The Impact of Smart City Pilots on Haze Pollution in China—An Empirical Test Based on Panel Data of 283 Prefecture-Level Cities

Abstract

The rapid pace of urbanization in China has led to a significant increase in haze pollution in its cities. However, there has been limited research on the dynamic impact and mechanisms of smart city pilots, which offer an innovative approach to urbanization, on haze pollution. This study selects panel data from 283 prefecture-level cities in China from 2007 to 2017 and uses a quasi-experimental approach based on the three batches of pilot construction of smart cities since 2012 to examine the impact of smart city pilots on haze pollution. The multi-phase difference-in-differences (DID) model is used for the analysis. The findings reveal: (1) Smart city pilots have a significant positive effect on reducing urban haze pollution. (2) Smart city pilots contribute to changes in the urban development model, where technological innovation, industrial structure adjustment, and resource allocation optimization under innovation-driven development significantly mitigate haze pollution. (3) Heterogeneity analysis shows regional differences in the effectiveness of smart city pilot policies in reducing haze pollution in China, with a decreasing trend from the eastern to the western regions. The haze-reducing effect of smart city pilots in the central region has yet to be observed. This research provides valuable theoretical and policy insights for improving urban ecological environments and promoting green transformations of production and lifestyle.

1. Introduction

In the wake of China’s reform and opening-up policies, the urbanization rate has been steadily increasing, accompanied by rapid economic growth. Drawing upon data disseminated by the National Bureau of Statistics of China, the urbanization rate concerning the permanent population has witnessed a remarkable increase, soaring from a modest 17.92% in 1978 to a significant 64.72% in 2021. It is projected that China will persist in this accelerated urbanization phase. However, the extensive economic development model that sacrifices the environment is no longer sustainable. While China’s urbanization process is accelerating, “urban diseases” such as haze pollution have become more frequent, with a wider impact range and longer duration, severely threatening people’s health and constraining the economy’s high-quality development [1]. Therefore, improving the urban development model and addressing environmental issues such as haze pollution have become urgent issues that need to be solved to facilitate the advancement of green development in China.

Haze is described as significant numbers of fine, dry particles floating in the atmosphere that result in reduced visibility and turbid air. A myriad of factors contribute to the genesis of haze, including air humidity, respirable particulate matter, fine particulate matter, sulfur dioxide, and nitrogen oxides. Notably, PM_2.5, denoting atmospheric aerosols with an aerodynamic diameter not exceeding 2.5 microns in the ambient air, is a prevalent metric in haze pollution quantification [2]. Given its capacity for prolonged suspension in the atmosphere, PM_2.5 serves as a critical determinant of air quality and visibility. The interactions between PM_2.5 and haze are mutual: PM_2.5 fosters hazy conditions, while the latter reciprocally intensifies PM_2.5 accumulation.

The expansion of cities and populations due to rapid urbanization is one of the reasons for the frequent occurrence of hazy weather [3]. However, the fundamental cause of this issue can be attributed to the distorted industrial structure characterized by an excessive focus on the secondary sector, a coal-based energy structure, rapid population aggregation, and increased intensity of road transportation. The combined effect of these factors caused haze pollution to quickly grow and spread across China [4]. In 2017, the Communist Party of China’s (CPC) 19th National Congress’s report clearly identified pollution prevention and control and elevated “green development” to the highest level of national development strategy for the first time. After years of governance led by the government and coordinated by the market, the haze pollution has been mitigated. As per the October 2022 report by China’s Ministry of Ecology and Environment, the annual average concentration of PM_2.5 in cities at the prefecture level and above has witnessed a decline, from 46 µg/m³ in 2015 to 30 µg/m³ in 2021. The 20th National Congress of the CPC, convened in 2022, reiterated the imperative for enhanced coordinated pollution control with an aim to substantially mitigate instances of severe pollution. Set against this backdrop, the urgency of investigating and addressing the issue of haze pollution becomes particularly pronounced.

The idea of smart cities was initially introduced in the Smart Planet vision framework by IBM in 2008. Smart city construction is expected to enhance national urban management capabilities and service levels and promote industrial transformation and development. Studies suggest that a smart city represents a new model for urban development that expedites the construction of innovative infrastructure [5], with technology as its fundamental driving force [6], and integrates technology into various areas of urban management and services [7]. It improves the level of intelligent urban living and city governance by using intelligent technology to enhance the flow of talents, information, and resources [8,9]. Several scholars perceive smart cities as a technological remedy to pressing urban challenges [10] and an important approach to achieving sustainable urban development [11]. In light of these developments, the Ministry of Housing and Urban-Rural Development issued the “Notice on Launching the National Smart City Pilot Work” in December 2012. This directive explicitly delineates a smart city as an innovative paradigm for urban planning, construction, and management that effectively leverages modern science and technology, assimilates information resources, orchestrates business application systems, and reinforces urban planning, construction, and management. The release of this document marked the initiation of the pilot phase for smart cities. So far, China has published a total of 290 lists of national smart city pilots in three batches. It is evident that smart cities, characterized by a novel model of technology-driven and innovative planning, construction, and management, point to a new direction for the future development of cities.

As a way to address the problem of urban economic growth, smart city construction has attracted increasing attention in developing countries [12]. Early research by some scholars on smart cities predominantly centered on qualitative analysis, exploring the theoretical content of smart cities, and actively exploring solutions to challenges encountered during smart city development [13]. Currently, some scholars have pointed out through empirical research that smart city development has the potential to considerably enhance urban economic efficiency [14], foster an augmentation in green total factor productivity [15], reduce environmental pollution [16,17], and enhance the quality of urban economic development [18]. As an important branch of environmental problems, there are few studies that empirically analyze the complex effects of smart cities on haze pollution. Therefore, as a novel urban development model, can smart cities facilitate the promotion of green development in cities? How does it affect the intensity of haze pollution? Will this effect differ depending on the region where the city is located? What are the underlying mechanisms? Clarifying these issues has significant value for exploring new models of urbanization in China, effectively promoting environmental pollution prevention and control, and promoting China’s comprehensive transformation to a green economy.

Therefore, based on the smart city pilot policy, this paper presents an empirical investigation into the effects and mechanisms of smart city development on haze pollution and examines whether the impact differs across regions. The study employs panel data covering 283 Chinese prefecture-level cities from 2007 to 2017 and takes the three batches of smart city pilot cities selected by the Ministry of Housing and Urban-Rural Development as quasi-experiments. The paper applies the multi-phase difference-in-difference (DID) model to estimate the causal effect of smart city development on haze pollution. The heterogeneity analysis is also conducted to explore the regional differences in the impact of smart city development on haze pollution.

The remainder of this paper is structured as follows: Section 2 provides an overview of the relevant literature and the policy backgrounds of smart city pilots are introduced. Section 3 proposes the impact mechanism along with several causal hypotheses. Section 4 outlines the variables, empirical model, and data used in this study. In Section 5, the primary empirical findings and a series of robustness tests are discussed, followed by an exploration of the heterogeneous effects of the smart city pilots. Finally, Section 6 summarizes the study’s conclusions and offers recommendations for policy.

2. Literature Review and Policy Background

2.1. Literature Review

As a response to the challenges of growing cities and increasing resource scarcity, the development of smart cities represents a promising strategy that aligns with new urban development trends, accelerates the transformation of urban development models, and effectively addresses various “urban diseases”, such as ecological degradation, environmental pollution, and traffic congestion, that arise from China’s new urbanization process. The objective of this paper is to explore the association between urbanization and the severity of haze pollution. Previous studies can be broadly categorized into three types: first, urbanization promotes environmental and haze pollution [19,20]. The expansion of urban scale triggers a siphoning effect on surrounding production factors, population, and labor force, leading to increased population density, encroachment of urban production land on green areas, and increased pollution emissions, thereby exacerbating environmental pollution and “urban diseases” [21]. The expansion of urban economies and production scales also contribute to increased industrial and transportation emissions [22]. Shao et al. (2019) point out that urbanization’s agglomeration effects promote economic development while exacerbating environmental pollution represented by haze [23].

Second, urbanization alleviates environmental and haze pollution [24]. Scholars such as Li (2017) and Hao (2018) argue that while urbanization promotes industrial agglomeration effects, evidence suggests a positive correlation between urban innovation capabilities and agglomeration levels, and technological innovation helps cities engage in environmental governance and reduce pollution [25,26]. Brajer et al. (2011) and Xie et al. (2019) argue that urbanization’s economic development improves per capita Gross Domestic Product (hereinafter referred to as GDP), and following economic growth, citizens will pay more attention to environmental quality, generating social pressure on the government to increase environmental prevention and control efforts, ultimately reducing environmental pollution [27,28]. Fan et al. (2023) have also found that increased urban density during urbanization can help reduce carbon intensity and improve the environment [29]. Third, the relationship between urbanization and environmental pollution is complex and non-linear. Xu et al. (2008) found through a variable intercept model analysis that the relationship between China’s urban scale and resource consumption follows an “N-shaped” curve, which implies that the Kuznets curve [30], which illustrates the “inverted U-shaped” relationship between urban development and environmental quality, is not a constant phenomenon.

Furthermore, academic investigations have delved into the underlying processes that lead to the formation of haze pollution. In reality, there are many factors that contribute to the formation of haze, such as respirable particulate matter, fine particulate matter, sulfur dioxide, nitrogen oxides, etc. It is important to reasonably measure the extent of haze pollution; some scholars use PM10 concentration to reflect haze, which is not as effective as PM_2.5 concentration in portraying haze pollution [31,32]. Qin et al. (2016) used global nighttime lighting data, global population distribution data, and PM_2.5 surface concentration data to comprehensively measure PM_2.5 concentrations and urban sprawl levels in Chinese cities and combined them with economic statistics to examine the influence of city size and its spatial structure on haze pollution [33]. Yu et al. (2015), in their examination of PM_2.5 data collated from 354 cities spanning the period 2001–2010, embarked on an exploration of local government strategies for haze pollution control [34]. Wu et al. (2016) investigated the influencing factors of haze pollution based on data from Chinese cities with PM_2.5 monitoring stations nationwide in 2014 [35]. Gradually, PM_2.5 concentration has established itself as the primary metric utilized for measuring haze pollution [2]. According to research, the contributing elements to the persistent rise in PM_2.5 concentrations encompass both natural phenomena and societal influences. In terms of nature factors, such as climate, wind speed, and humidity in the air [36,37], these are related to the concentration of PM_2.5. From the perspective of human activities, some researchers believe that the increase in PM_2.5 is related to economic growth [38], the proportion of secondary industries [39], coal consumption’s share in energy consumption structure [40], urbanization [41], openness [42], etc.

2.2. Policy Background

Since 2010, the Chinese government has proactively fostered the development of smart cities through the introduction of various policies, concentrating on top-level design, operation, management, service, and development. These policies have played a crucial role in guiding and encouraging the transition to smart urbanization and the adoption of next-generation information technologies in urban planning, construction, and management.

The National Smart City pilot program, initiated by the Ministry of Housing and Urban-Rural Development in 2012, introduced a holistic indicator system embracing infrastructure, intelligent construction and livability, smart management and services, along with intelligent industry and economy. This comprehensive schema laid the groundwork for the ensuing evolution and execution of smart city initiatives throughout China. Concurrently, pilot cities emerged as experimental grounds for pioneering solutions and optimal practices.

The pilot project consisted of three phases, involving a total of 277 regions across the country. The inaugural set of the pilot project, introduced in December 2012, covered 90 regions, encompassing 37 prefecture-level cities, 50 districts (counties), and 3 towns. Subsequently, in May 2013, the second phase incorporated 103 regions, encompassing 83 cities and districts, along with 20 counties and towns. By April 2015, the third batch had expanded to incorporate an additional 84 regions. The spatial-temporal fractal map of PM_2.5 concentration in 148 smart city pilots is shown in Figure 1. It is clear that most smart city pilots experienced a significant decline in terms of PM_2.5 concentration.

In 2015, multiple Chinese authorities underscored the significance of talent cultivation in smart city development. They jointly promulgated the “Guiding Opinions on the Construction and Application of Smart City Standard System and Evaluation Index System”, accentuating the need to nurture a competent workforce that would catalyze the digital metamorphosis of urban governance. In a novel move, the central government incorporated smart city evolution into a governmental work report for the inaugural time in 2015. Furthermore, the notion of a “smart society” garnered amplified endorsement at the 19th National Congress of the Communist Party of China in 2017.

Subsequently, in 2018, the “Smart City Top-Level Design Guide” elaborated on the guarantee measures, categorizing them into four domains: organizational, policy, talent, and funding. This comprehensive approach ensured that smart city development would be supported not only by robust infrastructure but also by a conducive policy environment, a skilled workforce, and adequate financial resources. The 14th Five-Year Plan, released in 2021, emphasized the importance of smart cities, prompting governments at all levels to commit to the advancement of digital society, the digital economy, and digital government initiatives. The overarching objective of these efforts is to promote green urbanization and enhance citizens’ quality of life through the proliferation of smart city infrastructure.

3. Theoretical Analysis and Hypotheses

The existing literature has predominantly centered on the link between conventional urban development models and the severity of haze pollution. However, the effect of smart city development on haze pollution remains inadequate. Due to the fact that smart city pilots were carried out in three batches in 2012, 2013, and 2014 in a total of 290 cities in China, previous studies, such as that conducted by Shi et al. (2018), have used the first batch of pilot cities in 2012 as a baseline for analysis in their empirical models and lacked consideration for the longitudinal dimension of the three batches of smart city pilots [16]. Additionally, various innovations generated under the framework of smart city pilots and innovation-driven development require time to be completed. The overall construction process also involves a long time span, and the implementation and utilization of innovative equipment and technologies also require considerable time. Furthermore, it takes time for investment to translate into actual results. Therefore, the impact of the smart city pilot on haze reduction is not immediately evident. Similarly, the effects of industrial structure adjustment and resource allocation optimization also require a long period to be fully implemented and matured. Moreover, as time progresses, the overall construction model and results of smart cities continue to evolve and improve, and the effect of haze reduction at different time points may not be the same. Based on the above-mentioned considerations of time lag, this paper adopts the multi-phase DID model to analyze the three batches of smart city pilots using different time baselines, preventing time-related biases in the research findings. Based on this, the study posits the following hypothesis:

Hypothesis 1.

Smart city pilots can effectively reduce haze pollution.

Wang et al. (2021) argued that technological innovation can foster the coordinated development of the economy and ecology [43]. Based on Schumpeter’s innovation theory and the content of smart city pilots, this paper classifies the innovation drivers generated during smart city construction into five distinct categories: technological innovation, market innovation, product innovation, resource allocation innovation, and organizational innovation [16]. Specifically, the definitions of these innovation drivers are as follows:

(1)

Technological innovation. Smart city pilots involve embedding various sensors into cities and applying them to various aspects such as railways, highways, water supply, pipelines, and buildings. Based on the Internet and the Internet of Things (IoT), information integration is achieved through cloud technology to realize intelligent urban management. Angelidou (2017) points out that information and communication technologies (hereinafter referred to as ICT) play an important and active role in urban innovation [44]. In the construction of smart cities and the management of haze, intelligent emission monitoring equipment consisting of various types of sensors and inductors is applied to the economic production activities of enterprises and the lives of citizens, providing an all-round technological upgrade to the traditional pollution emission monitoring, pollution management model, and technical means, allowing for real-time dynamic information collection, intelligent monitoring of emission alert lines, and early warning and control.

(2)

Market innovation. With the accelerated implementation of smart city pilots, the development of ubiquitous emerging industries such as productive services, new energy, intelligent vehicles, and new materials will be further promoted. It can be foreseen that as the construction of smart cities continues to deepen nationwide, emerging markets will surface more frequently and their maturity will increase day by day. Against this background, it is inevitable to develop industrial markets driven by new technologies and high-tech. The rise of emerging industries urgently requires new technologies and a large number of new professionals, which will drive the development of industries such as internet services, information communication, technology R&D, and derivative industries such as software, design, and commerce. Changing the market structure of traditional cities from the bottom-up logic will help the transformation and upgrading of traditional high-pollution and high-emission industries.

(3)

Product innovation. The progress of smart city development has the potential to facilitate the integration of intelligent technologies, such as next-generation information technology, cloud computing, and big data, into conventional urban industries and business products. This not only enhances enterprise information technology but also aligns with the growing focus on eco-friendliness and environmental protection. However, due to the short development period of smart city pilots in China and the lengthy technology transformation cycle, the present focus of product innovation is primarily on improving existing products. In other words, traditional products are being coupled with intelligent trends to achieve digital and intelligent upgrades, enabling function expansion and technological advancement. Although the level of new product innovation and research transformation is not yet sufficient, improving product innovation is still based on meeting or opening up new market demands, brainstorming, and researching under the smart city application framework to find a balance between technology and demand.

(4)

Resource allocation innovation. Building on the Internet of Things, smart cities can achieve comprehensive and real-time awareness of urban conditions through data mining and sharing. This awareness is then utilized to provide intelligent analysis and processing using cloud computing technology, thereby providing a reference for various stakeholders and policymakers. Specifically, with regard to the focus of this study, smart cities continuously and dynamically monitor market demands and group preferences to adjust the urban economic model, gradually shifting production factors such as human capital and financial capital away from traditional high-energy-consuming industries and toward emerging industries, thereby achieving energy conservation and emission reduction in cities.

(5)

Organizational innovation. The market innovation promoted by the construction of smart cities has contributed to the formation of new enterprises. However, innovative and knowledge-driven new enterprises differ from traditional enterprises in terms of management models, requiring a more flat, scientific, and efficient networked management approach. The transition from traditional management models to smart governance models has avoided the problems of wasted time and production resources resulting from the lagging efficiency of traditional management models.

Based on the above analysis, this paper uses Schumpeter’s innovation-driven theory to construct a theoretical framework, taking the five types of innovation as the starting point for smart city innovation drivers and establishing a connection between innovation drivers and the three major effects. On this basis, this study explores the role of smart city pilots in reducing haze pollution and its logical mechanism: smart city pilots—innovation driven—three major effects—reduction of haze pollution. In other words, smart city pilots achieve innovation drivers by producing technological innovation effects (TI), industrial structure adjustment effects (ISA), and resource allocation optimization effects (RAO), thereby achieving the goal of reducing haze pollution. To better understand the impact mechanism of this study, the detailed framework is shown in Figure 2. The confluence of the two distinct types of innovation delineated within the orange-dashed rectangle engenders effects attributable to technological innovation. Concurrently, the triad of innovative strategies enclosed by the green-dashed rectangle collectively precipitates effects consistent with the adjustment of the industrial structure. Lastly, the trio of innovations signified within the blue-dashed rectangle collectively prompts effects in alignment with the optimization of resource allocation.

The technological innovation effect is driven by both technological and product innovation. Research by Liu et al. (2012) shows that technological innovation is the main driving force behind China’s industrial pollution reduction [45]. The technological innovation brought about by smart city pilots fosters the aggregation of innovative factors, including talent, firms, and the capital flow of R&D, and enhances local innovation levels. In the stage of innovation-driven economic development, traditional enterprises increasingly find it difficult to achieve increasing returns to scale and face diminishing marginal returns through factor and investment paths. In this context, technological innovation plays a “catfish effect”, promoting enterprises to vigorously promote innovation and development. In addition, under the new requirements of promoting new urbanization, government departments and scientific research institutions are expected to intensify their research and adoption of clean energy and environmental protection technologies, enhance resource utilization efficiency, lower energy consumption per unit of GDP, and attain the objective of emissions control and reduction. Product innovation relies on the research and transformation of technological innovation to form optimized products with higher quality and technological content. During the use of products, loopholes and upgrade points are constantly discovered and fed back to the technological innovation end for iterative upgrading. Through market practice and feedback, technological innovation is constantly promoted and advanced. Specifically, the technological innovation effect improves energy efficiency, reduces energy consumption intensity, and continues to offset high-polluting emissions, achieving the effect of reducing haze pollution. Therefore, this paper proposes the following hypothesis:

Hypothesis 2.

Smart city pilots reduce haze pollution through technological innovation effects.

The industrial structure adjustment effect is jointly driven by technological innovation, product innovation, and market innovation. The smart city pilot can function as a starting point where technological innovation provides abundant ICT and information technology applications, accelerating the digital transformation of traditional enterprises and effectively enhancing their market competitiveness. At the same time, new markets bring new demands that need to be met through supply-side responses, leading to the emergence of more technology- and knowledge-intensive industries as new drivers of urban economic growth. These industries share the common features of being technology-intensive and low-pollution, thus driving the refinement and modernization of the regional industrial structure. The optimized innovative products introduced into the market imply the updating and upgrading of the production technology, equipment, management methods, and efficiency of the users, as well as changes in production patterns and the direction of resources towards more advanced forms of innovation-driven economic development. As the industry matures, leading enterprises will lead the entire industry towards transformation, transforming traditional industries into intensive technology- and knowledge-driven industries. In addition, the emerging markets exert pressure on traditional industries and firms with high energy consumption and pollution levels, promoting a benign competition mechanism and optimizing the industrial structure of the smart city. In summary, the effect of industrial structure adjustment based on smart city construction has changed the proportional relationship between the secondary and tertiary industries in the industrial structure through factor circulation and resource allocation, changing the traditional pattern of relying on high energy consumption and high emission industries to drive economic growth. High-polluting enterprises are upgraded iteratively or transferred to other industries, promoting the tertiary industry as the mainstay of the urban economy with high efficiency and low consumption, thereby generating a sustained haze reduction effect. Therefore, the following hypothesis is proposed in this paper:

Hypothesis 3.

Smart city pilots reduce haze pollution through industrial structure adjustment effects.

The optimization effect of resource allocation is driven by market innovation, resource allocation innovation, and organizational innovation. Smart cities use the Internet of Things network constructed by sensors and big data technology to achieve the coordinated operation of people, objects, information, and data in the city. The efficient operation of information flow, logistics, and transportation reduces the cost of exchange between various entities and improves the efficiency of exchange. At the same time, more varied and efficient channels of exchange are also advantageous for enhancing resource allocation efficiency and the flow of production factors. The goal of this study is to promote the flow of innovative resources and technological elements into the energy industry in order to improve energy efficiency, reduce energy consumption intensity, reduce haze pollution emissions, improve pollution treatment levels, and enhance enterprise environmental protection and governance capabilities. Smart cities, serving as a novel driver of economic growth, need to rely on new economies to drive urban development. Emerging industries, including big data, cloud computing, and the Internet of Things, have adapted to the development trend and become a new wave, attracting capital, labor, data information, and technology through the siphon effect and gathering element flow, thereby improving and optimizing the resource allocation structure of emerging and traditional industries. The emerging information technology applications in smart city construction are directly projected onto the fields of social governance and corporate management. Through ICT technology upgrades, they are promoted towards scientific and intelligent management. Governments and enterprises play an active role as dual subjects in resource allocation, increasing their sensitivity to market demand and response speed to enhance the efficiency of production resources and innovative factors. The resource allocation optimization effect is mainly manifested in the change in energy consumption structure and the flow of innovative factors into the energy industry. Smart city pilots improve the consumption structure of traditional and clean energy, and the influx of innovative factors effectively promotes the transformation and upgrading of traditional energy enterprises towards green development, thereby achieving a significant mitigating effect on haze pollution. In addition, in a fair and transparent mature market system, the development of emerging markets will likely attract the investment of capital and production factors to continuously withdraw from energy-intensive, high-emission, heavily polluting industries and pivot towards clean industries with advanced technology, high energy efficiency, and low emissions, achieving optimal factor allocation. Therefore, this paper proposes the following hypothesis:

Hypothesis 4.

Smart city pilots reduce haze pollution through resource allocation optimization effects.

There exist regional disparities in the process of urban growth and development, and the new urban development model based on smart city construction is equally affected by regional differences against the backdrop of the inherent physical and geographic reliance of traditional cities. In eastern and coastal areas of China, large amounts of production factors have been accumulated since the early stages of reform and opening-up due to their geographic advantages. These cities have developed rapidly through long-term capital investment, scientific technology, human resources, and advanced management models. As a result, many key urban clusters, such as Beijing-Tianjin-Hebei, the Yangtze River Delta, and the Pearl River Delta, have been formed, and their development levels are significantly higher than those in the central and western regions. However, there is a natural interaction between urbanization and environmental pollution. The pollution level in eastern and northern urban clusters is more severe than that in central and western regions, and the differences in their energy consumption structures further deepen this disparity in pollution levels. In China, an inverted U-shaped relationship has been observed between industrial agglomeration and haze pollution [46]. Environmental pollution, a product of the early stages of urbanization and industrialization, is often a vital barrier to urban agglomeration [47]. However, in the field of pollution treatment, there is also an effect of scale; as urbanization and industrialization progress, agglomeration becomes a way to effectively control environmental pollution [48]. The same is true at the level of haze pollution, where the effects of scale manifest themselves in the form of possible economies of scale in the treatment of gas emissions in areas of high agglomeration, leading to a reduction in PM_2.5 emissions per unit of input and output. In contrast, the congestion in the market caused by agglomeration and the competition and imitation between firms in the short term led to a sharp increase in regional emissions of fine particles, triggering a rise in regional air pollution. This has led to a sharp increase in regional emissions of fine particles, leading to an increase in PM_2.5 levels in the air. Therefore, the impact of geographical industrial agglomeration on haze pollution may have non-linear characteristics and trends. Currently, in the eastern and central regions of China, industrial agglomeration can effectively ease haze pollution, but the opposite is true in the western regions [49]. Compared with the relatively dispersed smart city pilots in central and western regions, smart city pilots in eastern regions are more likely to generate cluster effects, which can improve environmental governance efficiency and quality through spatial spillover effects and regional information exchange in the cluster. The above regional differences lead to higher homogeneity of smart cities in the same region or city cluster, and the construction effect of smart cities is also better, while the construction effect of smart cities often varies between different regions. Therefore, this study proposes the following hypothesis:

Hypothesis 5.

The degree of haze reduction effect of smart city pilots varies due to regional differences.

4. Methodology and Data

4.1. Model Setting

4.1.1. Baseline Regressing Model

In this paper, a total of 148 smart cities were identified after merging the districts and counties with their corresponding prefecture-level cities from the three batches of pilot city lists. This study employs a quasi-natural experiment design using the smart city pilot policy and adopts the multi-phase DID model to empirically investigate the effect of the policy on haze pollution. For sample selection, the study employs panel data encompassing 283 Chinese prefecture-level cities. The sample of cities with smart city construction is treated as the treatment group, while the sample of cities without smart city construction is treated as the control group. Considering the time differences in the implementation of smart city pilot policies across different cities, which were launched in three batches, this study follows the research design of Wang et al. (2022) [50] and uses the multi-phase DID method for evaluation. The baseline econometric model is specified as follows:

$${\mathrm{PM}}_{2.5it}={\beta }_0+{\beta }_1{\mathrm{DID}}_{it}+{\beta }_2{X}_{it}+{\gamma }_t+{\mu }_i+{\epsilon }_{it}$$

(1)

where PM_2.5it is the explained variable, representing the level of haze pollution in city i in t years. DID_it is the core explanatory variable, indicating whether city i is a smart city pilot in t years. β₁ is the primary coefficient of interest in this study, which quantifies the alleviating effect of smart city pilots on haze pollution, and a significantly negative coefficient value would suggest that smart city pilots have a beneficial impact on reducing haze pollution in urban areas. X_it is a group of control variables comprising economic scale, transportation scale, level of openness, information level, and population scale. γ_t represents the time-fixed effect, μ_i represents the individual-fixed effect for each city, and ε_it is the random error term.

4.1.2. Mediating Effect Model

As stated previously, the effect of smart city pilots on haze pollution is not direct but rather operates through the mechanisms of the smart city pilots—innovation driven—three major effects—reduction of haze pollution. To verify the plausibility of this logical mechanism, the principle of the mediating effect test in Wen et al. (2014) [51] was referenced to form a set of mediating effect test equations through Equations (2) and (3):

$${\mathrm{MD}}_{it}={\beta }_0+{\beta }_1{\mathrm{DID}}_{it}+{\beta }_2{X}_{it}+{\gamma }_t+{\mu }_i+{\epsilon }_{it}$$

(2)

In Equation (2), MD_it is used as the mediating variable, representing the effects of technological innovation, industrial structure adjustment, and resource allocation optimization, respectively. γ_t represents the time-fixed effect, μ_i represents the individual-fixed effect of each city, X_it is a series of control variables, and ε_it is the residual term. The three mediating variables are used for regression, respectively, to verify whether smart city construction drives the three main effects through innovation; this is step one.

If the coefficient is significant, it means that the smart city pilot does produce three major effects, and then the mediating variable MD_it is substituted into Formula (1) to form Formula (3) to test whether the haze reduction effect of the three major effects is significant; this is step two.

$${\mathrm{PM}}_{2.5it}={\beta }_0+{\beta }_1{\mathrm{DID}}_{it}+{\beta }_2{X}_{it}+{\beta }_3{\mathrm{MD}}_{it}+{\gamma }_t+{\mu }_i+{\epsilon }_{it}$$

(3)

4.2. Variable Definition

4.2.1. Dependent Variables

Haze Pollution (PM_2.5). As the main pollutant causing haze is PM_2.5, this paper uses the annual average PM_2.5 concentration of the city to express the degree of haze pollution. The PM_2.5 data in this paper is derived from the global annual average surface PM_2.5 concentration data provided by Columbia University’s International Earth Science Information Network, which is raster data from meteorological satellite monitoring. This article uses ArcGIS10.2 to analyze the remote sensing raster data to derive the annual average PM_2.5 concentration data for 283 cities.

4.2.2. Independent Variables

Smart City Pilots (DID), which is expressed by treat_i × time_t. treat_i is a dummy variable for the pilot cities, taking one for cities that established a smart city pilot within the sample period and 0 for those that did not. time_t is a dummy variable for the period before and after the pilot policy, taking zero before a city is approved for the smart city pilot and one after.

4.2.3. Mediating Variables

The mediating variables that reflect the three major effects of smart city construction are as follows: the technological innovation effect is measured by the number of local patent applications (Tech); the industrial structure adjustment effect is measured by the proportion of the output value of the tertiary industry to the city’s GDP (Indu); and the resource allocation optimization effect is measured by the proportion of coal consumption to total energy consumption (Engy). Coal consumption represents the traditional high-energy-consumption and high-pollution industries with high emissions, and the change in the proportion of coal consumption can indicate the application of other production factors and technological resources to the development of clean energy or the promotion of energy conservation, emission reduction, and energy efficiency improvement in the coal industry.

4.2.4. Control Variables

Economic scale (Per gdp), which is measured by the per capita GDP of each prefecture-level city; Transportation scale (Tra), which measures the intensity of urban transportation by road area, due to the emissions from high-intensity road transportation being an important source of haze pollution; Level of openness (Open), which measures a city’s level of openness to foreign direct investment. Information level (Inform) is measured by the total business volume of the postal and telecommunications industries. Population scale (Pop) is another control variable. However, as the population data provided in the China Urban Statistical Yearbook are related to registered population rather than resident population, the special household registration system in China often leads to a large deviation in measuring the actual population size of cities, which can cause significant bias in research conclusions. Therefore, this study uses the ratio of the city’s permanent population to its administrative area to measure the population scale of the city in order to make the research conclusion more accurate and realistic.

4.3. Data Description

The data in this paper were obtained from the China Statistical Yearbook and the China City Statistical Yearbook from 2008–2019. To avoid the issue of duplicate samples, where some cities were listed in the smart city pilot program at different administrative levels, the districts and counties were merged into the samples of the prefecture-level cities. Based on data availability, the final balanced panel dataset contained information for 283 prefecture-level cities in China from 2007 to 2017. Descriptions of the dependent, explanatory, and control variables, as well as their data sources, are provided in

Table 1. Data descriptions and source.

Variable Type	Indicator Selection	Symbol	Indicator Description	Data Source
Dependent Variable	Haze Pollution	PM2.5	Annual average concentration of PM2.5	Columbia University’s International Earth Science Information Network (https://beta.sedac.ciesin.columbia.edu/ accessed on 13 December 2022)
Explanatory Variable	Smart City Pilots	DID	DID = treati × timet	Patent search and analysis platform of the State Intellectual Property Office
Control Variable	Economic Scale	Per gdp	City GDP per capita	China Statistical Yearbook (2008–2019)
	Transportation Scale	Tra	Logarithm of the road area as a measure of urban transportation development
	Level of Openness	Open	Logarithm of foreign direct investment
	Information Level	Inform	Logarithm of the total business volume of the postal and telecommunications industries
	Population Scale	Pop	Logarithm of urban population per unit of administrative area
	Technological Innovation Effect	Tech	Logarithm of the number of invention patent applications
Mediating Variable	industrial structure adjustment effect	Indu	Proportion of the output value of the tertiary industry to GDP	China Urban Statistical Yearbook (2008–2019)
	Resource Allocation Optimization Effect	Engy	Coal consumption as a proportion of total energy consumption

. Descriptive statistics for each variable can be found in

Table 2. Descriptive Statistics.

Vars	Obs	Mean	SD	Min	Max
PM2.5	3113	3.5054	0.5491	1.5516	4.4629
Tra	3113	8.7277	0.8835	5.8464	11.4617
Per gdp	3113	10.3046	0.7518	7.9255	16.4854
Open	3113	9.6796	1.9762	1.5041	14.9413
Inform	3113	12.3730	1.0258	9.3917	16.5024
Pop	3113	5.7384	0.9075	1.5476	7.8816

.

5. Results and Analysis

5.1. Baseline Regression Results

A stepwise regression approach was used to examine whether smart city pilot policies contribute to haze control according to the econometric model in the study design. Given that different regions may have varying economic growth trends, this study controls for provincial-level time trends in the baseline model and utilizes city-cluster robust standard errors in models (1) to (6) (see

Table 3. Basic return.

Vars	(1)	(2)	(3)	(4)	(5)	(6)
DID	−0.0753 *** (−13.53)	0.0240 *** (−3.76)	−0.0172 *** (−2.69)	−0.0170 *** (−2.74)	−0.0165 ** (−2.56)	−0.0160 ** (−2.47)
Per gdp		−0.0739 *** (−14.73)	−0.0563 *** (−10.01)	−0.0582 *** (−9.87)	−0.0550 *** (−8.92)	−0.0546 *** (−8.85)
Tra			−0.0343 *** (−6.71)	−0.0347 *** (−6.77)	−0.0335 *** (−6.49)	−0.0332 *** (−6.42)
Open				0.0028 (1.10)	0.0033 (1.26)	0.0033 (1.26)
Inform					−0.0099 * (1.80)	−0.0100 * (−1.81)
Pop						−0.0262 * (−1.64)
Constant	3.5353 *** (20.95)	4.3734 *** (17.05)	4.3904 *** (15.24)	4.3863 *** (14.97)	4.4610 *** (13.81)	4.6053 *** (11.53)
City-FE	Yes	Yes	Yes	Yes	Yes	Yes
Year-FE	Yes	Yes	Yes	Yes	Yes	Yes
R2	0.3081	0.3841	0.4021	0.4134	0.5029	0.5536
Obs	3113	3113	3113	3113	3113	3113

Notes: ***, **, and * denote significance levels of 1%, 5%, and 10%, respectively; values in brackets are robust standard errors of clustering; and “YES” indicates controlling for firm and year-fixed effects.

). All models include individual and time-fixed effects, and the goodness of fit improves as additional control variables are introduced, indicating that all models possess good explanatory power. The results shown in Table 3 reveal several findings. First, regardless of the inclusion of control variables, smart city construction has a significant effect on reducing haze pollution. Second, in the model without any control variables (column 1), the coefficient of the core variable DID is significant at the 99% confidence interval, indicating that smart cities exhibit the most significant reduction in haze pollution, with a decrease of 7.53% in PM_2.5 concentration compared to non-smart cities. Third, in models (2) to (6), the haze-reducing effect of smart city construction gradually diminishes, but the effect remains significant. Fourth, the effects of economic scale (Per gdp), transportation scale (Tra), information level (Inform), and population scale (Pop) are consistently significant, with economic scale and transportation scale being significant at the 99% confidence interval and information level and population scale being significant at the 95% confidence interval. In contrast, the impact of the level of openness (Open) is consistently insignificant. Overall, these findings suggest that smart city construction plays a critical role in mitigating haze pollution, but its effect is moderated by other contextual factors.

5.2. Parallel Trend Test

The fundamental assumption of the multi-phase DID approach is the parallel trends hypothesis, which assumes that the experimental group and the control group had the same pre-existing trends before the smart city pilot policy impact. Therefore, it is necessary to conduct a parallel trend test on the dependent variable to meet the requirements of using the DID method to evaluate the effectiveness of the policy. Thus, this study follows the practical research method proposed by Jacobson et al. (1993) [52] to conduct a parallel trend test. The method can be expressed as follows:

$${\mathrm{PM}}_{2.5it}={\alpha }_0+{\displaystyle \sum _{t=-7}^5{\alpha }_2{D}_{it}}+{\alpha }_3{X}_{it}+{\gamma }_t+{\mu }_i+{\epsilon }_{it}$$

(4)

where D_it is a set of dummy variables, taking the value of 1 if city i implements the smart city pilot policy in year t and 0 otherwise. α₁ is a constant term, α₃ represents the coefficient of the control variables, and the meanings for other variables have the same meaning as those in Equation (1). This study focuses on the coefficient of α₂, which reflects the difference in haze pollution between pilot and non-pilot cities in the year t when the smart city pilot policy is implemented. Additionally, this study takes the 7th period before the implementation of the smart city pilot policy as the baseline period.

The results of the parallel trends test, shown in Figure 3 below, indicate that the coefficient estimates for most years before the implementation of the smart city pilot policy are not significant. This suggests that there are no significant differences between pilot and non-pilot cities before policy implementation, and the research sample passes the parallel trend test.

5.3. Placebo Tests

5.3.1. Time Placebo Test

The differences in haze pollution between cities in the treatment and control groups are due to time variations. In this paper, false policy implementation times were constructed by advancing the smart city pilot policy implementation time by 4 years, 3 years, 2 years, and 1 year, denoted as smartcitypost^-false4, smartcitypost^-false3, smartcitypost^-false2, and smartcitypost^-false1, respectively, and regressing Equation (1) with these false policy implementation times. The results (shown in

Table 4. Regression results of time-placebo test.

Vars	PM2.5 (−4)	PM2.5 (−3)	PM2.5 (−2)	PM2.5 (−1)
Smartcitypost-false	−0.0046 (0.0081)	−0.0075 (0.0073)	−0.0089 (0.0070)	−0.0105 (0.0068)
Per gdp	−0.0364 *** (0.0078)	−0.0363 (0.0078)	−0.0363 (0.0078)	−0.0362 (0.0078)
Tra	−0.0235 *** (0.0051)	−0.0235 (0.0051)	−0.0234 (0.0051)	−0.0234 (0.0051)
Open	0.0032 (0.0025)	0.0031 (0.0025)	0.0031 (0.0025)	0.0031 (0.0025)
Inform	−0.0086 (0.0053)	−0.0085 (0.0053)	−0.0085 (0.0053)	−0.0085 (0.0053)
Pop	−0.0264 * (0.0147)	−0.0264 (0.0147)	−0.2562 (0.0147)	−0.2548 (0.0147)
Constant	4.3507 (0.0142)	4.3487 (0.1311)	4.3447 (0.1332)	4.3427 (0.1332)
City-FE	Yes	Yes	Yes	Yes
Year-FE	Yes	Yes	Yes	Yes
R2	0.676	0.676	0.676	0.676
Obs	3113	3113	3113	3113

Notes: *** and * denote significance levels of 1% and 10%; values in brackets are robust standard errors of clustering; and “YES” indicates controlling for firm and year-fixed effects.

) indicate that the coefficient estimates for smartcitypost^-false4, smartcitypost^-false3, smartcitypost^-false2, and smartcitypost^-false1 are not significant at the 10% level. This indicates that there is no systematic difference in the time trends between the treatment and control groups of cities and again demonstrates that the smart city pilot policy promotes the reduction of haze pollution in cities.

5.3.2. City Placebo Test

To avoid the potential bias resulting from unobserved variables in the baseline regression, this study conducted a city placebo test by replacing the treatment group cities with a randomly selected group of 148 cities from the sample and treating the remaining cities as the control group (results shown in Figure 4). The coefficients estimated from this placebo test provide an estimation of the impact of the smart city pilot policy on haze pollution in cities. This process was repeated 500 times to obtain 500 regression coefficients and their corresponding p-values. The kernel density distribution and p-values of these 500 regression coefficients indicate that the majority of the coefficients are not significant and are distributed around the value of zero, following a normal distribution. The estimated coefficient from the baseline regression is located in the high tail of the distribution of the placebo regression coefficients, and it is a rare event in the city placebo test. Therefore, the possibility of the baseline results being affected by unobserved factors can be ruled out.

5.4. Robustness Tests

5.4.1. PSM-DID

To address the potential bias resulting from systematic differences in the trend of smart cities and other cities, a propensity score matching difference-in-differences (PSM-DID) approach was employed for robustness testing. First, a logistic model and nearest neighbor matching method were used to conduct 1:3 with replacement matching and 1:1 without replacement matching for the 148 treatment cities in each year during the study period. Second, the balance of the matched sample was checked, and the results showed that the standard deviation of the matching variables decreased significantly after matching, and there were no significant differences in observable characteristics between the treatment and control groups. These results indicate that the matching variables and matching methods used in this section were appropriate. Finally, the matched control group and treatment group were used in the baseline difference-in-differences regression model, and the results are shown in

Table 5. Robustness tests.

Vars	(1)	(2)	(3)	(4)
DID	−0.0163 *** (−2.72)	−0.0155 *** (−2.83)	−0.0160 ** (−2.47)	−0.0172 ** (−2.88)
Per gdp	−0.0232 *** (−5.33)	−0.0312 *** (−4.98)	−0.0550 *** (−8.90)	−0.0121 *** (−3.43)
Tra	−0.0293 *** (−5.63)	−0.0217 *** (−6.18)	−0.0334 *** (−6.50)	−0.0378 *** (−7.86)
Open	0.0043 (0.56)	0.0032 (0.42)	0.0035 (1.26)	0.0056 (1.41)
Inform	−0.0062 ** (−2.11)	−0.0057 ** (−2.02)	−0.0101 * (−1.79)	−0.0102 * (−1.75)
Pop	−0.0255 ** (−1.90)	−0.0241 ** (−1.92)	−0.0238 * (−1.65)	−0.0242 * (−1.71)
Constant	4.4633 *** (10.22)	4.5531 *** (10.43)	4.5344 *** (12.02)	4.6445 *** (12.55)
City-FE	Yes	Yes	Yes	Yes
Year-FE	Yes	Yes	Yes	Yes
R2	0.5430	0.5234	0.5535	0.5320
Obs	3113	3113	3113	3113

Notes: ***, **, and * denote significance levels of 1%, 5%, and 10%, respectively; values in brackets are robust standard errors.

. Column (1) shows the regression results of 1:3 with replacement matching for the treatment group, and column (2) shows the results of 1:1 without replacement matching for the treatment group.

5.4.2. Excluding the Effects of Special Cities

To address the potential bias from China’s first-tier cities with high levels of digitalization and favorable development conditions, the baseline difference-in-differences model was re-estimated by excluding four first-tier cities: Beijing, Shanghai, Guangzhou, and Shenzhen. The results, presented in column (3) of Table 3, demonstrate the robustness of our findings. Specifically, it finds out that the smart city pilot policy continues to have a significant negative impact on haze pollution despite the exclusion of these four cities, indicating that the baseline results are not driven solely by the unique characteristics of first-tier cities in China and further supporting the effectiveness of the smart city pilot policy in reducing haze pollution.

5.4.3. Replacing the Explanatory Variables

To test the robustness of the regression results, urban sulfur dioxide annual concentration was used instead of PM_2.5 to examine whether the smart city pilot policy is effective in reducing sulfur dioxide pollution. The regression results are shown in column (4) of Table 3. From Table 3, it can be seen that the smart city pilot policy still has a significant negative impact on pollution, indicating that the baseline regression results are robust.

5.5. Mediating Mechanisms Tests

In accordance with the principle of mediation analysis, this paper used the effects of technological innovation, industrial structure adjustment, and resource allocation optimization as mediating variables to test their effects on the relationship between smart city construction and haze pollution reduction as presented in Formula (2) (results shown in

Table 6. Mediating effects test.

TI			ISA			RAO
Formula	(2)	(3)	Formula	(2)	(3)	Formula	(2)	(3)
DID	0.2305 ** (7.05)	−0.0087 * (−1.71)	DID	2.2158 *** (8.88)	−0.0159 ** (−2.43)	DID	−0.0021 * (−1.72)	−0.0160 ** (−2.47)
Tech		−0.0338 ** (−9.30)	Indu		−0.0002 (−0.58)	Engy		0.1982 *** (5.07)
Control	Yes	Yes	Control	Yes	Yes	Control	Yes	Yes
R2	0.5832	0.6917	R2	0.4426	0.5310	R2	0.5032	0.5573
Obs	3113	3113	Obs	3113	3113	Obs	3113	3113

Notes: ***, **, and * denote significance levels of 1%, 5%, and 10%, respectively; values in brackets are robust standard errors.

). The first step’s regression results showed a significant positive correlation between smart city construction and the three effects. The coefficients for technological innovation, industrial structure adjustment, and resource allocation optimization were all positive, indicating that smart city innovation drives the five innovations and subsequently produces the three effects, which is consistent with the analysis and summary of the mechanism in the previous section. While this paper has confirmed that smart city construction helps reduce haze pollution, further investigation is still needed to examine its impact mechanism. The step-two regression results showed that the coefficient for the impact of technological innovation on haze pollution was significant, whereas the coefficient for industrial structure adjustment was not significant, and the coefficient for resource allocation optimization was negative and significant, indicating that adjusting energy consumption structure did not directly reduce haze pollution.

Based on the test results, this paper further analyzed the relationships between the three effects and haze pollution. It can be found that the technological innovation effect is the most significant and effective, in line with the previous analysis and hypothesis; that is to say, the technological innovation effect driven by technological innovation and product innovation can be directly applied to the field of haze control, whether it is intelligent monitoring, early warning, and control of pollution emissions by governing entities or the application of emerging technologies by production entities, all of which achieve the purpose of increasing energy efficiency and reducing pollution and are directly reflected in the reduction of haze pollutants emissions.

The test results showed that there was a significant positive correlation between smart city construction and industrial structure adjustment effects, indicating that smart city construction helps to optimize and upgrade urban industrial structures. However, the regression analysis in step two showed that there was no significant correlation between industrial structure adjustment effects and the reduction of haze pollution. The reason may be that the adjustment of industrial structure by smart city construction is still in its early stages, with most cities still relying on the secondary industry as their economic backbone and development mainstay, and the haze reduction effect of the tertiary industry has a lag and has not yet been fully realized in the sample cities of this study. Step one showed that smart city innovation drives the optimization of resource allocation, which was significant, indicating that market innovation, resource allocation innovation, and organizational innovation are positively correlated with the optimization of resource allocation effects.

5.6. Heterogeneity Analysis

Regional differences in urban development across China are significant, resulting from various factors such as geographical location, natural conditions, and government policies. Cities in the eastern region hold significant advantages in policy, economic conditions, talent level, industrial capital, and technological innovation when compared to cities in the central and western regions. Consequently, the construction of smart cities may vary due to regional disparities. This paper argues that the efficacy of smart cities in reducing haze will vary due to regional differences in the three major regions of eastern, central, and western China. In this paper, the cities were divided into three categories—eastern, central, and western—to conduct regression analysis, and the results are detailed in

Table 7. Test for heterogeneity of test area categories.

Vars	Eastern Region	Central Region	Western Region
DID	−0.0385 *** (−4.33)	0.0120 (1.00)	−0.0243 * (−1.87)
Per gdp	−0.0212 *** (−3.14)	−0.0948 *** (−5.80)	−0.1284 *** (−9.12)
Tra	−0.0686 *** (−8.25)	−0.0344 *** (−3.67)	0.0156 * (1.64)
Open	−0.0038 (−1.02)	0.0186 *** (3.08)	0.0081 * (1.82)
Inform	0.0026 (0.30)	−0.0047 ** (−0.43)	−0.0045 (−0.50)
Pop	−0.0136 ** (−0.85)	−0.0646 ** (−1.96)	0.1563 (1.55)
Constant	4.6190 *** (8.24)	5.1527 *** (8.52)	3.4712 *** (6.99)
City-FE	Yes	Yes	Yes
Year-FE	Yes	Yes	Yes
R2	0.6842	0.5039	0.6685
Obs	1111	1091	911

Notes: ***, **, and * denote significance levels of 1%, 5%, and 10%, respectively; values in brackets are robust standard errors.

.

The heterogeneity test shows that the development of smart cities in the eastern and western regions has resulted in significant haze reduction, with the reduction effect of smart cities in the east being higher than that in the west. However, the development of smart cities in the central region has not only failed to reduce haze but has even worsened haze pollution. This finding supports the hypothesis that the effect of smart city construction on haze reduction has regional differences.

The reason why cities in the eastern part of China are better positioned for the construction of smart cities is due to their geographical advantages and policy incentives. Prior to the implementation of smart city development, the eastern cities had already accumulated a strong foundation in terms of talent, capital, technology, information exchange, advanced management models, and experience in urban development and production factors. As a result, they occupy a more advantageous position in the construction of smart cities compared to cities in the central and western regions. The eastern cities, relying on their accumulated production and technological factors, have already entered a higher stage of innovation-driven development and have demonstrated more significant effects on environmental governance through their smart city construction. In contrast, cities in the central region, located between the east and west, lack the advanced economy and innovation foundation of the eastern region, as well as the abundant natural resources of the western region. Specifically, the scale and quality of urban development in the central region have not reached the level of the eastern region, yet they face greater population pressure compared to the western region. The industrial structure in the central region is still driven by labor and other traditional production factors, and it has undertaken the transfer of a significant amount of the high-energy-consuming and high-emission industries from the eastern region. In addition, the central region has a relatively weak foundation for building smart cities due to its lack of attraction to talent and innovation factors. Since the central region was established as a pilot city for smart cities, the central cities have been required to prioritize the completion of new infrastructure and promote the application of high-tech industries and smart technologies in industrial parks, office buildings, road facilities, and so on. However, most of the central cities are still in the process of expanding their urban areas and improving their public transport systems, so the difficulties faced by cities in the central region in building smart cities can be described as “inherent deficiencies and acquired deformities”, and currently, the efficacy of their smart city construction in reducing haze has not been fully realized.

Although the development level of the western region is not as high as that of the central region, the heterogeneity test results show that the effect of smart city construction on reducing haze is still significant in the western region. This is due to the fact that western cities are more geographically dispersed, and the spatial spillover effects of haze pollution have a relatively small intra-regional and inter-city impact. Furthermore, from the perspective of geographical location and meteorological conditions, the western region is less likely to form secondary pollution caused by the accumulation of haze pollution, thus favoring the decomposition of primary haze pollution. At the same time, the size and population density of cities in the western region are generally smaller than those in the central and eastern regions, and the industrial structure is less dependent on labor-driven, energy-intensive industries. In addition, the western region actively utilizes its rich natural and cultural landscapes to develop low-emission cultural tourism industries. From the perspective of energy consumption structure, the western region has abundant clean energy sources, such as wind and hydropower, which restrain pollution emissions caused by fossil energy. In recent years, with the maturity and popularization of big data and cloud computing technologies, branches of multinational companies in China have often set up their cloud computing centers in Guizhou, and Chinese domestic leading technology companies such as Tencent and Alibaba have also similarly placed their big data centers and servers in the western region. The deployment and settlement of these enterprises not only promote local development but also facilitate the circulation and application of technology in the local area, thereby helping the western region achieve low-pollution growth dynamics.

Overall, the findings of this study suggest that regional differences in urban development, talent level, and technological innovation significantly influence the efficacy of smart city construction in reducing haze. The eastern region has taken the lead in smart city construction due to its strong foundation in urban development and production factors. The western region has achieved significant haze reduction effects due to its advantageous geographical location, low dependence on high-energy-consuming industries, abundant clean energy sources, and low population density. However, the central region faces significant difficulties due to its weak foundation in smart city construction, high dependence on labor-driven, high-energy-consuming industries, and population pressure. Cities in the central region should prioritize the completion of new infrastructure construction and promote the application of high-tech industries and intelligent technologies to achieve low-pollution growth.

6. Conclusions and Policy Recommendations

Based on three batches of smart city pilots to construct a quasi-natural experiment, this paper selected panel data from 283 prefecture-level cities in China from 2007 to 2017 to test the impact of smart city pilots on haze pollution through an empirical model by utilizing the multi-phase difference-in-differences method. The research findings show that, compared with non-pilot cities, smart city pilots have a significant positive impact on reducing urban haze pollution. The mechanism tests show that smart city pilots, coupled with ICT technology driven by innovation, have lowered the annual average concentration levels of urban PM_2.5 pollutants through the effects of technological innovation, industrial structure adjustment, and resource allocation optimization. Heterogeneity analysis shows that there are regional differences in the haze reduction effect of smart cities. The intensity of haze reduction decreases from the eastern region to the western region of China, but the smart city pilots in the central region have not yet exerted their haze reduction effect. This paper is an important reference for improving the urban environment and ecology, creating an environmentally friendly city, and creating a green living and production environment. In conjunction with the findings of the study, this paper makes the following policy recommendations:

First, foster innovation and technology integration, invest in the growth of human capital, and support clean energy industries. To support China’s "dual carbon" goals and strengthen the haze-mitigating effects of smart city construction, the government should boost funding for research and development, encourage the fusion of technology and urban management, and support the emergence of new eco-friendly industries. In order to lessen dependency on fossil fuels, this entails supporting clean energy technology and renewable energy sources, as well as investing in human capital development to create a workforce with the necessary skills to support the advancement of these technologies.

Second, investment in infrastructure development should be amplified to advance sustainable transportation and foster low-carbon urban planning. To actualize the vision of sustainable smart cities, it is crucial for the government to channel substantial resources into infrastructure development, with a concentrated emphasis on sustainable transportation and low-carbon urban planning. These initiatives play a pivotal role in reducing carbon emissions and promoting eco-friendly transportation alternatives. In light of this, it is advised that governmental investment in infrastructure projects, particularly in the central region, be escalated to bolster the evolution of sustainable transport systems and advance low-carbon urban planning. This infrastructure development should prioritize the utilization of low-carbon materials and the promotion of energy efficiency. Taking into account the importance of segregating industrial activity from city areas, the urban planning process should ensure that the locations for such activities are strategically planned. This consideration aids in decreasing the potential negative impact on air quality and residents’ health.

Third, tailor policies to regional conditions and promote public participation. It is recommended that the government develop regional policies that consider the variability of mitigating haze effects across different regions. For example, in regions with high levels of industrial pollution, the emphasis should be on pollution control measures, while in regions with abundant natural resources, the emphasis should be on promoting green development. Additionally, the government should encourage public participation in the planning and implementation of smart city pilots, enhance public awareness of the benefits of smart city pilots for the environment, encourage public feedback and suggestions on pollution control measures, and promote a culture of environmental responsibility among citizens.

Fourth, smart city construction is an important way to innovate in urban governance. At present, China is in a difficult period of climbing over the hurdles of low-carbon energy transformation, and many structural contradictions such as coal-dependent energy, heavy structure, and low efficiency are still prominent. The proposed smart city has an important role in promoting the improvement of urban energy structures. On the one hand, through digital technology to accelerate the deep integration of traditional energy industry and digital industry, optimize energy production and sales, supply and demand on both sides, and improve energy utilization efficiency; on the other hand, relying on big data, cloud computing, the Internet of Things, and other technologies to strengthen the government’s collection, analysis, and processing of haze pollution data, enhance the government’s supervision of haze pollution and early warning capability, and force the energy structure adjustment of enterprises with digital technology supervision.

Fifth, smart city construction will promote the digitalization and intelligence of urban production and life. On the one hand, relying on convenient and efficient internet technology and numerous self-media platforms, it will strengthen the publicity and education of environmental protection concepts, with the goal of improving the quality of the urban population and creating a culture of low-carbon and green urban development, reducing the urban pollution problems brought about by high population density. On the other hand, the construction of smart cities has created a large number of digital industries with a strong demand for electricity, while the western region is rich in clean energy and the price of electricity is more favorable compared to the eastern and central regions. Domestic head technology companies have strengthened the industrial layout of the western region, with industrial migration driving the flow of population, which is conducive to alleviating the problem of high urban pollution caused by high population density in the eastern and central regions.