The Impact of the Circular Economy Pilot Policy on Carbon Emissions in Chinese Cities and Its Underlying Mechanisms

Abstract

The development of the circular economy is an important way for China to achieve its carbon peak and carbon neutrality goals on schedule. In this paper, we use the construction of circular economy demonstration cities as a quasi-natural experiment to systematically evaluate the carbon reduction effect of the circular economy pilot policy using the multi-period Differences-in-Differences (DID) model and Spatial-Differences-in-Differences (Spatial-DID) model. The research findings indicate that the circular economy pilot policy may effectively restrain the intensity of carbon emissions and the volume of carbon emissions, primarily driving carbon reduction in resource-based cities, old industrial base cities, and cities in the central regions, rather than universally exerting a significant impact on energy conservation and carbon reduction in all cities. Government investment in technology and the improvement of factor allocation structure play a mediating role in the carbon reduction effect of the circular economy pilot policy, while the intensity of public management plays a moderating role. When considering the externalities of urban networks, it becomes evident that the policy exhibits a notable spatial spillover effect. This not only significantly propels local efforts to reduce carbon emissions but also exerts a “demonstration effect” on the surrounding areas. The spillover effect on carbon emissions volume surpasses that on carbon emission intensity. This study offers empirical evidence for the ongoing promotion of the circular economy pilot policy nationwide, facilitating the achievement of cities’ dual carbon goals.

1. Introduction

The extensive development model, which has long relied on material and resource inputs, has created miraculous economic growth in China, but it has also led to insufficient momentum for sustainable urban development. In contrast, the circular economy, with its core focus on resource efficiency and recycling, represents a sustainable development model characterized by the material closed-loop flow of “resources-products-consumption-regenerated resources”. Therefore, the transition from an extensive development model to a circular economy model is imperative. Against this backdrop, the policy of piloting circular economy demonstration cities has emerged. Two batches of pilot cities for the circular economy have been launched, expanding from 40 cities (counties) in 2013 to 101 cities (counties) in 2015. This policy primarily revolves around transformation tasks such as industrial structure, urban construction, public services, resources, and environmental protection, and undertakes the important responsibility of promoting low-carbon and sustainable urban development. The circular economy pilot policy aims to promote the recycling and reuse of renewable resources, with each type of renewable resource reuse capable of reducing carbon emissions in the “extracting raw materials and initial processing of raw materials” stages. Further subdividing into different industries, the recycling of renewable resources can include the reuse of scrap steel, recycled aluminum, recycled paper, recycled plastics, etc., reducing carbon emissions during the production of primary resources. According to data from the Renewable Resources Information Network, in 2019, China recycled 2.34 million tons of scrap steel, 0.07 million tons of scrap aluminum, 0.5 million tons of scrap paper, and 0.19 million tons of scrap plastic, with a cumulative emission reduction of 1.1 billion tons of CO₂ (https://mp.weixin.qq.com/s?__biz=MzA3MDE5NDAzNA==&mid=2651349413&idx=1&sn=5fcbc8f3148e4b184e0e97407c15256d&chksm=853c47b8b24bceaec70251f79f449910d0a972c936f0889e2cef7e63b21fdf388a6e55bbd41a&scene=27, accessed on 10 June 2024). Upon the completion of the pilot work in the two batches of circular economy demonstration cities and with the imminent commencement of nationwide circular economy construction, it is of paramount practical significance to elucidate the carbon reduction effect, operational mechanism, and heterogeneity of the regional circular economy pilot policy. Furthermore, it is imperative to examine the spillover effects of the circular economy pilot policy to advance nationwide circular economy construction and achieve the “dual carbon” goals.

There are two categories of literature related to this subject, one of which is research literature related to the circular economy. Primarily, it revolves around the conceptual definition and limitations of the circular economy [1], the measurement framework and indicator system of the circular economy [2,3], the goals or strategic orientation of the circular economy [4], the driving factors of the circular economy [5,6,7], and the industrial processes and technological solutions for circular recycling [8]. Among them, the study of circular economy involves research subjects such as the recycling of batteries and the reuse of household waste [9,10]. Regarding the systematic evaluation and causal investigation of the implementation effects and influence mechanisms of the circular economy pilot policy, including their impact on economic growth, technological innovation, pollution emissions, and environmental performance. The literature research reveals that the implementation of the circular economy pilot policy can lead to a slowdown in economic growth [11], technological innovation [12], and reduction in air pollutants [13]. The effectiveness of circular economy policies in different countries depends on green public procurement, advocates of circular economy, recycling technologies, and centralized governance, as demonstrated by Iannone [14] who examined European Union countries as well as China, Japan, and the United States and found that green public procurement contributes to advancing circular economy policies in more developed economies. Lazarevic et al. [15] discovered that the effectiveness of Finland’s circular economy policy is related to recycling technologies and the optimization of resource-intensive system configurations. Droege et al. [16] argue that the implementation and evaluation of Portugal’s circular economy policy require politically astute advocates of circular economy who can openly and resolutely implement projects related to circular economy policies. Alberich et al. [17] focused on the European Union’s circular economy policies from 2011 to 2022 and found that the effectiveness of policies depends on recycling technologies and centralized governance. The impact of the circular economy pilot policy on environmental performance is twofold; on the one hand, it may enhance environmental performance improvement, increasing resource utilization efficiency [18,19,20], while on the other hand, it may compel businesses to acquire resource recycling equipment, squeezing production costs and potentially weakening market competitiveness, which is not conducive to the adoption of resource recycling technologies and enhancing environmental performance [21,22,23,24]. The reasons for the uncertainty of policy effects are closely linked to government financial support for R&D [25,26], public governance [27], and the efficiency of renewable resource markets [7], among other factors.

Similarly, the impact of the circular economy pilot policy on carbon emissions is uncertain and closely related to national and institutional quality. Wang et al. [28] utilized the Augmented Mean Group (AMG) technique and found a close relationship between urban waste recycling policies and carbon emission intensity in the top seven carbon-emitting countries in the world. Mawutor et al. [29] analyzed Ghanan data from 2006 to 2017 and discovered a non-linear inverted U-shaped relationship between the circular economy pilot policy and carbon emissions, with the inflection point of the inverted U-shape being related to robust and efficient institutional governance quality; that is, when the institutional governance quality threshold exceeds 1.155, circular economy development can promote carbon reduction. Another category of literature pertains to scholarly works on urban carbon reduction, primarily encompassing the Copenhagen International Climate Accord [30], the Paris Accord [31], new energy initiatives [32], carbon taxes [33], the low-carbon city pilot policy [34,35], as well as the carbon trading policy [36,37] and other regulatory measures, all contributing to the reduction in urban carbon emissions.

Can the pilot policy of circular economy implemented by China in 2013 and 2015 achieve the decoupling of economic growth and carbon emissions, and thus achieve the goals of peaking carbon emissions and achieving carbon neutrality as planned? Few studies consider the construction of circular economy demonstration cities implemented by China as a quasi-natural experiment to explore the carbon emission reduction effects of the circular economy pilot policy. Therefore, the main contributions of this study include the following: firstly, utilizing a multi-period DID model to investigate the impact of the pilot policy on carbon emission intensity and carbon emissions, providing policy references for promoting the development of a circular economy at the city level. Secondly, comparing and analyzing the differences in the mechanisms of the pilot policy on carbon emission intensity and carbon emissions, and examining the heterogeneity of policy effects from the perspectives of resource endowment, industrialization level, and economic foundation, providing references for formulating a tailored circular economy pilot policy. Thirdly, in addition to considering the local effects of the pilot policy, this paper also studies the spillover effects of the circular economy pilot policy and applies a spatial DID model to examine the spatial spillover effects of this policy. This paper not only contributes to the comprehensive promotion of the circular economy pilot policy construction in China but also provides a theoretical basis for achieving the goals of peaking carbon emissions and achieving carbon neutrality.

2. Policy Background and Research Hypotheses

2.1. Policy Background

In 2005, the State Council issued the first policy document related to the circular economy, namely “Several Opinions of the State Council on Accelerating the Development of Circular Economy”, explicitly proposing the three fundamental principles of “reduction, reuse, and recycling”. It advocated adopting various effective measures to minimize resource consumption and environmental impact, thereby maximizing resource utilization efficiency. Subsequently, the “Circular Economy Promotion Law of the People’s Republic of China” was promulgated in 2008, further providing incentives and legal safeguards for the governance of the circular economy. To implement the practice of the circular economy pilot policy, the “Notice of the National Development and Reform Commission on Organizing the Creation of Demonstration Cities (Counties) for Circular Economy” issued in 2013 (hereinafter referred to as the “Notice”) elucidates the ten major evaluation indicators, including the socioeconomic development level, resource output level, and reduction measures of the demonstration cities. It is also planned to select approximately 100 circular economy demonstration cities (counties) in two batches during the “12th Five-Year Plan” period. In 2013, 40 cities and counties were selected as circular economy exemplars. In 2015, another 61 cities and counties were recognized for their circular economy efforts. The report from the 19th National Congress of the Communist Party of China in 2017 further advocated for the establishment of a comprehensive green and low-carbon circular economic system. To implement this strategic decision, the State Council’s 2021 “Guidance on Accelerating the Establishment of a Green, Low-Carbon Circular Economic System” proposed measures to perfect green, low-carbon, and circular development in production, distribution, consumption, and innovation. The report of the 20th National Congress of the Communist Party of China in 2022 (the complete text of the “Twenty Major Reports” can be found at the following website: https://www.gov.cn/xinwen/2022-10/25/content_5721685.htm, accessed on 10 June 2024) reiterated the commitment to the efficient and circular utilization of all types of resources. Since then, the circular economy pilot policy demonstration cities have gradually become an integral pathway for the advancement of circular economy development and the achievement of peak carbon and carbon neutrality goals.

2.2. Research Hypotheses

The effectiveness of environmental policies in achieving carbon emission reduction depends on the efficient transmission of policies among the central government, local governments, and enterprises. On one hand, there is the transmission mechanism between the central government and local governments. During the selection process of circular economy demonstration cities (counties), local governments voluntarily choose to participate in the selection after weighing officials’ performance evaluations, economic costs, and local ecological environments. After the selection process, the central government requires local governments to formulate a 3–5-year development plan for a circular economy, and then provide support through specific central financial subsidies and market incentive policies such as green financial support. On the other hand, there is a transmission mechanism between local governments and enterprises. Looking at the construction planning of various circular economy demonstration cities (counties), local governments’ incentive measures for enterprises mainly include financial subsidies for circular economy projects and the selection of circular economy enterprise titles. For enterprises engaged in the research and application of circular economy technologies, local governments provide a certain proportion of financial subsidies for the research and application of intermediate products. The widespread application of intermediate products can optimize the structure of factor allocation, and improve the efficiency of energy and resource allocation, and the research and innovation of intermediate products can reduce the resource input per unit output. Therefore, the carbon emission reduction effect of the circular economy pilot policy depends on financial and technological support from local governments, improvement in factor allocation structure, and strong government supervision.

Based on the aforementioned theoretical analysis, and referencing existing pieces of literature [7,25,26], this study employs the construction of pilot circular economy cities as a quasi-natural experiment to delineate the mechanisms by which the circular economy pilot policy impacts carbon emission reduction. This is achieved by considering three dimensions: government technology investment, the amelioration of factor allocation structure, and the intensity of public management. Furthermore, it scrutinizes the spatial spillover effects of such policies from the perspective of urban network externalities.

Firstly, the circular economy pilot policy fosters carbon emission reduction through the influx of government technological investments. Technological advancements and innovations are pivotal to the low-carbon transition [38]. Following the implementation of the circular economy pilot policy, circular economy enterprises may face financial pressures in the research and application of recycling technologies, potentially squeezing out some production cost inputs, reducing market share, and thereby weakening market competitiveness. Considering that the key extraction technologies, environmental product technologies, energy-saving technologies, and comprehensive utilization technologies for recycling are not highly advanced, with a weak foundation for innovation, it is difficult to achieve breakthroughs in recycling technologies solely relying on the efforts of circular economy enterprises. Therefore, the circular economy pilot policy needs to be supported by financial and technological investments to foster innovation platforms for the circular economy and to create a regulatory environment conducive to the development of the circular economy. To ensure the research and innovation of recycling technology, the “Notice” clearly states that under equal conditions, priority should be given to investments related to circular economy demonstration cities. It also focuses on the need to support local governments in issuing bonds for circular economy development, encourage financial institutions to innovate green finance for circular economy development according to actual conditions, and provide free usable knowledge to enterprises through government technology investment [39]. Circular economy enterprises utilizing these recycling technologies not only reduce production costs from the practical activities of the circular economy pilot policy [40] but also help promote carbon reduction.

Secondly, the circular economy pilot policy is spurring carbon emission reduction through the refinement of factor allocation structures. Due to the relatively weak innovative foundation of recycling technology, circular economy enterprises tend to prioritize reducing the input of resource elements by adjusting the structure of factor allocation. In light of this, the “Notice” incorporates the three major principles of “reduction, reuse, and recycling” into industrial development planning, maximizing the utilization of energy and other resources, reducing resource pressure from the source of economic activities, minimizing waste pollution, and thereby transforming the economic growth model driven by resource elements. Consequently, the circular economy pilot policy can achieve the “reduction” of resource elements through the adjustment of factor allocation structure, and further achieve energy conservation and sustainable economic development through the improvement of renewable resource efficiency [7]. Given the absence of environmental constraints in the past, technological advancements tended to favor labor savings and capital enhancement; however, when environmental constraints are incorporated, technological progress leans towards labor enhancement and capital conservation [41], which signifies that the change in per capita capital is the primary driving force behind the growth of energy consumption in China [42], and the change in per capita capital can be utilized to reflect the changes in factor allocation structure.

Thirdly, the intensity of public management influences the carbon emission reduction effects of the circular economy pilot policy. When the government’s financial investment and the title of circular economy enterprises are insufficient to incentivize participation in the construction of circular economy demonstration cities, robust government supervision becomes crucial. Therefore, the successful implementation of circular economy pilot policies is inseparable from government public governance [27]. In this regard, the “Notice” mandates adherence to the “PRC Promotion of Circular Economy Law” among other legal and administrative regulations, including advancing the implementation of circular economy projects, strengthening organizational leadership, and ensuring the distribution of responsibilities. Enhanced public management vigor encourages pilot cities to fortify organizational leadership. By innovating coordinated management measures, it effectively fosters the division of green circular industry development responsibilities. Consequently, this ensures stringent adherence to local energy-saving and emission reduction targets and is conducive to advancing carbon emission reduction through the innovation of urban management systems.

Fourthly, the circular economy pilot policy exerts an influence on carbon emission reduction in the surrounding non-pilot areas, manifesting as either a “demonstration effect” or a “displacement effect”. On the one hand, the “demonstration effect” emerges when considering the ease of policy information acquisition and the similar resource endowments, which facilitate policy emulation among neighboring cities [43]. As pioneers of the circular economy, these demonstration cities draw keen interest from governments of adjacent non-pilot areas, provoking a competitive emulation of policies that efficiently recycle resources, thereby fostering carbon emission reduction through the effective utilization of fossil fuels and the recycling of renewable energy. On the other hand, the “displacement effect” suggests that since the issuance of the “Notice” with its goal of “enhanced resource utilization efficiency”, the resource and environmental regulations in circular economy demonstration cities have intensified. According to the “pollution haven hypothesis”, high-energy-consuming and highly polluting enterprises from demonstration cities may relocate to surrounding areas with less stringent regulations, potentially hindering carbon emission reduction efforts in those non-pilot areas.

In light of the above, this study posits the following hypotheses:

H1. 

The circular economy pilot policy facilitates carbon emission reduction.

H2. 

Governmental investment in technology and the improvement of the structure of factor allocation are conduits through which the circular economy pilot policy exerts its influence on carbon emission reduction.

H3. 

The efficacy of the circular economy pilot policy on carbon emission reduction is moderated by the extent of public administration.

H4. 

The impact of the circular economy pilot policy on carbon emission reduction in surrounding non-pilot areas depends on the interplay of the “demonstration effect” and the “displacement effect”.

3. Research Methodology

3.1. Model Construction

3.1.1. Multi-Period DID Model

The DID model is commonly employed in empirical research for policy evaluation. Its advantages lie in effectively avoiding the endogeneity issues caused by reverse causality, and to some extent alleviating bias problems due to omitted variables. Comparing before and after policy implementation allows for a more precise estimation of policy effects. There are a total of two batches of circular economy demonstration cities. The first batch of demonstration cities was publicly announced in 2013, and included 40 cities and counties. This paper selects 19 prefecture-level cities from the first batch as sample pilot cities, as shown in Appendix A 

Table A1. The initial roster of 40 exemplary cities and counties in the field of circular economy.

City Level	Eastern Region	Central Region	Western Region
Prefecture-level city	Suzhou, Weifang, Guangzhou, Quzhou, Nanping, Chengde	Hebi, Loudi, Tongling, Huangshi, Jilin, Jincheng	Wuzhou, Jinchang, Wuhai, Guangan, Pu’er, Chongqing Dazu, Shangluo
County-level city	Ninghai, Yongkang, Xintai, Yanqing, Taixing, Shishi, Gaoyang	Jieshou, Guixi, Gucheng, Zixing, Diaobinshan, Boai, Xiaoyi	Tongwei, Tiandong, Longli, Huolinguole, Ge’ermu, Shanshanxian, Yimen

Source of information: National Development and Reform Commission.

. The second batch of demonstration cities was publicly announced in 2015, and included 61 cities and counties. This paper selects 24 prefecture-level cities from the second batch as sample pilot cities, as shown in Appendix A 

Table A2. The second roster of 61 exemplary cities and counties in the field of circular economy.

City Level	Eastern Region	Central Region	Western Region
Prefecture-level city	Taizhou, Liaocheng, Xuzhou, Zhanjiang, Yangzhou, Qindao, Tianjin	Jinmen, Luoyang, Shenyang, Xinxiang, Anshan, Ji’an, Fuyang, Changsha	Baotou, Liupanshui, Baiyin, Shizuishan, Chongqing Qijiang, Lasa, Tongren, Luzhou, Chongqing, Hechuan, Qujing, Liuzhou
County-level city	Pingyuan, Anji, Zhaoyuan, Guangning, Luoding, Haining, Danyang	Anhua, Fengyang, Fengcheng, Zhijiang, Zhangshu, Fanchang, Anxiang, Qianjiang, Tonghe, Jianping, Taonan, Changge	Hancheng, Cengong, Linxia, Pujiang, Fuchuanyaozhu, Jianshebintuan’ershi, Chongqing Liangping, Xichong, Tuoketuo, Jianshebintuanyishi, Rikazeshi, Manasi, Jingchuan, Yongning, Qingtongxia, Datong, Xiangyun

Source of information: National Development and Reform Commission.

. The DID model is primarily designed for situations involving multiple treatment group individuals and control group individuals, with a certain degree of similarity between the treatment and control groups. The synthetic control method is suitable for policy evaluation issues with fewer treatment group individuals. Therefore, this paper opts for a multi-period DID model based on the sample size of the treatment group. To examine the carbon emission reduction effects of the circular economy pilot policy, this study selects the lists of the first two batches of circular economy pilot cities based on the sample scope and constructs the following multi-period DID model:

$$Y_{it}={\alpha }_0+{\alpha }_1{D}_{it}+{\alpha }_{2}X+{\varphi }_i+{\eta }_t+{\epsilon }_{it}$$

(1)

where $i$ represents the city; $t$ denotes the year; $Y_{it}$ is the dependent variable indicating carbon emission intensity and total carbon emissions; $Trea{t}_i$ signifies whether a city is a circular economy demonstration city, with a value of 1 for demonstration cities, and 0 for others; and $Tim{e}_{it}$ indicates the inception of the circular economy pilot policy in demonstration cities, assigned a value of 1 upon policy implementation, and 0 otherwise. The explanatory variable ${\mathit{Treat}}_i\times {\mathit{Time}}_{it}$ is a dummy variable for the construction of circular economy demonstration cities, represented by $D_{it}$. ${\alpha }_1$ is the coefficient measuring the impact of the circular economy pilot policy; if ${\alpha }_1$ is significantly less than 0, it suggests that the circular economy pilot policy contributes to carbon emission reduction, thereby validating Hypothesis 1. ${\alpha }_0$ is the constant term, ${\alpha }_2$ is the coefficient for control variables, ${\varphi }_i$ represents regional fixed effects, ${\eta }_t$ accounts for time fixed effects, and ${\epsilon }_{it}$ denotes the random error term.

3.1.2. Mediation Effects Model

The mediating effect model is a common method for exploring causal mechanisms, consisting of a mediating variable, an independent variable, and a dependent variable. The relationship among the three is as follows: if the independent variable X exerts a certain influence on the dependent variable $Y_{it}$ through a variable $M_{it}$, then $M_{it}$ is considered a mediating variable between $D_{it}$ and $Y_{it}$. To delve deeper into the transmission mechanism between the circular economy pilot policy and urban carbon emissions, to explore the role played by intermediary variables such as government technology investment and improvements in factor allocation structure, this paper references the stepwise regression method proposed by Baron and Kenny [44] to construct specific mediation effect models, and refers to the stepwise regression coefficient testing method proposed by Qi et al. [45] to conduct specific model tests as follows:

$$M_{it}={\beta }_0+{\beta }_1{D}_{it}+{\beta }_{2}X+{\varphi }_i+{\eta }_t+{\epsilon }_{it}$$

(2)

$$Y_{it}={\gamma }_0+{\gamma }_1{D}_{it}+{\gamma }_2{M}_{it}+{\gamma }_{3}X+{\varphi }_i+{\eta }_t+{\epsilon }_{it}$$

(3)

where $M_{it}$ represents the mediating variable, denoted, respectively, by government technology investment ($TE$) and the structure of factor allocation ($FS$); ${\beta }_1$ stands for the impact coefficient of the circular economy pilot policy on the mediating variables, and a coefficient significantly greater than zero indicates that the pilot policy has a positive impetus on the mediating variables; ${\beta }_0$ is the constant term; and ${\beta }_2$ represents the coefficient for the control variables.

3.1.3. Moderation Effects Model

Based on the aforementioned analysis of mechanisms, the carbon emission reduction effect of the pilot policy in the realm of circular economy may be intricately linked to the magnitude of public administration efforts ($GME$). Consequently, this research incorporates $GME$ as a regulating variable, precisely integrating the product of $GME$ and $D$, along with $GME$ itself, as explanatory variables into Equation (1) to derive the following model.

$$Y_{it}={\theta }_0+{\theta }_1{D}_{it}+{\theta }_{2}GM{E}_{it}+{\theta }_{3}X+{\varphi }_i+{\eta }_t+{\epsilon }_{it}$$

(4)

$$Y_{it}={\lambda }_0+{\lambda }_1{D}_{it}+{\lambda }_{2}GM{E}_{it}+{\lambda }_3{D}_{it}\times GM{E}_{it}+{\lambda }_{4}X+{\varphi }_i+{\eta }_t+{\epsilon }_{it}$$

(5)

where if the coefficient ${\gamma }_3$ is positive, it signifies that $GME$ positively influences the carbon emission reduction effect of the circular economy pilot policy, hence confirming the validity of the moderating effect. ${\lambda }_0$ represents the constant term, ${\lambda }_1$ the impact coefficient of the circular economy pilot policy, ${\lambda }_2$ is the influence coefficient of public administration intensity, and ${\lambda }_4$ is the coefficient of the control variables. The remaining variables are consistent with those in Equation (1).

3.1.4. SDID Model

The utilization of Model (1) presupposes mutual independence among cities; specifically, the investigation samples of each city are unaffected by whether other cities have joined as demonstration cities for the circular economy. According to Hypothesis 3, the construction of demonstration cities for the circular economy may not only impact the pilot cities but also extend its influence to neighboring non-pilot cities, thus giving rise to spatial spillover effects. Consequently, the causal effects evaluated by the multi-period DID model will become ineffective, necessitating an expansion of the model. Therefore, this study draws inspiration from the methods proposed by Dubé et al. [46] and Li and Du [47] to construct the following SDID model for the assessment of spatial spillover effects of the circular economy pilot policy.

$$Y_{it}={\phi }_0+\rho W\times Y_{it}+{\phi }_1{D}_{it}+{\phi }_{2}W\times D_{it}+{\phi }_3{X}_{it}+{\phi }_{4}W\times X_{it}+{\varphi }_i+{\eta }_t+{\left(1-\delta W\right)}^{-1}{\epsilon }_{it}$$

(6)

where $W$ encompasses spatial adjacency, geographic distance, and economic distance spatial weight matrices; $\rho$ denotes the spatial autocorrelation coefficient of carbon emissions; ${\phi }_1$ signifies the impact coefficient of the circular economy pilot policy; ${\phi }_2$ represents the spatial spillover effects of the circular economy pilot policy; ${\phi }_3$ captures the spillover effects of control variables; ${\phi }_4$ portrays the spatial spillover effects of control variables; $\delta$ denotes the spatial autocorrelation coefficient of random errors; and the remaining variables correspond to Equation (1). Equation (4) presents the general form of the SDID model, and based on whether the correlation coefficients are zero, the model can be differentiated into the Spatial Error DID model (SEM-DID), the Spatial Lag DID model (SLM-DID), and the Spatial Durbin model (SDM-DID). The appropriate SDID model needs to be selected through correlation tests.

3.2. Variable Description

The urban carbon emissions calculated in this study encompass the sum of carbon emissions resulting from fossil fuel combustion and industrial production processes. Specifically, the carbon emissions from fossil fuel combustion are based on the carbon emissions caused by 17 types of energy consumption in 47 socio-economic sectors published in the CEADs database, as well as the carbon emissions generated by nine industrial production processes. The specific calculation Formulas (7) and (8) are as follows:

$$C{E}_{energy}={\displaystyle \sum _{m=1}^{47}{\displaystyle \sum _{n=1}^{17}C{E}_{mn}=}}{\displaystyle \sum _{m=1}^{47}{\displaystyle \sum _{n=1}^{17}\left(E_{mn}\times NC{V}_n\times E{F}_n\times O_{mn}\right)}}$$

(7)

$$C{E}_{process}={\displaystyle \sum _{t=1}^{9}C{E}_t=}{\displaystyle \sum _{t=1}^9\left(P_t\times E{F}_t\right)}$$

(8)

In the equation, $C{E}_{energy}$ represents the carbon emissions resulting from fossil fuel combustion, $m$ denotes the socio-economic sectors, $n$ signifies the types of energy, $E_{mn}$ stands for the consumption of type $m$ of energy sector $n$, $NC{V}_n$ represents the net calorific value of type $n$ of energy, $E{F}_n$ denotes the carbon emission coefficient of type $n$ of energy, and $O_{mn}$ symbolizes the oxidation rate of type $m$ of energy sector $n$. $C{E}_{process}$ represents the carbon emissions from industrial production processes, $t$ signifies the types of industrial production processes, $P_t$ represents the output of industrial products, and $E{F}_t$ denotes the carbon emission coefficient of the industrial production process.

Based on the calculated urban carbon emissions, the carbon emission intensity ($CI$) and carbon emissions ($CE$) are derived. These variables serve as the dependent variables. The carbon intensity ($CI$) is defined as the ratio of urban carbon emissions to GDP. Furthermore, the construction of a circular economy demonstration city is regarded as a quasi-natural experiment, serving as the explanatory variable ($D$). In cities or time points where the policy has not been implemented, $D$ equals 0, while in pilot cities where the policy has been implemented, $D$ equals 1. Considering other factors that may influence carbon emissions [48,49], the following control variables are selected: the level of economic development ($ECO$), represented by the ratio of urban GDP to the total population; the level of urbanization ($URB$), represented by the ratio of the urban population to the total population; financial leverage ($FIN$), represented by the ratio of urban financial institution loans to GDP; level of openness to foreign investment ($FDI$), represented by the ratio of actual foreign investment balance to GDP; and environmental regulations ($ER$), represented by the comprehensive utilization rate of industrial waste in cities. The specific meanings, processing, and descriptive statistics of each variable are shown in

Table 1. Descriptive statistics of key variables.

Variable	Description	Sample Size	Standard Deviation	Minimum Value	Median Value	Maximum Value
$CI$	Logarithm of carbon emissions as a percentage of GDP	4050	0.798	−12.956	−10.290	−4.785
$CE$	Logarithm of total carbon emissions	4050	1.123	2.521	6.072	10.060
$D$	Dummy variable for pilot cities post-policy implementation (1 if yes, 0 if no)	4050	0.252	0.000	0.068	1.000
$ECO$	Logarithm of per capita GDP	4050	0.999	13.461	16.362	19.774
$URB$	Proportion of urban population to total city population	4050	0.234	0.075	0.401	1.926
$FIN$	Ratio of financial institution loans to GDP	4050	0.576	0.075	0.917	9.622
$FDI$	Proportion of actual foreign capital used to GDP	4050	0.019	0.001	0.018	0.212
$ER$	Comprehensive utilization rate of industrial waste	4050	0.224	0.002	0.794	1.000

.

3.3. Sample Selection and Data Sources

The sample for this study encompasses data from 270 Chinese cities between 2006 and 2020. The urban carbon emissions data is calculated as the sum of fossil fuel carbon emissions and industrial production process carbon emissions based on Formulas (7) and (8). The consumption of different types of fossil energy comes from the CEADs database, while the output of industrial products is sourced from the “China Urban Statistical Yearbook”. Data on the net calorific value of different types of fossil energy, carbon emission factors, oxidation rates, and the carbon emission factors of different industrial products in the production process are derived from the “2006 IPCC Guidelines for National Greenhouse Gas Inventory” and the “2007 China Energy Development Report”. Additional variables are derived from the “China City Statistical Yearbook” (the “Chinese Urban Statistical Yearbook” can be accessed at the following web address: https://www.stats.gov.cn/zs/tjwh/tjkw/tjzl/202302/t20230220_1913734.html, accessed on 10 June 2024) and the “China Urban Construction Statistical Yearbook” (the “Statistical Yearbook of Urban Construction in China” can be accessed at the following web address: https://www.mohurd.gov.cn/gongkai/fdzdgknr/sjfb/tjxx/jstjnj/index.html, accessed on 10 June 2024). The list of pilot cities for circular economy initiatives is obtained from policy documents issued by the National Development and Reform Commission. Furthermore, all price-related variables have been deflated to the base year of 2006. The descriptive statistical analysis of the primary variables is illustrated in Table 1.

4. Empirical Results and Robustness Testing

4.1. Baseline Regression

Table 2. Baseline regression results.

Variable	$\mathit{C}\mathit{I}$		$\mathit{C}\mathit{E}$
$D_{\mathit{it}}$	−0.058 **	−0.062 **	−0.061 **	−0.060 **
	(0.039)	(0.020)	(0.025)	(0.026)
$ECO$		−0.601 ***		0.398 ***
		(0.000)		(0.000)
$URB$		0.165 **		0.169 **
		(0.034)		(0.029)
$FIN$		0.061 ***		0.061 ***
		(0.001)		(0.001)
$FDI$		−1.016 **		−1.025 **
		(0.016)		(0.015)
$ER$		0.096 ***		0.096 ***
		(0.009)		(0.009)
Constant	−9.804 ***	−0.672	5.648 ***	−0.660
	(0.000)	(0.213)	(0.000)	(0.220)
Control	N	Y	N	Y
Year FE	Y	Y	Y	Y
City FE	Y	Y	Y	Y
Observations	4050	4050	4050	4050
R-squared	0.391	0.450	0.340	0.365

Note: (1) The value within the parentheses represents the p-value; (2) *** and ** indicate statistical significance at the 0.01 and 0.05 levels, respectively.

illustrates the influence of the circular economy pilot policy on both the intensity of carbon emissions and overall carbon discharges. The findings indicate that irrespective of the inclusion of control variables, the circular economy pilot policy significantly reduces carbon emission intensity and total carbon emissions at the 0.04 level. Furthermore, the impact of the policy on the aggregate carbon emissions of pilot cities is more pronounced, while the suppressive effect on carbon emission intensity is somewhat less. In essence, the establishment of demonstrative cities for the circular economy has achieved the initial objective of energy conservation and carbon reduction. Firstly, by promoting the circular economy model, resource recycling and reuse reduce energy waste, drive technological innovation, and enhance energy efficiency. For instance, through the recycling and reuse of waste batteries, valuable materials can be effectively extracted, alleviating excessive mining and energy consumption of mineral resources. Secondly, the reuse of waste and the efficient use of energy significantly reduce the demand for and reliance on traditional energy sources such as coal and oil, while promoting the utilization of renewable energy sources like wind, solar, and hydroelectric power. This shift facilitates a low-carbon transformation of the energy structure, leading to a reduction in carbon intensity and emissions in Chinese cities. Although the overall sample suggests that the circular economy pilot policy is conducive to carbon reduction, it does not imply that such policies can effectively promote energy conservation and carbon reduction in all cities. Further analysis of the theoretical mechanisms in conjunction with the preceding text reveals that the circular economy pilot policy primarily impacts carbon emissions through government technology investment, the amelioration of factor allocation structure, and the intensity of public management. The effectiveness of this impact is closely related to the innovative foundation, resource endowment, and industrial structure of pilot cities. In other words, the better the innovative foundation of a city, the more effective government technology investment will be in the research and development innovation of resource recycling technologies. If a city’s resource endowment is abundant and the proportion of high-energy-consuming industries in the industrial structure is high, low-cost resource inputs may lead to lower resource allocation efficiency. In such cases, the increase in factor allocation efficiency will be more significant.

4.2. Parallel Trend Test

The premise for applying the DID model is the fulfillment of the parallel trends assumption, wherein the cities in the experimental and control groups exhibit similar trends in carbon emissions and carbon intensity when not affected by the policy intervention. Conversely, if there are pre-existing differences between the treatment and control groups, the estimated results of the DID model may not reflect the net effect of the policy, as there could be other factors influencing the variations in urban carbon emissions and carbon intensity. The limitations of the parallel trend test method lie in its emphasis on the absence of significant differences in the trends over time between the treatment group and the control group before policy implementation. In contrast, in the application of the DID model, it is more crucial to examine the absence of significant differences in the trends over time between the treatment group and the control group after policy implementation. Therefore, it is essential to carefully select potential influencing factors as control variables. To examine whether the target variables of the treatment and control groups satisfy the parallel trends assumption before the construction of circular economy demonstration cities, and to present the changing trend of policy effects over the years, this study employs event study methodology. Utilizing the parallel trend test method of a multi-period Difference-in-Differences (DID), we establish Equation (9) to test for parallel trends and dynamic effects:

$$Y_{\mathit{it}}={\alpha }_0+{\displaystyle \sum _{k=-5}^5{\mu }_k{D}_{t_p+k}}+{\alpha }_2{X}_t+{\varphi }_i+{\eta }_t+{\epsilon }_{it}$$

(9)

where $t_p$ denotes the years of circular economy demonstration city construction; and $D_{t_p+k}$ represents the interaction terms of the years before, during, and after the construction of circular economy demonstration cities with a dummy variable. Owing to the sample period spanning from 2006 to 2020, with seven years after the first batch of construction and nine years prior to the second batch, limitations in graphical representation necessitate the consolidation of interaction terms. The 5th to 9th years before construction were therefore consolidated into the 4th year before the pilot, and similarly, the 7th year after construction was consolidated into the 6th year post-pilot. To avoid multicollinearity, we set the 2nd year before construction as the baseline. That is, when $t-t_p=k$, the range of $k$ is [−4, 6], excluding $k$ = −2, then $D_{t_t+k}$ = 1; otherwise, $D_{t_t+k}$ = 0. Variable $k$ = −1 denotes the first year before the implementation of the energy rights trading system, $k$ = 0 marks the year of implementation, $k$ = 1 signifies the second year after implementation, and subsequent values of $k$ follow in this manner. The short dashed line in Figure 1 represents the confidence interval calculated at a 0.95 significance level by robust standard errors, the horizontal axis indicates the time interval between different years and the year of circular economy demonstration city construction, and the vertical axis reflects the carbon emission reduction effect of the pilot policy for circular economy. The results suggest that, before the construction of demonstration cities, the pilot policy for the circular economy had no significant estimated coefficients for carbon emission intensity and total carbon emissions, thereby meeting the parallel trends assumption. The absolute value of the estimated coefficient for carbon emission intensity from the first to the fifth year of demonstration city construction reached the 5% significance level, indicating that the pilot policy for the circular economy has a noticeable suppressive effect on carbon emission intensity and total carbon emissions, with a certain degree of lag. Moreover, the regression coefficients for carbon emission intensity and total carbon emissions due to the pilot policy were insignificant in the sixth year post-policy implementation, affecting the sustainability of the carbon reduction effect of the circular economy pilot policy.

4.3. Robust Test

4.3.1. Placebo Test

In the DID model, to eliminate the influence of non-policy factors on the research results, errors in the “policy effect” may arise. Specifically, by observing through the plotting of kernel density graphs, generally, the more concentrated the points are near the zero point on the horizontal axis, the more reliable the “policy effect” of the DID model is deemed, indicating the passage of a placebo test. To mitigate the contamination of our baseline regression conclusions by unobserved external variables, this study meticulously selected 101 cities to serve as nominal pilot sites, thereby constructing a novel policy dummy variable for regression analysis. This process was iteratively conducted 1000 times, with each iteration drawing a random selection of cities and executing regression analysis, culminating in a graphical representation of the distribution for the mean values of the 1000 policy dummy variable coefficients. Figure 2 demonstrates that the mean regression coefficients, with carbon emission intensity and total carbon emissions as dependent variables, approximate zero, and the majority of the coefficients’ p-values exceed 0.1. This indicates that the unaccounted external factors exert negligible interference with the baseline regression outcomes, thereby substantiating Hypothesis 1.

4.3.2. Correction of Sample Selection Bias

Despite the National Development and Reform Commission’s release of pilot cities for the circular economy, the selection of these cities may be influenced by economic conditions, endowment of resources, and industrial foundations, along with other socioeconomic factors. Hence, to correct estimation biases caused by differences in socioeconomic factors between treatment and control groups, this study employs the PSM-DID model for regression analysis once more. Specifically, a 1:1 nearest neighbor matching method is utilized to identify samples from the control group that most closely resemble those in the treatment group, followed by a reassessment using the PSM-DID model to investigate the carbon emission reduction effects of the circular economy pilot policy. Following one-to-one with replacement nearest neighbor PSM matching, the propensity score value probability density distributions of the urban samples in the experimental and control groups are closer after matching than before, indicating a favorable matching effect in this study. This suggests that the examination of the impact of the circular economy pilot policy on urban carbon emissions using the PSM-DID model is reliable. The results in

Table 3. Robustness test results.

Variables	PSM-DID		Introducing Interaction Effects		Excluding Interference from Related Policies
	$\mathit{C}\mathit{I}$	$\mathit{C}\mathit{E}$	$\mathit{C}\mathit{I}$	$\mathit{C}\mathit{E}$	$\mathit{C}\mathit{I}$	$\mathit{C}\mathit{E}$
$D_{it}$	−0.055 **	−0.053 **	−0.058 **	−0.055 **	−0.092 **	−0.090 **
	(0.032)	(0.040)	(0.036)	(0.045)	(0.028)	(0.032)
$ECO$	−0.540 ***	0.459 ***	−0.284 ***	0.714 ***	−0.498 ***	0.503 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
$URB$	0.093	0.096	0.042	0.046	0.172 *	0.176 *
	(0.226)	(0.209)	(0.607)	(0.577)	(0.098)	(0.089)
$FIN$	0.114 ***	0.115 ***	0.023	0.024	0.091 ***	0.092 ***
	(0.000)	(0.000)	(0.275)	(0.254)	(0.000)	(0.000)
$FDI$	−1.096 ***	−1.089 ***	−0.001	0.008	−0.537	−0.559
	(0.008)	(0.009)	(0.999)	(0.989)	(0.399)	(0.379)
$ER$	0.082 **	0.082 **	0.137 ***	0.137 ***	0.084 *	0.084 *
	(0.026)	(0.026)	(0.001)	(0.001)	(0.092)	(0.092)
Constant	−1.601 ***	−1.587 ***	−5.026 ***	−5.042 ***	−2.410 ***	−2.419 ***
	(0.005)	(0.005)	(0.000)	(0.000)	(0.000)	(0.000)
Control	Y	Y	Y	Y	Y	Y
Year FE	Y	Y	Y	Y	Y	Y
City FE	Y	Y	Y	Y	Y	Y
Province × Year FE			Y	Y
Observations	3828	3828	4050	4050	2310	2310
R-squared	0.471	0.373	0.525	0.451	0.349	0.364

Note: (1) The value within the parentheses represents the p-value; (2) ***, **, and * indicate statistical significance at the 0.01, 0.05, and 0.1 levels, respectively.

indicate that, after correcting for sample selection bias, the circular economy pilot policy continues to suppress the increase in carbon emission intensity and volume, thus confirming Hypothesis 1.

4.3.3. Inclusion of Interaction Fixed Effects

The formulation of policies in our country often takes provinces and time as the units. For instance, certain provinces have implemented environmental policies directly or indirectly related to carbon reduction in certain years, such as green financial reform and innovation pilot policies, green fiscal expenditures, and other provincial-level environmental policies. This may lead to the regression results in this study being influenced by other policies rather than the circular economy pilot policy. Therefore, this study incorporates provincial interactive fixed effects in the regression to better eliminate the interference of provincial-level environmental policy effects and enhance the robustness of the estimation results. This method has potential limitations, such as the possibility of multicollinearity or overfitting, and it is necessary to clarify the environmental policies at the provincial level that may affect carbon reduction. Table 3 shows that after including provincial interaction fixed effects, the circular economy pilot policy maintains a significant negative impact on carbon emission intensity and volume, thereby positively reaffirming Hypothesis 1.

4.3.4. Eliminate Interference from Other Policies

To further eliminate interference from other policies at the city level, a review of relevant literature reveals that urban-level policies such as the low-carbon city pilot policy [50], the carbon trading pilot policy [51], and the energy rights trading pilot policy [52] can significantly promote urban carbon reduction. Combining the robustness analysis with the inclusion of provincial interactive fixed effects, the estimation results remain significant after eliminating interference from provincial-level environmental policies. The sample excludes pilot cities involved in low-carbon city pilot policy, carbon trading pilot policy, and energy rights trading pilot policy. The results in Table 3 indicate that the circular economy pilot policy still exhibits a significant inhibitory effect on both carbon intensity and carbon emissions. This positively validates Hypothesis 1.

After placebo tests, correcting sample selection biases, incorporating interactive fixed effects, and eliminating other policy interferences through a series of robustness checks, the circular economy pilot policy still exhibits a clear inhibitory effect on urban carbon emissions and carbon intensity. In other words, the construction of the circular economy demonstration cities can promote carbon reduction by efficiently utilizing energy and breaking free from traditional energy dependencies.

5. Mechanism Testing and Heterogeneity Analysis

5.1. Impact Mechanism Testing

According to Hypotheses 2 and 3, the investigation of the influencing mechanisms includes the following: firstly, the intermediate effects of government technology investment, talent inflow, and per capita fixed capital; secondly, the moderating effects of public management intensity. For the intermediate variables and moderating variables, government technology investment ($TE$) is selected, expressed as the proportion of fiscal technology expenditure to GDP; per capita fixed capital ($FS$) is represented by the logarithm of the total fixed assets proportion to urban employment; public management intensity ($GME$) is represented by the logarithm of the number of public management and social organization employees as a proportion of the total population.

5.1.1. Mechanism Effects of Government Technology Investment

Table 4. Mediation mechanism regression results.

Variables	Government’s S and T Funding			Per Capita Fixed Capital
	$\mathit{T}\mathit{E}$	$\mathit{C}\mathit{I}$	$\mathit{C}\mathit{E}$	$\mathit{F}\mathit{S}$	$\mathit{C}\mathit{I}$	$\mathit{C}\mathit{E}$
$D_{\mathit{it}}$	0.007 ***	−0.059 **	−0.057 **	−0.144 ***	−0.056 **	−0.054 **
	(0.001)	(0.027)	(0.034)	(0.000)	(0.036)	(0.044)
$TE$		−0.393 *	−0.389 *
		(0.053)	(0.055)
$FS$					0.041 ***	0.041 ***
					(0.003)	(0.003)
Constant	−0.126 ***	−0.721	−0.709	6.522 ***	−0.939 *	−0.927 *
	(0.004)	(0.181)	(0.188)	(0.000)	(0.085)	(0.089)
Control	Y	Y	Y	Y	Y	Y
Year FE	Y	Y	Y	Y	Y	Y
City FE	Y	Y	Y	Y	Y	Y
Observations	4050	4050	4050	4050	4050	4050
R-squared	0.208	0.451	0.365	0.104	0.451	0.366

Note: (1) The value within the parentheses represents the p-value; (2) ***, **, and * indicate statistical significance at the 0.01, 0.05, and 0.1 levels, respectively.

shows that the coefficients of variable $D_{it}$ as the dependent variable $TE$ are all positive and significant at the 0.001 level. Regardless of the intermediate effects of government technology investment or talent inflow, the coefficients of variables $TE$ as the dependent variables $CI$ and $CE$ are all significantly negative at the 0.06 level. The absolute values and significances of the corresponding $D_{\mathit{it}}$ coefficients are lower than the estimated results of the benchmark regression in Table 2, without the inclusion of this intermediate variable. This indicates that after controlling for the effect of government technology investment, the impact of the circular economy pilot policy on carbon intensity and carbon emissions is significantly weakened. This suggests that the circular economy pilot policy will promote carbon emission reduction by strengthening government technology investment.

5.1.2. Mechanism Effects of Per Capita Fixed Capital

Table 4 shows that the coefficient of variable $D_{\mathit{it}}$ as the dependent variable $FS$ is significantly negative at the 0.001 level, and the coefficients of variables $FS$ as the dependent variables $CI$ and $CE$ are all significantly positive. The values and significances of the corresponding $D_{\mathit{it}}$ coefficients are lower than the estimated results of the benchmark regression in Table 2, without the inclusion of this intermediate variable. This indicates that after controlling for the effect of per capita fixed capital, the impact intensity of the circular economy pilot policy on carbon intensity and carbon emissions is weakened. This suggests that the circular economy pilot policy will achieve carbon emission reduction by suppressing the stock of per capita fixed capital. Hypothesis 2 has been positively verified.

5.1.3. The Regulatory Effect of Public Management

The results illustrated in

Table 5. Results of the regulatory effect test.

Variables	$\mathit{C}\mathit{I}$		$\mathit{C}\mathit{E}$
	(1)	(2)	(3)	(4)
$D_{\mathit{it}}$	−0.065 **	−0.968 ***	−0.063 **	−0.984 ***
	(0.015)	(0.002)	(0.019)	(0.001)
$GME$	0.097 **	0.111 ***	0.098 **	0.112 ***
	(0.023)	(0.010)	(0.022)	(0.009)
$D\mathrm{it}\times GME$		−0.205 ***		−0.209 ***
		(0.003)		(0.002)
Constant	−0.218	0.054	−0.202	0.075
	(0.705)	(0.926)	(0.724)	(0.897)
Control	Y	Y	Y	Y
Year fixed	Y	Y	Y	Y
City FE	Y	Y	Y	Y
Observations	4050	4050	4050	4050
R-squared	0.451	0.452	0.366	0.367

Note: (1) The value within the parentheses represents the p-value; (2) *** and ** indicate statistical significance at the 0.01 and 0.05 levels, respectively.

authenticate the regulatory effect of public governance. The estimated coefficients of variable $D_{\mathit{it}}$ for columns (1) to (4) all demonstrate significant negativity at a 0.02 significance level, elucidating that upon incorporating variable $GME$ as a moderating factor, the circular economy pilot policy continues to exert a conspicuous inhibitory influence on both carbon emission intensity and volume. The coefficient of interaction term $D_{it}\times GME$, with $CI$ and $CE$ as dependent variables, remains significantly negative at a 0.003 level. Furthermore, the magnitude of public governance’s impact on curbing the effects of $CI$ and $CE$ is quantified as 0.205 and 0.209, respectively. This signifies that public governance possesses a positive regulatory role in the carbon reduction effect of the circular economy pilot policy, with a stronger emphasis on intensifying the mitigation of carbon emission intensity rather than carbon volume. Hypothesis 3 is positively validated.

5.2. Heterogeneity Analysis

The preceding text examined the overall impact of the circular economy pilot policy on carbon emission reduction. Based on the conceptual framework of establishing pilot circular economy cities and counties, it becomes necessary for regions to implement circular economy development strategies tailored to their unique characteristics, such as resource endowments, industrial structures, and economic foundations. The carbon reduction impact of the circular economy pilot policy is intricately linked to urban resource endowment, industrial structure, and the foundation of innovation. In other words, the policy is not universally effective for carbon reduction in all cities. Consequently, this study delves into a nuanced analysis of the carbon reduction effects of the circular economy pilot policy, taking into account the variations in resource endowment, degree of industrialization, and economic bases among cities.

5.2.1. Diversity in Resource Endowment

The abundance of resource endowment influences whether a model circular economy city opts for cost-effective methods to exploit and utilize energy resources, thereby consistently advancing a low-carbon transformation. China boasts coal reserves exceeding 4.5 trillion tons, dispersed throughout the country, yet their distribution is markedly uneven. To investigate the differential impact of the circular economy pilot policy on carbon reduction across various resource endowment backgrounds, this study adheres to the “National Sustainable Development Plan for Resource-Based Cities (2013–2020)”, categorizing resource-based cities by their level of resource endowment into cities with high resource endowment (growth-type resource cities or G-Resource), medium resource endowment (mature-type resource cities or M-Resource), lower resource endowment (regenerating resource cities or R-Resource), and depleted resource endowment (declining resource cities or D-Resource), thereby conducting a heterogeneity test on resource endowment. The outcomes, as indicated in

Table 6. Heterogeneity of city characteristics.

Variables	G-Resource		M-Resource		R-Resource		D-Resource
	$\mathit{C}\mathit{I}$	$\mathit{C}\mathit{E}$	$\mathit{C}\mathit{I}$	$\mathit{C}\mathit{E}$	$\mathit{C}\mathit{I}$	$\mathit{C}\mathit{E}$	$\mathit{C}\mathit{I}$	$\mathit{C}\mathit{E}$
$D_{\mathit{it}}$	−0.098 **	−0.098 **	−1.538 ***	−1.531 ***	−0.116 *	−0.112 *	0.021	0.022
	(0.035)	(0.035)	(0.000)	(0.000)	(0.060)	(0.069)	(0.804)	(0.797)
Constant	0.552	0.383	−8.264 **	−8.188 **	−1.578	−1.545	−6.574 ***	−6.471 ***
	(0.671)	(0.768)	(0.010)	(0.010)	(0.190)	(0.200)	(0.002)	(0.002)
Control	Y	Y	Y	Y	Y	Y	Y	Y
Year FE	Y	Y	Y	Y	Y	Y	Y	Y
City FE	Y	Y	Y	Y	Y	Y	Y	Y
Observations	345	345	195	195	885	885	225	225
R-squared	0.797	0.289	0.300	0.521	0.424	0.313	0.586	0.437

Note: (1) The value within the parentheses represents the p-value; (2) ***, **, and * indicate statistical significance at the 0.01, 0.05, and 0.1 levels, respectively.

, exhibit that the estimated coefficients of variable $D_{\mathit{it}}$ in the group of cities with higher resource endowment are significantly negative, whereas the estimated coefficients of variable $D_{\mathit{it}}$ in the group of cities with resource depletion are not significant. Moreover, as the resource endowment increases, the significance of the estimated coefficients of variable $D_{\mathit{it}}$ strengthens. This signifies that the policy implementation of circular economy pilot projects can notably promote carbon emissions reduction in regions with a higher degree of resource abundance. Consequently, this implies that the circular economy pilot policy can be initially implemented in cities with a higher degree of resource abundance.

5.2.2. Heterogeneity of Industrialization Levels

Industrialization, as a major consumer of China’s energy resources, is influenced by economies of scale. Due to the lower operational costs of pollution reduction equipment, it exhibits a heightened marginal abatement effect [53]. The potential for energy conservation and carbon reduction is substantial, and the restraining effect of the circular economy pilot policy on carbon emissions is more prominent. To unveil the impact of industrialization level on the effectiveness of the circular economy pilot policy, this study, following the “National Plan for the Adjustment and Transformation of Old Industrial Bases (2013–2022)”, differentiates the samples into a group with a high degree of industrialization (cities in old industrial bases) and a group with a low degree of industrialization (cities not in old industrial bases), and subsequently conducts group regression.

Table 7. Heterogeneity test for industrialization level.

Variables	High Industrialization Level		Low Industrialization Level
	$\mathit{C}\mathit{I}$	$\mathit{C}\mathit{E}$	$\mathit{C}\mathit{I}$	$\mathit{C}\mathit{E}$
$D_{\mathit{it}}$	−0.133 ***	−0.130 ***	0.008	0.010
	(0.002)	(0.002)	(0.808)	(0.759)
Constant	0.737	0.698	−0.588	−0.563
	(0.439)	(0.463)	(0.426)	(−0.446)
Control	Y	Y	Y	Y
Year FE	Y	Y	Y	Y
City FE	Y	Y	Y	Y
Observations	1410	1410	2640	2640
R-squared	0.537	0.211	0.418	0.455

Note: (1) The value within the parentheses represents the p-value; (2) *** indicates statistical significance at the 0.01 level.

reveals that the estimated coefficients $D_{\mathit{it}}$ for the high industrialization group are all significantly negative, while the estimated coefficients $D_{\mathit{it}}$ for the low industrialization group are inconspicuous. This indicates that, in comparison to the cost and benefits of energy conservation and emission reduction pathways, it is advisable to prioritize the construction of circular economy demonstration cities in old industrial bases, thereby achieving carbon emission reduction at lower costs.

5.2.3. The Heterogeneity of Economic Foundations

The more robust a city’s economic foundation, the stronger its capacity for technological innovation on the supply side, and the more effective the public’s awareness and supervision of corporate environmental responsibility on the demand side. This study classifies the sample areas based on their economic foundations into Eastern, Central, Western, and Northeastern regions, and then conducts subgroup regression analysis. As shown in

Table 8. Test of economic foundation heterogeneity.

Variables	Eastern Region		Central Region		Western Region		Northeastern Region
	$\mathit{C}\mathit{I}$	$\mathit{C}\mathit{E}$	$\mathit{C}\mathit{I}$	$\mathit{C}\mathit{E}$	$\mathit{C}\mathit{I}$	$\mathit{C}\mathit{E}$	$\mathit{C}\mathit{I}$	$\mathit{C}\mathit{E}$
$D_{\mathit{it}}$	−0.049	−0.048	−0.098 **	−0.100 **	−0.056	−0.049	−0.099	−0.096
	(0.212)	(0.222)	(0.014)	(0.013)	(0.391)	(0.453)	(0.263)	(0.277)
Constant	1.094	1.079	−1.261	−1.225	−4.133 ***	−4.150 ***	1.533	1.606
	(0.365)	(0.370)	(0.277)	(0.290)	(0.003)	(0.003)	(0.294)	(0.271)
Control	Y	Y	Y	Y	Y	Y	Y	Y
Year FE	Y	Y	Y	Y	Y	Y	Y	Y
City FE	Y	Y	Y	Y	Y	Y	Y	Y
Observations	1290	1290	1170	1170	1095	1095	495	495
R-squared	0.445	0.512	0.660	0.431	0.337	0.304	0.484	0.248

Note: (1) The value within the parentheses represents the p-value; (2) *** and ** indicate statistical significance at the 0.01 and 0.05 levels, respectively.

, the circular economy pilot policy in the Central region exhibits a significant inhibitory effect on both carbon emission intensity and total carbon emissions, whereas the estimated coefficients for the other regions are not significant. This may be because the Eastern region has a solid foundation for technological innovation and higher resource utilization efficiency, leaving limited room for carbon emission reduction. Conversely, the Western and Northeastern regions have weaker foundations in technological innovation, which hampers the development and application of technologies for resource and waste recycling. Therefore, for the Central region, which has a certain foundation for innovation and potential for carbon emission reduction, the comprehensive implementation of the circular economy pilot policy based on the principles of “reduction, reuse, and resource recovery” can facilitate the establishment of circular production, distribution, and consumption systems, thereby promoting the cyclic use of urban resources and a reduction in carbon emissions.

6. Examination of the Spillover Effects of the Circular Economy Pilot Policy’s Spatial Dimensions

6.1. Spatial Autocorrelation Analysis

Before implementing the SDID model, it is imperative to authenticate the spatial correlation of carbon emissions. This involves utilizing Moran’s I index for the scrutiny of both the intensity and volume of carbon emissions, alongside their spatial correlation and spillover effects. The study has thus examined Moran’s I for both $CI$ and $CE$. Moran’s I reflect the spatial correlation between urban carbon emissions and carbon emission intensity, with a value distribution range of [−1, 1]. When the value is in the range of [−1, 0), it indicates a negative correlation between cities, while if the value is in the range of (0, 1], it indicates a positive correlation between cities. When the value is 0, it signifies no correlation between cities. The specific formula is as follows:

$$I={\displaystyle \frac{{\displaystyle \sum _{i=1}^k{\displaystyle \sum _{j=1}^k{w}_{ij}\left(x_i-\overline{x}\right)\left(x_j-\overline{x}\right)}}}{{\displaystyle \sum _{i=1}^k{\displaystyle \sum _{j=1}^k{w}_{ij}{\displaystyle \sum _{i=1}^k{\left(x_i-\overline{x}\right)}^2}}}}}$$

(10)

As indicated in

Table 9. Spatial autocorrelation indices.

Variable	$\mathit{C}\mathit{I}$			$\mathit{C}\mathit{E}$
	Moran’s I	z	p-Value	Moran’s I	z	p-Value
2006	0.168	4.127	0.000	0.209	5.125	0.000
2007	0.159	3.922	0.000	0.217	5.310	0.000
2008	0.155	3.827	0.000	0.213	5.206	0.000
2009	0.189	4.655	0.000	0.215	5.262	0.000
2010	0.234	5.716	0.000	0.241	5.890	0.000
2011	0.201	4.929	0.000	0.222	5.434	0.000
2012	0.148	3.645	0.000	0.235	5.753	0.000
2013	0.156	3.843	0.000	0.256	6.258	0.000
2014	0.203	4.993	0.000	0.244	5.963	0.000
2015	0.282	6.874	0.000	0.278	6.772	0.000
2016	0.310	7.565	0.000	0.271	6.605	0.000
2017	0.319	7.807	0.000	0.265	6.477	0.000
2018	0.327	8.023	0.000	0.257	6.285	0.000
2019	0.335	8.253	0.000	0.232	5.684	0.000
2020	0.328	8.084	0.000	0.229	5.612	0.000

, the values of Moran’s I are uniformly positive and predominantly significant at the 0.001 level, signifying a substantial positive spatial correlation in carbon emissions. Moreover, between the years 2006 and 2020, Moran’s I for both $CI$ and $CE$ in the regions has shown a fluctuating upward trend, suggesting that the spatial correlation for $CI$ and $CE$ in these areas is generally increasing. Additionally, the impact of $CI$ and $CE$ in pilot cities on their adjacent areas has been on an ascending trajectory.

6.2. The Examination Results of the SDID Model

The examination results of the Synthetic Difference in Differences (SDID) model are delineated in

Table 10. SDID model tests.

Variable	SLM-DID		SEM-DID
	$\mathit{C}\mathit{I}$	$\mathit{C}\mathit{E}$	$\mathit{C}\mathit{I}$	$\mathit{C}\mathit{E}$
LM-test	3.743 **	26.402 ***	23.921 ***	5.307 **
LR-test	90.35 ***	106.66 ***	99.39 ***	100.03 ***
Wald-test	91.21 ***	108.08 ***	100.11 ***	100.76 ***

Note: *** and ** indicate statistical significance at the 0.01 and 0.05 levels, respectively.

, which presents the regression outcomes. Notably, the Lagrange Multiplier (LM) test achieved a significance level of 0.05, suggesting the selection of an appropriate SDID model is substantiated. Further analysis through Likelihood Ratio (LR) and Wald tests indicates the superior performance of the Spatial Durbin Difference in Differences (SDM-DID) model over the Spatial Lag Model Difference in Differences (SLM-DID) and Spatial Error Model Difference in Differences (SEM-DID) models. Moreover, following the Hausman test, this paper incorporates fixed effects into the SDM-DID model.

Table 11. SDID model regression results.

Variable	Spatial Adjacency		Geographic Distance		Economic Distance
	$\mathit{C}\mathit{I}$	$\mathit{C}\mathit{E}$	$\mathit{C}\mathit{I}$	$\mathit{C}\mathit{E}$	$\mathit{C}\mathit{I}$	$\mathit{C}\mathit{E}$
$D_{\mathit{it}}$	−0.050 **	−0.047 *	−0.062 **	−0.059 **	−0.053 **	−0.050 **
	(0.046)	(0.060)	(0.015)	(0.020)	(0.036)	(0.048)
$W\times D_{\mathit{it}}$	−0.382 ***	−0.382 ***	−0.340 **	−0.339 **	−0.348 ***	−0.345 ***
	(0.000)	(0.000)	(0.016)	(0.016)	(0.000)	(0.000)
ρ	0.143 ***	0.143 ***	0.534 ***	0.539 ***	0.271 ***	0.273 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Direct Effect	−0.062 **	−0.059 **	−0.064 **	−0.061 **	−0.058 **	−0.055 **
	(0.018)	(0.023)	(0.016)	(0.020)	(0.028)	(0.035)
Indirect Effect	−0.430 ***	−0.428 ***	−0.819 **	−0.819 **	−0.491 ***	−0.487 ***
	(0.000)	(0.000)	(0.018)	(0.029)	(0.000)	(0.000)
Total Effect	−0.492 ***	−0.487 ***	−0.882 **	−0.880 **	−0.549 ***	−0.542 ***
	(0.000)	(0.000)	(0.012)	(0.020)	(0.000)	(0.000)
Control	Y	Y	Y	Y	Y	Y
Year FE	Y	Y	Y	Y	Y	Y
City FE	Y	Y	Y	Y	Y	Y
Observations	4050	4050	4050	4050	4050	4050
R-squared	0.104	0.520	0.121	0.609	0.128	0.617

Note: (1) The value within the parentheses represents the p-value; (2) ***, **, and * indicate statistical significance at the 0.01, 0.05, and 0.1 levels, respectively.

showcases the SDM-DID model’s conclusions, incorporating spatial adjacency, geographic proximity, and economic distance in the spatial weight matrix, thereby enhancing the robustness of the results. Upon the inclusion of control variables and fixed effects, the autocorrelation coefficient $\rho$ of the dependent variable is significantly positive at the 0.001 level, indicating the existence of spatial correlation in urban carbon emissions. Simultaneously, the coefficients $D_{it}$ for dependent variables $CI$ and $CE$ are significantly negative, with the spatially weighted term $W\times D_{it}$ coefficients significantly negative at both the 0.001 and 0.016 levels. Given that $W\times D_{it}$ regression coefficient does not directly reflect the degree of its influence, it is necessary to employ partial derivatives to decompose it into direct, indirect, and total effects. The direct effects not only encompass the influence of policy implementation on the pilot cities but also include the impact of the pilot cities on the surrounding non-pilot cities, that is, the feedback effect. Meanwhile, the indirect effects pertain to the influence of policy implementation on the surrounding non-pilot cities. The findings reveal that, on the one hand, the estimated coefficients for the direct effects are significantly negative at the 0.035 level, aligning with the conclusions of the non-spatial models and thereby reaffirming the robustness of Hypothesis 1. On the other hand, both the indirect and total effects’ estimated coefficients are significantly negative, indicating that the circular economy pilot policy exerts a negative spatial spillover effect on carbon emissions. The spillover effect is more pronounced in carbon emission intensity than in the total volume of carbon emissions, implying that the circular economy pilot policy significantly reduces both the carbon emission intensity and the total quantity of emissions in the surrounding areas of the pilot cities. This outcome is attributed to the neighboring cities adopting efficient resource utilization and recycling policies through emulation and learning from the pilot cities, fostering knowledge and technology spillovers conducive to energy conservation and emission reduction in the surrounding regions, thus positively validating Hypothesis 4.

7. Discussion

According to the aforementioned research findings, the previous literature primarily discussed the relationship between the circular economy pilot policy and technological innovation, as well as air pollutants. This article further explores the energy-saving and carbon-reducing effects of the circular economy pilot policy, manifested in the reduction in carbon emissions achieved through the recycling of resources and the resource utilization of waste. In other words, the concept and development model of circular economy can provide new insights for low-carbon development. Moreover, environmental policies often use pollutant emission levels as regulatory indicators, tending to prioritize end-of-pipe pollution control methods in production. In contrast, the circular economy pilot policy aims to reduce resource constraints and waste pollution from the source and production process, guiding enterprises to achieve carbon reduction through resource conservation and clean production, rather than primarily relying on end-of-pipe control methods. For resource-based cities and old industrial base cities, it is challenging to change from dependence on traditional energy sources such as oil and coal through end-of-pipe environmental governance. Therefore, the circular economy pilot policy promotes energy-saving and emission reduction technological innovation through government investment in science and technology, reducing reliance on traditional energy sources and effectively unlocking the “carbon lock-in” of resource-based cities and old industrial base cities. It is essential to emphasize that the participation of enterprises in the construction of circular economy demonstration cities relies on strong government guidance and regulation. This not only requires strengthening organizational leadership and ensuring the implementation of tasks but also entails increasing the allocation of green funds by the government to empower green financial capital for energy-saving and emission-reduction technological innovation in circular economy demonstration cities.

8. Conclusions and Policy Implications

Advancing the construction of demonstration cities for the circular economy is a pivotal strategy for China to achieve its carbon peak and neutrality goals on schedule. This study treats circular economy demonstration cities as a quasi-natural experiment, employing a multi-period Difference-in-Differences (DID) model to empirically assess the impact, mechanisms, and heterogeneity of the circular economy pilot policy on carbon emissions. The conclusions of the research are as follows: On the whole, the circular economy pilot policy may contribute to the reduction in carbon emissions and the mitigation of carbon emission intensity. However, it is essential to clarify that the circular economy pilot policy predominantly exerts a notable impetus on carbon reduction in resource-based cities, old industrial base cities, and central regions, rather than being universally effective for carbon reduction in all cities. This is closely intertwined with the operational mechanism of the circular economy pilot policy. Furthermore, the carbon reduction effects of the circular economy pilot policy are primarily realized through channels such as governmental investment in science and technology and the improvement of factor allocation structures, with public management efforts playing a moderating role. Moreover, the circular economy pilot policy not only significantly propels local carbon reduction efforts but also exerts a negative spillover effect on the carbon emissions of surrounding areas, manifesting as a “demonstrative effect”.

The limitations of this study lie in two aspects. Firstly, the performance evaluation of local governments is closely related to the effectiveness of environmental regulations. This study did not integrate this influencing factor into the empirical research on the circular economy pilot policy. In the future, it could further incorporate perspectives such as fiscal decentralization and official promotion to examine the role of government-enterprise relations in the carbon reduction effects of the circular economy pilot policy. Secondly, due to data availability, this study only selected prefecture-level cities as research samples and did not delve into research at the micro level of enterprises. In the future, it could further obtain micro-level data from listed company social responsibility reports and ESG reports to thoroughly investigate the impact of the circular economy pilot policy on carbon reduction at the micro level of enterprises. In light of the aforementioned conclusions, the policy implications are as follows:

• First, it is recommended to gradually expand the content and coverage of the circular economy pilot policy systematically and progressively. Adhering strictly to the three major principles of “reduction, reuse, and recycling”, tailored 3–5 year development plans for circular economy and their phased objectives should be formulated based on the resource utilization efficiency and innovation foundation of circular economy enterprises, with regular monitoring of the progress and completion of the phased objectives of the plan. In the short term, the focus should mainly be on improving the structure of factor allocation, using renewable resources to replace non-renewable resources, and achieving resource reduction. Meanwhile, in the long term, continuous and in-depth promotion of recycling technologies is needed, specifically guiding innovative resource flows towards the innovation and application fields of new energy and renewable energy, strengthening the large-scale use of solar energy, wind energy, biomass energy, geothermal energy, hydrogen energy, and other energy sources to facilitate the green and low-carbon transformation of urban energy systems and curb carbon emissions. Furthermore, the standardized governance model of circular economy in national demonstration cities should be summarized, replicated, and promoted to surrounding areas, fostering innovative cooperation between national circular economy demonstration cities and surrounding areas in the field of energy generation to fully demonstrate the “demonstration effect” of the circular economy pilot policy.
• Second, it is recommended that, in the governance of the circular economy, active participation from entities at all levels be galvanized, with a focus on green and low-carbon technological innovation, adjustment of the factor allocation structure, and the local government’s management of circular economy development. Due to the reliance of the circular economy pilot policy on the support of circular economy industries, the high initial investment level and long recovery period of circular economy industries, as well as the relatively low equipment levels of key technologies such as mining technology, environmental protection technology, energy-saving technology, and comprehensive utilization technology, there is a bottleneck phenomenon. At the local government level, it is necessary to increase economic incentives for circular economy industries, enhance interest subsidies for key technology development and promotion loans from fiscal funds, broaden the financing channels for key technologies through green financial innovation products such as green bonds, encourage enterprises participating in carbon emission reduction accounting to enter the carbon market trading and obtain market incentives for carbon emission reduction benefits, and strengthen the attention of local governments to the construction of circular economy and the support for circular economy innovation. On the legal front, the government responsibility system within the circular economy legal framework should be perfected, clarifying local government responsibilities for circular economy development goals, management scope, focus, and the specific procedures and timelines for responsibility implementation, thus ensuring local government’s governance intensity in managing the circular economy from its root. At the corporate level, the focus should be on reducing excess capacity, deleveraging, and facilitating an orderly exit for ‘zombie’ enterprises, optimizing the structure of corporate factor allocation, and steering social capital towards new energy or enterprises with higher energy efficiency.
• Third, it is advisable for the circular economy pilot policy to be tailored to the unique attributes of each city, identifying low-carbon transformation measures that align with the city’s specific traits. The circular economy pilot policy and regulations tend to be more macroscopic, lacking operational details, while the implementation of the circular economy pilot policy requires effective government supervision and public oversight. Therefore, to further refine the circular economy pilot policy and regulations, local governments can establish a circular economy carbon emission reduction accounting evaluation system, incorporate circular economy carbon emission reduction accounting into the local government assessment index system, and establish mechanisms and platforms for public participation in a circular economy. At the same time, there should be intensified policy guidance for key cities, implementing circular economy development strategies per differences in resource endowment, industrial structure, and economic foundation, such as currently prioritizing carbon reduction in resource-based cities, old industrial cities, and cities within the central region.
• Finally, the research findings indicate a significant “demonstration effect” of the circular economy pilot policy demonstration cities on the transformation of energy demand in surrounding areas. In response to this, this study suggests summarizing the standardized model of circular economy governance in national circular economy demonstration cities for replication and promotion in surrounding areas to enhance their demonstration effect, while also fostering innovative cooperation between national circular economy demonstration cities and surrounding areas in the field of energy generation.