Pollution and Carbon Emission Reduction Effects of Transit Metropolis Construction: Evidence from China

Abstract

The aim of this study was to empirically examine the effects of China’s Transit Metropolis Construction Pilot (TMCP) policy on pollution and carbon dioxide emission reductions based on annual panel data from 286 prefecture-level cities in China for the period 2011–2019, using a staggered difference-in-differences approach. The results show that the TMCP policy significantly reduced the annual total carbon monoxide and carbon dioxide emissions in pilot cities by approximately 1.624 million and 221.883 million tons, respectively. Further mechanism analysis demonstrated that the TMCP policy reduced pollution and carbon dioxide emissions by improving the operational efficiency of public transit, alleviating urban traffic congestion, and enhancing public environmental awareness. Finally, our heterogeneity analysis indicates that the pollution and carbon dioxide emission reduction effects of the TMCP policy were more pronounced in cities with poor public transit accessibility and low environmental regulation intensity. This study provides policymakers with valuable policy insights into effectively promoting public transit use, reducing urban air pollutants and carbon dioxide emissions, and developing a sustainable urban transportation system.

1. Introduction

The consequences of global warming, such as natural disasters, extreme weather conditions, and species extinction, pose a significant threat to the Earth’s ecological environment and human production and life [1]. As the world’s largest manufacturing country with the highest carbon emissions, China is fully committed to implementing pollution and carbon dioxide emission reduction measures, and has set a national strategic goal of peaking carbon emissions by 2030 and achieving carbon neutrality by 2060. The road transportation sector, a major contributor to fossil energy consumption and air pollutant emissions, produced approximately 1.228 billion tons of carbon dioxide emissions in 2023, accounting for approximately 9.8% of China’s total carbon dioxide emissions [2]. In particular, with the acceleration of urbanization and the rapid increase in the number of private cars, the challenge of reducing urban traffic-related pollutants and carbon dioxide emissions has become extremely severe. According to official data released by the Ministry of Ecology and Environment of China, the number of cars in China reached 336 million in 2023, an increase of 5.3% from that in the previous year, and has become a significant source of air pollution emissions [3].

To reduce urban residents’ reliance on car travel, alleviate traffic congestion, decrease transportation energy consumption and carbon dioxide emissions, and improve the quality of public transit services, the Ministry of Transport of China launched the Transit Metropolis Construction Pilot (TMCP) policy in November 2011. As of December 2024, the Ministry of Transport of China had cumulatively launched four batches of the TMCP project, covering 117 pilot cities, and had invested several billion US dollars in dedicated incentive funds [4]. Despite the significant investment and implementation over many years, only few empirical studies have assessed the pollution and carbon dioxide emission reduction effects of the TMCP policy. To fill this gap, in this study, we aimed to empirically examine the pollution and carbon dioxide emission reduction effects of the TMCP policy and thereby provide empirical evidence on how to synergistically promote both urban pollution and carbon dioxide emission reductions.

Currently, the research on the influencing factors of collaborative pollution reduction and carbon dioxide emission reduction mainly focuses on two dimensions: policy mechanism design [5,6] and digital technology empowerment [7,8]. Additionally, factors such as technological progress [9] and changes in urban spatial form [10] also impact the effectiveness of pollution and carbon dioxide emission reduction efforts. However, few studies have examined the impact of transportation on the combined effects of pollution reduction and carbon dioxide emission reduction, despite extensive research on its negative externalities, such as traffic congestion and carbon emissions [11,12,13]. In this regard, several studies have investigated the role of urban public transportation in reducing pollution and carbon dioxide emissions; however, these studies have primarily focused on individual cities [14,15,16] and have been limited to isolated analyses of either pollution reduction or carbon dioxide emission reduction effects [17,18,19]. Regarding the environmental benefits of the TMCP policy, which are most relevant to our study, Bi et al. (2024) recently conducted an empirical investigation into the impact of China’s TMCP policy on urban air quality [20]. Their results demonstrate that the implementation of the TMCP policy significantly reduced PM_2.5 concentrations. However, Bi et al. (2024) analyzed only the PM_2.5 concentration as a single pollutant indicator, without further examining its potential impact on carbon dioxide emission reduction. Moreover, their research samples were limited to the first two batches of pilot cities, which were mostly large cities with urban populations exceeding 1 million, lacking coverage of medium- and small-sized cities [20]. By contrast, our research covered the third batch of medium-sized cities that implemented the TMCP policy and simultaneously examined the impacts of the TMCP policy on urban pollution (i.e., carbon monoxide emissions) and carbon dioxide emissions.

The main contributions of this article to the existing literature are as follows. First, this article empirically examined the pollution and carbon dioxide emission reduction effects of China’s TMCP policy, thereby expanding the existing literature’s research on the single pollutant PM_2.5 and enriching the limited body of literature in this emerging field. Although a few studies have elaborated on the details [21], diffusion [4], and implementation effects of China’s TMCP policy [20,22], no empirical studies have been conducted to simultaneously examine its potential effects on reducing urban air pollution and carbon dioxide emissions. Second, through the use of a quasi-natural experiment based on the phased implementation of China’s TMCP policy, this article confirms that the development of public transit can achieve the synergy of urban pollution and carbon dioxide emission reductions. This further enriches and expands the existing assessment of the potential environmental benefits of public transit, especially by providing large-sample empirical evidence from China, establishing a reliable causal relationship, and compensating for the shortcomings of most existing literature, which is limited to a single city [14,15,16]. Third, by revealing the underlying mechanisms and heterogeneity of the pollution and carbon dioxide emission reduction effects of the TMCP policy, our study provides valuable insights into understanding how the development of a transit metropolis could achieve its environmental benefits and how policymakers could effectively utilize such a policy to reduce urban air pollution and carbon dioxide emissions.

This article is arranged as follows: Section 2 elaborates on the policy background, and Section 3 presents the research hypotheses. Section 4 describes the methods and data. Section 5 reports the empirical results and analysis, including the baseline regression results, parallel trend assumption, robustness checks, underlying mechanisms, and heterogeneity analysis. Finally, Section 6 provides conclusions and policy implications.

2. Policy Background

To address problems such as the rapid increases in urban population and the number of motor vehicles, the aggravated traffic congestion, the deteriorating environmental pollution, and the intensified pressure on energy consumption that arise from the rapid urbanization process, in March 2004, the Ministry of Construction of China issued the “Opinions on Giving Priority to the Development of Urban Public Transport,” initiating the implementation of the urban public transit priority development strategy to transform urban travel centered around cars into green travel oriented toward public transit. In 2011, the Ministry of Transport of China issued the “Notice on Matters Concerning the Implementation of the National Transit Metropolis City Demonstration Project,” marking the official launch of the TMCP policy. In December 2024, China launched a TMCP policy in 117 cities across four phases, gradually expanding from large cities and megacities with dense populations to medium- or small-sized cities with a population of less than 1 million. The list of specific pilot cities and their implementation dates is shown in

Table 1. List of cities that have implemented the TMCP policy.

Batch	Cities	Date
First batch (15 cities total)	Beijing, Shenzhen, Chongqing, Nanjing, Jinan, Zhengzhou, Wuhan, Xi’an, Changsha, Shijiazhuang, Taiyuan, Dalian, Harbin, Kunming, Urumqi.	2012
Second batch (22 cities total)	Shanghai, Tianjin, Guangzhou, Hangzhou, Ningbo, Hefei, Fuzhou, Nanchang, Qingdao, Shenyang, Changchun, Suzhou, Xinxiang, Zhuzhou, Liuzhou, Haikou, Guiyang, Lanzhou, Xining, Yinchuan, Baoding, Hohhot.	2013
Third batch (50 cities total)	Zhangjiakou, Linfen, Wuhai, Anshan, Panjin, Tonghua, Mudanjiang, Changzhou, Yangzhou, Kunshan, Huzhou, Jinhua, Suzhou, Fuyang, Bengbu, Wuhu, Shangrao, Zaozhuang, Yantai, Weifang, Weihai, Luoyang, Xuchang, Nanyang, Zhumadian, Xiangyang, Yichang, Changde, Zhangjiajie, Loudi, Foshan, Nanning, Guilin, Guigang, Sanya, Chengdu, Zigong, Luzhou, Meishan, Fuling, Zunyi, Kaili, Yuxi, Baoshan, Baoji, Xianyang, Tianshui, Guyuan, Kashgar, Yining.	2017
Fourth batch (30 cities total)	Tangshan, Cangzhou, Xingtai, Yangquan, Jinzhou, Baicheng, Wuxi, Xuzhou, Yancheng, Shaoxing, Taizhou, Yiwu, Jiujiang, Yichun, Jining, Rizhao, Hebi, Puyang, Luohe, Shiyan, Jingzhou, Xianning, Yueyang, Yongzhou, Duyun, Lijiang, Jiuquan, Wuzhong, Bole, Korla.	2023

Notes: The implementation date is defined as the date on which the Ministry of Transport of China publishes the list of pilot cities for the TMCP project.

.

To clarify the assessment targets for each pilot city and to scientifically evaluate the achievements in constructing a transit metropolis, the Ministry of Transport of China issued an evaluation index system for transit metropolis in June 2013. This evaluation system includes 30 specific indicators, such as the share of public transit trips, the coverage rate of transit stations within 500 m, the punctuality rate of public transit, and the average speed of buses. In March 2022, the Ministry of Transport of China updated the transit metropolis evaluation index system, streamlining it to 20 indicators. One indicator can be customized by pilot cities according to their local characteristics of urban public transit development. As of now, 74 pilot cities have passed the assessment conducted by the Ministry of Transport of China and have been awarded the title “National Transit Metropolis Demonstration City”.

3. Research Hypotheses

The implementation of the TMCP policy could reduce urban air pollution and carbon dioxide emissions through at least three mechanisms: improved transit operational efficiency, alleviated urban traffic congestion, and enhanced public environmental awareness. First, the TMCP policy would improve the operational efficiency of public transit in the pilot cities, and through such an operational efficiency improvement effect, its implementation would reduce the reliance of urban residents on private car travel, thereby reducing the emissions of pollutants and carbon dioxide from motor vehicles. A central task of China’s TMCP policy is to fully implement the strategy of prioritizing public transportation and improve the service quality and reliability of urban public transit. Therefore, the implementation of the TMCP policy would improve the public transit infrastructure and encourage people to choose and use public transit services, thereby increasing the ridership of public transit and enhancing the operational efficiency of public transit. For instance, during the five-year period when the TMCP policy was implemented in Hangzhou, 177 new bus routes were opened, 4030 new and updated buses were added or replaced, and the share of public transit in travel increased from 44.2% at the beginning of the project to 61.87% in 2018. Furthermore, the daily ridership of the urban rail transit system in Hangzhou has increased from less than 300,000 people to 1.5 million people, and the satisfaction rate for public transit has reached as high as 87.7% [23].

Second, China’s TMCP policy could alleviate urban traffic congestion. Through this congestion alleviation effect, the implementation of the TMCP policy could also reduce urban traffic pollutants and carbon dioxide emissions. Existing studies have shown that traffic congestion significantly increases carbon dioxide and pollutant emissions during car travel [24,25]. For instance, Sun et al. (2018) found that the carbon emissions of vehicles in congested conditions can be two to three times higher than those at normal driving speeds [26]. Therefore, through the construction of dedicated bus lanes, the promotion of public transit usage, and the optimization of intersection traffic signals, the implementation of the TMCP policy could alleviate urban traffic congestion, thereby further reducing urban air pollution and carbon dioxide emissions. For example, during the construction of Transit Metropolis, Qingdao added 101.8 km of dedicated bus lanes [27], while in Wuhu, the average bus speed during peak hours increased from 16 km per hour to 23.6 km per hour [28].

Finally, the implementation of the TMCP policy could also enhance public environmental awareness. Through this environmental concern effect, the construction of a transit metropolis could increase public awareness of green travel, further decreasing urban residents’ dependence on car travel and thereby reducing urban pollution and carbon dioxide emissions. One of the important tasks in the implementation of the TMCP policy is to enhance public awareness of green travel and increase the proportion of green travel. For instance, during the period of the construction of a Transit Metropolis, Shangrao actively advocated and promoted the development of the urban non-motorized transportation system, enhancing residents’ awareness of green travel, and a total length of 442.84 km of green paths and green corridors was constructed, and the green travel share reached 74.3% [29]. Taiyuan invested in 41,000 public bicycles with free one-hour rides, reaching a daily cycling volume of 568,500 [30].

In conclusion, through the efficiency enhancement, congestion alleviation, and environmental concern effects, the implementation of China’s TMCP policy could reduce urban air pollution and carbon dioxide emissions. Thus, we propose the following research hypotheses:

Hypothesis 1:

The TMCP policy significantly reduces urban pollution and carbon dioxide emissions.

4. Methods and Data

4.1. Model Specification

Our study employed a difference-in-differences (DID) method to empirically investigate the impact of China’s TMCP policy on urban air pollution and carbon dioxide emissions. The baseline regression model is specified as follows:

$$Y_{i,t}={\alpha }_0+{\beta }_{1}TMC{P}_{i,t}+\gamma X_{i,t}+{\mu }_i+{\theta }_t+{\epsilon }_{i,t},$$

(1)

where i and t represent the city and time, respectively; $Y_{i,t}$ represents the dependent variable, such as the per capita carbon monoxide (CO) and carbon dioxide (CO₂) emissions of city i at time t; ${\mathit{TMCP}}_{i,t}$ is a dummy variable that represents whether city i has implemented TMCP policy at time t. The value of ${\mathit{TMCP}}_{i,t}$ is 1 in the year of implementation and in subsequent years; otherwise, its value is 0. $X_{i,t}$ represents a series of control variables; ${\mu }_i$ represents the city fixed effect, which is used to control the inherent differences among cities that do not change over time; ${\theta }_t$ represents a fixed effect of time, which is used to control for factors that affect all cities over time; ${\epsilon }_{i,t}$ represents the random disturbance term; and ${\beta }_1$ is the parameter estimator of interest in this study, which reflects the average treatment effect of the implementation of the TMCP policy on the per capita CO and CO₂ emissions. A negative coefficient indicates that the implementation of the TMCP policy significantly reduced urban pollution and carbon dioxide emissions.

One key assumption that DID estimates to be effective is the parallel trend assumption, which means that the outcome variables in the treatment and control groups had the same trend of change before the implementation of the TMCP policy. To verify this hypothesis, this study followed the approach of Zhang et al. (2024) [31], drew on the ideas of the event study method, and constructed a dynamic DID model as follows:

$$Y_{i,t}={\alpha }_0+{\sum }_{s=-5}^{s=5}{\beta }_{s}TMC{P}_{i,t}^S+\gamma X_{i,t}+{\mu }_i+{\theta }_t+{\epsilon }_{i,t}$$

(2)

where ${\mathit{TMCP}}_{i,t}^s$ represents a series of relative time dummy variables. If the time distance between the implementation of the TMCP policy in city i and the date t is s, then the value is 1; otherwise, it is 0. Before implementation, s < 0; after implementation, s ≥ 0. ${\beta }_s$ reflects the dynamic treatment effect of implementing the TMCP policy on the per capita CO and CO₂ emissions. A statistically insignificant ${\mathit{TMCP}}_{i,t}^s$ coefficient (s < −1) indicates no significant difference in per capita CO and CO₂ emissions between the cities in the treatment and control groups before the implementation of the TMCP policy, thus supporting the parallel trend assumption.

In the context of staggered DID designs, the two-way fixed effects (TWFE) estimator has long been regarded as a straightforward and effective estimation method and has been widely applied in the existing literature. However, recent studies have demonstrated that when treatment effects are heterogeneous, the TWFE estimator may yield biased results [32,33]. To obtain an unbiased estimate of the average treatment effect, it is necessary not only to satisfy the parallel trends assumption but also to ensure that the treatment effect remains constant across groups and over time. Specifically, in the case of a staggered DID design where treatment times vary, using individuals who received treatment earlier as the control group for those treated later can lead to outcome variables that already reflect the effects of the prior treatment. In this situation, the greater the weight assigned to such control groups, the more likely the TWFE estimator will suffer from unexplained biases due to intertemporal cross-contamination [32]. Therefore, this paper first assesses the treatment effect heterogeneity of the TWFE estimator using the Goodman-Bacon decomposition [32], and then re-estimates Equation (2) with three alternative estimators proposed in the recent literature to enhance the credibility of the baseline results [33,34,35].

4.2. Main Variables and Data Sources

4.2.1. Dependent Variable

Considering that public transit is mainly used for the daily travel of urban residents, the impact of the TMCP policy on urban traffic pollutant and carbon dioxide emissions is largely reflected at the individual level. Therefore, following the approach of Li et al. (2025) [36], we used the natural logarithm transformation of per capita CO (lnPCO) and CO₂ emissions (lnPCO₂) as dependent variables to measure the levels of urban traffic pollutants and carbon dioxide emissions, respectively, to examine the pollution and carbon dioxide emission reduction effects of the TMCP policy. We used CO as a measurement indicator for pollutants because it is the main emission from vehicles, and most of it comes from passenger cars rather than trucks. In 2023, the emissions of carbon monoxide, hydrocarbons, nitrogen oxides, and particulate matter from China’s motor vehicles were 7.249 million, 1.872 million, 4.731 million, and 0.044 million tons, respectively. The CO emission from passenger cars was approximately 4.707 million tons, accounting for approximately 64.93% of the total carbon emission from all motor vehicles [3]. The annual emission data on CO and CO₂ were collected from the Emissions Database for Global Atmospheric Research (https://edgar.jrc.ec.europa.eu/, 1 August 2025), and the urban population data were obtained from the LandScan Global Population Dataset (https://landscan.ornl.gov/, 1 August 2025).

4.2.2. Independent Variables

The core independent variable in this study is a binary indicator representing whether the sample cities had implemented the TMCP policy (TMCP). If a city is designated as a pilot city for the TMCP policy in year t, the value of TMCP for that city in year t and subsequent years is 1; otherwise, it is 0. We obtained the pilot cities and dates of the TMCP policy implementation from the official website of the Ministry of Transport of China, as shown in Table 1.

4.2.3. Control Variables

We also controlled a series of time-varying factors that might have affected urban CO and CO₂ emissions. First, the level of economic development and the degree of population concentration are fundamental factors influencing urban pollutant and carbon dioxide emissions. Therefore, this study followed the approaches of Xu et al. (2022) [37] and Ren et al. (2022) [38], using the natural logarithm of per capita gross domestic product (lnPGDP) to measure the economic development level of a region and the natural logarithm of population density (lnPD) to measure the degree of population concentration in a city. Second, the energy intensity reflects the energy utilization efficiency of regional economic activities, and the industrial structure directly affects energy consumption intensity and emission characteristics. Following the approach of Ren et al. (2022) [38] and He and Wang (2025) [39], we measured energy intensity (lnEI) using the natural logarithm of the ratio of total energy consumption to the gross domestic product (GDP) and the industrial structure (IS) using the proportion of the second industry’s output value to the GDP. Moreover, technological innovation and human capital play potential supporting roles in reducing pollution and carbon dioxide emissions, and government actions influence ecological environment governance through the allocation of fiscal resources. Therefore, drawing on the research of Ren et al. (2022) [38] and He and Wang (2025) [39], we used the ratio of local total expenditure on science and education to local general public budget expenditure to measure innovation capacity (IC) and the fiscal freedom degree (FFD) to measure government behavior, which is calculated as the ratio of local general public budget revenue to the local general public budget expenditure. Except for the urban population data, which were sourced from the LandScan Global Population Dataset, the data of the remaining control variables were derived from the China Urban Statistical Yearbook. We excluded cities where the control variable data were severely lacking, such as Taiwan, Macao, and Hong Kong, and used the linear interpolation method to fill in the few missing values of some sample cities. Furthermore, to reduce the influence of outliers on the estimation results, we truncated the data of all continuous variables at the 1% and 99% percentiles. To avoid possible interference from the COVID-19 pandemic on the estimation results, our sample observation time window was limited to 2019. Eventually, a yearly panel dataset covering 286 prefecture-level cities from 2011 to 2019 was compiled and used for subsequent empirical analysis. The descriptive statistics of the main variables are presented in

Table 2. Descriptive statistics of the main variables.

Variables	Definitions	Mean	S.D.	Min	Max
lnPCO	Per capita CO emissions (kg)	4.210	0.786	2.652	9.201
lnPCO2	Per capita CO2 emissions (kg)	8.746	0.849	6.274	11.650
TMCP	Binary indicator of TMCP policy implementation	0.159	0.366	0	1
lnPGDP	Per capita GDP (Yuan)	10.691	0.569	9.394	12.051
lnPD	Population density (people per square kilometer)	5.407	0.956	2.659	7.716
lnEI	Energy intensity (tons of standard coal per million yuan)	4.139	0.484	3.044	5.383
IS	Industrial structure (%)	46.716	10.521	19.25	72.83
IC	Innovation capability	0.194	0.041	0.094	0.291
FFD	Financial freedom degree	0.462	0.221	0.102	1.007

Notes: Observations N = 2574.

.

5. Empirical Results and Analysis

5.1. Baseline Regression Results

The benchmark regression results based on Equation (1) are shown in

Table 3. Results of benchmark regression.

Variables	lnPCO	lnPCO2	lnPCO	lnPCO2
	(1)	(2)	(3)	(4)
TMCP	−0.055 ***	−0.073 ***	−0.028 **	−0.049 ***
	(0.014)	(0.013)	(0.013)	(0.012)
lnPGDP			0.051 *	0.034
			(0.029)	(0.031)
lnPD			−1.022 ***	−0.904 ***
			(0.069)	(0.080)
lnEI			0.025	0.055 **
			(0.022)	(0.023)
IS			−0.001	0.002
			(0.001)	(0.001)
IC			0.053	−0.209
			(0.123)	(0.199)
FFD			0.056	0.075
			(0.050)	(0.064)
Constant	4.218 ***	8.757 ***	9.088 ***	12.964 ***
	(0.002)	(0.002)	(0.349)	(0.542)
Observations	2574	2574	2574	2574
R-squared	0.986	0.987	0.987	0.988
Date FE	YES	YES	YES	YES
City FE	YES	YES	YES	YES
Control	NO	NO	YES	YES

Notes: Robust standard errors are reported in parentheses. FE, fixed effects. *** p < 0.01, ** p < 0.05, * p < 0.1.

, and the standard errors are clustered at the city level. Columns (1) and (2) present the regression results using per capita CO and CO₂ emissions as dependent variables, respectively, with controlled time and city fixed effects. Columns (3) and (4) further incorporate control variables on this basis. Regardless of whether control variables were included, the implementation of the TMCP policy significantly reduced the per capita CO and CO₂ emissions, indicating that the TMCP policy significantly reduced pollution and carbon dioxide emissions. Therefore, the proposed research hypothesis is supported. Since columns (3) and (4) included control variables, all subsequent analyses in this article are based on columns (3) and (4). Specifically, the implementation of the TMCP policy led to a 2.8% reduction in per capita CO emission and a 4.9% reduction in per capita CO₂ emission in the pilot cities. Given that the average per capita CO and CO₂ emissions in the pilot cities were 0.095 and 8.369 tons, respectively, the annual per capita CO and CO₂ emissions in the pilot cities decreased by 0.003 (0.095 × 0.028) tons and 0.410 (8.369 × 0.049) tons, respectively. On the basis of the total population of 541.18 million in the pilot cities in 2019, we roughly estimated that the implementation of the TMCP policy led to an annual total reduction of approximately 1.624 (0.003 × 541.18) million tons of CO emission and 221.883 (0.410 × 541.18) million tons of CO₂ emission in the sample cities. In terms of economic welfare, based on the assumptions in Cao and Shen [40], each liter of gasoline consumed by a car releases 2.230 kg of CO₂, and the average cost per liter of gasoline is 7.000 yuan. We estimate that the implementation of the TMCP policy has generated approximately 696.494 (221.883 × 7/2.230) billion yuan in economic benefits for the sample pilot cities by reducing carbon dioxide emissions.

5.2. Parallel Trend Assumption

An important condition for the validity of the DID estimation is to satisfy the parallel trend assumption. To this end, we employed the dynamic DID model of Equation (2) for parameter estimation and present the estimation results in Figure 1. Before the implementation of the TMCP policy, the ${\mathit{TMCP}}_{i,t}^s$ coefficient (s < −1) was not significant, which indicates no significant pretreatment trend in the per capita CO and CO₂ emissions between the cities in the treatment and control groups, supporting the parallel trend assumption. Furthermore, it can be observed that starting from the year when the TMCP policy was implemented, the ${\mathit{TMCP}}_{i,t}^s$ coefficient (s ≥ 0) began to be significantly negative. Moreover, this effect remained significant five years after the implementation of the TMCP policy. This indicates that the pollution and carbon dioxide emission reduction effects of the TMCP policy were continuous and long-lasting. Furthermore, it is worth noting that as the implementation of the TMCP policy progresses, its effects on pollution reduction and carbon dioxide emission reduction become increasingly significant, which further supports our baseline regression results.

5.3. Robustness Checks

To further enhance the reliability of the benchmark regression results, we also conducted a series of robustness checks, including propensity score matching (PSM), other related policies, alternative measures of the dependent variable, placebo tests, treatment effect heterogeneity, and double machine learning estimation.

5.3.1. PSM-DID Estimation

During the implementation of the TMCP policy, the government may give priority to selecting regions with high economic development levels and well-developed transportation infrastructure as pilot cities rather than randomly allocating them, which could lead to estimation bias. Therefore, before conducting the DID estimation, we first performed PSM of the sample cities for the PSM-DID estimation. Specifically, we used the per capita GDP, population density, energy intensity, industrial structure, innovation capability, and degree of financial freedom as matching variables, and estimated the propensity score using the probit model. The caliper nearest neighbor matching method was then applied to perform 1:1 matching with the replacement on a year-by-year basis. On the basis of the matched samples, we then conducted a DID estimation. The PSM-DID estimation results are shown in columns (1) and (2) of

Table 4. Results of robustness checks.

Variables	PSM-DID		Other Related Policies		Alternative Measures
	lnPCO	lnPCO2	lnPCO	lnPCO2	lnCOI	lnCEI
	(1)	(2)	(3)	(4)	(5)	(6)
TMCP	−0.027 **	−0.040 ***	−0.026 **	−0.048 ***	−0.039 **	−0.051 ***
	(0.013)	(0.012)	(0.013)	(0.013)	(0.015)	(0.014)
Constant	9.008 ***	12.927 ***	9.099 ***	13.070 ***	11.504 ***	15.581 ***
	(0.362)	(0.562)	(0.356)	(0.538)	(0.616)	(0.759)
Observations	2385	2385	2574	2574	2574	2574
R-squared	0.987	0.988	0.988	0.988	0.982	0.981
Date FE	YES	YES	YES	YES	YES	YES
City FE	YES	YES	YES	YES	YES	YES
Control	YES	YES	YES	YES	YES	YES

Notes: Robust standard errors are reported in parentheses. FE, fixed effects. *** p < 0.01, ** p < 0.05, * p < 0.1.

. The estimated TMCP coefficients are significantly negative and quantitatively similar to our baseline estimate.

5.3.2. Other Related Policies

During the implementation of the TMCP policy, the sample cities might have also implemented other policies at the same time, which might have an impact on the estimation results. Therefore, in this study, we also controlled for three other related policies that were simultaneously implemented during the same period: the Low-Carbon City Pilot, Broadband China Pilot Program, and Smart City Pilot. Specifically, we used a binary indicator to measure whether each sample city implemented the aforementioned policies, and we then added these dummy variables to the baseline regression model and reperformed DID estimation. As shown in columns (3) and (4) of Table 4, the estimated TMCP coefficients are all significantly negative, consistent with the results of the baseline regression analysis.

5.3.3. Alternative Measures of the Dependent Variable

Drawing on the practices in previous studies [39,41], we measured the intensity of CO emission by the amount of CO emission per unit of GDP (COI), and the intensity of CO₂ emission by the amount of CO₂ emission per unit of GDP (CEI). We then repeated the DID estimation. As shown in columns (5) and (6) of Table 4, the estimated TMCP coefficients remains significantly negative, consistent with the results of the baseline regression analysis. We also used nitrogen oxides instead of CO to measure pollutant emissions from motor vehicles and then re-estimated the DID model, with consistent results.

5.3.4. Placebo Test

To avoid the possibility that the regression results of the staggered DID might be caused by other unobserved factors, we conducted a placebo test. We randomly selected 106 cities from all sample cities to replace the actual cities implementing the TMCP policy and randomly assigned a fictitious policy implementation time. We then re-estimated the DID model, saved the estimation results, and repeated this process 1000 times. Finally, 1000 regression coefficients and their corresponding p values were obtained. The kernel density distribution graph of the 1000 coefficients estimates was plotted, as shown in Figure 2. The estimated placebo coefficient values are clustered around 0 and approximately follow a normal distribution. Therefore, the reduction effects on pollution and carbon dioxide emissions resulting from the implementation of the TMCP policy that were observed in the benchmark regression were unlikely to be caused by unobservable factors.

5.3.5. Treatment Effects Heterogeneity

To alleviate the potential bias of the TWFE staggered DID estimator in the context of treatment effect heterogeneity, we first performed the Goodman–Bacon decomposition [32]. As shown in

Table 5. Results of the Goodman–Bacon decomposition.

Comparison Groups	Coefficient	Weight
Panel A Dependent variable: lnPCO
Treatment vs. Never Treated	−0.028	0.876
Earlier Group Control vs. Later Group Treatment	−0.003	0.040
Later Group Control vs. Earlier Groups Treatment	0.001	0.084
Panel B Dependent variable: lnPCO2
Treatment vs. Never Treated	−0.049	0.876
Earlier Group Control vs. Later Group Treatment	−0.004	0.040
Later Group Control vs. Earlier Groups Treatment	−0.001	0.084

, most of the impact of per capita CO and CO₂ emissions on the final TWFE estimator was from the estimated coefficients of the control group, which consisted of cities that had never been affected by the TMCP policy, with a weight of 87.6%. Therefore, in the setting of this study, the bias caused by treatment effect heterogeneity is relatively small. In addition, we drew on previous studies [33,34,35], proposed treatment effect heterogeneity as an alternative estimator for the robustness test, and present the estimation results in Figure 3. Before the implementation of the TMCP policy, no significant difference in per capita CO and CO₂ emissions was found between the cities in the treatment and control groups, whereas after the policy was implemented, the per capita CO and CO₂ emissions in the cities in the treatment group decreased significantly, which further supports the robustness of the benchmark regression results.

5.3.6. Double Machine Learning Estimation

To further enhance the reliability of the results, we also conducted a robustness test using the double machine learning model proposed by Chernozhukov et al. (2018) [42]. The partial linear regression model was constructed as follows:

$$Y=D{\theta }_0+g_0\left(X\right)+U,E\left[U|X,D\right]=0$$

(3)

$$D=m_0\left(X\right)+V,E\left[V|X\right]=0$$

(4)

where Y represents the dependent variable, indicating the per capita CO and CO₂ emissions, and D is the treatment variable. If a city implemented the TMCP policy, its value is 1; otherwise, it is 0. $X=\left(X_1,\dots \dots ,X_p\right)$ is a control variable at the city level. $g_0$ and $m_0$ are unknown nuisance functions that can be estimated using the lasso (LASSO), gradient boosting (GB), neural network (NN), and random forest (RF) algorithms. U and V are random error terms. ${\theta }_0$ represents the average treatment effect. The estimation results of double machine learning are shown in

Table 6. Results of the double machine learning estimation.

Panel A Dependent Variable: lnPCO
Variables	LASSO	GB	NN	RF
	(1)	(2)	(3)	(4)
TMCP	−0.047 ***	−0.081 *	−0.058 ***	−0.108 **
	(0.008)	(0.042)	(0.018)	(0.045)
Constant	−0.000	0.003	0.015 ***	0.010
	(0.002)	(0.013)	(0.002)	(0.013)
Observations	2574	2574	2574	2574
Date FE	YSE	YSE	YSE	YSE
City FE	YES	YES	YES	YES
Control	YES	YES	YES	YES
Panel B Dependent variable: lnPCO2
Variables	LASSO	GB	NN	RF
	(1)	(2)	(3)	(4)
TMCP	−0.063 ***	−0.108 ***	−0.154 ***	−0.169 ***
	(0.008)	(0.041)	(0.021)	(0.042)
Constant	0.001	0.000	−0.005	−0.000
	(0.002)	(0.014)	(0.003)	(0.013)
Observations	2574	2574	2574	2574
Date FE	YSE	YSE	YSE	YSE
City FE	YES	YES	YES	YES
Control	YES	YES	YES	YES

Notes: Robust standard errors are reported in parentheses. FE, fixed effects. *** p < 0.01, ** p < 0.05, * p < 0.1.

. Regardless of which learning algorithm was used, the TMCP coefficients were all significantly negative. This indicates that the implementation of the TMCP policy significantly reduced the per capita CO and per capita CO₂ emissions, consistent with the baseline regression results. Furthermore, it can be observed that the estimated results from the double machine learning approach are larger than the baseline estimates, indicating that our baseline estimates are relatively conservative and that the pollution reduction and carbon dioxide emission reduction effects of the TMCP policy are greater than we initially estimated.

5.4. Underlying Mechanisms

As mentioned earlier, China’s TMCP policy could reduce pollution and carbon dioxide emissions by improving the operational efficiency of public transit, alleviating urban traffic congestion, and enhancing public environmental awareness. To verify these potential mechanisms, we first examined the impact of China’s TMCP policy on public transit ridership (PTR). As shown in column (1) of

Table 7. Results of the mechanism analysis.

Variables	PTR	CDI	SMOG-Search	EP-Search
	(1)	(2)	(3)	(4)
TMCP	0.390 **	−0.032 **	3.061 ***	8.619 ***
	(0.166)	(0.016)	(0.461)	(1.346)
Constant	−4.023	7.574	−102.972 ***	−196.122 ***
	(3.888)	(5.623)	(15.000)	(36.357)
Observations	2527	268	2544	2562
R-squared	0.990	0.837	0.860	0.940
Date FE	YES	YES	YES	YES
City FE	YES	YES	YES	YES
Control	YES	YES	YES	YES

Notes: Robust standard errors are reported in parentheses. FE, fixed effects. *** p < 0.01, ** p < 0.05, * p < 0.1.

, the TMCP coefficient is significantly positive at the 5% level, indicating that the implementation of the TMCP policy significantly increased the ridership of public transit. Second, we investigated the impact of the TMCP policy on urban traffic congestion. We measured the urban traffic conditions using the congestion delay index (CDI), with data from the Choice Database (https://choice.eastmoney.com/dataservice), which records the daily congestion delay index of 100 major cities in China released by Amap since October 2015. The higher the congestion delay index, the more severe the traffic congestion. Owing to the lack of congestion data, we excluded sample observations prior to 2015 and cities that had already implemented the TMCP policy before 2015 to avoid estimation bias caused by using the always-treated cities as the control groups [37]. As shown in column (2) of Table 7, the implementation of the TMCP policy significantly reduced the congestion delay index, indicating that it alleviated urban traffic congestion. Finally, following the approach of Dai et al. (2024) [43], we used the Baidu search index of urban residents for the keywords “smog” (SMOG-search) and “environmental pollution” (EP-search) to represent the degree of urban residents’ concern about the environment. As shown in columns (3) and (4) of Table 7, the TMCP coefficients are all significantly positive, indicating that the implementation of the TMCP policy significantly enhanced urban residents’ concern about the environment.

5.5. Heterogeneity Analysis

We next conducted a heterogeneity analysis to examine the potential differences in the pollution and carbon dioxide emission reduction effects of the TMCP policy among the sample cities. First, we examined the potential moderating effect of public transportation accessibility. Specifically, we divided the sample into two subsamples based on whether the per capita number of public transit vehicles in the cities in 2011 was above the median: high and low transit availability. We then conducted a DID estimation for each subsample, and the results are shown in Panel A of

Table 8. Results of heterogeneity analysis.

Panel A	High Transit Availability		Low Transit Availability
Variables	lnPCO	lnPCO2	lnPCO	lnPCO2
	(1)	(2)	(3)	(4)
TMCP	−0.021	−0.032 *	−0.047 *	−0.074 ***
	(0.013)	(0.018)	(0.027)	(0.020)
Constant	9.134 ***	12.896 ***	9.219 ***	12.995 ***
	(0.514)	(0.676)	(0.453)	(0.890)
Observations	1305	1305	1305	1305
R-squared	0.996	0.989	0.980	0.988
Date FE	YES	YES	YES	YES
City FE	YES	YES	YES	YES
Control	YES	YES	YES	YES
Panel B	High regulation intensity		Low regulation intensity
Variables	lnPCO	lnPCO2	lnPCO	lnPCO2
	(1)	(2)	(3)	(4)
TMCP	−0.027 *	−0.010	−0.065 ***	−0.038 *
	(0.014)	(0.014)	(0.020)	(0.020)
Constant	13.374 ***	10.285 ***	11.928 ***	8.523 ***
	(0.725)	(0.523)	(0.906)	(0.455)
Observations	972	972	1602	1602
R-squared	0.989	0.996	0.987	0.982
Date FE	YES	YES	YES	YES
City FE	YES	YES	YES	YES
Control	YES	YES	YES	YES

Notes: Robust standard errors are reported in parentheses. FE, fixed effects. *** p < 0.01, ** p < 0.05, * p < 0.1.

. The pollution and carbon dioxide emission reduction effects of the TMCP policy are more pronounced in cities with low transit availability. Second, we examined the moderating effect of the intensity of environmental regulations. To improve urban air quality and control greenhouse gas emissions, the State Council of China announced a list of 113 environmental protection priority cities in November 2007 [44]. Therefore, we divided the sample into two subsamples based on whether the cities are designated as environmental protection priority cities: low and high environmental regulation intensities. We then performed a DID estimation for each subsample. As shown in Panel B of Table 8, the implementation of the TMCP policy had a more significant effect on per capita CO and per capita CO₂ emission reductions in the cities with lower levels of environmental regulation.

6. Conclusions and Policy Implications

This study empirically examined the pollution and carbon dioxide emission reduction effects of China’s TMCP policy using a staggered DID method based on annual panel data of 286 prefecture-level cities in China from 2011 to 2019. Our results show that the implementation of the TMCP policy significantly reduced per capita CO and CO₂ emissions. This conclusion remains valid after a series of robustness checks, and such pollution and carbon dioxide emission reduction effects were long-lasting and even became stronger over time. The implementation of the TMCP policy led to an average reduction of approximately 2.8% in per capita CO emissions and 4.9% in per capita CO₂ emissions in the pilot cities. These figures imply that the annual total CO and CO₂ emissions in the sample cities decreased by approximately 1.624 million and 221.883 million tons, respectively. Furthermore, our mechanism analysis demonstrated that the implementation of the TMCP policy significantly increased public transit ridership, alleviated traffic congestion, and enhanced public environmental awareness. This supports the theoretical framework proposed in this study that the development of a transit metropolis could achieve pollution and carbon dioxide emission reductions through three potential mechanisms: efficiency improvement, congestion alleviation, and environmental concern effect. Finally, the results of our heterogeneity analysis showed that the pollution and carbon dioxide emission reduction effects of the TMCP policy were more significant in the cities with poor public transit accessibility and low environmental regulation intensity.

The results of this study have several significant implications. First, our study demonstrates that China’s TMCP policy has a significant effect on pollution and carbon dioxide emission reductions. Therefore, policymakers should comprehensively summarize and promote the successful experiences of pilot projects for developing a transit metropolis and further expand the coverage and number of pilot cities. Other countries and regions can also draw on the experiences and practices of China’s TMCP policy to reduce urban traffic pollutants and carbon dioxide emissions. Second, on the basis of the results of our mechanism analysis, policymakers can promote the use of urban public transit by developing a transit metropolis. This not only reduces urban traffic pollutants and carbon dioxide emissions but also alleviates urban traffic congestion and enhances public environmental awareness, thereby building a green, low-carbon, and sustainable urban transportation system. Finally, on the basis of the results of our heterogeneity analysis, policymakers can prioritize the implementation of the TMCP policy in cities with poor public transit accessibility and low environmental regulation intensity to achieve the strongest synergy effect on pollution and carbon dioxide emission reductions. In summary, the research findings presented in this paper have significant implications for advancing sustainable urban development. At the theoretical level, this study enriches existing theories of sustainable transportation and urban environmental governance by revealing the synergistic effects of the TMCP policy on reducing pollution and carbon dioxide emissions. It also provides new empirical evidence that enhances understanding of the role of public transportation systems in achieving urban sustainable development goals. At the practical level, our study offers policy guidance for Chinese government departments to promote sustainable development in the transportation sector and presents operational strategies and references for global cities to implement the United Nations Sustainable Development Goals, thereby fostering the development of green, low-carbon, and sustainable urban transportation systems.

This study has certain limitations that suggest directions for future research. First, due to the lack of data on the funds invested by various sample cities in implementing the TMCP policy, we are unable to conduct a comprehensive cost–benefit analysis of the policy’s effects on pollution and carbon dioxide emission reduction. Therefore, when cost data become available, future research can perform such analyses to more thoroughly and objectively evaluate the social and economic value of implementing the TMCP policy. Second, this study examined only the local effects of the TMCP policy on pollution reduction and carbon dioxide emission reduction, without considering its potential spillover effects on surrounding cities. Therefore, it is both interesting and important to further investigate these potential spatial spillover effects and comprehensively assess the social and environmental impacts of the TMCP policy. Third, this study empirically examined only the pollution reduction and carbon dioxide emission reduction effects of the TMCP policy implemented in Chinese cities. Future research could extend this analysis to other emerging economies or developed countries to generalize our findings. Finally, although we investigated the mechanisms underlying the pollution and carbon dioxide emission reduction effects of the TMCP policy, limited data availability prevented us from fully elucidating these mechanisms. For instance, using the Baidu search index for urban residents with the keyword “environmental pollution” as a proxy indicator of public environmental awareness may involve measurement errors. Future research could directly examine the impact of the TMCP policy on urban residents’ awareness of green travel or their green travel preferences, thereby further expanding our research.