The Impact of Carbon Emission Trading Policy on Industrial Structure Adjustment: A Perspective of Sustainable Development

Abstract

To mitigate the effects of climate change, carbon emission trading policy (CET) has emerged as a crucial policy instrument for nations. As the largest developing country, China confronts the pressing need to steer industrial restructuring and foster sustainable economic growth. Utilizing provincial panel data from 2005 to 2020, this study constructs a difference-in-differences model to examine the influence of CET on industrial structure adjustment (ISA) and corroborates these findings with robustness tests. The analysis reveals that: (1) CET substantially facilitates industrial restructuring; (2) CET inherently motivates enterprises towards technological innovation, thus advancing regional industrial restructuring; and (3) the effects of CET on industrial structures exhibit marked regional variability.

1. Introduction

Rising sea levels, the increased frequency of extreme weather events, and significant global ecological issues stemming from global warming are of considerable global concern. Likewise, the profound threats posed by global warming to societal development warrant serious attention [1]. Global warming not only reduces agricultural yields [2] but also destabilizes the labor market [3] and adversely affects human health [4]. Consequently, the imperative to address climate change and advance the reduction in greenhouse gas emissions has emerged as a widely acknowledged consensus in the 21st century [5]. The IPCC’s Fifth Assessment Report (2014) pinpointed the combustion of fossil fuels and industrial processes as the primary sources of the greenhouse gasses driving global warming [6]. The International Energy Agency’s Global Energy Review: 2021 Carbon Dioxide Emissions’ reports that global greenhouse gas emissions reached 40.8 billion tonnes of carbon dioxide equivalent in 2021 [7], underscoring escalating concerns about climate change. Carbon emissions, a major source of these gasses, have led the international community to adopt the Paris Agreement, which sets forth detailed emission reduction targets and strategies. A key component of this agreement, the carbon emission trading policy (CET), has emerged as a critical tool for nations combating climate change. Initiated by the Chinese government in 2011 and expanded in 2013 to include pilots in Beijing, Shanghai, Guangdong, Tianjin, Chongqing, Hubei, and Shenzhen, these trading systems have been refined over time to glean insights into effective CET market operations. By 2024, China had established CET markets across 31 provinces and regions, positioning its market as one of the largest worldwide. The Chinese government has committed to peaking carbon dioxide emissions by 2030 and achieving carbon neutrality by 2060 [8].

ISA is a crucial pathway to sustainable economic development [9]. Guiding this adjustment can significantly drive technological innovation, enhance regional production efficiency and competitiveness, and promote sustainable economic growth. As a market-incentive environmental regulation, CET raises several questions: What impact does it have on regional industrial structures? Do these impacts vary across different regions? What mechanisms drive these influences? Existing research has inadequately addressed the effects of CET on ISA. A more thorough investigation of these effects not only helps refine carbon market mechanisms but also provides a theoretical foundation for governmental policy-making, thus effectively fostering ISA and sustainable economic development [10]. This forms the crux of the current study.

Therefore, this study utilizes provincial panel data from China spanning from 2005 to 2020 to develop a difference-in-differences model, assessing the influence of CET on industrial structure adjustment (ISA) [11]. The robustness of the findings is confirmed through placebo tests and sample data reduction. Subsequently, this paper investigates the regional disparities in the relationship between CET and ISA [12], employing a stepwise regression method to elucidate the mechanisms that connect these variables. Distinct from previous research, this study presents several novel contributions: Firstly, it examines the effects of CET on industrial structure, thus enriching the body of theoretical research on these policies; secondly, this paper adopts the perspective of sustainable development and incorporates both the optimization of industrial structure (OIS) and rationalization of industrial structure (RIS) into the scope of its research, thereby expanding the theory of ISA. Thirdly, this paper incorporates spatial heterogeneity into its research and empirically tests the impact of CET on industrial structure.

The remaining part of the article in this study is structured as follows: Section 2 is a review of the relevant literature, Section 3 specifically describes the data and variable sources, Section 4 mainly presents the empirical results, and Section 5 presents the appropriate conclusions and recommendations based on the data results.

2. Literature Review

2.1. Research Related to the Carbon Emission Trading Policy

CET serves as a policy tool for controlling carbon emissions through the market-based trading of carbon dioxide emission rights, aiming to reduce emissions. Originating from the Coase property rights theory [13], this market is a direct application of Dales’s late 1960s concept of “pollution rights trading” [14]. Under this framework, the limited carrying capacity of Earth’s resources justifies the legalization and market-based trading of pollution rights, thus facilitating environmental governance [15]. The effectiveness of CET hinges on the development of a robust market mechanism that incentivizes firms to meet environmental targets cost-effectively, even in the absence of regulatory oversight. Established early, the EU carbon trading market initially focused on establishing market specifics such as industry coverage, carbon caps, quotas, and pricing. These foundational studies were pivotal for the market’s development. Subsequent research shifted towards assessing the market’s efficacy in reducing emissions. For instance, Martin (2015) found that the EU’s carbon trading system reduced corporate emissions by 3% and 10–26% during its first and second phases, respectively [16]. Moreover, Brink (2016) observed a negative correlation between carbon price fluctuations and different carbon taxes [17]. Other researchers have explored the economic implications of carbon trading. For example, Delarue noted an increase in electricity costs due to the EU ETS [18], whereas Hoffmann’s empirical analysis (2005–2007) highlighted minimal impacts on corporate investment decisions in Germany’s power sector [19]. Utilizing a numerical simulation, Abrell, J., demonstrated that a moderate minimum price within the EU Emissions Trading System could decrease the costs of the EU’s climate policy by up to 30% [20]. In contrast, China’s CET market is still evolving, showing promising results in emission reductions. Research by Zhang (2020) indicated a 16.2% decrease in emissions in pilot regions, notably in the economically advanced eastern areas [21]. Feng Y. (2020) analyzed the effects and mechanisms of carbon trading on urban carbon emission intensity using the DID model [22]. Additionally, Hübler (2014) discussed the macroeconomic impacts of China’s CET market in terms of industry competitiveness and economic growth uncertainty [23], while Wu identified potential GDP reductions due to carbon cuts, which could be partly offset by effective trading policies [24].

2.2. Research Related to ISA

Disparities in income across various industries facilitate the reallocation of production factors—such as labor—from less efficient to more developed sectors [25]. This process, known as the OIS, is intricately linked to regional economic growth [26]. It involves a reduction in the share of traditional agriculture and manufacturing, alongside a gradual increase in high-value and high-tech sectors including the service and creative industries [27]. Such optimization not only bolsters Total Factor Productivity (TFP) but also furthers sustainable economic development [28,29]. Further analysis reveals that economic growth driven by capital investment may create discrepancies between industrial and employment structures [30]. Notably, an excessive focus on tertiary industries harbors numerous risks; it may precipitate the decline of traditional sectors, leading to structural unemployment, and increase reliance on imported resources like finance and information technology, thus diminishing economic resilience and heightening vulnerability to external shocks [31]. In the era of the COVID-19 pandemic, enhancing economic resilience is critical for sustained development [32].

The objective of ISA is to optimize resource allocation efficiency and foster sustainable economic development [33]. This transformation encompasses two principal aspects: RIS and OIS, both critical for today’s sustainable economic progress [34,35]. Rationalization reflects the coordinated development among diverse industries and the input-output structural efficacy. Conversely, an irrational industrial structure can disrupt energy factor allocation and utilization, impeding the growth rate and imposing a significant ‘structural burden’ on the economy [36]. As an essential element of the economic framework, the advancement in industrial structure denotes the share of efficient production sectors. This not only impacts economic growth, resource allocation efficiency, and social welfare but also bolsters supply chains, enhances product quality, fosters innovation, and improves energy efficiency [37]. Moreover, this optimization often leads to the emergence of high-value-added industries, creating numerous job opportunities in high-skilled and research and development sectors, thereby elevating workforce competency and quality [38].

Current research regarding the effects of China’s CET on industrial structure remains inconclusive [39]. As a form of market-based environmental regulation, some scholars contend that this policy suppresses industrial transformation by increasing production, external, and opportunity costs for enterprises. This escalation in costs can lead to resource misallocation and a ‘crowding out effect’ that stifles innovation [40]. Furthermore, stringent environmental regulations may induce pollution-intensive enterprises to relocate to areas with more lenient regulations, a scenario described as the ‘pollution haven hypothesis’ [41]. From a corporate perspective, Yue Dai observed that China’s CET has failed to harmonize economic, social, and environmental development, exerting no significant impact on industrial structure [42]. Zhang identified spatial spillover effects associated with this policy [43], while Nie’s findings support the ‘pollution haven hypothesis’, indicating that the policy has prompted the relocation of pollution-intensive industries [44].

Several scholars contend that CET facilitates ISA [45]. First, stringent environmental regulations escalate market entry costs for pollution-intensive enterprises, resulting in a reduction in such enterprises and a proliferation of cleaner alternatives, thus establishing “green barriers” [46]. Second, while environmental regulations introduce additional costs for pollution control that may initially impede profitability, they ultimately encourage enterprises to innovate, thereby securing long-term competitive advantages and fostering the adjustment of regional industrial structures, in accordance with the Porter hypothesis [47,48]. Zhou’s empirical analysis indicates that CET has enhanced the green adjustment of China’s manufacturing industry [49]. Furthermore, Xu’s study from the perspective of the marine industrial structure reveals that carbon trading policies have not only upgraded the marine industrial structure but also exerted a positive influence in adjacent provinces [15]. Shen’s investigation into high carbon-emitting sectors, such as electricity, electronics, and machinery, demonstrates that the pilot CET has markedly curtailed energy-intensive industries in the industrial sector [50].

Researchers worldwide have thoroughly debated CET and its implications for ISA. However, existing studies present several limitations. Primarily, much of this research prioritizes the emission reduction and economic impacts of carbon trading, thus overlooking its effects on industrial structures. Furthermore, analyses often focus solely on OIS, neglecting the critical aspect of rationalization, which is vital for both the sustainability of industrial structures and economic development. Moreover, the discourse on the influence of China’s CET on industrial structures is fragmented and contentious, with many studies favoring theoretical approaches or simulations over empirical data analysis. This paper aims to contribute to the field by focusing on industrial structures to deepen the theoretical understanding of CET. It broadens the scope of industrial structure theory by integrating both advancement and rationalization, and it introduces spatial factors to empirically assess the impact of CET on industrial structures.

3. Date and Methods

3.1. Modeling

The difference-in-differences (DID) method applies two-fold differentiations to treatment and control groups before and after policy implementations [51], thereby effectively mitigating the influence of time trends by juxtaposing changes within the same timeframe for both groups. This approach consequently circumvents any potential confounding from temporal trends [45]. Moreover, DID addresses spatial confounders by analyzing variations across different regions, accounting for both regional discrepancies and unique area-specific factors. In empirical research, DID mitigates distortions in results stemming from selection biases. By simultaneously evaluating changes in both the treatment and control groups, DID enables a more precise estimation of treatment effects and successfully prevents potential issues with endogeneity, thus its prevalent use in policy evaluation. Notably, the validity of DID hinges on the parallel trends assumption; accordingly, this paper includes a sequence of robustness tests in later sections. This paper treats the CET as a quasi-natural experiment, using the difference-in-differences method to assess the impact of the CET on ISA. It constructs a difference-in-differences model with dual fixed effects as follows:

$$\begin{array}{c}{Y}_{it}={\gamma }_0+{\gamma }_1{did}_{it}+{\beta X}_{it}+{\mu }_i+{\tau }_t+{\epsilon }_{it}\end{array}$$

(1)

$${did}_{it}={treated}_i*{after}_t$$

(2)

where $Y_{it}$ is the explanatory variable, indicating the level of industrial structure of the ith province in year t. In this paper, we select two dimensions of RIS and OIS, to be measured; ${did}_{it}$ is the core explanatory variable, indicating whether region i has implemented the CET at time t, if it has implemented it ${did}_{it}=1$, and if it has not, then ${did}_{it}=0$. $X_{it}$ is the control variable, τ_t is the time-fixed effect, which can be used to regulate the impact of various policy factors that change over time, ${\mu }_i$ is the individual fixed effect, which can be used to regulate the impact of regional variability, and 6 is the error term. Variation of various policy factors impact, ${\mu }_i$ represents the individual fixed effect, can be used to regulate the differential impact of the region, ${\epsilon }_{it}$ is the error term. ${did}_{it}$ of the coefficient γ₁ is the core estimation parameter, indicating the net effect of CET on the industrial structure, if the coefficient of γ₁ is significant, it indicates that the CET on the regional industrial structure has a significant impact; on the contrary, it has no effect.

3.2. Variables and Data

3.2.1. Explained Variables

This paper considers industrial structure as the dependent variable, mainly evaluating it from the RIS and OIS dimensions [52,53,54]. RIS refers to adjusting the previously irrational industrial structure to achieve rational allocation of production factors and coordinated development among industries. This not only reflects the degree of coordination between industries but also shows the effectiveness of resource utilization and characterizes the quality of industry aggregation. To measure the rationalization level of industrial structure, this paper uses the Theil index as the evaluation metric. The calculation formula is as follows:

$$RIS={\sum }_{i=1}^n\left({\displaystyle \frac{Y_i}{Y}}\right)\ln \left({\displaystyle \frac{Y_i}{L_i}}/{\displaystyle \frac{Y}{L}}\right)$$

(3)

In Formula (3), ‘i’ represents the i-th industry, ‘n’ stands for the number of industrial sectors, ‘Y’ denotes employment output value, and ‘L’ represents the number of employees; thus, Y/L indicates the level of productivity. According to classical economic theory, when the economy is in a final state of equilibrium, the productivity levels among all sectors are the same, leading to: Yi/Li = Y/L, and therefore RIS = 0. Conversely, if the industrial structure deviates from this equilibrium state, the Theil index will not be zero, indicating an irrational industrial structure. Additionally, RIS can also reflect the coupling between the output structure and the employment structure. Due to data availability, this paper calculates the RIS index using data from the primary, secondary, and tertiary sectors of various cities. OIS refers to the transformation of leading industries as the economy develops, which in turn affects the overall industrial structure. The specific connotations include enhancements in technological levels within industries, increases in added value, and a rise in the degree of servitization. This paper uses Clark’s Law, refers to the methods of Shao [52] and Wang [53] to measure the level of OIS by the ratio of the tertiary to the secondary industry, and constructs a corresponding index of OIS. A higher index value indicates a higher level of OIS.

$$OIS={\displaystyle \frac{Y_3}{Y_2}}$$

(4)

3.2.2. Explanatory Variables

In 2011, China released a pilot policy document for CET, identifying Beijing, Tianjin, Chongqing, Hubei, Shanghai, Guangdong, and Shenzhen to pilot CET markets [11]. This provides a quasi-natural experiment opportunity for this paper’s study on CET, which could overcome endogeneity. The policy was not officially launched until 2013. Therefore, we set 2013 as the starting year for the experiment. The period before 2013 is considered the non-experimental period, and the period from 2013 onwards is considered the experimental period. A dummy variable was introduced for the pilot policy to simulate a natural experiment scenario. Given Shenzhen’s inclusion in Guangdong province, the study selected six regions—including Beijing, Tianjin, Chongqing, Hubei, Shanghai, and Guangdong—as the treatment group, with 25 other provinces forming the control group. This configuration yielded panel data spanning from 2005 to 2020 across 31 provinces and cities, thus underpinning the model’s empirical analysis (

Table 1. Treatment group and control group regions.

Group	Specific Provinces and Cities
Treatment group	Beijing, Tianjin, Chongqing, Hubei, Shanghai, Guangdong
Control group	Shandong, Shanxi, Inner Mongolia, Qinghai, Hebei, Liaoning, Jiangsu, Zhejiang, Sichuan, Anhui, Hunan, Jiangxi, Jilin, Fujian, Henan, Guangxi, Hainan, Xinjiang, Guizhou, Yunnan, Shaanxi, Gansu, Heilongjiang, Ningxia, Tibet

).

This article explains the variable as shown in Equation (2), where “did_it” represents the interaction term between the dummy variables “treated_i” and “after_t”. The construction of these two dummy variables is as follows: The regional dummy variable (“treated_i”) distinguishes areas based on whether they have implemented a CET market, with provinces and cities approved for pilot CET treated as the treatment group and assigned a value of 1, while other provinces and cities are set as the control group and assigned a value of 0. The time dummy variable (“after_t”) differentiates based on the implementation time of the CET market. This paper identifies 2013 as the year when the CET market began to have an effect, with years 2013 and onwards assigned a value of 1, and years before 2013 assigned a value of 0.

3.2.3. Control Variables

To accurately assess the impact of the CET on industrial structure and control for other interfering factors, while minimizing estimation bias caused by collinearity, we introduce the following control variables [55,56]: economic development level (pgdp), represented by per capital GDP; capital input level (invest), expressed as the ratio of fixed asset investment to GDP; urbanization level (urban) measured by the proportion of the urban population to the total population in the region; infrastructure status (infra), using the road mileage per square kilometer in each province to measure the impact of regional infrastructure levels on industrial upgrading; government regulation level (fiscal), quantified by the proportion of government fiscal expenditure to GDP in each region, serving as an indicator of the impact of government industrial policies; and information technology level (internet), measured by the ratio of the total postal and telecommunications business to GDP in each province. Specific variable selection and calculation methods are shown in

Table 2. Main variables and their calculation methods.

Variable Name	Variable Symbol	Calculation Method
Rationalization of industrial structure	RIS	Tyrell index
Optimization of industrial structure	OIS	Tertiary industry output/secondary industry output
Carbon emission trading	did	Whether to implement a carbon trading policy
Level of economic development	pgdp	Logarithmic GDP per capita
Capital input level	invest	Fixed asset investment/gross regional product
Urbanization level	urban	Urbanized population/total population
Infrastructure status	infra	Road mileage per square kilometer
Government regulation level	fiscal	Government fiscal expenditure/Gross domestic product
Information technology level	internet	Total postal and telecommunications business/GDP
Carbon price	price	Logarithmic annual average of daily transaction prices

.

In terms of data selection, this paper uses panel data from 31 provincial-level administrative regions in China from 2005 to 2020. The main sources of data are the ‘China Statistical Yearbook’, as well as statistical yearbooks from various provinces, cities, and departments. Missing values in the data were filled using linear interpolation and the ARIMA model method.

Table 3. Descriptive statistics of variables.

Variable	Median	Standard Deviation	Minimum	Maximum
RIS	8.279	11.01	1.173	126.6
OIS	1.222	0.669	0.527	5.244
did	0.0968	0.296	0	1
pgdp	10.49	0.660	8.528	12.01
invest	0.863	0.919	0.142	7.379
urban	0.542	0.147	0.208	0.896
fiscal	0.255	0.193	0.0798	1.379
infra	0.832	0.501	0.0360	2.205
internet	0.07	0.05	0.01	0.29
price	3.198	0.570	1.858	4.343

presents the descriptive statistics of the variables.

4. Empirical Analysis

4.1. Benchmark Model Regression

This paper uses the difference-in-differences method to conduct an empirical analysis of Model (1), with the specific results shown in

Table 4. Baseline regression of the impact of CET on ISA.

Variable	(1)	(2)	(3)	(4)
	RIS	OIS	RIS	OIS
did	5.2384 **	0.2955 *	6.3351 **	0.3322 **
	(2.2685)	(1.6501)	(2.4780)	(2.2586)
pgdp			0.5920	0.3346 **
			(0.4487)	(2.2715)
invest			−0.2890	0.0842
			(−0.2974)	(0.8295)
urban			23.2367 ***	−1.7583 *
			(4.9772)	(−1.7605)
fiscal			−0.1293 **	−0.8991 ***
			(−2.3827)	(−5.4010)
infra			−0.0530 **	−0.0615
			(−2.0952)	(−0.5392)
internet			27.9832 **	2.0865 ***
			(2.0909)	(4.7758)
control variable	No	No	yes	yes
Area-fixed effects	yes	yes	yes	yes
Time-fixed effects	yes	yes	yes	yes
Constant	4.3058 ***	0.9468 ***	1.4914 ***	5.1198 ***
	(8.0853)	(14.5837)	(3.5355)	(3.9574)
Sample size	496	496	496	496
R2	0.682	0.721	0.927	0.916

Note: t-statistic values are in parentheses, and ***, **, and * indicate 1%, 5%, and 10% significance levels, respectively.

. In the table, columns (1) and (2) present the regression results without control variables, while columns (3) and (4) consider control variables. It can be seen from Table 4 that, whether or not control variables are included, the core explanatory variable ‘did’ is significantly positive. CET has had a positive and effective impact on RIS and OIS. The above research results fully demonstrate that CET has played an active role in effectively promoting RIS and OIS. This policy has guided enterprises to gradually transition towards a low-carbon direction by adjusting and eliminating industries, producing significant effects.

4.2. Robustness Tests

4.2.1. Parallel Trend Test

Before applying the difference-in-differences model, it is crucial to ensure the validity of the parallel trend assumption. The parallel trends assumption requires that the trends in industrial structure changes for both the treatment and control groups must be consistent prior to the implementation of the policy. To test this hypothesis, this study established time dummy variables encompassing four years both prior to and following the exogenous shocks associated with the pilot policy’s implementation. Subsequently, the following model was constructed:

$$Y_{it}={\beta }_0+{\sum }_{t=2005}^{2020}{\tau }_t{\beta }_t{treat}_2+{\beta X}_{it}+{\mu }_i+{\tau }_t+{\epsilon }_{it}$$

(5)

The definition of variables is the same as in Formula (1). This paper selects the year before the implementation of the CET (2012) as the baseline year, with the related test results shown in Figure 1 and Figure 2. Upon examining Figure 1, it becomes clear that none of the dummy variables before the carbon emissions trading policy’s implementation met the 10% significance threshold. This suggests that the changes in RIS and OIS prior to the policy’s enactment were statistically insignificant, aligning the findings of this research with the parallel trends hypothesis. Following the implementation of the pilot policy, the dummy variables in the fourth year demonstrated significant positive outcomes at the 5% significance level, indicating that the CET markedly fostered ISA. The policy’s influence on industrial structure displayed a temporal delay, presumably because regions underwent an adaptation phase post-enactment, necessitating a period for the policy effects to manifest and significantly alter regional industrial structures. Consequently, these findings uphold the parallel trends hypothesis.

4.2.2. Reduction in Sample Data

To mitigate potential biases arising from anomalies in the Tibetan data, which may contain more extreme values, this study performed a robustness check. This involved excluding Tibetan data from the sample and conducting additional regression analyses using Formula (5). The results, detailed in columns (1) and (2) of

Table 5. The regression results with reduced sample data and inclusion of the dummy variable nonex.

Variable	(1)	(2)	(3)	(4)
	RIS	OIS	RIS	OIS
did	6.4108 **	0.3032 **	6.3206 **	0.3315 **
	(2.5195)	(2.1729)	(2.4935)	(2.2570)
pgdp			0.0173	−0.1106
			(0.2597)	(−1.1306)
invest	0.9976	0.3238 **	0.7153	0.3361 **
	(0.6411)	(2.2714)	(0.5465)	(2.2709)
urban	−0.1510	0.0610	−0.2793	0.0838
	(−0.1560)	(0.6112)	(−0.2846)	(0.8252)
fiscal	23.9452 ***	−2.1127 **	22.1817 ***	−1.7706 *
	(5.1670)	(−2.0924)	(4.8545)	(−1.7508)
infra	−0.8969 **	0.4063	2.0414	−0.6452
	(−5.3784)	(0.5681)	(0.4555)	(−1.4626)
internet	3.0516	−0.0565	3.5293	−0.0607
	(0.8298)	(−0.5558)	(1.0314)	(−0.5340)
control variable	29.9543 **	2.2867 ***	28.2559 **	2.0885 ***
Area-fixed effects	(2.0654)	(5.5206)	(2.1173)	(4.7380)
Time-fixed effects	Yes	Yes	Yes	Yes
Constant	Yes	Yes	Yes	Yes
	−19.3060	−1.6128 *	−17.0321	−1.6177
Sample size	(−1.4171)	(−1.6777)	(−1.4245)	(−1.5660)
R2	480	480	496	496

Note: t-statistic values are in parentheses, and ***, **, and * indicate 1%, 5%, and 10% significance levels, respectively.

, demonstrate that the positive effects of the CET are still evident, even without the Tibetan data. This robustness of the baseline model’s estimations suggests that outliers do not significantly impact the findings. Accordingly, this procedure increases the reliability and accuracy of the conclusions regarding the CET’s impact on regional industrial structures.

4.2.3. Impact of the CET Market in Non-Pilot Provinces

In December 2016, Sichuan and Fujian, two non-pilot provinces, inaugurated their carbon emissions trading markets. To evaluate the influence of these markets on the estimation outcomes, Formula (5) incorporated a dummy variable, ‘nonex’, to identify prefecture-level cities in these provinces and included an interaction term for the year the markets commenced (2017). The integration of the ‘nonex’ variable into Formula (5) facilitated the display of regression results in columns (3) and (4) of Table 5.

Analysis of the estimation results shows that the ‘did’ coefficient aligns with prior findings, affirming that the carbon trading pilot policy continues to significantly influence OIS, including in non-pilot provinces. This reinforces the robustness of the baseline model’s estimates, confirming that the paper’s conclusions are applicable not only to pilot provinces but also remain valid across non-pilot provinces with CET markets.

4.2.4. Placebo Test

To reduce the influence of potential confounders on the correlation between policy dummy variables and industrial structure, this study employs counterfactual assumptions for robustness checks. It introduces new ‘policy times’ by shifting the policy implementation timeline forward by two and four years. These new policy dummy variables are generated through interactions with the treatment group’s dummy variables; non-significant results suggest that the significant outcomes previously reported are not artifacts of random variation, thus affirming the robustness of these findings. As illustrated in

Table 6. Placebo test results.

Variable	Policy Two Years Ahead		Policy Four Years Ahead
	RIS	OIS	RIS	OIS
did	0.0016	0.0007	0.0019	0.0003
	(0.5327)	(0.2389)	(0.3159)	(0.2913)
pgdp	0.8724	0.4612 ***	0.7613	0.3238 **
	(0.5716)	(3.3119)	(0.3917)	(2.2412)
invest	−0.1826	0.0481	−0.3126	0.0724
	(−0.1637)	(0.4558)	(−0.2986)	(0.6154)
urban	19.8713 ***	−2.3159 **	21.2734 ***	−1.9531 *
	(4.6948)	(−2.1759)	(3.9716)	(−1.7249)
fiscal	−1.3297 ***	0.6179	0.9847	−0.5981
	(−4.8492)	(0.6651)	(0.2684)	(−1.3712)
infra	2.8914	−0.0792	3.1891	−0.0824
	(0.8721)	(−0.8972)	(1.0142)	(−0.7759)
internet	30.6102 **	2.0714 ***	34.6441 ***	2.0144 ***
	(2.1729)	(5.4416)	(3.0658)	(4.1298)
Area-fixed effects	Yes	Yes	Yes	Yes
Time-fixed effects	Yes	Yes	Yes	Yes
Constant	−17.5187	−1.5147	−16.5981	−1.6255
	(−1.3162)	(−1.2194)	(−1.3648)	(−1.2387)
Sample size	496	496	496	496
R2	0.872	0.901	0.897	0.911

Note: t-statistic values are in parentheses, and ***, **, and * indicate 1%, 5%, and 10% significance levels, respectively.

, the influence of the shifted policy dummy variables on industrial structure is inconsequential, thereby corroborating the study’s conclusions.

4.2.5. CET Market Mechanism Test

CET synergistically combines market mechanisms with government intervention, substantially enhancing the effectiveness of environmental regulation (refer to Figure 3). However, existing studies primarily illustrate the mean effects of these interventions on ISA. Comparative analyses reveal that, unlike the well-established European Union Emissions Trading Scheme (EU-ETS), China’s nascent CET market suffers from an underdeveloped operational mechanism [57]. Overly intrusive government intervention stifles market development, inhibiting the efficacy of market mechanisms. To meticulously demonstrate the influence of market forces within CET on ISA, this study quantifies the market’s role. Given that carbon prices are governed by market supply and demand, they serve as indirect indicators of market force potency [58]. Accordingly, this paper develops a bidirectional fixed-effects difference-in-differences model to evaluate the impact of market forces under CET on ISA.

$$Y_{it}={\gamma }_0+{\gamma }_1{did}_{it}+{\gamma }_2{did}_{it}*{price}_{it}+{\beta X}_{it}+{\mu }_i+{\tau }_t+{\epsilon }_{it}$$

(6)

In the above formula, price_it denotes the carbon price of region i at moment t. In this paper, we choose the logarithmic value of the annual average of the daily transacted price in the carbon trading market to measure it, and the data comes from the China Carbon Trading Platform, and the other symbols have the same meaning as above.

Columns one and two of

Table 7. Regression results of the market mechanism test and the expected effect test.

Variable	(1)	(2)	(3)	(4)
	RIS	OIS	RIS	OIS
did	6.1382 **	0.3117 *	6.0237 **	0.3984
	(2.1794)	(1.8129)	(2.2541)	(2.273)
Did * price	0.7491 **	0.4258 **
	(2.3482)	(2.0853)
pre1			0.0127	0.0081
			(0.6372)	(0.4839)
control variable	yes	yes	yes	yes
Area-fixed effects	yes	yes	yes	yes
Time-fixed effects	yes	yes	yes	yes
Constant	3.2912 ***	1.1476 ***	2.7459 ***	1.0116 ***
	(5.8436)	(7.2681)	(4.1679)	(5.8967)
Sample size	496	496	496	496
R2	0.764	0.811	0.791	0.742

Note: t-statistic values are in parentheses, and ***, **, and * indicate 1%, 5%, and 10% significance levels, respectively.

empirically tested the impact of carbon pricing on ISA using a difference-in-differences model. The regression results indicate that, whether for RIS or OIS, the did * price is significant at the 5% level. This implies that when carbon pricing represents the CET market mechanism, the carbon trading market mechanism can significantly promote the development of regional RIS and OIS.

4.2.6. Expected Effects Test

In reality, there is a possibility that market participants may anticipate the imminent implementation of the CET and alter their decision-making behaviors in advance, which could confound the effects of the CET implementation. In order to avoid the above situation, this paper examines whether the expected effect exists prior to the implementation of CET, and it further adds an interaction term between the policy grouping dummy variable and the dummy variable for 1 year prior to the implementation of CET to Formula 1 [59]. Columns (3) and (4) of

Table 8. Technological innovation mechanism test of CET affecting ISA.

Variant	(1)	(2)	(3)	(4)	(5)
	RIS	OIS	inno	RIS	OIS
did	6.3351 **	0.3322 **	0.3327 **	6.5729 **	0.3668 **
	(2.4780)	(2.2586)	(2.4255)	(2.5111)	(2.3912)
inno				1.2531 ***	8.8309 *
				(3.1129)	(1.7354)
pgdp	0.5920	0.3346 **	0.8608 ***	0.1382	0.3135 **
	(0.4487)	(2.2715)	(5.7469)	(0.1033)	(2.3111)
invest	−0.2890	0.0842	0.0291	−0.2582	0.0913
	(−0.2974)	(0.8295)	(0.5635)	(−0.2939)	(0.9150)
urban	23.2367 ***	−1.7583 *	6.4124 ***	24.9374 ***	−1.3367 *
	(4.9772)	(−1.7605)	(6.3923)	(4.6525)	(−1.6991)
fiscal	2.0278	−0.6363	0.2347	5.7182	−0.3995
	(0.4572)	(−1.4587)	(0.3642)	(1.4000)	(−0.9963)
infra	3.5203	−0.0615	0.4391 **	4.4878	−0.0843
	(1.0351)	(−0.5392)	(2.2507)	(1.3373)	(−0.6898)
internet	27.9832 **	2.0865 ***	1.4373 **	24.7001 *	1.8841 ***
	(2.0909)	(4.7758)	(2.5757)	(1.8903)	(4.9868)
Area-fixed effects	Yes	Yes	Yes	Yes	Yes
Time-fixed effect	Yes	Yes	Yes	Yes	Yes
Constant	−16.0818	−1.6196	−3.6164 ***	−15.6071	−1.6989 *
	(−1.3292)	(−1.5745)	(−3.0761)	(−1.3161)	(−1.7202)
Sample size	496	496	496	496	496
R2	0.876	0.891	0.943	0.874	0.892

Note: t-statistic values are in parentheses, and ***, **, and * indicate 1%, 5%, and 10% significance levels, respectively.

show that although the coefficient estimates of the anticipatory effect variables are positive, they are not significant and lack statistical significance. At the same time, the coefficient estimates of the core explanatory variables remain significantly positive. This suggests that the exogeneity of CET is strong and that it has not formed a significant anticipatory policy effect.

4.3. Analysis of Impact Mechanisms

The findings indicate that CET effectively fosters ISA, although the specific mechanism through which this policy influences industrial transformation remains unverified. As per the Porter hypothesis, moderately enhanced government environmental policies can catalyze technological innovation and optimize resource allocation, thus improving resource utilization efficiency and simultaneously advancing environmental protection and industrial restructuring [28]. Technological innovation acts as a primary driver of ISA, facilitating regional industrial upgrades through sustained research and development, the implementation of new technologies, scale adjustments, and production method enhancements. As a potent market-based environmental policy instrument, CET stimulates technological innovation within enterprises, promoting the transformation of regional industrial structures. To validate the described impact mechanism, this study applied stepwise regression to assess mediating effects [29] and developed a model based on Formula (1):

$$M_{i,t}={\gamma }_0+{\gamma }_1{did}_{i,t}+{\gamma }_2{X}_{i,t}+{\mu }_i+{\tau }_t+{\epsilon }_{i,t}$$

(7)

$$Y_{i,t}={\rho }_0+{\rho }_1{did}_{i,t}+{\rho }_2{M}_{i,t}+{\rho }_3{X}_{i,t}+{\mu }_i+{\tau }_t+{\epsilon }_{i,t}$$

(8)

In the formula, $M_{i,t}$ represents the mediating variable, specifically referring to the technology innovation variable (inno), which is expressed by the logarithm of the total number of patent grants. The data are sourced from the National Bureau of Statistics of China and the Statistical Yearbook of China. The definitions of other variables are the same as in Formula (1). If ${\alpha }_1$ in Formula (1), ${\gamma }_1$ in Formula (7), and ${\rho }_2$ in Formula (8) are all significant, this indicates the presence of a mediating effect. In this case, if the regression coefficient ${\rho }_1$ is also significant, it constitutes a partial mediation effect; if ${\rho }_1$ is not significant, it constitutes a complete mediation effect.

Regressions are performed sequentially on Formulas (1), (7), and (8), with the results shown in Table 8. As can be seen from column (3), ${\gamma }_1$ is significantly positive at the 5% level, indicating that the implementation of the CET has a positive impact on the technological innovation of regional enterprises. Furthermore, columns (4) and (5) indicate that the level of technological innovation has a significant positive effect on RIS and OIS. Therefore, CET can promote technological innovation in enterprises, thus facilitating the transformation of the regional industrial structure.

4.4. Heterogeneity Analysis

Given the vast expanse of China, the impact of CET can differ significantly across regions, influenced by diverse natural environments and varying levels of economic development. While previous research has validated the policy’s beneficial effects on transforming industrial structures and elucidated its mechanisms, regional heterogeneity demands further investigation to ascertain if significant disparities in its effectiveness exist among different areas.

In this study, China’s 31 provinces were categorized into three regions—East, Central, and West—to assess the impact of CET on regional industrial structures. Regression results are detailed in

Table 9. Regression results on the regional heterogeneity of CET.

Variable	RIS			OIS
Area	East	Central	West	East	Central	West
did	8.8142 ***	1.1032 **	1.9694 ***	0.4385 **	0.0390	−0.1221
	(3.5494)	(2.1518)	(4.9669)	(2.0897)	(0.8965)	(−0.9691)
pgdp	2.8026	4.0742 *	−0.9324 *	0.5054 **	0.1802	−0.1358
	(0.8310)	(1.8620)	(−1.7814)	(1.9803)	(1.4052)	(−1.0648)
invest	−0.3224	0.3104	0.6606	0.0332	−0.1215	−0.2800 *
	(−0.2534)	(0.1865)	(0.8263)	(0.4103)	(−0.7516)	(−1.7577)
urban	28.5462	−26.8691	11.4110 ***	−4.4294 **	−0.2392	0.9091
	(1.0464)	(−1.2594)	(2.9419)	(−2.1091)	(−0.1765)	(0.7467)
fiscal	−2.7221	5.9188	1.6675	2.0072	1.0200	0.6405 **
	(−0.0615)	(0.5021)	(1.2779)	(1.5468)	(0.5880)	(2.0728)
infra	9.7516	−1.5684	0.1634	0.0840	−0.2197 ***	0.1231 *
	(1.0581)	(−0.8296)	(0.6623)	(0.4367)	(−3.2601)	(1.9436)
internet	83.3363 *	32.9492 ***	10.7691 ***	3.2668 ***	2.2060 ***	0.6097 *
	(1.7558)	(3.2920)	(3.7864)	(5.0116)	(3.7210)	(1.7062)
Area-fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Time-fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Constant	−48.2838	−27.5263 **	5.4619	−2.4818	−0.9797	1.9183 **
	(−1.5129)	(−2.0667)	(1.4722)	(−1.5712)	(−0.9949)	(2.2265)
Sample size	192	144	160	192	144	160
R2	0.2900	0.4685	0.6537	0.7745	0.6177	0.3361

Note: t-statistic values are in parentheses, and ***, **, and * indicate 1%, 5%, and 10% significance levels, respectively.

. The analysis reveals that the policy significantly enhanced both RIS and OIS in the Eastern region. In contrast, although the policy supported the RIS in the Central and Western regions, its effect on their structural upgrading was not significant.

Specifically, the impact of CET is modulated by regional factors, including the natural environment and economic development levels. The Eastern region, with its advanced economic and technological innovation capabilities, stands in contrast to the Central and Western regions. Upon implementation of this policy, enterprises in the Eastern region can rapidly implement technological innovations, facilitating both the RIS and OIS. In contrast, although the policy helps eliminate outdated industries and streamlines the industrial structure in the Central and Western regions, enterprises there often lack the technological innovation capacity needed for advanced industrial development. Thus, the policy’s effectiveness in promoting high-level industrial upgrading in these regions is limited.

5. Conclusions and Discussion

5.1. Conclusions

This article uses provincial panel data from China spanning 2005 to 2020 and employs a difference-in-differences approach to empirically investigate the effects of CET on ISA. To validate the findings, the study conducts robustness checks, heterogeneity analyses, and mechanism tests. The key findings are as follows: First, CET significantly enhances ISA, facilitating both the RIS and OIS while effectively reducing carbon emissions. Second, these policies drive industrial adjustment through technological innovation, which serves as a mediating variable. This relationship is analyzed through stepwise regression, demonstrating that the policies incentivize technological innovation within enterprises, thus fostering regional industrial adjustment. Third, the impact of CET exhibits considerable regional variability. Differences in natural environments and economic development levels result in distinct outcomes in resource allocation via market mechanisms across the eastern, central, and western regions of China. Specifically, these policies considerably advance both the RIS and OIS in the eastern region, whereas in the central and western regions, they primarily enhance the RIS without significantly impacting their upgrading.

5.2. Policy Implications

Drawing on the conclusions outlined above, this paper suggests several policy implications. First, it recommends enhancing the CET mechanism and refining policy implementation details. While stronger local CET can catalyze ISA, variations in developmental stages and economic conditions across China necessitate that local governments customize these policies to fit regional contexts, systematically examining their implementation and impacts. This examination should encompass policy design as well as the identification and resolution of challenges encountered during execution, thereby optimizing the CET mechanism and capitalizing on opportunities for technological innovation within enterprises. Second, it is imperative to harness the synergistic effects of regional policies. The policy-making process should account for interactions and strategic behaviors among local governments, address the spatial spillover effects of the CET, and enhance cross-regional coordination. This approach encourages local governments to foster healthy competition in policy development and to establish robust compensation mechanisms. Local authorities should tailor their strategies to planned industrial layouts, aligning with regional realities and strengths, to facilitate the efficient redistribution of labor and capital, stimulate enterprise innovation, and advance OIS. Third, this paper advocates for the integration of technological innovation into the factor allocation process. Government agencies should vigorously pursue market reforms, streamline CET market mechanisms, ensure that markets play a pivotal role in resource distribution, increase the engagement of market actors, compel polluting enterprises to internalize environmental management costs, and improve overall resource allocation efficiency.

5.3. Limitations and Prospects

This study has the following limitations: First, due to the unavailability of certain data in China, this paper utilizes provincial panel data, preventing a focus on micro-level aspects. Future research should aim to evaluate the effects of CET from a micro perspective. Second, China’s national CET market was officially launched in July 2021. The relatively short establishment period may mean that the policy effects have not fully emerged. Additionally, the lack of timely data release by the Chinese government has led to the exclusion of the national CET market from this study. Future research should continuously monitor and analyze the long-term impact of the national CET market to provide more effective recommendations.