Coordinated reduction of pollutant and carbon emissions through smart city construction: TOE-based mechanisms and spatial spillovers in China

Coordinated reduction of pollutant and carbon emissions

The conceptual framework of “co-benefits” was pioneered by the Intergovernmental Panel on Climate Change (IPCC) in its Third Assessment Report, initially defined as the non-climate advantages - such as enhanced air quality - accruing from greenhouse gas mitigation policies. This definition has since evolved into a more dynamic understanding of the bidirectional synergy between climate mitigation and local pollutant control^[18]. CRPC has become an important theme in environmental economics and urban sustainability research. The core idea is that, because conventional air pollutants and greenhouse gas emissions often share common sources, especially fossil fuel combustion, policies that alter energy use, industrial production, and transport systems may generate simultaneous environmental benefits^[2]. In the context of China, this issue is particularly important because the country has long faced the dual challenge of carbon mitigation and local air pollution control, while its industrial structure and energy system create substantial overlap in the sources of both types of emissions^[19,20].

Existing studies show that the effectiveness of CRPC depends on a range of economic, technological, and institutional factors. Economically, the achievement of co-reduction is contingent upon robust financial investment and the fundamental configuration of the urban economy^[18]; specifically, industrial gross domestic product (GDP) scale and the sophistication of the industrial structure are decisive factors in determining emission intensity^[10,21,22]. Furthermore, the interaction between financing constraints and pollutant emission levies has been found to trigger a clean energy substitution effect, which further amplifies these environmental synergies^[20,23]. Within the energy dimension, there is a broad consensus that clean energy deployment, enhanced energy efficiency, and structural optimization of the energy mix constitute the core technical drivers for elevating synergistic performance^[22,24-27]. From a policy perspective, including low-carbon pilot programs, emissions trading systems, and air pollution prevention policies, have been shown to improve co-reduction performance by promoting cleaner production, optimizing energy allocation, and encouraging green innovation^[9-12].

Although this literature has generated important insights, two limitations remain. First, many studies focus on environmental policies explicitly designed for pollution control or carbon mitigation, while less attention has been paid to broader urban governance reforms that may generate indirect environmental benefits. Second, the mechanisms underlying co-reduction are often examined from a single perspective, such as energy structure or innovation, without integrating technological, organizational, and institutional dimensions into a unified analytical framework. These limitations suggest the need to examine whether urban digital governance, particularly smart city construction, can contribute to coordinated environmental improvement and through what channels.

Smart city construction and environmental performance

The smart city concept generally refers to the use of digital infrastructure, information networks, and intelligent management systems to improve urban operation, public services, and governance capacity^[28]. Recent studies suggest that smart city construction may also have environmental consequences. By strengthening information sharing, increasing monitoring capacity, improving resource allocation, and facilitating data-driven decision-making, smart city programs may reduce energy waste, improve pollution control efficiency, and support low-carbon urban development^[28-31].

Empirical research has provided preliminary evidence that smart city construction can improve environmental quality, promote green innovation, support industrial upgrading, and enhance energy efficiency^[32,33]. However, the existing evidence remains fragmented in three respects. First, many studies examine single outcomes, such as carbon emissions, air quality, or green total factor productivity, rather than CRPC. Second, the literature tends to emphasize average treatment effects while paying less attention to heterogeneity across cities with different regulatory conditions, growth stages, or resource endowments. Third, spatial interdependence has not been sufficiently explored, even though digital governance, industrial relocation, and technology diffusion may generate environmental spillovers across connected cities.

TOE framework and its relevance to this study

The TOE framework provides a useful perspective for understanding how policy interventions translate into measurable outcomes. The framework argues that the adoption and effectiveness of innovation are jointly shaped by technological conditions, organizational capacity, and the external environment^[34]. Because smart city construction is not merely a technological project but also a governance reform embedded in institutional and socioeconomic contexts, the TOE framework is well suited to analyze its environmental implications.

Unlike prescriptive models that dictate specific outcomes, the TOE framework serves as a versatile taxonomy, offering clear structural boundaries and highly extensible variables^[35]. In the environmental field, previous studies have applied the TOE framework to examine green innovation, low-carbon transition, energy efficiency improvement, and sustainable supply-chain management^[36-39]. These studies suggest that environmental outcomes are rarely driven by technology alone. Instead, they depend on the interactions among technological upgrading, organizational adaptation, and external environmental conditions^[40]. This insight is particularly relevant to the present study. Smart city construction may improve co-reduction not only by expanding digital infrastructure and innovation capacity, but also by reshaping industrial organization, enhancing human capital, and strengthening ecological governance and pollution control.

In summary, the existing literature indicates that smart city construction may influence environmental performance, but it does not yet provide sufficient evidence on whether such policies promote CRPC, through which mechanisms this occurs, and whether the effects extend beyond pilot cities. This study addresses these questions by combining a city-level quasi-natural experiment with a TOE-based analytical framework. Compared with conventional mechanism analyses that typically focus on a single transmission channel, such as technological innovation, industrial upgrading, or environmental regulation, the TOE framework provides a more comprehensive perspective by integrating technological conditions, organizational capacity, and external environmental support into a unified analytical system. Smart city construction is essentially a complex governance reform involving digital technologies, institutional coordination, industrial transformation, and environmental governance simultaneously. Therefore, relying on a single mechanism may overlook the interactions among different dimensions. The TOE framework enables this study to systematically examine multiple transmission channels and better explain how smart city construction promotes the coordinated reduction of pollutant and carbon emissions.

Direct effect of smart city construction on co-reduction

As a transformative strategic trajectory for modern urban evolution, smart city construction leverages cutting-edge information technologies to fundamentally reconfigure urban governance paradigms. This shift significantly enhances the efficiency of resource allocation and utilization, fostering agglomeration effects that provide tech-enabled solutions to deep-seated urban challenges^[3,41]. Consequently, this systemic optimization offers a potent catalyst for environmental governance, facilitating a transition toward the coordinated management of pollution and carbon emissions.

The impact of smart city construction on CRPC is manifested through four interconnected pillars: smart economy, smart ecology, smart governance, and smart living. Within the smart economy, the digitalization of urban financial systems and the growth of the digital economy have been empirically shown to bolster environmental sustainability^[42-44]. This economic vitality enhances a city’s capacity to invest in advanced transportation and telecommunication infrastructures, which underpin long-term green development^[45]. Parallel to this, smart ecology integrates intelligent energy systems and sophisticated regulatory strategies to accelerate renewable energy adoption and optimize energy efficiency. By curbing the environmental externalities of excessive energy consumption, these systems establish a rigorous technical foundation for urban CRPC^[28,46,47]. Furthermore, core ecological components such as expanded green coverage and intelligent waste management protocols provide natural pathways for simultaneous pollution and carbon mitigation^[48].

In the realm of smart governance, the effectiveness of CRPC is highly contingent upon governmental priorities and institutional motivation. Increased administrative emphasis on environmental protection, coupled with enhanced public service motivation among officials, has been shown to yield superior governance outcomes^[49,50]. The deployment of specialized technologies, such as remote sensing for aquatic ecosystem monitoring, further provides the precision required for sustainable environmental management^[50]. Finally, smart living initiatives drive a cultural shift toward ecological civilization. By integrating big data analytics into intelligent transportation systems, smart cities alleviate urban congestion and optimize public transit networks^[51]. These advancements, combined with digital platforms that promote sustainable household consumption and green travel preferences, effectively reduce indirect emissions and cultivate low-carbon lifestyles among residents^[4,52,53].

Nevertheless, the co-reduction effect of smart city construction is not unconditional. Improvements in digital infrastructure and energy efficiency may lower production costs and stimulate additional economic activities, thereby increasing energy demand and partially offsetting the expected environmental benefits through the energy rebound effect^[54]. Consequently, the net environmental effect of smart city construction depends on whether technological progress and governance improvements outweigh the potential rebound effect. In the Chinese context, where smart city policies emphasize green digital transformation, environmental regulation, and industrial restructuring simultaneously, the overall effect is expected to remain positive.

Taken together, these changes suggest that smart city construction may promote CRPC rather than improving only one environmental dimension. On this basis, we propose the following hypothesis:

Hypothesis 1: Smart city construction significantly promotes CRPC.

Indirect effects under the TOE framework

Environmental dimension

The third set of mechanisms operates through improvements in urban environmental governance. By integrating digital technologies into environmental management and public decision-making, smart city construction may create more favorable conditions for ecological governance and pollution control^[64]. Through enhanced planning capacity, improved information transparency, and more efficient allocation of public resources, smart city initiatives may strengthen ecological investment and support the implementation of urban greening programs^[65]. The application of digital governance tools further enables local governments to identify ecological weaknesses, monitor the implementation of environmental projects, and incorporate ecological objectives into urban development strategies in a more timely and coordinated manner. Increased investment in urban greening and ecological construction has the potential to improve environmental quality, enhance carbon sequestration capacity, and create conditions conducive to long-term low-carbon transformation^[46,66].

Smart city construction may also enhance pollution control capacity by supporting intelligent environmental governance^[31]. The deployment of real-time monitoring systems, intelligent sensing technologies, cross-departmental information sharing, and data-driven enforcement can improve the precision and responsiveness of environmental regulation while reducing the cost of environmental supervision^[67-69]. These improvements may encourage greater compliance among firms, strengthen the treatment of environmental externalities, and facilitate the more effective implementation of pollution control policies. As environmental governance becomes more timely, targeted, and transparent, cities may be better positioned to simultaneously reduce conventional air pollutants and carbon emissions. Taken together, by promoting ecological construction and strengthening pollution control capacity, smart city construction may enhance cities’ environmental governance capability^[70]. These environmental improvements are therefore expected to serve as potential transmission channels through which smart city construction facilitates CRPC.

Accordingly, we propose:

Hypothesis 4: Smart city construction may promote CRPC through the potential transmission channels of ecological construction and pollution control capacity.

The mechanism flowchart of this study is presented in Figure 1.

Figure 1. Research framework. TOE: Technology-organization-environment; CRPC: coordinated reduction of pollutant and carbon emissions.

Models

China’s smart city pilot program was organized into three annual pilot cohorts, officially designated as the 2012, 2013, and 2014 batches. This study treats these three batches of smart city pilots as quasi-natural experiments and adopts a multi-period DID model, following the approach of^[71], to empirically examine the impact of smart city construction on the coordinated reduction of pollutant and carbon emissions. The baseline econometric model is specified as follows:

We estimate the following baseline models:

where SO_2,it represents the SO₂ emission intensity of city i in year t; CO_2,it denotes the carbon emission intensity of city i in year t; DID_it is a dummy variable that equals 1 if city i started the smart city pilot in year t, and 0 otherwise; Control_it refers to a set of control variables; λ_i represents city fixed effects, μ_t denotes time fixed effects, ε_it is the stochastic error term. The core coefficients of interest are β₁ and β₂. During the empirical testing process, four possible scenarios may arise:

Scenario I: β₁ > 0, β₂ > 0;

Scenario II: β₁ > 0, β₂ < 0;

Scenario III: β₁ < 0, β₂ > 0;

Scenario IV: β₁ < 0, β₂ < 0.

A significant synergistic effect of pollution and carbon reduction is considered to exist only when Scenario IV is observed and the coefficients are statistically significant.

To further investigate the mechanism through which smart city construction affects the coordinated reduction of pollutant and carbon emissions, this study draws on the research of^[72] and establishes the following mediating effect model based on TOE framework:

where y_it represents the dependent variable, which includes SO_2,it sulfur dioxide emission intensity and CO_2,it carbon dioxide emission intensity; M_it denotes the mediating variable. All other variables remain consistent with those defined in the baseline regression.

Variables

Independent variable

The independent variable is Smart City Construction (DID), a policy dummy capturing the implementation of the smart city pilot program. Since 2012, China has integrated smart city construction into its national urbanization strategy, utilizing digital technologies to advance urban governance. The Ministry of Housing and Urban-Rural Development designated three batches of smart city pilots in 2012, 2013, and 2014, covering 290 cities, counties, and towns. Although the list of the third cohort was formally released in April 2015, it was officially titled the “2014 Annual National Smart City Pilot List”. Therefore, this study identifies the third cohort according to its official batch year, namely 2014. This study focuses on prefecture-level cities, excluding subordinate districts, county-level cities, and towns. After removing cities with incomplete or low-quality data, 92 prefecture-level cities are retained as the treatment group. The control group comprises 167 prefecture-level cities never selected as pilots and meeting the same data standards. A DID approach is adopted: the treatment group is assigned a value of 1 (others 0), and the policy period is coded as 1 starting from the approval year (0 before). The interaction term “ treatment group × policy time” captures the dynamic marginal effect of the policy shock.

This variable captures the average treatment effect of smart city construction on urban environmental performance under the multi-period DID framework.

Figure 2 is a geographical distribution map of smart city pilot policy.

Figure 2. The geographical distribution of smart city pilot policy. Base map source: Standard Map Service, Ministry of Natural Resources, China (http://bzdt.ch.mnr.gov.cn/), approval number GS(2024)0650. The base map has not been modified.

Sample and data

The sample covers 259 prefecture-level cities in China over the period 2006-2022. Following data availability and consistency, this study retains 92 pilot cities in the treatment group and 167 non-pilot cities in the control group. County-level cities, districts, and observations with severe data deficiencies are excluded to ensure comparability across cities. Specifically, CO₂ emission data are obtained from the China Emission Accounts and Datasets (CEADs), SO₂ emission data are sourced from the China City Statistical Yearbook, and macroeconomic variables are deflated using price indices from the National Bureau of Statistics to ensure temporal comparability.

The list of smart city pilots is compiled from official policy documents. Socioeconomic, environmental, and urban construction variables are collected from relevant official statistical yearbooks, city-level statistical sources, and other publicly available government publications. To ensure temporal comparability, monetary variables are deflated to constant prices. For a limited number of missing observations, the panel was completed using linear interpolation. Since the missing values do not exceed 2% of the sample, they will not have a significant impact on the empirical test results. All statistical analyses were conducted using Stata 18. Supplementary data processing and visualization were performed using Python 3.13, and spatial mapping was performed using ArcGIS.

Table 1 and Table 2 are the variable definitions and descriptive statistics of the variables, respectively.

Baseline results

The baseline estimates in Table 3 show that smart city construction significantly reduces both sulfur dioxide emission intensity and carbon emission intensity, indicating that the policy generates a clear co-reduction effect. This finding supports Hypothesis 1 and suggests that smart city construction has environmental consequences beyond its original role in digital urban governance.

The negative coefficients on the smart city policy variable imply that digital governance can contribute to the integrated control of conventional air pollutants and carbon emissions at the city level. A plausible explanation is that smart city construction improves environmental monitoring, strengthens interdepartmental coordination, and enhances the efficiency of resource allocation in areas such as energy use, transportation, and public management^[31].

Overall, the baseline results indicate that smart city policy should not be understood solely as a technological modernization strategy. It also functions as a governance instrument with measurable environmental benefits in the context of CRPC.

Parallel trend test

The parallel trends assumption test is a key prerequisite in policy evaluation methods such as the DID approach. Its core concept is that, prior to policy implementation, the dependent variable in the treatment and control groups should exhibit similar trends. To test this, this paper specifies the following model:

Since the smart city pilot policy was implemented in staggered batches, event time is defined relative to each city’s pilot year. In Equations (8) and (9), $$ D_{i t}^{k} $$ is an event-time indicator that equals 1 when city i is in the k-th year relative to policy implementation and 0 otherwise. The event window ranges from five years before to seven years after policy implementation (-5 to +7), with event time -1 omitted as the reference period. The coefficients β_1k and β_2k capture the dynamic policy effects on SO₂ and CO₂ emission intensity, respectively. All other variables are defined consistently with the baseline model.

Figures 3 and 4 present the estimated event-time coefficients over the window from -5 to +7. Except for an isolated SO₂ estimate at event time -2, the pre-policy coefficients are statistically insignificant and exhibit no systematic trend. For both outcomes, the policy effects are insignificant during the first two post-policy years, become significantly negative from the third year onward, and subsequently increase in absolute magnitude, indicating a delayed and gradually strengthening environmental effect.

Figure 3. Parallel trend test of SO₂ emission intensity. The horizontal dashed line denotes the zero-effect benchmark, the vertical dashed line marks the year of policy implementation (event time = 0), and the capped vertical bars represent 95% confidence intervals.

Figure 4. Parallel trend test of CO₂ emission intensity. The horizontal dashed line denotes the zero-effect benchmark, the vertical dashed line marks the year of policy implementation (event time = 0), and the capped vertical bars represent 95% confidence intervals.

To address potential heterogeneous treatment effects, we further conduct a Goodman-Bacon decomposition and apply the Callaway-Sant’Anna group-time average treatment effect on the treated (ATT) estimator. Table 4 shows that the two-way fixed-effects (TWFE) estimates are primarily driven by comparisons between pilot and never-treated cities, while potentially problematic comparisons receive limited weight. The Callaway-Sant’Anna estimates remain significantly negative for both outcomes, indicating that accounting for treatment-effect heterogeneity does not alter the main findings.

Robustness tests

Multi-period DID combined with placebo test

To avoid the potential impact of unobservable factors at the city and year levels on the regression results, this study randomly assigns smart city pilot cities and pilot years to create a pseudo-core explanatory variable for multi-period DID placebo tests. The model is structured as follows:

(10) $$ S O_{2,i t}=\alpha_{1}+\beta_{1} {DID}\_aw_{i t}+\gamma_{1} { Control }_{i t}+\lambda_{i}+\mu_{t}+\varepsilon_{i t} $$

(11) $$ C O_{2,i t}=\alpha_{2}+\beta_{2} {DID}\_aw_{i t}+\gamma_{2} { Control }_{i t}+\lambda_{i}+\mu_{t}+\varepsilon_{i t} $$

This study conducts 500 rounds of random sampling estimations based on Equations (10) and (11) respectively. As shown in Figure 5 and Table 5, the placebo coefficients are generally centered around zero, whereas the actual baseline estimates are located in the left tails of the corresponding distributions. The empirical p-values are 0.002 for SO₂ emission intensity and 0.0659 for CO₂ emission intensity. These findings provide strong placebo-test support for the SO₂ result and supportive evidence for the CO₂ result at the 10% level, suggesting that the baseline estimates are unlikely to be driven by random treatment assignment.

Figure 5. SO₂ and CO₂ emission intensity placebo test. The gray vertical dashed lines mark zero, the red vertical dashed lines indicate the actual baseline estimates, and the red horizontal dashed lines denote the P = 0.10 threshold.

Table 5

Placebo test results (500 iterations)

Statistic	SO2	CO2
Replications	500	500
Mean coefficient	-0.0006	0.0011
Standard deviation	0.0164	0.0269
5th percentile	-0.0267	-0.0431
95th percentile	0.0263	0.0448
Share of P-values < 0.10	0.1880	0.3460
Empirical P-value	0.0020	0.0659

Heterogeneity analysis

Potential mechanism analysis

Discussion

The empirical results enrich the literature in at least three respects. First, they show that a policy not originally designed as a direct environmental intervention can still generate substantial environmental gains when it reshapes urban governance capacity, data integration, and resource allocation. This finding extends the existing co-reduction literature, which has mainly focused on environmental policies with explicit regulatory objectives, by showing that digital governance reforms can also serve as effective instruments for CRPC.

Second, the results suggest that the environmental effects of smart city construction should be understood as a systemic governance outcome rather than a purely technological outcome. The mechanism analysis provides evidence that the co-reduction effect is not driven by digital infrastructure alone. Instead, it may operate through multiple technological, organizational, and environmental channels, although the strength of the evidence varies across individual mechanisms. This interpretation is consistent with the TOE framework and helps explain why smart city construction produces broader environmental consequences than those captured by single-indicator studies.

Third, the heterogeneity results indicate that the effectiveness of smart city construction depends on local governance capacity and structural conditions. The stronger effects observed in cities with more stringent environmental regulation suggest that digital governance complements rather than substitutes for formal regulation. Likewise, the stronger effects in growing and non-resource-based cities imply that the environmental returns to smart city construction are shaped by development stage and industrial structure. In this sense, smart city policy is not a uniform solution, and its effectiveness depends on whether cities possess the institutional and economic conditions required to translate digital capacity into environmental improvement.

More broadly, the findings have implications for how urban sustainability is evaluated. Existing studies often assess smart city policy through separate outcomes such as innovation, carbon emissions, or pollution abatement. The evidence presented here suggests that a more meaningful assessment should focus on whether digital transformation improves the joint governance of multiple environmental pressures. For rapidly urbanizing economies, this broader perspective is especially important because climate mitigation and local environmental quality are closely interconnected in practice.

Spatial autocorrelation

Prior to conducting spatial econometric analysis, it is necessary to test the dependent variables for spatial dependence. This study employs the Global Moran’s I index to test for spatial autocorrelation in SO₂ emission intensity and carbon emission intensity, respectively. The model is constructed as follows:

where n represents the total number of spatial units; x_i and x_j denote the observed values of spatial units i and j, respectively; $$ \bar{x} $$ represents the mean of all observed values. The value of Moran’s I typically ranges between -1 and 1. A value greater than 0 indicates positive spatial autocorrelation, meaning observed values tend to cluster in space; a value less than 0 indicates negative spatial autocorrelation, suggesting a dispersed pattern; and a value of 0 implies a random spatial distribution of observed values.

The calculation results are shown in Figure 6. The results show that the Global Moran’s I values for SO₂ and CO₂ emission intensities are consistently positive, with all corresponding P-values below 0.05. This indicates that cities with similar emission characteristics tend to cluster geographically, and spatial dependence should therefore be taken into account in the empirical analysis.

Figure 6. Global Moran’s I.

The local Moran scatterplots [Figures 7 and 8] further reveal clear spatial clustering patterns. Over time, the concentration of high-high clusters declines while the concentration of low-low clusters rises, suggesting that urban environmental performance is not randomly distributed across space and that environmental improvement may diffuse across cities.

Figure 7. Local Moran scatter plot of SO₂ emission intensity in certain years.

Figure 8. Local Moran scatter plot of CO₂ emission intensity in certain years.

Spatial model specification and estimation

This study employs Lagrange multiplier (LM) tests and likelihood ratio (LR) tests to select the appropriate spatial econometric model. Tables 15 and 16 present the LM test and LR test results, respectively. Both tests significantly reject the null hypothesis, indicating that the Spatial Durbin Model (SDM) should be adopted for analysis. Consequently, this study applies the Spatial Difference-in-Differences method combined with the SDM to examine whether smart city construction exerts spatial effects on the synergistic governance of pollution and carbon reduction at the urban level.

Compared with the traditional DID model, the Spatial Difference-in-Differences (SDID) model can examine the spatial spillover effects of smart city construction. Therefore, building upon the baseline regression model, this paper constructs the following SDM:

where W_ij represents the spatial weight matrix; $$ \delta_{1} {\textstyle \sum_{j=1}^{n}} W_{i j} S O_{2, j t} $$ and $$ \theta_{1} {\textstyle \sum_{j=1}^{n}} W_{i j} C O_{2, j t} $$ are spatial lag terms, primarily capturing spatial spillover effects - meaning that pollutant emissions in one region can be influenced by the emission levels of its neighboring areas. $$ \delta_{2} {\textstyle \sum_{j=1}^{n}} W_{i j} D I D_{j t} $$ and $$ \theta_{2} {\textstyle \sum_{j=1}^{n}} W_{i j} D I D_{j t} $$ denote the spatial interaction term, reflecting the impact of the weighted average of smart city construction in neighboring regions on the local area. This term allows for the analysis of how policy implementation in surrounding regions affects local pollutant emissions through spatial linkages. $$ \delta_{3} {\textstyle \sum_{j=1}^{n}} W_{i j} { Control }_{j t} $$ and $$ \theta_{3} {\textstyle \sum_{j=1}^{n}} W_{i j} { Control }_{j t} $$ represent the spatial lag term of the control variables, while all other variables retain the same definitions as in the baseline regression.

To capture different channels of intercity environmental spillovers, this study constructs a geographical adjacency matrix and an economic distance matrix. The geographical adjacency matrix assigns a value of 1 when two cities share a common administrative boundary and 0 otherwise. It is intended to capture cross-boundary pollution diffusion, regional environmental coordination, and policy imitation among neighboring cities. The economic distance matrix assigns greater weights to cities with more similar levels of economic development and is intended to capture policy learning, technology diffusion, and industrial linkages among economically connected cities. The diagonal elements of both matrices are set to zero, and both matrices are row-standardized before estimation.

A geographical adjacency matrix is selected instead of a geographical inverse-distance matrix because the spatial mechanisms emphasized in this study are primarily associated with direct cross-boundary pollution transmission and administrative coordination among contiguous cities. An inverse-distance matrix assumes a continuous distance-decay relationship and assigns positive connections to almost all city pairs, including geographically distant cities, which may obscure the boundary-based diffusion and governance mechanisms considered here. The economic distance matrix is included as a complementary specification to capture non-geographical linkages among cities.

where i and j are spatial unit identifiers, i,j ∈ [1,n], with n representing the total number of spatial units; W₁ takes the value of 1 if the two cities are spatially adjacent, and 0 otherwise; W₂ measures the magnitude of economic disparity between two regions, g_i denotes the mean value of real GDP for region i over the observation period.

Table 17 presents the empirical results of the SDM utilizing both geographical and economic distance matrices. The estimation results show that the coefficients of the smart city policy variable remain significantly negative under both the geographic and economic matrices. This indicates that smart city construction significantly reduces local sulfur dioxide emission intensity and carbon emission intensity even after spatial dependence is taken into account^[15].

More importantly, the coefficients of the spatial interaction term are also significantly negative. Specifically, geographical proximity generates stronger synergistic effects owing to the natural dispersion patterns of pollutants and the necessity of administrative coordination in neighboring regions^[75]. This suggests that smart city construction in one city can improve the co-reduction performance of neighboring or economically connected cities through spillover effects.

The spatial spillover effect under the geographical weight matrix is stronger than that under the economic weight matrix, suggesting that geographical proximity plays a more important role than economic linkages in transmitting the environmental benefits of smart city construction. This finding is consistent with the characteristics of pollution diffusion and regional environmental governance. Air pollutants such as SO₂ can spread across adjacent administrative boundaries through atmospheric transport, while neighboring cities are also more likely to engage in joint environmental governance, policy coordination, and information sharing^[76]. In contrast, although economically connected cities facilitate technology diffusion and industrial collaboration, these channels generally influence environmental performance more indirectly and require a longer period to generate observable co-reduction effects.

Spatial effect decomposition and discussion

To further clarify the magnitude of spatial effects, this study decomposes the estimated impacts into direct effects, indirect effects, and total effects. The decomposition results in Table 18 show that smart city construction has significantly negative direct, indirect, and total effects on both sulfur dioxide emission intensity and carbon emission intensity under the spatial adjacency matrix.

The direct effects indicate that smart city construction improves the local co-reduction performance of pilot cities. The indirect effects show that the policy also generates positive spillovers for surrounding cities. This means that the environmental benefits of smart city construction are not confined to the local jurisdiction in which the policy is implemented.

Several channels may explain these spillover effects. Geographic proximity may facilitate policy imitation, administrative coordination, and the diffusion of environmental governance practices. Economic linkages may promote spillovers through technology transfer, industrial coordination, factor mobility, and the convergence of environmental standards.

Overall, the spatial results suggest that smart city construction generates environmental benefits at both the local and regional levels. These findings highlight the importance of viewing digital urban governance not only as a city-specific reform, but also as a potential driver of CRPC.

Conclusions

This study examines whether smart city construction, originally introduced as a strategy for digital urban transformation, also contributes to CRPC. Using panel data for 259 prefecture-level cities in China from 2006 to 2022 and exploiting the staggered rollout of the smart city pilot policy, the empirical analysis shows that smart city construction significantly reduces both sulfur dioxide emission intensity and carbon emission intensity. These results indicate that smart city policy generates a clear CRPC effect. This finding extends the understanding of smart city construction beyond technological modernization alone. The environmental value of smart city policy is not confined to the deployment of digital infrastructure or the improvement of administrative efficiency. More fundamentally, it reflects the capacity of digital governance to reorganize information flows, strengthen regulatory responsiveness, and create conditions conducive to coordinated environmental improvement.

The mechanism analysis further shows that the co-reduction effect of smart city construction is generated through multiple and interconnected pathways. In the technology dimension, smart city construction promotes green innovation and digital infrastructure development. In the organization dimension, it facilitates industrial upgrading and improves human capital. In the environment dimension, it strengthens ecological investment and pollution control capacity. Taken together, these findings suggest that the environmental effects of smart city construction should be understood as the outcome of systemic urban transformation rather than isolated technological change.

The results also reveal that the effects of smart city construction are conditioned by differences in local governance and structural characteristics. The co-reduction effect is more pronounced in cities with stronger environmental regulation, in growing cities, and in non-resource-based cities, indicating that the environmental returns to digital transformation depend on supportive institutional and economic conditions. In addition, significant spatial spillover effects suggest that the environmental benefits of smart city construction are not confined to pilot cities themselves, but may also extend to geographically proximate and economically connected cities.

Overall, the findings support a broader conclusion: digital urban governance can become an important pathway toward integrated environmental governance. When combined with industrial transformation, ecological investment, and effective regulatory implementation, smart city construction can contribute not only to more efficient urban management, but also to CRPC.

Policy implications

Several policy implications follow from these findings. First, governments should continue promoting smart city construction by strengthening digital infrastructure, improving data-sharing mechanisms, and enhancing intelligent environmental governance platforms, thereby facilitating the coordinated reduction of air pollutants and carbon emissions.

Second, since the empirical results confirm the importance of technological, organizational, and environmental channels, policymakers should promote coordinated policies across these three dimensions. Specifically, greater emphasis should be placed on green technological innovation, industrial upgrading, human capital cultivation, ecological construction, and pollution control to maximize the environmental benefits of smart city initiatives.

Third, smart city construction should be better integrated with environmental regulation. In regions with stronger regulatory capacity, digital technologies can further improve environmental monitoring and enforcement efficiency. In regions with relatively weak environmental regulation, governments should first strengthen institutional capacity to ensure the effective implementation of digital governance.

Fourth, resource-based cities should formulate differentiated smart city strategies according to their development stage. Growing cities should continue integrating digital technologies with green industrial transformation. Mature and regenerative cities should accelerate energy structure adjustment and low-carbon technological innovation. Declining resource-based cities require stronger fiscal support, industrial revitalization, and innovation investment to overcome structural constraints before smart city construction can generate significant environmental benefits.

Finally, given the significant geographical spillover effects, neighboring cities should strengthen cross-regional cooperation in environmental monitoring, information sharing, and joint pollution prevention and control. Regional coordination mechanisms can amplify the environmental benefits of smart city construction beyond administrative boundaries.

Limitations and future research

This study has several limitations that should be acknowledged. First, the analysis focuses on prefecture-level cities in China, and the external validity of the findings should be assessed with caution in other institutional or urban contexts.

Second, although the study identifies several plausible channels, the mechanism analysis is based on observable city-level indicators and therefore cannot fully capture all dimensions of institutional change associated with smart city construction. Future research may combine micro-level firm or household data with policy evaluation designs to provide a more detailed account of how digital governance affects environmental behavior and performance.

Third, while the study documents spatial spillover effects, the precise pathways of intercity transmission require further examination. Future work could distinguish more clearly among technology diffusion, industrial relocation, infrastructure connectivity, and regional policy coordination as alternative channels of spillover.

Finally, although the TOE framework provides a comprehensive perspective for examining the potential mechanisms through which smart city construction promotes CRPC, the mediation analysis is based on observational data and therefore cannot fully establish the underlying causal transmission processes. Future studies may employ dynamic mediation models or quasi-experimental approaches to further verify these mechanisms.