Research on the Impact Effects of the Thermal Power Industry and Other High-Haze-Pollution Industries on the Atmospheric Environment

Abstract

As industrialization accelerates, China’s industrial development pace has been rapidly increasing. However, this growth has been accompanied by an increase in high-pollution and high-emission industries, leading to the release of a significant amount of air pollutants and exacerbating haze pollution nationwide. This article utilizes the spatial dynamic Durbin model and panel threshold regression model to analyze the impact of the thermal power industry and other high-haze-pollution industries on atmospheric environmental quality. The results indicate a negative correlation between the thermal power industry and other high-haze pollution industries and atmospheric environmental quality. There is a spatial spillover effect of the thermal power industry and other high-haze-pollution industries on air pollution. Environmental regulations have a single-threshold characteristic in their impact on atmospheric quality in the thermal power industry and other high-haze-pollution industries, as does green technology innovation. Additionally, cumulative rainfall has a significant single-threshold effect on the atmospheric environmental quality in regions with the thermal power industry and other high-haze-pollution industries. The article suggests policies for severely polluted areas, including reducing high-haze-pollution enterprises, optimizing industrial structures rationally, strengthening regional cooperation, enhancing regional haze pollution prevention and control coordination mechanisms, increasing the intensity of environmental regulations, utilizing the threshold effect of environmental regulations, promoting green technological innovation and application in heavily polluted areas, and exploring options to improve air pollution through increased rainfall. These recommendations aim to provide reference points for China to further optimize its industrial structure and comprehensively manage haze pollution.

1. Introduction

1.1. Research Background

China’s industrial development has witnessed a significant upward trend as the process of industrialization continues to accelerate. According to data from the National Bureau of Statistics of China, as of 2019, China’s industrial added value has surpassed the early days of New China by a staggering factor of 970, establishing an independent and comprehensive modern industrial system. China stands as the sole country globally to encompass all industrial categories outlined in the United Nations industrial classification. Since the initiation of the reform and opening-up policy in 1978, industries associated with industrialization have made remarkable contributions to China’s economic growth. According to World Bank data in 2019, China’s manufacturing added value first surpassed that of the United States in 2010, solidifying its position as the world’s leading manufacturing powerhouse. By 2018, China’s manufacturing added value accounted for over 28% of the global total.

However, beneath China’s rapid industrial growth lies significant environmental costs. In the World Development Report 2017 issued by the World Bank, China was classified alongside India as a country with high economic growth but also high environmental pollution. The increasing emissions of particulate-matter-containing waste gases resulting from industrial development have become a major contributing factor to the escalating severity of air pollution in China. The proliferation of highly polluting and high-emission enterprises such as petroleum and chemical industries and thermal power plants has led to the release of substantial amounts of harmful substances into the atmosphere, resulting in a sharp decline in air quality. Particularly after 2013, the central and eastern regions of China experienced numerous episodes of large-scale and persistent haze, severely impacting both atmospheric conditions and public health. Despite the introduction of policies such as the “Ten Articles on Air Quality” and the implementation of measures like the Environmental Protection Law of the People’s Republic of China by the central government, the problem of haze pollution in key regions such as the Beijing–Tianjin–Hebei area and its surrounding areas remains grave. From October 2018 to March 2019, the cumulative number of heavily polluted days in the Beijing–Tianjin–Hebei region and its neighboring cities reached 624 days. Adding to the perplexity, even after the comprehensive outbreak of the COVID-19 pandemic in early 2020, with the majority of motor vehicles, factories, and construction sites forced to suspend operations, and restaurants compelled to close, severe haze pollution persisted in most northern Chinese cities during periods such as 25–28 January and 11–13 February 2020. This is mainly due to winter heating demands and certain key industries such as steel, thermal power, and petrochemicals, which, despite facing the influence of winter heating and the pandemic, did not cease operations due to various reasons. In addition, the haze pollution in the United States is mainly caused by the exhaust emissions of fuel vehicles. The haze pollution in the United Kingdom is mainly caused by coal combustion, while the haze pollution in China is mainly caused by desulfurization, underselling, dust removal, and water vapor spraying in high-haze-pollution plants such as thermal power plants. Therefore, this paper selects China as a case study.

Consequently, investigating the impact of these industrial enterprises with high haze pollution of the atmospheric environment has become of paramount importance in China’s efforts to control atmospheric environmental pollution. This research aims to explore the negative effects of high-haze-pollution enterprises on the atmospheric environment and elucidate the specific pathways through which these impacts occur. The findings will provide better choices for administrative authorities to optimize current resource allocation. Moreover, through an in-depth analysis of the structure and characteristics of industrial pollution industries, this research seeks to understand their specific influence on the atmospheric environment, thereby offering targeted strategies to address the environmental challenges currently faced by China and the world. By doing so, China can achieve rapid and efficient economic growth while ensuring the realization of its goals in sustainable development and environmental protection.

1.2. Relevant Literature

1.2.1. Research on Haze Pollution

Regarding air pollution, haze mainly refers to the mixture of fog and smoke, with PM2.5 and other fine particles being its main components (Dong et al., 2019 [1]). Since the beginning of the 21st century, with the accelerated industrialization and rapid increase in the number of high-haze-pollution industries in China, the haze problem in many cities has become increasingly serious. In recent years, many scholars in China have conducted extensive research on the sources and formation mechanisms of haze pollution (Ren et al., 2014 [2]; Wang et al., 2013 [3]; Zhao et al., 2013 [4]; Huang et al., 2012 [5]). The main characteristics of haze pollution vary among different seasons, with coal combustion being the major factor influencing winter haze pollution (Sun et al., 2013 [6]; Li et al., 2010 [7]). In the formation process of summer haze, crop burning and meteorological factors play dominant roles (Huang et al., 2012 [8]).

The formation of haze is not only related to physical and chemical factors such as atmospheric conditions and climate change but also to human economic activities (Xu and Lin, 2016 [9]; Ma et al., 2016 [10]). Regarding the relationship between economic growth and haze pollution, both domestic and foreign scholars generally believe that economic growth is a decisive factor influencing PM2.5 emissions, and there exists a non-linear relationship between PM2.5 and GDP per capita (Hao and Liu, 2016 [11]; Keene and Deller, 2015 [12]; Ma et al., 2016 [10]). However, some scholars argue that economic growth will exacerbate PM2.5 emissions, and technological progress cannot offset this increase (Lyu et al., 2016 [13]; Fu et al., 2020 [14]).

1.2.2. Spatial Spillover Effects of Atmospheric Pollution

Currently, research on the spatial spillover of atmospheric pollution mainly includes the following aspects: Ma et al. (2016 [10]) pointed out that due to the geographical and regional economic development attributes, atmospheric pollution in China exhibits significant spatial clustering. Shao et al. (2016 [15]) used nighttime satellite data to verify the significant spatial spillover effects of atmospheric pollution in 30 provinces in China. In addition, when considering the spatial spillover effects of atmospheric pollution, both natural and socioeconomic factors have varying degrees of influence. Factors such as wind speed and direction, atmospheric humidity and pressure, geographical location, and season can affect the regional migration of pollutants (Liu et al., 2015 [16]; Liu Xiaorui, 2020 [17]). Moreover, a series of socioeconomic factors including economic level, production technology, industrial structure, population density, and foreign direct investment also influence the migration of air pollutants (Cheng et al., 2017 [18]; Liddle, 2014 [19]; Kouchaki-Penchah et al., 2016 [20]; Zhu et al., 2019 [21]).

1.2.3. The Impact of Environmental Regulations on the Atmospheric Environment

Currently, there is limited research on the impact of environmental regulations on atmospheric environmental quality. Greenstone (2001 [22]) studied the U.S. manufacturing industry and found that environmental regulations cannot effectively play a role in certain industries. Schou (2002 [23]) argued that environmental regulations are redundant environmental protection measures. Davis (2008 [24]) studied the effectiveness of traffic regulations in Mexico City and found insufficient evidence to prove that environmental regulations can effectively improve air quality. In recent years, studies on environmental regulations have gradually shown their effectiveness in improving atmospheric environmental quality. Environmental regulations have reversed the impact of industrial structure on atmospheric environmental quality and effectively improved it (Zhang et al., 2019 [25]). Guo and Lu studied the relationship between China’s air quality and jurisdictional fragmentation using a spatial autoregressive model and found that the number of regulatory units per square kilometer within prefecture-level cities has a significant negative impact on air quality (2019 [26]). In summary, the mechanisms by which environmental regulations affect atmospheric environmental quality still require further research. Furthermore, existing studies often overlook the influence of industrial upgrading and high-haze-pollution industries on PM2.5. Therefore, it is necessary to conduct more comprehensive research by incorporating a wider range of factors.

1.2.4. The Impact of Green Technological Innovation on Atmospheric Environmental Quality

Currently, there are two main perspectives regarding the research on the impact of green technological innovation on atmospheric environmental quality. One perspective posits that green technological innovation can effectively enhance atmospheric environmental quality (Zhou et al., 2014 [27]; Lu et al., 2020 [28]; Zhou et al., 2023 [29]). Conversely, some scholars hold a contrary viewpoint suggesting that green technological innovation may, to a certain extent, diminish atmospheric environmental quality. For instance, Li et al. (2017 [30]), through an analysis of panel data from 21 two-digit industrial sectors in China spanning the years 1999 to 2013, found that while technological innovation may enhance production efficiency, it concurrently depletes non-renewable resources and contributes to environmental pollution.

1.2.5. The Impact of High-Pollution Industries on the Atmospheric Environment

The atmospheric environment serves as the primary and most direct indicator of the state of the natural environment. Currently, there is a limited amount of research in the academic community, both domestically and internationally, regarding the relationship between high-haze-pollution industries and atmospheric environmental quality. The majority of studies have focused on the relationship between industrial structure and atmospheric environmental pollution. Chebbi (2010 [31]) argued that industrial structure plays a decisive role in the allocation of production resources such as labor, capital, technology, and energy among different industries, thus exerting a significant influence on the emission of atmospheric pollutants. Zhang et al. (2014 [32]) suggested that industrial structure can have a substantial impact on changes in the ecological environment. Industrial structure refers to the composition of various industries and the inter-relationships and proportions between them. Some key studies include Cui et al. (2015 [33]), who established an Ecological Environment Quality Index (EQI) model for Jinan City from 2000 to 2011 and found a negative correlation between the proportion of the primary industry and atmospheric environmental conditions. Wang et al. (2016 [34]) discovered that optimizing the industrial structure can significantly improve atmospheric environmental quality and break the vicious cycle of imbalanced development between the ecological environment and industrial structure.

To summarize, the current research in the domestic and international academic community on the relationship between industrial structure and atmospheric environmental quality is relatively comprehensive. However, there remains a lack of research on the relationship between high-haze-pollution enterprises and atmospheric environmental quality. In order to address this research gap, this research employs a spatial dynamic Durbin model and a panel threshold regression model to analyze the relationship between high-haze-pollution industries and atmospheric environmental quality. The aim is to provide empirical analysis results that can serve as a reference for the formulation of targeted policies by the government and contribute to future research in this field.

2. Research Design

2.1. Main Research Methodologies

2.1.1. Static Panel Model

This paper refers to the research of Huang et al. (2019) [35] and sets the static panel model formula as:

$$\begin{gathered} {\mathbf{G}\mathbf{A}\mathbf{P}\mathbf{M}}_{\mathbf{i}\mathbf{t}}={\mathsf{\beta }}_{\mathbf{1}}{\mathbf{I}\mathbf{G}\mathbf{S}}_{\mathbf{i}\mathbf{t}}+{\mathsf{\beta }}_{\mathbf{2}}{\mathit{P}\mathit{G}}_{\mathbf{i}\mathbf{t}}+{\mathsf{\beta }}_{\mathbf{3}}{\mathbf{C}\mathbf{D}}_{\mathbf{i}\mathbf{t}}+{\mathsf{\beta }}_{\mathbf{4}}{\mathbf{P}\mathbf{A}\mathbf{A}\mathbf{Q}}_{\mathbf{i}\mathbf{t}}+{\mathsf{\beta }}_{\mathbf{5}}{\mathbf{P}\mathbf{G}\mathbf{D}\mathbf{P}}_{\mathbf{i}\mathbf{t}} \\ +{\mathsf{\beta }}_{\mathbf{6}}{\mathbf{I}\mathbf{S}}_{\mathbf{i}\mathbf{t}}+{\mathsf{\beta }}_{\mathbf{7}}{\mathbf{F}\mathbf{O}\mathbf{L}}_{\mathbf{i}\mathbf{t}}+{\mathsf{\mu }}_{\mathbf{i}}+{\mathsf{\epsilon }}_{\mathbf{i}\mathbf{t}} \end{gathered}$$

where the subscripts i and t represent the ith province and the tth year, respectively. ${\mathsf{\mu }}_{\mathrm{i}}$ and ${\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}$ represent disturbance terms.

2.1.2. Panel Threshold Regression Model

According to the research results of Hansen (1999) [36] and Li (2015) [37], the mathematical expression for the panel single-threshold model of environmental regulatory intensity (lnk₁) is:

$$\begin{array}{l}{\mathit{l}\mathit{n}\mathit{y}}_{\mathit{i}\mathit{t}}={\mathsf{\beta }}_{\mathbf{1}}{\mathit{l}\mathit{n}\mathit{k}\mathbf{1}}_{\mathit{i}\mathit{t}}\left({\mathit{l}\mathit{n}\mathit{k}\mathbf{1}}_{\mathit{i}\mathit{t}}<\mathsf{\gamma }\right)+{\mathsf{\beta }}_{\mathbf{2}}{\mathit{l}\mathit{n}\mathit{k}\mathbf{1}}_{\mathit{i}\mathit{t}}\left({\mathit{l}\mathit{n}\mathit{k}\mathbf{1}}_{\mathit{i}\mathit{t}}\ge \mathsf{\gamma }\right)\\ +{\mathsf{\beta }}_{\mathbf{3}}{\mathit{l}\mathit{n}\mathit{x}\mathbf{1}}_{\mathit{i}\mathit{t}}+{\mathsf{\beta }}_{\mathbf{4}}{\mathit{l}\mathit{n}\mathit{m}\mathbf{1}}_{\mathit{i}\mathit{t}}+{\mathsf{\beta }}_{\mathbf{5}}{\mathit{l}\mathit{n}\mathit{m}\mathbf{2}}_{\mathit{i}\mathit{t}}+{\mathsf{\beta }}_{\mathbf{6}}{\mathit{l}\mathit{n}\mathit{m}\mathbf{3}}_{\mathit{i}\mathit{t}}\\ +{\mathsf{\beta }}_{\mathbf{7}}{\mathit{l}\mathit{n}\mathit{m}\mathbf{4}}_{\mathit{i}\mathit{t}}+{\mathsf{\beta }}_{\mathbf{8}}{\mathit{l}\mathit{n}\mathit{m}\mathbf{5}}_{\mathit{i}\mathit{t}}+{\mathsf{\beta }}_{\mathbf{9}}{\mathit{l}\mathit{n}\mathit{m}\mathbf{6}}_{\mathit{i}\mathit{t}}+{\mathit{\epsilon }}_{\mathit{i}\mathit{t}}\end{array}$$

The mathematical expression for the panel single-threshold model of green technological innovation (lnk₂) is:

$$\begin{array}{l}{\mathit{l}\mathit{n}\mathit{y}}_{\mathit{i}\mathit{t}}={\mathsf{\beta }}_{\mathbf{1}}{\mathit{l}\mathit{n}\mathit{k}\mathbf{2}}_{\mathit{i}\mathit{t}}\left({\mathit{l}\mathit{n}\mathit{k}\mathbf{2}}_{\mathit{i}\mathit{t}}<\mathsf{\gamma }\right)+{\mathsf{\beta }}_{\mathbf{2}}{\mathit{l}\mathit{n}\mathit{k}\mathbf{2}}_{\mathit{i}\mathit{t}}\left({\mathit{l}\mathit{n}\mathit{k}\mathbf{2}}_{\mathit{i}\mathit{t}}\ge \mathsf{\gamma }\right)\\ +{\mathsf{\beta }}_{\mathbf{3}}{\mathit{l}\mathit{n}\mathit{x}\mathbf{1}}_{\mathit{i}\mathit{t}}+{\mathsf{\beta }}_{\mathbf{4}}{\mathit{l}\mathit{n}\mathit{m}\mathbf{1}}_{\mathit{i}\mathit{t}}+{\mathsf{\beta }}_{\mathbf{5}}{\mathit{l}\mathit{n}\mathit{m}\mathbf{2}}_{\mathit{i}\mathit{t}}+{\mathsf{\beta }}_{\mathbf{6}}{\mathit{l}\mathit{n}\mathit{m}\mathbf{3}}_{\mathit{i}\mathit{t}}\\ +{\mathsf{\beta }}_{\mathbf{7}}{\mathit{l}\mathit{n}\mathit{m}\mathbf{4}}_{\mathit{i}\mathit{t}}+{\mathsf{\beta }}_{\mathbf{8}}{\mathit{l}\mathit{n}\mathit{m}\mathbf{5}}_{\mathit{i}\mathit{t}}+{\mathsf{\beta }}_{\mathbf{9}}{\mathit{l}\mathit{n}\mathit{m}\mathbf{6}}_{\mathit{i}\mathit{t}}+{\mathit{\epsilon }}_{\mathit{i}\mathit{t}}\end{array}$$

The mathematical expression for the panel single-threshold model of cumulative precipitation (lnk₃) is:

$$\begin{array}{l}{\mathit{l}\mathit{n}\mathit{y}}_{\mathit{i}\mathit{t}}={\mathsf{\beta }}_{\mathbf{1}}{\mathit{l}\mathit{n}\mathit{k}\mathbf{3}}_{\mathit{i}\mathit{t}}\left({\mathit{l}\mathit{n}\mathit{k}\mathbf{3}}_{\mathit{i}\mathit{t}}<\mathsf{\gamma }\right)+{\mathsf{\beta }}_{\mathbf{2}}{\mathit{l}\mathit{n}\mathit{k}\mathbf{3}}_{\mathit{i}\mathit{t}}\left({\mathit{l}\mathit{n}\mathit{k}\mathbf{3}}_{\mathit{i}\mathit{t}}\ge \mathsf{\gamma }\right)\\ +{\mathsf{\beta }}_{\mathbf{3}}{\mathit{l}\mathit{n}\mathit{x}\mathbf{1}}_{\mathit{i}\mathit{t}}+{\mathsf{\beta }}_{\mathbf{4}}{\mathit{l}\mathit{n}\mathit{m}\mathbf{1}}_{\mathit{i}\mathit{t}}+{\mathsf{\beta }}_{\mathbf{5}}{\mathit{l}\mathit{n}\mathit{m}\mathbf{2}}_{\mathit{i}\mathit{t}}+{\mathsf{\beta }}_{\mathbf{6}}{\mathit{l}\mathit{n}\mathit{m}\mathbf{3}}_{\mathit{i}\mathit{t}}\\ +{\mathsf{\beta }}_{\mathbf{7}}{\mathit{l}\mathit{n}\mathit{m}\mathbf{4}}_{\mathit{i}\mathit{t}}+{\mathsf{\beta }}_{\mathbf{8}}{\mathit{l}\mathit{n}\mathit{m}\mathbf{5}}_{\mathit{i}\mathit{t}}+{\mathsf{\beta }}_{\mathbf{9}}{\mathit{l}\mathit{n}\mathit{m}\mathbf{6}}_{\mathit{i}\mathit{t}}+{\mathit{\epsilon }}_{\mathit{i}\mathit{t}}\end{array}$$

2.1.3. Spatial Durbin Model

Spatial econometric models encompass the spatial lag model (SLM), spatial error model (SEM), and spatial Durbin model (SDM). The spatial lag model (SLM) is primarily employed to investigate whether the explanatory variables exhibit spillover effects across regions. The spatial error model (SEM), on the other hand, is used to examine the spatial effects of omitted variables or unobserved random shocks not accounted for in the explanatory variables. Given the focus of this research on the impact of the thermal power industry and other high-haze-pollution industries on atmospheric environmental quality, the spatial Durbin model (SDM) is selected for analysis. Furthermore, compared to other spatial models, the SDM includes spatially lagged variables of the dependent variable, which can partially mitigate omitted variable bias.

Referring to the research of Zhu et al. (2023) [38], the formula of the spatial Durbin model (SDM) is set as:

$$\begin{array}{l}{\mathbf{G}\mathbf{A}\mathbf{P}\mathbf{M}}_{\mathbf{i}\mathbf{t}}=\mathsf{\rho }{\displaystyle \sum _{\mathbf{j}=\mathbf{1}}^{\mathbf{N}}}{\mathbf{W}}_{\mathbf{i}\mathbf{t}}{\mathbf{G}\mathbf{A}\mathbf{P}\mathbf{M}}_{\mathbf{i}\mathbf{t}}+{\mathsf{\alpha }}_{\mathbf{1}}{\mathbf{I}\mathbf{G}\mathbf{S}}_{\mathbf{i}\mathbf{t}}+{\mathsf{\alpha }}_{\mathbf{2}}{\mathit{P}\mathit{G}}_{\mathbf{i}\mathbf{t}}+{\mathsf{\alpha }}_{\mathbf{3}}{\mathbf{C}\mathbf{D}}_{\mathbf{i}\mathbf{t}}\\ +{\mathsf{\alpha }}_{\mathbf{4}}{\mathbf{P}\mathbf{A}\mathbf{A}\mathbf{Q}}_{\mathbf{i}\mathbf{t}}+{\mathsf{\alpha }}_{\mathbf{5}}{\mathbf{P}\mathbf{G}\mathbf{D}\mathbf{P}}_{\mathbf{i}\mathbf{t}}+{\mathsf{\beta }}_{\mathbf{1}}{\displaystyle \sum _{\mathbf{j}=\mathbf{1}}^{\mathbf{N}}}{\mathbf{W}}_{\mathbf{i}\mathbf{t}}{\mathbf{I}\mathbf{G}\mathbf{S}}_{\mathbf{i}\mathbf{t}}\\ +{\mathsf{\beta }}_{\mathbf{2}}{\displaystyle \sum _{\mathbf{j}=\mathbf{1}}^{\mathbf{N}}}{\mathbf{W}}_{\mathbf{i}\mathbf{t}}{\mathit{P}\mathit{G}}_{\mathbf{i}\mathbf{t}}+{\mathsf{\beta }}_{\mathbf{3}}{\displaystyle \sum _{\mathbf{j}=\mathbf{1}}^{\mathbf{N}}}{\mathbf{W}}_{\mathbf{i}\mathbf{t}}{\mathbf{C}\mathbf{D}}_{\mathbf{i}\mathbf{t}}\\ +{\mathsf{\beta }}_{\mathbf{4}}{\displaystyle \sum _{\mathbf{j}=\mathbf{1}}^{\mathbf{N}}}{\mathbf{W}}_{\mathbf{i}\mathbf{t}}{\mathbf{P}\mathbf{A}\mathbf{A}\mathbf{Q}}_{\mathbf{i}\mathbf{t}}+{\mathsf{\beta }}_{\mathbf{5}}{\displaystyle \sum _{\mathbf{j}=\mathbf{1}}^{\mathbf{N}}}{\mathbf{W}}_{\mathbf{i}\mathbf{t}}{\mathbf{P}\mathbf{G}\mathbf{D}\mathbf{P}}_{\mathbf{i}\mathbf{t}}+{\mathsf{\mu }}_{\mathbf{i}}+{\mathsf{\lambda }}_{\mathbf{i}}\\ +{\mathsf{\epsilon }}_{\mathbf{i}\mathbf{t}}\end{array}$$

where ρ represents the spatial regression coefficient, and ${\mathrm{W}}_{\mathrm{i}\mathrm{t}}$ denotes the ith row and jth column element of the N × N standardized non-negative spatial weight matrix W. The subscripts i and t, respectively, indicate the ith province and the tth year. ${\mathsf{\mu }}_{\mathrm{i}}$ and ${\mathsf{\lambda }}_{\mathrm{i}}$ represent the spatial and time effects, respectively.

2.2. Research Hypotheses

In the research on the impact effects of the thermal power industry and other high-haze-pollution industries on atmospheric environmental quality, various factors need to be considered, such as environmental regulations, spatial spillover effects of air pollution, green technological innovation, and the relationship between the thermal power industry and other high-haze-pollution industries and atmospheric environmental quality. Therefore, this paper focuses on investigating the relationship between the thermal power industry and other high-haze-pollution industries and atmospheric environmental quality, the spatial spillover effects of air pollution, the threshold effects of environmental regulations on atmospheric environmental quality, and the threshold effects of green technological innovation.

2.2.1. Relationship Between Thermal Power Industry and Other High-Haze-Pollution Industries and Atmospheric Environmental Quality

He and Xu (2009) conducted empirical analysis using the Tobit model and found a positive correlation between industrial development and haze pollution [39]. Chen and Zhang (2010) analyzed carbon emissions in China and used an input–output model to process data on carbon emissions from 2002 to 2005. They concluded that a significant reduction in the proportion of pollution-intensive enterprises led to a significant decrease in carbon emissions, suggesting that industrial adjustments contribute to pollution reduction [40]. Based on the analysis of the relationship between the thermal power industry and other high-haze-pollution industries and environmental pollution, we propose Hypothesis 1:

H1. 

There is a negative correlation between the thermal power industry and other high-haze-pollution industries and environmental pollution.

2.2.2. Spatial Spillover Effects of Thermal Power Industry and Other High-Haze-Pollution Industries on Air Pollution

Liang et al. (2017) found that air pollution exhibits strong spatial correlation and regional transmission characteristics due to its high mobility [41]. Pan et al. (2015) discovered that air pollution is largely spread or transferred to neighboring areas through atmospheric circulation and industrial relocation, among other natural and economic mechanisms [42]. Ma et al. (2016) identified spatial spillover from pollutants in adjacent areas as one of the significant causes of air pollution [43]. Based on the aforementioned analysis, we propose Hypothesis 2:

H2. 

The impact of the thermal power industry and other high-haze-pollution industries on air pollution exhibits strong spatial spillover effects.

2.2.3. Threshold Effects of Environmental Regulations on Atmospheric Environmental Quality in Thermal Power Industry and Other High-Haze Pollution Industries

Environmental regulatory tools are usually classified into three categories: command and control, which involves administrative orders and laws and regulations; economic incentives and constraints, which utilize market-based regulations; and implicit environmental regulations related to environmental awareness, attitudes, and cognition (Testa et al., 2011 [44]). Wanlley and Whitehead (1994 [45]) suggested that when environmental regulations are stringent, they not only lead to higher production costs but also divert investment funds from prospective projects to pollution reduction and quality improvement. From a static perspective, environmental regulations have both constraint and crowding-out effects on highly polluting enterprises, thereby impacting their pollutant emissions. Li et al. (2013 [46]) found that when environmental regulations are relatively weak, the crowding-out effect dominates because the innovative behaviors adopted by companies in response to regulations often exhibit a lag. Moreover, complying with regulations incurs lower costs, and companies lack the motivation to implement technology innovations that help reduce atmospheric pollutant emissions. As a result, the environmental regulations have a limited impact on atmospheric environmental improvement. However, as the regulatory intensity strengthens, the costs of compliance also increase, and technological innovation gradually becomes the optimal choice for maximizing company profits. Environmental regulations progressively restrict the ability of the thermal power industry and other highly polluting enterprises to emit pollutants. Based on the above analysis, we propose Hypothesis 3:

H3. 

Environmental regulations have threshold effects on atmospheric environmental quality in the thermal power industry and other high-haze-pollution industries.

2.2.4. Threshold Effects of Green Technological Innovation on Atmospheric Environmental Quality in Thermal Power Industry and Other High-Haze-Pollution Industries

Green technological innovation prompts companies to shift away from the traditional production model of “high input–low output–high energy consumption–high pollution”. It disconnects resource and environmental pressures from socioeconomic activities by maximizing the use of raw materials to reduce energy consumption and adopting ecological cycle controls during operations to avoid or minimize pollution. Lan and Munro (2014 [47]) found that as the intensity of environmental regulations increases, the market demand for environmental technologies also rises. Consequently, companies allocate more resources to R&D in environmental technologies, leading to environmental quality improvements. Huang et al. (2020 [48]) suggested that technological innovation plays a promoting role in reducing environmental pollution under environmental regulations. Lu et al. (2020 [28]) found that green technological innovation has a significant inhibitory effect on haze pollution when the level of local environmental regulations is relatively high. Based on the above analysis, we propose Hypothesis 4:

H4. 

Green technological innovation has threshold effects on atmospheric environmental quality in the thermal power industry and other high-haze-pollution industries.

2.2.5. Threshold Effects of Rainfall on Atmospheric Environmental Quality in Thermal Power Industry and Other High-Haze-Pollution Industries

Rainfall is considered one of the main natural processes that improve air quality (Duhanyan and Roustan, 2011 [49]). It can significantly enhance the positive emission reductions achieved through human control measures (Leung and Gustafson, 2005 [50]). However, there is limited research in the academic community on the flushing effect of particulate matter and the improvement of air quality after a single rainfall event, as well as the corresponding lag effects. Previous research often combined multiple rainfall events into one sample (Guo et al., 2014 [51]). However, Yuan et al. (2014 [52]) found that after a low rainfall amount, rainfall may worsen air quality. This suggests that there might be a critical threshold between rainfall amount and changes in air quality. Based on this, we propose Hypothesis 5:

H5. 

Rainfall has threshold effects on atmospheric environmental quality in high-haze-pollution industries.

2.3. Variable Selection

2.3.1. Dependent Variable

Atmospheric environmental quality (measured by the Atmospheric Environmental Composite Pollution Index in this research). Environmental quality generally refers to the suitability of the overall environment or certain elements of the environment for human survival, reproduction, and socioeconomic development in a specific environment (Du, 2007 [53]). In this research, the Atmospheric Environmental Composite Pollution Index is calculated based on the emissions of three major pollutants: sulfur dioxide, nitrogen oxides, and geographical average PM2.5. The formula for the Atmospheric Pollution Composite Index is as follows: Atmospheric Pollution Composite Index = (sulfur dioxide emissions (t)/GDP + nitrogen oxide emissions (t)/GDP + geographical average PM2.5/GDP)/3.

2.3.2. Key Independent Variable

The total industrial output value of the thermal power industry and other high-haze-pollution industries represents the economic value of the thermal power industry and other five high-haze-pollution industries. The production methods of high-haze-pollution enterprises, such as thermal power companies and petrochemical industries, which primarily rely on coal, oil, and natural gas, determine their destructive impact on atmospheric environmental resources (Yang and Zhang, 2014 [54]). The economic activities of the thermal power industry and other high-haze-pollution industries, which are directly related to atmospheric pollution, can be better captured by using the total industrial output value of high-haze-pollution industries to measure their direct impact on atmospheric environmental quality.

2.3.3. Threshold Variables

Environmental regulations. Government environmental regulatory policies can promote technological innovation in the environmental field, thereby prompting producers to reduce pollution emissions (He et al., 2015 [55]). In this research, the removal amount of industrial sulfur dioxide is used as a measure.

Green technological innovation. “Green technology” refers to technology that allows for the efficient use of natural resources and is environmentally friendly in production and use (Ding, 1993 [56]). This variable reflects the impact of technological progress on atmospheric environmental quality.

Cumulative rainfall. Cumulative rainfall, as a natural factor, can reflect the dilution and deposition effects in the atmosphere.

2.3.4. Control Variables

Energy efficiency. This variable can reflect the efficiency of energy utilization in a region.

Built-up area. Larger built-up areas may have more industries and population, thereby affecting atmospheric pollution.

Level of economic development. Measured by GDP per capita, a higher level of economic development may be accompanied by more industrial activities and pollution.

Industrial structure. The proportion of the secondary industry in GDP is used to measure the industrial structure of a region, which reflects the impact of industrial structure on atmospheric pollution.

Level of openness. The ratio of foreign investment amount to GDP is used to measure the level of openness in a region, which can reflect the potential impact of external investment on the environment.

Population density. Higher population density may result in more residential and transportation activities, thereby affecting atmospheric pollution.

The definitions and explanations of all variables are shown in

Table 1. Variable Definitions and Explanations.

Variable Type	Abbreviation	Name	Calculation Method
Dependent Variable	y	Atmospheric environmental quality	Comprehensive Atmospheric Environmental Pollution Index
Independent Variable	x1	Total industrial output value of thermal power industry and other high-haze-pollution industries	Total economic value of thermal power industry and five other high-haze-pollution industries
	x2	Industrial sales value of thermal power industry and other high-haze-pollution industries	Industrial sales value of thermal power industry and five other high-haze-pollution industries
Threshold Variables	k1	Environmental regulation	Industrial sulfur dioxide removal quantity
	k2	Green technological innovation
	k3	Cumulative precipitation
Control Variables	m1	Energy efficiency
	m2	Built-up area
	m3	Level of economic development	GDP per capita
	m4	Industrial structure	Proportion of the secondary industry in GDP
	m5	Level of openness to the outside world	Actual utilized foreign investment amount/GDP
	m6	Population density	Population density (persons per square kilometer)

.

2.4. Descriptive Statistics of Variables

This research conducted a descriptive statistical analysis of several key variables.

Table 2. Descriptive Analysis of Variables.

Variables	Sample	Mean	Standard	Minimum	Maximum
lny	589	−2.286	0.023	−2.302	−2.083
lnx1	589	6.065	1.559	0.025	9.020
lnx2	589	6.087	1.482	−0.045	9.035
lnk1	589	−2.286	0.014	−2.302	−2.194
lnk2	589	5.825	2.043	−2.303	10.262
lnk3	589	6.647	0.649	4.971	7.825
lnm1	589	−1.533	0.301	−1.988	−0.376
lnm2	589	6.867	0.881	4.278	8.706
lnm3	589	10.058	0.854	7.887	11.925
lnm4	589	−0.605	0.173	−1.271	−0.336
lnm5	589	−0.902	0.672	−1.917	1.748
lnm6	589	5.289	1.523	0.789	8.250
wlny	589	−2.289	0.011	−2.300	−2.267
wlnx1	589	6.166	1.173	4.211	7.881
wlnx2	589	6.152	1.171	4.184	7.835
wlnk1	589	−2.288	0.009	−2.298	−2.270
wlnk2	589	5.923	1.555	3.500	8.284
wlnk3	589	6.660	0.569	5.707	7.421
wlnm1	589	−1.550	0.233	−1.851	−1.079
wlnm2	589	6.905	0.662	5.677	7.842
wlnm3	589	10.057	0.767	8.807	11.111
wlnm4	589	−0.586	0.116	−0.806	−0.440
wlnm5	589	−0.924	0.579	−1.606	0.139
wlnm6	589	5.353	1.193	3.007	6.903

presents the main statistical characteristics of these variables. From the descriptive statistics provided in Table 2, it is evident that there are differences in the magnitude of the variables, warranting the natural logarithm transformation for all variables. Based on these statistical characteristics, we can observe a certain degree of homogeneity among the cities in terms of atmospheric environmental quality, as most cities face similar air pollution issues. However, there are significant variations in the economic impact of the thermal power industry and other high-haze-pollution industries across different cities. These variations may be attributed to differences in economic structure, industrial positioning, and environmental policies of the cities. For instance, some cities might rely more heavily on the thermal power industry and other high-haze-pollution industries as their main economic pillar, while others may have undergone industrial upgrading or implemented strict environmental policies, thereby reducing the economic impact of these industries.

2.5. Data Source

The data utilized in this research were sourced from multiple authoritative publications and databases, including the China Statistical Yearbook on Environment, China Statistical Yearbook, China Industrial Statistical Yearbook, and the EPS database. The PM2.5 data were obtained from the Global (GL) Annual PM2.5 Grids released by Columbia University for the period between 1998 and 2016. To address missing values, a linear interpolation method was employed for data imputation. Given that the five-year target for air pollution control in China extended until the year 2017, and considering the administrative measures implemented by the Chinese government to shut down a considerable amount of relevant polluting enterprises in 2017 and 2018, this research focused on provincial-level annual data from 2001 to 2017 to explore the relationship between PM2.5 and industrial agglomeration. The data were analyzed using STATA 15.0 and GeoDa1.4.6.

3. Empirical Analysis

The following are the results of the spatial Durbin model analysis and panel threshold regression analysis of the collected data using software stata15.0 and GeoDa 1.4.6:

3.1. Analysis of Spatial Durbin Model

3.1.1. Main Regression Coefficients

Table 3. Main Regression Coefficient Table.

	(1)	(2)
	wlny	wlny
Main
wlnx1	−0.000988 ***
	(−2.649)
wlnk1	−0.102 ***	−0.0983 ***
	(−4.509)	(−4.391)
wlnm1	0.00767 ***	0.00753 ***
	(4.298)	(4.265)
wlnm2	−0.00915 ***	−0.00859 ***
	(−7.941)	(−7.312)
wlnm3	−0.00802 ***	−0.00800 ***
	(−11.850)	(−12.013)
wlnm4	0.00378 **	0.00451 **
	(1.967)	(2.344)
wlnm5	−0.00252 ***	−0.00252 ***
	(−3.484)	(−3.528)
wlnm6	0.0147 ***	0.0164 ***
	(3.052)	(3.454)
wlnx2		−0.00122 ***
		(−3.288)
Wx
wlnx1	−0.000148
	(−1.005)
wlnk1	0.0441 ***	0.0457 ***
	(5.274)	(5.491)
wlnm1	−0.00631 ***	−0.00618 ***
	(−7.072)	(−6.978)
wlnm2	0.00239 ***	0.00237 ***
	(4.113)	(4.198)
wlnm3	−0.000843 **	−0.000792 **
	(−2.368)	(−2.340)
wlnm4	0.00197 **	0.00230 ***
	(2.438)	(2.872)
wlnm5	0.000753 **	0.000615 **
	(2.508)	(2.033)
wlnm6	0.00353	0.00325
	(1.364)	(1.290)
wlnx2		−0.000200
		(−1.477)
Spatial
rho	0.0563 ***	0.0535 ***
	(4.727)	(4.467)
Variance
sigma2_e	0.00000773 ***	0.00000763 ***
	(17.046)	(17.058)
N	589	589

t statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

presents the results of the SDM Model (1) lny-lnx1-lnk1 and Model (2) lny-lnx2-lnk1.

For SDM Model (1), a significant negative correlation between lny and lnx1 is observed. This implies a positive correlation between the growth of the thermal power industry and other high-haze-pollution industries and the deterioration of atmospheric environmental quality. This result aligns with Hypothesis 1, as an increase in the thermal power industry and other high-haze-pollution industries leads to larger production scales and higher emissions of pollutants, thereby accelerating the degradation of atmospheric environmental quality. Furthermore, a significant negative correlation is found between lny and lnk1, indicating that the strengthening of environmental regulations contributes to the improvement of atmospheric environmental quality. According to Porter Hypothesis (1995 [57]), strict environmental regulation can force thermal power and other high-haze-polluting enterprises to actively carry out environmental management, research and develop more green environmental protection technologies for desulfurization, underselling, and dust removal in thermal power and other high-haze-polluting industries, promote the greening of production technology, and achieve a cleaner production process environment. This promotes the greening of production processes and facilitates a cleaner production environment. For instance, since the 11th Five-Year Plan of the Communist Party of China, the Chinese government has incorporated environmental indicators as binding criteria for assessing the performance of government officials. This shift from a sole focus on GDP-oriented achievements has compelled local governments, as the actual implementers of central environmental policies and responsible parties for environmental pollution, to place significant constraints on heavily polluting enterprises. As this policy has been implemented, national data from the China National Environmental Monitoring Centre indicates varying degrees of improvement in atmospheric environmental quality compared to previous periods. Moreover, the implementation of environmental regulations is mandatory, and non-compliant enterprises face severe penalties for violating emission standards. Polluting entities have no room for negotiation and must comply with the established standards. Additionally, as the stringency of environmental regulations increases, the costs associated with non-compliance also rise. Consequently, an increasing number of highly polluting enterprises are compelled to adjust their industrial structures, reduce pollutant emissions, and consequently improve atmospheric environmental quality. Despite recent efforts to increase the development and utilization of petroleum, natural gas, and hydropower in China, coal remains a predominant component of China’s energy composition. Compared to other energy sources, coal combustion significantly contributes to air pollution. Environmental regulations often constrain coal consumption by increasing its cost, thereby reducing the proportion of coal used for heating during winter and optimizing the energy consumption structure. These measures aim to improve atmospheric environmental quality.

Turning to the relationship between lny and lnm1, a significant positive correlation is observed, indicating that higher energy utilization efficiency leads to better improvement in atmospheric environmental quality. The reasons behind this relationship can be elucidated as follows: Firstly, improved energy utilization efficiency implies reduced pollutant emissions when producing an equivalent amount of energy. Technological advancements in energy utilization can result in fewer harmful substances being released into the atmosphere during combustion or production processes. Therefore, a highly efficient energy system directly contributes to improved air quality. Secondly, increased energy efficiency reduces the energy costs required to produce the same amount of products or services, bringing in substantial cost savings for businesses and governments. These savings can be reinvested in environmental protection measures or the development of more environmentally friendly technologies, thereby positively impacting atmospheric environmental quality. Lastly, since approximately half of China’s coal is used for power generation, it is more practical to seek green coal-fired power generation technologies as a substitute for coal resources. High energy efficiency not only signifies more efficient production processes but also indicates a reduced overall energy requirement for completing the same tasks. This reduction can further lower the demand for mineral fuels that may contribute to air pollution.

Regarding the relationship between lny and lnm3, a significant negative correlation is observed. This implies that as the level of economic development in a region increases, the atmospheric pollution situation improves. Referring to studies on the Environmental Kuznets Curve, this negative correlation can be attributed to technological advancements, increased environmental awareness, and stricter environmental regulations leading to reduced atmospheric pollution emissions as economic development progresses. Analyzing the spatial lag effects of the SDM model, the spatial lag of lny and lnx1 exhibits an insignificant negative correlation. This indicates that the growth of high-haze-pollution industries in one region does not significantly affect the atmospheric environmental quality of neighboring regions. This could be attributed to neighboring regions implementing certain pollution control measures, such as establishing air-quality-monitoring stations, establishing green belts, or employing other technical means to reduce external pollution intrusion. On the other hand, the spatial lag of lny and lnk1 shows a significant positive correlation. This suggests that when the level of environmental regulation in neighboring regions increases, atmospheric environmental quality in a particular region also improves. In other words, environmental regulations have a spillover effect in spatial terms. This effect is mainly due to stricter environmental regulations resulting in reduced pollutant emissions from pollution sources. Additionally, according to the “pollution haven” hypothesis, as environmental regulations become stricter, the main emitters of atmospheric pollutants tend to shift to regions with less stringent environmental regulations, thereby improving the local atmospheric environmental quality.

Similar to the aforementioned Model (2), similar conclusions can be inferred.

3.1.2. Indirect Effect Analysis

The indirect effect results of the SDM Model (1) presented in

Table 4. Indirect Effect Results.

	(1)	(2)
	wlny	wlny
LR_Indirect
wlnx1	−0.00114
	(−1.354)
wlnk1	0.229 ***	0.237 ***
	(4.755)	(5.031)
wlnm1	−0.0352 ***	−0.0340 ***
	(−6.458)	(−6.440)
wlnm2	0.0115 ***	0.0114 ***
	(3.177)	(3.334)
wlnm3	−0.00795 ***	−0.00735 ***
	(−4.558)	(−4.619)
wlnm4	0.0130 ***	0.0149 ***
	(2.675)	(3.145)
wlnm5	0.00363 **	0.00281 *
	(2.132)	(1.688)
wlnm6	0.0262 *	0.0245 *
	(1.862)	(1.813)
wlnx2		−0.00149 **
		(−1.976)
N	589	589

t statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

indicate a non-significant negative correlation between lny and lnx1. This suggests that the reduction of the thermal power industry and other high-haze-pollution industries in certain regions not only lowers their own atmospheric environmental pollution but also alleviates atmospheric pollution in neighboring areas. There exists a certain degree of spatial spillover effect, which represents an important form of spatial interaction and manifests as the diffusion of socioeconomic elements in space. This finding is consistent with research Hypothesis 2. The primary reason for this phenomenon is that when the thermal power industry and other high-haze-pollution industries in the surrounding areas diminish, the emission rates of atmospheric pollutants correspondingly decrease. The reduction in atmospheric pollution in neighboring regions indirectly contributes to the mitigation of local atmospheric environmental pollution. Furthermore, the decrease in high-haze-pollution enterprises in the surrounding areas reduces the management pressure on the local government regarding high-haze-pollution industries. Consequently, the local government can employ more effective measures and methods to regulate the existing thermal power industry and other high-haze-pollution industries, enhance environmental regulatory intensity, and implement corresponding policies for the control of atmospheric environmental pollution. As the policies in the surrounding areas are effectively implemented and yield positive outcomes, the local government naturally tends to learn from the environmental regulatory policies adopted by neighboring regions. Therefore, there is an indirect influence on the governance of local atmospheric environmental pollution.

Similar to the aforementioned Model (2), similar conclusions can be inferred.

3.2. Analysis of Panel Threshold Regression

3.2.1. wLNY-wlnx1-wLnk1-Threshold Test

First, the single and double thresholds for wlnk1 are presented, along with their 95% confidence intervals. The single-threshold value for wlnk1 is −2.2768, and the double-threshold values are −2.2768 and −2.2872. The detailed 95% confidence intervals can be found in

Table 5. Threshold Effects and 95% Confidence Intervals for wlnk1 Variables.

	Thresholds	95% CI
Single Model	−2.2768	−2.2771	−2.2767
Double Model
Ito1	−2.2768	−2.2771	−2.2767
Ito2	−2.2872	−2.2872	−2.2871

.

According to the results of the bootstrap test for threshold effects presented in

Table 6. Bootstrap Test Results for Threshold Effects of wlnk1 Variables.

Model	Threshold	F	p-Value	BS-Reps	10%	5%	1%
Single-threshold	Single	22.610	0.023	300	15.159	18.977	26.003
Dual-threshold	Single	22.610	0.020	300	12.824	18.044	27.199
	Double	4.600	0.633	300	11.072	13.839	17.117

, the single-threshold effect of wlnk1 passes the significance test at the 5% level, while the double-threshold effect does not pass the significance test at the 5% level. Therefore, the single-threshold effect of variable wlnk1 is significant.

Figure 1 and Figure 2 show the regression results for the single-threshold effect and double-threshold effect of wlnk1. According to the statistical results of the single-threshold effect (1) presented in

Table 7. Regression Results for Single- and Double-Threshold Effects of wlnk1.

	(1)	(2)
	wlny	Wlny
wlnx1	−0.002 ***	−0.002 ***
	(0.000)	(0.000)
wlnm1	0.002	0.002
	(0.002)	(0.002)
wlnm2	−0.005 ***	−0.006 ***
	(0.001)	(0.001)
wlnm3	−0.010 ***	−0.010 ***
	(0.001)	(0.001)
wlnm4	0.004 **	0.004 *
	(0.002)	(0.002)
wlnm5	−0.003 ***	−0.003 ***
	(0.001)	(0.001)
wlnm6	0.027 ***	0.027 ***
	(0.005)	(0.005)
0._cat#c.wlnk1	−0.139 ***	−0.206 ***
	(0.030)	(0.045)
1._cat#c.wlnk1	−0.141 ***	−0.207 ***
	(0.031)	(0.046)
2._cat#c.wlnk1		−0.208 ***
		(0.046)
_cons	−2.599 ***	−2.751 ***
	(0.070)	(0.104)
N	589	589

t statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

, when wlnk1 is less than −2.277, the regression coefficient of wlnk1 to wlny is significantly −0.139. When wlnk1 is greater than −2.277, the regression coefficient of wlnk1 to wlny is significantly −0.141. This indicates a significant threshold effect. The original intention of environmental regulation design is to reduce harmful emissions, improve air quality, and encourage enterprises to adopt cleaner and more efficient production technologies. From this perspective, we expect that implementing stricter environmental regulations on high-haze-pollution industries will lead to better air quality. However, referring to many academic studies on environmental economics and industrial organization, the effect of environmental regulation may not be linear. When the environmental regulation of the thermal power industry and other high-haze-pollution industries is below the threshold, it is too lenient for most thermal power industries and other high-haze-pollution enterprises to produce any significant atmospheric environmental quality improvement. However, once the regulation reaches the threshold, the high-intensity environmental regulation substantially increases the production costs of the thermal power industry and other high-haze-pollution enterprises, reducing their operating profits. In order to maintain short-term profits, enterprises tend to shift the R&D funds originally used for technological innovation to green technological innovations such as desulfurization, de-marketing and dust removal of high-polluting enterprises such as thermal power, which are more conducive to pollution control, thus effectively improving the quality of the atmospheric environment. Moreover, if the regulation becomes too stringent, then there may be a counterproductive effect as the thermal power industry and other high-haze-pollution enterprises may seek other ways to circumvent these regulations or engage in various other methods, such as bribing government officials, to resist the implementation and supervision of the regulations.

In Figure 1 and Figure 2, the dotted line of the threshold inspection chart with LR value of 7.35 represents the critical value at 95% significance level. This means that if the LR statistic exceeds 7.35 at a certain point, we can consider the threshold corresponding to that point to be significant.

The solid line usually represents the change trend of the likelihood ratio statistic (LR statistic) of the threshold variable.

3.2.2. LNY-lnx1-Lnk2-Threshold Test

First, the single and double thresholds for wlnk2 are provided, along with their 95% confidence intervals. The single-threshold value for wlnk2 is 4.737, and the double-threshold values are 4.861 and 4.079. The detailed 95% confidence intervals can be found in

Table 8. Threshold Effects and 95% Confidence Intervals for wlnk2 Variables.

	Thresholds	95% CI
Single Model	4.737	4.688	4.746
Double Model
Ito1	4.861	4.833	4.884
Ito2	4.079	4.036	4.113

.

According to the results of the bootstrap test for threshold effects presented in

Table 9. Bootstrap Test Results for Threshold Effects of wlnk2 Variables.

Model	Threshold	F	p-Value	BS-Reps	10%	5%	1%
Single-threshold	Single	126.650	0.000	300	30.563	37.812	54.126
Dual-threshold	Single	126.650	0.000	300	31.814	41.614	53.580
	Double	45.670	0.010	300	26.730	34.221	45.645

, the single-threshold effect of wlnk2 passes the significance test at the 5% level, and the double-threshold effect also passes the significance test at the 5% level. Therefore, the threshold effect of variable wlnk2 is significant. At the same time, the single-threshold graph and doubl- threshold graph of wlnk2 are shown in Figure 3 and Figure 4.

In Figure 3 and Figure 4, the dotted line of the threshold inspection chart with LR value of 7.35 represents the critical value at 95% significance level. This means that if the LR statistic exceeds 7.35 at a certain point, we can consider the threshold corresponding to that point to be significant.

The solid line usually represents the change trend of the likelihood ratio statistic (LR statistic) of the threshold variable.

Based on the regression results for the single-threshold effect and double-threshold effect of wlnk2, as presented in

Table 10. The Regression Results for the Single-Threshold and Double-Threshold Effects of wlnk2.

	(1)	(2)
	wlny	wlny
wlnx1	−0.002 ***	−0.002 ***
	(0.000)	(0.000)
wlnm1	0.003 *	0.006 ***
	(0.002)	(0.002)
wlnm2	−0.003 **	−0.003 **
	(0.001)	(0.001)
wlnm3	−0.010 ***	−0.010 ***
	(0.001)	(0.001)
wlnm4	0.008 ***	0.009 ***
	(0.002)	(0.002)
wlnm5	−0.003 ***	−0.003 ***
	(0.001)	(0.001)
wlnm6	0.026 ***	0.023 ***
	(0.004)	(0.004)
0._cat#c.wlnk2	0.002 ***	0.003 ***
	(0.000)	(0.000)
1._cat#c.wlnk2	0.001 **	0.002 ***
	(0.000)	(0.000)
2._cat#c.wlnk2		0.001 ***
		(0.000)
_cons	−2.295 ***	−2.280 ***
	(0.025)	(0.024)
N	589	589

t statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

, it is evident that when wlnk2 is less than 4.737, the regression coefficient of wlnk2 to wlny is significantly 0.002. When wlnk2 exceeds 4.737, the regression coefficient of wlnk2 to wlny is significantly 0.001, indicating a significant threshold effect, which aligns with the expectations. Green technological innovation aims to develop and adopt more environmentally friendly and efficient technologies, thereby reducing pollution emissions and environmental degradation. When green technological innovation in the thermal power industry and other high-haze-pollution industries falls below the threshold value, it primarily focuses on fundamental and foundational improvements. These improvements often result in a significant enhancement of atmospheric environmental quality at relatively low costs. Furthermore, in the early stages of technological innovation, the learning effect of green technological innovation in the thermal power industry and other high-haze-pollution industries is significant, enabling substantial improvements in atmospheric environmental conditions with relatively small investments. However, when green technological innovation surpasses the threshold value, due to the deepening of technological innovation, new technological improvements may require increased investments and time. At this stage, the weakened learning effect of green technological innovation in the thermal power industry and other high-haze-pollution industries diminishes the benefits brought about by further technological innovation. Consequently, the effect of further developing green technological innovation in the thermal power industry and other high-haze-pollution industries on improving atmospheric environmental quality is diminished.

3.2.3. LNY-lnx1-Lnk3-Threshold Test

First, the threshold values and their 95% confidence intervals for the single-threshold and double-threshold effects of wlnk3 are provided. The single-threshold value for wlnk3 is 5.949, while the double-threshold values are 5.949 and 7.177. Detailed information regarding the 95% confidence intervals can be found in

Table 11. The Threshold Effect Values and 95% Confidence Intervals for wlnk3 Variables.

	Thresholds	95% CI
Single Model	5.949	5.910	5.964
Double Model
Ito1	5.949	5.910	5.964
Ito2	7.177	7.158	7.178

.

Based on the results of the bootstrap test for variable threshold effects presented in

Table 12. The Results of the Bootstrap Test for wlnk3 Variable Threshold Effects.

Model	Threshold	F	p-Value	BS-Reps	10%	5%	1%
Single	Single	32.760	0.003	300	15.205	19.554	25.528
Double	Single	32.760	0.007	300	17.105	20.767	29.013
	Double	9.610	0.163	300	10.867	13.036	19.206

, it can be observed that the single-threshold effect of wlnk3 passes the significance test at the 5% level, while the double-threshold effect does not pass the significance test at the 5% level. Therefore, the single-threshold effect of the wlnk3 variable is considered significant. The single-threshold graph and doubl- threshold graph of wlnk3 are shown in Figure 5 and Figure 6.

According to the statistical results of the single-threshold effect (1) presented in

Table 13. Regression Results for the Single-Threshold and Double-Threshold Effects of wlnk3.

	(1)	(2)
	wlny	wlny
wlnx1	−0.0020 ***	−0.0019 ***
	(0.000)	(0.000)
wlnm1	−0.0003	−0.0002
	(0.002)	(0.002)
wlnm2	−0.0051 ***	−0.0051 ***
	(0.001)	(0.001)
wlnm3	−0.0102 ***	−0.0103 ***
	(0.001)	(0.001)
wlnm4	0.0042 **	0.0043 **
	(0.002)	(0.002)
wlnm5	−0.0030 ***	−0.0030 ***
	(0.001)	(0.001)
wlnm6	0.0307 ***	0.0304 ***
	(0.005)	(0.005)
0._cat#c.wlnk3	0.0019 *	0.0007
	(0.001)	(0.001)
1._cat#c.wlnk3	0.0011	−0.0001
	(0.001)	(0.001)
2._cat#c.wlnk3		0.0002
		(0.001)
_cons	−2.3123 ***	−2.3020 ***
	(0.025)	(0.025)
N	589	589

t statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

, it is evident that when wlnk3 is less than 5.949, the regression coefficient of wlnk3 to wlny is significantly 0.0019. However, when wlnk3 exceeds 5.949, the regression coefficient of wlnk3 to wlny is not significant and becomes 0.0011. This indicates a significant threshold effect. The observed phenomenon can be attributed to the effective “cleansing” of the air by rainfall when the cumulative precipitation is below the threshold value. During the rainfall process, wet deposition removes particulate matter and certain soluble pollutants from the atmosphere. Increasing the amount of precipitation significantly improves air quality. However, when the accumulated rainfall reaches the threshold value, the “cleansing” effect weakens as most of the removable particulate matter and pollutants in the air have already been cleared. Additionally, under high-precipitation conditions, other factors such as surface runoff and soil absorption start to dominate, gradually diminishing the effectiveness of improving atmospheric environmental quality.

4. Conclusions and Policy Recommendations

4.1. Conclusions

This research investigates the impact of China’s thermal power industry and other high-haze-pollution industries on the atmospheric environment, yielding significant findings regarding the negative correlation between these industries and air quality. The research conclusions not only play an important role in guiding the scientific planning and effective transfer of high-haze-pollution industries such as thermal power in China, but also provide reference for the scientific planning and effective transfer of high-haze-pollution industries such as thermal power in most developing countries in the world. The spatial dynamic Durbin model analysis used in this paper is also suitable for other similar scenes or locations to optimize the national industrial structure and scientifically control haze pollution. The specific conclusions are as follows:

(1) There exists a negative correlation between the thermal power industry and other high-haze-pollution industries and atmospheric environmental quality. Specifically, as these industries expand and their production scales increase, the emission levels of pollutants also rise, consequently leading to the deterioration of the atmospheric environment quality.

(2) The influence of the thermal power industry and other high-haze-pollution industries on atmospheric pollution exhibits spatial spillover effects. Specifically, the reduction of such industries in certain regions not only decreases their own atmospheric pollution but also alleviates the atmospheric pollution in surrounding areas, thereby mutually impacting each other.

(3) Environmental regulations demonstrate a singular threshold effect on atmospheric environmental quality concerning the thermal power industry and other high-haze-pollution industries. This research finds that when the environmental regulations targeting these industries fail to reach the threshold, strengthening them does not yield a sufficiently noticeable effect on improving atmospheric environmental quality. However, once the environmental regulations for these industries meet and surpass the threshold, strengthening the environmental regulations significantly enhances atmospheric environmental quality. This research conclusion aligns with the mainstream views in the academic community. Although environmental regulations may initially be influenced by economic growth and have an opposite effect on environmental quality, the emission reduction effectiveness of environmental regulations will gradually manifest as the economy develops and public environmental awareness improves.

(4) Green technological innovation exhibits a singular threshold effect on atmospheric environmental quality concerning the thermal power industry and other high-haze-pollution industries. This research reveals that below a certain threshold, due to learning and scale effects, the development of green technological innovation in the thermal power industry and other high-haze-pollution industries progressively enhances the improvement of atmospheric environmental quality. However, when green technological innovation surpasses this threshold, further development of such innovation gradually diminishes its effectiveness in improving atmospheric environmental quality.

(5) Cumulative precipitation demonstrates a singular threshold effect on atmospheric environmental quality concerning the thermal power industry and other high-haze-pollution industries. When the cumulative precipitation falls below a specific threshold, an increase in cumulative precipitation gradually strengthens its beneficial impact on air quality. Conversely, when cumulative precipitation surpasses this threshold, its positive effect on atmospheric environmental quality gradually diminishes.

4.2. Policy Recommendations

(1) In regions characterized by severe environmental pollution, it is advisable to implement appropriate measures to reduce the presence of the thermal power industry and other high-haze-pollution industries, while simultaneously optimizing the industrial structure. Firstly, it is imperative to optimize the industrial development layout of enterprises, ensuring harmony with the local natural environment. This entails reducing the proportion of industries that contribute significantly to the thermal power industry and other high-haze-pollution industries, while simultaneously bolstering support for innovative and high-tech industries, thus expediting the transition and upgrading of industries associated with high haze pollution. Secondly, expediting the adjustment of the industrial structure related to the thermal power industry and other high-haze-pollution industries is essential. This can be achieved through scientifically driven capacity reduction measures, such as bankruptcy reorganization and liquidation, aimed at eliminating polluting “zombie” enterprises in regions heavily burdened by haze pollution. Concurrently, industries possessing inherent advantages in terms of products, technology, resources, and location should be guided and enhanced by the government to bolster their competitiveness and foster their transformation and upgrading.

(2) It is advised to strengthen inter-regional cooperation and refine coordinated governance mechanisms for regional haze pollution prevention and control. Considering the phenomenon of spatial spillover resulting from the thermal power industry and other high-haze-pollution industries’ impact on atmospheric environmental quality, it is recommended that administrative authorities establish a legally binding regional leadership organization for joint prevention and control of smoke pollution. These institutions should develop scientifically grounded strategic plans to facilitate the unified management of haze pollution prevention and control within the region. Furthermore, it is essential to enhance policies and regulations for the coordinated prevention and control of haze pollution, clearly delineating responsibilities and authority among various departments. In light of the current state of environmental protection and regional atmospheric environmental quality outlined in the newly revised national policies, the effectiveness of policies for the coordinated prevention and control of regional haze pollution should be optimized.

(3) It is advised to enhance the stringency of environmental regulations in regions afflicted by severe environmental pollution and leverage the threshold effects of environmental regulations. Depending on the severity of pollution in different regions, local governments should implement tailored environmental regulations policies, differentiating the governance of atmospheric pollution. This approach ensures the effective formulation of region-specific policies to avoid the pitfalls of a “one-size-fits-all” approach. To capitalize on the role of diverse environmental regulatory tools in improving atmospheric environmental quality, the government should prioritize market-based incentive policies. Measures such as environmental taxes and emission trading permits should be fully utilized to maximize their external effects. In contrast, for mandatory regulatory tools, such as command-and-control measures that may encounter implementation challenges, the government should devise sound plans to achieve favorable anticipated governance outcomes.

(4) It is advised to facilitate green technological innovation and application in regions heavily burdened by pollution. Local governments should increase support for the research and application of green technologies in the thermal power industry and other high-haze-pollution industries. Firstly, corresponding fiscal subsidies or tax incentives should be provided to these industries, encouraging them to invest more in the R&D of green technologies. This will expedite the improvement of their green technological innovation progress, promote the application of green technologies in production, reduce the costs associated with green technological innovation, and maximize the learning effects derived from such innovations. Secondly, high-quality talents play an indispensable role in the development of green technological innovation. Therefore, local governments should introduce proactive policies to attract and retain high-quality talents, fostering an environment conducive to their retention and influx. Additionally, cultivating the green acumen of high-quality talents will further embed the concept of green development in the hearts and minds of individuals.

(5) It is advised to improve policies related to utilizing increased precipitation to mitigate air pollution. Administrative authorities can maximize the “cleansing” effect of rainfall on atmospheric pollution by implementing appropriate measures. Firstly, local governments must enact relevant policies to enhance the collection and utilization of rainwater, particularly in areas where precipitation falls below the threshold value. Secondly, administrative authorities can explore and promote techniques for artificial rainfall to augment the cumulative precipitation levels in areas failing to meet established standards. This approach aims to achieve the objective of “cleaner” air and improve local atmospheric environmental quality.