How Does the “Civilized City” Selection Affect Environmental Governance Performance? A Spatial DID Approach Based on Prefecture-Level Cities

Abstract

Our study employs panel data from 272 Chinese prefecture-level cities (2003–2020), leveraging the “Civilized City” selection campaign as a quasi-natural experiment. Using a Spatial Durbin Difference-in-Differences model, we systematically analyze the policy’s impact on local environmental governance performance and its spatial spillover effects, with rigorous robustness checks. Results reveal a significant positive spatial correlation in China’s environmental governance performance, indicating interdependence among cities rather than isolated decision-making. The “Civilized City” initiative not only improves local environmental governance but also generates spillover benefits for neighboring regions, thereby enhancing coordinated regional sustainability. Finally, we propose policy recommendations grounded in empirical findings and China’s governance context.

1. Introduction

Due to rapid economic growth, limitations in awareness and ideology, and delays or inadequacies in governance implementation, China has faced increasingly prominent issues of resource–environment conflicts and ecological degradation. According to the 2022 Environmental Performance Index (EPI) released by Yale University’s Center for Environmental Law and Policy (YCELP) in collaboration with Columbia University’s Center for International Earth Science Information Network (CIESIN), China’s EPI score has increased by 11.40 compared with a decade ago. EPI usually indicates the achievements and performance of an organization, enterprise, or country in the aspects of environmental protection and sustainable development. However, among 180 participating countries and regions, China is positioned in the 88.89th percentile in terms of environmental performance, reflecting a less favorable standing in environmental governance. Over the years, both central and local governments in China have implemented various policy measures and practices to improve environmental performance. However, China’s environmental policies have long struggled due to incompatibility with the promotion incentives of local officials and the motivations of corporate actors [1,2]. Because of environmental externalities and local self-interest, issues such as formalism, government–business collusion, and falsification of environmental data have emerged [3,4], significantly hampering the implementation of these policies and leading to a substantial gap between actual results and intended environmental governance goals.

Currently, “Civilized City” is a widely promoted comprehensive evaluation policy. The policy considers the comprehensive development achievements of a city from five perspectives: economic, political, cultural, social, and ecological. However, studies on the influence of this policy on environmental governance performance are limited. Barmé and Goldkorn (2013) considered the “Civilized City” initiative as a strategy for urban governance, which not only promotes urban economic development but also standardizes urban management [5]. Greenstone and Hanna (2014) and Liao and Wu (2025) argued that robust public support in weak institutional environments can enable the success of environmental regulations [6,7]. Hosseini et al. (2013) and Li et al. (2014) confirmed the spatial correlation of environmental pollution among nations and Chinese provinces/cities, respectively [8,9]. Hottenrott and Rexhäuser (2015) provided evidence supporting the idea that decentralized constraints prompt local governments to adopt active and efficient environmental policies, thereby improving regional environmental quality [10]. Zuo et al. (2017) developed a provincial environmental performance evaluation index system for China using thematic frameworks and the DPSIR model, and evaluated the eastern, central, and western regions [11]. Li et al. (2021) found that “Civilized City” initiatives improved public services in education, science, ecology, culture, and healthcare [12]. Zhang et al. (2021) analyzed corporate environmental performance under government intervention through “Civilized City” evaluations [13]. Chai et al. (2022) and Dong (2024) used quasi-natural experiments to demonstrate that “Civilized City” evaluations improve the social responsibility of enterprises and also improve environmental performance [14,15]. In 2023, Yang et al. began focusing on urban environmental governance and analyzed the influence of “Civilized City” evaluations on the environment from the perspective of urban “green energy innovation” [16,17,18]. By reviewing the existing literature, we find that research on the “Civilized City” selection policy and its impact on environmental governance performance has made considerable progress. It provides a solid foundation for this study. However, some shortcomings remain to be addressed.

First, as a typical tool for incentivizing local governments in urban governance, the influence of the “civilized city” selection on local environmental governance performance has received relatively little attention from scholars. Understanding whether this new method of authoritative recognition, such as the “Civilized City” selection, can produce different environmental governance effects and thus enrich the toolbox of environmental governance is both necessary and practically significant.

Second, existing studies use limited single-dimensional pollution indicators to measure regional environmental governance, and most focus on industrial pollution control, such as industrial “three wastes” [19,20] or industrial sulfur dioxide emissions [21], often neglecting research on pollution governance in residential life or ecological environmental governance. Third, current research on the impact of the “Civilized City” selection on environmental governance mainly employs the difference-in-differences (DID) method without considering spatial factors. The traditional DID method typically assumes that pollution emissions and environmental governance behaviors in different regions are independent of each other. However, this assumption clearly deviates from reality, as pollutants such as gases and liquids can flow across borders, and government governance behaviors may also involve “passing the buck” or “neighboring imitation”. Neglecting the spatial correlation may cause incorrect inferences and poor model examination [22].

Based on this, we take prefecture-level cities in China as the study object to analyze the influence of the “civilized city” selection on urban environmental governance performance. First, we conduct spatial autocorrelation tests to confirm the existence of spatial dependence in urban environmental governance performance, followed by a series of specification tests for econometric model identification. Second, using a Spatial Durbin Difference-in-Differences (SDM-DID) model, we obtain regression results and conduct spatial effect decomposition to examine the impact of the “Civilized City” campaign on environmental governance performance. We provide a detailed mechanistic analysis of these findings and perform multiple robustness tests to enhance the scientific validity and persuasiveness of our conclusions.

The main contributions are as follows: First, the existing literature has mainly used pollution emission indicators, such as industrial sulfur dioxide and industrial wastewater, to measure regional environmental governance. By contrast, this study employs the entropy method to derive a comprehensive indicator for measuring environmental governance. It selects relevant indicators from three areas—industrial pollution control, residential pollution control, and ecological environmental governance—to construct a comprehensive indicator system, thus providing a more realistic and comprehensive reflection of regional environmental governance performance. Second, the previous literature primarily uses the traditional difference-in-differences (DID) method to analyze the relationship between “Civilized City” selection and environmental governance. This paper, however, incorporates spatial factors and adopts the Spatial Durbin Difference-in-Differences (SDM-DID) method to investigate the spatial spillover effects of the “Civilized City” selection on environmental governance performance. DID and SDM-DID can be used to conduct quasi-natural experiments, which is a social science study method that uses naturally occurring events or policy changes to simulate experimental conditions, thereby providing a basis for causal inference. These experiments do not rely on the direct manipulation of experimental conditions by researchers, but rather identify causal relationships by observing the impact of exogenous events on the treatment and control groups.

2. Material and Methods

2.1. Theoretical Considerations and Development of Hypotheses

The “National Civilized City” title represents the highest honor in China’s comprehensive city evaluation system, offering a powerful incentive mechanism. “Prefecture Civilized City” is a part of the national civilized city evaluation system, specifically referring to the evaluation of cities at the prefecture level. Its definition is essentially the same as that of a nationally civilized city, but it places greater emphasis on the exemplary role of prefecture-level cities in regional development. As a visible acknowledgment of outstanding urban development performance, this designation can facilitate the promotion process of local government officials. One crucial aspect of urban civilization improvement is the protection of ecology and the environment. Therefore, it is reasonable to hypothesize that the “Civilized City” selection process could lead to significant environmental governance benefits. Under the incentive of promotion, through which channels does the “Civilized City” selection influence environmental governance performance? Based on a literature review, this study clarifies the transmission mechanisms as follows:

(1)

The Technological Progress Effect

The “Civilized City” evaluation improves environmental governance performance through technological progress. This effect operates through two main pathways: On one hand, the “Civilized City” selection, which aims to promote sustainable and high-quality urban development. Its evaluation system includes indicators related to technological progress [23], which impose implicit requirements on cities to focus on technological innovation. Participating cities are incentivized to invest in science and technology development, thereby fostering technological progress and providing robust technological support for green development. On the other hand, from the perspective of environmental governance in China, government departments are the primary regulators and responsible entities, while enterprises are the main targets of governance. To meet the “Civilized City” criteria, city governments have been incentivized to strengthen urban management and tighten environmental regulations. Therefore, the “Civilized City” selection campaign addresses environmental governance challenges by driving green technology advancement through technological progress, yielding tangible environmental improvements.

(2)

The Public Participation Effect

As public awareness of the desirability of harmonious development between humans and nature increases, the demand for higher environmental quality intensifies, resulting in more environmental protection appeals. These demands prompt local governments to implement more environmental initiatives such as energy-saving projects and green financial support. Given the complexity and wide scope of “Civilized City” initiatives, local governments may find it challenging to coordinate efforts using their resources alone. Thus, under constrained time and resources, governments must collaborate with existing community self-governance forces to form effective interaction and compensate for limited administrative resources. By mobilizing society adaptively to attract public participation, local governments can avoid inefficiencies in policy implementation [24]. The “Civilized City” evaluation, therefore, enhances urban environmental governance by prompting government actions through public participation.

Hypothesis 1:

The “Civilized City” evaluation positively influences environmental governance performance.

The cross-border movement of pollutants, such as wastewater and exhaust gas, creates spatial interconnections, meaning that the environmental quality of one region is inevitably affected by pollution emissions from neighboring regions. This finding underscores the strong spatial interdependence of environmental governance. Hosseini and Kaneko have confirmed a significant spatial correlation of environmental pollution across countries [8]. Similarly, studies have demonstrated substantial spatial associations of environmental pollution across provinces and cities in China [9,25]. Moreover, factors such as industrial layouts and public policies further intensify the spatial interdependence of environmental quality among regions [26,27].

Local governments with weaker governance capabilities are highly motivated to learn and emulate the governance concepts and methods of the leading regions. Neighboring “Civilized Cities” serve as exemplary models by exploring modern governance pathways, creating “replicable and scalable” approaches, and encouraging cross-regional learning and adoption of best practices. This leads to strengthened mutual learning and mimetic behaviors among cities, facilitating regional environmental governance. Thus, “Civilized Cities” act as “trailblazers,” promoting environmental governance improvements in neighboring regions through the imitation effect.

There is a significant interdependence among adjacent cities. Local governments not only engage in “yardstick competition” in terms of economic development but also compete in urban governance and other domains [21]. The “Civilized City” selection process provides measurable results for comparing governance levels among cities, intensifying competition among local governments. The top-down benchmarking competition initiated by the central government leverages a mechanism of winning official commendations and public approval to stimulate the formation of competitive relationships among cities. The “Civilized City” selection features an effective incentive mechanism, and under the allure of this honor, many cities not selected as “Civilized” strive to catch up by strengthening urban governance and continuously improving urban civilization. Consequently, cities that receive the “Civilized City” designation generate a competitive effect on other regions, compelling them to increase their pollution control efforts and improve their environmental governance. This results in the spatial spillover effect of the “Civilized City” selection, driving environmental governance activities and performance improvements in neighboring areas.

Hypothesis 2:

The “Civilized City” evaluation produces spatial spillover effects, enhancing regional environmental governance performance.

2.2. Sample Selection and Data Sources

In China, the hierarchy of cities is classified by administrative levels. For example, municipalities are directly managed by the central government and enjoy a high degree of administrative autonomy. They have extensive management authority over administrative, economic, and social affairs and can directly interact with central ministries and commissions. Prefecture-level cities are mainly focused on the management of local affairs and managed by the provincial government. This study employs panel data from 272 prefecture-level cities from 2003 to 2020 for empirical analysis. The choice of the study period is due to the lack of data at the prefecture-city level before 2003. Therefore, based on data consistency and availability, the research period spans from 2003 to 2020. Usually, we evaluate “National Civilized City” based on the “National Civilized City Evaluation System” released by the China Central Civilization Committee. The evaluation system mainly includes four parts: (1) urban civilization level, which covers urban cultural construction, the civility of citizens, urban environmental sanitation, and so on; (2) urban infrastructure construction, including transportation, housing, water conservancy, energy, and communications; (3) social development, including education, science and technology, health, culture, and sports; (4) improvement of people’s livelihoods, including residents’ income, social security, employment, healthcare, and elderly care. Specific evaluation steps are as follows. First, provinces, autonomous regions, and municipalities recommend a number of candidate cities for the “National Civilized City” title. Second, the Central Civilization Office organizes third-party assessments of these candidate cities every year. The assessment methods include document review, questionnaire surveys, and on-site inspections. Third, the weighted average score over three years of assessment is used as the basis for selecting the new batch of “National Civilized Cities”.

This study excludes samples from municipalities directly under the central government and county-level cities due to significant political and economic differences, as well as data limitations. Prefecture-level cities (e.g., Bijie City in Guizhou Province, Sansha City in Hainan Province, and Hami City in Xinjiang) established after 2003 and cities in Tibet with severe data deficiencies are also excluded. Additionally, the Central Civilization Office announced six batches of “Civilized Cities” in October 2005, January 2009, December 2011, February 2015, November 2017, and November 2020.

Since the evaluation criteria for the first batch of “Civilized Cities” differed significantly from subsequent batches and the number of selected cities was relatively small, this study excludes the first batch of “Civilized Cities” for sample validity [28]. The analysis focuses on the last five batches. The resulting sample consists of 135 prefecture-level cities designated as “Civilized Cities” (treatment group) and 137 prefecture-level cities that did not receive this designation (control group).

The core explanatory variable, data related to the “Civilized City” selection, was obtained from the China Civilization Network. Data on central fiscal expenditure and annual average exchange rates were from the National Bureau of Statistics. Other data were from the China Urban Statistical Yearbook, with missing data supplemented by provincial and municipal statistical yearbooks, the EPS database, and the CNRDS database. Data that could not be supplemented were interpolated to construct a balanced panel dataset of 272 prefecture-level cities from 2003 to 2020. Additionally, the study logs the number of green patent applications (adding 1 before taking the logarithm) and logarithmically transforms other variables, except for industrial agglomeration [29]. The dataset comprises 272 cities over 18 years, totaling 4896 observations, with descriptive statistics and correlation analysis results reported in

Table 1. Descriptive statistics.

Variables	Mean	Standard Deviation	Minimum Value	Maximum Value
EGP	0.302	0.069	0.022	0.999
DID	0.200	0.400	0	1
lnPGDP	9.930	0.800	7.545	12.363
AGGLO	0.928	0.286	0.144	1.948
lnFISC	1.140	0.595	−0.648	3.469
lnTECH	4.060	1.946	0	9.801
lnFDI	−0.214	1.778	−11.513	3.626

Note: ln indicates that the variable is logarithmically transformed.

and Figure 1.

2.3. Methodology

We take Chinese prefecture-level cities as the study object, designating the cities selected as the 2nd to 6th batch of “Civilized Cities” during the observation period as the treatment group, while other cities in the sample serve as the control group. To explore the influence of the “Civilized City” selection on China’s environmental governance performance, the baseline DID model is constructed as follows:

$${EGP}_{it}={\alpha }_0+{\alpha }_1{Treat}_i+{\alpha }_2{Post}_t+{\alpha }_3{Treat}_i\cdot {Post}_t+X_{it}^{\prime }\beta +{\epsilon }_{it}$$

(1)

Here, the subscripts $i$ and $t$ represent city and year, respectively; ${EGP}_{it}$ denotes environmental governance performance; ${Treat}_i$ is a grouping dummy variable, where if a city $i$ is selected as a “Civilized City”, ${\mathrm{T}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{t}}_i=1$, otherwise ${Treat}_i=0$; ${\mathrm{P}\mathrm{o}\mathrm{s}\mathrm{t}}_t$ is a time dummy variable, where if the time $\mathrm{t}$ is after the implementation of the “Civilized City” selection policy, ${Post}_t=1$, otherwise ${Post}_t=0$. The core explanatory variable is the interaction term of the grouping dummy variable and the time dummy variable ${Treat}_i\times {\mathrm{P}\mathrm{o}\mathrm{s}\mathrm{t}}_t$, and its coefficient ${\alpha }_3$ reflects the net effect of the “Civilized City” selection policy on environmental governance performance. $X_{it}^{\prime }$ represents other control variables that indicate the environmental governance performance of cities; ${\epsilon }_{it}$ represents the random error term; ${\alpha }_i\left(i=0,1,2,3\right)$ and $\beta$ are coefficients to be estimated.

The baseline DID model assumes that all individuals in the treatment group are simultaneously influenced by the policy, which violates the parallel trend assumption if it is not true. Since different prefecture-level cities are awarded “Civilized City” status at different times, we utilize a multi-period DID model based on the method of Beck, Levine, and Levkov (2010). It uses individual and time-fixed effects in the panel regression models [30]:

$$\begin{gathered} {EGP}_{it}={\alpha }_0+{\alpha }_1{DID}_{it}+{{\alpha }_{2}lnPGDP}_{it}+{{\alpha }_{3}AGGLO}_{it}+{{\alpha }_{4}lnFISC}_{it} \\ +{{\alpha }_{5}lnTECH}_{it}+{{\alpha }_{6}lnFDI}_{it}+{\mu }_i+{\gamma }_t+{\epsilon }_{it} \end{gathered}$$

(2)

where the dependent variable represents environmental governance performance in city $\mathrm{i}$ in year $\mathrm{t}$; and the core independent variable is a treatment variable that varies across time and individuals, replacing the interaction term in the baseline DID model. Its value is set as follows: if the city $\mathrm{i}$ is a “Civilized City” and the time is the first year or later of its three-year creation cycle, ${DID}_{it}=1$; otherwise, ${DID}_{it}=0;{\mathrm{l}\mathrm{n}\mathrm{P}\mathrm{G}\mathrm{D}\mathrm{P}}_{\mathrm{i}\mathrm{t}}$, ${\mathrm{A}\mathrm{G}\mathrm{G}\mathrm{L}\mathrm{O}}_{\mathrm{i}\mathrm{t}}$, ${\mathrm{l}\mathrm{n}\mathrm{F}\mathrm{I}\mathrm{S}\mathrm{C}}_{\mathrm{i}\mathrm{t}}$, ${\mathrm{l}\mathrm{n}\mathrm{T}\mathrm{E}\mathrm{C}\mathrm{H}}_{\mathrm{i}\mathrm{t}} \mathrm{a}\mathrm{n}\mathrm{d} {\mathrm{l}\mathrm{n}\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i}\mathrm{t}}$ represent the control variables in this study: economic growth, industrial agglomeration, fiscal decentralization, green technology, and foreign direct investment, respectively. ${\mu }_i$ represents the individual fixed effects, which replaces the rough group dummy variable ${\mathrm{T}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{t}}_{\mathrm{i}}\cdot {\mathrm{P}\mathrm{o}\mathrm{s}\mathrm{t}}_{\mathrm{t}}$ in the baseline DID model. The time-fixed effect ${\gamma }_t$ replaces the period dummy variable ${\mathrm{P}\mathrm{o}\mathrm{s}\mathrm{t}}_{\mathrm{t}}$. ${\alpha }_i\left(i=\mathrm{0,1},\dots ,6\right)$ represents the coefficients to be estimated, and the coefficient ${\mathsf{\alpha }}_1$ reflects the net effect of the “Civilized City” selection policy on environmental governance performance. Because a city’s environmental governance is not only related to its neighboring regions but also to other social and economic factors, this study employs the Spatial Durbin-Difference-in-Differences (SD-DID) model. This includes the spatial lags of the dependent and independent variables to analyze the spatial effects of “Civilized City” evaluations on environmental governance. The model is formulated as follows:

$$\begin{gathered} {EGP}_{it}=\rho {\sum }_{j=1}^n{w}_{ij}{EGP}_{jt}+{\alpha }_1{DID}_{it}+{{\alpha }_{2}lnPGDP}_{it}+{{\alpha }_{3}AGGLO}_{it}+{{\alpha }_{4}lnFISC}_{it} \\ \begin{array}{c}+{{\alpha }_{5}lnTECH}_{it}+{{\alpha }_{6}lnFDI}_{it}+{\beta }_1{\sum }_{j=1}^n{w}_{ij}{DID}_{jt}\\ +{\beta }_2{\sum }_{j=1}^n{w}_{ij}{lnPGDP}_{jt}+{\beta }_3{\sum }_{j=1}^n{w}_{ij}{AGGLO}_{jt}\\ +{\beta }_4{\sum }_{j=1}^n{w}_{ij}{lnFISC}_{jt}+{\beta }_5{\sum }_{j=1}^n{w}_{ij}{lnTECH}_{jt}\\ +{\beta }_6{\sum }_{j=1}^n{w}_{ij}{lnFDI}_{jt}+u_i+{\gamma }_t+{\epsilon }_{it}\end{array} \end{gathered}$$

(3)

where $\mathsf{\rho }$ represents the spatial autoregressive coefficient, and ${\alpha }_i\left(i=1,\dots ,6\right)$ and ${\beta }_i\left(i=1,\dots ,6\right)$ represent the coefficients of the explanatory variables and spatial lag coefficients, respectively. ${\mathrm{w}}_{\mathrm{i}\mathrm{j}}$ refers to the element in the $i$th row and $j$th column of the spatial weight matrix. $\mathsf{\rho }{\sum }_{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{w}}_{\mathrm{i}\mathrm{j}}{\mathrm{E}\mathrm{G}\mathrm{P}}_{\mathrm{j}\mathrm{t}}$ is the spatial lag of the dependent variable, environmental governance performance, while indicating the influence of the environmental governance performance of neighboring areas on the environmental governance performance of the region. ${\mathsf{\beta }}_1{\sum }_{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{w}}_{\mathrm{i}\mathrm{j}}{\mathrm{D}\mathrm{I}\mathrm{D}}_{\mathrm{j}\mathrm{t}}$ represents the spatial lag of the core explanatory variable, reflecting the impact of the “Civilized City” selection in neighboring areas on the environmental governance performance of the region. Other terms, such as $\mathsf{\beta }{\sum }_{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{w}}_{\mathrm{i}\mathrm{j}}{\mathrm{X}}_{\mathrm{j}\mathrm{t}}$, represent the spatial lags of the control variables, indicating the influence of neighboring areas’ control variables on the environmental governance performance of the region. The meanings of other parameters are consistent with those in Formula (2). This study, referencing the research of You and Lv (2018), constructs inverse geographical distance spatial weight matrices and k-nearest neighbor spatial weight matrices to perform robustness tests [31].

3. Results and Discussion

3.1. Spatial Autocorrelation Test

When there is strong spatial dependence among data, it is necessary to use spatial econometric methods. Scholars typically use Moran’s I to test for the spatial interdependence among observations. This study calculates Moran’s I for environmental governance performance data from 272 Chinese prefecture-level cities from 2003 to 2020. The results are presented in

Table 2. Moran’s I values for environmental governance performance in China (2003–2020).

Year	Moran’s I index	Statistic	p-Value
2003	0.077	2.467 **	0.014
2004	0.109	3.372 ***	0.001
2005	0.108	3.371 ***	0.001
2006	0.057	1.777 *	0.076
2007	0.090	2.688 ***	0.007
2008	0.082	2.462 **	0.014
2009	0.092	2.779 ***	0.006
2010	0.093	2.809 ***	0.005
2011	0.105	3.129 ***	0.002
2012	0.112	3.307 ***	0.001
2013	0.117	3.417 ***	0.001
2014	0.108	3.166 ***	0.002
2015	0.200	5.560 ***	0.000
2016	0.207	5.746 ***	0.000
2017	0.196	5.418 ***	0.000
2018	0.182	5.038 ***	0.000
2019	0.175	4.821 ***	0.000
2020	0.158	4.308 ***	0.000

Note: ***, **, * indicate significance at the 1%, 5%, and 10% levels.

.

Moran’s I values for environmental governance performance across Chinese prefecture-level cities from 2003 to 2020 are all greater than zero, showing a fluctuating trend. In 2006, the Moran’s I value reached its lowest point at 0.057, although it remained significant at the 10% level. Similarly, the values for 2003 and 2008 were relatively low but significant at the 5% level. For all the other years, the values were significant at the 1% level. This indicates a significant positive spatial correlation in the distribution of environmental governance performance across cities, reflecting the spatial clustering effect.

3.2. Spatial Panel Model Selection

First, the LM test was used to determine the presence of spatial effects in the econometric model. Then, LR and Wald tests were conducted to evaluate whether the chosen Spatial Durbin Model (SDM) could degenerate into the Spatial Autoregressive (SAR) model or the Spatial Error Model (SEM). The results are presented in

Table 3. Results of LM, LR, and Wald Tests.

		Statistic	p-Value
LM Test	LM-ERR	41.636 ***	0.000
	Robust LM-ERR	64.846 ***	0.000
	LM-LAG	0.395	0.530
	Robust LM-LAG	23.605 ***	0.000
LR Test	SDM → SAR	60.700 ***	0.000
	SDM → SEM	66.800 ***	0.000
Wald Test	SDM → SAR	61.030 ***	0.000
	SDM → SEM	67.010 ***	0.000

Note: a. For the LR and Wald tests, the null hypothesis H0: SDM can degenerate into SAR (SDM → SAR), or SDM can degenerate into SEM (SDM → SEM). b. *** indicates significance at the 1% level.

.

The results show that in the LM test, the null hypothesis of no spatial autocorrelation was rejected at the 1% significance level for spatial error, and for spatial lag, one test also rejected the null hypothesis. This finding confirms the necessity of using a spatial econometric model. In the LR and Wald tests, the p-values were all much lower than 0.01, indicating that the SDM could not be simplified into either the SEM or SAR model. Therefore, the SDM is suitable for this study.

After confirming the use of the SDM, the Hausman test was applied to determine whether to use fixed or random effects. If the Hausman test is passed, the LR joint significance test is conducted to decide whether to adopt individual fixed effects, time-fixed effects, or a two-way fixed effects model. The results are presented in

Table 4. Results of the Hausman Test and LR Joint Significance Test.

		Statistic	p-Value
HausmanTest		31.330 ***	0.000
LR Joint Significance Test	time vs. both	10,017.17 ***	0.000
	ind vs. both	82.650 ***	0.000

Note: a. In the LR joint significance test, the null hypothesis H0 for “time vs. both” is that the time fixed effect coefficients are jointly equal to 0, while the null hypothesis H0 for “ind vs. both” is that the individual fixed effect coefficients are jointly equal to 0. b. *** indicates significance at the 1% level.

. Based on the results, a spatial Durbin difference-in-differences (SDM-DID) model with both individual and time-fixed effects was chosen.

3.3. Parallel Trend Test

To verify whether “Civilized Cities” and non-“Civilized Cities” exhibited parallel trends in environmental governance performance before the policy intervention, this study follows the method of Beck, Levine, and Levkov (2010) [30], employing an event analysis method for the parallel trend test. Based on the multi-period DID model in Equation (2), the regression model is set as follows:

$$\begin{gathered} {EGP}_{it}=\mathsf{\alpha }+{\sum }_{T=1}^{14}{\beta }_{-T}{DID}_{i,t-T}+\beta {DID}_{it}+{\sum }_{T=1}^{12}{\beta }_{+T}{DID}_{i,t+T} \\ \begin{array}{c}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+{{\theta }_{1}lnPGDP}_{it}+{{\theta }_{2}AGGLO}_{it}+{{\theta }_{3}lnFISC}_{it}\\ +{{\theta }_{4}lnTECH}_{it}+{{\theta }_{5}lnFDI}_{it}+{\mu }_i+{\gamma }_t+{\epsilon }_{it}\end{array} \end{gathered}$$

(4)

Here, ${\mathsf{\beta }}_{-\mathrm{T}}$ represents the impact T years before policy intervention, $\mathsf{\beta }$ represents the year of policy impact, and ${\mathsf{\beta }}_{+\mathrm{T}}$ represents the impact T years after policy intervention. When the year is the period impacted by the “Civilized City” selection policy, ${\mathrm{D}\mathrm{I}\mathrm{D}}_{\mathrm{i}\mathrm{t}}$ takes the value of 1; otherwise, it takes the value of 0. The meanings of the other parameters are consistent with those in Formula (3). The results shown in

Table 5. Results of the parallel trend test.

	Coefficient		Coefficient		Coefficient
t − 14	−0.001	t − 5	−0.000	t + 4	0.008 *
	(−0.114)		(−0.018)		(1.944)
t − 13	−0.001	t − 4	−0.000	t + 5	0.011 **
	(−0.150)		(−0.044)		(2.489)
t − 12	−0.001	t − 3	−0.002	t + 6	0.018 ***
	(−0.355)		(−0.471)		(3.772)
t − 11	−0.002	t − 2	−0.000	t + 7	0.021 ***
	(−0.380)		(−0.009)		(4.121)
t − 10	−0.003	t − 1	−0.000	t + 8	0.025 ***
	(−0.725)		(−0.005)		(4.217)
t − 9	−0.002	t	0.003	t + 9	0.044 ***
	(−0.574)		(0.747)		(4.427)
t − 8	−0.001	t + 1	0.006	t + 10	0.052 ***
	(−0.168)		(1.463)		(4.522)
t − 7	0.000	t + 2	0.006	t + 11	0.051 ***
	(0.049)		(1.514)		(4.256)
t − 6	−0.000	t + 3	0.005	t + 12	0.063 ***
	(−0.101)		(1.149)		(3.558)

Note: a. The values in parentheses represent the t-statistics corresponding to each coefficient. b. ***, **, * indicate significance at the 1%, 5%, and 10% levels, respectively.

indicate that the estimated coefficients for all periods before the policy intervention were not statistically significant. This suggests that there are no significant differences in the environmental governance performance trends between the treatment and control groups prior to the policy, satisfying the parallel trend assumption.

3.4. Empirical Results

Table 6. Baseline regression.

	DID Model		SDM-DID Model
	(1)	(2)	(3)	(4)
DID	0.008 ***	0.007 ***	0.007 *	0.006 *
	(6.571)	(5.794)	(1.784)	(1.826)
lnPGDP		−0.009		0.011 *
		(−1.326)		(1.688)
AGGLO		0.004		0.0004
		(0.632)		(0.092)
lnFISC		−0.015 ***		−0.014 ***
		(−5.425)		(−2.953)
lnTECH		0.003 ***		0.002 ***
		(6.063)		(6.827)
lnFDI		−0.001 **		−0.001 ***
		(−2.343)		(−3.388)
W*DID			0.010 ***	0.008 ***
			(4.386)	(5.083)
W*lnPGDP				−0.039 ***
				(−3.824)
W*AGGLO				0.007
				(1.517)
W*lnFISC				0.002
				(0.864)
W*lnTECH				0.003 **
				(2.340)
W*lnFDI				0.002 ***
				(3.209)
ρ			0.126 ***	0.115 ***
			(3.503)	(3.888)
City FE	YES	YES	YES	YES
Year FE	YES	YES	YES	YES

Note: a. Variables starting with W represent the spatial lag terms of the corresponding explanatory variables. b. The values in parentheses are t-statistics, with the corresponding standard errors calculated using Driscoll–Kraay robust standard errors. c. ***, **, and * indicate significance levels at 1%, 5%, and 10%, respectively.

presents the regression parameter estimation results of the DID model without considering spatial effects, as well as the regression parameter estimation results of the SDM-DID model based on a 0–1 adjacency spatial weight matrix.

For the core explanatory variable, without considering spatial effects and only including the “Civilized City” selection as an explanatory variable in the regression model, results in Column (1) of Table 6 show that the estimated coefficient of the “Civilized City” selection variable is 0.008, which is significant at the 1% level. After including a set of control variables, the results in Column (2) indicate that the sign of the estimated coefficient for the “Civilized City” selection variable remains unchanged and remains significantly positive at the 1% level. This result suggests that the creation of “Civilized Cities” promotes environmental governance performance. This finding can be explained by the following reasons: The designation of “Civilized City” serves as a key performance indicator (KPI) in the central government’s evaluation system for local administrations, with direct implications for official promotion. Consequently, municipal governments mobilize concentrated resources to enhance environmental governance performance—both to secure this prestigious honor and to advance their political careers. Moreover, certain evaluation criteria (e.g., major pollution incidents, public satisfaction ratings) carry veto power, compelling the government to intensify environmental governance efforts.

After accounting for spatial effects, the regression results based on the SDM-DID model with a 0–1 adjacency spatial weight matrix show that under both conditions—without and with control variables—the coefficients for the “Civilized City” selection variable are positive, with p-values less than 0.1. This indicates that the “Civilized City” selection activity has a profound and positive effect on promoting environmental governance performance. In Column (4), which includes other control variables, the regression coefficient for the “Civilized City” selection policy is 0.006, suggesting that successfully creating a “Civilized City” can improve a city’s environmental governance performance by 0.006 units. This aligns with the findings of Lu et al. (2020), which indicate that the “Civilized City” selection activity effectively drives urban environmental governance [19]. This finding can be explained by the following reasons. According to the learning effect, when a city is successfully awarded the title of “Civilized City”, its environmental governance practices (such as waste sorting and smart environmental protection systems) are often adopted by neighboring cities. Thus, through imitating neighbors’ governance practices, the city is more likely to improve its environmental governance performance. Furthermore, when neighboring cities demonstrate environmental improvements, local residents tend to exert public pressure on their own municipal government to follow suit, which makes the government pay more attention to environmental governance performance improvements.

3.5. Regression Results of Spatial Lag Variables

The SDM-DID model results in Table 6 show the spatial lag effects of the various variables. First, the spatial autoregressive coefficients ρ in the models without and with the control variables are 0.126 and 0.115, respectively, and both are significant at the 1% level. This suggests that a city’s environmental governance performance is often positively influenced by that of neighboring cities. These findings highlight the strong spatial spillover effects of environmental governance performance across regions.

In Columns (3) and (4), the spatial lag coefficients for the “Civilized City” variable are 0.010 and 0.008, respectively, and both are significant at the 1% level. It indicates that the “Civilized City” policy not only improves local environmental governance performance but also has a significant driving impact on neighboring areas. When neighboring areas undertake the “Civilized City” initiative, local governments are likely to learn from their urban management practices, further strengthening environmental governance.

For the spatial lag of the control variables, the spatial lag coefficient of lnPGDP is −0.039, which is significant at the 1% level. It indicates that the economic growth of neighbors negatively influences local environmental governance. The spatial lag coefficient of lnTECH is 0.003 and significant at the 5% level, which suggests that neighbors’ green technology development can also benefit local environmental governance through technological spillover effects. As for lnFDI, the coefficient is 0.002, which is significant at the 1% level. It suggests that the spillover effects of environmentally friendly management and technologies brought by neighbors’ foreign direct investment can also benefit the local city.

3.6. Spatial Effect Decomposition

According to Lesage and Pace (2009), when the spatial lag of the dependent variable is significantly different from zero, its estimated coefficient cannot unbiasedly reflect the marginal effect of explanatory variables on the dependent variable [32]. Therefore, this paper further decomposes the effects of explanatory variables on environmental governance performance into direct impacts, indirect impacts, and total impacts for analysis. The spatial decomposition results for each explanatory variable at the full-sample level are reported in

Table 7. Spatial effect decomposition results based on full sample.

	Direct Effects	Indirect Effects	Total Effects
DID	0.006 *	0.010 ***	0.016 ***
	(1.881)	(5.683)	(3.427)
lnPGDP	0.010	−0.042 ***	−0.032 ***
	(1.598)	(−3.847)	(−5.180)
AGGLO	0.001	0.008 *	0.009 **
	(0.139)	(1.671)	(1.984)
lnFISC	−0.014 ***	0.0004	−0.014 ***
	(−2.912)	(0.208)	(−3.287)
lnTECH	0.002 ***	0.004 **	0.006 ***
	(8.727)	(2.495)	(3.949)
lnFDI	−0.001 ***	0.002 ***	0.001 **
	(−3.250)	(3.208)	(2.385)

Note: a. The values in parentheses are t-statistics, with their corresponding standard errors being Driscoll–Kraay robust standard errors. b. ***, **, and * denote significance levels at 1%, 5%, and 10%, respectively.

.

The results suggest that the “Civilized City” selection has significant positive direct, indirect, and total effects on environmental governance performance, with estimated values of 0.006, 0.010, and 0.016, respectively. This suggests that a region’s environmental governance performance is positively influenced not only by its own “Civilized City” creation activities but also by the positive spillover effects from neighboring regions.

The “Civilized City” initiative, widely recognized and supported by society, incentivizes local governments to improve environmental governance, encourages citizen participation in urban management, and promotes technological advancements among enterprises for green production and pollution control. Consequently, it enhances urban environmental governance performance while creating strong benchmark incentives for neighboring regions, leading to spatial spillover effects.

3.7. Robustness Tests

This section verifies the validity of the empirical results by altering the spatial weight matrix and controlling for other policy effects.

3.7.1. Replacing the Spatial Weight Matrix

A geographically based spatial weight matrix is more suitable for exploring the spatial effects of the “Civilized City” policy on environmental governance. While the study used a 0–1 adjacency matrix initially, alternative weight matrices based on inverse geographical distance and k-nearest neighbors were constructed for robustness checks, following You and Lv (2018) [31,33]. The results are reported in

Table 8. Spatial effect decomposition results based on inverse geographical distance spatial weight matrix.

	Direct Effects	Indirect Effects	Total Effects
DID	0.006 *	0.028 ***	0.034 ***
	(1.730)	(4.778)	(6.169)
lnPGDP	0.005	−0.053 ***	−0.048 ***
	(1.005)	(−2.986)	(−3.483)
AGGLO	−0.001	0.031 ***	0.030 ***
	(−0.223)	(2.809)	(2.639)
lnFISC	−0.014 ***	−0.0004	−0.014 ***
	(−2.919)	(−0.118)	(−5.431)
lnTECH	0.002 ***	0.005	0.007 **
	(5.725)	(1.488)	(2.132)
lnFDI	−0.001 **	0.002 *	0.001
	(−2.564)	(1.714)	(1.276)

Note: a. The values in parentheses are t-statistics, with their corresponding standard errors being Driscoll–Kraay robust standard errors. b. ***, **, and * denote significance levels at 1%, 5%, and 10%, respectively.

and

Table 9. Spatial effect decomposition results based on the k-Nearest Neighbor spatial weight matrix.

	Direct Effects	Indirect Effects	Total Effects
DID	0.006 *	0.014 ***	0.020 ***
	(1.794)	(6.940)	(6.057)
lnPGDP	0.004	−0.029 ***	−0.025 ***
	(0.892)	(−3.273)	(−4.729)
AGGLO	−0.0000	0.014 **	0.014 **
	(−0.013)	(2.165)	(2.057)
lnFISC	−0.015 ***	0.005	−0.010 ***
	(−2.864)	(1.441)	(−3.782)
lnTECH	0.002 ***	0.003 **	0.005 ***
	(7.829)	(2.115)	(3.694)
lnFDI	−0.001 ***	0.001	−0.0002
	(−2.722)	(0.784)	(−0.329)

Note: a. The values in parentheses are t-statistics, with their corresponding standard errors being Driscoll–Kraay robust standard errors. b. ***, **, and * denote significance levels at 1%, 5%, and 10%, respectively.

. These results indicate that the “Civilized City” policy’s direct, indirect, and total effects remain significantly positive under these alternative matrices, confirming the robustness of the empirical analysis.

3.7.2. Controlling for Other Policy Effects

Given that other policies implemented during the “Civilized City” initiative may have influenced the results, particularly pilot programs like “Low-Carbon Cities” and “Smart Cities” [19,34], these policies were considered for robustness checks. As there is overlap among cities affected by these policies, they might confound the effects of the “Civilized City” initiative.

Following Cao Qingfeng (2020) and Xu et al. (2024) [35,36], the baseline SDM-DID model (Equation (3)) was augmented with control variables for these policies. If the policy effect of the “Civilized City” initiative becomes insignificant after adding these variables, it suggests that the observed effects are due to other policies. However, if the policy effect remains significantly positive, the results will be robust. Table 5, Table 6, Table 7, Table 8 and Table 9 show that the conclusions are robust.

$$\begin{gathered} {EGP}_{it}=\rho {\sum }_{j=1}^n{w}_{ij}{EGP}_{jt}+{\alpha }_1{DID}_{it}+{\gamma }_1{DT}_{it}+{\gamma }_2{ZH}_{it} \\ \begin{array}{c}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em} +{{\alpha }_{2}lnPGDP}_{it}+{{\alpha }_{3}lnAGGLO}_{it}+{{\alpha }_{4}lnFISC}_{it}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em} +{{\alpha }_{5}lnTECH}_{it}+{{\alpha }_{6}lnFDI}_{it}+{\beta }_1{\sum }_{j=1}^n{w}_{ij}{DID}_{jt}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em} +{\delta }_1{\sum }_{j=1}^n{w}_{ij}{DT}_{jt}+{\delta }_2{\sum }_{j=1}^n{w}_{ij}{ZH}_{jt}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em} +{\beta }_2{\sum }_{j=1}^n{w}_{ij}{lnPGDP}_{jt}+{\beta }_3{\sum }_{j=1}^n{w}_{ij}{lnAGGLO}_{jt}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em} +{\beta }_4{\sum }_{j=1}^n{w}_{ij}{lnFISC}_{jt}+{\beta }_5{\sum }_{j=1}^n{w}_{ij}{lnTECH}_{jt}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em} +{\beta }_6{\sum }_{j=1}^n{w}_{ij}{lnFDI}_{jt}+u_i+{\gamma }_t+{\epsilon }_{it}\end{array} \end{gathered}$$

(5)

In this case, ${\mathrm{D}\mathrm{T}}_{\mathrm{i}\mathrm{t}}$, ${\mathrm{Z}\mathrm{H}}_{\mathrm{i}\mathrm{t}}$ represent the “Low Carbon City” policy and the “Smart City” policy, respectively. If city $\mathrm{i}$ is designated as a low-carbon city pilot in year $\mathrm{t}$, then ${\mathrm{D}\mathrm{T}}_{\mathrm{i}\mathrm{t}}=1$ takes the value of 1 for city $\mathrm{i}$ in year $\mathrm{t}$ and subsequent years; otherwise, it takes the value of 0. Similarly, if city $\mathrm{i}$ is designated as a smart city pilot in year $\mathrm{t}$, then ${\mathrm{Z}\mathrm{H}}_{\mathrm{i}\mathrm{t}}=1$ takes the value of 1 for city $\mathrm{i}$ in year $\mathrm{t}$ and subsequent years; otherwise, it takes the value of 0. The meanings of other parameters are consistent with those in Formula (5). The result can be observed in

Table 10. Excluding the influence of other policies.

	Direct Effects	Indirect Effects	Total Effects
DID	0.005 *	0.009 ***	0.014 ***
	(1.661)	(5.188)	(3.253)
DT	0.009 ***	−0.005 ***	0.004
	(4.585)	(−2.976)	(1.375)
ZH	0.002 **	0.003	0.005 *
	(2.056)	(1.434)	(1.817)
lnPGDP	0.011 *	−0.044 ***	−0.033 ***
	(1.819)	(−4.055)	(−4.850)
AGGLO	0.001	0.005	0.006
	(0.211)	(1.043)	(1.200)
lnFISC	−0.015 ***	0.001	−0.014 ***
	(−3.171)	(0.553)	(−3.378)
lnTECH	0.002 ***	0.004 **	0.006 ***
	(8.617)	(2.455)	(3.859)
lnFDI	−0.001 ***	0.002 ***	0.001 **
	(−3.267)	(3.017)	(2.166)

Note: a. The values in parentheses are t-statistics, with their corresponding standard errors being Driscoll–Kraay’s robust standard errors. b. ***, **, and * denote significance levels at 1%, 5%, and 10%, respectively.

.

4. Conclusions and Policy Recommendations

We use the “Civilized City” evaluation as a quasi-natural experiment, selecting panel data from 272 prefecture-level cities in China from 2003 to 2020. Based on the Spatial Durbin Model (SDM) and Difference-in-Differences (DID) approach, it empirically analyzes the influence of the “Civilized City” evaluation on environmental governance performance from both overall and regional perspectives.

The main findings of this study are as follows. First, China’s environmental governance performance demonstrates significant positive spatial dependence. During the 2003–2020 period, the Global Moran’s I index for regional environmental governance performance consistently showed statistically significant positive values, revealing strong spatial clustering patterns among cities. The environmental governance performance of cities is not independent but is influenced by the governance situations of neighboring regions. Therefore, spatial factors must be incorporated into environmental governance study, as inter-regional spillover effects cannot be overlooked.

Second, the policy for the selection of “civilized cities” can effectively contribute to urban environmental governance. Overall, the campaign to create civilized cities, as a typical incentive mechanism, can play a positive role in environmental governance. It can not only improve the environmental governance performance of the local area but also generate positive spatial spillover effects on neighboring areas, bringing environmental benefits to them. The National Civilized City is the highest honor reflecting the effectiveness of urban comprehensive governance. This policy can effectively motivate local governments. To meet the environmental requirements, local governments will strengthen environmental regulation, which in turn will promote technological progress in enterprises. Moreover, the campaign to create civilized cities is a participatory activity involving everyone. It will attract more public participation, thereby assisting urban environmental governance. Furthermore, the honor of being a “Civilized City” will create a powerful benchmark incentive, encouraging imitation and competition among neighboring areas, thus generating spatial spillover effects of this positive impact. It is worth mentioning that even after excluding the influence of similar policies such as “Smart City” and “Low-Carbon City,” the environmental governance benefits of the civilized city selection remain significant.

Based on these findings, and considering China’s specific context, this study proposes several policy recommendations. First, strengthen cross-regional cooperation for joint governance to achieve regional win-win outcomes. The study shows that environmental governance performance is significantly spatially dependent, indicating that a region’s environmental governance outcomes are closely related to its geographical location and the situation in neighboring cities. Therefore, relying solely on local conditions for environmental governance is often insufficient, and it is more effective to adopt a holistic approach and collaborate with other regions to promote environmental governance.

Second, improve the “Civilized City” evaluation mechanism to better leverage the policy’s effects. This study confirms that the “Civilized City” evaluation policy is effective in enhancing urban environmental governance, so it can be continuously optimized and developed. To win the battle against pollution and environmental degradation, China can further establish an environmental governance leadership responsibility system that balances both accountability and incentives, emphasizing responsibility while also harnessing the power of honor.

There are some limitations of our study. We only pay attention to the first-order neighborhood effect but ignore the second- or third-order neighborhood effects, especially given that the Durbin model inherently captures more diffusion layers. In future studies, we can further clarify how exactly a change in the “Civilized City” status of a city affects the environmental governance index of more distant cities.