Technological Innovation, Urban Spatial Structure, and Haze Pollution: Empirical Evidence from the Middle Reaches of the Yangtze River Urban Agglomeration

Abstract

The rapid economic growth of China has caused significant adverse effects on the environment. Meanwhile, technological innovation, a fundamental driver of economic development and social progress, plays a pivotal role in mitigating haze pollution. This study comprehensively examines the influence of technological innovation on haze pollution in the Yangtze River urban agglomeration, using the STIRPAT model and analyzing research data from 2004 to 2020. Additionally, the study investigates the specific moderating role of urban spatial structure in the relationship between technological innovation and haze pollution. The findings reveal that technological innovation in the middle reaches of the Yangtze River urban agglomeration can effectively curtail haze pollution, and its impact extends to surrounding areas through spillover effect. The polycentric urban spatial structure significantly enhances the haze-reducing effect of technological innovation in the region. Notably, in cities with a strong level of environmental regulation, the urban spatial structure plays a substantial role in augmenting the haze-reducing impact of technological innovation. The policy implications of this research underscore the importance of continuous improvement in technological innovation within the middle reaches of the Yangtze River urban agglomeration. To address haze pollution in future development, the adoption of a polycentric development strategy and the establishment of sound environmental governance policies are recommended.

1. Introduction

The urban agglomeration in the middle reaches of the Yangtze River is experiencing rapid acceleration in both urbanization and industrialization. As a result, the region exhibits high density and wide distribution of population and industries. The haze pollution issue in this region has escalated to a critical level, causing it to become one of the most heavily haze-polluted areas in China. This concerning development is confirmed by the National Urban Air Quality Report for February 2022, which was published by the Ministry of Ecology and Environment. The city clusters situated in the middle reaches of the Yangtze River have been placed among the lowest ranks of the overall index of environmental air quality. Notably, cities like Xiangyang, Jingmen, and Xiaogan occupied particularly unfavorable positions, ranking 163rd, 156th, and 151st, respectively, out of a total of 168 cities assessed across China.

These results demonstrate that atmospheric environmental problems are quite prominent in this region and need urgent attention. Numerous studies have confirmed that technological innovation can help improve the efficiency of energy use and the reduction of environmental pollution while still promoting economic development [1]. In this context, it is crucial to recognize the role of innovation as a pivotal process for achieving effective synergy between economic and social development and promoting environmental sustainability and low-carbon emissions. Consequently, scholars have extensively explored the impact of various urban spatial structures on urban economic growth and innovation efficiency. The debate often centers around the effectiveness of polycentric versus monocentric urban development patterns [2]. Some studies suggest that a polycentric urban spatial structure is effective in promoting technological innovation and knowledge spillovers [3]. Conversely, other studies argue that a polycentric urban spatial structure may hinder the concentration of innovation, potentially leading to a negative impact on technological development [4,5]. Additionally, the association between urban spatial structure and environmental pollution has been a significant area of scholarly concern. Certain research has affirmed that compact urban development patterns are effective in mitigating haze pollution [6]. However, the negative consequences of sprawling development patterns on environmental pollution can be offset by advancements of urban technology [7]. Notably, in China, variations in economic development levels and urban production and lifestyle across different regions introduce uncertainties regarding the influence of urban spatial structure on environmental pollution. Factors such as traffic function layout play a crucial role in shaping this relationship. Urban spatial structure has an impact on both technological innovation and haze pollution, but existing studies have only the relationship between pairs of these elements and have lacked analysis combining the three.

The occurrence of haze pollution is influenced not only by natural factors, such as atmospheric circulation and atmospheric chemistry, but also by anthropogenic factors, including economic and social development [8]. Relevant researchers have extensively explored the various factors influencing haze pollution, examining aspects such as international engagement, population concentration, urbanization, industrial structure, regional spillover, transportation intensity, energy efficiency, financial development, and environmental governance [9,10]. Regrettably, despite extensive discussions on innovation and environmental pollution in the literature, a profound examination of the interplay between urban spatial structure, technological innovation, and haze pollution is still lacking. While some studies have delved into polycentric structures and urban innovation development based on the theory of agglomeration externalities, they contend that polycentric spatial structures impede the agglomeration of intra-city factors, consequently hindering urban innovation and economic growth [5]. Innovation activities are mainly based on the high quality of sharing, matching, and knowledge spillover from agglomeration to enhance urban innovation capacity and promote technological innovation. Therefore, innovation has a higher standard for agglomeration, and the polycentric spatial development model may undermine the agglomeration effect and thus hinder technological innovation [11,12]. Consequently, there is no clear research conclusion on whether the polycentric or monocentric urban spatial structures can promote technological innovation and thus curb haze pollution, and the specific mechanisms involved have not been properly investigated. Currently, amid the continuous urban expansion and development in China, it is of paramount theoretical and practical significance to investigate the processes through which technological innovation and urban spatial structure impact haze pollution. This exploration aims to provide rational guidance for the development of cities and urban agglomerations, harness the potential of urban spatial structure in fostering technological innovation, and more effectively curb haze pollution, thus ameliorating the prevailing state of environmental pollution in urban agglomerations. Optimizing urban spatial structure is an efficacious approach to environmental protection and urban spatial structure offers a more attainable alternative than attempting to curtail pollution emissions by reducing economic development and living standards [13].

In light of the effect of urban spatial structure on technological innovation and haze pollution, this paper addresses the main limitation of existing studies, which solely investigate the relationship between technological innovation, urban spatial structure, and haze pollution without combining all three aspects. Consequently, this study endeavors to examine the influence of technological innovation on haze pollution, its interaction with urban spatial structure, and the variation in the effect of urban spatial structure under varying environmental regulations. Firstly, utilizing the WorldPop population distribution data and employing an exploratory spatial data analysis method (ESDA) to accurately identify primary and secondary population centers in cities, thus measuring the level of polycentric development in the middle reaches of the Yangtze River urban agglomeration. Secondly, a benchmark regression model based on the STIRPAT model was constructed to empirically test the impact of technological innovation on haze pollution and its interaction with urban spatial structure in the region. Thirdly, considering the spatial spillover effect of haze and technological innovation, the study further analyzes the direct and indirect effects of technological innovation and its interaction with urban spatial structure in the middle reaches of the Yangtze River were further analyzed using the spatial model. The aim of this comprehensive exploratory research is to elucidate the specific effects of technological innovation and urban spatial structure on haze pollution in the cities within the middle reaches of the Yangtze River. This endeavor seeks to identify urban spatial structure governance policies that best suit the region and provide the necessary empirical support and decision-making foundation for formulating and effectively implementing haze control policies.

The paper is organized as follows. Section 2 introduces the hypothesis, drawing from a comprehensive review of the related literature. Section 3 provides an overview of the research data and methodology. The empirical results are examined and analyzed in Section 4 and Section 5. Finally, Section 6 presents a summary of the primary research findings and offers pertinent policy recommendations.

2. Literature Review and Theoretical Hypothesis

2.1. The Influence Mechanism of Technological Innovation on Haze Pollution

Technological innovation is a significant driving force for promoting green and sustainable development, and plays a pivotal role in influencing efficient energy use developments [14]. Notably, green technological innovation contributes significantly to enhancing energy efficiency within the production process, leading to a total reduction in energy emissions [15,16]. From an energy perspective, technological innovation demonstrates a preventive effect on haze pollution by diminishing its occurrence and enhancing overall environmental quality. In the context of haze pollution control, technological innovation primarily curtails total haze pollution through two distinct avenues. Firstly, enterprises adopt green technological innovation to renovate their production methods, optimize production, develop clean production technologies, and enhance energy utilization rates in the production process. These measures effectively address the sources of haze pollution, thereby reducing the complexities of managing haze pollution at its later stages [17,18]. The second is the effective prevention of haze through technological innovation, for example, enterprises reduce future haze pollution by improving or retrofitting emission treatment equipment to achieve haze control effects [19]. Khazzoom [20] and Berkhout et al. [21] have pointed out that technological innovation, while improving energy efficiency by enhancing the performance of production equipment, can also result in reduced unit production costs and subsequently lead to increased demand and use of production equipment, thereby potentially elevating overall energy consumption. As a consequence, they argue that technological progress can promote haze pollution to some extent.

Moreover, technological innovation plays a role in influencing haze pollution by inducing changes in industrial processes. Through both supply and demand channels, technological innovation impacts the input and output of industries and the efficiency of production resource allocation and renovation. This influence facilitates the transformation and upgrading of the industrial process, subsequently leading to an indirect reduction in haze pollution. Technological innovation directly affects the industrial process from the supply side by providing a highly skilled workforce, upgrading the production technology base, and improving management capabilities. Indirectly, technological innovation affects the industrial structure through production demand, consumer demand, and export, thus influencing the proportion, scale, and development rate of industrial processes [22]. The optimization of industry emerges as a critical factor in haze pollution, as the proportionate representation and development rate of each industry closely relate to haze generation [23]. By optimizing the industrial process, the production process can be refined, resulting in the reduction of air pollutants and haze pollution [24].

As technological advancements and industrial structure upgrades occur in a city or region, a concentration of R&D personnel tends to form in the innovation department, leading to a notable population agglomeration effect. Previous studies have come to two different conclusions regarding the link between population agglomeration and haze. Some researchers have pointed out that innovative talents and enterprises tend to concentrate in innovative cities or urban centers with better transportation and infrastructure, therefore increasing the size and density of the regional population within the region. Considering the scale effect of population agglomeration, it is evident that such agglomeration results in increased demand for housing, home appliances, and motor vehicles, subsequently leading to amplified production output by enterprises and a consequent rise in energy consumption, thereby exacerbating the occurrence of haze pollution. Conversely, the agglomeration effect resulting from population concentration can promote a decrease in individual car usage by encouraging the use of public transportation, enhancing resource utilization efficiency, mitigating air pollutant emissions, and effectively alleviating haze pollution [6,8]. Zhu et al. further confirm that the scale effect of population agglomeration has significant importance in the development of urban agglomerations in the middle reaches of the Yangtze River [25].

In addition to the direct effects mentioned above, technological innovation also has indirect effects on haze pollution through technology spillover, which helps mitigate haze pollution in neighboring cities [16].

According to the influence mechanism mentioned above, we have proposed the following hypothesis.

Hypothesis 1. 

Technological innovation contributes to the reduction of haze pollution in urban agglomerations.

2.2. Moderation Effect of Urban Spatial Structure

Spatial structure concerns the distribution of population and economic activities within a specific area and constitutes a crucial determinant of resource utilization efficiency in a region [26]. Therefore, urban spatial structure pertains to the spatial distribution of population, resources, and economic activities within a city. It can be categorized into two types, which are monocentric and polycentric [13]. Monocentric spatial structure is characterized by an uneven distribution of economic activities, such as population, resources, and industries, with a dominant central core serving as the primary center of the city. By contrast, the polycentric spatial structure exhibits a more evenly distributed pattern, with fewer disparities in functional connections between different regions. The evolution of urban spatial structure directly impacts the city’s environment and neighboring areas. Notably, prior research has focused on exploring the impact of population and economic activity concentration as well as dispersion at various scales on haze pollution, aiming to investigate the relationship between urban spatial structure and haze pollution.

Some scholars argue that monocentric and compact spatial structures, with their more densely distributed population and economic activities, facilitate optimal resource agglomeration and allocation, leading to reduced household energy consumption and shorter commuting distances. Consequently, these factors effectively contribute to lower haze pollution levels within cities [6,27]. Conversely, other studies suggest that the concentration of population and economic activity resulting from monocentric development patterns hampers the reduction of haze pollution [28]. The polycentric city network can expand the geographical space of a single large city, creating dynamic externalities [29]. This facilitates improved energy efficiency and reduced pollution emissions through economies of scale and technological spillovers [30]. As a decentralized agglomeration, polycentric spatial structures enhance resource flow and market integration through inter-city interactions, thereby accelerating technology spillovers and knowledge sharing between regional cities. Consequently, this optimizes the efficiency of production resource allocation, reduces pollutant emissions, and mitigates haze pollution. Therefore, the polycentric development model not only fosters positive spillover effects by promoting regional market integration but also facilitates the optimization of industrial structure and production resource allocation efficiency, thus reducing pollutant emissions such as haze pollution and advancing haze control efforts [13]. Furthermore, it is evident that the influence of the urban spatial structure on haze pollution varies among cities of different regions [31]. This variation can be attributed to differences in urban production, lifestyle, industrial layout, transport planning, and land use among cities with different population densities [32]. Additionally, at larger spatial scales, such as city clusters and provinces, no consistent findings on the relationship between spatial structure and environmental pollution have been observed.

Regarding the correlation between urban spatial structure and innovation, some studies highlight that at the level of urban agglomerations, the polycentric spatial structure weakens market boundaries between cities, facilitating the flow of production resources and knowledge spillovers [33]. In a polycentric city network, polycentric agglomerations positively impact urban innovation by optimizing industrial structures and technological advancements. These agglomerations foster synergistic development of industries and efficient resource utilization, effectively enhancing the spillover and integration of information and technology, thereby improving the total productivity of all sectors.

Based on the previous discussion, the following hypothesis was formulated.

Hypothesis 2. 

Urban spatial structure has a moderating effect on technological innovation and haze pollution, with polycentric urban spatial structure especially enhancing the effect of technological innovation in reducing haze pollution.

2.3. Heterogeneity in the Impact of Technological Innovation on Haze Pollution

Governmental environmental regulation significantly influences regional environmental improvement [34]. On one hand, environmental regulations impose constraints on enterprises, compelling them to adhere to environmental standards and prioritize environmental protection measures. Some enterprises with advanced technological resources apply green innovation results to production through innovation compensation, so as to effectively avoid the compliance cost brought by environmental regulations.

On the other hand, government environmental regulations can lead to the imposition of emission fees and environmental taxes, which subsequently increase the pollution treatment costs for enterprises. This additional financial burden may prompt some companies to consider relocating to regions with less stringent environmental regulations [35,36,37]. As a consequence, local pollution-intensive production activities may be reduced, promoting the transformation and upgrading of the industrial process and enhancing environmental quality in the region. Moreover, this may also induce changes in the spatial structure of the city to some extent. Additionally, the intensification of environmental regulations can lead to the displacement of companies with lower technological capabilities, ultimately enhancing the concentration of industries in the region. This concentration promotes knowledge diffusion and spillover effects, thus contributing to an overall improvement in technological advancement. Consequently, the development of polycentric agglomerations can effectively mitigate the generation of haze pollution in regions with robust environmental regulations.

Based on the previous discussion, the following hypothesis was formulated.

Hypothesis 3. 

In regions with high levels of environmental regulation, the influence of technological innovation on haze pollution is relatively pronounced in areas where the spatial structure of cities tends to be polycentric.

3. Methods and Data

3.1. Model

The IPAT model, initially proposed by Ehrlich and Holdren [38], is a widely adopted framework for analyzing the influence of human activities on environmental pollution. This model is represented by the fundamental equation I = PAT, where I denotes pollution, P represents population, A signifies the level of affluence, and T signifies the level of technological advancement. Subsequently, Dietz and Rosa [39] extended the IPAT model to create the STIRPAT model, which is frequently employed for analyzing the factors that impact haze pollution. The foundational formula is as follows:

$$\mathrm{I}={\alpha \mathrm{P}}^{\beta 1}{\mathrm{A}}^{\beta 2}{\mathrm{T}}^{\beta 3}\epsilon$$

(1)

The annual average concentration of ${\mathrm{PM}}_{2.5}$ (PM) was chosen to measure the environmental impact ($\mathrm{I}$) in the model. The population (P), wealth (A), and technology level (T) are measured by population density (pop), GDP per capita (gdp), and technological innovation (tech). Hence, model (1) is extended according to the needs of this paper, and the model is adjusted as follows:

$$\ln P{M}_{it}={\alpha }_0+{\alpha }_1\ln tec{h}_{it}+{\alpha }_{2}pol{y}_{it}+{\alpha }_3(tec{h}_{it}\times pol{y}_{it})+{\alpha }_4{X}_{it}+{\epsilon }_{it}$$

(2)

In the formula, $i$ is the cross-sectional unit of the 27 prefecture-level cities in the middle reaches of the Yangtze River urban agglomeration, and $t$ denotes the year. $P{M}_{it}$ denotes the haze pollution in city i in year t, and $tec{h}_{it}$ denotes the level of technological innovation in city i in year t, and $pol{y}_{it}$ denotes the spatial structure of city i in year t. Technological innovation ($tec{h}_{it}$), urban spatial structure ($pol{y}_{it}$), and technological innovation and spatial structure ($tec{h}_{it}\times pol{y}_{it}$) interaction terms are the core explanatory variables. $X$ is a set of related control variables. ${\alpha }_0$–${\alpha }_4$ are the coefficients to be estimated. $\epsilon$ is the random disturbance term. Since both variables of the interaction term are continuous variables, in order to make the coefficient (${\alpha }_1$) of the first term have a more intuitive meaning, we centralized both variables of the interaction term [40]; therefore, model (2) was extended as follows:

$$\ln P{M}_{it}={\alpha }_0+{\alpha }_1\ln tec{h}_{it}+{\alpha }_{2}pol{y}_{it}+{\alpha }_3(tec{h}_{it}-{\overline{tech}}_{it})\times \left(pol{y}_{it}-{\overline{poly}}_{it}\right)+{\alpha }_4{X}_{it}+{\epsilon }_{it}$$

(3)

where ${\overline{tech}}_{it}, {\overline{poly}}_{it}$ denote the mean value of technological innovation ($tec{h}_{it}$) and the spatial structure of the city (${\mathit{poly}}_{\mathrm{it}}$). The interaction term after centralization is $(tec{h}_{it}-{\overline{tech}}_{it})\times \left(pol{y}_{it}-{\overline{poly}}_{it}\right)$.

Since the urban spatial structure of cities within the middle reaches of the Yangtze River urban agglomeration has shown minimal change over time, the OLS estimation that controls for individual fixed effects is adopted for empirical analysis.

Secondly, a spatial Durbin model was formulated by integrating the STIRPAT model, designed to examine the spatial spillover of technological innovation, and the interaction term between technological innovation and urban spatial structure on the impact of haze pollution. The equation is as follows:

$$y_{it}=\rho w_i{y}_i+x_{it}\beta +w_i{x}_{it}\delta +u_i+{\gamma }_t+{\epsilon }_{it}$$

(4)

According to the requirements of this paper, we expanded model (4) as follows:

$$\begin{array}{l}\ln P{M}_{it}=\alpha +{\rho }_1{\displaystyle {\displaystyle \sum }_{i=1}^n}{W}_{it}\ln P{M}_{it}+{\beta }_1\ln tec{h}_{it}+{\rho }_2{\displaystyle {\displaystyle \sum }_{i=1}^n}{W}_{it}\ln tec{h}_{it}+{\beta }_{2}pol{y}_{it}+{\rho }_3{\displaystyle {\displaystyle \sum }_{\mathrm{i}=1}^n}{W}_{it}pol{y}_{it}+{\beta }_3(tec{h}_{it}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}-{\overline{tech}}_{it})\times \left(pol{y}_{it}-{\overline{poly}}_{it}\right)+{\rho }_4{\displaystyle {\displaystyle \sum }_{\mathrm{i}=1}^n}{W}_{it}(tec{h}_{it}-{\overline{tech}}_{it})\times \left(pol{y}_{it}-{\overline{poly}}_{it}\right)+\theta \sum X_{it}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+\delta {\displaystyle {\displaystyle \sum }_{i=1}^n}{W}_{it}{X}_{it}+{\epsilon }_{it}\end{array}$$

(5)

where $W_{it}$ is a matrix of spatial weights, $X_{it}$ is the set of control variables selected for the study, $\beta$ and $\theta$ denote the influence coefficient of each variable, $\rho$ and $\delta$ are the spatial weight coefficients, and ${\epsilon }_{it}$ is the random disturbance term.

3.2. Variables

3.2.1. Explanatory Variables

The primary explanatory variable in this study is haze pollution, represented by ${\mathrm{PM}}_{2.5}$. Haze pollution is a fundamental indicator used to assess the severity of haze pollution within China [41]. To measure haze pollution in the middle reaches of the Yangtze River, the annual average concentration values of ${ \mathrm{PM}}_{2.5}$ for each prefecture-level city were selected from 2004 to 2020. These data were sourced from the Dalhousie University Atmospheric Composition Analysis Group (DAUAG), which combines direct ground-based data with indirect satellite data. This approach provides a highly accurate and objective measurement of haze pollution levels in the region. Using the published ${\mathrm{PM}}_{2.5}$ concentration grid data, the Arcgis 10.8 software was employed to analyze the corresponding data for each prefecture-level city. A higher value of ${\mathrm{PM}}_{2.5}$ indicates more severe levels of haze pollution. Figure 1 illustrates the haze pollution in the middle reaches of the Yangtze River, with most cities showing a decreasing trend in haze pollution over time. Due to space constraints, this paper only presents the calculation results for a four-year period.

3.2.2. Core Explanatory Variables

Technological innovation (tech) is typically assessed based on both input and output indicators [42,43]. Regarding output indicators, patent data represent the most commonly used index for measuring innovation output, which can be further divided into the number of patent applications and the number of patent grants. The number of patent applications reflects the level of active innovation activities and the willingness of individuals in a region to seek patent protection. Meanwhile, the number of patent grants indicates whether the declared patents meet certain technological innovation standards and have received official authorization. Therefore, the number of patent grants provides a more accurate reflection of the actual level of innovation capacity. In this study, the number of patent grants is chosen as the measurement index for technological innovation, and it is expected to have a negative coefficient.

In accordance with the research of Wang et al. [44], the method used to identify the spatial structure of the city (poly) is based on the morphological perspective, which is measured according to the population distribution. Based on the principle of data availability, raster data of the average distribution of population within the study area in each year were obtained from WorldPop. This population distribution data reflect the 24 h average population distribution within the city, represent information collected on natural conditions and human activities within the area, cover a range of information such as night-time lighting, employment, and traffic, and aggregate the economically active population in a geographical unit into a meta-grid of one square kilometer per unit surface area. Therefore, by observing the number of people in each grid, it can directly capture the population value of the subcenter inside the city. These data are extensively utilized by scholars in the field of urban economics and provide a highly accurate reflection of the distribution of economic activities within cities.

The identification of each population center in a city means the identification of the main and the subcenter of population in each city. According to the principle of non-parametric measurement, the interaction between urban center (main/subcenter) and nearby areas and the comparability of population center (main/subcenter) are considered, and the population center (main/subcenter) is determined by the ESDA method. At the city level, an analysis of the local spatial autocorrelation pattern of population density within each city allows for the measurement of the Local Moran’s I index for each raster. This index reflects the spatial clustering of WorldPop rasters with different population densities. Subsequently, the rasters with higher population density and higher surrounding population density are selected and retained as candidate rasters for identifying population centers. Finally, the population centers were established based on the principle that the raster clusters must be extensive and high in total population.

Using the methods mentioned above, this paper identifies the population centers, including main centers and subcenters, of each city in the middle reaches of the Yangtze River from 2004 to 2020. The grids contained in each population center are aggregated to calculate the population of these centers.

Furthermore, to assess the degree of urban polycentricity from a morphological perspective, the proportion of population in each subcenter ($po{p}_{subcenter}$) relative to all population centers is calculated. This proportion represents the importance of subcenters compared to the main centers of a city. A higher value of this indicator indicates a greater proportion of population in the city’s subcenters, signifying the greater importance of these subcenters and indicating the value of a more polycentric urban structure. The exact formula for calculating this indicator is as follows:

$$poly=\frac{{{\displaystyle pop}}_{subcenter}}{{{\displaystyle pop}}_{subcenter}+{{\displaystyle pop}}_{maincenter}}$$

(6)

Figure 2 and Figure 3 illustrate the trends of technological innovation and urban spatial structure in the middle reaches of the Yangtze River among different cities. Overall, technological innovation shows greater advancements in 2020 compared to 2004. Additionally, there have been some changes in urban spatial structure within the middle reaches of the Yangtze River over the same period.

3.2.3. Other Control Variables

Firstly, based on the STIRPAT model, economic growth and population were selected separately as the control variables for this study.

Economic development (eco) is recognized as one of the key contributors to haze pollution [45]. Therefore, the GDP per capita of each city in the middle reaches of the Yangtze River urban agglomeration during the study period from 2004 to 2020 was selected as the measurement index. It is anticipated that a larger economic volume will result in increased resource consumption and pollutant production; thus, the coefficient is expected to be negative.

Similarly, population level (pop) is identified as a significant factor influencing environmental pollution [46,47]. Higher population density can generate economies of scale, improve energy efficiency, and reduce pollutant emissions, ultimately leading to a reduction in haze pollution concentration. However, human socio-economic activities are also closely related to the increase in haze pollution. Consequently, the population measurement per unit area was chosen to represent population density, and the expected sign of the coefficient is uncertain.

Furthermore, several other factors closely related to haze pollution are taken into consideration as control variables in this paper.

It is indicated in the literature that pollution emissions from outdated and underdeveloped manufacturing industries are more extreme, while modern innovation in service industries can lead to emission-reduction effects. This demonstrates that low-quality industrial processes can cause excessive haze pollution, especially in secondary industries [8]. Consequently, the proportion of GDP from secondary industries was selected as a measure of urban industrial advancement (ind) and has an expected positive coefficient.

The degree of openness to the outside world (open) reflected by foreign direct investment (FDI) is a fundamental factor to be considered in the study of environmental pollution in China [48]. However, there are differing findings on the impact of openness on haze pollution. Some studies suggest that FDI leads to a “pollution paradise” by transferring highly polluting industries to less regulated countries, while others propose a “pollution halo” effect, indicating that FDI can improve environmental quality by introducing eco-friendly technologies and products [49]. Thus, the proportion of actual FDI utilized in a year to GDP was chosen as a measure of the level of opening up and has an uncertain expected coefficient.

To assess the intensity of environmental governance by local governments, the occurrence of environment-related words in the work reports of prefecture-level municipalities was selected as an environmental regulation (env) indicator [50,51]. Unlike previous studies that focus on specific indicators like the number of environmental protection personnel and R&D investment, this approach provides a more comprehensive view of local governments’ environmental governance policies. The frequency and proportion of environment-related words in the Government Work Reports of prefecture-level city governments are selected as indicators to measure environmental regulations and have an expected negative coefficient.

Motor vehicle emissions from transportation (tri) are a significant source of haze pollution [10,52]. Accordingly, the number of vehicles owned at the prefectural level was selected as the measurement index of transportation factors to observe its impact on haze pollution and has an expected positive coefficient.

3.2.4. Data Sources and Processing

Data on industrial structure, openness to the outside world, transportation, economic development, population density, and technological innovation were sourced from China City Statistical Yearbook (2004–2020) and local statistical bureaus of each city. Data on urban spatial structure, population, and haze pollution were obtained from the Atmospheric Composition Analysis Group of Dalhousie University, Canada and WorldPop raster data, respectively. Data on environmental regulations of each city were obtained from the government reports of prefecture-level cities. The descriptive statistics of the variables are shown in

Table 1. Descriptive statistics for each variable.

Variable	Obs	Mean	Std. Dev.	Min	Max	Metric or Description	Unit
PM2.5	459	47.69	10.94	22.96	79.04	Annual average PM2.5 values (in logarithms)	µg/${\mathrm{m}}^3$
Technological innovation (tech)	459	2311	5077	28	58,923	Number of patents granted (in logarithms)	Pieces
Urban spatial structure (poly)	459	0.317	0.233	0	0.822	Degree of urban population dispersion (not logarithmic)	%
Population density (pop)	459	431.2	175.0	136.4	985.1	Population density (in logarithms)	People/${\mathrm{km}}^2$
Economic development (eco)	459	39,197	27,957	4787	145,545	GDP per capita (in logarithms)	Yuan
Openness to the outside world (open)	459	2.430	1.553	0.0921	8.030	Share of actual foreign investment utilized in the year in GDP (not logarithmic)	%
Industrial structure (ind)	459	48.97	7.591	28.56	66.99	Secondary sector value added as a proportion of GDP (not logarithmic)	%
Environmental regulation (env)	459	0.00310	0.00158	0	0.0100	Word frequency statistics for environmental protection in prefecture-level municipal government work reports (not logarithmic)
Transportation (tri)	459	404.7	497.0	0	3269	Number of cars owned by prefecture-level municipalities nationwide (taken as logarithm)	Thousands of vehicles

below. In the empirical process, numerical data underwent logarithmic processing, while ratio data did not. Additionally, logarithmic processing was not applied to the urban spatial structure data due to some city values being zero.

4. Results

4.1. Spatial Autocorrelation Test

To conduct a comprehensive analysis of the spillover effects of haze pollution, a region-wide spatial correlation index was employed for testing. This involved detecting the spatial correlation of haze pollution among the 27 prefecture-level cities in the middle reaches of the Yangtze River urban agglomeration. For this purpose, two spatial weighting matrices were constructed.

The first weighting matrix, known as the geographical distance weight matrix, took into account the geographical distance between prefecture-level cities. Specifically, it used the inverse square of the distance between the geographical centers of city i and city j [53], which is expressed as follows:

$$W_1=\left\{\begin{array}{c}\frac{1}{d_{ij}^2} i\ne j\\ 0 i=j \end{array}\right.$$

(7)

where $d_{ij}$ is the distance between the geographical centers of province i and province j.

To account for the spatial correlations between regions arising from both geographic and economic factors, the nested weight matrices of geographic and economic distances were considered. This approach considers the radiative effects of geographic distance and economic factors [18]. The formula is as follows:

$$W_2=W_1·diag\left(\frac{\overline{Y_1}}{\overline{Y}},\frac{\overline{Y_2}}{\overline{Y}},\dots ,\frac{\overline{Y_n}}{\overline{Y}}\right)$$

(8)

To reveal the agglomeration pattern of local haze pollution and technological innovation level in the research area,$W_2$ was taken as an example. Local Moran scatterplots of ${\mathrm{PM}}_{2.5}$ and technological innovation level were analyzed, indicating that both haze pollution and technological innovation in the urban agglomeration in the middle reaches of the Yangtze River exhibit a significant spatial spillover effect.

4.2. Analysis of Benchmark Regression

4.2.1. Analysis of Technological Innovation and Haze Pollution

Based on the results of the Hausman test, a fixed-effects model is chosen for the benchmark regression. The estimated results of technological innovation and the impact of interaction terms between technological innovation and urban spatial structure on haze pollution are presented in

Table 2. Benchmark regression results.

Explanatory Variables	Explained Variable: lnPM2.5
	(1)	(2)	(3)	(4)
lnx1	−0.0940 ***	−0.0566 ***	−0.0585 ***	−0.0539 ***
	(0.0000)	(0.0002)	(0.0001)	(0.0003)
x2			−0.3219 *	−0.4100 **
			(0.0645)	(0.0191)
interact				−0.0495 ***
				(0.0022)
lnx3		0.3654 ***	0.3597 ***	0.3980 ***
		(0.0007)	(0.0008)	(0.0002)
lnx4		−0.1194 ***	−0.118 ***	−0.1227 ***
		(0.0001)	(0.0001)	(0.0001)
x5		−0.0235 ***	−0.0222 ***	−0.0170 **
		(0.0005)	(0.0010)	(0.0138)
x6		0.0191 ***	0.0189 ***	0.0195 ***
		(0.0000)	(0.0000)	(0.0000)
x7		10.8764 **	11.1286 ***	11.4486 ***
		(0.0067)	(0.0055)	(0.0039)
lnx8		0.0156	0.0178	0.0162
		(0.4381)	(0.3770)	(0.4168)
Constant	4.4594 ***	2.2586 ***	2.3902 ***	2.1664 ***
	(0.0000)	(0.0004)	(0.0002)	(0.0006)
Observations	459	459	459	459
$R^2$	0.420	0.724	0.726	0.732
adj. $R^2$	0.384	0.702	0.704	0.710
Urban effects	Control	Control	Control	Control

Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. The values in parentheses are p values.

, while individual fixed effects are controlled in each column. In column (1) of Table 2, the coefficient of technological innovation (lnx1) is −0.0940, signifying statistical significance at the 1% level without the inclusion of control variables. This result suggests that a 1 unit increase in the level of urban technological innovation is associated with a 9.4% reduction in haze pollution.

Subsequently, additional control variables are introduced in columns (2) and (3) of the regression analysis. Remarkably, the regression coefficients of urban technological innovation (lnx1) remain significantly negative at the 1% level. For instance, considering the regression coefficient of urban technological innovation in column (3), a 1% increase in the technological innovation index of each city in the middle reaches of the Yangtze River urban agglomeration is associated with a reduction in haze pollution of 0.0585%. This result confirms hypothesis 1, which suggests that technological innovation effectively reduces the generation of haze pollution, consistent with Shi and Zhang’s findings [16].

The control variables also yield some valuable insights. Population density, economic development, openness level, industrial structure, and environmental regulations all significantly affect haze pollution at the 1% level, while transport has no significant effect on haze pollution. This aligns with Chen et al.’s study, which indicated that regional differences within the study area led to transport having no significant effect on haze pollution [54]. The increase in economic development and openness of cities in the middle reaches of the Yangtze River can effectively suppress haze pollution, while population density, industrial structure, and environmental regulations exacerbate haze pollution in the region. Shao et al. also confirmed that the scale effect generated by population density and the increase in the proportion of secondary industries would significantly contribute to haze pollution. The “pollution haven” hypothesis for the openness to the outside world is not confirmed in China, but the improvement of openness to the outside world tends to benefit the control of haze pollution [8].

4.2.2. Analysis of Regression Results of Moderating Effects

In column (4) of Table 2, the regression results include the interaction term between technological innovation and urban spatial structure. Both the interaction term between technological innovation and urban spatial structure (interact) and technological innovation (lnx1) pass the significance test at the 1% level, suggesting that there is a significant moderating effect of urban spatial structure in the relationship between technological innovation and haze pollution. Moreover, the regression coefficient of the interaction term is the same sign as the regression coefficient of urban technological innovation (lnx1), which strongly indicates that urban spatial structure significantly enhances the influence of technological innovation on haze pollution, and has a significant positive moderating effect on the haze reduction effect of technological innovation. This finding validates hypothesis 2, suggesting that urban spatial polycentricity enhances the haze-reducing effect of technological innovation. This finding aligns with Wu et al.’s research, which also highlighted that a polycentric spatial structure can effectively mitigate haze pollution through technological innovation [55]. With the increase in urban scale in the middle reaches of the Yangtze River urban agglomeration, the gradual enrichment of innovation promotes the continuous development of technological advancement, thus mitigating to a certain extent the negative effect of polycentricity on technological innovation [44]. Meanwhile, the low-density development caused by the polycentric urban spatial structure reduces the concentration of people’s lives and economic activities, which directly reduces the generation of haze pollution. Moreover, polycentric urban spatial structure, congestion, and commuting times are reduced in the main centers within the city, which directly reduces haze pollution caused by transport [56]. A polycentric urban development pattern may also promote the formation of industrial agglomerations within the city, enhancing specialization and knowledge spillover from the same industry, thus promoting technological innovation [44]. With the development of multi-centers in cities in the region, technology spillover and knowledge sharing promote the optimization of production factor allocation efficiency [56]. Specialized division of labor and regional cooperation also contribute to the optimization of the efficiency of the allocation of production factors, thereby reducing haze emissions [57].

4.2.3. Analysis of Spatial Spillover Effects

Prior to estimating the spatial econometric model parameters, the (robust) LM test is usually used to ascertain whether the model corresponds to a spatial lag model or a spatial error model. The criterion used is that the model with the more significant LM statistic is the preferred one. In cases where both models exhibit the same level of significance in their LM statistics, the significance of the LM statistic becomes crucial in determining the model’s setup. Equation (3) underwent the LM test to determine the appropriate form of estimation for the spatial econometric model, and the results are presented in

Table 3. Results of the spatial correlation test.

	W1	W2
Moran’s I (Spatial error)	9.942 ***	10.518 ***
Lagrange multiplier (Spatial error)	80.556 ***	90.768 ***
Robust Lagrange multiplier (Spatial error)	26.428 ***	34.938 ***
Lagrange multiplier (Spatial lag)	277.016 ***	292.420 ***
Robust Lagrange multiplier (Spatial lag)	222.887 ***	236.590 ***
Hausman	437.73 ***	260.47 ***

Note: *** indicate significance at the 1% levels, respectively.

. The findings indicate that the null hypothesis of no spatial lag and no spatial lagged error term can be rejected at the 1% significance level. Therefore, it is evident that spatial lag and spatial error should be taken into consideration when constructing regression models with spatial effects. The Hausman test significantly rejects the null hypothesis, which indicates that the spatial Durbin model with fixed effects should be applied for the spatial analysis.

As haze pollution, technological innovation level, and urban spatial structure are influenced not only by geographical distance within the region but also by economic development; the spatial Durbin model under the nested weight matrix (W2) of geographical and economic distance is chosen for spatial effect analysis. The direct, indirect, and total effects of the interaction between technological innovation level, technological innovation level, and urban spatial structure are analyzed. On the other hand, as the spatial structure of cities in the middle reaches of the Yangtze River urban agglomeration varies less over time and the differences exist mainly among cities in the middle reaches of the Yangtze River urban agglomeration, the spatial spillover effect analysis is chosen to control for individual city effects, as in the previous benchmark regression.

The results of the spatial spillover effect are shown in

Table 4. Results of spatial spillover effects.

Explanatory Variables	(1)	(2)	(5)	(6)	(7)
	Main	Wx	LR_Direct	LR_Indirect	LR_Total
lnx1	−0.000	−0.174 ***	−0.009	−0.398 ***	−0.407 ***
	(0.952)	(0.000)	(0.249)	(0.000)	(0.000)
x2	−0.064	3.592 ***	0.124	8.116 ***	8.239 ***
	(0.445)	(0.000)	(0.267)	(0.000)	(0.001)
interact	−0.008	−0.476 **	−0.032 *	−1.058 *	−1.090 *
	(0.342)	(0.033)	(0.081)	(0.067)	(0.066)
lnx3	−0.035	−0.045	−0.041	−0.145	−0.187
	(0.555)	(0.922)	(0.548)	(0.898)	(0.875)
lnx4	−0.014	0.174 **	−0.005	0.374 **	0.368 **
	(0.617)	(0.017)	(0.837)	(0.023)	(0.030)
x5	−0.001	−0.015	−0.001	−0.036	−0.037
	(0.841)	(0.435)	(0.702)	(0.409)	(0.411)
x6	0.000	0.009 ***	0.001	0.021 ***	0.022 ***
	(0.692)	(0.000)	(0.350)	(0.000)	(0.000)
x7	−2.965	8.516	−2.632	17.072	14.440
	(0.111)	(0.149)	(0.149)	(0.207)	(0.307)
lnx8	0.002	0.031	0.004	0.076	0.080
	(0.850)	(0.598)	(0.688)	(0.591)	(0.587)
rho	0.575 *** (0.000)
sigma2_e	0.002 *** (0.000)
Observations	459	459	459	459	459
$R^2$	0.440	0.440	0.440	0.440	0.440
Number of id	27	27	27	27	27

Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. The values in parentheses are p values.

below. Although technological innovation in the middle reaches of the Yangtze River urban agglomeration does not have a significant impact on the reduction of local haze, the spillover effect of technological innovation from neighboring regions is more obvious. This phenomenon suggests that the haze reduction effect of technological innovation in the middle reaches of the Yangtze River urban agglomeration is mainly due to the technological spillover from neighboring regions. The technological innovation from neighboring regions provides a complement to the local technological innovation, while the local technological innovation level cannot achieve a better haze reduction effect, and then has a positive effect on the local haze reduction effect. Shi and Zhang also confirmed in their study that technological innovations can have a positive effect on haze pollution control in neighboring regions through “technological spillover effects” [16].

The interaction between local and neighboring urban spatial structures and technological innovation has a significant moderating effect, indicating that both local and neighboring polycentric urban spatial structures can effectively contribute to the haze reduction effect of technological innovation in the middle reaches of the Yangtze River urban agglomeration. Compared with the local polycentric spatial structure, the polycentric urban spatial structure in the neighboring areas can enhance the haze reduction effect of technological innovation in the middle reaches of the Yangtze River urban agglomeration. The reason for this is likely to be that the polycentric spatial structure of the city cluster in the middle reaches of the Yangtze River has weakened the boundaries between cities, strengthened the cooperation and division of labor between cities, and made the interactive flow of factors and the diffusion of knowledge spillover more frequent and the advantages of technological innovation more significant in the polycentric network [58,59]. Under the influence of the rationalization of industrial structure and the improvement of technological innovation level, pollutant emissions and haze pollution have been reduced [55].

4.2.4. Environmental Regulation Heterogeneity

In this study, the median environmental regulation-related word frequency was used to characterize environmental regulation. Cities with environmental regulation-related word frequency values greater than the median at the prefecture level were classified as cities with strong environmental regulation, while cities with environmental regulation-related word frequency less than the median at the prefecture level were classified as cities with weak environmental regulation. Regression analyses were conducted for each of them, and the regression results are shown in

Table 5. Baseline regression results under strong and weak environmental regulation.

Explanatory Variables	Explained Variable: lnPM2.5
	Weak Environmental Regulation	Strong Environmental Regulation
lnx1	−0.0516 **	−0.0511 **
	(0.0205)	(0.0193)
x2	−0.8504 **	−0.3960 *
	(0.0108)	(0.0995)
interact	−0.0021	−0.0925 ***
	(0.9324)	(0.0008)
lnx3	0.3114 **	0.1819
	(0.0376)	(0.2706)
lnx4	−0.1081 **	−0.1373 ***
	(0.0123)	(0.0087)
x5	−0.0227 **	−0.0200 *
	(0.0171)	(0.0728)
x6	0.0183 ***	0.0199 ***
	(0.0000)	(0.0000)
x7	36.0389 ***	5.4249
	(0.0006)	(0.3845)
lnx8	0.0171	−0.0187
	(0.4840)	(0.6757)
Constant	2.7040 ***	3.7923 ***
	(0.0020)	(0.0003)
Observations	230	229
$R^2$	0.717	0.781
adj. $R^2$	0.666	0.743
Urban effects	Control	Control

Note: *, **, and *** indicate significant at the 10%, 5%, and 1% levels, respectively. The values in parentheses are p values.

below. According to the results in column (1) of Table 5, the interaction term between technological innovation and urban spatial structure is not significant in areas with weak environmental regulation, while the results in column (2) demonstrate that urban spatial structure has a significant moderating effect in areas with strong environmental regulation. The regression coefficients of the interaction term and the regression coefficients of urban technological innovation (lnx1) are of the same sign, which fully illustrates that urban spatial structure significantly enhances the impact of technological innovation on haze pollution and exerts a significant positive moderating effect on the haze reduction effect of technological innovation under the strong environmental regulation. This finding verifies hypothesis 3, which states that urban polycentric spatial structure enhances the haze reduction effect of technological innovation in areas with strong environmental regulations. The likely reason for this is that the “compliance cost” generated by the increased environmental regulation may have squeezed out the lower-skilled firms in the region, reduced the number of firms in each industry in the region of industrial concentration, increased the concentration of industries in the region, and further promoted the diffusion and spillover of knowledge, which contributed to the improvement of the technological level and thus reduced the generation of haze pollution. For technology-intensive industries, in particular, cities with a polycentric development model are more conducive to technology-intensive firms acting as a synergistic mechanism to enhance technology spillovers through technological cooperation and complementarity [60].

4.3. Robustness Tests

4.3.1. Robustness Tests Based on Green Technological Innovation

In addition to the number of patents granted, technological innovation can also be measured by the number of green invention patent applications (lngreen) in the current year, in order to eliminate the bias in the findings caused by the selection of specific indicators. A new core explanatory variable (lngreen) was generated by taking the logarithm of the number of green invention patent applications, and a new interaction term (interact1) was formed in conjunction with the spatial structure of the city. According to the results in columns (1), (2), and (3) in

Table 6. Robustness tests based on green technological innovation.

Explanatory Variables	Explained Variable: lnPM2.5
	(1)	(2)	(3)	(4)
lngreen	−0.0797 ***	−0.0311 ***	−0.0316 ***	−0.0283 ***
	(0.0000)	(0.0004)	(0.0003)	(0.0013)
x2			−0.3223 *	−0.4048 **
			(0.0777)	(0.0278)
interact1				−0.0399 ***
				(0.0065)
lnx3		0.4236 ***	0.4239 ***	0.4462 ***
		(0.0001)	(0.0001)	(0.0000)
lnx4		−0.1432 ***	−0.1424 ***	−0.1489 ***
		(0.0000)	(0.0000)	(0.0000)
x5		−0.0230 ***	−0.0214 ***	−0.0168 ***
		(0.0008)	(0.0019)	(0.0173)
x6		0.0200 ***	0.0198 ***	0.0204 ***
		(0.0000)	(0.0000)	(0.0000)
x7		10.9630 ***	11.3955 ***	11.0792 ***
		(0.0072)	(0.0052)	(0.0062)
lnx8		−0.0044	−0.0038	−0.0016
		(0.8209)	(0.8433)	(0.9353)
Constant	4.1153 ***	1.9503 ***	2.0456 ***	1.9373 ***
	(0.0000)	(0.0027)	(0.0017)	(0.0027)
Observations	452	452	452	452
adj. $R^2$	0.336	0.704	0.705	0.710
Urban effects	Control	Control	Control	Control
Time effect	No control	No control	No control	No control

Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. The values in parentheses are p values.

below, the coefficients of green technological innovation in the middle reaches of the Yangtze River urban agglomeration are all significantly negative at the 1% level. This finding is consistent with the previous paper and still verifies hypothesis 1, which suggests that technological innovation can effectively reduce the generation of haze pollution. The results in column (4) of Table 6 are also consistent with the previous paper, as the coefficients of the interaction term between green technological innovation and urban spatial structure and the coefficient of green technological innovation are both significantly negative at the 1% level, both with the same sign, further verifying that the tendency of urban spatial structure to be polycentric in the middle reaches of the Yangtze River urban agglomeration can effectively promote the haze reduction effect of technological innovation.

Similarly, green technological innovation was used as the core explanatory variable to verify the environmental regulation heterogeneity, and the results are shown in

Table 7. Robustness tests under strong versus weak environmental regulation.

Explanatory Variables	Explained Variable: lnPM2.5
	Weak Environmental Regulation	Strong Environmental Regulation
lngreen	−0.0087	−0.0465 ***
	(0.4663)	(0.0009)
x2	−0.8052 **	−0.5794 **
	(0.0183)	(0.1200)
interact1	−0.0082	−0.0656 ***
	(0.7109)	(0.0039)
lnx3	0.3820 **	0.1941
	(0.0112)	(0.2305)
lnx4	−0.1695 ***	−0.1559 ***
	(0.0000)	(0.0029)
x5	−0.0192 *	−0.0183
	(0.0519)	(0.0985)
x6	0.0199 ***	0.0212 ***
	(0.0000)	(0.0000)
x7	36.0300 ***	7.5922
	(0.0010)	(0.2273)
lnx8	−0.0002	0.0011
	(0.9928)	(0.9799)
Constant	2.5931 ***	3.5389 ***
	(0.0043)	(0.0008)
Observations	205	227
adj. $R^2$	0.663	0.748
City Effect	Control	Control
Time effect	No control	No control

Note: *, **, and *** indicate significance at the 10%, 5%, and 1%, levels respectively. The values in parentheses are p values.

below. According to the results of Table 7, the interaction term between green technological innovation and urban space is not significant in areas with weak environmental regulation. By contrast, in areas with strong environmental regulation, urban spatial structure significantly enhances the haze reduction effect of technological innovation. This finding is also consistent with the previous one.

4.3.2. Robustness Tests for Replacing PM_2.5

In addition to the annual average value of PM_2.5, this paper also adopts the indicators of sulfur dioxide and industrial smoke dust as air-quality measures to further investigate the effects of technological innovation, urban spatial structure, and technological innovation interaction term on haze pollution, in order to eliminate the bias of the specific indicator selection. According to the results in

Table 8. Tests for replacement of PM_2.5.

Explanatory Variables	Explained Variables: In SO2		Explained Variable: Ln Industrial Smoke Dust Emissions
	(1)	(2)	(3)	(4)
lnx1	−0.0965	−0.0874	−0.2253 ***	−0.1830 **
	(0.2422)	(0.2917)	(0.0051)	(0.0192)
x2	−2.5162 ***	−2.6903 ***	−1.4620	−2.2718 **
	(0.0094)	(0.0062)	(0.1193)	(0.0139)
interact		−0.0978		−0.4549 **
		(0.2793)		(0.0000)
lnx3	1.3722 **	1.4480 **	1.6664 ***	2.0189 ***
	(0.0211)	(0.0157)	(0.0040)	(0.0004)
lnx4	−1.2247 ***	−1.2339 ***	−0.0150	−0.0577
	(0.0000)	(0.0000)	(0.9268)	(0.7151)
x5	−0.0867 **	−0.0764 **	−0.0768 **	−0.0289
	(0.0210)	(0.0484)	(0.0351)	(0.4270)
x6	0.0672 ***	0.0685 ***	0.0116 *	0.0176 ***
	(0.0000)	(0.0000)	(0.0704)	(0.0054)
x7	64.4242 ***	65.0567 ***	−26.5055	−23.5636
	(0.0038)	(0.0035)	(0.2181)	(0.2586)
lnx8	0.2526 **	0.2495 **	0.1229	0.1082
	(0.0244)	(0.0263)	(0.2584)	(0.3045)
Constant	11.4655 ***	11.0231 ***	1.0303	−1.0272
	(0.0012)	(0.0019)	(0.7627)	(0.7576)
Observations	459	459	459	459
adj. $R^2$	0.617	0.617	0.100	0.155
Urban effects	Control	Control	Control	Control
Time effect	No control	No control	No control	No control

Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. The values in parentheses are p values.

, technological innovation has a significant inhibitory effect on industrial smoke dust emissions, and the interaction term between urban spatial structure and technological innovation significantly contributes to the haze reduction effect of technological innovation, which is consistent with the previous conclusion.

4.3.3. Robustness Tests for Additional Control Variables

Considering that city size has an impact on the relationship between urban space and haze pollution as well as on haze pollution itself, population size was added as a control variable for the regression analysis, and the regression results are shown in

Table 9. Tests for inclusion of population size.

Explanatory Variables	Explained Variable: lnPM2.5
	(1)	(2)
lnx1	−0.0539 ***	−0.0617 ***
	(0.0003)	(0.0000)
x2	−0.4100 **	−0.4443 **
	(0.0191)	(0.0107)
interact	−0.0495 ***	−0.0456 ***
	(0.0022)	(0.0047)
lnx3	0.3980 ***	0.2652 **
	(0.0002)	(0.0225)
lnx4	−0.1227 ***	−0.1213 ***
	(0.0001)	(0.0001)
x5	−0.0170 **	−0.0192 ***
	(0.0138)	(0.0054)
x6	0.0195 ***	0.0195 ***
	(0.0000)	(0.0000)
x7	11.4486 ***	11.3394 ***
	(0.0039)	(0.0040)
lnx8	0.0162	0.0139
	(0.4168)	(0.4831)
City size		0.0010 ***
		(0.0055)
Constant	2.1664 ***	2.5799 ***
	(0.0006)	(0.0001)
Observations	459	459
adj. $R^2$	0.710	0.714
Urban effects	Control	Control
Time effect	No control	No control

Note: **, and *** indicate significant at the 5%, and 1% levels, respectively. The values in parentheses are p values.

. The regression results after adding population size as a control variable are generally consistent with the previous finding.

4.3.4. Robustness Tests Considering Lagged Effects

As the innovation index reflects short-term fluctuations, and direct regression of the interaction term between technological innovation, urban spatial structure, and technological innovation may have estimation bias, robustness tests of lagged effects should be considered. The results of the tests are shown in

Table 10. Robustness tests considering lagged effects.

Explanatory Variables	Explained Variable: L.lnPM2.5
	(1)	(2)	(3)	(4)
L.Lnx1	−0.0726 ***	−0.0306 *	−0.0357 ***	−0.0337 **
	(0.0000)	(0.0530)	(0.0252)	(0.0323)
x2			−0.3696 **	−0.4460 **
			(0.0340)	(0.0101)
L.interact				−0.0562 ***
				(0.0005)
lnx3		0.3182 ***	0.3082 ***	0.3519 ***
		(0.0021)	(0.0028)	(0.0006)
lnx4		−0.0932 ***	−0.0884 ***	−0.0927 ***
		(0.0046)	(0.0071)	(0.0042)
x5		−0.0166 ***	−0.0154 **	−0.0098
		(0.0163)	(0.0252)	(0.1578)
x6		0.0176 ***	0.0171 ***	0.0179 ***
		(0.0000)	(0.0000)	(0.0000)
x7		5.7923	6.0379	6.1388 *
		(0.1265)	(0.1098)	(0.0062)
lnx8		0.0041	0.0086	0.0121
		(0.8726)	(0.7393)	(0.6356)
Constant	4.3352 ***	2.2590 ***	2.4133 ***	2.1319 ***
	(0.0000)	(0.0002)	(0.0001)	(0.0005)
Observations	432	432	432	432
adj. $R^2$	0.277	0.633	0.636	0.646
Urban effects	Control	Control	Control	Control
Time effect	No control	No control	No control	No control

Note: *, **, and *** indicate significant at the 10%, 5%, and 1% levels, respectively. The values in parentheses are p values.

. According to the results in columns (1), (2), and (3) of Table 10, technological innovation in the middle reaches of the Yangtze River urban agglomeration can effectively reduce the generation of haze pollution, which is consistent with the previous section and confirms that technological innovation significantly contributes to the reduction of haze pollution, in accordance with hypothesis 1. According to the results in column (4), the coefficients of the interaction term between technological innovation and urban spatial structure and technological innovation are significant and of the same sign, which further verifies that the tendency of urban spatial structure to be polycentric in the middle reaches of the Yangtze River urban agglomeration can effectively promote the haze reduction effect of technological innovation and make a significant and positive contribution to the haze reduction effect of technological innovation.

5. Discussion

Due to the provinces in the middle reaches of the Yangtze River having different levels of economic development and social conditions, regional heterogeneity in the effect of technological innovation on haze pollution should be analyzed. The results of provincial heterogeneity are shown in

Table 11. Results of the test for heterogeneity by province.

Explanatory Variables	Explained Variable: lnPM2.5
	(1) Jiangxi Province	(2) Hubei Province	(3) Hunan Province
lnx1	−0.0133	−0.0364	−0.0960 ***
	(0.3672)	(0.2758)	(0.0000)
x2	−0.0660 **	−0.1983 ***	−0.1631 **
	(0.0271)	(0.0024)	(0.0126)
interact	0.0134	0.0407 *	0.1142 ***
	(0.4054)	(0.0936)	(0.0000)
lnx3	0.0648 ***	0.0612 **	0.2005 ***
	(0.0044)	(0.0001)	(0.0000)
lnx4	0.0243	0.0517	0.2967 ***
	(0.3753)	(0.2182)	(0.0000)
x5	0.0076	0.0027	−0.0285 ***
	(0.1535)	(0.8029)	(0.0072)
x6	0.0024	−0.0098 ***	0.0066 ***
	(0.2795)	(0.0001)	(0.0000)
x7	5.8104	−12.6880 *	−6.3481
	(0.2829)	(0.0837)	(0.1285)
lnx8	0.0420 **	0.0293	−0.0396
	(0.0110)	(0.4637)	(0.1062)
Constant	2.7581 ***	3.7261 ***	0.9267 ***
	(0.0000)	(0.0000)	(0.0017)
Observations	153	170	136
adj. $R^2$	0.361	0.232	0.535
Urban effects	Control	Control	Control
Time effect	No control	No control	No control

Note: *, **, and *** indicate significant at the 10%, 5%, and 1% levels, respectively. The values in parentheses are p values.

. According to the results in Table 11, from a provincial perspective, technological innovation in Hunan Province can significantly suppress the generation of haze pollution, while technological innovation in Jiangxi Province and Hubei Province does not significantly reduce the generation of regional haze pollution. Within Hubei Province and Hunan Province, the signs of the spatial structure of polycentric and technological innovation coefficients are different, suggesting that in Hubei Province and Hunan Province, urban polycentric development suppresses the haze reduction effect of technological innovation in the region. The reason for this is that when the degree of polycentric concentration is high, the overly dispersed and homogeneous spatial pattern means that the development stage and economic volume of several cities in the provinces are similar, which in turn tends to lead to more competition than cooperation between cities. Especially since a competitive mindset still prevails in China, competition among cities may inhibit innovation exchange and innovation improvement.

6. Conclusions and Policy Implications

Environmental protection and economic development are the eternal concerns of human beings. Environmental protection is likely to affect the pace of economic development in the short term. However, as the level of technological innovation increases, environmental protection and economic development become more effectively coordinated, and the goals of environmental protection and economic development can be achieved at the same time [61]. Theory and research have confirmed that technological innovation is a key driver for achieving long-term green economic development in the face of environmental protection requirements [62]. Therefore, how to take advantage of technological innovation and urban spatial structure to promote the level of technological innovation and achieve effective haze control is one of the urgent issues to be solved in the middle reaches of the Yangtze River urban agglomeration.

Based on this consideration, this paper examines the impact of technological innovation on haze pollution in the context of urban spatial structure, using technological innovation as the main research perspective. Specifically, this paper examines the direct link between technological innovation and haze pollution and the moderating effect of urban spatial structure on technological innovation and haze pollution, as well as testing the moderating effect of urban spatial structure in two different types of cities with high and low environmental regulations, specifically through the use of basic data and urban spatial structure data from 2004 to 2020 as research samples. The main findings are as follows. Firstly, technological innovation in the middle reaches of the Yangtze River urban agglomeration has a significant inhibitory effect on haze pollution. Secondly, in the middle reaches of the Yangtze River urban agglomeration, urban spatial structure significantly enhances the impact of technological innovation on haze pollution and has a moderating effect on the haze reduction effect of technological innovation. Thirdly, in cities with a higher level of environmental regulation, the tendency of urban spatial structure to be polycentric makes a substantial contribution to the mitigating effect of technological innovation on haze pollution.

The policy implications of the findings are that in the current urbanization process, more attention should be paid to the impact of urban spatial structure on urban innovation activities as well as environmental pollution for the middle reaches of the Yangtze River urban agglomeration, and reasonably plan the region and the different urban development patterns within the region to further enhance urban innovation and improve the urban agglomeration environment. Based on these conclusions, the research proposes the following policy recommendations.

Firstly, science and technology policies should prioritize the advancement of technological innovation, with a particular emphasis on supporting green innovation activities. This approach can facilitate economic development and effectively reduce haze pollution at its source within the middle reaches of the Yangtze River urban agglomeration. Additionally, recognizing that the spillover effect of technological innovation from neighboring regions, the Yangtze River urban agglomeration should continue its coordinated regional development strategy and leverage the spillover effect in the region.

Secondly, regional policies should be oriented towards implementing a polycentric development strategy within the midstream Yangtze River urban agglomeration. This involves decentralizing the non-core functions from the main urban centers and central cities to mitigate haze pollution resulting from the over-concentration of population and economic activities in specific areas. To ensure successful implementation, efforts should be directed at promoting positive interactions between the main and secondary centers. This can be achieved through enhancing urban transport infrastructure, facilitating the movement of innovation factors within the agglomeration, encouraging city-to-city exchanges, and promoting the flow and overflow of technological innovation between cities. By fostering the coordinated development of each subcenter and strategically coordinating urban spatial structure, the region can effectively boost technological advancement and alleviate haze pollution.

Thirdly, in terms of environmental governance, the industrial layout of cities should be rationalized through environmental regulation to avoid the negative externalities brought about by agglomeration economies. On the other hand, under the multi-center structure of urban agglomeration, environmental constraints can encourage cities to compete healthily in order to obtain high-quality resources, and eliminate more traditional polluting enterprises and introduce high-tech enterprises to improve their competitiveness.

This study centers around addressing the issue of haze pollution within the midstream Yangtze River urban agglomeration. The primary focus is to explore the impact of technological innovation on haze pollution and its interaction with urban spatial structure. Moreover, the research delves into analyzing how the effectiveness of urban spatial structure varies under the different environmental regulations. The findings of this study serve as valuable reference for enhancing both urban spatial structure and haze pollution control within the midstream Yangtze River urban agglomeration. Furthermore, the study’s recommendations can be applied to other urban agglomerations which are similar to the Yangtze River urban agglomeration, offering valuable insights for their environmental improvement strategies.

However, this study has some limitations. We know that morphological polycentricity is likely to be distinguished from functional polycentricity [63], and knowledge polycentricity is one of the concrete manifestations of functional polycentricity [64]. However, due to the availability of data, this paper only measures urban spatial structure from the perspective of urban morphological polycentricity, and does not analyze and explore the impact of functional polycentricity. The analysis of the mechanisms of the impact of technological innovation and spatial structure on environmental pollution at different scales and heterogeneity can be strengthened in the future. Moreover, urban spatial form should consider all of a city’s compactness, density, and external shape characteristics in development [65]; therefore, further analysis should consider all of these factors as well.