Does Industrial Symbiosis Improve Carbon Emission Efficiency? Evidence from Chinese National Demonstration Eco-Industrial Parks

Abstract

Improving carbon emission efficiency (CEE) is a necessary consideration in response to economic downturn and climate change. In this case, industrial symbiosis paves the way for cities to conserve energy, reduce carbon emissions, and upgrade economic development patterns. This paper verifies the influence of industrial symbiosis policies on CEE, represented by National Demonstration Eco-industrial Parks (NDEPs). The difference-in-differences (DID) and spatial DID (SDID) models, as well as panel data of 282 Chinese cities during 2003–2019, were used to complete this argumentation process. The results show that NDEP policy dramatically increases the CEE of pilot cities. Compared with cities without NDEP, pilot cities climb by 3.49% on CEE, mainly due to industrial structure upgrading and green technological innovation. Moreover, eastern, non-resource-based, and multi-NDEP cities experience a noticeable increase in CEE. NDEP increases not only the CEE of pilot cities but also that of their neighboring cities within 450 km through diffusion and demonstration effects. All these findings help promote China’s NDEP construction and offer decision-making guidance for climate governance and low-carbon transition in China and others following a similar pathway.

1. Introduction

Rapid industrialization leading to overexploitation and unsustainable resource use makes it much harder to reach Sustainable Development Goal 7. There has been a global focus on how to ensure energy security, affordability, and sustainability [1]. Worse still, while addressing the energy trilemma, greenhouse-gas-induced global warming has an adverse effect on human well-being. As required by the Paris Agreement, countries should achieve net-zero carbon emissions in depth and quickly. Industrial production accounts for about 30% of global carbon emissions [2], and promoting industrial cleaner production deserves high attention. China releases more carbon dioxide than others because of its highly carbonized energy mix. Similar to global industrial carbon emissions, over 30% and 60% of carbon emissions nationwide come from industrial parks and industries in China, respectively [3,4]. Industrial parks, as one of the key drivers of urban economic development, are gathering areas for energy consumption and carbon emissions. Hence, there is a need to promote low carbon at the industrial park level to put the world back on a sustainable path. To that end, China has put in place a number of measures to translate its commitments to carbon peaking and carbon neutrality into tangible progress. Among them, it has gradually become a common practice to introduce the concept of industrial symbiosis in industrial parks to establish eco-industrial parks (EIPs). Compared with traditional industrial agglomeration, industrial symbiosis is a more advanced form of industrial organization. It provides the possibility of collaboration among traditionally independent industries through byproduct exchange, as well as energy and water recycling [5].

EIPs function as carriers of industrial symbiosis, deemed as an efficient path to reach the dual goals of circular development and lower carbon emissions [6]. To leverage the competitive advantages of industrial symbiosis in resource sharing, China launched the National Demonstration Eco-industrial Parks (NDEPs) project following the examples of developed countries such as the United Kingdom, Japan, and the Republic of Korea. In 2001, the Guangxi Guitang Group was approved as the first NDEP. The establishment of NDEPs needs to proceed through the following links: declaration, planning and demonstration, review and naming, supervision, and management. Parks with a certain economic strength and embryonic eco-industry are more likely to successfully apply for NDEP establishment. In fact, most NDEPs are upgraded from provincial or national development zones. As of the end of 2022, China had constructed 95 NDEPs in 24 provinces. Besides the NDEP, the central government also introduced policies from various perspectives to promote the low-carbon transformation of parks, including ecological civilization construction, circular economy, green manufacturing, and so on. Local governments implement these policies creatively in parks, considering local factor endowments and development levels [7]. Additionally, the focus of NDEP policy has also changed over time. Early NDEP construction mainly focused on facilitating economic advancement and reducing environmental pollution. After setting the goal of controlling greenhouse gas emissions in 2009, the Chinese government began to make developing a low-carbon economy a new goal for NDEP construction. In December 2020, working toward carbon peaking and neutrality was listed as one of the eight 2021 national priorities. Subsequently, these two goals have also become the NDEPs’ key evaluation content. NDEP policy is the concrete embodiment of the industrial symbiosis concept in China. Rigid restrictions on high-carbon industries in the NDEP will steer the industrial structure toward low-carbon emissions. At the same time, NDEP inevitably affects the carbon emission reduction activities in neighboring regions through the flow and exchange of regional resources and information. This paper proposes the following questions accordingly: Does the construction of NDEPs improve urban CEE? If so, how exactly does it exert an influence? In addition, does the influence further expand to the neighboring cities? Clarification of the impact of NDEP construction on carbon emissions not only enriches relevant research on NDEP policy effects, but also provides a reference scheme for the green transformation of traditional industrial parks.

Some scholars prefer single-factor indicators to measure the efforts to reduce carbon emissions, such as the carbonization index [8], total carbon dioxide emissions [9], and carbon intensity [10]. Such indicators are easy to measure, yet they ignore other factors related to carbon emissions, such as energy input. Energy input is a major source of carbon emissions (undesirable output) and economic benefits (desirable output). Thus, carbon emission efficiency (CEE), a total-factor index reflecting the development stage of urban low-carbon transition, has gradually vaulted into the mainstream vernacular [11]. Based on relevant studies, this study defines CEE as maximizing economic output with minimum inputs and carbon emissions [12], to help us more comprehensively assess NDEP policy effects under the net-zero target. DEA, as a non-parametric method, was widely applied to regional or industrial CEE measurement. At present, DEA models for CEE measurement mainly include the following: The first is the slacks-based measurement (SBM) model with undesired outputs, which is a non-radial model and may compensate for the traditional radial model’s inability to measure slackness variables and undesired outputs. The second considers the global reference on the basis of the former. Zhang et al. (2022) and Wang et al. (2020) used the SBM model and the global SBM model, respectively, to study the static carbon emission efficiency of the Yangtze River Economic Belt and airlines [13,14]. Both models above have a maximum efficiency value of 1, which means that no comparison can be made between effective decision-making units. To overcome this shortcoming, the introduction of super-efficiency in the SBM model has become a common practice for many scholars to calculate CEE [15,16]. On the whole, most studies used single-factor indicators to measure NDEP’s influence on lowering carbon emissions. By considering both input and output factors, a more accurate result can be produced to manifest how NDEP affects carbon emission reduction in this paper.

More and more Chinese scholars are also paying attention to the achievements of NDEP policy, with different research perspectives. Wu and Gao (2022) and Wu et al. (2023) probed into how NDEP policy affects green innovation and technological advancement through the difference-in-differences (DID) model [17,18]. Nie et al. (2022) analyzed the relationship between NDEP and carbon emission, claiming that green innovation is significant in facilitating low-carbon development of cities [19]. The above studies have only examined the direct impact of NDEP from an economic or environmental perspective, and much of the attention has been focused on a single indicator. Liu et al. (2022) demonstrated that NDEP has a positive effect on improving urban green total factor productivity, but overlooked the potential spatial impact of this policy on surrounding areas [20]. Furthermore, Liu et al. (2022) and Zhang and Wei (2023) dissected how NDEP promotes the green economy directly or indirectly using the spatial DID (SDID) model [7,21]. Chen et al. (2023) and Liu et al. (2023) proved that NDEP reduces pollution and carbon emission and improves CEE of the pilot cities and their neighboring cities [22,23]. A few scholars have noticed the spatial dependence between NDEP and CEE, but the range of NDEP’s spatial spillover effect is yet to be clarified. This paper constructs a spatial weight matrix via different geographical distance thresholds to fill this research gap.

This paper makes the following contributions: (1) Current studies provide important references for understanding industrial symbiosis and CEE, but there is still a need for further improvement. Our empirical study of the impact of industrial symbiosis on CEE can enhance our understanding of the role of industrial symbiosis in promoting carbon emission efficiency. Since NDEP is seen as an important vehicle for industrial symbiosis in practice, this paper provides direct evidence for industrial symbiosis to improve CEE. (2) This paper elucidates the differences in the impact of NDEP on CEE among cities with different resource endowments, geographical locations, and NDEP quantity. These findings can provide a valuable reference for the Chinese government because they can tailor differentiated policies to local conditions to fully leverage the positive externality of industrial symbiosis. (3) By constructing a spatial weight matrix via different geographical distance thresholds, this paper obtains the range of NDEP policy’s influence on the neighboring cities, offering decision support for formulating and implementing industrial symbiosis policies for different cities in the region.

2. Theoretical Hypothesis

Optimizing the production efficiency and energy consumption structure are the two most direct ways to improve CEE. Enterprises in an industrial park share infrastructure and consume energy in a gradient manner, thus forming industrial symbiosis. According to their various needs for resources and energy, the waste heat and excessive heat and pressure generated by upstream enterprises during production are recycled as resources [17]. These resources are transported as production factors to downstream enterprises, which reduces unnecessary production costs and drastically increases the proportion of renewable energy and production efficiency. The Chinese government reviews the officially named NDEPs every three years based on evaluation indicators such as economic growth, resource-saving, and environmental protection. This is beneficial to force companies to facilitate green innovation, and reduce pollutant discharge. NDEP’s carbon-reducing impact is not limited to parks, but also contributes to urban decarbonization through the eco-industrial network formed by linking low-carbon industries.

A spatial correlation exists between NDEP and CEE. There are various economic links among cities, including industrial agglomeration, technology diffusion, and diversified competition in environmental governance. These connections will lead to the inevitable influence of local city policies on the development direction of neighboring cities. A promotion system for local government officials motivates cities to learn from each other on strategies for carbon emission reduction and low-carbon technology innovation. Cities with higher CEE will be imitated by surrounding areas. Meanwhile, cities strengthen their competitive edge by reforming park regulations to encourage innovative elements to cluster locally. NDEP policy drives the interregional flow of production factors like the labor force and technology [24,25], and strengthens spatial correlation of carbon emission between neighboring cities by triggering technological spillover from industrial symbiosis. As the distance threshold widens, spatial correlations turn from agglomeration to dispersion, and the spatial impact of NDEP on CEE will gradually diminish. On such a basis, we propose Hypothesis 1:

Hypothesis 1 (H1).

NDEP improves CEE of local cities and their neighboring cities.

Infrastructure sharing would reduce repetitive investment and wasting of resources, and dramatically increase resource utilization efficiency. The efficiency of resource utilization can also be boosted via gradient utilization of energy and cogeneration [26]. The high demands of NDEP on the environment would set limits to carbon-intensive industries, and preferentially allocate production factors to low-carbon industries. Moreover, guided by the concept of industrial symbiosis, the share of renewable and clean energy has increased, transforming energy consumption patterns and strengthening the green economy effect of the park. As economic exchanges continue to expand and deepen, NDEP is in a strong position to engage in local competition and even take the lead in industrial structure upgrading in the surrounding areas because of its green and recyclable development pattern. To maintain their dominance, local governments are more willing to optimize the spatial distribution of industries. It will then be harder for carbon-intensive industries to enter the market than it is now [27]. In short, from promoting infrastructure sharing within parks to creating low-carbon industries, the entire urban decarbonization process develops along the “park—industry—city” vein. Accordingly, Hypothesis 2 is presented as follows:

Hypothesis 2 (H2).

NDEP construction improves CEE by optimizing industrial structures.

Local environmental regulation offers essential guidance for enterprises to take the road of green innovation. Preferential policies issued by the local government motivate enterprises in NDEP to participate in green innovation and offset the marginal cost of developing low-carbon products or technologies. Such an external incentive internalizes the environmental cost and forces enterprises to upgrade specific technologies to raise their CEE. Re-engineering existing production processes would optimize the energy structure, address the root cause of excessive fossil energy consumption, and improve the productivity of green products [28]. The region where a large number of innovation factors gather is bound to be the high-frequency region where innovation occurs. The innovation effect brought by the concentration of R&D elements in the region would help the park improve the speed of green technology upgrading and promotion, and lower pollution control costs for enterprises [29]. Moreover, the gathering and flow of talents give birth to new knowledge-intensive industries, which will undoubtedly optimize the industrial structure of cities, create a sound environment for knowledge spillover, and ultimately reduce carbon emissions. On such a basis, this paper proposes the last hypothesis.

Hypothesis 3 (H3).

NDEP encourages enterprises to innovate green technologies to improve CEE.

The influence of EIP, a regional industrial symbiosis system, on CEE is shown in Figure 1.

3. Materials and Methods

3.1. Model Construction

In this work, NDEP was used as a quasi-natural experiment to determine how NDEP impacts urban CEE. According to the NDEP lists approved and released by the Ministry of Ecology and Environment of China in batches, this paper constructed a multi-period DID model to assess the policy effect. The specific econometric model settings are the following:

$${ CEE}_{it}={\alpha }_0+{\alpha }_1{NDEP}_{it}+{\alpha }_2{Control}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(1)

where $i$ is different prefecture-level cities, $t$ denotes the year; ${CEE}_{it}$ represents the CEE; ${NDEP}_{it}$ is the dummy variable of the NDEP policy. In this paper, the core coefficient ${\alpha }_1$ is the emphasis and implies the actual changes before and after NDEP on urban CEE. If ${\alpha }_1>0$, then NDEP policy could improve CEE. ${Control}_{it}$ stands for a set of control variables. ${\mu }_i$ and ${\lambda }_t$ represent the city and time-fixed effects, respectively. ${\epsilon }_{it}$ is the random disturbance term.

Given the existence of spatial economic correlations, NDEP construction may have an impact on the CEE of nearby cities. Such a case, however, breaks the stable unit treatment value assumption of the conventional DID method and may result in bias in policy estimation. Therefore, this paper combined the spatial Dubin model (SDM) and DID model with multiple periods into the following SDID [30,31].

$$\begin{array}{c}{ CEE}_{it}={\alpha }_0+\rho {\displaystyle \sum _j}{w}_{ij}{CEE}_{it}+{\alpha }_1{NDEP}_{it}+\theta {\displaystyle \sum _j}{w}_{ij}{NDEP}_{it}+{\alpha }_2{Control}_{it}\\ +\psi {\displaystyle \sum _j}{w}_{ij}{Control}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it} \end{array}$$

(2)

where $w_{ij}$ is the element of the inverse-distance spatial weight matrix; $\sum _j{w}_{ij}{CEE}_{it}$, $\sum _j{w}_{ij}{NDEP}_{it}$, $\sum _j{w}_{ij}{Control}_{it}$ represent the spatial lag terms of urban CEE, NDEP construction, and control variables, respectively, and $\rho ,\theta ,\psi$ are coefficients to be estimated. The meanings of other variables are similar to Equation (1).

3.2. Variable and Data Descriptions

3.2.1. Explained Variable

DEA, as the most efficient and extensive method for efficiency calculation, is still limited in intertemporal incomparability and unable to rank the efficiencies of effective decision-making units. For that, a super-efficiency SBM model with undesirable outputs is constructed for CEE measurement in the case of global reference. Equation (3) shows the detailed model.

$$\begin{gathered} CEE=\mathrm{m}\mathrm{i}\mathrm{n}\frac{1+\frac{1}{m}\sum _{i=1}^m\frac{s_i^{-}}{x_{ik}}}{1-\frac{1}{q+h}\left({\sum }_{r=1}^q\frac{s_r^{d+}}{y_{rk}^d}+{\sum }_{t=1}^h\frac{s_t^{u-}}{y_{tk}^u}\right)} \\ s.t.\left\{\begin{array}{l}{x}_{ik}\ge {\displaystyle \sum _{j=1,j\ne k}^n}{x}_{ij}{\lambda }_j-s_i^{-}\\ y_{rk}^d\le {\displaystyle \sum _{j=1,j\ne k}^n}{y}_{rj}^d{\lambda }_j+s_r^{d+}\\ y_{tk}^u\ge {\displaystyle \sum _{j=1,j\ne k}^n}{y}_{tj}^u{\lambda }_j-s_t^{u-}\\ 1-\frac{1}{q+h}\left({\sum }_{r=1}^q\frac{s_r^{d+}}{y_{rk}^d}+{\sum }_{t=1}^h\frac{s_t^{u-}}{y_{tk}^u}\right)>0\\ s_i^{-}\ge 0,s_r^{d+}\ge 0,s_t^{u-}\ge 0,{\lambda }_j\ge 0\\ i=\mathrm{1,2}\dots ,m;r=\mathrm{1,2}\dots ,q;t=\mathrm{1,2}\dots ,h;j=\mathrm{1,2}\dots ,k\left(j\ne k\right)\end{array}\right. \end{gathered}$$

(3)

where $CEE$ stands for the CEE of each city; it can be valued above 1 in this model; $x,y^d,y^u$ are the input indicator, desirable output indicator, and undesirable output indicator, respectively; $s^{-},s^{d+},s^{u-}$ are the slack variables of input, desirable output, and undesirable output, respectively; and $\mathsf{\lambda }$ is the weight vector.

In this paper, real GDP and carbon dioxide emissions represent desirable and undesirable outputs, respectively, while the capital, labor force, and energy represent inputs.

(1)

Capital stock: based on fixed asset investment, capital stock was calculated via the perpetual inventory method with a depreciation rate set as 10.96%, then converted into the 2003 constant price [32].

(2)

Labor force: the year-end employee number was collected as the input indicator of the labor force.

(3)

Energy: we used the total annual electricity consumption in every city as a substitute for energy input [33].

(4)

Desirable output: urban real GDP was chosen as a proxy for desirable output and converted into the 2003 constant price.

(5)

Undesirable output: the undesirable output is the carbon emissions of prefecture-level cities. With reference to previous studies, this part of the data was obtained by aggregating the carbon emissions of the counties under their authority [34,35].

3.2.2. Core Explanatory Variable

The key independent variable is the interaction term ${NDEP}_{it}=treat\times post$. In detail, both $treat$ and $post$ are dummy variables. The former indicates whether the city participates in NDEP construction, and the latter represents when the pilot policy gets implemented. According to the NDEP list, in this paper, the treatment group consists of 33 prefectural-level cities, which have officially named NDEPs during the study period. If city $i$ has an NDEP as early as year $t$, ${NDEP}_{it}$ takes the value of 1 in year $t$ and later, and 0 otherwise.

3.2.3. Control Variables

(1)

Real GDP per capita (lnpgdp). This indicator reflects the urban economic development stage, expressed in terms of GDP per capita of each city at 2003 prices. In the stage of extensive growth, the consumption of traditional energy like fossil fuel accelerates carbon emission of cities, curbing the improvement in emission efficiency. Nonetheless, as economics advance, people’s demand of a clean environment grows, urging the government to increase input in technological innovation related to cleaner production [36].

(2)

Foreign direct investment (lnfdi). This variable is measured using the logarithm of actual foreign direct investment (FDI) in each city, converting dollars into yuan based on the average exchange rate over the years. It yields a pollution halo or pollution haven influence on CEE. Some hold that FDI promotes carbon emission reduction through low-carbon technology transfer and knowledge spillover [37]. Others believe that it may encourage the local government to usher in pollution-intensive industries to promote economic growth, hence curbing the improvement in CEE [38].

(3)

Governmental intervention (gov). This indicator is represented by local fiscal expenditure to GDP. The government plays an indispensable part in issuing environmental policies, enhancing pollution treatment, and guiding emission reduction plans, which can promote cleaner production and allocate resources in a reasonable manner to a certain extent [39]. However, excessive governmental intervention may hinder market vitality and result in cut-throat competition within industries, thereby wasting resources and lowering CEE.

(4)

Urbanization (urban). This variable is expressed by the proportion of urban population in the total regional population. Scholars hold different opinions on how urbanization affects CEE. One theory is that they are in a U-shaped relationship [40], while another suggests the opposite. The latter theory holds that during the initial expansion of the city, the two change in the same direction. But when the city grows beyond a certain limit, it would hinder the green transformation of cities.

(5)

Science and technology expense (science). This indicator is presented by the proportion of government science and technology expenditure to fiscal expenditure. For one thing, increased expenditure in science and technology greatly lowers carbon emissions, primarily by optimizing energy structure and facilitating the adoption of green technologies [41]. For another, the technological advance enlarges the scale of production, resulting in increased energy consumption and carbon emissions.

3.2.4. Mediating Variables

(1)

Industrial structure (ins). Excessive industrialization results in extensive energy consumption, and hinders the industrial structure from transforming and upgrading and the CEE from increasing. In this paper, the proportion of added value of secondary industry to GDP is employed to represent industrial structure.

(2)

Green technological innovation (inno). The number of green patent applications is used as an indicator to measure the level of urban green technology innovation. We searched the relevant data of green inventions and utility model patents from the China National Intellectual Property Administration (CNIPA), in accordance with the International Patent Classification (IPC) Green Inventory of World Intellectual Property Organization (WIPO) [30]. The number of these two types of patents in each prefecture-level city was summed to obtain the quantity of annual green patents applied for.

3.2.5. Data Sources

This paper took 282 prefecture-level cities from 2003 to 2019 as the study objects, excluding Bijie, Tongren, Laiwu, and other cities involved in administrative division adjustment during the study period, as well as Lhasa with severe data loss. The officially named NDEP list was collected from the official website of the Ministry of Ecology and Environment of China. Carbon emissions and green patent application data are from Carbon Emission Accounts & Datasets (CEADs) and CNIPA, respectively. Other city-level data not mentioned above are from the China City Statistical Yearbook, China Urban-Rural Construction Statistical Yearbook, and China Statistical Yearbook for Regional Economy. Some missing values were calculated according to yearbooks and Statistical Communiques or handled via linear interpolation.

Table 1. Descriptive statistics.

Variable	Obs	Mean	Std. Dev.	Min	Max
CEE	4794	0.3425	0.1478	0.0639	1.6283
lnpgdp	4794	9.9941	0.853	7.545	12.6968
lnfdi	4794	11.1406	2.1953	0	16.5832
gov	4794	0.1669	0.1017	0.0313	2.2794
urban	4794	50.4396	16.4345	16.69	118.84
science	4794	0.0125	0.0143	0	0.2068
ins	4794	47.5497	11.1599	2.66	90.97
inno	4794	0.4493	1.6245	0	34.67

presents the descriptive statistics of variables.

4. Results

4.1. Benchmark Regression

Based on Equation (1),

Table 2. Benchmark regression results.

Variables	(1)	(2)	(3)
NDEP	0.0958 *** (9.7696)	0.0486 *** (3.1884)	0.0349 ** (2.4103)
lnpgdp			0.0944 ** (2.2287)
lnfdi			−0.0054 *** (−3.0817)
gov			−0.1063 ** (−2.5085)
urban			−0.0031 *** (−4.2824)
science			0.5695 * (1.8326)
Constant	0.3379 *** (155.9235)	0.3015 *** (68.2941)	−0.3520 (−0.9094)
Observations	4794	4794	4794
City FE	NO	YES	YES
Year FE	NO	YES	YES
R-squared	0.0195	0.1705	0.2145

Note: *, **, and *** represent significance levels of 10%, 5%, and 1%, respectively; the values in parentheses are t-statistics.

reports the benchmark regression results. In detail, control variables and fixed effects are not included in the model in column (1). For robustness, we added two-way fixed effects in column (2). Based on column (2), we further introduced control variables in column (3) for the regression. Taken together, the coefficient of NDEP is significantly positive with or without control variables, meaning that NDEP does increase urban CEE. As presented in column (3), the estimated coefficient of NDEP is 0.0349. It shows that NDEP construction increases urban CEE by 3.49%, validating Hypothesis 1. Liu et al. (2023) found that cities with NDEP have a 3.1% higher CEE than cities without NDEP [23]. It once again proves that NDEP has a promoting effect on CEE.

4.2. Parallel Trend Test

It must pass the parallel trend test for the DID method to identify causality. To put it another way, before NDEP construction is initiated, the CEEs of both the treatment group and control group must follow a relatively consistent trend. Otherwise, it could be doubted that the net effect estimated using the model is not attributed to policy, but to the differentiation between the two groups. Therefore, this paper turned to an event study method to run an empirical analysis of the dynamic changes in urban CEE before and after the NDEP policy [42].

$${CEE}_{it}={\alpha }_0+\sum _{j=-5}^8{\beta }_j{NDEP}_{i,t+j}+{\alpha }_2{Control}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(4)

where ${\beta }_j$ is the influence coefficient of the NDEP time window, and measures the policy effect of NDEP construction in period $j$. ${NDEP}_{i,t+j}$ is the dummy variable. This variable takes the value 1 if city $i$ builds an NDEP in period $t+j$ and 0 otherwise. $t$ is the year of NDEP construction, $j$ is the year before and after NDEP construction, and other variables are seen in Equation (1).

According to Figure 2, during the five years before the NDEP policy, the estimation coefficient of NDEP in each period was basically positive but not obvious, indicating that the treatment and control groups showed a parallel time trend before the policy. In the current year of the policy and the three years thereafter, the difference between the two groups is not statistically significant, meaning the facilitating effect of NDEP on CEE did not show up in the third year. Until the fourth year, the coefficient of NDEP is significantly positive and continues climbing up, proving that NDEP increases CEE, but with some lag effect. One reason may be that the NDEP policy, along with supporting preferential policies, cannot instantly lower carbon emissions through green technological innovation. Nonetheless, as time goes by, green industry chains will be established, leading to improved energy efficiency. The impact of the policy to increase CEE will be apparent.

4.3. Robustness Test

4.3.1. Placebo Test

The influence of NDEP policy on CEE may be interfered with by unobservable missing variables and other random factors; hence, a placebo test is necessary. In light of previous studies, 33 cities were randomly selected from the 282 sample cities to be the pseudo-treatment group and the rest were the pseudo-control group [43]. Simultaneously, the policy implementation time was randomly generated for each pseudo-treatment group, constructing a pseudo-NDEP variable (random time for city and policy implementation). On such a basis, we rebuilt the benchmark regression model and repeated it randomly 1000 times. As the pseudo-treatment group was generated at random, the estimated coefficients on the pseudo-NDEP variables would not significantly deviate from 0 if they were not influenced by other unobservable factors. As shown in Figure 3, most estimated coefficients are distributed around 0 and generally exhibit a normal distribution. The vertical dashed line on the left represents the mean value of the coefficients of the pseudo-NDEP variable (0.0003969), which is smaller than the estimation coefficient of 0.0349 for the NDEP policy represented by the vertical line on the right. Thus, the improvement in urban CEE is indeed the result of the NDEP policy, not caused by other random factors, testifying to the robustness of the result.

4.3.2. PSM-DID

The NDEP policy is treated as an exogenous shock in our research. For a city to carry out NDEP construction, the local government must go through a procedure of application, acceptance check, nomination, and approval from the central government. The cities are selected according to multiple factors, including geographical location, industrial structure, economic development level, and opening-up level. In fact, most NDEPs are upgraded from provincial or national development zones, leading to bias in selecting sample cities. This paper used propensity score matching with DID (PSM-DID) to reduce the potential selection [44]. Specifically, the control variables mentioned above were used as matching variables to find the most similar control group using caliper 1:1 nearest neighbor matching.

Table 3. Balance test results.

Variable	Unmatched or Matched	Mean		%Bias	%Reduct |Bias|	t-Test		V(T)/ V(C)
		Treated	Control			t	p > |t|
lnpgdp	U	10.7	9.9006	104.0		21.87	0.000	0.73
	M	10.676	10.703	−3.5	96.6	−0.55	0.583	0.64
lnfdi	U	13.578	10.817	157.5		30.60	0.000	0.42
	M	13.466	13.461	0.3	99.8	0.06	0.951	0.95
gov	U	0.1176	0.1735	−63.1		−12.41	0.000	0.46
	M	0.1163	0.1172	-1.0	98.4	−0.23	0.817	1.63
urban	U	64.196	48.616	103.6		22.15	0.000	0.80
	M	63.437	64.543	−7.4	92.9	−1.12	0.261	0.63
science	U	0.0240	0.9110	81.5		21.21	0.000	2.08
	M	0.0239	0.0229	6.5	92.0	0.84	0.403	0.72

reports the results of the balance test. After matching, the t-test results of all control variables were not significant, and the standard deviation was controlled below 10%. It indicates that the matching method adopted in this paper and the matching results obtained are credible. After matching propensity scores, this paper runs regression via Equation (1). Column (1) in

Table 4. Robustness test results.

Variables	(1)	(2)	(3)	(4)	(5)	(6)
NDEP	0.0340 ** (2.2675)	0.0319 ** (2.0577)	0.0378 ** (2.1989)	0.0338 ** (2.3185)
LCC					0.0063 (0.7354)
EDC						0.0085 (0.7506)
Control variables	YES	YES	YES	YES	YES	YES
Constant	−0.2834 (−0.7112)	−0.3743 (−0.9484)	−0.3299 (−0.7886)	−0.3535 (−0.9206)	−0.3296 (−0.8625)	−0.3471 (−0.9223)
Observations	4730	4726	4352	4794	4794	4794
City FE	YES	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES	YES
R-squared	0.2174	0.2091	0.2014	0.2150	0.2101	0.2101

Note: ** represents significance level of 5%; the values in parentheses are t-statistics.

shows PSM-DID results. The coefficient of NDEP is positive at the 5% significance level, nearly the same as the benchmark regression result, proving that the NDEP policy robustly raises urban CEE.

4.3.3. Excluding Municipalities and Provincial Capital Cities

Compared with other prefecture-level cities, municipalities and provincial capital cities may have higher economic development and energy use efficiency. Thus, to accurately measure the effect of policy, this paper carried out regression after excluding data from the following cities, respectively: four directly administered municipalities—Beijing, Tianjin, Shanghai, and Chongqing—as well as 26 provincial capital cities [45]. Columns (2) and (3) in Table 4 represent the results after regression. The significant positive estimation coefficient of the variable NDEP indicates that the benchmark regression result is robust and free of any significant effect from the special samples.

4.3.4. Excluding Interference from Other Policies

Within the sample period, CEE may be affected by other city-level policies, such as low-carbon city (LCC) and new energy demonstration city pilot (EDC) policies [30,46]. Eliminating the interference of the above policies on the NDEP policy effect is an essential step to check the robust results. We chose the following two approaches for this test. First, we multiplied the dummy variables of relevant policy pilot cities and years to obtain the interaction term and added both interaction terms as control variables to Equation (1) simultaneously. The regression results are represented in column (4) in Table 4. It can be found that after controlling the above policies, the coefficients of NDEP are still significantly positive. Second, the core explanatory variable in the baseline regression model was replaced with dummy variables for each of the two contemporaneous policies mentioned above. As shown in columns (5) and (6), the coefficients for both policies are not significant. Thus, it is reasonable to believe that urban CEE is raised by NDEP policy, not by other policies.

4.3.5. Bacon Decomposition of DID with Multiple Periods

DID with multiple periods may be subject to the heterogeneity of the treatment effect, resulting in biased estimation results. For this reason, the two-way fixed effects are decomposed with reference to Goodman-Bacon (2021) [47]. Figure 4 shows the Bacon decomposition results of the NDEP policy on CEE. It can be seen that the overall DD estimation is 0.0349, positive at the 5% significance level, in line with the benchmark regression result. The timing groups included pre-treatment and post-treatment groups, and post-treatment and pre-treatment groups account for a small weight (3.57%), while the weight of the non-treatment group and treatment group is as high as 92.07%. It means the heterogeneous treatment effect exerts a relatively small influence on the estimation result, and the baseline regression result is robust.

4.4. Mechanism Analysis

From the above empirical analysis, the NDEP policy greatly improves CEE. Next, we probed into how NDEP exerts its impact. To this end, we constructed the following model to discuss the potential mechanisms of NDEP policy from two perspectives: industrial structure and green technological innovation [48].

$$M_{it}={\alpha }_0+{\alpha }_1{NDEP}_{it}+{\alpha }_2{Control}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(5)

$${CEE}_{it}={\gamma }_0+{\gamma }_1{NDEP}_{it}+{\gamma }_2{M}_{it}+\beta {Control}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(6)

where $M_{it}$ stands for the mediating variable, ${\alpha }_1$ denotes the effect of NDEP on the mediating variable, and ${\gamma }_2$ denotes the impact of the mediating variable on CEE. The meanings of other variables are similar to Equation (1).

Table 5. Mechanism test results.

Variables	(1)	(2)	(3)	(4)
	ins	CEE	inno	CEE
NDEP	−3.7877 *** (−5.2608)	0.0282 ** (1.9700)	2.4976 *** (4.4161)	0.0024 (0.1565)
ins		−0.0018 *** (−3.6666)
inno				0.0130 *** (3.3495)
Constant	−133.7926 *** (−5.9653)	−0.5868 (−1.5029)	18.6435 ** (2.3769)	−0.5941 * (−1.8965)
Control variables	YES	YES	YES	YES
Observations	4794	4794	4794	4794
R-squared	0.4199	0.2253	0.3945	0.2376
City FE	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Sobel Z	0.0283 *** (9.981)		0.0265 *** (6.067)
ACME	0.0281 [0.0227, 0.0340]		0.0264 [0.0182, 0.0348]
ρ at which ACME = 0	−0.253		0.089

Note: *, **, and *** represent significance levels of 10%, 5%, and 1%, respectively; the values in parentheses are t-statistics; the value within [] is a 95% confidence interval.

represents the results of Equations (5) and (6). In column (1), the NDEP coefficient is negative at the 1% significance level, implying that NDEP lowers the share of secondary industry in pilot cities. After introducing the mediating variable ins in Equation (6), the estimation coefficient of ins stays negative at the 1% significance level, while that of NDEP is 0.0282, greatly less than that in Equation (1). It means that industrial structure mediates between NDEP construction and CEE, which is consistent with Hypothesis 2. In other words, NDEP optimizes the industrial structure by restraining over-industrialization, thus maximizing resource environment and economic benefits. Some carbon-intensive enterprises, unable to afford the environmental costs, have been forced to relocate out of NDEP and cities. Associated industries moved to neighboring regions [7]. In addition, NDEP’s incentives encourage more resources to flow toward clean enterprises, indirectly leading to carbon-intensive enterprises out of the local market. The combination of these two mechanisms results in the carbon-reducing effect of NDEP.

Column (3) shows that after NDEP construction, the city drastically improves in green technology innovation. In column (4), the estimation coefficients of the mediating variable and the Sobel test are positive at the 1% significance level. It proves that NDEP increases CEE via green technological innovation and Hypothesis 3 is verified. As the actual objects for achieving NDEP’s development goals, enterprises are pushed to innovate more environmentally in production activities to maintain their original profitability. NDEP provides a rapid and sustainable pathway for the exchange and output of innovation factors in enterprises through industrial symbiosis networks. The concentration of innovative elements attracts both external and internal enterprises of NDEP to form cooperative and competitive relations for technological exchange. Through demonstration and imitation, technology catch-up traps are avoided, thereby achieving breakthroughs in green technology throughout the city. In addition, the policy tilt for NDEP key projects enables enterprises committed to innovative production technologies to enjoy timely policy dividends and provide technical support for urban low-carbon transformation from the side.

Due to the significant causal inference flaws in the traditional three-step mediation mechanism, a sensitivity analysis was conducted on the aforementioned mediating variables to verify the robustness of the mediation effect [49]. Table 5 shows the results of the sensitivity test. For industrial structure, when the sensitivity test parameter ρ is −0.253, the average causal mediation effect (ACME) is equal to 0. When −0.253 < ρ <1, the sign of ACME always remains positive. It means that the ACME of industrial structure is robust even if it deviates greatly from the sequential ignorability assumption. Similarly, for green technology innovation, when the sensitivity test parameter ρ is 0.089, the ACME is equal to 0. When −1 < ρ < 0.089, the sign of ACME always remains positive. The mediating effect results are equally robust.

4.5. Heterogeneity Analysis

In this section, the heterogeneity of NDEP policy effects is further explored from multiple angles, such as the urban geographical location, resource endowment, and NDEP quantity. To this end, drawing from the method of Wang et al. (2022) [50], the following model was constructed:

$${CEE}_{it}={\alpha }_0+{\beta }_1{NDEP}_{it}\times group+{\alpha }_2{Control}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(7)

where $group$ is the grouping variable standing for the dummy variables of the geographical location, resource endowment, and NDEP quantity, and ${\beta }_1$ is the coefficient of ${NDEP}_{it}\times group$, the key of this section. The meanings of other variables are similar to Equation (1).

4.5.1. Geographical Location

Cities differ drastically in the social and economic environment, which further leads to spatial heterogeneity of the NDEP policy effects. The sample cities fell into two categories, the eastern group and the non-eastern group, based on their geographical locations. For any city, if it falls into the eastern group, then $group=1$, otherwise $group=0$. In accordance with column (1) in

Table 6. Heterogeneity analysis results.

Variables	(1)	(2)	(3)
	Geographical Location	Resource Endowment	Park Quantity
NDEP × Eastern	0.0569 ** (2.2734)
NDEP × Central	−0.0221 (−0.7534)
NDEP × Western	−0.0812 *** (−3.0818)
NDEP × Resource		0.0099 (0.6253)
NDEP × Park			0.0572 ** (2.1227)
Constant	−0.3618 (−0.9360)	−0.3517 (−0.9087)	−0.1012 (−0.1134)
Observations	4794	4794	561
R-squared	0.2168	0.2145	0.5231
City FE	YES	YES	YES
Year FE	YES	YES	YES

Note: **, and *** represent significance levels of 5%, and 1%, respectively; the values in parentheses are t-statistics.

, the estimation coefficient of ${NDEP}_{it}\times group$ is positive at the 5% significance level. Therefore, NDEP mostly increases CEE of eastern cities, probably because NDEPs are mainly distributed in eastern areas, bringing in advantages in the talent pool, business environment, opening-up, and green technology. NDEP construction pools creative talents and resources, and promotes green technological innovation, thereby facilitating low-carbon transition. To further analyze the regional differences in the influence of NDEP on CEE in eastern, central, and western cities, the above steps were repeated to obtain the interaction terms between dummy variables and core explanatory variables in central and western regions, respectively. According to the regression results in column (1) of Table 6, NDEP has no significant effect on CEE in central cities, and has a reverse effect on CEE improvement in western cities. The likely explanation is that the central areas are dominated by the secondary industry, lagging in resource allocation and independent research and development, which slows down the speed of urban low-carbon transition in the region. Meanwhile, western areas fall behind in infrastructure and innovation incentives; thus, indiscriminative construction of EIP would waste resources and inhibit the improvement in CEE [42].

4.5.2. Resource Endowment

This paper divided sample cities into resource-based and non-resource-based cities, to testify whether NDEP policy effects would vary greatly according to resource endowment. For each city, if it falls into resource-based, then $group=1$, otherwise $group=0$. Column (2) of Table 6 represents that the estimation coefficient of ${NDEP}_{it}\times group$ is positive yet of no statistical significance, indicating that NDEP policy fails to significantly increase CEE. It is probably because resource-based cities may increase carbon emissions as they are highly dependent on mineral resources, curbing the progress of low-carbon industries. In this case, the current industrial structure and production model cannot adjust quickly, hence undermining the contribution of NDEP policy. Meanwhile, non-resource-based cities focus more on developing resource-related industries, increasing energy utilization efficiency, and making their industrial layout more reasonable. Relevant studies also confirm our conclusion that the more dependent cities are on natural resources, the more likely they are to suffer the “CEE curse” [51].

4.5.3. NDEP Number

The number of NDEPs indirectly reflects the intensity of local environmental regulation. Compared with traditional industrial parks, NDEP must meet higher requirements of energy saving and lowering carbon emissions. A city with more NDEPs would establish stricter environmental regulations and receive more subsidies and financial support from the local government. As a result, it will pool creative resources and raise the R&D level of green technology, leading to increased CEE. Thus, the influence of NDEP on CEE may vary according to the local NDEP number. For that, this paper divided the sample into single-NDEP cities and multi-NDEP cities, and denoted the $group$ of the former 0 and the latter 1. As presented in column (3) of Table 6, the influence is significant at the 5% level. Multi-NDEP cities benefit more from the NDEP policy than single-NDEP cities, as proved by the findings of Qian et al. (2022) [48].

4.6. Analysis on Spatial Spillover Effect

Before analyzing the influence of NDEP on CEE of surrounding areas via Equation (2), the global Moran’I index should be used to check if there is a spatial autocorrelation in CEE.

Table 7. Global Moran’I index.

Year	Moran’I	p-Value	Year	Moran’I	p-Value
2003	0.053	0.000	2012	0.072	0.000
2004	0.032	0.000	2013	0.062	0.000
2005	0.052	0.000	2014	0.064	0.000
2006	0.053	0.000	2015	0.064	0.000
2007	0.059	0.000	2016	0.058	0.000
2008	0.059	0.000	2017	0.090	0.000
2009	0.063	0.000	2018	0.088	0.000
2010	0.081	0.000	2019	0.087	0.000
2011	0.080	0.000

presents the index results. It is evident that during 2003–2019, the global Moran’I of CEE stayed positive at the 1% significance level, proving the necessity of running an analysis based on SDID.

To figure out the spatial agglomeration of urban CEE, we drew the local Moran’I scatter plot of CEE in 2004, 2009, 2014, and 2019, as shown in Figure 5. Most cities are distributed in quadrants 1 and 3, meaning that the CEE of most cities showed significantly positive spatial autocorrelation in these four years. Next, this paper ran the Lagrange multiplier (LM) test, Wald test, likelihood ratio (LR) test, and Hausman test to further look into the fit of the SDM model, as shown in

Table 8. SDID regression results.

Variables	(1)	(2)	(3)	(4)	(5)
	Main	Wx	Direct	Indirect	Total
NDEP	0.028 *** (4.32)	0.222 *** (3.80)	0.030 *** (4.67)	0.479 *** (3.41)	0.509 *** (3.67)
rho		0.499 *** (6.08)
sigma		0.004 *** (48.88)
Control variables	YES	YES	YES	YES	YES
City FE	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES
Observations	4794	4794	4794	4794	4794
Robust LM-error			1569.143 ***
Robust LM-lag			43.346 ***
Wald error			72.18 ***
Wald lag			64.75 ***
LR error			74.75 ***
LR lag			65.68 ***
Hausman			200.94 ***

Note: *** represent significance level of 1%; the values in parentheses are t-statistics.

. It manifests that all the tests rejected the null hypothesis at the 1% level, implying that our analysis of the spatial impact of NDEP policy using only spatial lag and spatial error models may lead to errors. Thus, this paper decided on the SDM model for further analysis.

Based on Equation (2), Table 8 demonstrates the regression results with spatial spillover effects. As displayed in columns (1) and (2), the coefficients of NDEP and its lag term are both positive at the 1% significance level. That is, NDEP policy increases the CEE of local and neighboring cities via positive spillover, which preliminarily proves Hypothesis 1. Further, after decomposing the spatial effects of NDEP on CEE, we obtained direct, indirect, and total effects. Columns (3)–(5) demonstrate the results of the effect decomposition. We find that NDEP’s indirect effect (0.479) on CEE is much larger than its direct effect (0.030), suggesting that NDEP has significant positive spatial spillovers to CEE. It indicates that NDEP increases the CEE of pilot cities and their surrounding cities. Meanwhile, the coefficient of the total effect is 0.509, meaning that NDEP increases the average CEE by a whopping 50.9% in general. Hypothesis 1 is finally validated.

According to the First Law of Geography, as cities get farther away from one another, their spatial correlation dwindles. Thus, having confirmed the effect of NDEP on nearby CEE, we further investigated the boundary of space attenuation of such an effect. Drawing from the method of Jia et al. (2021) [52], we set various distance thresholds to construct a spatial weight matrix as follows:

$$w_d=\left\{\begin{array}{c}\frac{1}{d_{ij}},if d_{ij}>d\\ 0,if d_{ij}\le d\end{array}\right.$$

(8)

where $d_{ij}$ represents the geographical distance between two cities, and $d$ is the distance threshold. We first set the initial distance threshold as 50 km and gradually added, then put the obtained $w_d$ into Equation (2) to attain the estimation coefficient of the indirect spillover effect and 95% confidence interval.

In accordance with Figure 6, as cities become farther away, the spatial spillover of NDEP policy on CEE generally gets attenuated. Within 50–450 km, the coefficient of the indirect effect is significantly positive, showing an inverted V-shaped tendency—first increasing, then declining, and peaking at 100 km of the distance threshold. It means that within such distance, NDEP drives the CEE of neighboring cities to boost, while such an effect declines as cities become farther away. The reason may be that industrial symbiosis pools elements of innovation, and a closer distance gives full play to technology diffusion and knowledge spillover. In the meantime, environmental protection constitutes a major part of local governmental officials’ accomplishments. Hence, cities would learn from one another in terms of management methods and experience in NDEP construction. However, when two cities are over 450 km away from each other, the spillover effect of NDEP becomes insignificant, and the estimation coefficient even turns negative when the distance surpasses the threshold of 500 km. That is, the spillover effect of NDEP policy is affected by the administrative division and geographical location. NDEP policy mostly affects neighboring cities within a province, not including cities over 450 km away from each other in the same provinces. The findings of Qian et al. (2022) support this conclusion [48].

5. Discussions

Based on modes of formation, industrial symbiosis falls into two categories: top–down, designed and promoted by the government, and bottom-up, the synergy arising spontaneously from the demands of enterprises [53]. At present, the government-led top–down approach prevails in China. The Chinese government achieves industrial symbiosis by directly transforming traditional industrial parks into EIPs or building new EIPs considering symbiotic relationships between enterprises. That means industrial symbiosis requires corresponding policy support to play a more effective role in China [54]. NDEP is one of the practical forms of industrial symbiosis under the leadership of the government.

Achieving the carbon peak and carbon neutrality in the coming period are commitments the country has made to the world [55]. In this context, NDEP is about extending and developing traditional industrial parks. Our study found that NDEP can indeed promote local urban CEE. In agglomeration areas, industries shifted from one-way pollution control to resource recycling after symbiotic transformation, thereby reducing carbon emissions and alleviating the negative environmental externality. Wang et al. (2021), Chen et al. (2022), and Yu et al. (2023) drew similar conclusions via DEA, an input–output model, an emergy analysis, and a linear programming model [56,57,58]. However, they mainly looked into industrial symbiosis and carbon emissions by selecting a certain industrial park as a case, with insufficient and relatively subjective samples. Our empirical study of industrial symbiosis and carbon emissions is an important addition to the relevant literature on environmental effects of industrial symbiosis.

Although both this study and Liu et al. (2023) used DID and SDID models to explore the carbon abatement effects of NDEP [23], there are still differences in the research perspective and depth of the subsequent analyses. First, their research topic is green location policy, not industrial symbiosis. Different entry points lead to different emphases of policy analyses. Second, they do not give the specific scope of the spatial spillover effect of NDEP. By constructing a spatial weight matrix via different geographical distance thresholds, this paper obtains the range of NDEP policy’s influence on the neighboring cities. Finally, this paper reveals the impact of NDEP on CEE in cities with different geographical locations and NDEP quantities. The number of NDEPs indirectly reflects the intensity of local environmental regulation. Further attention to the influence of the number of NDEP on CEE is helpful for local governments to grasp the policy intensity when formulating policies.

6. Conclusions and Policy Suggestions

6.1. Conclusions

Industrial symbiosis helps form a material closed-loop, reduces energy consumption and greenhouse gas emissions, as well as promotes sustainable development in the region. This paper collected 282 cities from 2003 to 2019 and constructed DID with multiple periods and SDID models, to explore the influence of industrial symbiosis policies represented by NDEP policy on CEE. Our key findings include the following:

First, NDEP policy increases urban CEE by 3.49%. After a set of checks such as the placebo test, PSM-DID, exclusion of special samples and other policies, and bacon decomposition, the result stays robust. It can be seen that industrial symbiosis exerts a significantly positive influence on CEE. Second, the result of the mechanism analysis proves that NDEP policy mainly increases CEE by means of industrial structure upgrading and green technological innovation. Third, factors like geographical location, resource endowment, and NDEP number exert a heterogeneous impact on CEE. In non-resource-based, eastern, and multi-NDEP cities, CEE sees some significant improvement. Finally, NDEP policy increases not only the CEE of pilot cities but also that of their neighboring cities within 450 km via diffusion and demonstration effects.

6.2. Policy Suggestions

One of the major realities China faces today is how to go about lowering carbon emissions. The above conclusions provide a new way to solve this problem. To this end, the following suggestions are made for the attention of the authorities:

(1)

The Chinese government should work to transform traditional industrial parks into EIPs to enhance energy efficiency and lower carbon emissions. Currently, less than 20% of cities have adopted NDEP policy, leaving a huge potential for policy promotion. Therefore, according to its own economic development, the government should increase financial support, and promote NDEP application and construction in an orderly manner. In addition, further improving the NDEP supervision and assessment system and highlighting the importance of carbon emission reduction are also important measures to promote NDEP development.

(2)

Green technological innovation is key to promoting urban CEE. Hence, the government should create incentives for enterprises to contribute to green innovation. First, infrastructure is updated and a stage is set for industrial symbiosis. Second, subsidies and tax preferences are offered to pool high-tech enterprises and innovative talents. Last, the protection of intellectual properties is enhanced and a sound environment for green technological innovation is created.

(3)

The differences in geographical location, resource endowments, and environmental regulations are what the Chinese government needs to focus on. Governments can make specific strategies for park adjustment targeted at different cities. Resource-based cities must gradually rule out carbon-intensive industries, and increase energy use efficiency by changing the conventional structure of fossil-fuel-based energy resources. Western areas lag in economic development, so they should avoid indiscriminative construction of EIPs. Instead, work should be done to improve the market environment and invest in information infrastructure.

(4)

NDEP not only increases the CEE of pilot cities but also improves CEE in their nearby cities. Hence, the Chinese government must rationally plan the NDEPs’ spatial distribution and fully leverage the diffusion and demonstration effects of NDEP. In addition, to build NDEPs, we must consider the spillover range of spatial effects to facilitate the exchange and cooperation of knowledge and technologies between administrative regions.

6.3. Future Considerations

Although this study captures the spatial spillover effect of an NDEP pilot on CEE, the following three issues remain to be considered: First, the SDID model constructed in this paper is only used to identify the impact of NDEP on CEE in and around the pilot cities. Some scholars, however, have used this model to complete the whole process from parallel trend testing to a heterogeneity analysis, especially to examine whether pilot policies of different types of cities radiate to neighboring cities. Therefore, the future is to apply the SDID model to related research topics based on overcoming the model’s endogeneity. Furthermore, in addition to the intensity of local government environmental regulation, factors such as urban resource allocation and management efficiency also have different degrees of influence on the effectiveness of NDEP policy. In the future, we will use case studies or in-depth interviews to explore how NDEP policy implementation differs across cities under the influence of these factors. Finally, the super-efficiency SBM model used in this paper can determine the direction of efficiency improvement based on slack variables and accurately reflect the differences in effective decision-making units. However, it is only one of the methods for calculating CEE, and future researchers can try to measure CEE using other methods, such as the super-efficiency epsilon-based measure model.