The Impact of Carbon Emission Trading on Industrial Green Total Factor Productivity

Abstract

The impact of carbon emission trading (ETS) has been widely investigated. In contrast to the existing studies, this article explores for the first time the theoretical relationship between the ETS and industrial green total factor productivity (IGTFP) and tests it empirically. Furthermore, the article investigates the mediating mechanisms and possible regional heterogeneity of the influence of the ETS on IGTFP. To this end, a time-varying difference-in-differences technique is employed, drawing on panel data of 276 cities in China from 2005 to 2019. The results show that the ETS significantly and robustly increases IGTFP in pilot regions. Green technological innovation and industrial structure have a significant mediating effect on the nexus of the ETS and IGTFP. However, there exists no mediating mechanism of general technological innovation. In addition, economic development and energy consumption positively moderate the effect of the ETS on IGTFP, and industrial dependence negatively moderates such an effect.

1. Introduction

The emissions trading scheme (ETS) is the world’s leading emissions reduction policy [1,2], including the E.U. ETS, Korean ETS, Australian ETS, and Chinese ETS, to name a few. ETSs are an important institutional innovation that use market mechanisms to control and reduce greenhouse gas emissions and promote the green and low-carbon transformation of economic development methods. Generally speaking, regulators set and allocate emission allowances to participants under a system of aggregate control and allowance trading. They combine environmental performance and flexibility through market-based trading instruments to make it possible for participants to achieve emission reduction goals at the least cost. Extensive research has confirmed the significant impact of ETSs on carbon emissions and energy consumption, such as the findings of Bel and Joseph [3], Lise et al. [4], and Schäfer [5]. At the same time, the economic impact of ETSs has also received considerable attention from the scientific community, including authors such as Choi et al. [6], Koch and Themann [7], and Nong et al. [8].

The dual role in economic growth and emissions reduction organically links ETSs to green total factor productivity (GTFP) [9]. Subject to limited resources and increasingly serious environmental pollution, resources and environment are no longer only endogenous variables to promote but are also constraints to limit economic development, so the concept of green development is proposed. Therefore, GTFP, which takes pollutant emissions into account in the output indicator system, is more acceptable than total factor productivity, which only considers economic output. Moreover, ETSs mainly cover industrial enterprises [10,11], which are the primary sources of carbon emissions [12]. In this sense, the ETS has the most direct impact on industrial enterprises, thus leading to a new important question that is different from regional GTFP: whether an ETS has influenced the industrial green total factor productivity (IGTFP). In this context, Chen and Hibiki [10] explored the effect of ETSs on IGTFP and IGTFP’s calculated indicators. However, the underlying remains unknown. Moreover, in the heterogeneity analysis of the effect of ETSs on IGTFP, it is important to figure out which factors are driving such a heterogeneous effect, but this has also been neglected in existing studies. Given this background, this paper investigates the effects of ETSs on IGTFP and the underlying mechanisms. In other words, this study answers two fundamental theoretical questions: one is the effects of an ETS on IGTFP, and the other is the association mechanism between an ETS and IGTFP.

Unlike the existing studies, this study has the following significant theoretical and practical contributions. Firstly, it investigates the impact of an ETS on IGTFP, which enriches the research spectrum of ETS impact. Secondly, and more importantly, the paper discusses the possible mediating mechanisms and regional heterogeneity of the ETS’s effects on IGTFP, thus enriching and deepening the theoretical understanding of the link between the ETS and IGTFP. Specifically, we reveal how the ETS enhances IGTFP through changes in technological innovation, green technology innovation, and industrial structure. We also explore the effects of economic development, energy consumption, and industrial dependence on the nexus of the ETS on IGTFP. Finally, this study provides key policy recommendations for enhancing the positive impact of ETSs on IGTFP, which provides a solid reference for the broader diffusion of ETSs and the green transition of the industrial sectors.

2. Literature Review and Theoretical Hypotheses

The E.U. ETS was implemented earlier and has more influence among the global ETS systems. De Perthuis and Trotignon [13], Flachsland et al. [14], and Verde et al. [15] discussed the working mechanisms of the E.U. ETS in terms of subsidy allocation, international convergence, environmental policy convergence, and price floors. In addition, Brohé and Burniaux [16], Borghesi et al. [17], Koch and Themann [7], and Stuhlmacher et al. [18] estimated the impact of the E.U. ETS on carbon emissions, low-carbon investment and innovation, and firm productivity. Apart from the E.U. ETS, Chevallier [19], Choi et al. [6], Nong et al. [8], and Wakabayashi and Kimura [20] confirmed the economic and carbon emission impacts of ETSs (including related information) in Australia, Korea, Vietnam, and metropolitan Tokyo (Japan), respectively. With the promotion of the ETS in China, research on ETSs in China has continued to grow in recent years. These studies have also shown that the ETS reduces China’s carbon emissions and energy consumption [21]. In addition, the research also refers to the effects on corporate profitability [22], low-carbon investment [23], income inequality [24], and haze pollution concentration [25].

Unlike total factor productivity, GTFP indicates the harmonious development of regional resources, the environment, and the economy and guides economic development to incorporate the reasonable consumption of energy and resources and the reduction in environmental pollution [26,27]. Since the core concept of low-carbon transition is to reduce energy consumption and carbon emissions, green total factor productivity becomes increasingly important in the ETS context. Therefore, the accounting of GTFP and factors influencing it have also received considerable scholars’ attention. For example, Lin and Chen [26], Lee and Lee [27], Song et al. [28], Cao et al. [29], Xie et al. [30], and Jin et al. [31], respectively, evidenced the effects of factor market distortion, green finance, fiscal decentralization, e-commerce development, energy consumption transition, and political competition on GTFP. Furthermore, given the significant economic and environmental impacts of the ETS on the pilot regions, Zhang et al. [32] confirmed that the DEA efficiency of the carbon trading market in the ETS pilot regions increases significantly. Li et al. [9] recently showed that the ETS significantly increases GTFP in the pilot regions.

Due to the different positions and rhythms of various economic sectors in green development, some scholars have calculated the GTFP of different sectors separately, such as extractive industries [33], agriculture [34], and the metal industry [35]. In addition, Zhong et al. [36] calculated the GTFP of the overall industrial sector in the Chinese Yangtze River economic belt. Although its coverage is expanding, the Chinese pilot ETS covers mainly the industrial sector [37,38,39,40,41], which accounts for the biggest part of China’s carbon emissions. The national ETS launched in 2021 also focuses only on the crucial industrial sector: the power sector. Moreover, the future expansion of the national ETS will also be focused on the industrial sector [42,43]. Therefore, the economic and environmental impacts of China’s ETS are actually more concentrated in the industrial sector [44,45,46]. Thus, the change in the industrial sector is significant when looking at the impact of the ETS. Thus, combining the above significant effects of the ETS, the ETS coverage of the industrial sector, and the definition of GTFP, the nexus of IGTFP and the ETS constitutes a fundamental theoretical and practical question in the context of green development. Considering the existing discoveries and the practice of ETSs in China, we attempt to explain the theoretical relationship between ETSs and IGTFP and thus derive the basic hypothesis of this article:

Hypothesis 1 (H1).

The ETS significantly affects IGTFP.

Technological innovation is an essential mechanism in the process of ETSs. Theoretically, an ETS stimulates the development of new technologies in manufacturing engineering, energy efficiency, and carbon sequestration to reduce carbon emissions. New technological advances likewise increase productivity and reduce environmental pollution, thus helping to improve IGTFP. Technological innovation is also a means to enhance resource allocation and production efficiency and thus effectively increases the level of green total factor productivity [47,48]. However, Bel and Joseph [49] asserted that a large surplus of quota allocation would lead to a very limited or even negative impact on technological innovation. In particular, the advancement of green and low-carbon technologies is one of the more desired outputs of ETSs, as opposed to general technological innovation. Technological innovation, which refers to the innovation of production technologies, includes new technologies development, or the application of existing technologies to innovation. The purpose of green technology innovation is to preserve the environment, also known as ecotechnology innovation. Thus, the impact of ETSs on green technological innovation has been widely confirmed by existing studies.

For instance, Rogge and Hoffmann [50] found that the E.U. ETS influences the speed and direction of technological innovation in power generation technologies. Teixidó et al. [51] identified a greater increase in innovation and adoption of low-carbon technologies due to the reform of the E.U. ETS. Similarly, Gao and Wang [52] argued that appropriate ETS design can effectively encourage firms to invest in low-carbon technologies, and Liu and Sun [53] concluded that China’s pilot ETS could significantly contribute to low-carbon technology innovation. However, some studies deviate from these above findings. For example, in some resource-based industrial businesses, Zhao et al. [54] discovered significant green innovation improvements rather than any evidence that the pilot ETS had an effect on the industry or all manufacturing industries. Chen et al. [55] even pointed out that the Chinese ETS inhibits the development of green innovation. Their argument was that businesses primarily choose to reduce output under the ETS framework rather than promote green technology innovation. Therefore, we further hypothesize:

Hypothesis 2 (H2).

Technological innovation mediates the effects of ETSs on IGTFP significantly.

Hypothesis 3 (H3).

Green technological innovation mediates the effects of ETSs on IGTFP significantly.

Obviously, ETSs bring additional operating costs to industrial enterprises, which will cause a certain shift of resources and investments away from industrial enterprises to other industries, mainly to the tertiary sector, i.e., changing the industrial structure. This change will lead to a decrease in industrial value added and industrial inputs such as human, capital, and energy, which are the main input–output indicators of IGTFP; therefore, a change in industrial structure may lead to the IGTFP change. On the other hand, ETSs may also lead to the industrial upgrading of industrial firms due to technological, managerial, and product improvements. Several studies have confirmed the above-mentioned impact of ETSs on industrial structure. For example, as a result of the implementation of the E.U. ETS, Zang et al. [56] discovered that the industrial structure is being upgraded and that its impact is increasing.

Additionally, Hu et al. [44] and Liu et al. [57] discovered that ETSs reduce emissions by restructuring industries; Tan et al. [58] figured out that ETSs improve energy use efficiency by changing industrial structures as well. As an important stimulus for economic growth, industrial restructuring determines the direction of economic development and the quality of environmental protection, which in turn becomes a meaningful way to enhance green total factor productivity [59]. Meanwhile, upgrading and adjusting industrial structures will promote the continuous optimization of resource allocation and improvement of production efficiency, effectively supporting the increase in GTFP [60]. Therefore, we also hypothesize:

Hypothesis 4 (H4).

Industrial structures significantly mediate the effects of ETSs on industrial green total factor productivity.

In addition to the four hypotheses mentioned above, the variability of the operating mechanisms of the ETS in different pilot regions may cause its impact on the IGTFP to be regionally different. For example, in Beijing and Shanghai, the ETS covers enterprises (units) that emit 10,000 tons or more of CO₂ directly or indirectly per year. In Tianjin, enterprises with annual integrated energy use of more than 10,000 tons of standard coal will be included in the ETS. Differently, Shenzhen has fewer high-carbon industries, and the standard setting of ETS-covered industries is mainly focused on the construction industry. In addition, with respect to quota allocation, significant differences still exist between ETS pilot regions, with multiple coexisting allocation methods, including grandfathering, benchmarking, and auctions [61].

Regional heterogeneity of the impact of the ETS on IGTFP is also theoretically caused by the different socioeconomic characteristics of the pilot regions, such as economic development, industrial characteristics, and energy use. The heterogeneous effect of the ETS has already received extensive attention, such as the research by Li et al. [9] and Zhang and Zhang [24]. Li et al. [9] analyzed the heterogeneous effects of ETSs on regional green total factor productivity in cities with different types of industrial structures, while Zhang and Zhang [24] examined the heterogeneous effects of ETSs on income inequality based on different spatial clusters. In contrast to existing studies, we examine whether the impact of ETSs on IGTFP varies significantly across different cities with various economic growth, energy use, and industrial dependence, additionally considering that the operational mechanisms of ETSs in different regions are also influenced by regional socioeconomic characteristics [42]. Conclusively, Figure 1 shows the theoretical framework of this article.

3. Methodology

Similar to several studies that have looked at the effects of ETSs on IGTFP, we employ a quasinatural, empirical, difference-in-differences (DID) approach. In October 2011, seven provinces and cities launched local pilot projects for carbon emission trading. However, it was not until 2013 that China’s ETS was first officially piloted in Shenzhen, Guangdong Province, and later extended to the above-mentioned provinces and cities and Fujian Province. The regional pilot ETS ran until July 2021, when China’s nationwide ETS for the power sector went live. During the pilot period, China’s ETS had a significant impact on the economic and environmental aspects of the pilot regions. Considering the COVID-19 pandemic, we examine the impact of the ETS on the IGTFP as of 2019. We develop the following time-varying DID model to estimate the ETS’s impact on the IGTFP:

$$y_{i,t}=\alpha +\theta di{d}_{i,t}+{\displaystyle \sum _{k=1}^n{\beta }_{k}co{n}_{k,it}}+{\gamma }_i+{\lambda }_t+{\epsilon }_{i,t},$$

(1)

where y denotes the IGTFP; i and t denote the city and time, respectively; γ_i and λ_t represent the region-fixed effects and time-fixed effects, respectively; did denotes the policy dummy variable; and con denotes control variables. The coefficient θ is the policy effect to be estimated, i.e., the effect of the ETS on the IGTFP. According to the idea of the DID method and the actual implementation of the ETS policy in China, seven provinces were selected as the experimental group, while other regional cities constitute the control group. In addition, since the ETS policy was implemented at different times in different regions, the did is 1 if the city belongs to the experimental group in the treatment period; otherwise, the did is 0.

Since the IGTFP is not reported in the existing statistical system, we manually measured the IGTFP for each city. First, similar to Lee and Lee [27], Li et al. [9], and Zhong et al. [36], we identified GDP as the desired output indicator; various pollutant emissions as the undesired output indicator; and labor, capital, and energy as input indicators. Accordingly, the input–output variables involved in the calculation of IGTFP included industrial value added, wastewater emissions, SO₂ emissions, dust emissions, labor inputs, capital inputs, and energy inputs. Industrial-value-added data were obtained from iFinD (http://www.51ifind.cn/index.php?c=index&a=home accessed on 17 March 2022). Since there is no GDP price index for prefecture-level cities, we used the GDP price index of each province to obtain the real industrial value added, using 2005 as the base period. We used the entropy method to integrate three pollutant emissions as the nondesired output. Industrial labor is the number of persons employed in industrial sectors at year end. We used industrial fixed capital stock to represent capital inputs that are estimated using the following perpetual inventory method:

$$ICA{P}_t=I_t+(1-\delta )ICA{P}_{t-1}$$

(2)

where ICAP_t is the annual fixed capital stock of each city, I_t is the annual new industrial fixed asset investment of each city, δ is the fixed capital depreciation rate, and ICAP_t−1 is the fixed capital stock of each city in the previous year. Using a 10% division of the city’s 2005 industrial fixed asset investment, each city’s initial industrial capital stock is measured. The annual amount of industrial fixed assets investment in each city is determined by the product of the city’s total social fixed assets investment and the ratio of industrial value added to the GDP in that year. As a rule, the capital natural depreciation rate is 9.6%. The fixed capital stock of each city also depreciated according to the fixed asset investment price index of each province for the base period of 2005. We use industrial electricity consumption to represent industrial energy input.

Based on the above input–output indicators, we refer to Feng et al. [35] and Zhong et al. [36] to integrate the metafrontier approach and the Malmquist–Luenberger productivity index to measure the IGTFP of each city. All calculations were conducted in MAXDEA8.0. software.

To avoid as much as possible the endogeneity problem of the ETS due to omitted variables in the model (1), we introduced various factors affecting IGTFP as control variables. First, referring to Song et al. [28], the control variables contain foreign investment, resident population, and residents’ education. To prevent other possible omitted variables, we further added the economic development, the proportion of value added in the tertiary sector, and the proportion in the secondary sector as control variables. Foreign investment is measured as the ratio of total foreign investment utilized to GDP, population density is measured as the amount of resident population per square kilometer, residents’ education is measured as the ratio of urban education financial expenditure to GDP, and economic development is measured by GDP per capita. GDP per capita is also converted to real values based on 2005 constant prices.

Technological innovation and green technology innovation are indicated by the number of patents granted and green patents granted, respectively. We collected green patents from the patent database of the China National Intellectual Property Administration. The industrial structure is expressed as the ratio of the tertiary sector’s value added to the secondary sector’s value added.

Table 1. Relevant variables and explanations.

Variable	Abbreviation	Measurement	Data Source
Output indicator of IGTFP
Industrial value added	IWA	Industrial value added (¥ 108)	iFinD
Industrial wastewater discharged	IWW	Industrial wastewater discharged (10 000 tons)	China City Statistical Yearbook
Industrial SO2 emission	ISO2	SO2 Emission (ton)
Industrial dust emission	IDUST	Industrial Soot(dust) Emission (ton)
Input indicator of IGTFP
Industrial employment	IEM	Persons employed in urban units of mining; manufacturing; and production and distribution of electricity, gas, and water at year end (10,000 persons)	China City Statistical Yearbook
Industrial capital	ICAP	Industrial fixed capital stock (¥ 108)
Industrial electricity consumption	IELE	Industrial electricity consumption (10,000 kwh)
Dependent variable
Industrial green total factor productivity	IGTFP	/	Authors’ calculation
Independent variable
did	did	/	Authors’ calculation
Control variable
Residents’ education	EDU	Ratio of education expenditure to GDP	China City Statistical Yearbook
foreign investment	FDI	Ratio of total actual foreign capital utilized to GDP
Population density	POP	Number of resident population per square kilometer (person/Km2)	Provincial statistical yearbook and statistical bulletin of national economic and social development of each city
Economic growth	GDP	GDP per capita (¥)	China City Statistical Yearbook
Tertiary industry	TER	Proportion of the tertiary sectors value added to GDP
Secondary industry	SEC	Proportion of the secondary sectors value added to GDP
Mediating variable
Technological innovation	TEC	Number of patents granted	China City Statistical Yearbook
Green technological innovation	GTEC	Number of green patents granted	Patent Database of the China National Intellectual Property Administration
Industrial structure	IND	Ratio of tertiary sectors added value to secondary sectors added value	China City Statistical Yearbook

summarizes the above key variables, their measurements, and their data sources. We used linear interpolation to estimate the missing data for individual cities with a few particular years.

According to Table 1, we obtained a total of 4140 observations, including 690 samples in the experimental group and 3450 samples in the control group from 276 cities between 2005 and 2019.

Table 2. Descriptive statistics.

Variable	Whole Sample				Experimental Group				Control Group
	Mean	Max.	Min.	Std. Dev.	Mean	Max.	Min.	Std. Dev.	Mean	Max.	Min.	Std. Dev.
IWA	399	4175	8.6	472	618	4175	70	761	355	2862	8.6	375
IWW	6996	96,501	7	9187	10,496	96,501	303	13,254	6296	85,735	7	7953
ISO2	507,196	683,162	2	56,309	48,804	683,162	65	81,800	51,102	496,377	2	49,673
IDUST	30,255	5,168,812	34	116,369	19,921	214,774	70	25,443	32,322	5,168,812	34	126,869
IEM	17.792	260.924	0.140	26.017	36.374	260.925	2.420	47.816	14.076	221.231	0.140	16.514
ICAP	2656	25,592	44	2980	3295	25,592	233	3914	2528	19,885	44	2738
IELE	722,837	12,277,696	1016	1,109,646	1,199,180	8,057,600	11,167	1,670,300	627,569	12,277,696	1016	930,507
IGTFP	0.711	5.056	0.004	0.572	0.906	5.057	0.165	0.769	0.673	4.568	0.004	0.516
did	0.065	1	0	0.248	0.396	1	0	0.489	0	0	0	0
EDU	0.031	0.195	0.001	0.018	0.026	0.075	0.003	0.011	0.033	0.194	0.001	0.020
FDI	0.018	0.200	0	0.019	0.024	0.124	0	0.022	0.017	0.200	0.000	0.019
POP	460.638	6729.494	5.819	522.679	858.097	6729.494	99.696	1037.070	381.066	1424.573	5.819	273.861
GDP	32,129	198,005	2730	23,225	39,261	148,756	4799	26,632	30,696	198,005	2730	22,213
TER	0.389	0.835	0.034	0.096	0.414	0.835	0.272	0.098	0.384	0.792	0.034	0.096
SEC	0.478	0.909	0.090	0.110	0.470	0.656	0.162	0.085	0.479	0.910	0.090	0.114
GTEC	287.545	16,882	0	939.626	702.771	16,882	0	1898.319	204.475	7411	0	545.965
TEC	3768.96	166,609	0	10,315.56	8726.304	166,609	32	18,944.442	2776.418	91,023	0	7076.390
IND	0.908	5.154	0.075	0.494	0.952	5.154	0.431	0.508	0.899	4.946	0.075	0.490
Observations	4140				690				3450

reports the descriptive statistics.

4. Results and Discussion

4.1. ETS’s Impact on IGTFP

Based on model (1), we examine the ETS’s effects on the IGTFP. Column one in

Table 3. Benchmark effects of ETS on China’s IGTFP.

Variable	Explained Variable: IGTFP
did	0.1559 ***	0.1572 ***	0.1645 ***	0.1531 ***	0.1224 ***	0.1185 ***	0.1186 ***
	(0.0161)	(0.0159)	(0.0161)	(0.0162)	(0.0158)	(0.0159)	(0.0160)
EDU		−3.8881 ***	−3.9744 ***	−3.6926 ***	−0.4010	−0.3809	−0.3808
		(0.5306)	(0.5308)	(0.5306)	(0.5563)	(0.5562)	(0.5586)
FDI			0.7342 ***	0.9547 ***	0.5413 **	0.5300 **	0.5300 **
			(0.2453)	(0.2470)	(0.2410)	(0.2410)	(0.2411)
LNPOP				0.3014 ***	0.5406 ***	0.5420 ***	0.5420 ***
				(0.0507)	(0.0515)	(0.0515)	(0.0519)
LNGDP					0.3370 ***	0.3232 ***	0.3232 ***
					(0.0216)	(0.0228)	(0.0260)
TER						−0.1620 *	−0.1622
						(0.0878)	(0.1312)
SEC							−0.0002
							(0.1184)
Constant	0.7013 ***	0.8242 ***	0.8129 ***	−0.9247 ***	−5.8072 ***	−5.6122 ***	−5.6122 ***
	(0.0031)	(0.0170)	(0.0174)	(0.2930)	(0.4230)	(0.4359)	(0.4359)
City fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	4140	4140	4140	4140	4140	4140	4140
R2	0.9000	0.9014	0.9016	0.9025	0.9083	0.9084	0.9084

Robust standard errors are in brackets; *, **, and ***, respectively, indicate significance levels of 10%, 5%, and 1%.

shows the results without considering the control variables. The did is positive at the 1% significance level, implying that the ETS significantly increases the IGTFP, which is consistent with the above theoretical analysis. Columns two to seven show that the ETS still significantly positively affects the IGTFP when control variables are gradually introduced, thus confirming hypothesis 1. A relatively robust result is shown in column seven. The coefficient of did is 0.1186 at the 1% significance level, indicating that the ETS increases the IGTFP of the pilot cities by about 11.86% relative to the nonpilot cities. Therefore, implementing the ETS policy contributes to the green development of Chinese industry. This finding is in line with Li et al. [9], who found that the ETS contributes to the regional GTFP in China. In addition, our findings support to some extent Hu et al. [44] and Zhang and Zhang [21], who argued that the ETS contributes to reducing energy use and carbon emissions. Although Zhang and Zhang [61] and Zhang and Duan [11] evidenced the adverse economic impact of the ETS on the pilot region (industrial sector) in China, considering that the green total factor productivity incorporates economic growth and environmental and resource enhancements, the increase in ecological performance caused by the ETS compensates to some extent for the loss of economic performance and ultimately leads to the rise in overall green development performance.

4.2. Robustness Check

4.2.1. Parallel Trend Hypothesis Test

The DID approach to policy effect assessment must satisfy a basic assumption: the independent variables in the experimental and control groups maintain essentially the same evolutionary trend before being subjected to a policy shock, which is an important prerequisite for judging the validity of the DID method. Therefore, we refer to Sun and Abraham [62], who used an event-study approach to verify this parallel trend hypothesis based on model (3):

$$y_{i,t}=\alpha +{\displaystyle \sum _{\zeta =-5}^6{\theta }_t{D}_{2013+\zeta }}+{\displaystyle \sum _{k=1}^n{\beta }_{k}co{n}_{k,it}}+{\mu }_i+{\lambda }_t+{\epsilon }_{i,t},$$

(3)

where D denotes the ETS policy dummy variable; −5 and 6, respectively, indicate the fifth period before the ETS implementation and the sixth period after the ETS implementation. That is, the parallel trend hypothesis spans from 2008 to 2019. θ_t denotes the annual policy effect from 2008 to 2019. Figure 2 illustrates the parallel trend test’s findings. Apparently, none of the coefficients of the explanatory variable before the ETS implementation pass the significance test, suggesting that prior to the ETS policy, the IGTFP was evolving consistently in the treatment and control groups. However, after the policy was implemented, the impact increases significantly. This indicates that our DID method passes the parallel trend test.

In addition, the coefficients of the policy dummy variable in 2013–2015 are statistically insignificant despite being positive, i.e., the ETS has no immediate effect on improving the IGTFP. The main reason is that the ETS, as an emerging policy instrument, initially covered relatively few industrial sectors and the quota system was relatively lenient, so there was no significant emission reduction pressure on many included enterprises, thus leading to a lag in the effect of the ETS. However, starting from 2016, the sizes and significance of the coefficients significantly increase, indicating that the positive impact of the ETS on the IGTFP has expanded substantially. For example, the coefficient increased to 0.285 in 2019 at the significance level of 1%. This conclusion is in line with those of Li et al. [9] and Zhang and Zhang [63]. The increase in the impact of the ETS on the IGTFP is largely determined by the expansion of the ETS in the pilot areas and the changes in its working mechanism. As Li et al. [9] and Zhang and Zhang [63] indicated, the ETS in China covers more and more industrial sectors on the one hand, and on the other hand, the policy implementation is becoming increasingly strict, which leads to a gradual increase in the various impacts of the ETS [21,24,61], including on the IGTFP in this paper.

4.2.2. Placebo Test

Other exogenous factors unrelated to the ETS policy may affect the IGTFP during the treatment period. To identify these exogenous effects, we refer to Zhang and Zhang [24] for a placebo test using a counterfactual approach. First, we assume that the ETS runs two years earlier. If other exogenous factors work, the coefficient θ of the model (1) should remain significant; otherwise, the core conclusion of this paper is robust, i.e., the change in the IGTFP comes from the operation of the ETS. Another counterfactual test is constructing virtual experimental and control groups [63] to re-estimate the model (1). Suppose the coefficient θ is not significant in the dummy experimental group. In that case, there is no systematic difference in the IGTFP changes between the experimental and control groups in the absence of an ETS policy, which indirectly confirms the robustness of the positive impact of the ETS on IGTFP. Following Zhang and Zhang [63], we construct a counter fact as follows: if the city belongs to the developed regions, it belongs to the experimental group; otherwise, the city belongs to the control group. We identify nine provinces as developed regions.

Table 4. Robustness check: Results of placebo test.

Variable	Explained Variable: IGTFP
	ETS Policy Advanced by 2 Years		Fictional Experimental Group
did	0.1238	0.0893	0.0750	0.0716
	(0.1158)	(0.0756)	(0.1128)	(0.1127)
Control variable	NO	Yes	NO	Yes
Constant	0.7007 ***	−5.7397 ***	0.7215 ***	−6.1605 ***
	(0.0032)	(0.4365)	(0.0033)	(0.4342)
City fixed effects	Yes	Yes	Yes	Yes
Time fixed effects	Yes	Yes	Yes	Yes
Observations	4140	4140	4140	4140
R2	0.8992	0.9078	0.8984	0.9078

Robust standard errors can be found in brackets; *** denotes significance level of 1%.

reports the results of the counterfactual tests and shows that the new coefficients of did are insignificant, which proves the robustness of the core findings of this paper.

4.2.3. PSM-DID Method

Although the use of the DID method has been justified earlier by parallel trend tests, considering that the nonrandom establishment of the ETS pilot provinces may lead to biased estimation results, this paper further employs the propensity score-matching (PSM) method to find a suitable control group for estimation. We selected residents’ education, foreign investment, population density, economic development, tertiary industry development level, secondary industry development level, industrial electricity consumption, and economic growth rate (GDPg) as matching variables to construct the logit regression model and used the nearest-neighbor matching and kernel matching to match the control group. The balance test’s results for covariates in

Table 5. Balance test for covariates.

Covariate	Sample	Nearest Neighbor Matching			Kernel Matching
		Bias (%)	Reduction Bias (%)	T-Test (p-Value)	Bias (%)	Reduction Bias (%)	T-test (p-Value)
EDU	Unmatched	−21.51	77.14	0.000	−21.51	67.14	0.000
	Matched	4.43		0.726	3.85		0.513
FDI	Unmatched	49.97	12.86	0.720	49.97	1.43	0.720
	Matched	24.98		0.846	2.78		0.169
POP	Unmatched	3.86	5.67	0.860	3.86	0.93	0.860
	Matched	0.32		0.819	0.05		0.436
GDP	Unmatched	−2.44	57.34	0.000	−2.44	36.09	0.000
	Matched	−0.16		0.281	−0.10		0.209
TER	Unmatched	9.34	23.00	0.028	9.34	28.00	0.028
	Matched	0.33		0.436	0.41		0.541
SEC	Unmatched	−1.73	−283.33	0.901	−1.73	−258.89	0.901
	Matched	−7.99		0.753	−7.30		0.797
IELE	Unmatched	−23.17	24.28	0.037	−23.17	14.17	0.037
	Matched	−0.41		0.187	−0.24		0.122
GDPg	Unmatched	−44.66	56.37	0.000	−44.66	56.95	0.000
	Matched	−2.19		0.316	−2.21		0.371

indicate that most covariates are significantly different between the experimental and control groups prior to matching. However, the deviations of the means of most covariates were considerably reduced after matching. In conclusion, before and after matching, the differences in most covariates between the experimental and control groups changed from significant to nonsignificant, thus achieving greater homogeneity between the experimental and control groups.

The matched data were used to re-estimate model (1), and the results are shown in

Table 6. Robustness check: Results of the PSM-DID method.

Variable	Explained Variable: IGTFP
	Nearest-Neighbor Matching		Kernel Matching
did	0.0795 ***	0.0635 **	0.1162 ***	0.1208 ***
	(0.0255)	(0.0255)	(0.0281)	(0.0281)
Control variable	NO	Yes	NO	Yes
Constant	0.6989 ***	−8.4938 ***	0.7299 ***	−10.0407 ***
	(0.0082)	(1.1494)	(0.0091)	(1.2556)
City fixed effects	Yes	Yes	Yes	Yes
Time fixed effects	Yes	Yes	Yes	Yes
Observations	1342	1342	1380	1380
R2	0.8735	0.8849	0.8733	0.8837

Robust standard errors can be found in brackets; ** and *** denote significance levels of 5% and 1%, respectively.

. Again, the coefficients of all dids are significantly positive at the 1% significance level, whether using nearest-neighbor matching or kernel matching methods. Therefore, the impact of the ETS on China’s IGTFP is positive and robust.

4.3. Mediating Mechanism Analysis

To examine hypotheses 2 to 4, we built the following mediating effects models and combined them with the model (1).

$$TE{C}_{i,t}=\alpha +\theta di{d}_{i,t}+{\displaystyle \sum _{k=1}^n{\beta }_{k}co{n}_{k,it}}+{\gamma }_i+{\lambda }_t+{\epsilon }_{i,t},$$

(4)

$$GTE{C}_{i,t}=\alpha +\theta di{d}_{i,t}+{\displaystyle \sum _{k=1}^n{\beta }_{k}co{n}_{k,it}}+{\gamma }_i+{\lambda }_t+{\epsilon }_{i,t},$$

(5)

$$IN{D}_{i,t}=\alpha +\theta di{d}_{i,t}+{\displaystyle \sum _{k=1}^n{\beta }_{k}co{n}_{k,it}}+{\gamma }_i+{\lambda }_t+{\epsilon }_{i,t},$$

(6)

$$y_{i,t}=\alpha +\theta di{d}_{i,t}+{\varphi }_{1}TE{C}_{i,t}+{\varphi }_{2}GTE{C}_{i,t}+{\varphi }_{3}IN{D}_{i,t}+{\displaystyle \sum _{k=1}^n{\beta }_{k}co{n}_{k,it}}+{\gamma }_i+{\lambda }_t+{\epsilon }_{i,t}$$

(7)

Preacher and Hayes [64] proposed the bootstrap method to test the mediating effects, which has become a widely accepted approach. We also applied this method to test for possible mediating effects of technological innovation, green technological innovation, and industrial structure. A bias-corrected bootstrap confidence interval typically has 5000 bootstrap samples, and all output confidence intervals have 95% confidence level. Considering the significant magnitude differences between technological innovation or green technological innovation and other variables, they are taken as natural logarithms in models (4)–(7). In addition, since the minimum value of the numbers of patents and green patents is zero (see Table 2), their natural logarithms are taken by adding one to each value. The mediating effects using the bootstrap method are reported in

Table 7. Mediating mechanism of ETS’s effects on IGTFP.

Variable	IGTFP			lnGTEC			lnTEC			IND
	Coefficient	Confidence Intervals		Coefficient	Confidence Intervals		Coefficient	Confidence Intervals		Coefficient	Confidence Intervals
Constant	−1.8264 ***	−2.1579	−1.4950	−16.5169 ***	−17.1132	−15.9206	−14.2346 ***	−14.8322	−13.6370	0.4216 ***	0.2852	0.5581
did	0.0969 ***	0.0344	0.1594	0.1240 ***	0.0352	0.2128	0.0227	−0.1468	0.1922	0.0633 ***	0.0104	0.1162
lnGTEC	0.1032 ***	0.0798	0.1267
lnTEC	0.1055 ***	0.0821	0.1289
IND	0.1027 ***	0.0450	0.1604
Control variable	Yes			Yes			Yes			Yes
Time-fixed effects	Yes			Yes			Yes			Yes
Region-fixed effects	Yes			Yes			Yes			Yes
Observations	4140			4140			4140			4140
Adjusted R-squared	0.7487			0.8741			0.8734			0.9095
Indirect effect of ETS on IGTFP	Effect			Boot SE			BootLLCI			BootULCI
TOTAL	0.0217			0.0053			0.0017			0.0417
GTEC	0.0128			0.0061			0.0020			0.0236
TEC	0.0024			0.0075			−0.0100			0.0148
IND	0.0065			0.0028			0.0001			0.0129

*** denotes significance level of 1%.

.

The results demonstrate that the mediating effect of technological innovation on the ETS’s affecting of the IGTFP is not significant because its confidence interval includes zero. Therefore, hypothesis 2 is not valid. Table 7 indicates that, although technological innovation significantly positively affects the IGTFP with a coefficient of 0.1055 and a confidence interval without zero, the ETS does not significantly contribute to the increase in technological innovation with a coefficient of 0.0227 and a confidence interval with zero. The correlation mechanism between the ETS and technological progress has not been established. The role of the ETS is limited in the current overall technological progress in China. The possible reasons lie in the inefficient domestic carbon trading market and overall weak environmental regulation. Most ETS-covered companies see participation in carbon trading mechanisms as a means to improve relations with governments and gaining a good social reputation, rather than as a cost-effective method to reduce greenhouse gas emissions [65].

In contrast, consistent with the theoretical hypothesis, green technology innovation and industrial structure significantly mediate the effect of the ETS on the IGTFP, as their confidence intervals both contain zero. The partial mediation effect of green technology innovation is 0.0128, and that of industrial structure is 0.0065, accounting for 10.79% and 5.48% of the total effect, respectively. Therefore, hypotheses 3 and 4 are valid. The technology driving effect of the ETS is more involved in developing and applying for green patents due to its own objectives in energy saving and emission reduction. In order to promote low-carbon and green technology innovation, China’s industrial sector has also produced significant achievements that are subject to the ETS, echoing the findings of Teixidó et al. [48], Gao and Wang [49], and Liu and Sun [50], but contradicting those of Zhao et al. [51] and Chen et al. [52].

Another important point to note is that the ETS significantly increases the industrial structure. That is, the secondary sector shrinks significantly relative to the tertiary sector in China under the ETS framework. This confirms our previous speculation that the ETS will lead to a shift of resources and investments to the tertiary sector. As Chen et al. [52] indicated, the industrial sector shrinks its output, especially for high-emitting products, as a result of the establishment of the carbon market. Although the ETS has led to an increase in the IGTFP, it is generally detrimental to the industrial sector’s sustainable development and the healthy development of the Chinese economy due to the absolute pillar position of the industrial sector in the Chinese economic system. Moreover, in addition to our hypothetical green-technology-driven industrial productivity gains, the management and product upgrading resulting from the ETS may be limited, which also leads to an economic contraction in the industrial sector. Notably, a desirable increase in the IGTFP should be achieved by both pollution reduction and economic output increase rather than a relatively abnormal IGTFP increase based on output reduction, which should capture policymakers’ attention.

4.4. Heterogeneity Analysis

According to the previous theoretical explanations, we examined the heterogeneity of the effects of the ETS on the IGTFP at various levels of economic growth, energy use, and industrial dependence using the following moderating effects model:

$$y_{i,t}=\alpha +\theta di{d}_{i,t}+{\varphi }_{1}mo{d}_{i,t}+{\varphi }_{2}di{d}_{i,t}\times mo{d}_{i,t}+{\displaystyle \sum _{k=1}^n{\beta }_{k}co{n}_{k,it}}+{\gamma }_i+{\lambda }_t+{\epsilon }_{i,t},$$

(8)

where mod denotes a set of moderating variables, and the coefficient $\varphi$₂ of the interaction term reflects the heterogenous effects of the ETS on the IGTFP. Development of the economy is still viewed as the GDP per capita, energy consumption is expressed by the overall electricity consumption of the city, and the level of industrial dependence is expressed by the proportion of value added in the secondary industry; these data also come from China City Statistical Yearbook.

Table 8. Results of heterogeneity analysis.

Variable	Explained Variable: IGTFP
	(1)	(2)	(3)
did	−2.4244 ***	−1.4213 ***	0.9814 ***
	(0.2764)	(0.1521)	(0.0731)
lnGDP	0.3198 ***
	(0.0257)
did *lnGDP	0.2377 ***
	(0.0258)
lnelectricity		−0.0755 ***
		(0.0058)
did *lnelectricity		0.1123 ***
		(0.0111)
Sec			0.0583
			(0.1164)
didr *sec			−1.8834 ***
			(0.1559)
Control variable	Yes	Yes	Yes
Constant	−5.2230 ***	−4.9395 ***	−5.2273 ***
	(0.4333)	(0.4229)	(0.4291)
City-fixed effects	Yes	Yes	Yes
Time-fixed effects	Yes	Yes	Yes
Observations	4140	4140	4140
R2	0.9104	0.9146	0.9117

Robust standard errors can be found in brackets. *** denotes significance level of 1%.

reports the regression results according to model (8). Our results confirm the heterogeneous effect of the ETS in Li et al. [9] and Zhang and Zhang [24]. On the contrary, we identified specific factors that produce heterogeneous effects. Notably, columns one and two have significantly negative coefficients of did. When the interaction term is introduced, there may be a problem of multicollinearity between independent variables, the interaction term, and moderating variables in the model, i.e., duplication or overinterpretation between variables in explaining the dependent variable, so that the main effect in this scenario may change. However, it should be noted that on this occasion, we are still focusing only on the results of Table 3 rather than Table 8 regarding the effect of the ETS on the IGTFP and the coefficients of the interaction terms.

According to column 1, the 1% level of the interaction term exhibits a significant positive coefficient. Since the ETS significantly and positively affects the IGTFP in Table 3, economic development reinforces the effect of the ETS on the IGTFP. This suggests that the positive effect of the ETS on the IGTFP becomes more significant as the level of economic development increases. Similarly, the coefficient of the interaction term between urban energy consumption and did is also significantly positive (see column 2). Hence, an increase in urban energy consumption also implies the ETS’s more significant positive effect on the IGTFP. These results suggest that in economically developed regions with a higher GDP per capita and energy consumption, the ETS contributes more to the IGTFP. The possible reason is that economically developed regions tend to enforce a stricter ETS, and local governments are capable of subsidizing industrial enterprises more for low-carbon and energy transition, so that industrial enterprises have more pressure and motivation for the green transition. In less developed regions, the ETS policy is relatively lenient, and the energy mix is more dependent on traditional fossil energy, so the green transition process of industrial enterprises is relatively slow.

Furthermore, column three demonstrates that, at the 1% level, there is a significant negative correlation between industrial dependence and did, which indicates that industrial dependence inhibits the positive effect of the ETS on IGTFP. Notably, industrial dependence significantly and positively affects the IGTFP, indicating a significant substitution relationship between the ETS and the secondary sector in affecting the IGTFP. The positive impact of the ETS on the IGTFP decreases as cities’ industrial dependence level increases. Usually, industrial-dependent cities have higher carbon emissions, but we find that Shenzhen, Beijing, and Shanghai, where the Chinese ETSs were first piloted, have relatively low industrial dependence and more developed tertiary industries. This indicates that China’s ETS does not pressure industrial-dependent cities significantly. In other words, the impact of the ETS in industrial-dependent cities is relatively weak, even though the ETS mainly covers industrial enterprises.

5. Conclusions and Policy Implications

The ETS is the leading policy for achieving a low-carbon transition. This paper investigates the effects of the ETS on the IGTFP and its dynamics by using Chinese prefecture-level panel data from 2005 to 2019 before the national ETS run. The results demonstrate that the ETS contributes to the increase in IGTFP in the pilot regions, and the growth becomes progressively significant over time. However, the positive effect does not become significant until 2016. The ETS raises the IGTFP through improving green technological innovation and industrial structure, but the expected mediating mechanism of general innovation is absent. In addition, the ETS contributes more to the IGTFP in cities with higher economic growth and energy use levels. However, the higher proportion of secondary industry in the city means a smaller optimistic effect of the ETS on the IGTFP. A number of significant policy implications are highlighted in the article based on the aforementioned finding.

First, the article provides scientific evidence for the national expansion of the ETS in China beyond the power sector. For China’s low-carbon transition and the absolute strategic foundation of its economy, the industrial sector is essential. The positive impact of the ETS on China’s IGTFP will reinforce China’s confidence that the ETS could be expanded to include more industrial sectors such as mining and manufacturing, in addition to the power sector.

Second, green technological innovation should be significantly improved. Specifically, it is proposed that quotas and subsidies be provided for green technological innovation in the industrial sector within the ETS framework. The use of renewable electricity should be encouraged in industrial enterprises and in the development of carbon sequestration technologies, carbon conversion technologies, and emission reduction technologies in the industrial sector, especially in the power sector. In addition, in order to force industrial enterprises to achieve technological emission reduction, a modest decrease in the overall free quota of the ETS is suggested; on the contrary, the auction mechanism should be strengthened. In addition, the price of carbon should be moderately increased under the market framework to exceed the cost of carbon reduction.

Third, the ETS causes a decrease in the secondary sector (mainly the industrial sector). As the main covered sector of the ETS, industrial sector development requires protection under the ETS framework. In addition to technological innovation, management and product upgrading in the industrial sector are also necessary and urgent. In addition, although ETS policies should be strict and prudent, a national unified carbon market should still maintain regional flexibility with regard to industrial friendliness, especially in industrial-dependent cities.

Fourth, the improvement in regional economic performance contributes to the positive effect of the ETS on the IGTFP. Economically developed regions can provide more flexible policy packages for the ETS implementation due to their solid socioeconomic support. Notably, economic growth remains a priority for regional development while the ETS is being implemented. Therefore, it is recommended that a coordinated growth mechanism be built for the ETS and the IGTFP based on economic development.

Despite the above findings, the paper has some limitations that warrant further exploration in the future. For example, in calculating the IGTFP, we assume that undesired output indicators include three pollutants. Due to the lack of statistical data, we do not consider another important indicator of green development, namely, industrial CO₂ emissions. In the future, we can consider more nondesired output indicators to better characterize the IGTFP, provided data are available. In addition, this article only explores the direct effects of the ETS on the IGTFP, without considering possible spatial spillover effects. We can continue to explore whether the ETS also has indirect effects on the IGTFP in neighboring regions in the future.