Does Low-Carbon Pilot City Policy Reduce Transportation CO₂ Emissions? Evidence from China

Abstract

Transportation is one of the major carbon dioxide (CO₂)-emitting industries, facing substantial reduction pressure under low-carbon sustainable development. Cities are key to reducing transportation CO₂ emissions, and the Low-Carbon City Pilot Policy (LCCPP) is essential to advance the development of low-carbon cities and achieve peak-carbon and carbon-neutral targets. In this paper, we analyse the effect of the LCCP on transportation CO₂ emissions using a multiperiod difference-in-differences (DID) method with data from 284 Chinese cities between 2006 and 2020. The results indicate a substantial reduction in urban transportation CO₂ emissions through the LCCP, and that the enhancement of urban public transportation levels and residents’ green mobility are effective ways to accomplish this. This conclusion is upheld after conducting various robustness tests. Examination of the heterogeneity of the results and spatial analysis revealed that the LCCPP significantly reduced transportation CO₂ emissions in eastern, western, and low-economy cities in China, but not in central and high-economy cities, that the reduction effect was better for southern, non-resource-based cities than for northern, resource-producing cities, and that it exerted notable spillover effects in surrounding cities. The results of this paper offer valid policy insights and practical guidance to maximise the CO₂ reduction effects of the LCCP in the transportation sector.

1. Introduction

Greenhouse gas emissions, particularly CO₂, significantly impact the environment and humanity [1,2]. The transportation industry accounts for about 24% of global GHG emissions [3], a share expected to rise, with China being the top CO₂ emitter, where transport ranks third among emission sources [4,5]. Reducing transport sector emissions is crucial for meeting carbon reduction and neutrality goals [6].

Currently, governments have adopted a range of policies to control GHG emissions [7] and growing numbers of cities are taking low-carbon development as a key central strategy to address climate change. The inhibitory effect of environmental regulation on CO₂ emissions is empirically demonstrated through numerous scholars [8,9,10,11], and the LCCP, as a key environmental regulation, has become an important policy for coping with climate warming and reducing CO₂ emissions. In 2010, China formally launched its Low-Carbon City Pilot policy (LCCPP), subsequently broadening the scope in 2012 and 2017. The three batches of low-carbon pilot cities, as released by the National Development and Reform Commission (NDRC) of China, are spatially distributed and visualised using ArcGIS 10.8 software, as shown in Figure 1.

The LCCPP clearly indicates specific tasks, emphasising the acceleration of developing low-carbon industrial, energy, building, and transportation sectors, as well as changing individuals’ lifestyles. It actively explores innovative experiences and practices, and plans and builds urban transportation, energy, and other infrastructures in accordance with low-carbon concepts. At present, China has launched pilot programmes in 6 provinces and 81 cities [12,13], which have had obvious effects on reducing total urban carbon emissions [14], as well as the industrial [15] and hotel sectors’ carbon emissions, and have promoted the tourism industry’s carbon emission efficiency [16,17]. However, studies on the specific impacts of the policy on CO₂ reduction in the transportation sector are limited; there is still important space for exploration.

To fill this gap, this paper evaluates the contribution of the LCCPP to urban transportation CO₂ emissions and responds to the following research questions: (1) Does the LCCPP reduce transportation CO₂ emissions in cities? (2) Does the effectiveness of the LCCPP vary according to the characteristics of different cities? (3) Does the implementation of the LCCPP impact the transportation CO₂ emissions in the surrounding cities?

Regarding the above questions, this paper analyses the impacts of the LCCPP on transportation CO₂ emissions using a multiperiod difference-in-differences (DID) approach with the panel data for 284 Chinese cities from 2006 to 2020. The primary novel contributions of this paper are highlighted in three main areas. First, in terms of policy evaluation, this study takes the Low-Carbon City Pilot Policy (LCCPP) as a quasi-natural experiment and focuses specifically on the city level to investigate its impact on transportation CO₂ emissions. Most existing research on transport-related carbon emissions has been conducted at the provincial or regional scale, whereas this study provides new city-level empirical evidence that reveals more precise spatial and policy effects. In addition, by integrating the propensity score matching (PSM) method with the multi-period difference-in-differences (DID) model, the analysis effectively mitigates potential sample self-selection bias, thereby improving the robustness and credibility of the estimated results. Second, the study explores the heterogeneous impacts of the LCCPP across cities with different characteristics and regions, uncovering policy effects that vary according to city type and developmental context. These findings differ from prior studies that often emphasise uniform policy impacts. Third, this paper empirically verifies the spatial spillover effects of the LCCPP on transportation CO₂ emissions in neighbouring cities, a dimension rarely examined in existing literature. Furthermore, the mechanism analysis reveals that while the LCCPP effectively promotes emission reduction through improvements in public transport and green mobility, technological innovation in transportation has not yet shown a significant carbon reduction effect, providing new insights that enrich and extend the existing understanding of low-carbon policy impacts. The detailed content framework is shown in Figure 2.

The remaining part of this paper is organised as follows: Section 2 provides a comprehensive review of the relevant literature. Section 3 outlines the methodology and details the data used in the analysis. Section 4 gives baseline regression results as well as the indirect mechanism thereof, complemented by a series of robustness checks. Section 5 examines heterogeneity as well as spatial spillover. Finally, Section 6 concludes with findings and policy implications.

2. Literature Review and Theoretical Hypotheses

2.1. Literature Review

The orderly advancement of low-carbon pilot work has brought analysis of the LCCPP’s effects to the attention of a wide range of scholars. By combing previous studies, we determined that the relevant literature focuses on two main areas.

One pertinent branch of literature examines the effects of the LCCPP. For instance, Zhang et al. [18] used a DID approach to evaluate its effect on urban total factor productivity (TFP) and concluded that LCCPP significantly enhanced TFP; however, the impacts on the various components of TFP varied. This conclusion parallels the study of Cheng et al. [19], who utilised a DID approach to determine that the policy significantly and consistently favours total green factor productivity. Consistent with these studies, Song et al. [20] and Zou et al. [21] demonstrated that LCCPP greatly improves the overall eco-efficiency and technological innovation capacity of cities; however, the degree of impact varied across different cities. In addition, scholars found that LCCPP can reduce urban energy poverty [22], improve urban air quality [23], and enhance corporate ESG practices [24], increasing urban employment [25].

The other branch of related literature explores the LCCPP’s effects on CO₂ emissions, covering two main aspects. On one hand, there are some divergent findings regarding the impact of the LCCPP on CO₂ emissions. For instance, Liu et al. [26] utilised data from 285 Chinese cities, concluding that the LCCPP significantly decreased carbon intensity but did not have notable effects on the total carbon emissions. Conversely, Lyu et al. [27] assessed the policy impacts of three batches of pilot cities based on 331 cities, and utilising DID modelling, they found that the LCCPP significantly lowered total and per capita CO₂ emissions. In addition, Du et al. [28] and Yu et al. [29] evaluated the effects of pilot policies on CO₂ emission efficiency using a DID model and PSM–DID model, respectively. The former discovered that policies greatly enhanced CO₂ emission efficiency and produced lasting effects. The latter study revealed that, overall, the LCCPP notably enhanced CO₂ emission efficiency in pilot scheme cities; however, the impact was significant in the second cohort but not in the first cohort. On the other hand, there have also been specialised studies that discovered that the LCCPP affects urban CO₂ emissions by influencing residents’ travel patterns. Liu et al. [30] found that utilising DID and synthetic control approaches, the LCCPP enhanced residential lifestyle green transition by reducing domestic carbon emissions by approximately 15.3%. Li et al. [31] empirically investigated how the LCCPP affects public transit utilisation and found that developing low-carbon cities significantly increased transit ridership and the frequency of transit trips per capita.

To conclude, the LCCPP has demonstrated significant effectiveness in reducing total urban carbon emissions, with extensive research supporting its positive impact on decarbonizing the industrial and building sectors [15,32]. In recent years, a growing body of literature has also analysed its role in lowering carbon emissions from residents [33], tourism, the hotel industry [16], and the trade circulation industry [34]. Notably, a substantial portion of these emissions—whether total, residential, tourism-related, or trade circulation—is closely tied to transportation activities. These relationships underscore the indirect ways in which the LCCPP mitigates urban emissions through cascading environmental and behavioural changes, highlighting the policy’s critical role in achieving broader urban decarbonization goals. However, despite these insights, the existing literature has yet to systematically investigate the direct impact of LCCPP on transportation sector CO₂ emissions. Most studies have concentrated on evaluating the policy’s effects on overall urban environmental outcomes and aggregate CO₂ reductions, leaving a significant research gap regarding its specific influence on transportation decarbonization.

In view of this, we incorporate the implementation of the three batches of the LCCPP into a comprehensive analytical framework and employ a multiperiod difference-in-differences (DID) approach to rigorously evaluate the policy’s impact on transportation CO₂ emissions. To enhance the robustness of our findings and mitigate potential selection bias among pilot cities, we incorporate controlled city attributes and apply the propensity score matching PSM–DID method. Additionally, recognising the interconnected nature of urban systems [35], we investigate the spatial spillover effects of the LCCPP on transportation CO₂ emissions in surrounding cities. This study not only fills the current gap in the literature but also provides a more holistic understanding of the policy’s direct and indirect effects on transportation decarbonization. These contributions aim to inform more targeted and effective low-carbon strategies, advancing the broader goal of sustainable urban development.

2.2. Theoretical Hypotheses

The LCCPP clearly indicates specific tasks, emphasising the acceleration of developing low-carbon industrial, energy, building, and transportation sectors, as well as changing individuals’ lifestyles. It actively explores innovative experiences and practices, and plans and builds urban transportation, energy, and other infrastructures in accordance with low-carbon concepts. Thus, it is meaningful to study its effect on urban transportation CO₂ emissions and its mechanisms of action on the relevant departments to assess the implementation effects and illustrate the experience of pilot cities.

Hypothesis 1.

Implementation of LCCP policy is conducive to urban transportation CO₂ emissions reduction.

The inhibitory effect of environmental regulation on CO₂ emissions has been empirically demonstrated by numerous scholars [36,37,38], For example, Sun (2025) found that the implementation of China’s Low-Carbon Transportation System (LCTS) pilot policy significantly reduced urban carbon emission intensity by an average of 17.3%, highlighting its effectiveness in promoting urban carbon reduction [39]. And the LCCPP, as a key environmental regulation, has become an important policy for coping with climate warming and reducing CO₂ emissions. Existing studies have confirmed that the LCCPP can directly reduce urban CO₂ emissions [40] and improve emission efficiency [41]. Considering that transportation is one of the major sources of urban carbon emissions, and its emission characteristics are closely linked to energy consumption and travel demand, it is reasonable to hypothesise that the implementation of the LCCPP will effectively promote reductions in transportation-related CO₂ emissions.

Hypothesis 2.

The LCCP policy reduces urban transportation CO₂ emissions by improving urban public transportation levels, green technology innovations, and residents’ green mobility.

The LCCPP promotes CO₂ emission reduction through several pathways. On the one hand, it encourages cities to develop public transportation systems and increase the number of buses to reduce CO₂ emissions [30]. On the other hand, it relies on measures including green technology innovation, industrial restructuring, and energy consumption reduction to facilitate carbon emission reduction targets [42,43,44]. For example, Huang Huan [45] argues the CO₂ emission reduction effect of pilot policies from multiple perspectives. However, the extent to which technological advancements specifically contribute to transportation sector decarbonization remains underexplored and requires further investigation. In addition, green travel has been widely recognised a valid way to reduce CO₂ emissions [46,47,48], and the LCCPP emphasises changing commuting habits to promote low-carbon urban development. Studies have shown that the pilot policy raises citizens’ awareness of low-carbon cities [49]. Some scholars have explored carbon reduction pathways in pilot policies from different perspectives, such as the public transportation environment [50] and resident participation [51]. Therefore, based on the above evidence, we propose that the LCCPP reduces transportation’s CO₂ emissions by enhancing urban public transportation, green technology innovations, and residents’ green travel.

Hypothesis 3.

The effect of the LCCP policy on urban transportation CO₂ emissions differs across cities and may exhibit spatial spillover effects.

Cities in China differ greatly in economic development, government administrative capacity, transportation infrastructure, and travel demand patterns. These variations shape the composition and efficiency of urban transportation systems, resulting in heterogeneous responses to environmental policies. Previous studies have shown that although the LCCPP effectively reduces CO₂ emissions overall, its impact varies substantially among regions due to differences in local development contexts and policy implementation capacity [38,52,53,54,55]. Moreover, since transportation systems are spatially connected through intercity travel, logistics, and regional cooperation, the implementation of the LCCPP in one city may also influence transportation-related CO₂ emissions in neighbouring cities through spillover effects. Therefore, it is reasonable to hypothesise that the LCCPP’s effects on transportation CO₂ emissions not only differ across cities but also extend spatially to surrounding areas.

3. Methodology and Data

3.1. Multiperiod DID Model

Because this policy was initiated in three stages, we use a multiperiod DID approach to examine whether the policy in China reduces urban transportation CO₂ emissions. The multiperiod DID baseline model is defined by the following:

$$Y_{it}=\alpha +\beta DID+\gamma Contro{l}_{it}+{\sigma }_t+{\eta }_i+{\epsilon }_{it}$$

(1)

where $i$ and $t$ are the observation city and observation year, respectively, and $Y_{it}$ is the urban transportation CO₂ emissions. DID denotes the virtual variable of the LCCP. $Contro{l}_{it}$ denotes control variables. $\alpha$ is a constant term, $\gamma$ is the estimated coefficient of the control variable, ${\sigma }_t$ is a city-fixed effect, ${\eta }_i$ is a time-fixed effect, and ${\epsilon }_{it}$ is a random disturbance term. The parameter $\beta$ is the LCCPP’s effect on urban transportation CO₂ emissions, where a significant positive value indicates that it has increased urban transportation CO₂ emissions, whereas a significant negative value implies that it has successfully decreased urban transportation CO₂ emissions.

3.2. Mediating Effects Model

The mediating effect model [56] is introduced to verify the LCCP’s indirect effect mechanisms on urban transportation CO₂ emissions. The specific model as follows:

$$M_{it}={\alpha }_1+{\beta }_{1}DID+{\gamma }_{1}Contro{l}_{it}+{\sigma }_t+{\eta }_i+{\epsilon }_{it}$$

(2)

$$Y_{it}={\alpha }_2+{\beta }_{2}DID+\rho M+{\gamma }_{2}Contro{l}_{it}+{\sigma }_t+{\eta }_i+{\epsilon }_{it}$$

(3)

where $M_{it}$ is a mediating variable, $\rho$ is the estimated coefficient of $M_{it}$. The remaining variables remain as defined in Model (1).

3.3. Data and Variables

Dependent variable

Herein, transportation CO₂ emissions data for each city are sourced from the MEIC website (http://meicmodel.org.cn/, accessed on 9 April 2025), which is based on the Multi-resolution Emission Inventory Model for Climate and Air Pollution Research (MEIC) developed by Tsinghua University. The MEIC model generates high-resolution carbon emission data for five sectors: transportation, power, industry, residential, and agriculture. Transportation emissions data cover various modes such as rail, bus, truck, and motorcycle. Compared to traditional ‘top-down’ methods, this model offers a more comprehensive and detailed dataset.

Independent variable

The independent variable indicates whether a city is categorised as a low-carbon pilot region; it is the DID variable in the empirical model, DID being the cross-multiplier of two variables: Treated and Period. The value of the Treated variable is 1 when the city is chosen as a pilot city and 0 otherwise. The value of the Period variable is 1 in the present year and all years thereafter if the city is categorised as a pilot city in that year, and 0 otherwise.

Mediating variables

According to the relevant literature review and theoretical hypotheses, urban public transportation, green technology innovations, and residents’ green travel may be a valid way to reduce CO₂ emissions in transportation. Consequently, we select these factors as mediating variables to examine their roles in the LCCPP’s impact on transportation CO₂ emissions.

Control variables

Referring to existing research [57,58,59,60] and combining the current developments in the transportation industry, this paper mainly controls the following variables: economic development, population, industrial structure, consumption level, urban density, infrastructure level, foreign investment level, and environmental pollution index.

Data specification

Owing to data availability limitations, 284 prefecture-level Chinese cities were included, and the duration of the research spans from 2006 to 2020. Data for all indicators were obtained from the China Urban Statistical Yearbook, the China Statistical Yearbook and the EPS database.

Table 1. Definition of variables.

Variable	Definition or Measurement	Unit
Urban transportation carbon emission (TCE)	CO2 emission of urban transportation industry measured by MEIC model	Tons
Difference-in-Difference (DID)	The cross-multiplier of the two variables Treated and Period	/
Economic development (ECO)	GDP per capita (GDP divided by the total population)	CNY per 10,000 people
Population (POP)	Total urban population at the end of year.	10,000 people
Industrial structure (INS)	Proportion of the tertiary industry GDP/GDP.	/
Consumption Level (CON)	Total retail sales of consumer goods/GDP	/
Urban density (URD)	Urban total population/built-up area.	10,000 people per km2
Infrastructure level (INF)	Urban road area.	10,000 square metres
Foreign investment level (FOI)	Overseas foreign direct investment/GDP	Ten thousand passengers per car
Environmental pollution index (EPI)	Calculated by entropy method (including industrial SO2 emissions, industrial wastewater emissions, industrial solid waste emissions)	/
Public transportation levels (PUT)	Public buses/Private car ownership.	/
Green technology innovations (GRT)	Number of green patent applications.	Pieces
Resident’s green mobility (REG)	Public transportation ridership	10,000 people

presents the definitions of variables used in this paper,

Table 2. Descriptive statistics.

Variables Set	Variables	Observation	Mean	Standard Deviation	Minimum	Maximum
Dependent variables	CO2	4260	254.7	268.6	8.972	2196
Independent variables	DID	4260	0.225	0.417	0	1
Control variables	ECO	4260	4.836	5.001	0.276	50.63
	POP	4260	440.7	313.4	16.41	3416
	INS	4260	39.90	10.05	8.580	83.87
	CON	4260	0.367	0.107	3.11 × 10−5	0.996
	URD	4260	0.354	0.425	0.022	8.749
	INF	4260	1816	2483	14	31,012
	FOI	4260	0.0190	0.0239	1.77 × 10−6	0.390
	EPI	4260	0.0897	0.0890	0.001	0.939
Mediating variables	PUT	4260	0.00512	0.00683	7.00 × 10−5	0.105
	GRT	4260	250.7	797.2	2	12,534
	REG	4260	20,473	43,290	18	516,517

presents the descriptive statistics, and

Table 3. Test for multicollinearity.

	ECO	POP	INS	CON	URD	INF	FOI	EPI
Tolerance	0.331	0.186	0.683	0.708	0.175	0.143	0.856	0.753
Variance inflation factor (VIF)	3.020	5.370	1.460	1.410	5.720	6.970	1.170	1.330

Note: Tolerances less than 0.1 and variance inflation factor values greater than 10 indicate possible multicollinearity.

presents the tests for multicollinearity of variables. Meanwhile, all variables other than the dummy variables were converted to natural logarithms.

4. Empirical Analysis

4.1. Parallel Trend Test and Dynamic Effect Analysis

An essential premise of the DID approach is the need to satisfy the parallel trend. Following the approaches in the relevant literature [61], event analysis is employed in this paper to examine parallel trends. The approach used for testing the parallel trend hypothesis is as follows:

$$Y_{it}=\alpha +{\displaystyle \sum _{k=1}^p{\beta }_{-k}}DI{D}_{it-k}+\beta DI{D}_{it}+{\displaystyle \sum _{k=1}^q{\beta }_{+k}DI{D}_{it+k}}+\gamma Contro{l}_{it}+{\sigma }_t+{\eta }_i+{\epsilon }_{it}$$

(4)

where ${\beta }_{-k}$ and ${\beta }_{+k}$ are the magnitude of the impact in k periods before and after the enforcement of the policy, respectively. The remaining variables remain as defined in Equation (1).

The consequences for the parallel trend test are shown in Figure 3, which summarises data for the three years before the start of the policy (including year three) as period 3, and data for the six years after the start of the policy (including year six) as period 6. Before the start of the pilot policy (period −3 to period −1), none of the policy shock coefficients significantly deviated from zero and all of them failed the significance test. This suggests that before the start of the LCCPP, a significant difference was not observed between treated and control groups, with the empirical model satisfying the parallel trend.

At the start of the policy (period 0), the policy shock coefficients are not statistically significant, reflecting that its impact is delayed and has not yet manifested itself in the current period. However, when the policy is implemented for 1 year and beyond (periods 1 to 6), the magnitude of the policy shock coefficients gradually increases, all of which achieve statistical significance. Therefore, with respect to dynamic effects, the LCCPP has increasingly contributed toward lowering transportation CO₂ emissions for pilot cities.

4.2. Baseline Results

Table 4. The DID regression results.

Variables	(1)	(2)
	CO2	CO2
DID	−0.048 ***	−0.055 ***
	(0.013)	(0.012)
ECO		0.198 ***
		(0.023)
POP		0.186 ***
		(0.057)
INS		0.008
		(0.032)
CON		0.001
		(0.007)
URD		0.037 **
		(0.018)
INF		0.015
		(0.010)
FOI		0.006 *
		(0.003)
EPI		0.003
		(0.008)
Constant	4.690 ***	3.775 ***
	(0.012)	(0.373)
Observations	4260	4260
R-squared	0.823	0.990
Year fix	Yes	Yes
City fix	Yes	Yes

Notes: Robust standard errors in parentheses, *** p < 0.01, ** p < 0.05, * p < 0.1.

shows the regression estimates of the LCCPP’s effect on transportation CO₂ emissions. Columns 1 and 2 present findings from estimating policy effects without and with control variables, respectively, with city–time-fixed effects. In both scenarios, the regression coefficients corresponding to the main explaining variables became negative with a significance level of 0.01, indicating that the LCCP significantly contributed to reducing transportation CO₂ emissions, demonstrating its full effectiveness in urban transportation CO₂ reduction. Specifically, the implementation of the LCCPP reduced urban transportation CO₂ emissions by 5.5%, which is likely attributed to the LCCP encouraging the growth of low-carbon transportation in cities, thereby reducing transportation CO₂ emissions.

In the control variables, ECO, POP, URD, and FOI are the main reasons for increasing urban transportation CO₂ emissions. For URD, some studies have shown that it can effectively reduce urban CO₂ emissions [62], while conversely, some studies have found that it is a major cause of increased CO₂ emissions in Chinese cities [58]. The effect of urban density on transportation CO₂ emissions in China needs to be seriously considered by the relevant authorities, considering that the density of Chinese cities is much higher than that of Western cities.

4.3. Indirect Mechanism

To further validate the indirect mechanisms of the effects of the LCCPP on transportation CO₂ emissions, its effect is estimated based on models (3) and (4). Since the key to mechanism identification is the effect of local independent variables on its results, we only show the direct effects of key variables. The results are shown in Figure 4, and the overall test results are presented in

Table A1. Test for the mediation mechanism.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
	PUT			GRT			REG
Variables	lnCO2	lnPUT	lnCO2	lnCO2	lnGRT	lnCO2	lnCO2	lnREG	lnCO2
			−0.023 ***			−0.011			0.010 ***
			(0.005)			(0.018)			(0.004)
did	−0.055 ***	0.072 ***	−0.053 ***	−0.055 ***	0.020	−0.055 ***	−0.055 ***	0.036 ***	−0.056 ***
	(0.006)	(0.024)	(0.006)	(0.006)	(0.024)	(0.006)	(0.006)	(0.019)	(0.006)
lnx1	0.198 ***	−0.023	0.198 ***	0.198 ***	0.028	0.198 ***	0.198 ***	0.446 ***	0.170 ***
	(0.014)	(0.046)	(0.014)	(0.014)	(0.049)	(0.014)	(0.014)	(0.055)	(0.014)
lnx2	0.186 ***	−0.048	0.184 ***	0.186 ***	−0.836 ***	0.178 ***	0.186 ***	1.459 ***	0.156 ***
	(0.028)	(0.121)	(0.028)	(0.028)	(0.109)	(0.029)	(0.028)	(0.128)	(0.028)
lnx3	0.008	0.036	0.009	0.008	−0.139 **	0.007	0.008	0.060*	−0.004
	(0.018)	(0.067)	(0.018)	(0.018)	(0.070)	(0.018)	(0.018)	(0.032)	(0.007)
lnx4	0.001	−0.076 ***	−0.001	0.001	0.057	0.001	0.001	−0.047	0.002
	(0.005)	(0.024)	(0.005)	(0.005)	(0.037)	(0.005)	(0.005)	(0.052)	(0.005)
lnx5	0.037 ***	0.150 ***	0.041 ***	0.037 ***	−0.104 **	0.036 ***	0.037 ***	0.301 **	0.165 ***
	(0.010)	(0.037)	(0.010)	(0.010)	(0.052)	(0.010)	(0.010)	(0.126)	(0.041)
lnx6	0.015 **	0.065 **	0.017 ***	0.015 **	−0.066	0.015 **	0.015 **	0.114 ***	0.019 ***
	(0.006)	(0.033)	(0.006)	(0.006)	(0.044)	(0.006)	(0.006)	(0.032)	(0.006)
lnx7	0.006 ***	0.020 **	0.006 ***	0.006 ***	0.017 *	0.006 ***	0.006 ***	−0.010	0.005 ***
	(0.002)	(0.008)	(0.002)	(0.002)	(0.009)	(0.002)	(0.002)	(0.008)	(0.002)
lnx8	0.003	0.030 *	0.004	0.003	−0.005	0.003	0.003	0.013	0.010 **
	(0.004)	(0.017)	(0.004)	(0.004)	(0.019)	(0.004)	(0.004)	(0.016)	(0.004)
Constant	3.775 ***	−5.714 ***	3.643 ***	3.775 ***	8.301 ***	3.854 ***	3.775 ***	−7.187 ***	3.291 ***
	(0.193)	(0.796)	(0.190)	(0.193)	(0.700)	(0.203)	(0.193)	(0.924)	(0.250)
Observations	4252	4252	4252	4260	4260	4260	4260	4260	4260
R-squared	0.990	0.841	0.990	0.990	0.626	0.990	0.990	0.953	0.990
Year fix	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
City fix	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

Notes: *** p < 0.01, ** p < 0.05, * p < 0.1.

in Appendix A.

According to Figure 4a, the direct impact coefficient γ is −0.054, which indicates the direct effect of the LCCP on urban transportation CO₂ emissions with a significance level of 1%. From Figure 4a,b, it can be found that the LCCP significantly improves the level of public transportation in a given city (α = 0.076 ***). The public transportation level variable reduces urban transportation CO₂ emissions (β = −0.022 ***), and after identifying and stripping out the mediating effect, the marginal effect of the LCCPP on urban transportation CO₂ emissions γ’ is −0.052 ***, which is smaller than the direct effect γ before the mediating effect is stripped out (−0.054 ***).

In Figure 4a,c, urban green technology innovation is improved by the implementation of the LCCP (α = 0.020), and the green technology innovation variable also reduces urban transportation CO₂ emissions (β = −0.011), but neither of them is significant. This shows that the green technology innovation effect does not play a significant mediating role in the implementation of the LCCP, and the direct effect of urban transportation CO₂ emissions is not realised through this mechanism. However, studies have shown that green technology innovations can contribute to a reduction in the overall CO₂ emissions of cities [63]. This may be due to technology innovations in the industrial sector that have significantly reduced industrial energy emissions. However, transportation CO₂ emissions are mainly determined by vehicle emissions, and technologies such as new energy vehicles may face higher costs and lower market acceptance at the initial stage, resulting in slower penetration and thus not yet contributing to a significant CO₂ reduction.

Figure 4a,d shows that the LCCP significantly increases residents’ green mobility (α = 0.036 ***). The residents’ green mobility variable shows its influence on reducing urban transportation CO₂ emissions (β = −0.010 ***), and after identifying and stripping off the mediating effect, the marginal effect of the LCCPP on urban transportation CO₂ emissions γ’ is −0.054 ***, which is smaller than the direct effect of the mediating effect before stripping off the mediating effect γ (−0.055 ***). It shows that the residents’ green mobility effect has a significant mediating effect; the direct effect of urban transportation CO₂ emissions can be realised through this mechanism.

In summary, from the results presented in Figure 4, it is clear that the LCCPP not only directly decreases transportation CO₂ emissions; it also promotes a decrease in urban transportation CO₂ emissions by enhancing the PUT and REG.

4.4. Robustness Tests

4.4.1. Placebo Test

To eliminate the influence of unobservable factors on results according to time and region, a placebo test was conducted with reference to the study of Xiao et al. [64] and Li et al. [15]. This was performed as follows: pseudo-processing group ‘Treated’ was randomly generated and regressed by a computer, and then 500 random simulations were performed, and the coefficients and p values of $Period\times Treated$ were extracted and plotted as scatter plots, as shown in Figure 5.

As shown in Figure 5, estimated coefficients for the interaction terms are mainly clustered around 0, with most simulated coefficients having p values above 0.05, indicating insignificance at the 1% level. This finding suggests that baseline estimates are likely not random and thus unaffected by omitted variables, affirming the robustness of the core conclusions.

4.4.2. PSM–DID Estimation Results

Despite its important role in causality judgements, the DID approach still suffers from the problem of sample self-selectivity, such that the sample is not random, which leads to biased estimation results [65]. Therefore, rematching the sample using PSM methods can minimise differences among treated and control groups across all characteristics [66].

Considering the weak constraints of the LCCPP and the randomness in the selection of pilot cities, we utilise the PSM–DID approach to examine the baseline results, thereby mitigating the impact of these factors on the research results. During the matching process, the control variables in the treated and control groups were first tested for balance, as shown in Figure 6.

As shown in Figure 6, differences in control variables among the treated and control groups before and after matching are presented. Significant differences among the covariates were evident before matching, but post matching, the standardised bias of each covariate was largely balanced around the zero line. This indicates that PSM effectively mitigates systematic differences among the control variables in the two groups, achieving balance.

In addition, to avoid the potential impact of the matching method on the robustness of the conclusion, we refer to the existing literature to use nearest neighbour matching [54], radius matching, and nuclear matching methods [67] separately to match the samples, all fixing time and region effects.

Table 5. Robustness tests.

	(1)	(2)	(3)
Variables	Nearest Neighbour Matching	Radius Matching	Nuclear Matching
	CO2	CO2	CO2
DID	−0.030 ***	−0.054 ***	−0.053 ***
	(0.009)	(0.006)	(0.006)
CTPP
Control variables	Yes	Yes	Yes
Constant	3.917 ***	3.858 ***	3.865 ***
	(0.213)	(0.198)	(0.198)
Observations	1843	4253	4248
R-squared	0.997	0.990	0.990
Year fix	Yes	Yes	Yes
City fix	Yes	Yes	Yes

Notes: *** p < 0.01.

presents DID estimation results after matching, which indicates that DID coefficients in columns 1–3 are all significantly negative, suggesting that the test conclusions are consistent with the previous ones after the PSM treatment eliminates self-selection bias among the samples, which further improves the robustness of the empirical conclusions.

5. Further Analysis

5.1. Heterogeneity Analysis

5.1.1. Heterogeneity in Geographic Location

The policy implementation and transportation CO₂ emission levels of the sample cities may differ greatly [68,69,70]. Therefore, the sample cities were first divided into eastern, central, and western regions according to the classification of the National Bureau of Statistics of China. They were also divided into northern and southern cities based on the Qinling–Huaihe River Line, which is the natural and climatic boundary between northern and southern China. Regional heterogeneity tests were then conducted separately.

Table 6. Heterogeneity regression.

Variables	(1)	(2)	(3)	(4)	(5)
	East	Central	West	North	South
	CO2	CO2	CO2	CO2	CO2
DID	−0.076 ***	−0.008	−0.065 ***	−0.017 *	−0.085 ***
	(0.008)	(0.010)	(0.010)	(0.009)	(0.007)
Control variables	Yes	Yes	Yes	Yes	Yes
Constant	4.672 ***	3.734 ***	3.734 ***	3.064 ***	4.086 ***
	(0.348)	(0.212)	(0.212)	(0.273)	(0.255)
Observations	1500	1500	1260	1950	2310
R-squared	0.875	0.865	0.865	0.818	0.874
Number of cities	100	100	84	130	154
Year fix	Yes	Yes	Yes	Yes	Yes
City fix	Yes	Yes	Yes	Yes	Yes

Notes: *** p < 0.01, * p < 0.1.

gives the regression results.

As shown in columns 1–3, it is clear that a gradient pattern is formed in the three major regions, namely, east (−0.076 ***), west (−0.065 ***), and central (−0.008). This indicates that the LCCPP as implemented in the east, central, and west regions has negative effects on transportation CO₂ emissions in all cases, but the effects differ, with transport carbon reduction in the east showing an optimal effect with 7.6%, followed by the west with 6.5%, with central being the worst with only 0.8%. The reason is that eastern cities are generally more economically developed, their public transportation systems are more complete, and the popularity of green mobility among residents is higher, so the policy is more effective. Additionally, we found that DID coefficients in the central region’s cities were smaller and failed significance tests. This may reflect rapid traffic growth outpacing lagging public transit development coupled with resource allocation prioritising economic expansion over low-carbon initiatives, thus weakening the policy’s impact amid rising car ownership and structural constraints. For western cities, the LCCP’s CO₂ reduction is also significant, which may be due to western cities’ smaller traffic volume and smaller transportation CO₂ emissions, which makes the “carbon lock-in effect” weaker [71], enabling the policy effect to be exerted more quickly.

Columns 4 and 5 show that the LCCP reduces transportation CO₂ emissions significantly in both southern and northern cities. According to estimated coefficients, the policy’s effect is significantly larger in the southern (−0.085 ***) than in the northern (−0.017 *) region. This indicates that the LCCP is more effective in the southern region. This is because the popularity of public transportation faces greater challenges in winter in the northern region, where winters are cold and long and people tend to use private cars to travel.

5.1.2. Heterogeneity in City Type

The effectiveness of LCCP policies may differ across cities [72,73]. Therefore, with reference to Lu et al. [74], cities were categorised into high-economy and low-economy cities based on gross domestic product per capita, and their economic levels’ heterogeneity was revisited.

Column 1 of

Table 7. Heterogeneous results based on city type.

Variables	(1)	(2)	(3)	(4)
	High-Economy Cities	Low-Economy Cities	Resource-Based Cities	Non-Resource-Based Cities
	CO2	CO2	CO2	CO2
DID	−0.039	−0.147 **	−0.029 ***	−0.064 ***
	(0.079)	(0.066)	(0.010)	(0.007)
Control variables	Yes	Yes	Yes	Yes
Constant	−1.832 *	0.980	3.022 ***	3.966 ***
	(1.016)	(1.132)	(0.240)	(0.232)
Number of cities	99	185	113	171
R-squared	0.888	0.662	0.848	0.848
Year fix	Yes	Yes	Yes	Yes
City fix	Yes	Yes	Yes	Yes

Notes: *** p < 0.01, ** p < 0.05, * p < 0.1.

indicates that the LCCPP coefficient for the sample of high-economy cities is −0.039; however, it is not significant. Column 2 indicates that the LCCPP coefficient in the low-economy cities sample is −0.147 and is significant at a 5% level. This suggests that the LCCPP regarding transportation CO₂ emission reductions has a significant role in low-economy cities, whereas it does not exert a significant role in high-economy cities. This may be due to the fact that in high-economy cities beyond leading eastern hubs, the policy’s impact is limited, as rapid urbanisation drives traffic demand and car ownership, outpacing public transit development. Resources often prioritise industrial growth over low-carbon infrastructure, and residents’ reliance on convenience reduces responsiveness to green incentives. Conversely, low-economy cities, with lower traffic volumes and less-entrenched high-carbon systems, face fewer transition barriers. Their residents, more sensitive to cost-saving policies (e.g., subsidised transit), adopt green travel readily, amplifying reductions from a lower baseline.

The sample was divided into resource-based and non-resource-based cities based on the categorisation criteria of the State Council’s National Sustainable Development Plan for Resource Cities (2013–2020) and then tested. Columns 3 and 4 of Table 7 of the DID-estimated coefficients are both significant at the 1% level, implying that in both types of cities, the LCCPP effectively decreases transportation CO₂ emissions. However, regarding the values of the estimated DID coefficients, the estimated coefficient value is −0.064 for non-resource-based cities and −0.029 for resource-based cities, with the former being more than twice as high as the latter; thus the effect of the policy is better utilised in non-resource-based cities. This may be due to resource-based cities’ reliance on carbon-intensive industries and logistics, limiting transit shifts, while non-resource-based cities, with diversified economies and flexible infrastructure, more readily adopt green travel, enhancing transportation CO₂ emission reductions.

5.2. Spatial Spillover Analysis

When variables are spatially correlated, ignoring spatial correlation can lead to bias [75]. Therefore, this paper further utilises the spatial estimation models to evaluate the impact of spatial spillovers of the LCCPP. The formula is as follows:

$$Y_{it}=\phi {\displaystyle \sum _{j=1}^n{w}_{ij}}{Y}_{it}+\alpha +\beta DID+\gamma Contro{l}_{it}+\theta {\displaystyle \sum _{j=1}^n{w}_{ij}}DID+\lambda {\displaystyle \sum _{j=1}^n{w}_{ij}}Contro{l}_{it}+{\sigma }_t+{\eta }_i+{\epsilon }_{it}$$

(5)

where $\phi$ is the spatial autoregressive coefficient for the dependent variable, $w_{ij}$ is the spatial weight matrix, and $\theta$ is the coefficient of spatial spillover. Other letters have the same implications as in Equation (1).

5.2.1. Spatial Autocorrelation and SDM Applicability Tests

The spatial aggregation of transportation CO₂ emissions in 284 cities across China from 2006 to 2020 was examined using Moran’s I index. As shown in

Table 8. Results of Moran’s I index measurement.

Year	Moran’s I	E (I)	SD (I)	Z-Statistic	p-Value
2006	0.180 ***	−0.004	0.034	5.426	0.000
2007	0.177 ***	−0.004	0.034	5.317	0.000
2008	0.172 ***	−0.004	0.034	5.171	0.001
2009	0.193 ***	−0.004	0.034	5.743	0.000
2010	0.212 ***	−0.004	0.034	6.312	0.000
2011	0.219 ***	−0.004	0.034	6.468	0.000
2012	0.223 ***	−0.004	0.035	6.560	0.000
2013	0.226 ***	−0.004	0.035	6.617	0.000
2014	0.214 ***	−0.004	0.035	6.217	0.000
2015	0.206 ***	−0.004	0.035	6.055	0.000
2016	0.211 ***	−0.004	0.035	6.185	0.000
2017	0.211 ***	−0.004	0.035	6.181	0.000
2018	0.209 ***	−0.004	0.035	6.109	0.000
2019	0.207 ***	−0.004	0.035	6.059	0.000
2020	0.211 ***	−0.004	0.035	6.181	0.000

Note: *** p < 0.01.

, Moran’s I index is positive and passes the 1% significance level test, confirming the significant positive spatial correlation of transportation CO₂ emissions in China, which can be used for the subsequent spatial econometric model applicability test. Meanwhile, Figure 7 shows the distribution of local Moran’s I, with the horizontal axis representing urban transportation CO₂ emissions and the vertical axis representing their spatially lagged values. It can be clearly seen that the spatial dependence of urban transportation CO₂ emissions has always existed, and most of the cities have clustering characteristics of ‘high–high’ and ‘low–low’.

Further testing and analysis are required to identify the most suitable spatial econometric model. The results of LM tests (Lagrange Multiplier tests), as shown in

Table 9. LM and LR tests.

Test			Test Statistics	p-Value
LM test	LM	Spatial error	1738.7770 ***	0.0000
		Spatial lag	145.2860 ***	0.0000
	Robust LM	Spatial error	1599.3900 ***	0.0000
		Spatial lag	5.9000 ***	0.0150
LR test	H0: SAR model		102.7500 ***	0.0000
	H0: SEM model		169.7400 ***	0.0000
Hausman test	Test of H0: Difference in coefficients not systematic
	chi2(8) = 80.19		Prob >= chi2 = 0.0000

Note: *** p < 0.01.

, demonstrate that the LM error and lag tests and the robust LM error and robust LM lag tests reject the original hypothesis at the 1% level. This reveals that spatial substantial correlation and spatial perturbation coexist, i.e., the use of SDM to estimate urban transportation CO₂ emissions is better than SEM and SAR. Moreover, LR tests (Likelihood Ratio tests), which determine whether SDM degenerates into SEM and SAR, rejected the original hypothesis at a 1% significance level, suggesting that the SDM model is more appropriate. Furthermore, the Hausmann test results favour the fixed-effect model. Based on these findings, this paper employs fixed-effect SDM to further examine the spatial spillover effects of the LCCPP on transportation CO₂ emissions.

5.2.2. Spatial Panel Regression Results

Table 10. Spatial spillover effects of the LCCPP.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	Main	WDID	Spatial	Variance	Direct Effect	Indirect Effect	Total Effect
	CO2	CO2	CO2	CO2	CO2	CO2	CO2
DID	−0.059 ***	−0.051 **			−0.065 ***	−0.131 ***	−0.196 ***
	(0.0162)	(0.0253)			(0.0161)	(0.0378)	(0.0377)
Control variables	Yes	Yes			Yes	Yes	Yes
rho			0.453 ***
			(0.0278)
sigma2_e				0.108 ***
				(0.00257)
Observations	4260	4260	4260	4260	4260	4260	4260
R-squared	0.683	0.683	0.683	0.683	0.683	0.683	0.683
Year fix	Yes	Yes	Yes	Yes	Yes	Yes	Yes
City fix	Yes	Yes	Yes	Yes	Yes	Yes	Yes

Notes: *** p < 0.01, ** p < 0.05.

presents the experimental findings concerning spatial impacts on transportation CO₂ emissions by the LCCPP based on the SDM–DID model. The R2 and Sigma2 values of the SDM model are satisfactory, and DID coefficients at 1% level are significantly negative, which means that the transportation CO₂ emission reduction by the LCCPP is obvious in the local area. In column 2, the coefficient of the spatial lag term WDID is −0.051 with significance at a 5% level, suggesting that the LCCPP exerts a notable spatial spillover effect toward reducing urban transportation CO₂ emissions.

Comparing the regression coefficients of the base regression with those of the spatial Durbin regression coefficients, it can be found that ignoring the spatial spillover effect will underestimate the negative inhibitory impact of the policy to some degree. A possible explanation is that the policy’s effects stimulate transportation CO₂ emission control measures in the region, where pilot cities can be regarded as the pioneer zones for CO₂ emission control, under the incentive of a ‘demonstration effect’ [76]; the neighbouring cities will learn from their advanced management mode, which will promote the transportation CO₂ emission reduction. Therefore, a comprehensive consideration of the spatial spillover effect of the LCCPP may improve the conclusions drawn from this study and draw them closer to the objective reality.

Because the estimated parameters of the SDM only explain the direct spatial effects of the LCCPP on transportation CO₂ emissions, not the true partial regression coefficients, the marginal effects cannot be judged. Therefore, it is necessary to further decompose the policy’s effects into the direct, indirect, and total effects by utilising the partial differential decomposition method. Columns 5–7 in Table 10 present the detailed spatial estimation results.

First, in a spatial model, the estimated parameter of the direct effect is −0.065, and all of the results pass a 1% significance level test, which proves that the LCCPP at the city level exerts a significant suppression effect on transportation CO₂ emissions. Second, the indirect effect reflects spatial spillover generated by the LCCPP on urban transportation CO₂ emissions, and its coefficient is −0.131 and is significant at a 1% level, which proves that the LCCPP reduces the transportation CO₂ emissions in the region and exerts a notable inhibitory effect on the transportation CO₂ emission of the surrounding cities.

6. Conclusions and Policy Implications

Considering the LCCPP and transportation CO₂ emissions in the same research framework, this paper uses 284 cities in China as its study object to examine the effects of the LCCP on urban transportation CO₂ emissions. Relevant empirical tests are conducted through DID methods, and the following conclusions are drawn:

First, the baseline regression analysis and mechanism analysis findings show that the LCCP has directly and significantly promoted transportation CO₂ emission reduction in cities, and the role of the LCCPP as a ‘booster’ of transportation CO₂ emission reduction has been fully manifested. Second, the results of the indirect mechanism identification show that the LCCPP can effectively reduce CO₂ emissions by upgrading public transportation levels and residents’ green mobility. However, green technological innovations in urban transportation, such as new energy vehicles (including battery electric, plug-in hybrid, and fuel cell vehicles), have not yet shown significant carbon reduction effects due to their higher costs and challenges in large-scale adoption. Therefore, the mechanism in Hypothesis 2 is partially verified. Third, the heterogeneity in the LCCPP is remarkable. Because of the higher public transportation level and green mobility penetration rate among residents in eastern cities, the negative LCCPP effects on transportation CO₂ emissions are most obvious in eastern regions of China. Moreover, owing to lower emission baselines, higher policy sensitivity, favourable economic and geographic conditions, and less carbon-intensive logistics dependence, the policy is more effective in low-economy cities, southern cities, and non-resource-based cities. Finally, there are remarkable spatial spillover effects of the LCCPP on urban transportation CO₂ emission control. Ignoring spatial spillover effects will to some extent underestimate the suppression of transportation CO₂ emissions by the policy, as the contribution of the indirect effects to the total effect is higher than that of the direct effects, confirming that the LCCPP not only decreases transportation CO₂ emissions in its local settings but also produces a radiation effect and a penetrating effect on CO₂ emission control in other surrounding areas. Thus Hypothesis 3 is confirmed. Based on these conclusions, we suggest several policy implications.

(1)

We support the strong evidence that the LCCP can restrain transportation CO₂ emissions. Local governments can apply this policy to reduce transportation CO₂ emissions. Therefore, according to the overall concept of ‘expanding point by point,’ the carbon emission reduction experiences of pilot cities can be organised on a regular basis so as to form a replicable and effective model to facilitate the widespread promotion of low-carbon cities.

(2)

The results of the indirect mechanism identification show that the LCCPP can reduce transportation CO₂ emissions through the improvement of urban public transportation levels and the popularisation of green mobility among residents. Therefore, more focus should be placed on improving urban public transportation levels in the construction of low-carbon cities and further increasing the incentives for low-carbon travel, such as the provision of subsidies and the construction of green transportation infrastructure. Simultaneously, it will be essential for governments and communities to enhance promotion and guidance to elevate the residents’ consciousness and sense of responsibility for low-carbon travel. In addition, the promotion and localised application of new energy vehicles should be strengthened to compensate for the lack of effectiveness of technological innovations at the city level.

(3)

The findings of the heterogeneity analysis indicate that the effectiveness of the LCCP toward lowering transportation CO₂ emissions varies across different regions, with ineffective performance in central and high-economy cities and lower mitigation effects in northern and resource-based cities than in southern and non-resource-based cities. Therefore, authorities should prioritise these cities, scrutinise the reasons behind the LCCPP’s ineffectiveness and enhance policy supervision. Differentiated policy measures should be implemented, avoiding a uniform approach, and instead, tailored policy implementation plans should be developed that align with the specific characteristics of each city.

(4)

Finally, because there are obvious spatial spillover effects on transportation CO₂ emission reduction when using the LCCPP, a cross-regional management coordination mechanism should be established based on the LCCPP to enhance interregional collaboration. Nonpilot cities should actively learn from the transportation management experience of pilot cities to reduce local transportation CO₂ emissions.

7. Discussion

Although this study provides robust evidence that the Low-Carbon City Pilot Policy (LCCPP) significantly reduces carbon emissions from the transport sector, several limitations should be acknowledged. First, the carbon emission data used are derived from sectoral energy consumption and do not account for the full life-cycle emissions (LCA) associated with transport infrastructure, such as vehicle manufacturing, road construction, and maintenance. Future research could integrate LCA data to capture embodied emissions and improve the comprehensiveness of the assessment. Moreover, this study faces several other limitations. Potential interactions with concurrent environmental policies (e.g., carbon trading pilots, fuel standards, or new-energy vehicle subsidies) may partially affect the estimated effects. In addition, due to data availability, this study covers the period up to 2020, which may constrain the assessment of long-term policy impacts. Compared with previous research that generally finds technological innovation to be an effective driver of overall urban carbon reduction, our results suggest that such innovation has not yet played a significant role in reducing transportation-related CO₂ emissions. This difference may be related to the relatively high costs, limited charging and supporting infrastructure, and the slow market penetration of new energy vehicles during the study period. As data on technological diffusion and policy performance continue to improve, future studies could update the dataset and conduct further empirical investigations to verify and extend this finding, while also employing micro-level travel or firm data to better capture spillover effects and long-term policy implications.