Decoupling relationship between economic growth and PM2.5 emissions in the transportation sector in China: regional differences and influencing factors

PM2.5 emissions from the transportation sector are a source of haze pollution in China, to which, however, less attention is paid by society. The decoupling relationships between PM2.5 emissions and economic growth from the transportation sector in the eastern, central, and western regions of China from 2010 to 2017 are analyzed by using the Tapio decoupling model. On this basis, in the transportation sector, socioeconomic factors influencing PM2.5 emissions and effective means of controlling PM2.5 emissions are studied by using a logarithmic mean Divisia index model. The results indicate that: (a) in China’s transportation sector, the decoupling relationships of the two aspects in the eastern, central, and western regions show an N-shaped trend, that is, the rate of change in PM2.5 emissions from the transportation sector gradually exceeds that of economic development. The strong decoupling changes into an expansive coupling in the eastern and central regions, while the strong decoupling becomes an expansive negative decoupling in the western region. (b) Economic growth and population growth mainly contribute to the increase of PM2.5 emissions. Improvements of the energy structure and a decrease in transport intensity are the main factors driving a reduction in PM2.5 emissions. (c) Due to regional differences in the ‘rebound effect’ and ‘technological effect’, technological progress has increased PM2.5 emissions from the transportation sector in the central region, while reduced such emissions in the eastern and western regions. This research provides targeted policy reference for regional governance of PM2.5 emissions from the transportation sector.


Introduction
(generally refers to particulate matter with an aerodynamic equivalent diameter less than or equal to 2.5 µms), as a main component of haze, has seriously affected people's productivity and lives [1][2][3]. To control PM 2.5 pollution, the Chinese Government promulgated an Action Plan for Air Pollution Prevention and Control in 2013 [4]. Benefiting from the regulation of the policy, PM 2.5 emissions from the industry, power, and household sectors were separately decreased by 40%, 31%, and 28% from 2013 to 2017 [5]. Meanwhile, PM 2.5 emissions from the transportation sector were increased by 3%, and its proportion of the total societal emissions was increased from 4.6% in 2013 to 7.1% in 2017. Compared with other sectors, emissions from transportation increase, therefore, PM 2.5 emissions from the transportation sector have hindered the emission-reduction process insofar as it relates to PM 2.5 across the whole of society. More significantly, the transportation sector has become the primary source of PM 2.5 emissions in some cities. For example, in 2017, PM 2.5 emissions from the transportation sector in cities, such as Shenzhen, Beijing, Jinan, Shanghai, Hangzhou, and Guangzhou accounted for 52.1%, 45%, 32.6%, 29.2%, 28%, and 21.7% of the total emissions in these cities, respectively [6]. In the future, if no other intervention is taken, with the continuous improvement of people's living standards and the constant increase of car ownership, the pressure on controlling PM 2.5 emissions from the transportation sector will be mounting.
Due to the vast areal extent of China, there are regional imbalances in transportation services, vehicle ownership and transportation infrastructure in the east, central, and western regions, as well as different emission reduction potentials and environmental supervision efforts, which are prone to problems of regional environmental imbalances [7]. Historical data indicate that economic development, transportation energy consumption, and pollution emissions tend to be high in the east and low in the west [8]. Among them, in the transportation sector, PM 2.5 emissions in eastern China were about 48% of total emissions (central and western China accounted for about 31% and 21% thereof). To achieve the win-win goal of economic development and haze control, it is necessary to evaluate the ability of the regional transportation sector to grow without increasing PM 2.5 emissions (i.e. the decoupling effect) [9] and its related drivers at the regional level from the perspective of the transportation sector.
The present study is designed to answer the following questions: in the transportation sector, what is the ability of eastern, central and western China to grow economically without increasing PM 2.5 emissions? What are the socioeconomic factors affecting PM 2.5 emissions and their influencing mechanisms? To answer these questions, the Tapio decoupling model [9] are combined with the logarithmic mean Divisia index (LMDI) model [10] to reveal the decoupling relationships between PM 2.5 emissions from the transportation sector and economic growth in the eastern, central, and western regions of China. Furthermore, factors influencing the decoupling relationship, including the energy structure, technological progress, transport intensity, economic growth, and population are analyzed. This study has important practical significance for appropriately setting the goal of PM 2.5 emission reductions from the transportation sector. The results provide an intuitive perspective for promoting the coordinated growth of PM 2.5 emission reductions and the regional economy.
The contributions of this research are mainly reflected in the following two aspects: (a) Different from previous studies that use PM 2.5 concentration as the monitoring indictor [11,12], this research is designed so as to adopt data related to PM 2.5 emissions. (b) Previous studies focus on CO 2 emissions [7,8,13], energy consumption [14,15], and chemical mechanisms of PM 2.5 pollution [3] from the transportation sector. Unlike these, the present research attempts to ascertain differences in PM 2.5 emissions from a regional perspective and reveal the underlying mechanisms of any socioeconomic impacts, so as to fill the research gap in the understanding of air pollution in the transportation sector. It should be noted: although the PM 2.5 concentration data show advantages such as intuitiveness and being easy to be monitored at the city level, they are greatly affected by factors such as meteorology and geography [16]. The use of PM 2.5 emission data can avoid errors of the aforementioned factors in socio-economic research. In this case, we focus on primary PM 2.5 emissions, which account for a large portion of PM 2.5 pollution in China [17]. The main source of primary PM 2.5 emissions is fossil fuel consumption [18], which is directly affected by socioeconomic factors. Secondary PM 2.5 emissions come from the oxidation of other air pollutants and contribute to the total amount of PM 2.5 , but the formation of secondary PM 2.5 emissions shows complex chemical changes and has certain uncertainties [19]. Therefore, we do not further consider secondary PM 2.5 emissions.

Literature review
Haze pollution, as a natural phenomenon, is one of the most serious environmental problems in China [20], but its causes lie in socioeconomic activities [21]. Some studies show that haze pollution is related to economic growth [22,23]. According to the theory behind the environmental Kuznets curve, economic growth usually causes an increase in pollution emissions, however after implementing effective mitigation policies, higher economic growth will be accompanied by lower pollution emissions, which can be regarded as a decoupling [24]. The decoupling implies that economic growth is divorced from environmental pollution, which allows analysis of the dilemma at the nexus between them [25]. Decoupling analysis can elucidate the relationship between research objects through elasticity and shows simple calculation and intuitive results [26]. It is considered a good choice [27] when studying the relationship between economic development and pollution emissions, which has been widely used in the research of energy [14,28] and carbon emissions [7,8,13,29]. However, at present, the PM 2.5 decoupling has been mostly studied at the national [25,30] or regional level [11,31], while is less often used at the sectoral level. Although Fang et al [24] assessed the decoupling relationship between PM 2.5 emissions from industry and industrial growth in China, there is no research on PM 2.5 emissions from the transportation sector.
There are four key methods used to evaluate factors influencing PM 2.5 emissions: the Stochastic Impacts by Regression on Population, Affluence, and Technology model (STIRPAT), Production Decomposition Analysis (PDA), Structural Decomposition Analysis (SDA), and the logarithmic mean Divisia index (LMDI) method (table 1).
Compared with other methods, although the LMDI model, as a branch of index decomposition analysis, cannot decompose multiple influencing factors [38], its decomposition factors are easy The selection of factors is subjective and has the influence of random errors.
Xu and Lin [32] Economic growth Zou and Shi [12] Urbanization & Population Lu et al [2] Industrial agglomeration PDA Can analyze micro-factors related to 'efficiency' .
Efficiency estimation affects the accuracy of results; lack of discussion on structural components (energy structure, etc.).
Xu et al [33] Output efficiency & Technological progress Wang et al [34] Efficiency improvement & Technological progress SDA Can identify direct and indirect emissions from economic activities.
Requires high-quality data of input-output tables; exist problem of 'path dependence' .
Guan et al [17] Capital formation & Exports Yang et al [35] Consumption & Consumption pattern Li et al [36] Emission intensity LMDI Can provide perfect decomposition; data requirements are low; results are easy to interpret.
The number of decomposing factors is limited.
Liu and Wang [37] Freight volume Dong et al [22] Synergistic emissions reduction Zhang et al [21] Emission intensity & Energy intensity Dong et al [31] Emission coefficient & Economic growth Zhang et al [11] Investment scale & Emission intensity Fang and Yu [24] Industrial development & Energy intensity to explain, thus making the model more applicable [26]. Some scholars have employed the LMDI model to estimate the effects of drivers on PM 2.5 pollution, such as the emission coefficient, energy structure, economic output, energy efficiency, and population [21,31]. In particular, Zhang et al [11] found that PM 2.5 emission intensity and investment efficiency accelerate the regional decoupling. Dong et al [22] proposed that the synergistic effect of carbon emissions is the main factor promoting reductions in PM 2.5 emissions. A few scholars have analyzed factors influencing for PM 2.5 emissions from the industry [24] and freight [37] of China and considered that improving energy efficiency will promote emission reductions. Most studies suggest that the LMDI model can be used to reveal the mechanisms of influence, and extent of, key socioeconomic factors and their effects on changes in PM 2.5 emissions. Therefore, after comparing the advantages and disadvantages of the four models, the LMDI model is chosen for the analysis of those factors affecting PM 2.5 emissions from the transportation sector in the eastern, central, and western regions of China.
Existing studies mainly focus on the whole-ofsociety, industry, or other sources, and few scholars have studied PM 2.5 emissions from the transportation sector of China. To explore regional differences in PM 2.5 emissions from the transportation sector, the decoupling state between PM 2.5 emissions from the transportation sector and economic growth is revealed based on decoupling theory. The internal drivers of PM 2.5 emissions from the transportation sector are researched by combining such data with an LMDI model. The research provides some new insights into high-quality development of the transportation sector in each region.

Tapio decoupling model
The decoupling model is introduced to determine the degree of dependence of PM 2.5 emissions from the transportation sector on economic growth. The decoupling model includes the Organization for Economic Cooperation and Development index-based decoupling method, Velma decoupling index method, and Tapio elastic coefficient method [14,39]. The Tapio model enables further subdivision of the decoupling state, with more accurate results. Therefore, the Tapio model is selected to study the decoupling state between PM 2.5 emissions from the transportation sector and economic output. The governing formula is as follows: where, PM t and PM 0 represent the PM 2.5 emissions at time t and the base period t 0 , respectively; TA t and TA 0 indicate the added values in the transportation sector at times t and t 0 , respectively. Tapio [9] introduced an elastic coefficient into the decoupling model and defined that the increase and  Figure 1 shows classification standards for eight decoupling state and its economic meaning.

LMDI decomposition model
In order to study socioeconomic impacts of PM 2.5 emissions from the transportation sector, a decomposition model is introduced for quantitative analysis. The LMDI decomposition model can flexibly set influencing factors, decompose the residual, and avoid the problems of zero (or negative) values, so it has been widely used in the environmental research [26]. By using the LMDI method, the contributions of factors influencing PM 2.5 emissions from the transportation sector are analyzed. The governing formula is as follows: where, subscript i indicates a province of China; n represents the total number of provinces in the region; PM 2.5i denotes the PM 2.5 emissions from the transportation sector in the ith province; e i represents the total consumption of fossil energy of the transportation sector in the ith province; E i indicates the total energy consumption of the transportation sector in the ith province; V i denotes the total turnover of the transportation sector in the ith province; TA i indicates the value added by the transportation sector in the ith province; P i denotes the total population of the ith province. Transport turnover is the product of transport volume (persons or tons) and average distance (km). It reflects the amount of transportation workload, directly affects the energy consumption of the transportation process, and influences pollution emissions. To obtain the total turnover of transportation, it is necessary to convert the passenger turnover and then add it to freight turnover, as follows: where, V t indicates the total turnover (ton-km); V t p represents passenger turnover; V t f denotes freight turnover; C is a conversion coefficient. The conversion coefficient is determined through comparing turnover per person-km (moving one person one km) with turnover of moving one ton of goods one km. The conversion coefficients for each mode of transport can be obtained China's current statistical system [40] and relevant literature [41], and are listed in table 2.
Formula (2) is simplified into formula (4): According to the model calculation, the change of PM 2.5 emissions from the transportation sector from the base year to the target year can be decomposed into six factors: To obtain the influences of each effect in the period [0, t], the following formula is used: where, X as a variable represents EF, ES, EI, TI, EG, P. Among them, ∆EF represents the emission coefficient effect and reflects influences of change of the energy emission coefficient on PM 2.5 emissions; ∆ES indicates the energy structure effect, which reveals influences of the change in the proportion of fossil energy (oil, natural gas, and coal) consumed by the transportation sector in total energy on PM 2.5 emissions; ∆EI denotes the technological progress effect, which is used to evaluate the effects of improvements in energy efficiency on PM 2.5 emissions; ∆TI denotes the transport intensity effect, which measures influences of improvement of transportation efficiency on PM 2.5 emissions; ∆EG represents the economic growth effect and reflects influences of economic growth on PM 2.5 emissions; ∆P indicates the population size effect and measures changes to PM 2.5 emissions due to changes in the population. It should be noted: on the one hand, because the data are selected from a short research interval, PM 2.5 emission factors associated with each separate source of energy will not fluctuate greatly due to the increase of energy utilization [42,43]; on the other hand, for the time being, there remains a lack of a clear standard for PM 2.5 emissions from vehicles in China. The change of emission factors from vehicles mainly results from the variation in the energy structure of vehicles, which is consistent with the effect of energy structure. Therefore, it is assumed that the emission factor is constant and the influence of the emission coefficient effect on PM 2.5 emissions is always zero.   the decoupling and mitigate the continuous growth of PM 2.5 emissions in the future. As shown in figure 3, from the time dimension analysis in the medium term (2-3 years), the decoupling state of the transportation sector has mostly changed from strong decoupling to weak decoupling, that is, the growth rate of PM 2.5 emissions becomes higher than that of the transportation sector. The decoupling state deteriorates, which can further prove the correctness of the short-term decoupling results.

Influence factors for growth of PM 2.5 emissions in regions
To elucidate the mechanisms of influence on PM 2.5 emissions from the transportation sector, the growth of PM 2.5 emissions is decomposed into an energy structure effect (∆ES), technological progress effect (∆EI), transport intensity effect (∆TI), economic growth effect (∆EG), and population size effect (∆P) from 2010 to 2017. As demonstrated in figure 4, emission incremental effects in the eastern and central regions are greater than emission reduction effects, PM 2.5 emissions thus increase; in the western region, each effect is relatively small and finally the emission reduction effect is greater than the emission increment effect. Specifically, the energy structure effect and transport intensity effect are emission reduction effects, while the economic growth effect and population size effect are emission increment effects. In eastern and western regions, the technological progress effect is one of emission reduction, while it increases emissions in the central region.
Due to the different levels of economic development in eastern, central, and western China, there may be differences in the economic implications reflected by the absolute and relative values (contribution rate) of the LMDI decomposition results. Herein, the contribution rate index is used to compare the influences between regions, thus eliminating the uncertainty of the decomposition results caused by different levels of economic development. In addition, different LMDI decomposition periods will also produce a little differences in decomposition results (see figure 5), but this does not change the main conclusions of the article. Next, we will analyze the mechanism of the influencing factors of long-term results.

Emission reduction effect
The emission reduction effects of the energy structure effect on PM 2.5 emissions from the transportation sector mainly benefit from the optimization of the energy consumption structure. The emission reductions arising therefrom in the eastern, central, and western regions are 55, 89, and 37 thousand tons, respectively. The contribution rates of energy structure effect on PM 2.5 emissions in the eastern, central, and western regions of China are −1272%, −802%, −722% respectively. The energy structure refers to the proportion of fossil energy (oil, gas, and coal) in the total energy mix and describes the cleanliness  of the energy consumption structure. As shown in figure 6, the proportions of fossil energy in the eastern, central, and western regions of China separately were decreased by 5%, 7%, and 3% from 2010 to 2017. The lower the proportion of fossil energy in the transportation sector, the cleaner the energy consumption structure, and the lower its pollution emissions.
In addition, PM 2.5 emissions contain many harmful substances, which are mainly divided into heavy metal pollution (such as Cd, Hg, As, Cr, etc) and organic pollution (such as polycyclic aromatic hydrocarbons). Different proportions of harmful components in PM 2.5 will cause different degrees of harm to human health [46]. Studies have shown that the  harmful substances and health hazards caused by fuel oil are generally greater than those caused by burning coal [47]. In China's transportation sector, the proportion of fuel used in the east is larger than that in the central and western regions. Therefore, in the case of the same total emission, PM 2.5 produced by the eastern transportation sector may cause greater health hazards. Fully adjusting the energy structure cannot only reduce emissions, but also reduce health risks.
The emission reduction effects of the transport intensity effect on PM 2.5 emissions from the transportation sector are mainly due to the improvement of transportation efficiency. The emission reductions due to the effect in the eastern, central, and western regions reach 69, 59, and 37 thousand tons, respectively. The contribution rates of transport intensity effect on PM 2.5 emissions in the eastern, central, and western regions of China were −1603%, −534%, −705% respectively. The transport intensity is defined as the turnover per unit of economic output. If this factor promotes emission reductions, it means that the turnover required to produce economic output per unit decreases, the transportation efficiency increases, and energy consumption and pollution per unit transportation decrease [15]. The possible reasons for this are that, on the one hand, the energy consumption per capita associated with buses and subway systems in China is 17% and 2.4% of that of cars, respectively [48]. The development of public transportation saves energy costs, effectively diverges private transportation and realizes environmental protection to a certain extent; on the other hand, with the continuous integration of 'Internet + Logistics' , transportation resources are integrated to reduce unnecessary transportation service and improve transportation efficiency. The improvement of the transportation efficiency implies a concomitant decrease in transportation costs. The proportion of transportation costs in China in gross domestic product decreased from 21.3% in 2010 to 14.7% in 2019. However, in 2019, the proportion still showed a gap with 7.6% of the USA [49]. This result implies that there is still considerable scope for optimizing transportation routes in China.

Emission increment effect
The emission increment effects of the economic growth effect on PM 2.5 emissions from the transportation sector are mainly caused by rapid economic development, which is consistent with conclusions reached by Chen et al [1]. Rapid economic development inevitably consumes much energy and materials and promotes continuous growth in emissions Because development in the eastern region is always outpacing than that in the central and western regions, the increase of car ownership and construction of transportation infrastructures are fast in the east and slow in the west [8]. Therefore, the energy consumption and pollution emissions caused by PM 2.5 emissions in the transportation sector are high in the east and low in the west of China.
The emission increment effects of the population size effect on PM 2.5 emissions from the transportation sector mainly result from population inflow. More population inflows mean higher proportional travel and transportation demands, which accelerates energy consumption and PM 2.5 emissions. The increment of PM 2.5 emissions in the eastern region is 11 thousand tons, which is much higher than those in the central (4 thousand tons) and western (5 thousand tons) regions. And the contribution rates of population size effect on PM 2.5 emissions in the eastern, central, and western regions of China were −251%, −37%, −98% respectively. The main reason for this is that the eastern region shows advantages in economy, technology, and policy terms [22], and therefore accrues a greater population influx than central and western regions, resulting in more pollution emissions.

Uncertain effect: the technological progress effect
Technological progress is measured by energy consumption per unit turnover, and the lower the energy consumption per unit turnover, the higher the technical level of the region [50]. The technological progress effect is divided into the 'technological effect' and 'rebound effect' . On the one hand, technological progress can improve the energy efficiency and reduce unit energy consumption [31] and pollution emissions in the transportation sector (the so-called 'technological effect'). On the other hand, technological progress can reduce the real price of energy [51], which stimulates new demand for transportation services [52] and leads to a new round of energy consumption and pollution emissions (the 'rebound effect'). Therefore, the influences of technological progress are finally evaluated by the relative sizes of the 'technological effect' and 'rebound effect' .
In the transportation sector, technological progress separately reduces PM 2.5 emissions by 31 and 11 thousand tons in the eastern and western regions, while increasing them by 47 thousand tons in the central region. The contribution rates of technological progress effect on PM 2.5 emissions in eastern, central and western China are −727%, 427% and −203% respectively. This finding indicates that the 'technological effects' are greater in the eastern and western regions, while the 'rebound effect' is greater in the central region. The main reason for this is that the research and development (R & D) capability is strong in the eastern region, whereas there is much room for improvement with respect to technological progress in the western region, so the 'technological effect' is dominant in the two regions [52,53]. However, the R & D capacity and space for technological progress are limited in the central region, market-oriented reform of energy resources lags, and the oil price is lower than that of other markets [54]. As shown in figure 7, the average price of gasoline in central China is always lower than that in the east and west. In this case, it is more vulnerable to the influence of oil price to produce new transportation demand, resulting in the significant rebound effect in the central region of China. In research into urban transportation [51,55], the presence of this significant rebound effect in the central region of China has been verified.

Conclusions and policy suggestions
The decoupling relationships between PM 2. Based on these conclusions, PM 2.5 emissions should be reduced as much as possible while maintaining economic growth. The following policy suggestions are proposed: (a) Improving the energy structure of vehicles. In 2020, the prevalence of new-energy vehicles is low (5.4%) in China, so government should strengthen efforts to shift hydrocarbon-fuelled vehicle use to new-energy vehicle use as much as possible. Specifically, government can increase the prevalence of new-energy vehicles through tax reduction, subsidies, and policy support. (b) Improving the efficiency of the transportation system. On the one hand, government should reasonably add buses, subways, and urban railways, and increase commuting concessions for public transportation. Meanwhile, the public is called on to adopt low-energy travel modes, such as walking, cycling, and on-line interaction. On the other hand, with the development of e-commerce, there is significant scope for the improvement of logistics service demands. Government needs to optimize the storage system, ameliorate transportation paths for inventory and shorten the supply chain, so that excessive transportation is avoided. In addition, government departments should also optimize traffic layouts, increase linear traffic routes, and reduce unnecessary transportation in combination with urban and rural policies and the current land-use situation.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).