMM method are not valid. Therefore, the fixed effect regression of the static panel is more reliable than the other regression results.

However, it can be seen in Table 3 that regardless of the model, the coefficients of CE are significant and negative in all specifications, implying that the negative correlation between clean energy and CO₂ emissions is robust Taken together, the results in Table 3 indicate that there is an association between clean energy and carbon dioxide emissions. Our empirical findings confirm that advancing clean energy use can help reduce carbon dioxide emissions. The results in Table 3 imply important policy implications for conducting clean energy-use strategies in fighting carbon dioxide emissions.

4.2. The effect of emission trading policy on clean energy

Table 3 shows that the fixed effect regression of the static panel is more reliable than the other regression results. Thus, in Table 4, we exploit the fixed effect regression of the static panel to explore the correlation between emission trading policy and clean energy. As the control variable increases, the coefficient value of the CORE gradually decreases. Table 4 compares the summary statistics for the relationship between emission trading policy and clean energy. The CORE coefficients in Table 4 are 1.278, 1.072, 0.950, 0.945, and 0.926. The results of Table 4 display that the CORE coefficients are significant and positive in all specifications. The interaction term $C O R E = d μ * y e a r$ denotes the “policy treatment effect.” In 2011, China approved seven provinces and cities to carry out carbon trading pilots and started emissions trading pilot work. Thus, this paper sets 2011 as the initial pilot time (year = 1, if t ≥ 2011; year = 0, if t < 2011). Furthermore, we choose provinces (or municipalities) that have participated in the pilot policy since 2011 as pilot areas ( $d μ = 1$ , if $μ$ is the pilot province; otherwise, $d μ = 0$ ).

Table 4.
The impact of emission trading policy on clean energy.

Clean energy

(1) (2) (3) (4) (5)

CORE 1.278* (0.111) 1.072* (0.092) 0.950* (0.091) 0.945* (0.099) 0.926*** (0.106)

RGC −3.329* (0.603) −3.669* (0.550) −3.975* (0.632) −3.579* (0.692)

EDU 0.116* (0.023) 0.116* (0.023) 0.116*** (0.024)

GIN −0.052 (0.020) −0.046 (0.019)

R&D 0.279*** (0.099)

Intercept 5.984* (0.008) 6.484* (0.075) 6.511* (0.067) 6.589* (0.087) 3.059** (1.272)

N 448 448 448 448 448

F 133.499 87.667 61.768 41.235 35.637

	Clean energy
CORE	1.278*** (0.111)	1.072*** (0.092)	0.950*** (0.091)	0.945*** (0.099)	0.926*** (0.106)
RGC		−3.329*** (0.603)	−3.669*** (0.550)	−3.975*** (0.632)	−3.579*** (0.692)
EDU			0.116*** (0.023)	0.116*** (0.023)	0.116*** (0.024)
GIN				−0.052** (0.020)	−0.046** (0.019)
R&D					0.279*** (0.099)
Intercept	5.984*** (0.008)	6.484*** (0.075)	6.511*** (0.067)	6.589*** (0.087)	3.059** (1.272)
N	448	448	448	448	448
F	133.499	87.667	61.768	41.235	35.637

Note: Reported standard errors in parentheses. We employ ***, **, and * to represent statistical significance at the 0.01, 0.05 and 0.1 levels.

Our findings indicate a positive correlation between clean energy and the emission trading policy, implying that the emission trading policy promotes clean energy development. The policy pressure of carbon trading in the pilot areas promotes the development of clean energy technologies in these areas, gradually changing the traditional energy consumption structure and accelerating the replacement of traditional energy by clean energy. Therefore, the trading policy is conducive to the development of clean energy. In addition, this paper has proven that clean energy can be conducive to carbon dioxide emissions reduction, as shown in Table 3. Thus, we find that the carbon emissions trading policy can indirectly reduce CO₂ emissions by promoting clean energy development. In the next section, this paper further uses the DID model and the PSM-DID method to examine the effect of China’s current emission trading policy on carbon dioxide emissions and explore whether the trading policy can help mitigate carbon dioxide emissions. Our empirical findings are helpful for the government in making decisions regarding carbon dioxide emissions based on greenhouse gas control.

4.3. Quasi-natural experiments of Co₂ emissions

The carbon emission trading system in China has made significant progress.³⁶ In 2011, China approved seven provinces and cities to carry out carbon trading pilots and started emissions trading pilot work.⁵ In policy effects research, the DID model can identify the “policy effect.” In this section, provinces (or municipalities) launching carbon trading pilots are chosen as the treatment group; others are chosen as the control group.³⁷ If a province (or a municipality) belongs to the pilot areas, then $d μ = 1$ ; otherwise, $d μ = 0$ . This paper sets 2011 as the beginning of the pilot time. If $y e a r$ is in the policy implementation period, then $y e a r = 1$ ; otherwise, $y e a r = 0$ . Thus, for time before 2011, $y e a r = 0$ ; otherwise, $y e a r = 1$ . The interaction term $C O R E = d μ * y e a r$ denotes the “policy treatment effect.” The constructed DID model is as follows:

c a r b o n_{i t} = α_{0} + α_{1} d μ_{i t} + γ_{0} y e a r_{i t} + γ_{1} d μ_{i t} * y e a r_{i t} + Σ b_{j} c o t_{i t} + ε_{i t}

(1)

The general DID model essentially regresses the interaction terms of time and regional dummy variables, so it can effectively avoid the problem of endogeneity, but the model cannot address sample selection bias. However, propensity score matching (PSM) can effectively avoid the problem of sample selection bias, but it cannot deal with the problem of endogeneity.^38,39 Thus, this paper exploits the PSM-DID method to acquire more robust conclusions. The constructed PSM-DID model is:

c a r b o n_{i t}^{P S M} = α_{0} + α_{1} d μ_{i t} + γ_{0} y e a r_{i t} + γ_{1} d μ_{i t} * y e a r_{i t} + Σ b_{j} c o t_{i t} + ε_{i t}

(2)

c a r b o n_{i t}

denotes the carbon dioxide emissions indicator without PSM;

c a r b o n_{i t}^{P S M}

denotes the carbon dioxide emissions indicator with PSM;

c o t_{i t}

represents a set of control variables;

α_{1},

γ_{0},

γ_{1},

and

Σ b_{j}

are estimated parameters;

d μ_{i t}

is a dummy variable for the pilot areas;

y e a r_{i t}

is a dummy variable for the policy time;

d μ_{i t} * y e a r_{i t}

denotes a double-difference variable; and

ε_{i t}

denotes the error term.

The key premise of the DID model is the parallel trend hypothesis. This paper conducts a parallel trend test on the selected treatment group and control group of CO₂ emissions, and the method is as follows:

① Graphical method of intuitive judgement. A time trend graph of CO₂ emissions in pilot and no pilot areas was drawn.

② Annual dynamic effect analysis. The event research method is employed to test the annual dynamic effects of policy. The specific approach is to take the previous period of 2011 as the base period, construct period dummy variables, multiply each period dummy variable with the policy dummy variable, and incorporate it into the DID model for estimation. The empirical results are shown in Figure 1, Figure 2, and Table 5. The above two inspection methods both indicate that the carbon dioxide emissions of the treatment group and the control group have no difference with the year before the emission trading policy is issued.

Figure 1.

Parallel trend chart.

Figure 2.

Annual dynamic effect.

Table 5.

Annual dynamic effect test.

	Coef.	Std. Err.	T-Value	P-Value
Pre_4	0.007	0.081	0.56	0.765
Pre_3	0.009	0.052	0.18	0.862
Pre_2	0.079	0.055	1.44	0.163
Current	−0.047	0.028	−1.67	0.107
Post_1	−0.109***	0.031	−3.50	0.02
Post_2	−0.202***	0.063	−3.23	0.003
Post_3	−0.247***	0.075	−3.27	0.003
Post_4	−0.283***	0.078	−3.61	0.001

Note: We set 2011 as “the year of the policy occurred”, 2008 as Pre_4, 2009 as Pre_3, 2010 as Pre_2, 2011 as Current, 2012 as Post_1, 2013 as Post_2, 2014, 2014 as Post_3, and 2015 as Post_4.

Table 5 displays that the regression results before 2011 (current) are not significant, which meets the assumption that the treatment group and the control group have a similar trend before the policy begins. The treatment group after 2011 shows that the coefficients are −0.109, −0.202, −0.247, and −0.283. The results of Table 5 verified the basic regression results. Before 2011, the treatment group and control group had similar development trends.

The dependent variable of Models (1) to (5) is carbon dioxide emissions. Table 6 indicates that the coefficients of $D i D$ are significant and negative in all specifications. The estimated coefficients imply that emission trading policy induces a decline in carbon dioxide emissions. The coefficients of $D i D$ in Columns 1 to 6 are negative and significant, indicating that the emission trading market pilot has significantly reduced carbon dioxide emissions. Our findings imply that the current policy effectively controls CO₂ emissions and positively impacts environmental protection.

Table 6.

Difference in differences (DID) regression results.

Variable	(1)	(2)	(3)	(4)	(5)
$D i D$	−0.354** (0.141)	−0.332** (0.144)	−0.352** (0.145)	−0.315** (0.146)	−0.311** (0.147)
RGC		0.361 (0.615)	0.134 (0.604)	−0.356 (0.650)	−0.270 (0.653)
EDU			0.050*** (0.015)	0.054*** (0.015)	0.053*** (0.015)
GIN				−0.063* (0.033)	−0.063* (0.033)
R&D					0.071*** (0.026)
Intercept	5.098*** (0.051)	5.102*** (0.106)	5.135*** (0.105)	5.265*** (0.126)	4.380*** (0.338)
N	448	448	448	448	448
F	51.238	34.261	31.456	26.718	23.188

Note: Reported standard errors in parentheses. We employ ***, **, and * to represent statistical significance at the 0.01, 0.05 and 0.1 levels.

This paper uses the PSM-DID method for further robustness tests to address sample selection bias. The PSM approach was employed to look for the control group with the features closest to the treatment group. First, this paper matches samples from 2011 to 2016 and creates the different groups. Second, by exploiting grouping variables and characteristic variables, this paper matches the data. Third, this paper employs “kernel matching” and uses the logit model to appraise propensity scores. R&D, GIN, EDU, and RGC are set as the matching characteristic variables.

Figure 3 (a) depicts the core density map of the different groups’ propensity scores before matching. Figure 3(b) depicts the propensity scores of the two groups after matching. By comparing Figure 3(a) and (b), we find that the common tendency hypothesis can be met. Because the different groups’ propensity scores overlapped after matching, their distribution trends tend to be consistent.

Figure 3.

Kernel density distribution.

In Table 7, the PSM-DID results display a significant negative correlation between carbon dioxide emissions and China’s emissions trading policy (t = 3.32, p = 0.001). We compare the results of Table 7 and Table 6 and find that the DID results in Table 6 are robust The results of Table 6 and Table 7 show that the emission trading policy has indeed achieved CO₂ emissions reduction effects. Considering China’s current situation, the empirical findings might be caused by the following factors. The emission caps in pilot provinces are strict, which would deliver pressure to reduce emissions more directly to policy receivers. Regulated industries face tremendous pressure to decrease emissions in pilot provinces. In addition, the trading policy would facilitate the withdrawal of companies with severe pollution and accelerate the flow of capital to industries with high efficiency, thus realizing carbon dioxide emission reduction and energy conversion. Firms are granted agility in their compliance strategy to achieve carbon dioxide reduction missions through purchase quotas, and these efforts could optimize the allocation of factors and resources between different industries.

Table 7.

Propensity score matching-DID.

	Before Diff	After Diff	DID
Diff	−0.340	−0.805	−0.465***
St errors	0.085	0.112	0.140
t	−4.020	7.22	3.32
p	0.000	0.000	0.001

Note: *** p < 0.01.

5. Conclusion and policy implications

5.1. Conclusion

The empirical results of static and dynamic models show a negative association between clean energy and CO₂ emissions. Our findings imply that clean energy use validly controls CO₂ emissions and positively impacts environmental protection. Empirical results further indicate a positive association between clean energy and emission trading policy, implying that emission trading promotes clean energy development. The empirical results of quasi-natural experiments show a negative correlation between the “policy treatment effect” and CO₂ emissions; the emission trading policy has indeed achieved CO₂ emissions reduction effects.

Overall, the unique finding of this paper is that not only does the ETS pilot policy reduce carbon dioxide emissions, but it can reduce carbon dioxide emissions by stimulating clean energy development. In addition, both emission trading policies and clean energy use can mitigate carbon dioxide emissions. The results of this paper can provide useful inspiration for the formulation of emission trading policies and the development of clean energy, helping the government to build a carbon trading market and promote green energy transition, as well as helping green innovation for high-polluting enterprises, so as to achieve a low-carbon economy and green environmental protection target. Our findings are helpful for gaining insight into the trade-off among clean energy use, ETCs, and a low-carbon economy.

5.2. Policy implications

Enterprises should actively respond to the government’s carbon emission trading policies and clean energy policies, promote green technology innovation, exploit energy-efficient technologies, and employ clean energy sources to reduce carbon dioxide emissions. The Porter hypothesis states that environmental regulations that are rigid and flexible may create an incentive for firms to innovate. As a result of the compliance pressure on the carbon emissions trading policy and the additional income generated from the sale of emission allowances, enterprises might be more motivated to invest in clean energy technologies, leading to a reduction in traditional energy use and carbon dioxide emissions. In addition, enterprises can optimize resource allocation through energy conversion and allowance trading, thereby affecting the energy structure to achieve emission reduction targets.

The government should implement emission trading policies reasonably, encourage clean energy, develop a low-carbon economy, protect the environment, and take the path of sustainable development. Next, we put forward the following policy recommendations for the government: First, the government should reduce emission trading market risk, improve emission trading market efficiency, implement a more efficient emission trading policy, exploit a low-carbon economy, and take the path of sustainable development. Second, government departments should strengthen the management and guidance of the carbon emission trading market. In further improving the design and supporting policies of the carbon emission trading market, attention should be given to the coordinated development of carbon emission trading policies and clean energy policies. The synergy between the two should be effectively utilized to avoid potential policy conflicts. Third, a developed clean energy system that can promote CO₂ emissions reduction should be encouraged in each region and the surrounding areas. Government departments should heighten energy use efficiency and expand clean energy provision channels to relieve the pressure on the energy demand of individuals or enterprises. Finally, government departments should optimize their energy consumption system, reduce the dependence on traditional energy sources, and vigorously develop clean energy.

Footnotes

Declaration of conflicting interests.

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding.

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Fei Yang

Notes

Fei Yang is a PhD student in the School of Economics and Management at Wuhan University. His research interests are focused on quantitative economics. He has been working on combining economics, mathematics, statistics, and computer technology and uses economic mathematical models to study quantitative economic relationships.

Chunchen Wang belongs to the School of Information Management at Jiangxi University of Finance and Economics. Her research interests are quantitative economic analysis of innovation and economic growth.

References

Ren

Wang

, et al. Can carbon emission trading scheme achieve energy conservation and emission reduction? Evidence from the industrial sector in China. Energy Econ 2020; 85: 104590.

Hammoudeh

Nguyen

Sousa

. What explain the short-term dynamics of the prices of CO₂ emissions? Energy Econ 2014; 46: 122–135.

Yang

Wang

. Clean energy, financial development, and economic growth: evidence from spatial spillover effects and quasi-natural experiments. J Cleaner Prod 2021; 322: 129045.

Chen

Shi

. Clean energy development research. Shanghai: Shanghai University of Finance and Economics Press, 2011, p. 22.

Zhang

, et al. Emission reduction effect and carbon market efficiency of carbon emissions trading policy in China. Energy 2020a; 196: 117117.

Streimikiene

Roos

. GHG Emission trading implications on energy sector in Baltic States. Renew Sustain Energy Rev 2009; 13: 854–862.

Möst

Fichtner

. Renewable energy sources in European energy supply and interactions with emission trading. Energy Pol 2010; 38: 2898–2910.

Lin

Jia

. Is emission trading scheme an opportunity for renewable energy in China? A perspective of ETS revenue redistributions. Appl Energy 2020; 263: 114605.

Wang

Zhang

. Research on the dynamic linkage among the carbon emission trading, energy and capital markets. J Cleaner Prod 2020; 272: 122717.

10.

Al-mulali

Binti Che Sab

. The impact of energy consumption and CO₂ emission on the economic and financial development in 19 selected countries. Renew Sustain Energy Rev 2012; 16: 4365–4369.

11.

Shahbaz

Hye

QMA

Tiwari

, et al. Economic growth, energy consumption, financial development, international trade and CO₂ emissions in Indonesia. Renew Sustain Energy Rev 2013; 25: 109–121.

12.

Mahdi Ziaei

. Effects of financial development indicators on energy consumption and CO₂ emission of European, East Asian and Oceania countries. Renew Sustain Energy Rev 2015; 42: 752–759.

13.

Shahzad

SJH

Kumar

Zakaria

, et al. Carbon emission, energy consumption, trade openness and financial development in Pakistan: a revisit. Renew Sustain Energy Rev 2017; 70: 185–192.

14.

Ehigiamusoe

Lean

. Effects of energy consumption, economic growth, and financial development on carbon emissions: evidence from heterogeneous income groups. Environ Sci Pollut Res 2019; 26: 22611–22624.

15.

Khan

Teng

J-Z

Khan

, et al. Impact of globalization, economic factors and energy consumption on CO₂ emissions in Pakistan. Sci Total Environ 2019; 688: 424–436.

16.

Ahmed

Apergis

Bhattacharya

, et al. Electricity consumption in Australia: the role of clean energy in reducing CO₂ emissions. Appl Econ 2021; 53: 5535–5548.

17.

Maji

. Impact of clean energy and inclusive development on CO₂ emissions in sub-Saharan Africa. J Cleaner Prod 2019; 240: 118186.

18.

Ummalla

Goyari

. The impact of clean energy consumption on economic growth and CO₂ emissions in BRICS countries: does the environmental Kuznets curve exist? J Public Aff 2021; 21: e2126.

19.

Cai

Sam

Chang

. Nexus between clean energy consumption, economic growth and CO2 emissions. J Cleaner Prod 2018; 182: 1001–1011.

20.

Fei

Rasiah

. The clean energy-growth nexus with CO₂ emissions and technological innovation in Norway and New Zealand. Energy Environ 2014; 25: 1323–1344.

21.

Rahman

Alam

. Clean energy, population density, urbanization and environmental pollution nexus: evidence from Bangladesh. Renew Energy 2021; 172: 1063–1072.

22.

Anger

. Including aviation in the European emissions trading scheme: impacts on the industry, CO2 emissions and macroeconomic activity in the EU. J Air Transp Manag 2010; 16: 100–105.

23.

Chappin

EJL

Dijkema

GPJ

. On the impact of CO₂ emission-trading on power generation emissions. Technol Forecast Soc Change 2009; 76: 358–370.

24.

Jiang

. The construction of Shenzhen’s carbon emission trading scheme. Energy Pol 2014; 75: 17–21.

25.

Nguyen

Chapman

Farabi-Asl

. Nation-wide emission trading model for economically feasible carbon reduction in Japan. Appl Energy 2019; 255: 113869.

26.

Wallace

Wang

. Can an Emission Trading Scheme really reduce CO₂ emissions in the short term? Evidence from a maritime fleet composition and deployment model. Transp Res D, Transp Environ 2019; 74: 318–338.

27.

Shen

Tang

Zeng

. Does China’s carbon emission trading reduce carbon emissions? Evidence from listed firms. Energy Sustain Dev 2020; 59: 120–129.

28.

Zhang

Wang

Hao

, et al.

Shooting two hawks with one arrow: could China’s emission trading scheme promote green development efficiency and regional carbon equality?

Energy Econ 2021; 101: 105412.

29.

Chen

Lin

. Towards carbon neutrality by implementing carbon emissions trading scheme: policy evaluation in China. Energy Pol 2021; 157: 112510.

30.

Chen

Shen

. Clean energy development, carbon dioxide emission reduction and regional economic growth. Econ Res J 2019; 7: 188–202.

31.

Zhao

Jiang

Dong

, et al. Would environmental regulation improve the greenhouse gas benefits of natural gas use? A Chinese case study. Energy Econ 2020; 87: 104712.

32.

Sun

, et al. Green innovation effect of emission trading policy on pilot areas and neighboring areas: an analysis based on the spatial econometric model. Energy Pol 2021; 156: 112431.

33.

Coe

Helpman

. International R&D spillovers. Eur Econ Rev 1995; 39: 859–887.

34.

Al-mulali

Tang

. Investigating the validity of pollution haven hypothesis in the gulf cooperation council (GCC) countries. Energy Pol 2013; 60: 813–819.

35.

Carvallo

Kasman

Kontbay-Busun

. The latin American bank capital buffers and business cycle: are they pro-cyclical? J Int Financial Mark Inst Money 2015; 36: 148–160.

36.

Zhang

Luo

, et al. The effect of emission trading policy on carbon emission reduction: evidence from an integrated study of pilot regions in China. J Cleaner Prod 2020b; 265: 121843.

37.

Shaw

S-L

Fang

, et al. Impacts of high speed rail on railroad network accessibility in China. J Transp Geogr 2014; 40: 112–122.

38.

Heckman

. The common structure of statistical models of truncation, sample selection and limited dependent variables and a simple estimator for such models. Ann Econ Soc Meas 1976; 5: 475–492.

39.

Rosenbaum

Rubin

. The central role of the propensity score in observational studies for causal effects. Biometrika 1983; 70: 41–55.

	CO₂ emissions
	RE (1)	RE (2)	FE (3)	FE (4)	DGMM (5)	SGMM (6)
CE	−0.269** (0.132)	−0.225* (0.117)	−0.283** (0.134)	−0.236* (0.116)	−0.124*** (0.042)	−0.134*** (0.049)
RGC		−1.254*** (0.450)		−1.106** (0.487)	0.397*** (0.093)	0.351** (0.175)
EDU		0.067*** (0.012)		0.066*** (0.012)	0.005** (0.002)	0.004 (0.003)
GIN		−0.022*** (0.008)		−0.018** (0.008)	−0.001 (0.006)	−0.003 (0.007)
R&D		0.226** (0.095)		0.310** (0.147)	−0.068** (0.029)	−0.036* (0.021)
L1. y					0.799*** (0.067)	0.860*** (0.065)
L2. y					0.101* (0.057)	0.053 (0.065)
Intercept	5.361*** (0.136)	2.698** (1.175)	5.363*** (0.019)	1.631 (1.862)	1.429*** (0.340)	0.968*** (0.267)
Hausman			5.350	21.030
p			0.069	0.002
Sargan					88.250	123.820
p					0.000	0.000
AR(2)					1.025	1.091
p					0.305	0.275
N	448	448	448	448	448	448

Clean energy,emission trading policy,and CO 2 emissions: Evidence from China

Abstract

Keywords

1. Introduction

2. Related literature

2.1. Emission trading scheme, energy, and Co2 emissions

2.2. Clean energy and Co2 emissions

2.3. Emission trading scheme and Co2 emissions

2.4. Discussion

3. Data description

3.1. Main variables

3.2. Control variables

4.1. Clean energy promotes Co2 emission reduction

5.1. Conclusion

5.2. Policy implications

Footnotes

Declaration of conflicting interests.

Funding.

ORCID iD

Notes

References

2.1. Emission trading scheme, energy, and Co₂ emissions

2.2. Clean energy and Co₂ emissions

2.3. Emission trading scheme and Co₂ emissions

4.1. Clean energy promotes Co₂ emission reduction