Peer effects of environmental regulation on sulfur dioxide emission intensity: Empirical evidence from China

Peer effect and decision-making

Peer effect theory is well-known and has been widely considered by scholars in academia, which has defined peer effect as the phenomenon in which decision-making individuals are influenced by other individuals in the process of behavioral decisions.^13,21 Conceptually, it is closely relayed with the limited rationality of decision-makers and the inherent uncertainty characteristic of the decision-making process. ²² In simple words, the peer effect is in essence a manifestation of the spillover effect that is produced by peer behaviors among decision-makers, which also emphasized reflecting the strategic interactions between economic agents. ²³ The academic world has conducted thorough discussions on the peer effect in the diversified fields, including entrepreneurship, ²⁴ education behavior, ²⁵ competition, ²⁶ individual decision-making,^27,28 corporate governance, ²⁹ R&D investment, ³⁰ corporate innovation,^31–33 etc. Nevertheless, there is still gap on whether the peer effect applies to government actions, especially whether the peer effect in environmental decision-making among governments exists is still lacking comprehensive consideration.

Undoubtedly, government initiatives and policy decisions among different authorities are not entirely independent; rather, they are inevitably interdependent. ³⁴ Apparently, extensive studies have explored the government's actions and provide substantial evidence for the phenomenon of “race to the bottom,”^35–37 “race to the top,”^38,39 and “imitative learning,”^40,41 which has given some support to the peer effect of government actions. By contrast, the current works that investigate the interaction of government actions have ignored the possible peer effects in environmental policies decisions, especially the linkage and interactions of ERs. To our best knowledge, the literature related to the peer effect on ecological protection is mainly filed into two main streams. From one aspect, some scholars concentrated on the peer effects on corporate environmental protection and proposed several reasons for peer effects on environmental protection, such as the competition-based theory and the information cascades theory. ⁴² From another aspect, there is a preliminary consensus on peer effects on government environmental governance. For instance, Woods ⁴³ explicitly confirmed the existence of the policy and cost competition effect among a broader set of economic peers, meanwhile proposing that the relaxed regulatory standards may result in a race to the bottom to a large extent. Moreover, the similarity of political resources, the proximity of geographical location, and the ease of interregional associations are crucial considerations for green governance among local authorities. ⁴⁴ In this regard, Xu et al. ⁴⁵ pointed out that the central governments may consider the relative performance of peer regions as a baseline for assessing the policies’ intensity, resulting in similar behavioral decisions. In addition, despite continuous attention being paid to the peer effect theory in the issue of regional ER and confirmed an obvious peer effect among urban ER systems, ¹⁸ has overlooked to the connection between ER and pollutant governance within the framework of peer effect theory. Most existing literature focuses only on the peer effects regarding ER decisions as a single element, ¹⁸ which overlooks the potential influence of pollution emission control on ER decisions based on the peer effects theory, that is, the possible peer effects between ER and pollutant emissions.

ER and pollutant discharge

ER is widely considered a fundamental tool for managing pollutant discharge, which refers to the official laws, policies, and standards set by governments to manage and control the effect of human activities on the environment quality to achieve the goals of sustainable development. Indeed, it can be classified into three broad categories including command-and-control, market-based, and voluntary ER. Generally, the current research literature in this field primarily focuses on the following representative hypotheses to explain its possible effects on pollutant discharge, including the cost of compliance hypothesis, Porter hypothesis, green paradox, and pollution haven hypothesis. ¹⁰ First of all, in terms of the green paradox, environmental problems may be exacerbated to some extent by the signal of ER implementation, namely companies may increase the extraction of resources and increase pollution emissions before the ERs are implemented in order to avoid higher costs and restrictions. ⁴⁶ Apparently, Liu et al. ⁴⁷ confirmed the appearance of the green paradox in legal and supervised ER on energy consumption in China based on panel data from 1997 to 2015. Meanwhile, Tavares and Robaina ⁴⁸ explicitly proposed that the intervals and delays in the implementation of ER appear to have had a negative influence on environmental quality, supporting the green paradox in Europe. Secondly, the pollution haven and Porter hypothesis have been widely explored and verified in the existing literature. The former argues that strict ER may lead high-polluting industries to transfer to countries with relatively low ER levels,^17,49 which places an environmental burden on developing countries in particular.^50,51 For example, Nie et al. ⁵² examined the nexus between ER and the transfer of pollution-intensive industries, which confirmed the “pollution haven hypothesis” relying on the carbon trading pilot policies in China. The latter considered that the strict ER standards are conducive to promoting corporates’ innovation and technological progress, which can be regarded as the innovation compensation effect. In this regard, environmental quality can be efficiently improved by the behavior of companies seeking greener production methods and products. ⁵³ Lastly, based on the cost of compliance hypothesis, it is argued ER may cause additional costs of pollution control and production in the manufacturing industries, because of the distorting effect on resource allocation. That means stringent ERs may hinder innovation and technological progress in enterprises, as companies often opt to comply with minimum requirements rather than pursue more innovative strategies due to concerns about cost and risk. For instance, Becker et al. ⁵⁴ explored whether the distinct effect of ER may exist with the change in firm size, revealing that pollution abatement expenditures increase with the firm size. Meanwhile, Liu et al. ⁵⁵ found that legal and supervised ER have a distinct “cost effect” on energy consumption in China. Relevant literature has assessed its applicability for TFP and enterprises’ profit, the inhibitory effect on the company's profit and firm productivity has been proved by many scholars so far.^56–58

Against this background, prior literatures have widely investigated the nexus between ER and pollutant emissions, nevertheless, there are no consistent conclusions. On the one hand, the positive effects of ER on pollutant discharge are recognized. ⁵⁹ Especially, ER can mitigate pollution discharge through diversified channels such as industrial structure adjustment, ⁶⁰ technological innovations, ⁶¹ foreign direct investment, ⁶² industrial transformations, ⁶³ etc. On the other hand, some researchers argued that the nexus between ER and pollutant discharge may be more complex. According to the environmental Kuznets curve hypothesis, it is argued that there is an inverted U-shaped nexus between the economy and the environment. ⁶⁴ In this regard, most of the research has tried to explore the potential role of influences such as trade, urbanization, ⁶⁴ human capital, ⁶⁵ income inequality, ⁶⁶ government forces, etc. in the framework of economic development and ecological quality. For instance, Yin et al. ⁶⁷ investigated the moderating effect of ER and technical innovation on the environmental Kuznets curve, which revealed strict ER is contributed to bringing forward the inflection point of the CO₂ emission Kuznets curve. Moreover, some scholars also described that the nexus between ER and environmental pollution is inverted U-shaped.^68–70 Thereby, it is clear that most of the existing papers focus on the nexus between ER and pollution emissions, but few studies have explored the peer effect between them from the possible interaction between ER decisions in different governments.

ER and sulfur dioxide emissions

SO₂ is one of the primary air pollutants and a main focus of ER. The effectiveness of different countries’ ER policies in reducing SO₂ emissions has received widespread attention in the academic community. After in-depth exploration, some scholars have affirmed the effectiveness of ER policies in reducing sulfur dioxide emissions in their countries.^71,72 For instance, Hancevic ⁷³ indicated that the enactment of ERs regarding sulfur dioxide promotes the use of low-sulfur coal in coal-fired power generation, thereby reducing sulfur dioxide emissions. Similarly, after studying ER policies in Northern Europe, Korhonen et al. ⁷⁴ pointed out that ER and monitoring are effective means of reducing SO₂ discharge and improving environmental quality in the long term. In China, it has been proven that well-developed ER measures have significantly positive impacts on both the total amount and intensity of SO₂. ⁷⁵ This viewpoint is also confirmed by the research of Huang. ⁷⁶ However, some scholars indicated that the nexus between ER and SO₂ emissions is not simply linear but rather complex and nonlinear. Dong et al. ⁷⁷ suggested that the emission reduction effect of ER is related to the level of human capital, only when its level exceeds a certain threshold can ER effectively reduce SO₂ emissions. Moreover, the level of economic development is also a crucial threshold that affects the effectiveness of ER. ER only becomes effective when per capita GDP exceeds a certain level. ⁷⁸ It should be noted that there are also studies suggesting that ER does not significantly promote SO₂ emissions reduction. ⁷⁹ In summary, current literatures have not obtained a consistent conclusion, likely due to differences in factors such as research objects and study periods. Thus, it is necessary to carry out a detailed and comprehensive analysis of the peer effect between ER and SO₂ emission to draw a specific conclusion on this issue, so as to provide an empirical reference for targeted policy formulation in this area.

Methodology

In order to check whether the spatial econometric approach is suitable to evaluate the strong spatial interaction of ERI and air pollutant emissions in the empirical study, the spatial diagnostic test is carried out based on the research thought of Elhorst. ¹⁰¹ As shown in Table 1, the results of classic Lagrange's multiple (LM) show that models of pooled OLS, space-fixed effects, time-fixed effects, and time-space-fixed effects are all significant, which refused the null hypothesis without a spatially lagged item and spatial autocorrection error item. Meanwhile, in conjunction with the results of the likelihood ratio (LR) test, the spatial Durbin model (SDM) cannot be degenerated to either the spatial autoregressive (SAR) model or the spatial error model (SEM) individually, indicating that SDM is the most appropriate. Furthermore, the fixed effect and random effect are essential considerations in the SDM model. The Hausman statistic under the geographical inverse distance square matrix is < 0 (−21.99), revealing the fixed effect model is adopted. Simultaneously, the estimation of the joint test demonstrates that the separate space-fixed effect model and the time-fixed effect model fail to accurately reflect the relationships between the core variables, so the adoption of a double-fixed effect model is necessary. Taking into consideration the aforementioned points in model screening, the dual-fixed effect SDM model is employed to identify the peer effect of ERI on Iso₂, and the model equation is specified in equation (1).

I s o 2 i t = ρ W ⋅ I s o 2 i t + α 1 E R I i t + α 2 W ⋅ E R I i t + θ 1 X i t + θ 2 W ⋅ X i t + μ i + ν i + ε i t

(1)

OPEN IN VIEWER

Table 1.

Results of spatial diagnostic tests.

Test statistics	Panel OLS	Space-fixed effect	Time-fixed effect	Time-space fixed effect
LM lag	151.147***	310.909***	67.047***	116.334***
Robust LM lag	48.803***	111.529***	24.890***	0.930
LM error	110.087***	247.042***	47.205***	124.309***
Robust LM error	7.743***	47.661***	5.048**	8.902***
LR test	LR spatial lag	235.150***	LR spatial error	279.780***
Joint test	Time-fixed	4248.010***	Space-fixed	235.320***

Note: LM: Lagrange’s multiple; OLS: ordinary least squared; LR: likelihood ratio. *** p < 0.01, ** p < 0.05, * p < 0.1.

Diverging from existing literature that primarily examines the interplay of ERs across regions ¹⁸ (WERI on ERI), the core focus of the baseline regression in this study lies in investigating the impact of interregional ERI on pollution emissions (WERI on Iso₂). Additionally, we explore the peer effects of ERI under varying intensities of environmental regulatory measures. To differentiate the impact between cities with different levels of ERI, this study categorizes the sample cities into three levels based on the Jenks natural breaks method via ArcGIS software: high-intensity cities (denoted as “H”), medium-intensity cities (denoted as “M”), and low-intensity cities (denoted as “L”). Accordingly, the spatial lag variable WERI at t time period can be expressed as equation (2). Moreover, we further decompose equation (2) by utilizing dummy variables for the three categories of sample cities, resulting in spatial lag variables representing the influence of high-intensity cities (ERI^H_L, ERI^H_M, and ERI^H_H), medium-intensity cities (ERI^M_L, ERI^M_M, and ERI^M_H), and low-intensity cities (ERI^L_L, ERI^L_M, and ERI^L_H) on the three categories of cities, respectively. Subsequently, incorporating the decomposed WERI into the baseline regression equation (1), we obtain equation (3) as an elemental regression model for complementarity analysis. What's more, drawing upon the research framework of existing studies, we construct a mechanism test model, as presented in equation (4) to examine the potential mechanisms of peer effects of ERI and the impact of external event shocks.

W ⋅ E R I i t = E R I i t H + E R I i t M + E R I i t L = ( E R I H H + E R I M H + E R I L H ) + ( E R I H M + E R I M M + E R I L M ) + ( E R I H L + E R I M L + E R I L L )

(2)

I s o 2 i t = ρ W ⋅ I s o 2 i t + α 1 E R I i t + ( β 1 E R I H H + β 2 E R I M H + β 3 E R I L H ) + ( β 4 E R I H M + β 5 E R I M M + β 6 E R I L M ) + ( β 7 E R I H L + β 8 E R I M L + β 9 E R I L L ) + θ 1 X i t + θ 2 W X i t + μ i + ν i + ε i t

(3)

W ⋅ E R I i t = δ S i t + θ 1 X i t + θ 2 W ⋅ X i t + ε i t

(4)

Spatial weight matrices

Considering the long-distance dispersion of pollutants between cities can cause multiple pollutants to overlap in space and time to a certain extent, creating a serious regional pollution problem. This phenomenon increases the pressure for ER in a specific area or between regions. Based on the objective role of geographical distance between regions, we introduce the geographical inverse distance square matrix (W₁). And the geographical adjacent matrix (W₂) is constructed to conduct the robustness test. The detailed matrices are structured as follows.

W i j 1 = { 1 ( d i j ) 2 , i ≠ j 0 , i = j

(5)

W i j 2 = { 1 , if city i and city j are adjacent 0 , otherwise

(6)

Spatial autocorrelation analysis

Referring to the research, the spatial applicability test is first conducted to judge the existence of spatial correlations of Iso₂ and the suitability of the spatial econometric model. The special note is that Moran's I is the more common index of spatial autocorrelation in available literature, ¹⁰² whose values are distributed in [–1, 1]. When its value is larger than 0, indicating a positive spatial correlation and the objective variable shows a negative spatial correlation if its statistic value is less than 0, the absolute value reflects the degree of spatial difference. Besides, there is a random spatial correlation if its value equals zero. Moran's I index can be divided into global Moran's I and local Moran's I. The former simply shows whether there is clustering or dispersion in the whole space, and the latter can detect the range and location of special outliers or clusters. Moreover, Geary's C is also an important spatial proxy, which mainly focuses on the comparison between adjacent data points and is more sensitive to local spatial autocorrelation. Its value range is distributed in [0, 2]. In order to comprehensively investigate the spatial distribution pattern, we introduce both two spatial proxies. The specific calculation equations are shown below:

G l o b a l ( M o r a n ′ s I ) = ∑ i = 1 n ∑ j = 1 n W i j ( x i − x ¯ ) ( x j − x ¯ ) S 2 ∑ i = 1 n ∑ j = 1 n W i j

(7)

L o c a l ( M o r a n ′ s I ) = ( x i − x ¯ ) S 2 ∑ j = 1 n W i j ( x i − x ¯ )

(8)

Variable descriptions

Explained variable: sulfur dioxide emission intensity ( Iso ₂ ). Increased energy consumption is often accompanied by economic growth, which is one of the crucial factors contributing to increased air pollution. In this case, we use Iso₂ to represent the actual level of regional sulfur dioxide emissions, which is calculated by the share of total industrial SO₂ emissions to gross domestic production (GDP), and the values are logarithmically processed. Besides, the pollution discharge per unit of gross output can also eliminate the effect of size differences on industry emission levels, which is more accurate and reasonable.

Explanatory variable: ERI . Based on the current research of Li et al., ¹⁰³ this article constructs a systematic and comprehensive proxy to evaluate the ERI relying on the entropy method. That is, the specific and representative indexes included the treatment rate of centralized wastewater treatment plants (%), the harmless treatment rate of domestic garbage (%), and the comprehensive utilization rate of industrial solid waste (%).

Control variables. Based on the current studies,^18,104,105 the following variables are controlled to mitigate the possibility of bias in the estimated results: industrial structure ( IS ) is calculated by the sum of the output value of primary, secondary, and tertiary industries multiplied by 1, 2, 3, respectively. Population density ( PD ) equals the share of the total population at year-end to the total administrative area. External openness ( EO ) is described by the amount of foreign direct investment. Internal openness ( IO ) is expressed by the ratio of total retail sales of consumer goods to regional gross domestic production. Fiscal autonomy ( FA ) is calculated by the share of government fiscal revenue to government fiscal expenditure. The economic development level ( EL ) equals to the values of per capita GDP in logarithmic form.

Mechanism variables. Relying on the theoretical analysis, we tend to verify the potential influence of three mechanisms on the peer effect of ERI: learning mechanism, competition mechanism, and shock mechanism, namely the following mechanism variables are constructed. Firstly, in terms of the learning mechanism, it is expressed by internal learning behavior ( ILB ) and external learning behavior ( ELB ) comprehensively. The former is expressed by the dummy variable, namely the value is assigned to 1 if the ERI level of the city i in the current period is greater than the previous period, otherwise the value is 0. The latter is represented by the number of cities of the same ERI level as a city i in the same province. Secondly, in terms of the competitive mechanism, it is expressed by the fiscal competition behavior ( FCB ) and technological competition behavior ( TCB ) reasonably. The former is described by the ranking of the share of urban fiscal expenditure to fiscal revenue at the same ERI level, the latter is measured by total expenditure on science and technology in a city. Thirdly, as for the external shocks’ mechanism, we analyze it from the realistic characteristics of the city including infrastructure development level, digitalization degree, regional honors, and emission limits. In detail, these diversified perspectives are considered by the opening of the high-speed railway ( OHR ), “Broadband China” policy pilot ( BCP ), “National Civilized City” awards ( CTA ), and the implementation of special emission limits for air pollutants policy ( SELAP ). These shock mechanism variables are represented using dummy variables, where a value of 1 is assigned in the establishment year and subsequent years, and 0 otherwise.

Heterogeneity analysis variables. The marketization level is calculated by the Fan index and government intervention is represented by the share of fiscal expenditure to real domestic production. The officer characteristics are expressed by the length of the governor's tenure of office. The variable depictions and descriptive statistics are presented in Table 2, it can be viewed that the data of all core variables were within a reasonable value range, which provides a guarantee for the subsequent empirical processes.

OPEN IN VIEWER

Table 2.

Variable depictions and descriptive statistics.

Variables	Symbols	Obs.	Mean	SD	Min	Max
Explained variable
Sulfur dioxide emission intensity	Iso 2	3420	0.015	0.023	0.000	0.243
Explanatory variable
Environmental regulation intensity	ERI	3420	0.795	0.154	0.195	0.998
Control variables
Industrial structure	IS	3420	227.037	14.367	183.120	283.200
Population density	PD	3420	5.739	0.919	1.597	7.922
External openness	EO	3420	0.021	0.028	0.000	0.775
Internal openness	IO	3420	0.374	0.110	0.026	1.013
Fiscal autonomy	FA	3420	0.466	0.228	0.054	1.541
Economic development level	EL	3420	10.538	0.639	8.189	13.056

ERI: environmental regulation intensity; IS: industrial structure; PD: population density; EO: external openness; IO: internal openness; FA: Fiscal autonomy; EL: economic development level.

Data declaration

The panel dataset of 285 Chinese cities at the prefecture level and above from 2008 to 2019 is used in this work, and the original data is collected from the official statistical publication of the China Statistics Yearbook (2009–2020) and China City Statistics Yearbook (2009–2020). Other data sources include the Digital Finance Center of Peking University, the China Urban Construction Statistical Yearbook, the National Railway Administration, the Office of the Central Steering Committee on Spiritual Civilization Construction, the website of the National Ministry of Industry and Information, and prefectural government websites. Some missing data have been supplemented by manual queries and only a very small part of missing values was calculated by the interpolation approach.

Estimated results of benchmark regression

As for the spatial distribution pattern of the two study subjects, we choose Moran's I and Geary's C proxies to represent the characteristic of spatial agglomeration. As shown in Table 3, it is clear that Moran's I statistic values of Iso₂ and ERI are all greater than zero and are significant at the 1% confidence interval, reflecting the low–low and high–high agglomeration patterns in the sample period. Meanwhile, the values of Geary's C are all lower than one, ensuring positive spatial autocorrelations. Thereby, there is a significant spatial correlation between ERI and Iso₂, which provides the possibility for the establishment of peer effects.

OPEN IN VIEWER

Table 3.

The estimated results of spatial correlation test.

Year	Iso2		ERI		Year	Iso2		ERI
	Moran's I	Geary's C	Moran's I	Geary's C		Moran's I	Geary's C	Moran's I	Geary's C
	(1)	(2)	(3)	(4)		(5)	(6)	(7)	(8)
2008	0.145***	0.746***	0.286***	0.640***	2014	0.244***	0.659***	0.317***	0.642***
	(0.049)	(0.098)	(0.051)	(0.055)		(0.049)	(0.087)	(0.051)	(0.057)
2009	0.141***	0.746***	0.236***	0.696***	2015	0.224***	0.675***	0.269***	0.690***
	(0.048)	(0.101)	(0.051)	(0.057)		(0.049)	(0.091)	(0.051)	(0.058)
2010	0.166***	0.833**	0.283***	0.642***	2016	0.300***	0.674***	0.275***	0.696***
	(0.048)	(0.100)	(0.051)	(0.056)		(0.050)	(0.072)	(0.051)	(0.056)
2011	0.173***	0.803***	0.232***	0.717***	2017	0.199***	0.750***	0.314***	0.647***
	(0.050)	(0.079)	(0.051)	(0.056)		(0.050)	(0.076)	(0.051)	(0.055)
2012	0.236***	0.671***	0.226***	0.735***	2018	0.182***	0.746***	0.335***	0.625***
	(0.050)	(0.083)	(0.051)	(0.056)		(0.050)	(0.081)	(0.051)	(0.055)
2013	0.243***	0.661***	0.245***	0.702***	2019	0.125***	0.797***	0.308***	0.653***
	(0.049)	(0.086)	(0.051)	(0.056)		(0.050)	(0.083)	(0.051)	(0.056)

Note: ERI: environmental regulation intensity; Iso₂: sulfur dioxide emission intensity. *** p < 0.01, ** p < 0.05. Standard error in parentheses.

The result of baseline regression is reported in Table 4. In columns (1) and (2), the outcomes of uncontrolled and controlled univariate analyses confirm the presence of positive spatial autocorrelation for Iso₂, underscoring the significance of the collaborative control of multiple pollutants in the atmosphere and regional collaborative management. Meanwhile, the significant positive coefficients of WERI are ∼ 0.015 and significant at the 1% or 10% confidence interval, which verifies the positive peer effect of ERI on the Iso₂ discharge. In order to mitigate the issue of heteroscedasticity, we introduce robust standard errors to conduct the re-regressions, and the estimations are presented in columns (3) and (4). It is evident that the re-regression results after addressing heteroscedasticity remain highly consistent with the aforementioned results, reaffirming the presence of the positive peer effect of ERI on the Iso₂ emissions. This “beggar-thy-neighbor” phenomenon has also been confirmed in existing research, suggesting that strengthening ERs may lead to the migration of pollution-intensive industries and thus intensify environmental pollution in neighboring regions. ¹⁰⁶ The implementation of ERs introduces various restrictions on industrial production, targeting aspects like air and soil quality, waste emissions, and more, which inadvertently places a higher cost burden on enterprises. In response, industries tend to adjust their production behaviors and industrial structures over time. On the one hand, polluting and high-emission enterprises are incentivized to engage in technological innovation and shift their product mix to offset the increased production costs resulting from higher-priced factors. This reliance on innovation leads to enhanced productivity and a transition toward industrial decarbonization. On the other hand, the promulgation of a series of ER policies has led to a rise in survival costs for local industrial agglomerations, prompting enterprises to continually optimize their structures and consider relocating production outside the region. Consequently, traditional polluting industries tend to move to neighboring areas, exacerbating pollutant issues in these regions, and demonstrating the positive peer effect of ERI. In conclusion, the inhibitory effect and positive peer effects of ERI on Iso₂ are apparent, thus establishing Hypothesis 1.

OPEN IN VIEWER

Table 4.

The estimated results of peer effects.

Variables	(1)	(2)	(3)	(4)
WIso2	0.242***	0.222***	0.242***	0.222***
	(0.020)	(0.020)	(0.086)	(0.080)
ERI	−0.003*	−0.004**	−0.003	−0.004
	(0.002)	(0.002)	(0.003)	(0.003)
WERI	0.021***	0.015***	0.021**	0.015*
	(0.003)	(0.003)	(0.009)	(0.008)
Controls	No	Yes	No	Yes
City-fixed	Yes	Yes	Yes	Yes
Year-fixed	Yes	Yes	Yes	Yes
Observations	3420	3420	3420	3420
R 2	0.324	0.361	0.324	0.361
F-value	107.0	67.50	27.51	21.02
Cities	285	285	285	285

Note: ERI: environmental regulation intensity. ***p < 0.01, **p < 0.05, *p < 0.1. Standard error in parentheses. To alleviate the heteroscedasticity problem, the results of robust standard error of regional clustering heteroscedasticity are shown in columns (3) and (4).

Robustness test

To guarantee the validity of the empirical conclusion, we conduct diversified robustness tests including replacing explained variables, replacing spatial weight matrix, replacing samples, and replacing the empirical model. The detailed estimations of re-regressions are represented in Table 5. Firstly, substituting the original explained variable with SO₂ emission per capita (Pso₂), which both consider the pollutant emissions and population agglomeration and is widely used in current academia. Secondly, this paper introduces the geographical adjacent matrix (W₂) to verify the nexus between key research objects. Thirdly, the reality that municipalities and provincial capitals differ greatly from other cities in terms of the economic foundation, cultural heritage, advanced technology, resource allocation, and so on cannot be ignored, this paper substitutes sample interval to conduct re-regressions. Finally, considering the underlying advantages of the spatial lag X (SLX) model in more realistic estimations of local spillover effects,^107,108 this paper adopts it to further confirm the robustness of baseline results. Based on the re-regression results in Table 5, the direct mitigation effect and positive peer effect of ERI on Iso₂ emissions are both established evidently, verifying the credibility and reliability of baseline highlights.

OPEN IN VIEWER

Table 5.

The estimated results of the robustness test.

Variables	Replace explained variable	Replace weight matrix	Replace samples		Replace model
	Pso 2	Geographical adjacent matrix	Eliminate municipalities	Eliminate provincial capital	SLX
	(1)	(3)	(4)	(5)	(6)
WIso 2	0.232***	0.499***	0.219***	0.256***
	(0.022)	(0.026)	(0.020)	(0.023)
ERI	−39.900***	−47.342**	−0.005**	−0.004*	−0.004**
	(6.937)	(18.690)	(0.002)	(0.002)	(0.002)
WERI	31.319***	58.316*	0.015***	0.018***	0.015**
	(10.860)	(34.791)	(0.003)	(0.003)	(0.003)
Controls	Yes	Yes	Yes	Yes	Yes
City-fixed	Yes	Yes	Yes	Yes	Yes
Year-fixed	Yes	Yes	Yes	Yes	Yes
Observations	3420	3420	3372	3108	3420
Cities	285	285	281	259	285
F-value	83.31	82.39	68.38	61.91	63.02
R 2	0.411	0.401	0.367	0.363	0.336

Note: ERI: environmental regulation intensity. ***p < 0.01, **p < 0.05, *p < 0.1. Standard error in parentheses.

Potential endogeneity solving

Considering the potential endogeneity bias, this subsection conducts the generalized spatial two-stage least squares (GS2SLS) and system generalized method of moments (GMM) regression to address the endogeneity concerns. On the one hand, the GS2SLS method can estimate robust results even in the case of heteroscedasticity and large sample size issues.^109,110 On the other hand, the GMM estimation does not need to presuppose the distribution of error terms in advance, ¹¹¹ which consists of two main aspects including the difference method and Hansen estimation. The former is used to test the suitability of the model, the latter is adopted to test whether the instrumental variable is over-identified. ¹¹² The results of endogeneity solving are shown in Table 6.

OPEN IN VIEWER

Table 6.

Estimated results of endogeneity test.

Variable	GS2SLS	System GMM
	(1)	(2)	(3)
ERI	−0.004***	−0.009**	−0.069***
	(0.002)	(0.004)	(0.004)
WERI	0.015**	0.062***	0.103***
	(0.003)	(0.007)	(0.004)
City-fixed	Yes	Yes	Yes
Year-fixed	Yes	Yes	Yes
Controls	Yes	Yes	Yes
Observations	3420	3135	3135
R 2	0.825	–	–
AR (1)	–	0.001	0.000
AR (2)	–	0.483	0.343
Hansen test	–	0.108	0.151

Note: ERI: environmental regulation intensity; GS2SLS: generalized spatial two-stage least squares; GMM: generalized method of moments. *** p < 0.01, ** p < 0.05. Standard error in parentheses.

As shown in column (1), the coefficient of ERI and WERI is obviously negative, which confirms our above finding that the environmental effectively hinders the local Iso₂, and the “beggar-thy-neighbor” dilemma is also supported. It is undoubted that introducing the lagged terms of the dependent and endogenous variables can address the endogeneity problem to a large extent. ¹¹³ First, the lagged four and five-period variables of the dependent variable and spatially lagged terms of dependent variables, and lagged three and four-period variables of independent variables and spatially lagged terms of independent variables are used as instrumental variables in the regression model. As shown in column (2), the Hansen test shows the p-value is more than 0.10, indicating the instrumental variable is an exogenous variable, meanwhile, the result of the AR test reflects that a second-order autocorrelation does not exist. Thereby, the result of endogeneity solving is consistent with benchmark regression, confirming the obtained findings are robust. Secondly, considering the ventilation coefficient does not directly affect Iso₂ but is highly related to ERI, we further introduce it as the instrumental variable for ER. ¹¹⁴ Ventilation coefficient is calculated as the product of mixing height and wind speed and suits the principles of the instrumental variable. The region with poorer pollution dispersion is more likely to be confined to the air pollution issue, the local governments tend to pay more attention to ecological issues and strengthen their regulatory efforts. In this regard, based on the study of Chen et al., ¹¹⁵ we introduce the interaction term of ventilation coefficient with year as the instrumental variable. Similarly, the result in column (3) confirms the inhibitory effect and positive spatial overflow effect of ERI, thus the highlights in benchmark regression are valid.

Heterogeneity test

Heterogeneity of geographical region

Considering that cities in different areas have differences in geographical location, natural conditions, economic conditions, and policy tendencies, the formulation methods and implementation intensity of ER policies in different regions are also different. Therefore, the sample is classified into eastern, central, and western cities to deeply explore the impact of ERI on Iso₂. According to Table 7, the increase of ERI in the western region can obviously reduce the Iso₂, and there is a distinct aggravating effect of ERI on Iso₂ in the neighboring areas in the central and western regions. Possible reasons lie in the following aspects: The eastern region boasts a relatively complete industrial, technological, and talent base, and its industrial structure has gradually become dominated by high-end and sophisticated industries with low energy consumption and minimal pollutant emissions. ¹¹⁶ Therefore, ERI cannot have a great impact on them. The development of the western region is relatively backward, the agriculture, animal husbandry, and traditional industries accounting for a relatively high proportion. Thus, the improvement of ERI can facilitate the transformation and upgrading of traditional industries, so as to reduce the intensity of Iso₂ in these regions. Although some studies have come to similar conclusions,^117,118 they often overlook the heterogeneity of spatial spillover effects in different regions and our research complements this. From our results in Table 6, the strengthening of ERI in the central and western regions can also prompt some industries with high pollution and high energy consumption to move to the surrounding areas, thus posing challenges to their environmental quality.

OPEN IN VIEWER

Table 7.

The estimated results of heterogeneity analysis.

Variables	Geographical characteristic			Marketization degree				Official characteristic
	Geographical region			Overall market		Govern market		Governor's tenure
	Eastern	Central	Western	High-level	Low-level	High-level	Low-level	Long-term	Short-term
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
WIso2	0.156***	0.476***	0.023	0.276***	0.171***	0.125***	0.310***	0.102***	0.467***
	(0.048)	(0.034)	(0.038)	(0.032)	(0.028)	(0.023)	(0.030)	(0.021)	(0.037)
ERI	−0.002	−0.002	−0.009*	−0.006**	−0.004	−0.000	−0.007**	−0.002	−0.006**
	(0.002)	(0.003)	(0.005)	(0.003)	(0.003)	(0.002)	(0.003)	(0.003)	(0.003)
WERI	0.001	0.017***	0.020***	0.005	0.023***	0.008**	0.019***	0.004	0.020***
	(0.003)	(0.004)	(0.008)	(0.004)	(0.004)	(0.004)	(0.004)	(0.004)	(0.004)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
City-fixed	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year-fixed	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	1212	1308	900	1692	1728	1296	2124	1224	2196
Cities	101	109	75	141	144	108	177	102	183
R 2	0.616	0.562	0.270	0.360	0.382	0.406	0.396	0.423	0.376
F-value	66.98	57.81	11.37	33.04	36.99	30.53	48.36	30.92	46.01

Note: ERI: environmental regulation intensity. ***p < 0.01, **p < 0.05, *p < 0.1. Standard error in parentheses.

Complementarity test

As analyzed above, the existence of positive peer effects of ERI on Iso₂ in China is confirmed, according to the research of Xu et al., ¹⁸ this paper tends to investigate these peer effects among cities with different regulation levels. What is different is that we do not focus on the peer effect of ERI between different cities, but further consider the peer effect of ERI on Iso₂ in cities with different regulation levels. Based on the constructed model of equation (3), the results of complementarity tests under W₁ are represented in Table 8. According to the coefficients in column (1), it is undoubted that the peer effects of ERI in different cities are relatively distinct with different ERI levels. From the results of high-intensity cities on others, the parameters of ER^H_M and ER^H_H are both obviously positive, and the peer effect is more pronounced with medium-level cities. Besides, the coefficients of ER^M_H, ER^M_M, and ER^M_L are all positive at 1% or 10% confidence intervals, suggesting the peer effects of medium-level cities on cities with different intensity levels are all significant, and the peer effects are in descending order of low-level cities, high-level cities, and medium-level cities. Additionally, in terms of the peer effects of low-intensity cities on other cities, it is evident that it exerts a clear peer effect among low-level cities and medium-level cities. By contrast, the coefficients of ER^H_L and ER^L_H are both positive but not significant, namely there is no peer effect between high-regulation intensity cities and low-regulation intensity cities.

OPEN IN VIEWER

Table 8.

The estimated results of the complementarity test.

Variables	Geographical inverse distance square matrix (W1)
	Coefficients	Std. Err.
	(1)	(2)
WIso2	0.316***	0.018
ERI	−0.003	0.002
High–low
ERIHL	0.005	0.007
ERIHM	0.014***	0.005
ERIML	0.060***	0.006
Low–high
ERILM	0.035***	0.007
ERILH	0.018	0.011
ERIMH	0.008*	0.005
Same intensity
ERILL	0.015**	0.007
ERIMM	0.023***	0.004
ERIHH	0.013***	0.004
Controls	Yes	Yes
Observations	3420	3420
Cities	285	285
R2	0.332	0.332

Note: ERI: environmental regulation intensity. ***p < 0.01, **p < 0.05, *p < 0.1. “Std.Err.” is the abbreviation for standard error.

Taken together, there is a combination of “top-down” and “bottom-up” environmental regulatory interaction between cities in China, and the association of ER strategies between regions is more sensitive. Cities with high-level of ERI have a distinct positive radiating effect on cities with low regulation levels, namely high-level regulation cities are more likely to become objects of imitation and learning for other cities, helping them to achieve optimal policy regulatory effects. Moreover, cities with low-level of ERI also pose a distinct positive spillover influence on cities with high regulation levels, that is, the areas with high ERI may use the areas with lower regulation levels as the bottom line for environmental management and engage in strategic competition and differentiated competitive behavior to avoid falling into the same environmental pollution dilemma. However, the interaction between cities with large environmental regulation gaps is insignificant, this phenomenon relies on the fact that these regions are generally farther apart spatially, making it difficult to generate strategic interactions and complementary behaviors. Some existing papers have also confirmed the similar peer effects of ERI, but most of them only study the peer effects of environmental policies in different regions from the perspective of government decision-making, ignoring the peer effects of ERI on pollution reduction, ⁴⁴ that is, what is the peer effect of environmental decisions in the surrounding area on local emissions. Therefore, compared to previous works on the peer effect of ERI on pollution reduction, the conclusions drawn in this article offer more comprehensive and in-depth empirical support for this study.

Mechanism of learning effect

Based on equation (4), the learning mechanism in the peer effect of ERI is explored (Table 9). At the level of the sample cities as a whole, the regression coefficients for internal and external learning mechanisms are both significantly positive, indicating that self-improving behaviors (i.e. self-learning, self-correcting, and self-accumulating behaviors) and interactive behaviors (i.e. imitation behavior, demonstrative motivation, and radiative driving) in the process of regional environmental management are conducive to strengthening the peer effect of ERI. Especially, the role of the internal learning mechanism in regional environmental protection is more pronounced, revealing it is more effective to continually optimize regulatory tools from own past governance experience than to imitate external strategic behaviors.

OPEN IN VIEWER

Table 9.

Mechanism analysis of learning effect.

Variables	Learning mechanism
	All sample	Low	Middle	High	All sample	Low	Middle	High
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
ILB	0.008***	0.008	0.007*	0.009***
	(0.002)	(0.008)	(0.004)	(0.003)
ELB					0.003***	−0.025**	−0.000	0.004***
					(0.001)	(0.010)	(0.003)	(0.001)
Control	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	3420	480	1188	1752	3420	480	1188	1752
Cities	285	40	99	146	285	40	99	146
R 2	0.432	0.144	0.511	0.389	0.450	0.141	0.508	0.438

Note: ILB: internal learning behavior; ELB: external learning behavior. ***p < 0.01, **p < 0.05, *p < 0.1. Standard error in parentheses.

Furthermore, the influence of the learning mechanism on ERI exhibits distinctive patterns across cities with varying degrees of regulation intensity. Firstly, considering the internal learning mechanism, the promotion impact of internal learning behaviors on the peer effect of ERI manifests more prominently in high-intensity regulation cities. In contrast, a statistically significant correlation is lacking between internal learning behavior and low-intensity regulation cities. Regions with subdued environmental oversight often encounter deficiencies in pollution control experience and inadequate regulation frameworks. Moreover, their governmental priorities tend to favor economic efficiency over environmental preservation, this scenario contrasts with municipalities possessing elevated ERI levels. Secondly, within the context of the external learning mechanism, it becomes evident that among low-intensity regulation cities, external learning behaviors weaken the peer effects of ERI. Conversely, high-intensity regulation cities experience a pronounced positive external stimulatory effect. These empirical findings underscore that municipalities characterized by lower ERI need to enhance their investments in ecological development and environmental safeguarding. In contrast, high ERI cities exhibit a significant positive strategic interplay in their environmental management efforts, embodying a trend of reciprocal influence and underscoring a foundation for cooperative “neighbor and partner” strategies in practical contexts.

As evident from the above, the internal and external learning behaviors both can facilitate positive peer effects in the majority of cities in China, which is consistent with the research of Xu et al. ¹⁸ As such, the learning behavior of local governments may diverge based on the interplay of ERI levels within each locale, proving ERI measures in China are objectively not independent, especially in the high ERI cities.

Mechanism of competition effect

Taking the reality of competitive behaviors among local governments into consideration, the competition mechanism in the peer effect of ERI is also analyzed (Table 10). Different from the existing studies,^44,124 this paper mainly explored the impact of competition effect from the two aspects of fiscal competition and technology competition, which are the key to promoting the environmental governance work of a region, and also the main means of competition among different local governments.

OPEN IN VIEWER

Table 10.

Mechanism analysis of competition effect.

Variables	Competition mechanism
	All sample	Low	Middle	High	All sample	Low	Middle	High
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
FCB	0.001***	0.005***	0.001***	0.001***
	(0.000)	(0.002)	(0.000)	(0.000)
TCB					−0.000	−0.006	−0.002	0.002
					(0.002)	(0.009)	(0.004)	(0.003)
Control	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	3420	480	1188	1752	3420	480	1188	1752
Cities	285	40	99	146	285	40	99	146
R 2	0.455	0.158	0.507	0.377	0.429	0.146	0.507	0.390

Note: FCB: fiscal competition behavior; TCB: technological competition behavior. ***p < 0.01. Standard error in parentheses.

The analysis reveals that fiscal competition behaviors play a pivotal role in enhancing the peer effect of ERI. Across the entire samples and sub-samples categorized by varying ERIs, the coefficients consistently display significant positive values, particularly pronounced in cities with low ERI. Sufficient financial backing forms the foundational assurance for fostering the harmonized growth of both environmental and economic domains. The fiscal competition among local governments emerges as a compelling catalyst for strategic interactions and the phenomena of “race to the top” or “race to the bottom” among regions. This phenomenon prompts governmental authorities to continuously recalibrate their ERs, aligning them with local ecological contexts and economic development plans. This adaptability and adjustment hold the potential to augment the peer effect of ERI. In contrast, the relationship between technology competition behaviors and the peer effect of ERI appears less conclusive. Technological competition has the potential to stimulate innovation within a region, consequently driving the evolution of industries towards higher-end sectors characterized by diminished energy consumption. This transition, in turn, alleviates the strain on regional environmental management at its roots and mitigates the imperative for stringent ER policies.

Hence, this observation underscores that the fiscal competition avenues emerge as pivotal conduits for streamlined urban environmental governance, rather than technology competition. Taking into account the results mentioned above, it can be observed that intergovernmental strategic learning behaviors and competitive behaviors have both contributed to the peer effects of ERI to a certain extent, thereby affirming the validity of Hypothesis 2.

Mechanism of external shocks

The consensus is that the macropolicies and events may pose underlying shocks to the effectiveness of the ER process, which may alter the concrete trend and enforcement of ER measures. In this regard, we take characteristic exogenous events as external shocks to investigate their possible mechanism on the peer effect of ERI, and the relevant results are exhibited in Table 11. In terms of findings, the impact of external shocks is partially established.

OPEN IN VIEWER

Table 11.

Mechanism analysis of external shocks.

Variables	Explained variable: WERI
	All sample	Low	Middle	High	All sample	Low	Middle	High
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Panel A. The opening of high-speed railway and the “Broadband China” policy pilot
OHR	0.003	−0.038***	0.006	0.009**
	(0.004)	(0.013)	(0.006)	(0.004)
BCP					−0.018***	−0.048***	−0.034***	−0.005
					(0.004)	(0.015)	(0.007)	(0.005)
Control	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	3420	480	1188	1752	3420	480	1188	1752
Sample cities	285	40	99	146	285	40	99	146
R2	0.431	0.159	0.515	0.389	0.428	0.165	0.492	0.392
Panel B. “National Civilized City” awards and the implementation of special emission limits for air pollutants policy
CTA	−0.010**	−0.064***	−0.021**	−0.006
	(0.004)	(0.020)	(0.010)	(0.005)
SELAP					−0.001	0.164***	−0.011	−0.009*
					(0.005)	(0.044)	(0.012)	(0.006)
Control	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	3420	480	1188	1752	3420	480	1188	1752
Sample cities	285	40	99	146	285	40	99	146
R2	0.427	0.164	0.502	0.391	0.430	0.163	0.508	0.398

Note: OHR: opening of the high-speed railway; BCP: “Broadband China” policy pilot; CTA: “National Civilized City” awards; SELAP: special emission limits for air pollutants policy. ***p < 0.01, **p < 0.05, *p < 0.1. Standard error in parentheses.

Firstly, the parameters of the OHR are obviously positive in the high-level ERI cities, signifying that enhanced infrastructure development triggers significant labor and expertise migration. The ease of information exchange and collaborative environmental governance across regions significantly amplifies the peer effect of ERI. Furthermore, the extensive reach of high-speed railway networks bolsters market convergence, stimulating competition among cities and intensifying peer effects of ERI.

Conversely, the rise in digitization levels diminishes peer effects of ERI between regions, notably in the low ERI level cities as affirmed by the city-wide sample. In the digital era, digital technologies have empowered ecological governance, reshaping its landscape. Yang et al. ¹²⁵ suggested that the fusion of digital tools with ecological preservation enhances green technology innovation through information sharing and knowledge integration, ultimately reducing interregional environmental governance interactions by emphasizing local ER.

Furthermore, a significant negative correlation between National Civilized City awards and peer effects of ERI is observed, except in cities with high ERI levels. In terms of the characteristics of the “civilized city” selection mechanism, these long-term activities are subject to frequent reviews by higher authorities. As a result, regional enterprises prioritize green and sustainable development at the source under the profit maximization principle. In the context of a “civilized city” competition resembling a “promotion tournament,” local officials strive to enhance their ERs, aiming for superior environmental performance. ⁹⁴ As regional environmental quality improves, the propensity of local governments to engage in learning, imitation, and competitive behaviors towards other regions’ ERs may gradually wane.

In addition, official emission limit policies distinctly impact peer effects of ERI across cities with varying regulation intensities. For low ERI cities, the implementation of SELAP intensifies the pressure to manage pollutants while encouraging learning, imitation, and regulation exchange from higher-level cities. Conversely, high ERI cities experience the opposite trend. In conclusion, external shocks like high-speed railway openings, digital technology advancement, “civilized city” awards, and emission restriction policies significantly differentiate peer effects of ERI, particularly evident in cities with low levels of environmental regulations, thereby affirming the validity of Hypothesis 3.

Further analysis of the moderating effect

Given the potential for complex influences of macroexternal shocks and strategic policies on intergovernmental ERs and regional pollution emissions, this study employs a moderated model to introduce interaction terms between the aforementioned variables and WERI relying on existing research.^126,127 Thereby further examining their moderating effects on the peer effect of ERI on Iso₂. The results are represented in Table 12.

OPEN IN VIEWER

Table 12.

Moderating effect of external shocks.

Variables	Explained variable: Iso2
	(1)	(2)	(3)	(4)
Interact-OHR	0.005
	(0.004)
Interact-BCP		0.025***
		(0.005)
Interact- CTA			0.024***
			(0.006)
Interact-SELAP				−0.020**
				(0.009)
Controls	Yes	Yes	Yes	Yes
City-fixed	Yes	Yes	Yes	Yes
Year-fixed	Yes	Yes	Yes	Yes
Observations	3420	3420	3420	3420
R 2	0.361	0.367	0.365	0.363

Note: OHR: opening of the high-speed railway; BCP: “Broadband China” policy pilot; CTA: “National Civilized City” awards; SELAP: special emission limits for air pollutants policy. *** p < 0.01, ** p < 0.05. Standard error in parentheses.

It is obvious that external shocks exert a pronounced differentiated moderating effect on the interaction between ERI and Iso₂. The interaction term between WERI and the opening of high-speed railways (coded as Interact-OHR) is positive but is not significant, revealing that the enhancement of regional infrastructure does not significantly contribute to the manifestation of the positive peer effect of ERI on sulfur dioxide discharge. The interaction terms between WERI and the “Broadband China” policy pilot or “National Civilized City” awards (coded as Interact-BCP and Interact-CTA, respectively) are both significantly positive. From this, it can be inferred that as two national-level policy measures, the selection of “Civilized Cities” and the implementation of the “Broadband China” pilot program indirectly aggravate the occurrence of the “shifting pollution to neighboring regions” phenomenon in China's pollution management efforts to a certain extent. From one aspect, in order to meet the requirements of the selection and compete for the honors, some governments may adopt superficial and short-term regulation measures. These measures may not fundamentally address pollution issues and could potentially lead to the transfer or postponement of pollution to other regions or times. From another aspect, the increase in digitalization is a key factor contributing to the escalation of regional resource consumption and pollution emissions, meanwhile, the imbalance in regional development and unfair environmental governance drive certain areas to shift their pollution burden to other regions. By contrast, the interaction term between WERI and pollution limitation policies (coded as Interact-SELAP) is significantly negative at the 5% confidence interval, confirming its negative moderating effect on the positive peer effect of ERI. This provides evidence that the implementation of SELAP contributes to strengthening regional joint prevention and control mechanisms, enhancing the coordinated capacity for interregional pollution management.

Conclusion remarks

Can ER reduce the intensity of sulfur dioxide emissions? The answer to this question holds crucial significance and empirical implications for ecological civilization and green quality development in China and even globally. To address this, we utilize panel data from 285 prefecture-level cities in China to analyze the direct and spatial effects of ERI on Iso₂, and further investigate the source mechanism of peer effects in ERI. Additionally, we conduct a complementary test to explore the difference in peer effects in ERI among cities with different levels of ERI, meanwhile estimating the potential shocks of macroevents and strategic policies. The main findings are as follows:

Firstly, China's ER exhibits clear positive spatial autocorrelation. Cities with high (low) ERI are surrounded by cities with high (low) levels of regulatory oversight. From a local perspective, ER significantly reduces regional Iso₂, indicating the foundational role of ERI in improving the ecological environment. Simultaneously, from a spatial perspective, positive spatial spillover effects confirm the dilemma of China's “beggar-the-neighbor” phenomenon. These highlights remain robust after conducting endogeneity tests and various robustness tests, such as alternative variable specifications, spatial weight matrices, empirical samples, and economic models. Furthermore, heterogeneity analysis reveals that the impact of ERI is heterogeneous, with spatial spillover effects being more pronounced in western cities and cities with lower levels of marketization, weaker government intervention, and shorter terms of provincial governors. Secondly, the positive peer effects of ERI in China have been established, revealing the remarkable pattern of regulatory interaction, and coordinated ecological management among cities. Furthermore, under different levels of ERI, the degree of actions varies, confirming the existence of both “top-down” and “bottom-up” modes of environmental regulatory interaction. In particular, the interaction between cities with significant differences in ER is relatively inconspicuous. Thirdly, the roles of learning, competition, and external shocks in the peer effects of ER have been partially supported. Specifically, both internal and external learning behaviors contribute to the spatial correlation of ER across different regions, especially through self-learning, self-correction, and self-accumulation behaviors. Fiscal competition facilitates interaction among neighboring areas in terms of environmental regulatory strategies, rather than technological competition. Furthermore, various external shocks, such as the establishment of high-speed railway networks, advancements in digital level, the recognition of “Civilized Cities,” and policies restricting pollution emissions, exert diverse influences on the peer effects of ERI in the majority of Chinese cities (especially in lower-level urban areas). Simultaneously, these external shocks introduce distinct moderating effects on the interaction between ERs and sulfur dioxide emissions.

The empirical findings of this study contribute to the existing research in the following ways. Firstly, this study provides a new perspective on research related to peer effects. Existing research on the peer effect between different individuals of the same subject, such as the peer effects of ERI decisions among different regional governments. ¹⁸ In contrast, this study explores the peer effects of ERI on Iso₂ between different regions, extending the theory of peer effects to two objects that have a certain correlation between different individuals. Secondly, the empirical results of this article provide theoretical support for the interactive behavior of environmental regulatory decisions among different countries, especially among local governments in developing countries. The findings regarding learning effects and competition effects reveal the interactive relationship of ER decisions between different local governments, while exogenous shocks further confirm that macropolicies and events are key influencing factors in the decision-making interaction between governments. Additionally, heterogeneity tests provide a theoretical basis for the formulation of differentiated policies across different regions.

Policy recommendations

Based on the obtained critical conclusions, the following suggestions are put forward to facilitate the collaborative and sustainable development of the environment and economy.

Firstly, a sound regional cooperation mechanism for urban environmental pollution control should be established to promote high-quality, coordinated, and sustainable development of the regional environment. On the one hand, communication and collaboration mechanisms for environmental governance should be established among local governments to avoid unilateral actions and address the fundamental issues of fragmented governance, policy diversity, and lack of coordination among local governments. On the other hand, the national government should establish a robust environmental management and regulatory system that promotes a collaborative mechanism for environmental pollution control through shared responsibilities, information sharing, and coordinated prevention efforts across regions, thereby enabling collective governance and mutual development.

Secondly, the government should implement tailored ER policies, and establish a differentiated mechanism for the implementation of environmental policies to promote the coordinated development of green and high-quality regions. On the one hand, regional governments should formulate pollution control and environmental protection standards that align with their own development characteristics and practical needs. It is crucial to avoid blindly following the trend to increase or reduce the ER intensity with other cities. For developed countries or regions, governments should actively utilize advanced technologies for green transformation, strengthening their leading and exemplary role in achieving coordinated development between the economy and the environment. They should also enhance technological assistance and financial support while respecting the developmental needs of developing nations. It is imperative to avoid indiscriminately designating underdeveloped regions as mere havens for pollution, in order to achieve sustainable global development. For developing countries or regions, governments should actively seek to learn experiences and advanced technologies in environmental governance from developed countries and increase support in terms of funding, technology, and talent for environmental protection. Meanwhile, they should also strengthen environmental law enforcement, avoiding the blind introduction of polluting enterprises from developed countries for the sake of economic growth, while neglecting environmental sustainability. Additionally, cities with different levels of ER should form a complementary and mutually beneficial development pattern with surrounding areas. Meanwhile, it is necessary to actively foster a favorable interaction between “top-down” and “bottom-up” development patterns to build a two-way channel for mutual promotion and coordinated development.

Thirdly, the government should improve the comprehensive policy system for ER and establish the appropriate and sustainable environmental protection mechanism. On the one hand, local governments in remote and underdeveloped regions should prioritize public education and awareness campaigns to promote sustainable development practices among the public and industries. This can help prevent one-sided, cost-ignoring, and short-sighted development solely driven by economic growth. On the other hand, the national government should also strengthen supervision and inspection in key regions with lower ERI and key sectors with severe pollution, rigorously penalizing violations such as illegal emissions and fraudulent pollution control practices, which can encourage the green transformation of polluting enterprises and the establishment of effective environmental governance systems. Finally, the national government, particularly in developing countries, should strengthen the top-level design of environmental governance, improve relevant legal and supportive policies from the views of climate change, ecological deterioration, environmental pollution, and other environmental issues and effectively promote the fulfillment of responsibilities by local governments, thus driving coordinated development between the economy and environmental protection at a holistic level.

Limitations and future works

The theoretical analysis and empirical tests provide some bright contributions to the existing literature, and the research highlights confirm the inhibitory effect and positive spatial overflow effect of ERI on environmental Iso₂. Indeed, the peer effect of ERI in China's environmental governance system is also evidently supported. By contrast, there are still some shortcomings in this research that needs to be further improved in future studies, which are embodied in the following dual aspects:

Firstly, although we strive to accurately estimate the nexus between ERI and Iso₂, and effectively test the existence of peer effects of ERI among local governments, the availability of data makes the empirical research still somewhat biased. Considering the objective fact that the number of years of macro-official data statistics, this study uses up-to-date statistical data to conduct empirical research between key variables (2008–2019). Thereby, this research provides important research directions for our next research, namely, we will keep the focus on the update of statistics to do further exploration to ensure the timeliness of the research data. Secondly, in terms of the endogeneity issues, although we adopt the generalized spatial two-stage least squares and system GMM regression to address the endogeneity concerns, we fail to find appropriate instrumental variables in the process of empirical testing relying on the instrumental variable method. In this regard, this part of the discussion is flawed to some extent. Thereby, we will try to look for strictly exogenous variables as instrumental variables in future research and re-examine the “inhibitory effect” and “peer effect” of ER.