How do China’s low-carbon pilot policies reshape urban ecological performance across time and space?

1 Introduction

China’s rapid urbanization has generated substantial resource consumption and environmental degradation, challenging sustainable development. To fulfill its “dual carbon” commitment, the Chinese government launched three batches of low-carbon pilot projects across 81 cities in 2010. While existing evaluations demonstrate that these pilots effectively reduce carbon emission intensity and promote energy conservation, their broader social impacts, particularly regarding citizens’ wellbeing, remain insufficiently examined. As China increasingly emphasizes a people-centric urbanization paradigm that prioritizes citizens’ welfare as a core development objective, assessing whether low-carbon development enhances quality of life becomes imperative.

Ecological welfare performance (EWP), defined as welfare output per unit of ecological consumption, serves as a critical indicator of sustainable development (Long et al., 2017). Conceptually, EWP captures the symbiotic relationship among ecological resource consumption, economic growth, and welfare levels, making it suitable for evaluating whether low-carbon pilot policies (LCPPs) enhance human wellbeing. In China’s pursuit of peak carbon emissions, transforming economic growth into higher welfare output represents a crucial strategy to overcome resource constraints and enhance urban EWP. Examining whether low-carbon pilot policies (LCPPs) improve urban residents’ quality of life thus becomes essential for facilitating coordinated economic, environmental, and social development.

Scholars generally agree that low-carbon pilot cities have achieved remarkable results in reducing energy consumption and carbon emissions since the National Development and Reform Commission launched the first batch in 2010. Environmental effects include substantial reductions in carbon emission intensity across pilot batches (Dai and Cao, 2015; Liu et al., 2022), improvements in urban energy efficiency (Fan and Liu, 2022), and air quality (Song et al., 2019). These findings align with research on European low-emission zone policies demonstrating local air quality improvements (Gehrsitz, 2017) and pollution reduction from urban traffic congestion (Bernardo et al., 2021). Economically, LCPPs positively impact green economic growth (Wei and Gu, 2021) and facilitate harmonization between economic development and ecological environment (Zhang and Yan, 2023). Additionally, LCPPs promote green technological innovation at both firm and city levels (Xu and Cui, 2020; Deng et al., 2022), reduce economic disparities between peripheral and central cities (Wang and Wang, 2024), mitigate regional energy poverty (Song et al., 2024), and enhance urban sustainable development performance (Liu et al., 2025).

The concept of ecological welfare performance (EWP), rooted in the steady-state economic theory proposed by Daly (1974), reflects welfare output per unit of ecological consumption and serves as a key indicator of sustainable development. Empirical evidence indicates that China’s EWP exhibits significant regional heterogeneity, following a spatial provincial gradient of strong east, moderate center, and weak west (Deng et al., 2021), with eastern regions witnessing declining trajectories while central and western regions show fluctuating yet upward trends (Guo et al., 2022). Scholars’ investigations at both city and urban agglomeration levels further confirm that coordination within and between urban agglomerations represents an effective pathway toward holistic EWP enhancement (Wang et al., 2023). Environmental policy intensity (Wang and Luo, 2021), regulatory instruments (Ren et al., 2016), and fiscal decentralization (Li et al., 2022) significantly influence EWP with pronounced regional variation.

As the central government’s assessment targets shift towards achieving a balance between economic growth and environmental protection, local governments’ competitive objectives are also evolving, with a shift towards coordinated economic and environmental development. Drawing upon the perspective of governments’ multiple governance tasks, some scholars have examined the synergistic effects between LCPPs and other initiatives such as innovative city pilots and smart city pilots, verifying that the superposition of multiple governance objectives amplifies the efficacy of low-carbon pilot policies, thereby generating significant synergistic carbon emission reduction effects (Wang et al., 2022; Hao et al., 2023; Gong et al., 2025).

While extant literature has established the efficacy of LCPPs in curbing pollution, reducing emissions, and fostering economic growth, three critical questions await empirical scrutiny. Do these policies enhance urban EWP? What mechanisms drive such effects amid China’s new-type urbanization? And do they induce spatial spillovers? Although green technological innovation is recognized as an EWP determinant, its mediating role within the environment-economy-society nexus awaits systematic examination. This study addresses these gaps through three contributions. First, moving beyond prior research on economic and environmental effects, this study evaluates the social consequences of LCPPs by assessing their impacts on citizens’ quality of life and welfare, thereby extending the analytical lens to the social dimension of urban low-carbon development. Second, this study identifies urbanization as a catalyst unlocking policy potential: in highly urbanized settings, optimized resource allocation and environmental quality improvements create a virtuous cycle amplifying policy impact on EWP. Third, spillover analysis reveals that deepening low-carbon pilot construction generates intensifying positive spillovers on neighboring cities’ ecological welfare performance, as expanding pilot coverage and increasingly balanced spatial distribution amplify policy diffusion effects. This validates the pivotal role of pilot cities in advancing nationwide population health and wellbeing through green urban transition.

The remainder of this paper is organized as follows. Section 2 outlines the theoretical framework. Section 3 describes the data and methodology. Section 4 presents the empirical results, heterogeneity, and robustness analyses. Section 5 examines mechanisms and moderating effects. Section 6 analyzes spatial spillovers. Section 7 concludes with policy implications.

2 Theoretical analysis

Taking the low-carbon city pilot program as an opportunity, local governments aim to explore green and low-carbon transformation models at the city level and achieve synergy among environmental protection, economic growth, and social development goals. Enhancing urban EWP is essential for maximizing welfare output with minimal natural resource consumption. To this end, local governments implement LCPPs to guide enterprises across production and distribution sectors in achieving energy savings and emission reductions, thereby promoting green transformation at the production end. Research shows that LCPPs effectively reduce carbon emissions (Zhang, 2020) while significantly diminishing urban air pollution. The benefits of these policies far exceed the economic losses resulting from emission restrictions, thereby facilitating win-win outcomes in both environmental protection and high-quality economic development (Song et al., 2019). By adapting energy structures and improving energy and material use efficiency, enterprises transform resource inputs into economic outputs, facilitating the development of services such as healthcare and education, thereby improving quality of life and social welfare. Additionally, reducing pollutants and restoring ecological systems yields environmental enhancements that bolster public health standards, promote human capital accumulation, and ultimately elevate overall social welfare. Based on the foregoing analysis, Hypothesis 1 is proposed.

Hypothesis 1 Low-carbon city pilot policies can improve EWP.

A growing body of literature has examined the drivers of urban EWP, with particular attention to the role of environmental regulation. Empirical evidence suggests that stringent environmental policies can significantly shape EWP outcomes (Liu and Zhuang, 2022), though the relationship appears contingent on regulatory intensity, with moderate levels functioning as a necessary condition for substantive improvements (Guo et al., 2020). Su et al. (2022) further demonstrate that LCPPs significantly curb carbon emissions by accelerating green technological innovation. Extant literature suggests that green technological innovation enhances urban EWP by fulfilling three complementary functions. Environmentally, such innovation optimizes the energy mix, reduces reliance on high-polluting conventional sources, improves resource efficiency, and mitigates pollution exacerbated by energy rebound effects (Ouyang et al., 2022). Prior research has demonstrated that green technology transfer significantly restrains energy consumption in developing regions by influencing industrial structure and energy prices (Zhang et al., 2025). From a production and consumption perspective, it drives the green transformation of production sectors, expands the supply of green products, and elevates living standards in pilot cities, thereby better satisfying residents’ intrinsic demand for healthy and sustainable lifestyles (Cao and Gao, 2021). Regarding human capital, it enhances urban livability, generates wage premiums, and fosters innovation-driven development, which collectively facilitate the agglomeration of skilled technological talent (Jiang and Zhong, 2026). These agglomeration effects, in turn, generate technological and knowledge spillovers that ultimately contribute to EWP enhancement. The study by Gao et al. (2026) observed that the implementation of the LCPPs strengthened inter-city synergy in the field of green innovation; however, the policy effects exhibited structural asymmetry across different cities, with core cities benefiting disproportionately more. Accordingly, Hypothesis 2 is proposed.

Hypothesis 2 Low-carbon city pilot policies improve urban EWP through the promotion of green technology innovation.

The low-carbon pilot initiative seeks to catalyze economy-wide green transitions across production and consumption systems. Against the backdrop of rapidly evolving digital technologies and the expanding digital economy, digital transformation has increasingly become a pivotal enabler of cost efficiencies, operational optimization, and structural upgrading for enterprises, industries, and urban economies alike. The low-carbon pilot policy is found to significantly facilitate corporate digital transformation, thereby enhancing both environmental performance and market value (Tong et al., 2025). The digital economy is theorized to catalyze industrial structural upgrading through industrial digitalization, operationalized via the integration of the digital and real economies, resource allocation optimization, and transaction cost reduction (Chen and Yang, 2021). On the other hand, the digital economy is theorized to facilitate the precision-matching of consumer preferences, thereby elevating consumption quality, enhancing user experience, and ultimately augmenting household welfare. Empirical studies have consistently demonstrated that digital economy development exerts a mitigating effect on both carbon emissions (Bai, 2023), and the discharge of diverse urban pollutants (Deng and Zhang, 2022), thereby ameliorating environmental quality. Empirical research has further documented the welfare-enhancing effects of digital economy development, including public service equalization, quality-of-life improvements (Wang et al., 2023), and income inequality reduction (Luo and Wei, 2023).

Hypothesis 3 Low-carbon city pilot policies improve urban EWP through the promotion of digital economy development.

The low-carbon pilot construction process coincides with China’s accelerated advancement of new-type urbanization. By 2023, pilot cities had attained an average urbanization rate of 69.9%, signifying their collective entry into the middle-to-late urbanization stage; nevertheless, pronounced heterogeneity persists, with rates ranging from 37.5% to 99.8%. China’s National New-Type Urbanization Plan (2014–2020) mandates that low-carbon pilot policies be calibrated to urbanization development stages. Rapid urbanization elevates energy consumption and carbon emissions; yet spatial agglomeration of population, industry, and economic activity concurrently fosters improved social security and infrastructure, thereby enhancing resource utilization efficiency. Conversely, excessive concentration may engender traffic congestion and ecological degradation, potentially impeding ecological welfare improvement. Although scholars have affirmed the salutary effects of new-type urbanization in curbing urban carbon emissions (Li and Tang, 2024) and enhancing carbon emission efficiency (Zheng et al., 2026), extant research remains inconclusive regarding its impact on ecological welfare performance. Zhu and Xia (2019) identify urbanization as a threshold variable in the industrial agglomeration-environmental pollution nexus; in mature urbanization stages, economies of scale and technology spillovers attenuate or even reverse the pollution effects of industrial agglomeration. Chen and Liu (2023) document an inverted U-shaped relationship between urbanization and EWP at the provincial level, albeit with regional heterogeneity. Li and Hu (2023) identify a U-shaped effect in Jiangsu Province cities, whereas Li et al. (2025) conclude that urbanization exerts adverse effects on ecological welfare performance in both local and surrounding regions in Gansu Province. Based on this analysis, this study posits that urbanization levels moderate the effect of low-carbon pilot construction on urban EWP. Accordingly, Hypothesis 4 is proposed.

Hypothesis 4 Urbanization levels moderate the impact of low-carbon pilot policies on urban EWP.

Research indicates significant regional variations in EWP across China. Geographically, EWP is highest in the eastern regions, followed by the central regions, and lowest in the western regions (Long, 2019), while also exhibiting a “high-south, low-north” spatial pattern (Wang and Duan, 2023). Furthermore, inter-zonal EWP disparities exceed intra-zonal differences (Wang et al., 2022). Empirical evidence from Fang and Xiao (2019) demonstrates strong spatial correlation and relatively stable path dependence in provincial EWP across China. The study by Tian and Liu (2021) indicates that the LCPP in 2022 significantly stimulated green innovation among enterprises in peer cities, with spillover effects substantially exceeding those on pilot cities themselves. Chen (2025) documents considerable stability and positive spatial autocorrelation in urban EWP, whereby higher EWP levels facilitate upward transitions in adjacent cities, whereas lower levels impede such transitions. Consequently, low-carbon pilot policies (LCPPs) exert not only direct effects on pilot cities’ EWP but also spatial spillover effects on neighboring cities. Based on these considerations, Hypothesis 5 is proposed.

Hypothesis 5 The low-carbon pilot policy exerts spatial spillover effects on urban EWP.

Figure 1 illustrates the mechanism through which low-carbon city pilot policies affect EWP.

3 Methodology and data

3.1 Model

As previously noted, China’s National Development and Reform Commission (NDRC) has released three batches of low-carbon pilot cities since 2010. In the process of implementing the LCPPs, the non-random selection of pilot cities, coupled with regional development disparities and unobservable heterogeneity, may confound the policy’s causal effect on EWP. To address these endogeneity concerns, this study employs a multi-period difference-in-differences (DID) model, specified as shown in Equation 1.

E W P i t = α 1 + β 1 P L C P i t + φ X i t + μ i + γ t + ε i t (1)

Where E W P i t denotes the ecological welfare performance (EWP)of city i at time t. P L C P i t is the main explanatory variable of the low-carbon city pilot policy and is a cross-multiplier of pilot city and pilot time. It equals one if city i implements the low-carbon city pilot policy at time t, and 0 otherwise. X i t refers to a set of control variables; μ i refers to the city fixed effect, while γ t denotes the time fixed effects; ε i t represents the random disturbance term. β 1 reflects the impact of the pilot low-carbon city policy on EWP, with a positive value indicating that low-carbon city pilot policy significantly improves EWP.

Drawing on the theoretical framework outlined in Section 2, this study posits that LCPPs can enhance EWP through two primary channels: green technological innovation and digital economy development. Specifically, LCPPs can stimulate green patenting activities and accelerate digital transformation, thereby improving resource efficiency and welfare outcomes. To empirically test these mechanisms, the following mediation models are specified:

Channel 1: Green Technological Innovation

ln ⁡ g p a t i t = α 2 + β 2 P L C P i t + φ X i t + μ i + γ t + ε i t (2)

E W P i t = α 3 + β 3 P L C P i t + β 4 ⁡ ln ⁡ g p a t i t + δ X i t + μ i + γ t + ε i t (3)

In Equations 2, 3, ln ⁡ g p a t i t represents the mediating variable of green technology innovation level, and the logarithm of the number of green invention patents and the number of green utility models at each city-year level is used as a proxy of the green technology innovation level, and the data are obtained from the Green Patent Research Database (GPRD) of the China Research Data Service Platform. Since some cities report zero values in green patent counts during the study period, this paper follows Guo et al. (2022) by taking the natural logarithm of the sum of these two indicators plus one. If the coefficients β 1, β 2, and β 4 in Equations 1–3 are significant and the coefficients of β 3 become smaller or the level of significance decreases, this indicates that low-carbon pilot policies influence urban EWP via the mediating mechanism of green technological innovation.

Channel 2: Digital Economy Development

ln ⁡ g d i g i i t = α 2 + β 2 P L C P i t + φ X i t + μ i + γ t + ε i t (4)

E W P i t = α 3 + β 3 P L C P i t + β 4 ⁡ ln ⁡ g d i g i i t + δ X i t + μ i + γ t + ε i t (5)

In Equations 4, 5, where ln ⁡ g d i g i i t denotes the mediating variable for digitalization level of city i at time t. If the coefficients β 1, β 2, and β 4 in Equations 1, 4, 5 are significant and the coefficients of β 3 become smaller or the level of significance decreases, this indicates that LCPPs influence urban EWP via the mediating mechanism of digital economy development.

The theoretical framework further suggests that the impact of LCPPs on EWP may be contingent on urbanization levels. Highly urbanized cities possess superior infrastructure, human capital, and governance capacity, which may amplify policy effectiveness through resource optimization and service upgrading. To test this hypothesis, an interaction term between the policy dummy (DID) and urbanization level (URB) is introduced into the baseline model, as specified in Equation 6.

E W P i t = θ 0 + θ 1 P L C P i t + θ 2 P L C P i t × U R B + θ 3 U R B + δ X i t + μ i + γ t + ε i t (6)

The coefficient θ 2 is the parameter of primary interest: a positive and statistically significant estimate indicates that urbanization positively moderates the policy effect.

Finally, given the spatial interdependence of urban development, LCPPs may generate spillover effects on neighboring cities through knowledge diffusion, factor mobility, and demonstration effects. To capture these spatial dynamics, a spatial Durbin model (SDM) with individual fixed effects is employed, as specified in Equation 7.

E W P i t = C + ρ W E W P i t + φ 0 P L C P i t + σ 0 W P L C P i t + φ 1 X i t + σ 1 W X i t + μ i + r t + ε i t (7)

In Equation 7, where C denotes the constant term, W represents the spatial weight matrix, ρ is the spatial autoregressive coefficient. The spatial weight matrix employed in this study is the spatial economic matrix, and W E W P i t, W P L C P i t, and W X i t denote the spatially lagged terms of the dependent variable, core explanatory variable, and control variables, respectively. φ i and σ i are parameters to be estimated, while μ i , γ t , a n d ε i t represent city fixed effects, time fixed effects, and the stochastic error term, respectively.

3.2 Variable selection and description

3.2.1 Dependent variable

The dependent variable is ecological welfare performance (EWP). Beyond reducing carbon emissions, the goal of LCPPs is to explore the path of low-carbon transformation and development, which seeks to achieve higher quality of life with reduced resource input while promoting sustainable urban development. The EWP of pilot cities is an important indicator of the effectiveness of low-carbon city construction.

Two primary methods exist for measuring EWP: the single index ratio method and the composite index method. The former method employs the ratio of wellbeing level to ecological consumption to measure EWP (Jorgenson and Dietz, 2015), while the latter includes the Stochastic Frontier Production Model (Dietz et al., 2009), the Data Envelopment Analysis (DEA) method (Long et al., 2017), and a comprehensive factor evaluation index system based on an input-output perspective (Fang and Xiao, 2019).

In this paper, three resource consumption indicators—namely, land, water, and energy—are used as proxies for resource inputs. Given that the ecological footprint is generally assessed at the national level, substantial statistical inaccuracies often arise due to methodological inconsistencies and divergent classifications across provinces and municipalities. The assessment of social welfare, an output indicator of EWP, encompasses objective welfare indicators (e.g., life expectancy at birth and infant mortality), comprehensive objective welfare indicators (represented by the Human Development Index, HDI), and subjective welfare indicators. Nonetheless, subjective welfare indicators, such as life satisfaction, are influenced by individual perceptions and may not accurately reflect reality. Consequently, they are not suitable as a standalone measure of social welfare. While the Human Development Index (HDI) captures crucial social welfare dimensions such as longevity and health, it disregards the contributions of convenient transportation, ample employment opportunities, and high-quality public services such as education and medical care to people’s wellbeing. These elements are pivotal for urban development, as they foster a sense of fulfillment and satisfaction among individuals. Drawing on the extant research, this paper proposes a comprehensive evaluation system of EWP from the perspective of input-output.

As illustrated in Table 1, the EWP evaluation system is predicated on the indicator construction framework proposed by Xiao and Ji (2018) and Long (2019), treating resource consumption as input, welfare level as the desired output, and environmental pollution as the undesired output. The specific components of the system are delineated in Table 1.

TABLE 1. Index system of EWP.

As demonstrated in Table 1, the selection of social welfare indicators in this study is grounded in a comprehensive understanding of China’s unique context, encompassing three dimensions that are critical to measuring wellbeing: the standard of living of the population, the accessibility of public services, and the quality of the environment. Concerning the standard of living, factors such as income level, employment stability, consumption level, and consumption structure of the population are examined. The selection of key indicators encompasses pensions, healthcare, education, and associated metrics such as the coverage rate of basic urban insurance, the number of college students per 10,000 inhabitants, and the number of physicians per 10,000 inhabitants. This selection is informed by the findings of Cheng et al. (2019), who assert that the availability of urban infrastructure plays a substantial role in the green transformation of low-carbon cities. Public green space per capita and mechanized street sweeping and cleaning area per capita are chosen to evaluate environmental quality, following Dong et al. (2023).

To control for population size variations across cities and ensure data comparability, industrial wastewater discharges per capita, industrial sulfur dioxide emissions per capita, and industrial soot emissions per capita are selected as indicators of undesired outputs.

In this study, resource inputs and undesired outputs (e.g., pollution) are regarded as the costs of producing welfare outputs (e.g., input indicators). Conversely, desired outputs (e.g., the standard of living of the population, the level of public services, and the quality of the environment) are regarded as output indicators. The unemployment rate and Engel’s coefficient are considered adverse performance indicators, with lower values indicating higher performance levels. To avoid the subjectivity of manually assigning indicator weights, this paper adopts the entropy weight method to calculate the EWP level of each city.

3.2.2 The core explanatory variable

The main explanatory variable is the low-carbon city pilot policies (LCPPs). China’s low-carbon pilot program was launched in three batches. The first batch, announced in 2010, included both provincial-level pilots and municipal-level pilots, while the second and third batches, announced in 2012 and 2017, respectively, were implemented primarily at the city level. Because we use prefecture-level cities as the unit of analysis, the definition of treatments is based on explicit municipal-level pilot designation.

According to the official pilot lists, the three batches cover 81 pilot jurisdictions, including prefecture-level cities, county-level cities, specially administered or non-standard regions, and Lhasa, the capital of the Tibet Autonomous Region. Among these pilot jurisdictions, 71 are prefecture-level cities with comparable city-level statistical data. The remaining jurisdictions are excluded from the empirical sample because they are either below the prefecture level, not comparable with ordinary prefecture-level cities in terms of administrative and statistical systems, or subject to substantial data limitations. Specifically, we exclude seven county-level cities, namely, Jiyuan, Qiongzhong, Xunke, Changyang, Changji, Yining, and Hetian, because comparable city-level socioeconomic and environmental indicators are unavailable. We also exclude Daxing’anling Prefecture and the Xinjiang Production and Construction Corps because their administrative and statistical systems are not fully comparable with ordinary prefecture-level cities. Lhasa is excluded because several key indicators, including built-up land area per capita, water resources, industrial wastewater discharge, industrial SO₂ emissions, industrial soot emissions, and total electricity consumption, are missing for a substantial part of the sample period. After these exclusions, the treatment group consists of 71 prefecture-level low-carbon pilot cities with available data from 2006 to 2023.

Because the first batch of low-carbon pilots included both provincial-level and municipal-level pilots, the treatment definition requires particular care in a city-level DID design. Following city-level evaluations of China’s low-carbon pilot city policy, we define treated units based on explicit municipal-level pilot designation rather than automatically coding all cities located within first-batch pilot provinces as treated units (Ren et al., 2024). Cities in first-batch pilot provinces that were never designated as city-level pilots remain in the control group, while those later selected as city-level pilots are assigned the provincial entry year as their treatment year. Given that most of these cities are provincial capitals or key municipalities already affected by the provincial pilot policy, this classification better captures the actual policy rollout. For cities appearing in more than one batch, the treatment year is defined as the first year in which the city was explicitly designated as a municipal-level low-carbon pilot. Because the second batch of pilot cities was announced on 26 November 2012, near the end of the calendar year, we code 2013 as the implementation year for the second-batch pilot cities, following Song et al. (2019).

3.2.3 Intermediary variables

The theoretical analysis highlights technological innovation as a primary means for LCPPs to achieve carbon emission reduction (Zhang, 2020) and improve eco-efficiency in low-carbon pilot cities (Song et al., 2020). Green technological innovation is measured using green patents. Green patents refer to patents related to environmentally friendly technologies, including energy conservation and emission reduction, pollution control, clean energy, carbon mitigation, resource recycling, and ecological improvement (Xu et al., 2024). They are widely used as observable indicators of eco-innovation because they capture technological efforts aimed at reducing environmental impacts and improving resource-use efficiency (Favot et al., 2023). This definition is also consistent with the logic of the WIPO IPC Green Inventory, which identifies patent information related to environmentally sound technologies. In the Chinese context, green patents can be divided into green invention patents and green utility model patents. Green invention patents capture more substantive technological breakthroughs, whereas green utility models mainly reflect incremental improvements in product structure, equipment design, or practical applications. Therefore, we measure the level of green innovation using the logarithm of the number of green invention patents (GIP) and green utility model patents (GUP) granted in each city-year.

Drawing on Zhang et al. (2019), who posit that digital economic development, propelled by the internet revolution and manifested through digital finance, plays a pivotal role in narrowing regional and urban-rural disparities, this study employs the digitalization level sub-index from the Digital Inclusive Finance Index published by Peking University, a key metric gauging the digital and mobile penetration of urban financial services, to proxy for urban digitalization level (DL), thereby testing whether digitalization serves as a salient mediator in the mechanism through which low-carbon pilot policies enhance EWP.

3.2.4 Moderating variable

Following Gong et al. (2025), this study adopts urbanization level (URB) as the moderator. The urbanization rate of permanent residents, sourced from the CEIC database, serves as the operational measure.

3.2.5 Control variables

This paper selects the following control variables based on the studies of Zhu and Liu (2017), Li et al. (2019), Guo et al. (2022), and Li and Hu (2023), consider the interference caused by the problem of other omitted variables that may affect the selection of low-carbon pilot cities and EWP factors.

• Average GDP per capita of the city (PGDP) is measured in 10,000 yuan per person, indicating the level of economic development.
• Industrial structure (STR) is measured by expressing the ratio of value added in tertiary industry to that in secondary industry. A pivotal function of environmental governance is to compel industrial restructuring and enhance resource utilization efficiency. The implementation of LCPPs has significantly promoted the upgrading of industrial structures (Lu et al., 2020).
• Energy intensity (EI) is a measure of the efficiency with which energy consumption is converted into economic output. It is calculated by dividing electricity consumption by GDP.
• The size of government (SG) is expressed as the share of local fiscal expenditures in GDP, which represents the financial strength of the government. A stronger financial strength allows for easier implementation of pilot low-carbon policies and measures.
• The strength of support for science and education (SSE) is measured as the ratio of science and education expenditure to total local fiscal expenditure, following Dong et al. (2022). It represents the government’s emphasis on social efforts and people’s welfare, which significantly affects the transformation of economic performance into social welfare. Table 2 describes the main variables.

TABLE 2. Specification of main variables.

3.3 Data sources

In 2007, the Chinese government introduces the “Measures for the Assessment of Total Emission Reduction of Major Pollutants,” which incorporates environmental indicators into the performance evaluation system for local government officials and implements the environmental governance accountability system and the one-vote veto. This is an important signal from the central government to weaken the GDP evaluation and strengthen the weight of environmental protection evaluation. Considering the influence of this factor, this paper selects the panel data of 273 prefecture-level cities in China from 2006 to 2023. All the data are taken from the China Urban Statistical Yearbook, China Urban Construction Statistical Yearbook, and provincial statistical yearbooks, and the specific data sources are listed in Table 1.

Some of the missing data are supplemented by consulting the corresponding statistical yearbook of each city and the national economic and social development statistical bulletin of the corresponding year. Missing data for very few years are filled by interpolation. Due to substantial missing data in some cities, a total of twelve cities are excluded from the study sample: Shuozhou and Xinzhou in Shanxi Province; Heihe and Suihua in Heilongjiang Province; Yangjiang in Guangdong Province; Guigang in Guangxi Zhuang Autonomous Region; Bazhong and Ziyang in Sichuan Province; Lhasa in the Tibet Autonomous Region; Jinchang and Baiyin in Gansu Province; and Zhongwei in the Ningxia Hui Autonomous Region, a total of twelve cities, are excluded from the study sample. Table 3 presents the descriptive statistics of the variables.

TABLE 3. Descriptive statistics of the variables.

As can be seen from Table 3, the maximum value of the ecological welfare performance (EWP) of the 273 cities during the 18 years is 0.694 the minimum value is 0.316, the mean value is 0.501, and the standard deviation is 0.067, which indicates that there are obvious differences between the EWP of different cities. This explains the wide variation in city types, resource endowments, etc. within the scope of this study, suggesting a good representation of the cities covered. To conserve space, the correlation matrix and multicollinearity diagnostics are reported in Supplementary Appendix A. The results suggest that multicollinearity is unlikely to affect the baseline estimates.

4 Estimation and results analysis of benchmark model

4.1 Measurement of EWP

The entropy method is employed to assess the EWP of 273 Chinese cities from 2006 to 2023. The finding reveals a consistent upward trend in the overall EWP of China’s 273 cities over the 18 years, as depicted in Figure 2.

In addition, as demonstrated in Table 4 the mean, minimum, and maximum values of each city’s EWP in 2023 surpass those of 2006. However, it is noteworthy that the standard deviation for 2023 is smaller than that observed in 2006. This finding indicates an overall upward trend in the EWP of Chinese cities, accompanied by a narrowing gap between cities. A similar upward trend is observed either in pilot cities or non-pilot cities, with the gap in EWP between pilot cities being slightly pronounced than that between non-pilot cities. The highest, lowest, and average EWP values in pilot cities exceed those in non-pilot cities.

TABLE 4. Descriptive statistics of the variables (EWP).

Figure 3 illustrates the changes in the geographical distribution of the EWP of the cities over the 18 years within the scope of this study. In 2006, cities demonstrating high ecological welfare performance are predominantly located in coastal regions and along the Yellow River. In 2023, high performance is patchy and continuous. Cities in the eastern coastal provinces demonstrate a high level of performance and a balanced distribution, with a significant EWP improvement in cities of the central region. In contrast, cities in the northwest, northeast, and southwest regions exhibit comparatively modest growth in EWP.

4.2 Results analysis

As previously elaborated, this study examines three cohorts of low-carbon pilot cities launched in China since 2010, drawing on a sample of 71 prefecture-level and higher administrative cities with comparable data from 2006 to 2023, with ecological welfare performance as the dependent variable. In consideration of the temporal lag between the implementation of the LCPPs and its observable impact on EWP, this paper employs the EWP in the lag period as the dependent variable. The low-carbon pilot policy shocks are denoted as multi-period DID and the estimated results are shown in Table 5.

TABLE 5. The estimated results of the benchmark model.

^*,^**,^*** denote 10%, 5%, and 1% significance levels, respectively, with robust standard errors in brackets (the same below).

Table 5 reports the estimation results of Model (1). The results indicate that the low-carbon pilot policy shock (DID) exerts a statistically positive and significant impact on ecological welfare performance (EWP), corroborating the beneficial effect of low-carbon pilot initiatives on urban ecological welfare. Column 1 and Column 2 present the contemporaneous effects of the low-carbon pilot policy shock on ecological welfare performance. Regardless of whether control variables are included, the LCPPs coefficients are 0.086 and 0.011, both significant at the 1% level, indicating that LCPPs have significantly promoted the improvement of ecological welfare performance. Columns (3) and (4) report estimates with one-period and two-period lagged EWP with control variables, and the coefficient are 0.010 and 0.009, also significant at the 1% level. This pattern suggests a persistent yet gradually diminishing policy effect, validating Hypothesis 1.

Regarding control variables, PGDP and STR exert increasingly positive effects over time, suggesting that economic development and industrial upgrading toward tertiary sectors provide growing support for welfare gains. By contrast, EI demonstrates a negative effect, underscoring that energy intensity reduction is the most critical lever for preventing EWP decline. The positive effect of government size on ecological welfare performance attenuates over time as low-carbon pilot measures advance. This suggests that while fiscal capacity is initially essential for local governments to steer resource allocation toward sustainable urban development, market-based mechanisms become increasingly indispensable in the long run. The benchmark model reveals that science and education expenditure exerts a substantial positive effect on ecological welfare performance, with this beneficial impact amplifying markedly over time. This renders it the second most influential driver of ecological welfare improvement, surpassed only by economic development factors.

During the sample period, several urban pilot policies overlapped with the implementation of LCPPs, including the Smart City Pilot Policy (SCPP), Innovative City Pilot Policy (ICPP), and New-Type Urbanization Pilot Policy (NTUPP). These concurrent policies may affect EWP through digital governance, innovation incentives, infrastructure upgrading, industrial restructuring, and public service improvement. If not properly accounted for, their effects may be partly captured by the estimated coefficient of the LCPP indicator.

To address this concern, we augment the baseline specification by including policy dummies for SCPP, ICPP, and NTUPP. We further introduce interaction terms between the LCPP indicator and these concurrent policy dummies to examine whether the estimated effect of LCPPs varies with overlapping policy exposure. The results are reported in Table 6.

TABLE 6. Concurrent policy control and synergy effects.

*, , and * denote statistical significance at the 10%, 5%, and 1% levels, respectively.

Column (1) replicates the original specification, with the LCPP coefficient at 0.010 (p < 0.05). Column (2) adds three concurrent policy dummies. The coefficient of the LCPP indicator remains positive but becomes smaller and less precisely estimated after these policies are included, suggesting that the baseline estimate partly reflects the broader urban policy environment in which LCPP was implemented. Among the concurrent policies, SCPP and NTUPP are positively associated with EWP, while ICPP is statistically insignificant. Column (3) further introduces the interaction between the LCPP indicator and NTUPP. The interaction term is positive and marginally significant, indicating that LCPP is more strongly associated with EWP improvements when combined with new-type urbanization policies. Column (4) provides a four-group comparison by distinguishing LCPP-only cities, NTUPP-only cities, dual-treatment cities exposed to both LCPP and NTUPP, and cities exposed to neither policy. The dual-treatment group exhibits the largest positive coefficient, while the LCPP-only coefficient is positive but statistically insignificant.

These results suggest that LCPPs should not be interpreted in isolation from the broader urban policy environment. Rather, the estimated effect of LCPPs appears to be strengthened when low-carbon governance is implemented together with complementary new-type urbanization policies. This finding is consistent with the idea that LCPPs and NTUPP may jointly improve infrastructure, resource allocation, environmental quality, and welfare-related public services. Additional interaction tests with SCPP and ICPP, as well as batch-specific estimates for LCPP-only cities, are reported in Supplementary Tables E1,E2.

The robustness of this finding is further examined in Supplementary Table E1, which rules out synergy effects with other concurrent pilots, and Supplementary Table E2, which documents batch heterogeneity in standalone LCPP, effects. Collectively, these analyses corroborate that policy synergy between low-carbon and urbanization initiatives constitutes a key mechanism for ecological welfare improvement.

4.3 Robustness check

4.3.1 Heterogeneity treatment bias test

In policy research, if the policy shocks experienced by the sample are not synchronised, the impact of these shocks on cities that implemented the policy earlier or later may vary due to factors such as spillover effects. Additionally, factors such as population density and levels of economic development in different cities can influence their response to the policy. In other words, when individual treatment effects are not homogeneous, i.e., when there is temporal and spatial heterogeneity, traditional two-way fixed DID methods may produce erroneous results, so heterogeneity treatment bias tests are necessary.

According to the Goodman-Bacon decomposition results reported in Table 7, the comparison between never-treated units and timing groups accounts for 92% of the total weight, with a coefficient of 0.003. The Late_v_Early component, though small in weight at 4.65%, exhibits a positive coefficient of 0.002, indicating that heterogeneous treatment effects across cohorts do not introduce negative weighting bias following the indicator system adjustment. Nevertheless, given the theoretical risk of heterogeneity bias inherent in staggered difference-in-differences settings (Goodman-Bacon, 2021), this study employs the CSDID and SDID estimators as complementary robustness checks.

TABLE 7. Goodman-Bacon decomposition results.

Table 8 reports the CSDID estimates. The overall ATT is 0.006, significant at the 1% level. Dynamic effect analysis confirms the parallel trends assumption: pre-treatment coefficients are all insignificant, while the post-treatment average effect is 0.015 (significant at 5%). Cohort-specific estimates reveal marked heterogeneity. The first and second batches generate positive, marginally significant effects (ATT = 0.017 and 0.007 respectively), whereas the third batch is insignificant (ATT = 0.004). This pattern likely reflects two factors. First, early cohorts served as pioneering demonstration zones with greater policy resources and implementation intensity. Second, the third batch’s shorter implementation window (from 2017 to 2023, with only 7 years) may have precluded full policy effect materialization.

TABLE 8. The estimated results of the benchmark model using CSDID.

*, , and * denote statistical significance at the 10%, 5%, and 1% levels, respectively.

As a robustness check, we employ the SDID estimator, which constructs optimal synthetic control groups for each treated unit as reported in Table 9. Despite the relatively large number of treated units (71 cities), the SDID estimate (ATT = 0.006) is highly consistent with the CSDID estimate (ATT = 0.006) reported in Table 8, further corroborating the robustness of our baseline findings. The 95% confidence interval ranges from 0.000 to 0.012, indicating a positive and statistically significant effect of the low-carbon pilot policy on ecological welfare performance.

TABLE 9. The estimated results of the benchmark model using SDID.

Standard errors in parentheses.

p < 0.05.

4.3.2 Parallel trend test

Following Callaway and Sant Anna (2021), this study employs the CSDID estimator to test parallel trends and estimate dynamic treatment effects, with the year of policy announcement as the base period. Results are presented in Figure 4. All pre-treatment coefficients (Tm3 = 0.002, p = 0.278; Tm2 = 0.000, p = 0.823; Tm1 = −0.002, p = 0.465) are statistically insignificant, confirming the validity of the parallel trends assumption. Post-treatment coefficients exhibit a pronounced lagged structure: the policy effect remains insignificant during Tp1–Tp3 but turns significant at Tp4 and Tp5. This delayed emergence of positive effects, becoming evident only from the fourth year onward, suggests that the impact of LCPPs on ecological welfare performance is cumulative rather than instantaneous, consistent with the institutional nature of environmental policies that operate through gradual governance adjustments and infrastructure transformation.

4.3.3 Placebo test

In addition, this paper conducts a placebo test by randomly assigning treatment and control groups to rule out the effects of factors other than low-carbon city building on the EWP of the pilot cities. Since there are 71 low-carbon pilot cities in the sample of this study, 71 cities are randomly selected in each of the 1000 sample regressions. The analysis is replicated 1000 times for each batch of LCPPs shocks. The results are shown in Figure 5.

Figure 5 reports the coefficients of the interaction term between the pilot city dummy and the pilot time dummy P L C P i t are very close to 0, substantially smaller in magnitude than 0.086 in Table 5, regardless of whether it is the second or the third batch. The distribution of the estimated coefficients of the cross-multiplier term is close to the normal distribution with the mean of 0, and the P-values of 1000 estimated coefficients of the two policy shocks are mostly greater than 0.1. Consequently, the results of this paper pass the placebo test that the carbon pilots do not coincidentally promote the EWP of the pilot cities, confirming the robustness of the main regression results.

We conduct two additional robustness checks and report the full results in the Supplementary Appendix to conserve space. First, because the sample period includes the COVID-19 pandemic and the subsequent recovery period, we re-estimate the baseline model using the pre-pandemic sample from 2006 to 2019. The results, reported in Supplementary Appendix C1, show that the DID coefficients remain positive and statistically significant for contemporaneous EWP, one-period-lag EWP, and two-period-lag EWP. Second, to address the concern that pollution-related components may mechanically affect the EWP index, we reconstruct an alternative EWP index excluding industrial wastewater discharge, industrial SO₂ emissions, and industrial soot emissions. The results, reported in Supplementary Appendix B1, are consistent with the baseline estimates. These additional tests suggest that the main findings are not driven by the COVID-19 period or by the inclusion of pollution-related undesirable-output indicators in the EWP index.

4.4 Heterogeneity analysis

4.4.1 Regional heterogeneity

Due to variations in resource endowments and industrial bases, LCPPs may exert divergent impacts on the EWP of cities across diverse geographical locations. For instance, the economically developed and densely populated eastern regions of China have fewer natural resources than those available in the central and western regions. This study employs the group regression method, based on the NDRC’s categorization of Chinese cities into eastern, central, and western regions, to assess the impact of low-carbon city location heterogeneity on EWP. The estimation results are presented in Table 10. As the preceding analysis reveals, the impact of LCPPs on EWP is subject to a time lag; hence, the dependent variable in Table 10 is specified as the two-period lagged value of ecological welfare performance. In this study, the regression results of the groups are tested using SUEST and Fisher’s permutation test, both of which confirm significant differences in the impact of LCPPs between groups of cities in different regions.

TABLE 10. Estimated results of pilot policies for low-carbon cities in different regions.

*, , and * denote statistical significance at the 10%, 5%, and 1% levels, respectively.

Table 10 presents the regional heterogeneity of LCPPs effects. The policy exerts a significantly positive effect on EWP in eastern cities, a suppressive effect in central cities, and an insignificant effect in western cities. These divergent patterns indicate that low-carbon pilot policies exert differentiated impacts on urban EWP across regions. These divergent patterns suggest that low-carbon pilot policies are not universally effective but depend on regional institutional and economic contexts.

Regarding control variables, economic development exerts no significant effect on ecological welfare performance in eastern cities but yields a positive impact in central and western regions, indicating that sustained economic growth during the sample period has contributed to ecological welfare improvements in central and western China. Energy intensity poses a formidable barrier across all regions, with its suppressive effect most acute in central cities, followed by western and eastern counterparts. This gradient reflects disparate energy consumption patterns: eastern cities offset high demand through superior energy efficiency attributable to technological advancement; conversely, central and western cities, endowed with abundant resources, suffer from extensive consumption and lower efficiency due to constrained economic and technological development. Industrial structure significantly enhances EWP only in the eastern region, where a favorable tertiary-sector foundation serves as the second key driver of welfare improvement. This effect is absent in the central and western regions, suggesting that rapid economic growth there has not been accompanied by structural upgrading sufficient to advance EWP; future low-carbon policies should therefore prioritize industrial rationalization in these areas. Government size positively affects EWP in eastern and central cities, underscoring the critical role of fiscal capacity in implementing low-carbon initiatives. In western cities, this effect disappears, likely reflecting a heavier reliance on central government transfers rather than locally generated fiscal revenue. Science and education expenditure yields substantial welfare dividends in central and western cities, yet proves insignificant in the east, where existing investment levels already saturate marginal returns to EWP enhancement.

4.4.2 Heterogeneity of timetable for peaking carbon dioxide emissions

The National Development and Reform Commission (NDRC) of China does not establish a definitive timeline for peaking carbon emissions for each pilot city in the initial batch of low-carbon city pilots. The second batch of low-carbon pilot cities implements the voluntary principle of determining the carbon peaking schedule. In contrast, the third batch of pilot cities are instructed to establish targets of specific timelines for peaking carbon emission during the declaration stage. This paper conducts a group regression analysis based on the year of peaking carbon emission announced by each pilot city. The estimation results are presented in Table 11. In this study, the regression results of the groups are tested using SUEST and Fisher’s permutation test, and both tests confirm that there are significant differences in the coefficients between groups of cities with explicit target years of peaking carbon dioxide emissions.

TABLE 11. Estimated results of policy impacts in low-carbon pilot cities with explicit target year of peaking carbon dioxide emissions.

*, , and * denote statistical significance at the 10%, 5%, and 1% levels, respectively.

As shown in Table 11, the positive impact of LCPPs on the improvement of EWP is evident when the target year of peaking carbon dioxide emissions is set between 2021 and 2025. In pilot cities with carbon peaking targets before 2020, economic development, industrial structure, government size, and science and education expenditure all significantly promote ecological welfare performance, whereas energy intensity exerts the strongest negative effect. The compressed policy to peak timeframe leaves insufficient room for energy intensity adjustment, thereby constraining welfare gains. The LCPP effect is statistically insignificant in cities with earlier peaking targets, consistent with the policy lag documented above, as the compressed peaking timeframe leaves insufficient room for EWP improvements.

Cross group comparisons reveal that the later the peaking target, the more pronounced the positive effect of economic development, underscoring its foundational role across all scenarios. Industrial structure exerts its strongest positive effect in cities with the earliest peaking targets, which typically possess more advanced industrial bases; it becomes insignificant in cities with intermediate targets between 2021 and 2025, but re emerges significantly in cities with the latest peaking targets, where ample time permits structural optimization. Science and education expenditure consistently enhances welfare performance: cities with the earliest peaking targets leverage robust technological innovation to rapidly deploy green technologies, yielding marked emission reductions and welfare gains. In cities with later peaking targets, the benefits of science and education investment gradually materialize, with technological progress boosting total factor productivity and human capital accumulation jointly driving long-term EWP improvements. The later the peaking target, the more severely energy intensity suppresses welfare performance, indicating that cities with the latest peaking targets face considerable drag from elevated energy intensity. The effect of government size proves significant only in cities with the earliest peaking targets, where aggressive fiscal commitment is required to expedite policy implementation; as the peaking horizon extends, policy guidance and inter policy coordination substitute for direct fiscal outlays, rendering the effect of government size insignificant in cities with the latest peaking targets.

5 Test of the mediating role and moderating effect

5.1 Mediation effect analysis of green technology innovation

In accordance with the preceding theoretical analysis and literature review, it can be posited that the LCPPs may indeed promote the EWP of the pilot cities by enhancing their level of green technology innovation. Consequently, this study introduces green technology innovation as a mediating variable into the original model, as depicted in Equations 2, 3, to examine whether green technology innovation serves as a prerequisite for LCPPs to facilitate EWP enhancement. Table 12 reports the estimation results of the mediating effect of green invention patents and green utility models patents.

TABLE 12. Estimation result of intermediary mechanism (GIP&GUP).

*, , and * denote statistical significance at the 10%, 5%, and 1% levels, respectively.

Column (1) reports the estimates of Equation 1 on EWP, under the policy shocks of low-carbon city pilots in Table 12. Columns (2) and (4) contain estimates of Equation 2, while Columns (3) and (5) contain estimates of Equation 3. As reported in Column (2), LCPPs significantly increase green invention patents, while also promoting green utility model patents as shown in Column (4), albeit with a weaker effect—the former being approximately 1.6 times the latter. Columns (3) and (5) further reveal that incorporating either patent type into the baseline model substantially attenuates the DID coefficient, indicating that both serve as partial mediators. Collectively, these results suggest that LCPPs, enhance EWP, not only directly but also indirectly through a broad green innovation channel—encompassing substantive technological breakthroughs (green invention patents) and incremental technological adaptations (green utility model patents), thereby corroborating Hypothesis 2.

5.2 Analysis of the mediating effect of digitalization level

As theorized above, low-carbon pilot initiatives foster the digital transformation of traditional industries and stimulate nascent digital sectors, thereby accelerating economy-wide digital upgrading. Enhanced digitalization concurrently improves urban governance efficacy and advances digital-real economy integration in key sectors. These synergistic effects facilitate the low-carbon retrofitting of energy- and pollution-intensive industries, optimize energy structure, and ultimately augment environmental quality and living standards.

Consequently, this study incorporates digitalization level as a mediating variable into the baseline model, as specified in Equations 1, 4, 5, to examine whether digitalization is a prerequisite mechanism through which LCPPs enhance EWP. In particular, the digitalization variable is drawn from the Digitalization Level sub-index of the Peking University Digital Inclusive Finance Index, which is available from the year 2011 onward. Consequently, the mechanism test using digitalization as the mediator variable covers the period from 2011 to 2023, yielding 3,549 observations. Table 13 reports the estimation results for the mediating effect of digitalization level.

TABLE 13. Estimation result of intermediary mechanism (Digitalization Level).

*, , and * denote statistical significance at the 10%, 5%, and 1% levels, respectively.

Table 13 presents the mediation analysis results for digitalization. Column (1) reports the baseline impact of LCPPs on urban EWP. Column (2) presents the estimates of Equation 4, and Column (3) presents the estimates of Equation 5. The low-carbon policy exerts a positive and statistically significant effect on digitalization level, indicating that policy implementation considerably accelerates urban digital transformation. The results in Column (3) reveal that with digitalization introduced as a mediator, the DID coefficient decreases from 0.011 to 0.007, while the mediating variable itself remains statistically significant. This pattern indicates that digitalization serves as a partial mediator: the indirect effect via digitalization accounts for approximately 32% of the total policy effect, with the direct effect and other channels explaining the remaining 68%. The adjusted R-squared registers a modest improvement from 0.539 to 0.565, reflecting the explanatory contribution of digitalization. These findings indicate that the LCPPs accelerate urban digital transformation, which subsequently ameliorates ecological welfare outcomes in pilot cities, thereby substantiating Hypothesis 3.

A structural equation model with city-specific linear trends is employed to corroborate the above mediation mechanisms, as shown in Table 14. By allowing each city to follow its own temporal trajectory, this specification partially absorbs unobserved time-varying heterogeneity. Following standard practice, 1-year lagged green patents are adopted as the mediator, given that the environmental benefits of green technology innovation require a reasonable time lag to materialize. The sample periods align with those in Tables 12, 13 to ensure comparability. The results indicate that indirect effects through both green innovation and digital innovation channels are highly significant (0.0013 and 0.0027, respectively), whereas the direct DID effect is statistically insignificant, suggesting a pattern of full mediation. These findings lend more credible support to the causal mediation pathways underlying Hypotheses 2 and 3.

TABLE 14. Structural equation modeling of mediation effects with city-specific linear trends.

*, , and * denote statistical significance at the 10%, 5%, and 1% levels, respectively.

5.3 Joint mediation and interaction effects of green innovation and digital economy

In order to investigate whether green technology innovation and the digital economy operate synergistically, we estimate a joint mediation model incorporating both mediators. Consistent with the digital economy data coverage (2011–2023), the sample in this analysis is restricted to the same period. Results in Table 15 show that when both mediators are included simultaneously, the policy effect declines by approximately 56% (from 0.0093 to 0.0041): green innovation turns insignificant while the digital economy remains significant, indicating partial absorption. The interaction term is positive and significant, yet the policy main effect becomes insignificant, suggesting that green innovation enhances EWP primarily by enabling digital transformation. Excluding the digital economy restores policy significance (0.0063) and green innovation significance, with the interaction term remaining significant—confirming the absorption mechanism. These findings demonstrate that the two mediators are not parallel but complementary: the digital economy serves as an independent channel and amplifies green innovation’s effect through synergistic interaction.

TABLE 15. Joint mediation and interaction analysis.

*, , and * denote statistical significance at the 10%, 5%, and 1% levels, respectively.

5.4 Analysis of moderating effects

Drawing on the preceding theoretical analysis, this study incorporates urbanization level as a moderating variable into the baseline model to examine its moderating role in the nexus between low-carbon pilot policy and EWP, and to ascertain whether this moderating effect evolves dynamically as the policy is implemented. Table 16 reports the corresponding estimation results.

TABLE 16. The moderating effect of urbanization level on EWP across different periods.

*, , and * denote statistical significance at the 10%, 5%, and 1% levels, respectively.

Table 16 reports the dynamic moderating effects of urbanization rate on low-carbon pilot policy efficacy of Model (6), where Columns (1), (2), and (3) present the impacts on contemporaneous, one-period-lagged, and two-period-lagged EWP, respectively. The results indicate that the interaction term between the low-carbon pilot policy shock and urbanization rate is positive and statistically significant across all examined periods, exhibiting a pronounced increasing trend: from 0.039 in the contemporaneous period to 0.049 in the one-period-lagged specification, and further to 0.064 in the two-period-lagged specification. This substantiates that elevated urbanization rates significantly amplify the efficacy of low-carbon pilot policies in improving EWP, with such amplification effects demonstrating temporal cumulativeness. Cities with higher urbanization levels generate more pronounced ecological welfare improvements in the process of advancing low-carbon pilot initiatives. This aligns with the findings of Zhu and Xia (2019), who demonstrate that cities with higher urbanization levels possess superior environmental governance capacity, green technology innovation capabilities, and environmental regulation enforcement efficiency, thereby corroborating the thesis that urbanization ameliorates environmental performance.

Meanwhile, the main effect of urbanization rate exhibits a marked increasing trend, suggesting that urbanization’s beneficial impact on ecological welfare performance intensifies over time. This temporal amplification likely reflects the maturation of economies of scale, the upgrading of industrial structure, and technological advancements that enhance resource and environmental efficiency in later urbanization stages. As pilot cities advance through urbanization, concomitant improvements in resource utilization efficiency, public service provision, and environmental quality collectively enhance EWP.

6 Test of spatial spillover effects

As the preceding theoretical analysis suggests, LCPPs influence the EWP of pilot cities themselves while simultaneously generating spillover effects on neighboring cities. Given that the pilot cities entered the program at different times and that the sample involves a substantial number of cities, spatial spillover effects are examined separately for different batches to facilitate the analysis.

This study employs Python to calculate the global Moran’s Index of urban EWP for Chinese cities during 2006–2023, with results reported in Table 17. Given the extensive sample period, only representative years are listed due to space constraints. The findings indicate that the global Moran’s Index remains consistently positive throughout the sample period and is statistically significant at the 1% level, confirming significant positive spatial autocorrelation and pronounced spatial clustering in EWP across Chinese cities.

TABLE 17. Global Moran’s index of ecological welfare performance in Chinese cities.

This study employs a comprehensive suite of diagnostic tests—including the Lagrange Multiplier (LM) test, Likelihood Ratio (LR) test, Wald test, and Hausman test to identify the appropriate model specification for spatial spillover estimation, as shown in Table 18. All diagnostic tests reject their respective null hypotheses at least at the 5% significance level. The LM tests confirm the presence of both spatial lag dependence and spatial error dependence. The Wald and LR tests reject the null hypothesis that the Spatial Durbin Model (SDM) can be reduced to the Spatial Lag Model (SAR) or the Spatial Error Model (SEM). The Hausman test favors fixed effects over random effects (chi-square = 36.51, p = 0.037). These diagnostic results collectively establish the Spatial Durbin Model (SDM) with individual fixed effects as the most appropriate specification.

TABLE 18. Diagnostic tests for spatial econometric model selection.

The regression results of Model (7) are reported in Table 19. Columns (1), (2) and (3) present estimates using the spatial economic matrix for the first, second, and third batches of pilot cities, respectively, while Columns (4), (5) and (6) report corresponding estimates using the inverse distance matrix. All specifications include control variables and individual fixed effects. The spatial autoregressive coefficient ρ ranges from 0.448 to 0.497 across all specifications, indicating moderate spatial dependence in urban EWP.

TABLE 19. Spatial spillover effects of EWP across pilot batches: spatial economic benchmark and inverse distance comparison.

*, , and * denote statistical significance at the 10%, 5%, and 1% levels, respectively.

To examine whether the spatial spillover estimates are sensitive to the choice of spatial weight matrix, we further re-estimate the SDM using six alternative spatial weight matrices, including the inverse distance matrix, administrative adjacency matrix, economic-geographic nested matrix, economic-geographic weight matrix, spatial economic matrix, and spatial economic-geographic matrix. The results are reported in Supplementary Appendix F. Across these alternative specifications, the full-sample indirect effects remain positive and statistically significant, suggesting that the spatial spillover estimates are not driven by a particular definition of intercity connectivity.

This study decomposes SDM estimates to quantify the direct and indirect effects of the low-carbon pilot policy. The direct effects on local EWP are generally modest and statistically insignificant. By contrast, the indirect effects exhibit substantial batch heterogeneity. For the first batch, the indirect effects are negative and highly significant. Predominantly comprising regional central cities and provincial capitals, these early pilots likely triggered pollution leakage to neighboring cities through abruptly tightened environmental regulations, thereby generating negative spillovers. For the second and third batches, the indirect effects turn positive and significant (0.054–0.097), with the spatial economic matrix consistently outperforming the inverse distance matrix (third batch: 0.097 vs. 0.067). Strikingly, the third-batch spillover is approximately 1.8 times the second-batch effect under the spatial economic matrix and 1.3 times under the inverse distance matrix, underscoring marked amplification as the policy matures. The total effects similarly exhibit batch heterogeneity: negative for the first batch, yet positive for both the second and third.

These findings reveal a clear evolution from negative to positive spatial spillovers as the policy matures. The spatial economic matrix captures stronger effects than the inverse distance matrix, particularly for the third batch, indicating that economic linkages and geographic proximity both influence policy benefit transmission. Consequently, LCPPs catalyze regional EWP improvements, yet spillover direction and magnitude critically depend on interurban economic connections. As regional integration deepens, the policy generates progressively stronger spatial diffusion, substantiating Hypothesis 5.

7 Conclusion, discussion and limitations

7.1 Conclusion

The low-carbon city pilot program represents a significant initiative by the Chinese government to explore low-carbon and sustainable urban development models. It aims to balance economic growth with enhanced citizen welfare during a period of rapid urbanization and industrialization, in line with the goals of carbon peak and carbon neutrality. Grounded in the development of the EWP index system tailored to the Chinese context, this study assesses the trajectory of EWP from 2006 to 2023. Taking a low-carbon pilot city as a quasi-natural experiment, this paper uses a multi-period Difference-in-Difference model to study the impact of the LCPPs on urban EWP in China and its mechanism. Drawing on an 18-year panel dataset spanning 273 cities, this study confirms that low-carbon pilot construction positively affects urban EWP through enhanced green technological innovation and digital economy development. Furthermore, low-carbon pilot construction not only improves local EWP but also generates spatial spillovers on neighboring cities, although spatial spillover effects exhibit heterogeneity across pilot batches. The moderating role of urbanization strengthens progressively with advancing urbanization and deepening low-carbon pilot implementation. Heterogeneity analyses by city location and carbon peaking schedule further reveal contextual challenges in leveraging low-carbon city construction for EWP improvement. This paper concludes with corresponding policy recommendations.

7.2 Discussion and policy implications

Drawing on the main findings, the following policy recommendations are proposed to assist low-carbon pilot cities in further advancing urban ecological welfare performance.

Firstly, the transformation of economic development patterns and the reduction of energy intensity are identified as the most critical determinants influencing the effectiveness of low-carbon pilot policies in enhancing EWP. This study substantiates that LCPPs exert a significant positive influence on the EWP of pilot cities, and the robustness of these conclusions is corroborated through placebo testing, parallel trend analysis, and a comprehensive consideration of the heterogeneity inherent in treatment effects. Both the full national sample and the regional subsamples identify economic development as a critical determinant of ecological welfare performance improvement. Economic growth in the central region robustly promotes EWP improvement; however, low-carbon pilot construction exerts a negative effect on EWP in this region, whereas its impact remains insignificant in the western region. This indicates that economic development in the central region has not been accompanied by synchronous improvements in ecological welfare performance. This is attributable to the extensive economic growth mode currently prevalent in the region. Cities in central China should therefore adopt targeted energy efficiency initiatives and sustainable development strategies to address the adverse impact of elevated energy intensity on ecological welfare performance. Western pilot cities must be particularly vigilant against non-sustainable growth characterized by high energy consumption and high investment. They should proactively capitalize on their favorable resource endowments to foster the development of new energy sources and green industries, thereby ensuring the concurrent enhancement of economic development and ecological welfare performance. Cities in the eastern region should prioritize rational urban planning and industrial restructuring to mitigate the severe resource depletion and environmental degradation associated with excessive population concentration.

Secondly, the carbon peak target should be set in a manner that is commensurate with the stage and characteristics of urban development. The implementation of low-carbon pilot projects significantly enhances EWP exclusively for cities with carbon peaking targets set between 2021 and 2025, yielding no significant effects for cities with either earlier or later peaking schedules. In cities with earlier carbon peaking targets, LCPPs exert an insignificant effect on EWP improvement; such improvement is driven by the combined forces of fiscal capacity, science and education investment, industrial structure, and economic development. The longer the peaking horizon, the more severely energy intensity suppresses EWP. Consequently, economically underdeveloped cities with delayed peaking targets risk missing green transformation opportunities, thereby falling into the dilemma of high energy consumption coupled with constrained public welfare improvement. Pilot cities are thus urged to accelerate carbon dioxide emission peaking, innovate institutional mechanisms, and fully leverage policy opportunities to accelerate the implementation of carbon development plans.

This study identifies urbanization as a pivotal catalyst for low-carbon pilot construction and ecological welfare improvement. Building on Liu Y. et al. (2025), who demonstrate that new-type urbanization markedly enhances urban air quality, this paper proposes that local governments integrate low-carbon pilot initiatives with new urbanization and green transition strategies. Rational spatial planning, infrastructure upgrading, and public service enhancement should synergistically facilitate orderly resource agglomeration while advancing energy conservation and emission reduction targets. This study further reveals pronounced heterogeneity in spatial spillovers across cohorts. The first batch, designated top-down and relying on mandatory environmental regulations, triggered abrupt emission cuts that displaced carbon and pollution to neighboring jurisdictions, yielding insignificant local effects but negative spatial spillovers. Conversely, subsequent batches, adopting voluntary application with enhanced local autonomy, better aligned with municipal governance objectives. As pilot expansion deepened regional economic integration and inter-city policy learning, synergies between human capital and environmental improvement generated progressively amplified positive spillovers. Consequently, accelerated low-carbon pilot construction has substantially propelled region-wide ecological welfare gains, laying the groundwork for nationwide urban green transformation.

Finally, this study concludes through mediation analysis that low-carbon pilot policies effectively enhance urban EWP by fostering green technological innovation and digital economy development. This paper confirms that fiscal expenditures on science, technology, and education are advantageous to EWP on a national scale. Despite the limited impact of science and education expenditures on ecological welfare performance in the eastern region, such investments remain the most powerful driver of ecological welfare improvements in low-carbon pilot cities in the central and western regions. Although the low-carbon pilot policy in the western region has not yet demonstrated significant effects, sustained increases in science and education spending will significantly boost local talent accumulation and innovation capacity. Empirical evidence demonstrates that low-carbon pilot construction significantly drives digital economy expansion, which in turn exerts a notable mediating effect on ecological welfare performance improvement. Western regions are thus advised to leverage abundant energy endowments, scale science and education investment, upgrade digital infrastructure, and capitalize on the ‘East Data West Computing’ initiative to amplify the digital economy’s spillover effects, thereby enabling leapfrog advancement in EWP.

7.3 Limitations

This study constructs an ecological welfare performance index system at the city level and provides an exploratory assessment of the overall effect and transmission mechanisms through which the low-carbon pilot policy influences urban EWP. However, constrained by data availability, this study does not incorporate the perspectives of enterprises and public responses. Additionally, the quantity of green technology innovation is not disaggregated by source (enterprises, universities, or other organizations), rendering the mechanism analysis insufficiently comprehensive. Therefore, future research exploring the collaborative governance among government, enterprises, and the public during the implementation of low-carbon pilot policies holds significant implications for leveraging the leading role of low-carbon pilot cities in driving economy-wide green transformation. Although we provide mechanism evidence on green technological innovation and digitalization, it does not further identify causal mediation effects using instrumental-variable estimation because valid instruments for city-level green patents and digitalization are difficult to obtain. Future research may address this limitation by exploiting natural experiments, constructing credible exogenous instruments, or using micro-level data to identify the causal transmission channels more precisely.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fenvs.2026.1883687/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Shulei Cheng, Southwestern University of Finance and Economics, China

Reviewed by: Xuan Xie, Hunan Normal University, China

Daleng Xin, Shandong Normal University, China